article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
free - free emission from hii regions at radio wavelengths is a good tracer of young and heavily embedded regions of massive star formation such as luminous starbursts and nuclear star forming regions , where visual extinctions can be tens @xcite to hundreds of magnitudes , affecting even near and mid - infrared observations .
extragalactic continuum emission is complicated by the presence of non - thermal synchrotron emission within the beam , particularly at wavelengths longer than 6 cm . while it is possible in theory to do a spectral and spatial separation of free - free and synchrotron emission @xcite with multi - wavelength observations , the free - free emission can be most easily measured at millimeter wavelengths , where radio synchrotron has declined and
dust has not yet begun to take over .
the enhanced high frequency capabilities of the very large array ( vla ) improved k and q band receivers , fast - switching calibration techniques allow us to detect and resolve the free - free emission from large hii regions in nearby galaxies .
we report vla observations of the 7 mm continuum emission of the starburst galaxy m82 .
an interacting dwarf with a bar and gas ring in the central kpc @xcite , m82 is a prodigious former of stars , at a rate of @xmath6 .
the burst of star formation may have been triggered by the interaction with m81 @xcite , or by infalling molecular gas driven by the stellar bar .
its current active star formation can be traced back to 50 myr ago .
the strong star formation activity is probably responsible for the hot gas outflows found in optical , mm , and x - ray studies .
our goal is to use the 7 mm maps to isolate compact hii regions in m82 , and to determine their free - free flux density , from which we can infer lyman continuum rates , or @xmath7 . with the subarcsecond resolution afforded by the vla
we can determine locations and sizes of the bright hii regions on scales of a few pc .
extinction is high and patchy in m82 @xcite , estimated to be @xmath8 @xcite on large scales , and as much as several hundred mag behind the molecular clouds @xcite , thus radio free - free emission is a particularly valuable tracer of ionized gas in this dusty starburst galaxy .
spectral energy distributions ( seds ) of the compact radio sources at longer , cm wavelengths @xcite in m82 have shown them to be mostly non - thermal synchrotron sources , probably supernova remnants ( snr ) but possibly synchrotron - emitting wind driven bubbles @xcite .
the structures and the expansion properties of these non - thermal sources have been revealed by very long baseline interferometry ( vlbi ) and multi - element radio linked interferometry ( merlin ) with 3 35 mas resolutions and multi - epoch monitoring at 18 and 6 cm @xcite .
in addition to the radio continuum work , @xcite studied the h92@xmath9 and h52@xmath9 radio recombination lines in m82 , including the 7 mm ( 42.952 ghz ) continuum , in @xmath10 or @xmath4 10 pc resolution . in this paper , we have 7 mm images at two resolutions : high resolution , @xmath11 , for compact structures on scales of 3 pc , and low resolution , @xmath12 , for structure on scales of @xmath13 25 pc .
the map with @xmath14 synthesized beam is the highest resolution map of m82 at millimeter wavelengths so far .
lower resolution maps are important for judging the amount of missing , undersampled emission in higher resolution vla images , which act as spatial high - pass filters .
the distance to m82 was recently determined to be [email protected] mpc from the optical study of cepheids in m81 using _
hst_@xcite , assuming the same distance to both galaxies .
observations of the tip of the red giant branch in m82 itself suggests 3.9 @xmath150.3 ( sys ) @xmath150.3 ( random ) mpc @xcite . for consistency with previous work , we adopt the 3.6 mpc distance , with 1@xmath1617 pc . at this distance a 1 mjy thermal source at 7 mm represents an hii region ionized by the equivalent of @xmath4 150 o7 stars .
the radio data were acquired with the nrao vla using a- and b- configurations at 43 ghz ( nrao program i d : at303 ) and d - configuration at 45 ghz ( am839 , pi : meier , d. s. ) .
weather during the a - array observations in october 2004 was not good enough for useful 7 mm work .
the `` high resolution '' ( sub - arcsecond ) maps we present in this paper were made in the b - configuration on 22 april 2005 in continuum observation mode ( effective bandwidth @xmath4172 mhz ) .
the calibration cycle time for fast - switching was set to 100120 seconds .
3c147 was used for calibration of the absolute flux scale and 0958 + 655 was the phase calibrator .
uncertainty in the absolute flux density scale is estimated to be @xmath175% , based on vla estimates of the bootstrap accuracy at 7 mm . for most of our sources ,
statistical uncertainty due to noise dominates the uncertainty in flux density scale .
total on source time is 1.3 hours .
the ( _ u , v _ ) sampling covers 301610 @xmath18 , resulting in a @xmath40.2 beam , the size of which varies slightly with weighting . for our b - array map , the estimated @xmath19 ,
the largest angular scale of structures that our observation is sensitive to , is about @xmath20 .
lower resolution 7 mm continuum observations ( henceforth `` low resolution maps '' ) were made with the vla d - array at 45.47 ghz on november 27 2005 in spectral line mode , as part of a spectral line program ( am839 ) .
the setup was two 25 mhz spectrometer tunings per if set side - by - side .
the uv dataset was generated by fitting the uv spectrometer data minus the edge channel with a first order polynomial using uvlsf task .
the net bandwidth is @xmath4 31 mhz . because if 2 has contamination from a recombination line , only if 1 data is included .
the observations used fast - switching with 180 seconds cycle time .
3c147 and 3c286 were used as flux density calibrators , and 0958 + 655 was the phase calibrator .
total integration time on source is 2.2 hours .
the sampled ( _ u , v _ ) plane covers 5 - 157 @xmath18 , resulting in a @xmath41.5beam .
we estimate the @xmath19 is @xmath4 @xmath21 in our d - array map .
the maps were calibrated and reduced in aips following high frequency reduction procedures , and were deconvolved with clean .
the noise level in the maps is @xmath22 for the high resolution ( b - configuration ) maps and @xmath23 for the lower resolution ( d - configuration ) maps , but varies slightly depending on weighting .
flux measurements and positions were obtained with imfit task in aips unless otherwise noted .
the aips task pbcor was used to correct the flux density scale of these data for the primary beam efficiency of the instrument .
the high - resolution b - array map ( fig . [ figure : na_maps]a ) covers the inner 700 pc @xmath24 350 pc ( @xmath25 ) of m82 with a @xmath26 ( @xmath27 pc ) beam .
structures larger than @xmath20 are suppressed in this map .
we identify 14 compact sources with flux densities @xmath28 0.5 mjy ( @xmath29=12 k ; 5 @xmath30 ) , distributed in a @xmath4 @xmath31 east - west swath with position angle along the optical major axis @xcite .
the compact sources are high brightness , dense , and luminous hii regions and young supernova remnants ( snr ) . for the hii regions ,
our 5@xmath30 detection limit corresponds to @xmath32 ( assuming optically thin free - free emission ) , corresponding to about 80 o stars , or a cluster such as r136 .
the compact optical hii region m82-a1 studied by @xcite falls just below our detection limit .
properties of the compact sources in natural weighted map are given in table [ table : flux_compact ] .
we adopt the source designations of @xcite , using the b1950 coordinate of each source relative to @xmath33 , to facilitate comparisons with earlier studies , but the j2000 coordinates are given also . in table
[ table : flux_compact_200mas ] , we have collected the radio continuum flux densities of all 14 compact sources from the literature ( 18 1.3 cm * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) and this work .
the flux density measurements at different wavelengths are done at similar resolutions .
the beam sizes of maps at 18 cm to 1.3 cm range from @xmath14 to @xmath34 .
the 7 mm measurements quoted in table [ table : flux_compact_200mas ] ( as well as data shown in figure [ figure : sed_hii ] , [ figure : sed_snr ] ) are made on a tapered map of @xmath14 resolution in order to match beam sizes at other wavelengths from the literature .
the quoted numbers in table [ table : flux_compact_200mas ] are total integrated flux densities of compact sources , and the effect of different beam is less than 30% . in figure
[ figure : sed_hii ] , [ figure : sed_snr ] , and table [ table : flux_compact_200mas ] , we show the seds of all 14 compact sources at wavelength 18 cm to 7 mm .
the details of model fitting are included in table [ table : compact_em ] .
our discussion of properties of these compact sources , including their seds , can be found in
[ section : compact_thermal ] and [ section : nonthermal_compact_rc ] .
the high - resolution b - array map also reveals four regions of comparatively extended emission ( @xmath35 ) .
the bright , high frequency free - free emission detected here arises from gas with very high emission measure , so it is surprising to find such extended sources .
they are 12 across , or 1735 pc .
they are similar in extent to the large hii region associated with ngc 3603 in our galaxy , but with a much higher brightness and em .
we discuss the significance of these high brightness , extended regions in [ section : extended_em ] .
the low resolution d - array map is shown in figure [ figure : na_maps]b .
this map , with a beam of 199 @xmath24147 p.a .
17 , is sensitive to lower brightness emission , @xmath36 k rms ( @xmath370.5 mjy / bm ) and also more extended emission structures ( up to 60 , covering the entire field shown ) than the high resolution map . the low resolution image shows four major peaks with intensity 12 mjy / bm or higher .
the integrated flux density in the naturally weighted d - array map is @xmath38 jy . the total flux reported for m82 from _ wmap _ at 41 ghz flux density is @xmath39 jy @xcite .
thus our low resolution d array map recovers , within the uncertainties , all the 7 mm flux density .
the diffuse emission structure is elongated in the same direction and over the same extent as the swath of compact sources , and agrees with the orientation of the stellar elliptical component seen in the deep near infrared images @xcite .
we have fitted 17 peaks in the low resolution map with gaussians , and the corresponding flux densities , locations , and the sizes of the peaks are listed in table [ table : peaks_darray ] .
all of these peaks are found in the longer wavelength vla images of @xcite except peaks w4a and e1a , which are newly identified in this work . the peaks sum to @xmath4 0.65 jy at 7 mm .
if this flux density associated with low resolution peaks is all optically thin free - free emission , and @xmath40 , the inferred lyman continuum photon rate ( @xmath7 ) is @xmath41 .
( for @xmath42 of 5,000 , @xmath7 would increase by 30 40% . )
this value is in good agreement with that estimated from the free - free continuum at longer wavelengths by ( * ? ? ?
* @xmath43 , from 1.3 20 cm multi - wavelength spectral decomposition modeling ) , and at shorter wavelengths by @xcite and @xcite , who find @xmath44 and @xmath45 at 3 mm and 2.6 mm , respectively ( these numbers are interferometric and therefore are lower limits )
. it also agrees with that suggested from recombination line studies @xcite and the extinction - corrected mid - infrared fluxes of @xcite .
the good agreement in @xmath7 between our analysis and previous studies indicates that the peaks we see in the low resolution map of figure [ figure : na_maps]b dominate the total free - free emission .
non - thermal emission at 43 ghz is weak , but contributes a non - negligible fraction of the integrated flux density in m82 .
the estimate of the non - thermal component is 35% of the total at 7 mm , extrapolating from vla cm - wave flux densities of the same region @xcite .
this would give a thermal flux density of @xmath46 , 15 20% higher than 0.65 jy we obtain from summing the peaks in the d - array map of figure [ figure : na_maps]b .
we attribute the discrepancy to thermal emission from some hii regions distributed on too large a spatial scale to be seen in peaks in these images .
the 14 compact sources of the high resolution map ( fig.[figure : na_maps]a ) contribute @xmath47 of the total 7 mm flux density in the low resolution d - array map ( fig.[figure : na_maps]b ) .
these compact sources in the 7 mm high resolution b - array map also appear in lower frequency radio continuum maps @xcite .
we have computed radio seds for the 14 sources , which are shown in figures [ figure : sed_hii ] and [ figure : sed_snr ] , with details of the sed modeling given in table [ table : compact_em ] .
we identify nine regions with spectral index @xmath48 ( @xmath49 ) as hii regions , and the others as snr .
the flux densities at 43 ghz and the spectral indices are consistent with previous seds and source characterizations @xcite except for the two non - thermal sources 43.992 + 59.70 and 41.942 + 57.62 , which are discussed in
[ section : agn ] and [ section : individual_nonthermal ] , respectively . of the nine hii regions in the high resolution map , two have flux densities at more than five frequencies .
fitting of the spectra of these two sources suggests that they have free - free turnover frequencies ( where @xmath50 ) @xmath512.55 ghz . to have @xmath52 a few ghz requires emission measures of @xmath53 pc @xmath2 .
the other sources have flux densities at only three frequencies , and a single power law represents their observed flux densities fairly well ( fig.[figure : sed_hii ] ) . some of the deviations may be due to small amounts of extended emission in the sources and the slightly ( 15 - 20% ) differing beams of the flux densities from the literature .
four sources have rising sed ( @xmath54 ) .
rising spectrum radio sources have been found in many extragalactic systems , and are thought to be giant compact hii regions with high electron density @xcite . these are much larger than galactic compact hii regions because the super star clusters ( sscs ) , the clusters exciting them , are much larger both in excitation and in physical size . for the sources here ,
the emission appears to remain partially optically thick at frequencies up to 43 ghz , so we expect to see brightness temperatures of @xmath55 k. their observed brightness temperatures are lower than this , and imply ems @xmath56 ( derived from peak flux density ) , consistent with turnover frequencies of 3 ghz , an order of magnitude lower than observed .
the low brightness temperatures indicate small filling factors , and signal the presence of very compact sources of significantly less than 3 pc extent , which we can not resolve . considering the radio continuum luminosity , brightness of the emission , and the turnover frequency ,
these appear to be young hiiregions , and we infer that the star clusters exciting these nebulae are also relatively young .
the total flux density of just the detected compact hii regions comes to @xmath4 38 mjy , which is @xmath4 6% of the total thermal flux density at 7 mm ( at the same sensitivity ) , in other words , @xmath4 6% of the ionizing flux density .
but the high resolution 7 mm image fails to detect all of the compact hii regions in m82 , even fairly optically luminous ones such as m82-a1 @xcite , so we believe that there are more compact regions below our current detection limit . in this case
the hii region phase in m82 must last at least @xmath4 6% of the @xmath410 myr lifetime of o stars , and possibly longer , to a myr or more . to restrain the hii regions from expanding
requires external or internal confinements , perhaps external high pressures @xcite , or in the larger clusters , potentially slowing by gravity @xcite .
the compact sources are shown in figure [ figure : dust_lanes ] overlaid on a color ( vri ) image from the _ spitzer _ legacy project sings @xcite .
nearly all of the compact sources are found in dust lanes .
the only exception is source 43.922 + 59.70 , which is located near the dynamical center , and a potential agn candidate ( @xmath576 . )
we have attempted to align the compact radio sources identified in this and previous work ( allen 1999 , mcdonald et al 2002 , and references therein ) with the published _ hst _ near - ir continuum images of @xcite and @xcite , the catalogued optical clusters @xcite , and the archived _ hst _ optical measurements .
there is no consistent alignment of the radio and other sources within the _ hst _ pointing uncertainty .
the hii regions we see are presumably associated with young sscs ; that the clusters are not seen in the optical / near - ir continuum argues that the extinction must be high ( @xmath58 ) .
the molecular clouds alone can contribute @xmath59 of @xmath60@xmath61 if they happen to be along the line of sight to the hii regions .
if the average extinction at arcsecond resolution is @xmath62 ( from h - k color by * ? ?
* assuming @xmath63 ) , the line - of - sight extinctions implied in the lack of near - ir sources to these young clusters must be 4 times higher .
this is consistent with the high average ( @xmath64 ) and patchy extinction derived on large ( 11 , or @xmath4 200 pc ) scales toward the hii regions in m82 from brackett line observations by ( * ? ? ?
* @xmath65 ) , ( * ? ? ?
* @xmath66 ) , ( * ? ? ?
* @xmath67 ) , and ( * ? ? ?
* @xmath68 ) , much of which appears to be internal to the hii regions themselves . in the @xmath69 image of @xcite
, we identify 3 ( 39.227 + 54.32 , 40.933 + 58.98 , and 42.072 + 58.52 ) out of our 9 compact thermal sources , and one non - thermal source ( 43.992 + 59.70 ) as associated with compact @xmath69 structures .
if the @xmath69 emission is associated with the corresponding radio hii regions , we estimate , based on the inferred lyman continuum flux densities , that the extinction toward these ir - visible nebulae is @xmath70 .
none of these three sources are apparent in the _ hst _ near - ir continuum images , and it agrees with the high extinction suggested in the @xmath69 image . besides the compact dense sources in table [ table : compact_em ] , we detect extended regions in m82 with relatively high em . in figure
[ figure : extended_b - array ] we show b - array maps of free - free emission regions with integrated flux density of 50 70 mjy in 5 9 area .
these regions have @xmath71
. this agrees with the d - array measurements shown in table [ table : peaks_darray ] .
high em can be produced either by high rms electron density or by a long sightline ( @xmath72 ) of ionized gas .
if high electron density is responsible , then to reconcile the high em with the observed flux ( @xmath7 ) requires that the line of sight length be much smaller than the observed transverse distance .
for example , to maintain the ionization of a spherical volume of radius 15 pc , the observed transverse diameter of source in fig [ figure : extended_b - array]a , and rms electron density @xmath73 ( @xmath74 was inferred by @xcite in a 2 resolution ) requires a lyman continuum rate of @xmath75 .
the observed flux of this structure indicates @xmath76 .
this implies that the actual emitting volume is smaller , i.e. , the emitting region is sheetlike .
there are a number of such patches evident in figure [ figure : extended_b - array ] . while these large regions of high electron density are dynamically stable in the high pressure @xmath77 environment of the m82 starburst , the excitation requirements are inconsistent with other star formation indicators , with the possible exception of the `` yin - yang '' shaped source ( fig .
[ figure : extended_b - array]c ) located at the peak of the low resolution map ( fig .
[ figure : na_maps ] ) .
these structures could be the ionized surfaces of chimneys @xcite or superbubbles .
the coincidence of their locations with dust lane and co gas distribution implies that they are associated with in the molecular clouds , which would be consistent with an origin in the outer , ionized layers of large molecular clouds . despite its location near the edge of the superbubble ,
the `` yin - yang '' source is an exception in that it does appear to have sufficient flux to support the ionization , and there is a bright mid - ir source here too @xcite .
however this source has an unusual structure that is coherent and bubble - like on a scale of @xmath12 = 26 pc .
if this were a bubble , an energy of @xmath78 ergs would be required to clear the inner region of gas .
another explanation for these extended regions is that they represent a collection of smaller hii regions representing distributed star formation at the end of the stellar bar in m82 .
regions ( b ) - ( d ) in figure [ figure : extended_b - array ] are distributed on the 10 ionized ring of m82 @xcite , which may represent the @xmath79-orbits of a bar .
the region ( a ) is on the east ionized ridge , and probably near the end of the @xmath79-orbits .
star formation is observed to be near bar ends , perhaps facilitated by cloud - cloud collisions .
the dense star forming regions traced by high brightness 7 mm free - free emission are likely to be still close to their natal giant molecular clouds ( gmcs ) .
we therefore expect a high correlation of thermal free - free emission and molecular line emission , although at some sufficiently high resolution , they should begin to anti - correlate . in figure [ figure : molecular_gas ] , we show the 7 mm map with contours of co(1 - 0 ) line emission from the ovro map of @xcite and an hcn(1 - 0 ) map from the plateau de bure interferometer ( schilke , private communication ) .
the molecular gas tracer co(1 - 0 ) is more extended in the plane of the galaxy , while the dense molecular gas tracer hcn(1 - 0 ) emission is confined to the star formation area .
it is apparent that the correlation of the 7 mm continuum emission , largely free - free emission , with hcn gas is excellent .
a quantitative comparison of the molecular line fluxes and the free free flux confirms this conclusion .
figure [ figure : hcn_7 mm ] shows the beam - to - beam comparison between the low resolution 7 mm emission and hcn .
the linear correlation between hcn and radio continuum shown in figure [ figure : hcn_7 mm ] agrees with the tight correlation between global hcn emission and star formation tracers ( such as far - ir emission ) discovered in local normal spiral and starburst galaxies @xcite , in galactic molecular dense cores @xcite , and in high redshift galaxies @xcite .
our plot shows that this good correlation between dense molecular gas and star formation extends down to scales of tens of pc .
to more directly compare our results with @xcite who determined the correlation in terms of infrared luminosity , @xmath80 and hcn line flux , we have translated the 7 mm flux density into @xmath80 .
we assume an o - star - dominated stellar population which yields the relation of @xmath81 , following @xcite and @xcite , or @xmath82 at 3.6 mpc .
this relation could result in underestimates if evolved stellar populations contribute significantly to @xmath80 .
our 7 mm values are plotted with the data of @xcite on figure [ figure : gao_diagram ] .
further discussion of this correlation is in [ section : sfe ] . the polycyclic aromatic hydrocarbon ( pah ) emission features in the mid - infrared are expected to trace photo - dissociation regions ( pdrs ) generated by the softer uv photons produced by ob stars .
emission in the @xmath83 band of the irac camera is dominated by aromatic feature emission @xcite , probably pahs .
the contribution of starlight to the @xmath83 band is estimated to be minimal , at @xmath8410% ( based on irac 3.6 flux ; * ? ? ?
the resolution of irac at this wavelength , @xmath85 , is comparable to our low resolution 7 mm image .
m82 was observed at @xmath86 m by irac on _ spitzer _ as part of the _ spitzer _ infrared nearby galaxy survey ( sings ; * ? ? ? * ) .
the 7 mm contours of our low resolution d - array map are shown overlaid on the irac @xmath5 map in figure [ figure : pah ] .
the 7 mm and @xmath83 emission have very similar morphologies .
we have compared the corrected flux density in irac @xmath87 and at vla d - array 7 mm beam - by - beam .
the irac @xmath87 flux density is corrected to remove the stellar emission in the irac @xmath87 band , following @xcite .
the 7 mm map is convolved to @xmath88 to match the psf of irac channel 4 ( @xmath87 ) . the @xmath89 v.s .
@xmath90 plot is shown in figure [ figure : pah_free - free ] .
we limit the data points shown to those with @xmath91 detection at 7 mm .
this process reduces the contribution of non - thermal radio emission which is fainter at high frequency and more extended .
the ratio is linear , with @xmath92 and a @xmath89 zero intercept @xmath93 in 2.2 beam .
the high correlation of @xmath5 and 7 mm emission may mean that they are excited by the same relatively strong radiation sources on the @xmath94 length scale .
the @xmath89 that appears even at zero @xmath90 , about 40% of the total , could be excited by the relatively older population of b stars that do not produce strong radio continuum emission .
integrating the correlated ( above the offset level ) @xmath5 flux density gives about 30 jy .
our beam - to - beam linear correlation between @xmath89 and @xmath90 suggests that the lyman continuum rate is @xmath95 for @xmath96 .
other relations between @xmath97 and @xmath98 have been found from work at different wavelengths .
deriving a @xmath99 flux from our @xmath98 ( using @xmath100 from @xcite and @xcite ) , the flux density expected from our 7 mm derived @xmath7 using the @xcite relation for @xmath101 would give about a factor of 5 more @xmath97 than observed .
this may not be outside the large scatter in the observed relationship , which @xcite discuss .
another prediction of the @xmath87 flux density can be made from the empirical relations @xmath102 ( * ? ? ?
* based on studies on normal galaxies , starburst galaxies , and hii regions in the local group ) and @xmath103 @xcite .
this would give @xmath97 about a factor of 10 lower than observed .
all relations of the pah luminosity to the @xmath98 are uncertain and have large scatter because the pah flux will depend on the geometry of the radiation field , the role of dust , the presence of b stars , and possibly other factors not yet considered .
the spatial correlation of the pah feature and the ionization in m82 is clear ; the quantitative relation is not yet so .
pah emission may be associated with a b star population @xcite because the strong uv field of o stars could destroy pahs .
if we constrain the ionization sources of the radio continuum emission in hii regions to be b stars , forcing the initial mass function ( imf ) to be deficient in o stars , then to produce the observed @xmath7 would require the total mass of young stars to increase by 2 orders of magnitude relative to a normal imf .
this leads to a star formation rate of few hundred solar mass per year for the past 100 myr , which is highly unlikely . why do the radio continuum emission peaks correlate with @xmath5 peaks , where the pah carriers should be destroyed by the uv
it is possible that dust within the hii regions around the massive stars absorbs a significant fraction of the uv flux @xcite . or the production rate of pahs may be higher than the destruction . to examine the later scenario , we estimate the lower limit to the number density of hydrogen atoms , @xmath104 , for pah survival in the uv photon fields observed : @xmath105 where @xmath106 is the speed of light , @xmath107 is the photon flux rate [ @xmath108 for photons with energy between 513.6 ev , and the dimensionless @xmath109 is defined as the habing photodissociation parameter '' represents the ratio of local @xmath107 photo density and the density of atomic hydrogen .
] , following @xmath5712.7 of @xcite .
@xcite suggest @xmath110 , an upper limit from line - of - sight average from observation , for pah molecule destruction . the starburst99 code ( with salpeter imf )
gives a ratio of the far uv ionizing flux , @xmath111 , to @xmath98 of @xmath112 for o - type dominated stellar population , thus we find @xmath113 \sim 1.3 \times 10^{12}~\frac{i_{7mm}}{mjy / bm}.\ ] ] so the lower limit of hydrogen atom number density , @xmath114 for pah survival in the localized region where @xmath90 is observed . in m82 ,
the peak 7 mm vla flux density in a @xmath115 beam @xmath116 , requiring @xmath117 .
if @xmath118 which ensures the pah survival in the pdrs but not inside the ionized media @xcite , @xmath119 is required to supply the observed pah , close to the critical density for hcn excitation .
this agrees with the good correlation between intensities of pah with hcn emission in m82 found in @xmath120 beam ( not shown in this paper ) .
@xcite proposed the star formation history of selected starburst regions in m82 based on iso mid - ir imaging and spectroscopy for @xmath121@xmath122 . for the central @xmath123 region , they found star formation had peaked near 5 and 9 myr ago with amplitudes of 6.3 and 18.5 @xmath124 , respectively , and had dipped to less than 0.1 @xmath124 at 6 myr or so .
previously @xcite had commented on the lack of the wolf - rayet ( w - r ) spectral signatures in m82 , which may be due to high extinction toward the `` young '' star forming regions ( 36 myr ) where the w - r stars live , but could also reflect the star formation rate ( sfr ) dip near 3 + myr ago .
our observations suggest that high extinction is the reason for the absence of wr features , since all of the young radio free - free sources are found in heavily visually obscured regions .
the integrated flux density at 7 mm of the full region @xcite mapped is about 340 mjy in the low resolution 7 mm map .
this equates to a total mass in young stars of @xmath125 formed within the average lifetime of o stars of 13 myr , for a salpeter imf of 1 - 100 @xmath126 .
the contribution of the previous two starburst events at 5 myr and 9 myr to 7 mm flux density would be negligible . the average sfr over this region in about the past 3 myr ( before o stars enter the w - r phase ) , based on inferred total mass of young star population from 7 mm flux density , is thus estimated as 4 @xmath127 ( salpeter imf of 1 - 100 @xmath126 ) , in agreement with the radio recombination line study by @xcite and with the rate estimated based on numbers of radio snr @xcite .
our 7 mm continuum images give a picture of where the young star - forming regions are in m82 , particularly the compact 7 mm sources , which are likely very young embedded clusters .
what is the spatial relation of these sources to the molecular gas ?
is there evidence in the spatial distribution for how the star formation was triggered ?
we can estimate @xmath128 , the mass of new stars formed , from @xmath80 , @xmath129 , the total mass of gas , from @xmath130 , and the mass of dense ( @xmath131 ; @xmath132 ) gas from @xmath133 .
the dense gas fraction is therefore @xmath134 and the star formation efficiency , the fraction of total gas converted to newly formed stars , is @xmath135 .
this is the star formation efficiency for the current generation of stars ( i.e. , on time scales of a few myr ) .
we convert our observed @xmath136 in lower resolution map to @xmath137 in the 2.7 beam , take that @xmath137 to be @xmath80 , and find @xmath128 from @xmath138 , which assumes an o star dominated population with salpeter imf between 1 - 100 @xmath126 . since we are using the total 7 mm flux density for this computation rather than the thermal 7 mm flux density , the @xmath137 and @xmath80 values could be overestimated by 30% due to the nonthermal contribution .
the full translation is uncertain by a factor of 3 , mainly corresponding to the difference in lower mass cutoff in imf .
we find the dense gas mass fraction from @xmath139 .
for the molecular gas mass , we adopt the conversion of @xmath140 , and @xmath141 from @xcite .
these conversions for molecular gas are uncertain by a factor of 23 , discussed in @xcite . in figure [
figure : gao_diagram ] we plot the observables @xmath142 vs. @xmath143 .
the corresponding sfe against the dense gas fraction ( dgf ) are shown on the top and right .
the data points of m82 represent the results on a 2.7 scale ( or @xmath4 46 pc ) .
we show local normal spirals , luminous infrared galaxies ( lirgs ) , ultra - luminous infrared galaxies ( ulirgs ) , and high - z starburst galaxies from @xcite spirals , luminous infrared galaxies ( lirgs ) , ultra - luminous infrared galaxies ( ulirgs ) , and high - z starburst galaxies from @xcite on the same plot .
the sfe and dgf track in m82 roughly as they do in all sources , although the size scales of the observations can be very different .
all m82 data area lie under the sfe : dgf=1:1 line , which suggests that less than 50% of the molecular gas has been turned into stars .
most of the data points lie above the sfe : dgf = 1:10 line , implying a limit of @xmath4 10% sfe for dense molecular gas .
we have detected 5 non - thermal sources with flux densities higher than 5 @xmath30 .
their flux densities at wavelengths between 7 mm and 18 cm in @xmath144 beams are listed in table [ table : flux_compact_200mas ] and plotted in figure [ figure : sed_snr ] .
we overlay on the measured sed the best fit to a model of a single power law with free - free absorption .
the model results are shown in table [ table : compact_em ] .
about one third of the compact 7 mm sources we detected ( table 1 ) are non - thermal .
these sources have been known for some time and are believed to be snr @xcite .
however , they do not decline with time as expected , so it has been suggested that they may be wind - driven bubbles instead @xcite .
the brightest 7 mm continuum compact source is 43.992 + 59.70 , which has been suggested to be the nucleus of m82 @xcite . on the basis of european vlbi network and vla images , @xcite suggest this source could be a candidate for the agn in m82 based on the jet - like feature , although they detected only the northern part .
the study of two radio oh satellite lines also suggests the oh features of this source could be associated with a circumnuclear disk or torus @xcite .
43.992 + 59.70 is also within 3@xmath145 of the 2 @xmath146 peak of m82 @xcite that is presumably in the center of the mass distribution .
in addition , [ ne ii ] line @xcite , and oh maser kinematics @xcite , molecular gas investigation @xcite , and radio recombination line results @xcite also suggest that this source is close to the dynamical center of m82 , which supports the agn scenario .
the suggestion that 43.992 + 59.70 is an agn can be also tested by looking for variability .
long - term radio flux monitoring by @xcite suggests that the source has a stable flux density , with variability @xmath147 5% , which is the uncertainty in the absolute flux scale calibration .
our 7 mm flux density is about 63% of the archival centimeter - wave sed model s prediction .
it is possible that 43.992 + 59.70 is variable at the current epoch , or that the spectrum is not power law to 7 mm .
a discrepancy from the agn picture is that chandra high resolution x - ray images do not show 43.992 + 59.70 as a strong source .
also , another argument against the agn nature of 43.992 + 59.70 can be based on the recent high resolution vlbi and merlin images of this source @xcite , where it has a resolved round structure , consistent with an snr as suggested by its spectral index ( @xmath148 ) @xcite .
however , the measured 7 mm flux density appears to miss a significant amount of flux density from the sed model of snr , and it is hard to match the 7 mm flux density into the snr picture .
this bright non - thermal source also coincides with a compact @xmath69 and 1.644 [ feii ] emitter .
@xcite modeled the radio emission of non - thermal sources in m 82 , including this one , as wind - driven bubbles .
this scenario hypothesizes that non - thermal synchrotron emitting particles are produced at the shock between the wind of young star clusters and hot bubble gas .
however , @xcite measure the expension velocity of this source to be at 2700 km / s , which is about 50 times higher than the expect wind - driven shell @xcite , and it is more closer to the free - expending snr . in summary , these three models involving agn , snr , and non - thermal bubble of a young cluster could all explain the strong 7 mm emission of 43.992 + 59.70 . despite some suggestive dynamical evidences ,
the morphology and expansion revealed by the vlbi and merlin work strongly indicate that the source is an snr .
however , the snr scenario could not explain the decline of flux density at 7 mm from the expectation value of sed modeling .
the nature of this source might be still open to debate .
@xcite reported 9 x - ray sources in m82 .
the source cxo m82 j095550.2 + 694047 , or m82 x-1 ( @xmath149 , @xmath150 ; * ? ? ?
* ) , has been categorized as an ultraluminous x - ray source ( ulx ) and a 54 mhz quasi - periodic oscillation has been found in x - ray by @xcite and a possible 62 day x - ray luminosity variation by @xcite .
it is suggested to host an intermediate mass black hole @xcite .
many attempts have been made to find a radio counterpart of m82 x-1 .
the source 41.30 + 59.6 or 41.5 + 59.7 has been suggested to be a possible counterpart .
we did not detect a significant radio counterpart of m82 x-1 at 7 mm to a limit of 0.7 mjy / bm .
the seds of the nearby 7 mm sources 40.933 + 58.98 and 41.628 + 58.01 are clearly thermal and are also not candidates for the radio counterpart of m82 x-1 .
the flux density of nonthermal source 41.942 + 57.62 ( 41.9 + 58 in @xcite and @xcite ) at 7 mm is significantly below the power law model ( fig .
[ figure : sed_snr ] ) .
this source is known to vary @xcite ; its flux density declines at @xmath4 8.4
8.8% per year at @xmath4 5 ghz , as has been shown and discussed by several authors @xcite .
the cm data listed in table [ table : flux_compact_200mas ] and plotted in figure [ figure : sed_snr ] were obtained before 1996 @xcite . based on the epoch - corrected 1993 values of @xcite for the flux density of 41.942 + 57.62 , and for a spectral index of -0.72 ( from best - fit result without 7 mm data point )
, we estimate that the flux density of 41.942 + 57.62 would have been 8 mjy at 7 mm in 1993 , three times higher than the observed value .
this is consistent with a decline of 8.4%/year found at 5 ghz .
@xcite first reported the supernova sn2004am ( 9:55:46.61 ; + 69:40:48.1 ) , which was classified spectroscopically as type ii @xcite .
@xcite report 3-@xmath30 upper limits for this source of 0.34 mjy / bm ( 8.4 ghz ) and 1 mjy / bm ( 15 ghz ) on 2003 november 14 .
the null detection in their merlin observations at 5 ghz on march 9 2004 gave an upper limit of 0.18 mjy / bm ( 3 @xmath30 ) , and suggested that the sn2004am should reach the radio emission peak with a significant delay ( up to about 1 year ) .
we can place a 3 @xmath30 upper limit of @xmath151 mjy / bm to the 7 mm flux density of sn2004am in our observation of 2005 april 03 .
there are at least 7 known sources for which the 7 mm flux density predicted from their cm - wave flux densities should be larger than 1 mjy ( @xmath152 ) .
they are 40.68 + 551 , 43.82 + 628 , 45.41 + 637 , 45.48 + 647 , 46.18 + 677 , 46.52 + 639 , and 47.37 + 680 .
the last four of these are at @xmath153 from the phase center where the noise level is a factor of 2 higher than elsewhere and these may be down in the noise .
sources 40.68 + 551 and 43.82 + 628 are , however , within the central primary beam in a low - noise region .
both 40.68 + 551 and 43.82 + 628 are snr .
source 40.68 + 551 has minor variability in the long term radio continuum flux experiment of @xcite , but no variability of 43.82 + 628 has been reported , and this may be the first indication . for three of the 7 mm sources ( 39.659 + 55.73 , 43.992 + 59.70 , and 45.153 + 61.34 ) the 7
mm flux densities are stronger at the epoch of our observation ( 2005 ) with @xmath154 beam than at year 2000 - 2001 in the observation of @xcite with a @xmath10 beam .
the flux density differences are not conclusive ( @xmath155 of the c - array ) , but suggest that these sources may be varying .
the latter two sources are believed to be snr , based on their radio seds .
neither of them vary strongly at 6 cm in the 12-year monitoring program of @xcite between 19811992 .
source 39.659 + 55.73 is known to vary at 6 cm , fluctuating 20% continuously during this time period .
but we have characterized this source as an hii region based on its rising spectrum , and it therefore should not vary .
this source might be associated with a wind - driven bubble , as suggested for the non - thermal sources by @xcite .
we present 7 mm vla continuum maps of the starburst in m82 at two resolutions , 0.2 ( b - configuration ) and 2 ( d - configuration ) .
we recover essentially all of the 7 mm flux density of m82 in the lower resolution d - array map .
an estimated 2/3 of this flux density is thermal free - free emission from the starburst .
we detect less than 10% of the total 7 mm flux density in the high resolution
b - array map , which is sensitive to only bright , compact emission sources .
we identify 14 compact sources greater than 0.5 mjy / bm ( 5@xmath30 ) in the high resolution image , and have constructed seds using archival vla data .
nine of the compact sources are compact hii regions , with emission measures @xmath156 sizes of @xmath157 pc , and rms densities @xmath158 .
these compact sources are located in regions of heavy visual obscuration , with no associated near - ir continuum counterparts .
this could explain the lack of w - r features in m82 .
our detection limit is @xmath159 , so the smallest of these clusters has @xmath160 o stars , and the largest , @xmath4 1200 o stars ( @xmath161 ) .
the detected compact thermal sources account for @xmath4 6% of the total 7 mm free - free flux density .
since the maps detect only the most luminous compact hii regions , this ratio should be higher if fainter ones are taken into account .
we therefore infer that the compact hii region phase in m82 must last at least 6% of the hii region lifetime , @xmath41 myr or longer .
in addition to the compact hii regions , the high resolution map reveals extended ( 2 , or 30 pc ) regions of free - free emission with @xmath162 .
the comparison of intensity ( em ) and flux ( @xmath7 ) suggests that these regions are likely to be non - spherical in geometry .
we suggest two possibilities , either these regions indicate regions of distributed star formation , with many small hii regions possibly reflecting enhanced star formation at the ends of the bar in m82 , or these regions are sheet - like structures , possibly caused by winds or supernovae .
we have compared lower resolution 7 mm vla maps to maps of molecular gas tracers and dust .
the good correlation between the 7 mm emission and _ spitzer _
irac channel 4 map ( dominated by @xmath5 pah features ) suggests a relation @xmath163 , for intensities @xmath164 .
this indicates that the pah feature is a good tracer of star formation in m82 , although we obtain a different relation between @xmath7 and the @xmath5 emission than previous work .
the correlation of 7 mm emission and molecular gas tracers ( co and hcn ) at 2 scales is excellent , indicating that the relation of star formation and hcn emission found on global scales by @xcite holds down to gmc scales .
we estimate that the sfe in m82 on gmc scales is @xmath4 110% , similar to galactic values , and that the regions of highest sfe also have the highest fraction of dense ( @xmath165 ) gas .
the authors would like to thank almudena alonso - herrero , sarah lipscy , and peter schilke for their generosity in providing their data .
c.w.t . appreciates helpful discussions with yu gao .
some of the data used in this paper were obtained from the multimission archive at the space telescope science institute ( mast ) .
stsci is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
this research has also made use of the nasa / ipac extragalactic database ( ned ) and the nasa/ ipac infrared science archive ( irsa ) which are operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
achtermann , j. m. , & lacy , j. h. 1995 , , 439 , 163 allain , t. , leach , s. , & sedlmayr , e. 1996 , , 305 , 602 allen , m. l. , & kronberg , p. p. 1998
, , 502 , 218 allen , m. l. 1999 , ph.d .
thesis , alonso - herrero , a. , rieke , g. h. , rieke , m. j. , & kelly , d. m. 2003 , , 125 , 1210 beck , s. c. , turner , j. l. , langland - shula , l. e. , meier , d. s. , crosthwaite , l. p. , & gorjian , v. 2002 , , 124 , 2516 bennett , c. l. , et al .
2003 , , 148 , 97 beswick , r. j. , muxlow , t. w. b. , argo , m. k. , & pedlar , a. 2004 , , 8332 , 2 beswick , r. j. , et al .
2006 , , 369 , 1221 calzetti , d. , et al .
2007 , , 666 , 870 carlstrom , j. e. , & kronberg , p. p. 1991
, , 366 , 422 condon , j. j. 1992 , , 30 , 575 de grijs , r. 2001 , astronomy and geophysics , 42 , 12 de pree , c. g. , rodriguez , l. f. , & goss , w. m. 1995 , revista mexicana de astronomia y astrofisica , 31 , 39 dopita , m. a. , & sutherland , r. s. 2003 , astrophysics of the diffuse universe , berlin , new york : springer , 2003 .
astronomy and astrophysics library , isbn 3540433627 , dopita , m. a. , et al .
2005 , , 619 , 755 ebisuzaki , t. , et al .
2001 , , 562 , l19 engelbracht , c. w. , et al .
2006 , , 642 , l127 fenech , d. m. , muxlow , t. w. b. , beswick , r. j. , pedlar , a. , & argo , m. k. 2008 , , 391 , 1384 frster schreiber , n. m. , genzel , r. , lutz , d. , kunze , d. , & sternberg , a. 2001 , , 552 , 544 frster schreiber , n. m. , genzel , r. , lutz , d. , & sternberg , a. 2003 , , 599 , 193 freedman , w. l. , et al .
1994 , , 427 , 628 gao , y. , carilli , c. l. , solomon , p. m. , & vanden bout , p. a. 2007 , , 660 , l93 gao , y. , & solomon , p. m. 2004a , , 606 , 271 gao , y. , & solomon , p. m. 2004b , , 152 , 63 greve , a. , wills , k. a. , neininger , n. , & pedlar , a. 2002 , , 383 , 56 helou , g. , et al .
2004 , , 154 , 253 ho , p. t. p. , beck , s. c. , & turner , j. l. 1990 , , 349 , 57 hollenbach , d. j. , & tielens , a. g. g. m. 1997 , , 35 , 179 huang , z. p. , thuan , t. x. , chevalier , r. a. , condon , j. j. , & yin , q. f. 1994 , , 424 , 114 krding , e. , colbert , e. , & falcke , h. 2005 , , 436 , 427 kaaret , p. , simet , m. g. , & lang , c. c. 2006 , , 646 , 174 kennicutt , r. c. , jr . , et al .
2003 , , 115 , 928 keto , e. , ho , l. c. , & lo , k .- y . 2005 , , 635 , 1062 kronberg , p. p. , biermann , p. , & schwab , f. r. 1985 , , 291 , 693 kronberg , p. p. , pritchet , c. j. , & van den bergh , s. 1972 , , 173 , l47 kronberg , p. p. , & sramek , r. a. 1985 , science , 227 , 28 kronberg , p. p. , & sramek , r. a. 1992 , x - ray emission from active galactic nuclei and the cosmic x - ray background , 247 kronberg , p. p. , sramek , r. a. , birk , g. t. , dufton , q. w. , clarke , t. e. , & allen , m. l. 2000 , , 535 , 706 kronberg , p. p. , & wilkinson , p. n. 1975 , , 200 , 430 larkin , j. e. , graham , j. r. , matthews , k. , soifer , b. t. , beckwith , s. , herbst , t. m. , & quillen , a. c. 1994 , , 420 , 159 lester , d. f. , gaffney , n. , carr , j. s. , & joy , m. 1990 , , 352 , 544 lipscy , s. j. , & plavchan , p. 2004
, , 603 , 82 loiseau , n. , nakai , n. , sofue , y. , wielebinski , r. , reuter , h .-
, & klein , u. 1990 , , 228 , 331 mao , r. q. , henkel , c. , schulz , a. , zielinsky , m. , mauersberger , r. , strzer , h. , wilson , t. l. , & gensheimer , p. 2000 , , 358 , 433 matsumoto , h. , tsuru , t. g. , koyama , k. , awaki , h. , canizares , c. r. , kawai , n. , matsushita , s. , & kawabe , r. 2001 , , 547 , l25 matsumoto , h. , tsuru , t. g. , matsushita , s. , kawabe , r. , & prestwich , a. 2000 , bulletin of the american astronomical society , 32 , 751 matsushita , s. , kawabe , r. , kohno , k. , matsumoto , h. , tsuru , t. g. , & vila - vilar , b. 2005 , , 618 , 712 matsushita , s. , kawabe , r. , matsumoto , h. , tsuru , t. g. , kohno , k. , morita , k .-
i . , okumura , s. k. , & vila - vilar , b. 2000 , , 545 , l107 mattila , s. , meikle , w. p. s. , groeningsson , p. , greimel , r. , schirmer , m. , acosta - pulido , j. a. , & li , w. 2004 , , 8299 , 2 mayya , y. d. , carrasco , l. , & luna , a. 2005 , , 628 , l33 mccrady , n. , gilbert , a. m. , & graham , j. r. 2003 , , 596 , 240 mcdonald , a. r. , muxlow , t. w. b. , pedlar , a. , garrett , m. a. , wills , k. a. , garrington , s. t. , diamond , p. j. , & wilkinson , p. n. 2001 , , 322 , 100 mcdonald , a. r. , muxlow , t. w. b. , wills , k. a. , pedlar , a. , & beswick , r. j. 2002 , , 334 , 912 mcleod , k. k. , rieke , g. h. , rieke , m. j. , & kelly , d. m. 1993 , , 412 , 111 melo , v. p. , muoz - tun , c. , maz - apellniz , j. , & tenorio - tagle , g. 2005 , , 619 , 270 muxlow , t. w. b. , pedlar , a. , wilkinson , p. n. , axon , d. j. , sanders , e. m. , & de bruyn , a. g. 1994 , , 266 , 455 muxlow , t. w. b. , pedlar , a. , beswick , r. j. , argo , m. k. , obrien , t. j. , fenech , d. , & trotman , w. 2005 , memorie della societa astronomica italiana , 76 , 586 nakai , n. , hayashi , m. , handa , t. , sofue , y. , hasegawa , t. , & sasaki , m. 1987 , , 39 , 685 neininger , n. , guelin , m. , klein , u. , garcia - burillo , s. , & wielebinski , r. 1998 , , 339 , 737 oconnell , r. w. , & mangano , j. j. 1978 , , 221 , 62 pedlar , a. , muxlow , t. w. b. , garrett , m. a. , diamond , p. , wills , k. a. , wilkinson , p. n. , & alef , w. 1999 , , 307 , 761 peeters , e. , spoon , h. w. w. , & tielens , a. g. g. m. 2004 , , 613 , 986 portegies zwart , s. f. , baumgardt , h. , hut , p. , makino , j. , & mcmillan , s. l. w. 2004 , , 428 , 724 rieke , g. h. , lebofsky , m. j. , thompson , r. i. , low , f. j. , & tokunaga , a. t. 1980 , , 238 , 24 rodriguez - rico , c. a. , viallefond , f. , zhao , j .- h . , goss , w. m. , & anantharamaiah , k. r. 2004 , , 616 , 783 sakai , s. , & madore , b. f. 1999 , , 526 , 599 seaquist , e. r. , bell , m. b. , & bignell , r. c. 1985 , , 294 , 546 seaquist , e. r. , carlstrom , j. e. , bryant , p. m. , & bell , m. b. 1996 , , 465 , 691 seaquist , e. r. , frayer , d. t. , & frail , d. a. 1997 , , 487 , l131 seaquist , e. r. , kerton , c. r. , & bell , m. b. 1994 , , 429 , 612 seaquist , e. r. , & stankovi , m. 2007 , , 659 , 347 shen , j. , & lo , k. y. 1995 , , 445 , l99 simon , m. , simon , t. , & joyce , r. r. 1979 , , 227 , 64 singer , d. , pugh , h. , & li , w. 2004 , , 8297 , 2 smith , l. j. , & gallagher , j. s. 2001 , , 326 , 1027 smith , j. d. t. , et al .
2007 , , 656 , 770 smith , l. j. , westmoquette , m. s. , gallagher , j. s. , oconnell , r. w. , rosario , d. j. , & de grijs , r. 2006 , , 370 , 513 stevens , i. r. , read , a. m. , & bravo - guerrero , j. 2003 , , 343 , l47 storey , p. j. , & hummer , d. g. 1995 , , 272 , 41 strohmayer , t. e. , & mushotzky , r. f. 2003 , , 586 , l61 telesco , c. m. , joy , m. , dietz , k. , decher , r. , & campins , h. 1991 , , 369 , 135 telesco , c. m. 1988 , , 26 , 343 turner , j. l. , beck , s. c. , crosthwaite , l. p. , larkin , j. e. , mclean , i. s. , & meier , d. s. 2003 , , 423 , 621 turner , j. l. , & ho , p. t. p. 1983
, , 268 , l79 turner , j. l. , beck , s. c. , & ho , p. t. p. 2000 , , 532 , l109 turner , j. l. , & ho , p. t. p. 1994
, , 421 , 122 turner , j. l. , ho , p. t. p. , & beck , s. c. 1998 , , 116 , 1212 unger , s. w. , pedlar , a. , axon , d. j. , wilkinson , p. n. , & appleton , p. n. 1984 , , 211 , 783 wei , a. , walter , f. , neininger , n. , & klein , u. 1999 , , 345 , l23 willner , s. p. , soifer , b. t. , russell , r. w. , joyce , r. r. , & gillett , f. c. 1977 , , 217 , l121 wills , k. a. , pedlar , a. , muxlow , t. w. b. , & wilkinson , p. n. 1997 , , 291 , 517 wills , k. a. , pedlar , a. , muxlow , t. w. b. , & stevens , i. r. 1999 , , 305 , 680 wills , k. a. , redman , m. p. , muxlow , t. w. b. , & pedlar , a. 1999 , , 309 , 395 wu , j. , evans , n. j. , ii , gao , y. , solomon , p. m. , shirley , y. l. , & vanden bout , p. a. 2005 , , 635 , l173 yun , m. s. , ho , p. t. p. , & lo , k. y. 1993 , , 411 , l17 llccccl 39.277 + 54.32 & 39.29 + 54.2 & 48.051 & 40.64 & 0.9 & [email protected] & 0.31 , 0.25 , 60 + 39.659 + 55.73 & 39.68 + 55.6 & 48.427 & 41.97 & 3.6 & [email protected] & 0.21 , 0.18 , 170 + 40.933 + 58.98 & 40.94 + 58.8 & 49.694 & 45.20 & 0.6 & [email protected] & 0.47 , 0.35 , 70 + 41.145 + 56.29 & 41.17 + 56.2 & 49.904 & 42.51 & 0.8 & [email protected] & 0.38 , 0.30 , 170 + 41.628 + 58.01 & 41.64 + 58.4 & 50.375 & 44.22 & 1.2 & [email protected] & 0.33 , 0.26 , 0 + 41.942 + 57.62 & 41.95 + 57.5 & 50.686 & 43.77 & 2.6 & [email protected] & 0.18 , 0.13 , 180 + 42.072 + 58.52 & 42.58 + 58.4 & 50.817 & 44.69 & 1.2 & [email protected] & 0.28 , 0.19 , 10 + 42.193 + 59.13 & 42.21 + 59.2 & 50.945 & 45.32 & 1.4 & [email protected] & 0.26 , 0.21 , 130 + 42.549 + 58.13 & 42.56 + 58.0 & 51.291 & 44.29 & 0.9 & [email protected] & 0.68 , 0.44 , 140 + 42.673 + 58.32 & 42.69 + 58.2 & 51.414 & 44.47 & 0.6 & [email protected] & 0.82 , 0.52 , 140 + 43.173 + 58.43 & 43.18 + 58.3 & 51.908 & 44.56 & 0.7 & [email protected] & 0.25 , 0.19 , 90 + 43.295 + 59.27 & 43.31 + 59.2 & 52.028 & 45.42 & 1.6 & [email protected] & 0.22 , 0.17 , 170 + 43.992 + 59.70 & 44.01 + 59.6 & 52.723 & 45.78 & 5.2 & [email protected] & 0.20 , 0.16 , 180 + 45.153 + 61.34 & 45.17 + 61.2 & 53.875 & 47.41 & 1.1 & [email protected] & 0.23 , 0.21 , 170 .
crrrrrrl 39.277 + 54.32 & & 1.7 @xmath15 0.2 & & 1.5 @xmath15 0.2 & & 1.5 @xmath15 0.5 & m02 + 39.659 + 55.73 & & 1.5 @xmath15 0.2 & & 3.2 @xmath15 0.2 & & 8.0 @xmath15 0.6 & m02 + 40.933 + 58.98 & & 1.9 @xmath15 0.2 & 2.9 @xmath15 0.1 & 2.8 @xmath15 0.3 & 2.2 @xmath15 0.3 & 2.5 @xmath15 0.4 & a99 + 41.145 + 56.29 & & 2.0 @xmath15 0.3 & & 2.8 @xmath15 0.2 & & 5.1 @xmath15 0.6 & m02 + 41.628 + 58.01 & & 1.0 @xmath15 0.2 & & 1.8 @xmath15 0.2 & & 6.1 @xmath15 0.5 & m02 + 41.942 + 57.62 & 79.0 @xmath15 4.0 & 40.0 @xmath15 0.9 & 26.8 @xmath15 0.5 & 17.9 @xmath15 1.7 & 12.0 @xmath15 0.6 & 2.4 @xmath15 0.3 & ak98 + 42.072 + 58.52 & & 1.2 @xmath15 0.4 & & 1.7 @xmath15 0.2 & & 3.6 @xmath15 0.6 & m02 + 42.193 + 59.13 & 2.5 @xmath15 1.0 & 4.1 @xmath15 0.5 & 4.3 @xmath15 0.3 & 4.5 @xmath15 0.3 & 4.1 @xmath15 0.3 & 4.0 @xmath15 0.5 & ak98 + 42.549 + 58.13 & & 2.7 @xmath15 0.4 & & 1.8 @xmath15 0.2 & & 1.7 @xmath15 0.2 & m02 + 42.673 + 58.32 & & 1.0 @xmath15 0.4 & & 1.8 @xmath15 0.2 & & 1.1 @xmath15 0.2 & m02 + 43.173 + 58.43 & 8.5 @xmath15 1.0 & 5.2 @xmath15 0.2 & 3.8 @xmath15 0.2 & 2.4 @xmath15 0.3 & 1.8 @xmath15 0.3 & 1.6 @xmath15 0.5 & ak98
+ 43.295 + 59.27 & 19.5 @xmath15 1.0 & 10.3 @xmath15 0.2 & 8.0 @xmath15 0.1 & 5.5 @xmath15 0.3 & 4.1 @xmath15 0.4 & 2.5 @xmath15 0.5 & ak98 + 43.992 + 59.70 & 9.5 @xmath15 1.0 & 21.8 @xmath15 0.9 & 20.2 @xmath15 0.1 & 15.8 @xmath15 0.3 & 12.5 @xmath15 0.3 & 5.4 @xmath15 0.4 & ak98 + 45.153 + 61.34 & 12.5 @xmath15 1.0 & 7.9 @xmath15 0.3 & 5.7 @xmath15 0.1 & 3.9 @xmath15 0.3 & 2.5 @xmath15 0.3 & 2.2 @xmath15 0.5 & ak98 cclcrrrrc
39.277 + 54.32 & hii & pl & -0.1 & @xmath166 & @xmath167 & 22 & 220 & 0.4 + 39.659 + 55.73 & hii & pl & 0.8 & @xmath168 & @xmath169 & 116 & 1160 & 2.3 + 40.933 + 58.98 & hii & ffsa & & 4.9 & @xmath170 & 36 & 360 & 0.7 + 41.145 + 56.29 & hii & pl & 0.5 & @xmath168 & @xmath169 & 74 & 740 & 1.5 + 41.628 + 58.01 & hii & pl & 1.0 & @xmath168 & @xmath169 & 88 & 880 & 1.8 + 42.072 + 58.52 & hii & pl & 0.6 & @xmath168 & @xmath169 & 52 & 520 & 1.0 + 42.193 + 59.13 & hii & ffsa & & 2.5 & @xmath171 & 58 & 580 & 1.2 + 42.549 + 58.13 & hii & pl & -0.2 & @xmath166 & @xmath167 & 25 & 250 & 0.5 + 42.673 + 58.32 & hii & pl & -0.1 & @xmath166 & @xmath167 & 16 & 160 & 0.3 + 41.942 + 57.62 & snr & pl+ffa & -0.7/-0.9 & & + 43.173 + 58.43 & snr & pl+ffa & -0.7 & 0.9 & @xmath172 + 43.295 + 59.27 & snr & pl+ffa & -0.6 & & + 43.992 + 59.70 & snr & pl+ffa & -0.6 & 2.2 & 6.6 + 45.153 + 61.34 & snr & pl+ffa & -0.7 & 1.0 & 3.0 lcccccccccc w1 & 48.08 , 40.8 & 2.4 & 4 & 16.0 & 0.6 & & & 6 & 1.2 &
+ w2 & 48.44 , 42.0 & 4.1 & 8 & 10.5 & 1.2 & & & 12 & 2.3 &
+ w3 & 49.77 , 44.3 & 8.2 & 39 & 12.5 & 5.7 & & 0.08 & 57 & 11 & 0.2 + w4 & 49.96 , 42.9 & 8.1 & 38 & 12.9 & 5.5 & 2.8 & & 55 & 11 &
+ w4a & 49.50 , 42.3 & 4.6 & 19 & 7.4 & 2.8 & & & 28 & 5.5 & + w5 & 50.29 , 44.1 & 11 & 65 & 15.5 & 9.4 & 3.6 & 2.4 & 94 & 19 & 5.9 + r1 & 50.98 , 45.2 & 13 & 81 & 18.5 & 11.8 & 3.8 & 2.2 & 12 & 24 & 5.5 + r2 & 51.33 , 44.4 & 15 & 82 & 22.3 & 11.9 & 4.2 & 1.3 & 118 & 24 & 3.3 + r3 & 51.30 , 46.2 & 9.5 & 65 & 13.0 & 9.4 & & & 94 & 19 &
+ r4 & 52.06 , 48.1 & 6.9 & 41 & 10.4 & 6.0 & 2.3 & 0.7 & 60 & 12 & 1.7 + r5 & 52.55 , 48.8 & 7.3 & 40 & 10.6 & 5.8 & & 0.9 & 58 & 12 & 2.3 + r6 & 52.77 , 46.2 & 6.0 & 23 & 9.5 & 3.3 & & & 33 & 6.7 & + r7 & 53.06 , 48.1 & 5.9 & 39 & 8.4 & 5.7 & 2.2 & 2.0 & 57 & 11 & 4.9 + e1 & 53.66 , 49.7 & 4.3 & 29 & 6.1 & 4.2 & & & 42 & 8.4 & + e1a & 53.86 , 48.1 & 3.4 & 10 & 6.4 & 1.5 & & & 15 & 2.9 & + e2 & 54.00 , 51.4 & 8.1 & 54 & 11.5 & 7.8 & 2.4 & & 78 & 16 & + e3 & 54.82 , 53.4 & 2.7 & 16 & 3.8 & 2.3 & & & 23 & 4.6 & . | we present 7 mm very large array continuum images of the starburst galaxy m82 . on arcsecond scales ,
two - thirds of the 7 mm continuum consists of free - free emission from hii regions . in the subarcsecond resolution map ,
we identify 14 compact sources , including 9 bright hii regions with @xmath0 .
four of the hii regions have rising spectra , implying emission measures @xmath1 @xmath2 pc . except for one compact source with peculiar features ,
all other compact radio sources are found in dust lanes and do not have optical or near - infrared continuum counterparts .
four regions of extended , high brightness ( em @xmath3 ) radio emission are found in our high resolution map , including some as large as @xmath42 , or 30 pc , representing either associations of small hii regions , or sheetlike structures of denser gas .
the good correlation between 7 mm emission and _ spitzer _
irac @xmath5 continuum - removed pah feature suggests that pah emission may track the recently formed ob stars .
we find an excellent correlation between molecular gas and star formation , particularly dense gas traced by hcn , down to the @xmath4 45 pc scale in m82 .
we also find star formation efficiencies ( sfes ) of 110% on the same scale , based on co maps .
the highest sfe are found in regions with the highest dense gas fractions . |
during the past twenty years , there has been a developing appreciation of the strong biases against finding galaxies of low surface brightness .
these biases arise because the night sky is not particularly dark .
airglow , zodiacal light , and undetected stars and galaxies combine to create a optical background whose surface brightness is as bright as the extrapolated central surface brightnesses of large spiral galaxy disks .
with such a bright background , the ability to detect a galaxy depends not only upon the integrated luminosity of the galaxy , but also upon the contrast with which the galaxy stands out above the poisson fluctuations in the background .
a compact , high - surface brightness galaxy might be quite easy to detect , while another galaxy with the same total luminosity but with a much more extended , low - surface brightness structure would be difficult to find . while astronomers routinely consider the limiting magnitude of their galaxy catalogs , only recently have they begun to consider the ways in which surface brightness selection effects shape existing galaxy catalogs . as an example , freeman ( 1970 ) had initially showed that spiral galaxies share the same central surface brightness , with very little dispersion : @xmath11 .
disney ( 1976 ) , however , convincingly argued that because of the limiting surface brightness of the sample , one would not expect to have detected galaxies with lower surface brightnesses , and thus that there was no particularly compelling reason to believe that the freeman law reflected the intrinsic properties of spiral galaxies . while allen & shu ( 1979 ) later showed that the cutoff at bright surface brightnesses does indeed reflect an actual fall - off in the number of galaxies with increasing surface brightness , they concurred with disney s appraisal of the possible role of selection biases in shaping the faint end of freeman s surface brightness distribution .
thus , disney demonstrated that what was thought to be a general property of disk galaxies could in large part be explained by selection biases against finding low surface brightness galaxies ( lsbs ) .
disney s paper was the genesis of the modern efforts both toward understanding the role that surface brightness selection effects play in shaping existing galaxy catalogs and toward overcoming those biases in newer surveys .
although disney s work strongly suggested that selection biases could reproduce the freeman law , it presupposes that there exists a population of low surface brightness galaxies to be biased against .
it may have been that there really were almost no galaxies with central surface brightnesses fainter than @xmath12 . over the following years and extending through the present , however , there has been an enormous body of observational work that has conclusively demonstrated the existence of low surface brightness galaxies .
indeed , every time a new survey has been extend to fainter surface brightnesses , new lsbs have been discovered .
previous surveys which have been sensitive to lsbs have fallen into two basic types : large area photographic surveys and deep , small area ccd surveys . of the first type , the older ,
diameter - limited upsalla general catalog of galaxies ( ugc ) ( nilson 1973 ) had been the workhorse through most of the 70 s and 80 s .
it has been recently supplemented by a catalog of lsbs selected visually from the second palomar sky survey using the same selection criteria as the ugc catalog but with deeper plates ( schombert et al .
1992 , schombert & bothun 1988 , herein referred to as the `` poss - ii '' catalog ) , and now by the apm lsb survey which selected lsbs automatically and by eye using scanned plates from the apm galaxy survey ( impey et al . 1996 ) .
other deep photographic work has been used to search for lsbs in clusters ( impey et al .
1988 , davies et al .
1988 , irwin et al . 1990 , ferguson & sandage 1988 , sandage & binggeli 1984 ) . with the exception of the deeper cluster surveys
, the photographic work finds very few galaxies whose peak surface brightness is fainter than @xmath13 in @xmath14 ( corresponding to @xmath15 in @xmath1 , using mean colors from de blok 1995 , mcgaugh 1994 ) , unless they also have high surface brightness bulges .
the second type of lsb survey , the ccd surveys , have concentrated primarily on identifying lsbs within clusters of galaxies ( turner et al .
1993 ) . due to their small area ,
these surveys are restricted to finding galaxies of small angular size , which leads to a large degree of confusion between intrinsically low surface brightness galaxies and either high - redshift normal galaxies which appear to have low surface brightness because of @xmath16 cosmological dimming , or pairs of barely resolved galaxies .
a similar field survey by schwartzenberg et al ( 1995 ) also suffers from the same problems , although to a slightly lesser extent due to a larger angular size cutoff .
however , the difficulties in untangling the true surface brightness and angular size distributions are compounded by effects of poor seeing and the large pixel size of these surveys
. there has also been a promising recent foray into using ccds to survey for faint field lsbs through cross - correlating an image with a model galaxy profile ( davies et al.1994 )
. however , the areas surveyed so far are still too small to do more than to verify the promise of this method . in spite of the limitations ,
existing surveys have begun to advance our understanding of the density of the lsb population .
they have both proven the existence of lsbs , and provided a rich data set for studying the properties of individual lsbs .
they have also begun to yield quantitative results about the size of the lsb population .
first , several existing photographic and ccd cluster surveys have be used to estimate the relative number density as a function of surface brightness .
mcgaugh ( 1996 ) has used the selection criteria of the davies ( 1990 ) sample to estimate the relative survey volume as a function of central surface brightness , and shown that the distribution of disk surface brightnesses is very broad .
the derived distribution of surface brightnesses cuts off sharply brighter than the freeman surface brightness value , but falls only slowly with decreasing central surface brightness ; by @xmath17 , the number density drops by only a factor of 2 - 4 , but is statistically consistent with a flat distribution in surface brightness .
however , this analysis assumes that the distribution of intrinsic disk scale lengths is independent of surface brightness , which does not seem to be true ( i.e. the increase in the maximum scale length with decreasing surface brightness
see de jong & van der kruit 1994 , for example ) .
thus , there may be systematic offsets in the accessible survey volume as a function of surface brightness , leading to substantial uncertainties in the relative number density as a function of surface brightness .
more recently , the large apm lsb survey has reconstructed the luminosity function of lsbs with central surface brightnesses of @xmath18 , using a detailed quantification of the automated detection algorithm ( sprayberry 1997 ) .
the survey has 332 galaxies with redshifts , and has @xmath19% completeness in redshift identification for galaxies with @xmath20 .
they find that , when compared to the cfa survey luminosity function ( marzke et al 1994 ) which covers a nearly disjoint surface brightness range , the lsbs in their survey add rougly a third to the known luminosity density in galaxies with @xmath21 , and more than double the known luminosity density contributed by known galaxies with irr / im morphologies . in the survey described in this paper , we attempt to combining the best features of previous surveys , and to push observations to lower surface brightnesses than have been previously surveyed in the field .
we use ccds to observe a large area of the sky , obtain extremely accurate flat - fielding through the use of drift scanning , use completely automated detection to allow for a complete and detailed understanding of the selection biases in the survey , and search for galaxies with large angular sizes to reduce the background confusion and to identify lsbs which are most comparable to well - cataloged normal galaxies in large local surveys . in the following section , [ strategy ]
, we outline the basic strategy of our survey , and discuss how it is optimized for reaching the above goals . in [ datareduction ]
we describe the data we have used , and the data processing and selection criteria in detail . in [ properties ]
we discuss the properties of the entire sample of objects detected in the survey , and in [ lsbsamp ] we focus our attention on the properties of the subset of lsb galaxies .
finally , in [ numberden ] we use the lsb subsample to measure the surface density , number density , and luminosity density of lsb galaxies , and compare to the population to normal galaxies .
we discuss our estimation of errors in appendix a. in appendix b is a general discussion of reconstructing the number and luminosity density of angular diameter limited samples from the observed surface density .
we use @xmath22 throughout .
there are two essential elements to the survey which we have carried out , namely the use of time - delay - and - integrate ( tdi ) observations to achieve extremely accurate flat - fielding over large areas and the use of optimal smoothing filters to enhance the detection of large lsbs . in a tdi scan ( also known as `` strip'',``drift '' , or `` transit '' scanning ) ,
the telescope is held fixed , while the sky drifts across the field of view of the ccd camera , while lines of charge are stepped from row to row of the ccd to keep track with the sky .
drift scanning averages out the flat - fielding variations along the columns of the chip , and means that each line of the resulting strip of data has the same flat - fielding ( or `` flat - lining '' ) characteristics , up to the temporal stability of the chip response . millions of lines of the data can be combined to create an extremely accurate one - dimensional sky flat for the entire data set , much in the same manner that shift - and - stare techniques can be used to make two - dimensional sky flats .
drift - scanning therefore provides an extremely efficient way of using ccds to observe large areas of the sky while providing excellent flat fielding . with
the advantage of extremely accurate large scale flat fielding , we can smooth the data over large scales to reveal low surface brightness galaxies which are typically at the limits of detectability , due to poisson pixel - to - pixel fluctuations in the sky level . to increase sensitivity to galaxies whose surface brightnesses are comparable to the sky brightness
, the pixel - to - pixel noise can also be reduced by averaging together many pixels , increasing the effective area of the pixels . while this decrease in noise may be achieved by observing with large ccd pixels or by binning down the charge on the chip , it
may also be reached by smoothing high resolution data .
there are several advantages to using smoothing to reduce the pixel - to - pixel noise .
first , the smoothing may be done with a filter that is matched to the shape and size of the galaxy for which one is looking ; this produces the maximum gain in signal - to - noise for the targeted galaxies .
second , because the original , high - resolution data still exists after smoothing , no information is really lost ; the original data can provide additional information about the sources of any low - surface brightness features detected after smoothing .
if the smoothing were to have been done by binning the ccd during readout , and not during post - processing , the increase in signal - to - noise would have been optimized for square galaxies , not realistic ones , and most importantly , there would be much more confusion about the source of any fluctuations .
finally , because smoothing allows any filter to be used , one data set may potentially be used to independently survey for galaxies of many different shapes and angular sizes . to avoid contributions from bright foreground objects ,
we have cleaned images of bright foreground objects , such that subsequent smoothing truly recovers large scale fluctuations in the optical background .
because the known population of lsbs seems to be well fit by exponential disks ( de jong 1996 , de blok et al 1995 , mcgaugh & bothun 1994 , knezek 1993 ) , we have chosen to use an exponential profile as the filter kernel .
we have smoothed with an exponential filter of scale length @xmath23 , which drops the level of fluctuations by over a factor of ten , a gain that would only have been achieved with over a hundred - fold increase in exposure time . by choosing such a large scale length
, we eliminate any confusion between low surface brightness galaxies and distant , cosmologically - dimmed high surface brightness galaxies . for example , galaxies with an apparent scale length of @xmath24 are likely to be within @xmath25 ( @xmath26 , assuming an intrinsic scale length of @xmath27 ) and are near enough that @xmath28 surface brightness dimming is negligible .
this choice of scale length also identifies lsbs within a volume comparable to that which is occupied by ngc and ugc galaxies ; a normal galaxy with a @xmath23 exponential scale length has an angular diameter of close to @xmath29 ( assuming @xmath30 and @xmath31 ) , similar to the angular diameter limit of large photographic surveys . to identify candidate lsbs
, we identify all large regions of connected pixels in the smoothed image which are significantly above the level of the remaining fluctuations in the sky . the entire process
cleaning , smoothing , and searching is entirely automated , and is repeated with artificial galaxies added to the data to allow a thorough investigation of the selection biases of the survey .
we then visually inspect all candidates to identify the subset which are lsbs .
we perform additional artificial galaxy tests to test the limits of our ability to classify lsbs .
we use these limits to further restrict the lsb subset to the region where our detection and classification efficiencies are well - determined .
finally , we measure redshifts for the final lsb subset , allowing us to measure the absolute number density of the lsbs .
the survey utilizes existing imaging data , originally taken for the palomar transit grism survey ( schneider , schmidt , & gunn 1994 ; hereafter ssg ) for high redshift quasars
. we will give a brief summary of the initial data processing of this data set ; see ssg for a detailed description of the survey .
the data were taken with the 4-shooter camera ( gunn et al .
1987 ) at the cassegrain focus of the palomar 200-inch telescope .
the 4-shooter detector consists of 4 800x800 texas instrument ccds , arranged in a 2x2 grid , with a pixel scale of @xmath32 .
the telescope was operated in transit mode , holding the telescope fixed and allowing the sky to drift across the ccds while the ccd was clocked to keep pace with the passage of the sky .
the drift scan produces four very long images at a fixed declination covering a wide range in right ascension , each strip corresponding to one of the four chips .
the strip is cut into 400x1600 images with 120 rows of overlap between adjacent images .
because of difficulties with the high data flow rate , the ccds were double - clocked and the columns rebinned by a factor of 2 , to increase the effect of pixel scale to @xmath33 .
the seeing was @xmath34 .
the two leading ccds used the @xmath35 `` wide @xmath1 '' filter , while the two trailing ccds used @xmath36 ( griffiths 1990 ) . following postman et al .
( 1996 ) , we will use the shorthand notation of @xmath37 and @xmath38 for the combination of the hst filters and the 4-shooter response function .
only the @xmath37 strips were used , because of the greater susceptibility of the @xmath38 band to fringing and to low level fluctuations in the sky brightness .
we specifically use the qr and mn strips from the ssg survey , observed in april 1987 and 1988 , respectively .
( the two - letter notation reflects the fact that each strip is actually composed of two separate strips , corresponding to the two halves of the 4-shooter s ccd array .
we analyze each of the four substrips ( m , n , q , & r ) independently . )
the qr strip is at a declination of @xmath39 , and extends from @xmath40 in right ascension .
the mn strip is at a declination of @xmath41 , and extends from @xmath42 in right ascension . as described in more detail in ssg , a single bias level , measured in the overscan region ,
was subtracted from each image . following this ,
the lines of all of the images were combined to create a sky - flat for the ccd .
the flat did not change significantly over the course of three years .
after each image was flattened , a cubic polynomial was fit to the median sky level along the scanning direction and then subtracted to remove any large temporal fluctuations in the image .
further corrections of the temporal variations are described in [ flattening ] .
finally , the two substrips were calibrated against each other by requiring the same mean sky level in each pair of adjacent images .
each image was further overscan corrected by using a five piece cubic spline fit to the overscan region . following this ,
the images were trimmed to remove the bias strip , leaving 372x1598 images .
the imaging data were not taken in perfectly photometric conditions , and were initially not photometrically calibrated . however , strip scans allow one to easily detect the presence of cirrus . first , because the strip is a scan both in time and in right ascension , any temporal variations due to the presence of cirrus is easily recognized as structured banding within an individual image , due to the fluctuating brightness of the sky produced by the changing degree of reflection of the light from san diego .
second , the passage of cirrus dramatically drops the number of objects seen in a given image . based upon the number counts of postman et al .
( 1996 ) using the identical observational system , if the depth were to change by 1 magnitude , the number of detected objects would change by a factor 3 .
in contrast , in the data which we used , the number of objects never varies by more than 25% ( either rms or systematically ) , suggesting that the depth never varies by more than 0.3 magnitudes .
in fact , the variation in the depth is certainly even smaller , given that half of the fluctuation amplitude is due to simple poisson counting , and that there are additional fluctuations in the area lost to bright foreground stars . using the constancy of the sky brightness and of the number of objects per field as a measure of atmospheric clarity ,
we have restricted the data to the portions of the night which are most likely to be photometric . when reduced to the most photometric portions of the night ,
the qr substrips consist of 179 400x1600 images in each band between @xmath43 .
the mn substrips consist of 283 400x1600 images in each band .
twenty - eight of the mn images near @xmath44 are unusable due to the passage of clouds .
a total of 868 images were used in the survey .
the total area covered by the 4 individual strips is @xmath45 . as a test of our judgement , and as a means to calibrate the zero point of the palomar data , we have taken calibration images of many fields within the strip , under photometric conditions , and find no significant variation in the zero point of the strip scan data in the portions of the night to which we have restricted the data .
we have calculated the zero points for the mn and qr strips using calibration data taken in johnson @xmath1 with the kitt peak 0.9 m telescope in march of 1994 .
the johnson @xmath1 zero point of the kitt peak data was calibrated and airmass corrected using 6 independent observations of 8 different standard stars in two separate landolt ( 1992 ) fields , with an 0.015 magnitude rms in the airmass corrected zero point . for the mn strip
, we have johnson @xmath1 images of 7 different fields throughout the strip , 3 of which were observed more than once .
these images were used to calculate @xmath1 band magnitudes for objects which fell in the mn strip . comparing to the count rate for objects in the ssg data to their @xmath1 magnitudes in the kpno data , we calculate a johnson @xmath1 zero point for the mn strip of @xmath46magnitudes per data number , with no systematic difference between the zero points of the two halves of the strip .
we will quote johnson @xmath1 magnitudes for the remainder of the paper , for ease of comparison to other lsb samples ; although the data was taken in @xmath37 , the color term to correct from @xmath37 to johnson @xmath1 is small , @xmath47 , for @xmath48 ( postman et al .
the rms scatter among the 10 calculated zero points of the ssg data agreed with the expected uncertainty in the mean , based upon the uncertainties in the zero points calculated for the individual fields , which were in turn based upon the rms scatter in the airmass corrected zero point of the kpno data and in the zero points calculated for the stars in the kpno transfer standards .
the stability of the zero point throughout the mn strip suggests that the atmosphere was uniformly transparent for the portion of the night which met our criteria for being most photometric .
the zero point for the mn strip is also consistent with the conversion used for the palomar distant cluster survey ( postman et al .
1996 ) , which uses the same instrumental configuration . for the qr strip
, we have only one calibrated field , giving a zero point of @xmath49 magnitudes per data number in johnson @xmath1 for the qr strip . however , rather than use a single measurement to calibrate the entire qr strip , we will use the zero point of the mn strip for the qr data as well , and recognize the possibility of an 0.2 magnitude offset between the two data sets .
we note that the surface brightnesses probed in this paper are up to @xmath50 fainter than normal galaxies , and thus an error in the photometric calibration of less than half of a magnitude will do little to change the conclusions .
furthermore , given that the number of detected objects per frame and the sky brightness both vary as smoothly in the qr strip as in the mn strip , we do not anticipate finding field - to - field variations in the qr strip zero point which are significantly greater than those seen in the mn strip .
producing a map of the background sky requires the removal of all of the high - surface brightness stars and galaxies which are obstructing the view of the background . to identify the intervening objects , focas was used to identify all regions of connected pixels greater than 4 sigma above the mean sky , with a detected area of greater than six pixels after smoothing with focas s built - in filter ( jarvis & tyson 1981 , valdes 1982 ) .
the number of objects found per unit magnitudes began to deviate from a power law at @xmath51 , suggesting that the focas selection is incomplete beyond this point . in preparation for cleaning , the region
that focas had associated with each object was extended by an amount derived from the object s detected area and magnitude .
this ensured that any low - surface brightness halo , due to either seeing or to a slowly falling galaxy profile , was included with the central bright regions .
the pixels within the object were replaced with values drawn from the local sky histogram , created by sampling within a buffer around the extended region and clipping at 4 sigma above the local mean .
this cleaning procedure preserves the large scale distribution of surface brightness .
objects which have large areas for their magnitude ( or low magnitudes for their size ) are excluded from cleaning to avoid cleaning any brighter lsbs which focas may have been able to detect without additional processing . while replacing objects with the local sky histogram proved to be effective for objects fainter than roughly 19th magnitude , cleaning brighter , larger objects ,
would have invested tremendous amounts of computational time in a region guaranteed to be useless for detecting background objects . instead , large regions around bright objects were removed from the survey by creating masks for each image , which were used throughout processing . occasionally , the area of the masked region was sometimes too small to completely block out the halo of the masked object .
this problem was most commonly manifested in objects on the edges of images , and very occasionally in highly elongated non - boundary objects such as edge - on galaxies and meteor trails .
cases where the masking was insufficient were easily identified in the final catalog and removed from consideration .
the two columns on the left and right edges , and fifteen and thirty rows on bottom and top edges respectively were masked as well , to eliminate the most common regions of incomplete focas identification from consideration .
roughly 39% of the survey area is lost to the masked regions . in the final cleaned image ,
the masked regions were set to the median sky level of the unmasked regions .
the mask and the cleaned image were then rebinned by a factor of two to save space
. in drift scans , there are low amplitude row - to - row fluctuations due to temporal variations in the sky brightness ( @xmath52 ( 2 - 4 adu ) peak - to - peak , when smoothed on scales of @xmath53 ) .
these fluctuations are constant across a given row , and vary smoothly from row to row , creating very low level bright and dark `` bands '' across the strip scan .
the efficiency of the survey is compromised by the fluctuations ; significant low - surface brightness objects that lie in low points are more likely to be missed , while fainter objects that lie on the high points will be detected more frequently .
( although monte carlo simulations can measure the variation in efficiency , they can not compensate for the lost survey area . )
we flatten the overall sky background in each image by : ( 1 ) setting the masked regions to zero , ( 2 ) taking the mean of each row of the masked image and of the mask itself , ( 3 ) smoothing each 1-dimensional profile with a boxcar filter @xmath54 wide ( roughly 13 seconds , temporally ) , ( 4 ) dividing the smoothed image profile with the smoothed masked profile , and ( 5 ) subtracting the resulting 1-d profile from each column of the original image .
this produces a significantly flatter image with a roughly zero sky level .
correcting for large scale variations reduces the survey s sensitivity to objects of similar scales .
we have not lost much sensitivity in the case of this survey however , as the geometry of the strip makes it unlikely that many objects with scales of @xmath55 or larger could be detected in the first place .
because of the difficulty in treating boundaries in any smoothing problem , the flattening tends to become inaccurate at the beginning and ending of the image , and adjacent to regions where the entire width of the image has been masked out .
these regions were obvious in the final smoothed image , and were eliminated from consideration to avoid false detections . occasionally there is residual row - to - row structure left in the image after flattening , on scales smaller than the flattening scale .
these are also due to temporal variations on scales smaller than 13s , and are typically fainter than @xmath56 . after removing temporal variations ,
occasionally there are diagonal structures which can not be due to temporal variations , at the @xmath57 level .
the most likely sources of this excess structure are halos of stars off the field of view and emission from galactic cirrus .
our current data are not adequate for addressing these questions , but repeated , multi - color transit scan observations of a single region , with small random offsets in declination between scans and a larger fov , would help to constrain the origin of any residual structure . the cleaned , rebinned , and flattened images produced by the previous steps are remarkably featureless . any residual flattening inaccuracies are invisible , buried well below the pixel - to - pixel variations in the image .
smoothing the cleaned image reduces the pixel - to - pixel noise dramatically , uncovering low surface brightness features . after replacing the masked regions with the image median
, we smooth the image with a radially symmetric exponential profile ( @xmath58 ) with @xmath59 , which optimizes the signal - to - noise ratio increase for objects with exponential profiles identical in size and shape to the smoothing filter . smoothing with the @xmath23 filter
reduces the pixel - to - pixel noise by roughly a factor of twelve ( to @xmath60 adu ) . smoothing with
a larger smoothing length of @xmath53 reduced the level of fluctuations by an additional factor of only 1.6 , indicating that on the scale of the smoothing kernel ( @xmath61 ) the fluctuations in the images are not random , due to the very low level fluctuations discussed in flattening .
the histogram of pixel values for the @xmath53 smoothed images also show deviations from the expected poisson distribution . because of this indication of residual structure on @xmath62 scales ,
we chose to use only the single scale length to smooth our data ; with a larger field of view and reduced temporal fluctuations and scattering , a larger range of smoothing lengths would be more useful than with this particular data set .
a modified version of focas was used to detect all regions in the smoothed image greater than @xmath63 above the image median over an area of 25 pixels , corresponding to 11.2 arcsec@xmath64 .
the resulting focas catalog was restricted to eliminate any object touching a boundary ( the only negative `` dark '' objects found were also boundary objects ) or whose center was within @xmath65 ( where @xmath66 is the intensity weighted second moment radius ) of the left or right edges , or within @xmath67 of bottom edge or @xmath68 of the top edge .
the buffer on the upper edges is larger to avoid detecting objects in the overlap region in more than one image . with these selection criteria ,
868 objects were detected , giving an average of 1 candidate per each frame .
we expect the total catalog to include only one or fewer spurious detections due solely to poisson fluctuations ; this is based upon the number of independent detection areas in the survey , after correcting the survey area for the masked areas and the buffered regions along the edges of the images , and upon the peak height of an exponential profile which is @xmath69 above the mean at a radius of @xmath70 . the absence of `` dark '' features and the low poisson probability of false positives suggests that effectively all of the detected features are due to true enhancements in the the mean sky level ( although those may in some cases be due to other than astronomical sources , as discussed below ) .
a single image was created for each candidate object , consisting of three adjacent panels ( see figure [ lsbfig ] for an example ) .
the first panel is taken from the original uncleaned , unprocessed image .
the second panel is taken from the cleaned image , and has been expanded by a factor of two to match the angular scale of the first panel .
the third panel is taken from the smoothed image in which the object was detected , also adjusted to the same angular size as the other two panels .
all three panels are centered on the object .
these composite images allow easy browsing of the catalog ( with much less stringent disk space requirements than saving all the processed data ) , and were used to assign a classification to the objects , if possible .
classifications were based entirely upon visual inspection of the images .
the classification was given as combinations of the categories below , based upon the associated criteria .
the percentages in the brackets are the percentage of objects that have the category as a possible classification and the percentage of objects that have the category as their only possible classification .
\(l ) low surface brightness galaxy exhibits smooth , extended emission in the original image .
[ 4.1%,2.4% ] \(c ) cluster shows evidence that the low surface brightness emission is associated with several compact clumps , either clearly visible or just below the threshold for detection ( see dalcanton 1996 ) .
the classification typically reflects one of three cases , corresponding to clusters at different distances : 1 ) the light is the low - surface brightness component of a nearby cluster that can be clearly identified by large numbers of galaxies in the field ; 2 ) the emission is centered on a single galaxy with several , barely visible companions , probably corresponding to a cd galaxy of a more distant cluster ; 3 ) the emission is centered on a clump of several faint knots of emission , which are probably the centers of distant cluster galaxies just below the threshold of detection .
objects classified as clusters tend to be near other clumps in the smoothed sky maps . in the case of the low redshift clusters , the correlation is due to detecting pockets of dwarf galaxies and tidal debris throughout the cluster region . in more distant clusters ,
the correlation is most likely due to the presence of subclustering , or to the high amplitude of the cluster - cluster correlation function .
this clustering behavior was clearly seen in the recovery of the @xmath71 cluster ms1054 in dalcanton ( 1996 ) , where two or three strong peaks were seen in the smoothed sky background , centered on regions of truly high galaxy concentration .
[ 24.3%,19.9% ] \(t ) tidal features
shows evidence that the low surface brightness emission might be due to interactions between galaxies .
used if the emission lies between two close galaxies or is adjacent to a single galaxy that shows signs of being disturbed , especially if there is no other evidence that the light might come from galaxies below the detection limit .
the boundary between this class and the cluster class is not always distinct . [
2.0%,1.1% ] \(g ) groups of galaxies tight groups of foreground galaxies without extended low surface brightness emission . these probably entered the catalog through contamination of the sky histogram of one galaxy with the bright pixels of an adjacent galaxy .
however , they form a potentially interesting subclass , and may actually contain true members of the cluster and tidal classes , and thus are not classified as errors .
[ 5.4%,2.8% ] ( ? ) uncertainty represents uncertainty in the classification .
( ) unclassified no evidence for making a classification .
typically has no strong associated foreground emission , or no compelling reason to group the object with one class over another .
[ 52.3%,52.3% ] ( * ) errors cases where the object s inclusion in the catalog was more a product of the processing than the object s internal properties .
these cases include incompletely cleaned or incompletely masked objects , or objects which have only been included because they were superimposed on either a halo from an object off the field of view or on regions of inaccurate flattening in the upper and lower edges of the image .
[ 16.2%,16.2% ] many of the detected objects have more than one possible classification .
this is an unavoidable consequence of the detection procedure .
while smoothing the cleaned images increases the signal - to - noise to the point where faint features can be detected , the signal - to - noise of the original image , which is used for classification , remains unchanged .
thus , in many cases there is no unambiguous classification , in spite of a significant detection of an ebl fluctuation .
for example , there are cases where the feature in the smoothed image does not correspond to any obvious emission in the foreground image .
there are other cases where , although the detected region corresponds to one or more high surface brightness features in the original image , the low - surface brightness component responsible for the detection is too faint for any morphology to be visible which might provide a clue as to the proper classification .
note that because classification requires more signal - to - noise than detection , classified objects are even less likely to be spurious .
there are many cases where , in spite of reasonable signal - to - noise , the classification is ambiguous .
an lsb galaxy with a faint bulge or with several hii regions is possibly similar in appearance to a distant cluster with a faint cd galaxy , or even to a tidally disturbed normal galaxy .
because of this confusion , when we create a final catalog of objects which have lsb as their only possible classification , we are underestimating the true contribution of lsbs .
because of their ambiguous classification , we systematically exclude lsbs with high - surface brightness features ( i.e. bulges , resolved stars , hii regions ) from our catalog of definite lsbs , leaving us with a `` pure disk '' subsample of the larger lsb population ( e.g. see profiles in figure [ profilefig ] ) .
this is in contrast to visually selected , angular - diameter limited catalogs of lsbs , such as the possii and ugc catalogs ; a substantial fraction of the lsbs in these catalogs have strong central bulges ( for example , see the profiles in mcgaugh & bothun 1994 ) .
for the remainder of the paper , we will concentrate on the 21 objects which were classified as candidate lsbs based upon their optical morphology . of these
, 14 were classified as definite lsbs , and an additional 7 were classified as probable lsbs . in order to determine the central surface brightness and scale length of the lsb candidates , we fit the 21 lsb candidates interactively with iraf s `` ellipse '' surface brightness photometry package .
we then further restricted the lsb subsample to include only the 14 lsb candidates whose extrapolated central surface brightnesses were fainter than @xmath72 in @xmath1 ( corresponding to roughly @xmath73 in @xmath14 , using the median lsb color of @xmath74 from mcgaugh & bothun ( 1994 ) and de blok et al .
( 1995 ) ) ; because we did not use any photometric information in the initial classification , galaxies classified as potential lsbs were sometimes not dramatically fainter than the freeman value , and within the range of surface brightnesses that can be better constrained with the much larger apm survey ( impey et al .
1996 , sprayberry et al .
the remaining sample spans a range in central surface brightness of @xmath75 in @xmath1 and scale lengths of @xmath76 .
while the profiles are noisy ( particularly at large radii ) due to the very low signal - to - noise in the unsmoothed image , they are consistently better fit by exponential surface brightness profiles than by devaucouleur s profiles .
we estimated the uncertainty in the measured central surface brightness and exponential scale length by performing the identical ellipse fitting and profile fitting on artificial galaxies with the measured properties of each of the lsbs . for galaxies with @xmath77 , @xmath78 , and for galaxies with @xmath79 , @xmath80 . for galaxies with @xmath81 ,
@xmath82/\alpha = -0.006\pm0.09 $ ] , and for @xmath83 , @xmath84/\alpha = -0.1\pm0.12 $ ] . to measure the number density of lsbs
, we must determine the efficiency with which we can both detect and identify lsbs as a function of surface brightness and size . as discussed above ( [ properties ] ) , because our classification procedure effectively eliminates lsbs with bulges from the final sample , we need only to test the recovery of pure disk exponential profile galaxies in order to measure our efficiency . while doing so means that we will not be measuring the efficiency of detecting all lsbs ( i.e. lsbs with and without bulges and/or bright hii regions , which are seen to exist in brighter lsb samples ) , we know _ a priori _ that our classification procedure reduces the efficiency of finding such galaxies in our final sample to zero , given that the surface brightness profiles and images demonstrate that none of these galaxies with more complicated morphologies have made it into the final sample .
thus , the population of lsb disks with associated bulges and bright hii regions is entirely unconstrained by this paper , and the measured number density will be an underestimate of the true lsb population . furthermore , because the galaxies in our sample are all smooth exponential disks with low ellipticities ( with the possible exception of q-42 - 1 ) , measuring the efficiency of the survey requires only that we test the recovery of similarly smooth , face - on , exponential disks . to first measure our ability to classify lsbs , we added 100 artificial lsbs to various positions throughout the original imaging data and then clipped out the region containing the lsb , creating an image identical in size to the images which were used for classification in the actual survey .
the pure exponential face - on disks used as test lsbs were morphologically indistinguishable from the lsbs in the sample , which also have pure exponential profiles and low ellipticity .
as a result , the images of the artificial lsbs appeared remarkably similar to the true lsb images ( see figure [ lsbfig ] ) .
we inspected and classified the artificial lsbs , in the same manner as the actual survey data .
figure [ classfig](a ) shows the resulting classifications as a function of exponential scale length and surface brightness ; the triangles were classified as lsbs and the diagonal lines are lines of constant signal to noise ( i.e.@xmath85 , with varying @xmath86 , and @xmath7 measured in arcseconds ) .
figure [ classfig](b ) shows the resulting classification efficiency as a function of signal - to - noise ( characterized by @xmath86 ) , for @xmath87 ( for scale lengths smaller than @xmath88 , the classification ability drops dramatically , due to confusion between single , low redshift lsbs and pairs of faint , distant normal galaxies ) .
our classification efficiency is essentially 100% for @xmath89 , and zero for @xmath90 .
we approximate the efficiency in figure [ classfig](b ) analytically as @xmath91 which has a residual rms of 0.094 for @xmath92 . based upon equation [ classeqn ] , we make the final restriction on the subset of lsbs , retaining only those galaxies for which @xmath87 and @xmath93 .
this leaves a final lsb catalog of 7 galaxies which have securely measured classification efficiencies .
the classification images of these lsbs are shown in figure [ lsbfig ] ( including also the cleaned and smoothed images , as described in [ properties ] ) .
their radial profiles are shown in figure [ profilefig ] , and their structural properties are listed in table 1 .
in addition to our efficiency in classifying lsbs , we must also quantify our ability to detect lsb candidates in the first place . to do so , we used monte carlo simulations to measure the efficiency with which our automatic detection software identified the smoothed fluctuations produced by artificial galaxies , as a function of central surface brightness and exponential scale length . first , a single artificial , face - on galaxy with an exponential profile was added to an image from the survey .
the artificial galaxies were drawn from a uniform distribution in @xmath94 and central surface brightness , between @xmath95 and @xmath96 , above a fixed signal - to - noise ( @xmath97 ) .
the altered image was processed in an identical manner to the original survey data .
the resulting catalog of low surface brightness objects was searched for any detection that coincided with the artificial galaxy .
this process was repeated five times for each image in the survey .
the small surface density of true objects ( @xmath98 per frame ) implies that confusion and overlapping of low surface brightness objects is not a source of selection biases in the survey .
thus , the monte carlo simulations were designed to test the recovery of isolated galaxies , by requiring that there was at least a @xmath62 buffer between an artificial galaxy and the nearest candidate lsb . however , the detection efficiencies are modulated by any residual large scale variations in the background sky level , with more objects being detected in areas where the local sky is higher than average . therefore ,
if too large a buffer is required between the artificial galaxies and the real objects , then the artificial galaxies will be systematically biased away from these regions of increased detection efficiency .
the largest fluctuations occur on scales of @xmath99 , so by limiting the buffer to @xmath62 ( substantially smaller than the width of the strip ) we have allowed the artificial galaxies to sample the same sky levels as the real detections .
figure [ detectionfig](a ) shows the resulting detection efficiency for our survey , as a function of surface brightness and disk scale length .
superimposed on the contours of constant efficiency are all the lsbs with @xmath100 , as well as the regions of 100% and @xmath101% classification efficiency . within the region where lsbs can be reliably classified ,
the detection efficiency is well behaved , as can be seen in figure [ detectionfig](b ) , where we have plotted the detection efficiency as a function of scale length , for different central surface brightnesses .
the distribution is reasonably well fit by a gaussian , leading us to use @xmath102}\ ] ] as a reasonable approximation of the detection efficiency for the final lsb subsample .
the residual rms of this approximation is 0.056 .
the maximum detection efficiency is roughly 35% , which results largely from the loss in area behind masks , behind other stars and galaxies , and around the edges of the images .
the detection efficiency peaks close to the scale length of the @xmath23 exponential smoothing filter used in processing .
the contours of constant detection efficiency suggest that lsbs with @xmath103 may be in the full catalog , but be currently unclassified or misclassified as clusters .
we have not to attempted to model edge - on lsbs . including edge - on lsbs could have easily introduced more uncertainty than it would have removed ; we know very little about the vertical scale heights and opacities of lsb disks , particularly at the extreme surface brightnesses explored by this survey . furthermore , none of our detected lsbs is substantially elongated , and thus the detection efficiency for edge - on lsbs is not an essential ingredient for recovering the lsb number density of the smooth face - on lsbs in our sample .
however , since we are strongly biased against finding edge - on lsbs , due to our choice of a circularly symmetric smoothing filter , we may well be underestimating the number density of lsbs by effectively restricting our sample to include mostly face - on galaxies .
follow - up spectroscopy of the final lsb subsample was carried out in may of 1996 using the cryocam on the mayall 4 m at kitt peak national observatories . the 650 grism ( 400 line / mm ) was used in first order with a @xmath104 slit , giving @xmath105 resolution ( fwhm ) between 3500 and 7000 , with an average of 3 per pixel .
the slit was oriented along the major axis of the lsb , using the richley prisms to correct for the differential refraction of the atmosphere .
exposure times ranged from 2700s to 12,000s giving a signal - to - noise of 10 - 40 per resolution element longwards of the the 4000 break , with a mean signal - to - noise of 15 - 20 .
the spectra of the lsbs were overscan corrected , dark subtracted , flat - fielded using domeflats , illumination corrected with twilight flats , and then extracted , sky subtracted , and wavelength calibrated using he+ne+ar spectra taken adjacent to , or sandwiched between , each of the lsb spectra .
the focus of the spectrograph was a strong function of wavelength , and degrades by nearly a factor of two by 3600 .
the resulting lsb spectra were initially cross - correlated with a template spiral galaxy spectrum using iraf s `` rvsao '' package , and searched for emission lines .
the cross - correlation was limited to between 3700 and 5000 to isolate the 4000 break , the balmer lines , and the strong g - band .
since all galaxies were at very low redshift , the correlation included nearly the same region of the rest - frame spectra for all galaxies . using the resulting redshifts ,
the spectra were doppler corrected back to zero redshift .
then , to refine the redshift measurement , we created an improved correlation template for each galaxy , by coadding all the other lsb spectra , excluding the one galaxy spectra to be cross - correlated and the spectra for r-26 - 1 , which has low signal - to - noise and thus no convincing redshift .
the new templates were then cross - correlated with the correct galaxy spectra and a new redshift derived .
the process was repeated a second time using templates derived from the refined redshifts . on the final cross - correlation ,
no redshift changed by more than @xmath106 , and all correlation coefficients had @xmath107 , with the exception of m-232 - 1 , which had @xmath108 , and r-26 - 1 , for which we have no convincing redshift .
the emission line redshifts for r-27 - 1 and q-129 - 2 were within @xmath109 of the cross correlation absorption line redshifts .
we take this velocity difference to be representative of the @xmath110 uncertainty in the individual redshifts , although the formal error from the cross - correlation is smaller ( @xmath111 ) .
similar changes in the velocity were produced by changing the wavelength range and filter parameters used for the cross - correlation , again suggesting that the redshifts given in table 1 are accurate only to several @xmath112 .
in all cases , however , the cross - correlation functions do not have additional peaks which could be considered as alternative plausible redshifts .
the redshift for m-232 - 1 is more uncertain , due to the lower signal - to - noise and as can be seen from the lower correlation coefficient . by eye
, we find a limit of @xmath113 on the error in the redshift .
the final redshifts are listed in table 1 , and the spectra are shown in figure [ spectrafig ] , along with the template spectra used for cross - correlation , all shifted into the rest frame ; because of the low redshifts , the shift was only @xmath114 . for the purposes of analysis ,
we assign a redshift to r-26 - 1 by assuming it has the mean physical scale length of galaxies in the sample ( @xmath115 ) , giving an assumed recessional velocity of @xmath116 .
if r-26 - 1 were to have the smallest or largest physical scale length of the sample , it would have recessional velocities of @xmath117 or @xmath118 respectively .
the lsbs have recessional velocities of 4000 - 9000@xmath119 , and disk scale lengths of @xmath120 ( table 1 , excluding r-26 - 1 , due to its uncertain redshift ) , comparable to the disk scale lengths seen in angular diameter limited field surveys of brighter lsbs ( mcgaugh & bothun 1994 , de blok et al 1995 , de jong 1996 ) and of normal galaxies .
the lsbs in the final subsample are also at comparable distances to normal galaxies catalogued in large local surveys .
the similarity can be seen most readily in figure [ piefig ] , where we have plotted a `` pie '' redshift - position diagram for the final lsb sample ( solid triangles ) and for the galaxies in the zcat catalog of galaxies ( open circles ) .
the lsbs also have scale lengths which are similar to normal galaxies .
both the lsb and normal galaxy samples will be biased towards finding physically large galaxies , which have the largest probability of being identified in angular diameter limited surveys ( see appendix b ) .
figure [ piefig ] shows that there is no strong tendency for the lsbs in our sample to be physically associated with normal galaxies . between @xmath121 and @xmath122 ,
11 out of 23 zcat galaxies are in close associations , whereas 2 in 6 of the lsbs has a near neighbor . because of the small numbers of lsbs and the non - uniform selection criteria for the zcat galaxies , our sample does not provide definitive evidence for the low correlation between lsbs and normal galaxies .
however , it is suggestive of the general trend seen in brighter lsb samples for the small - scale correlation of lsbs to be weak compared to normal galaxies ( bothun et al .
1993 , mo et al .
1995 ) .
the follow - up , spectroscopic observations also provide validation for our classification procedure .
all of the candidate lsbs in the final sample which we observed spectroscopically have characteristic galaxy spectra , smooth light profiles perpendicular to the dispersion direction , and low redshifts , proving that the final sample includes only the nearby lsbs which the survey was designed to find . while our classification procedure may have missed lsbs ( i.e. false negatives ) , our strict selection criteria ( [ detection ] & [ classification ] ) successfully insured that we did not include any non - lsbs in our final candidate list ( i.e. false positives ) .
using the classification and detection efficiencies from equations [ classeqn ] & [ detectioneqn ] we can correct for the area that was effectively lost from the survey and calculate the surface density of lsbs with @xmath9 .
the total efficiency @xmath123 is a measure of the fraction of the total @xmath124 survey area which was actually accessible for detecting an lsb with central surface brightness @xmath6 and exponential disk scale length @xmath7 . using the formalism developed in appendix a
, the surface density of lsbs in our survey with @xmath125 and @xmath87 is : @xmath126 where the errors enclose 90% confidence .
the uncertainty is dominated by the poisson probability for 7 galaxies and a detailed calculation of the confidence interval is provided in appendix a. while a naive calculation of the surface density ( 7 galaxies in 17.5 square degrees ) gives a smller surface density , inspection of figure [ detectionfig](a ) for @xmath87 shows that the detection and classification efficiencies are small over much of region , leading to a large correction .
as shown in appendix b , the surface density of galaxies on the sky increases as @xmath127 towards small angular sizes , and thus most of the correction applied to the measured surface density is a correction for the lost classification efficiency of small galaxies with @xmath128 .
furthermore , because of the steep increase in the number of galaxies with small apparent scale lengths , the surface density is a strong function of the exact lower limit on @xmath7 , and thus lsb surveys with different angular selection criteria will derive very different surface densities . although the surface density may at first seem small , it is in fact significant , given that the large scale lengths to which the sample was restricted effectively confines the survey to a small , nearby volume .
the low surface density also explains why there is a consistent impression that lsbs are rare .
first , one would have to observe a square degree of the sky to be reasonable certain of finding one large lsb .
second , the lsb would appear much smaller than a normal galaxy of comparable scale length , because it would only be visible out to roughly one scale length before its surface brightness drops below the pixel - to - pixel variations in the sky ; normal galaxies are visible out to several scale lengths and thus are more noticible ( disney & phillips 1983 ) . using the distances for the final lsb subsample , the efficiency for detection and classification and the selection criteria for the final sample , we can calculate the effective volume in which each lsb could be detected .
an lsb at a distance @xmath129 , with angular scale length @xmath130 and central surface brightness @xmath131 could have been detected out to a distance of @xmath132 the effective volume of the survey is different for each galaxy , because the detection efficiencies are a strong function of the angular scale length and central surface brightness of the individual galaxies .
using @xmath133 , the volume that the @xmath134th galaxy could be detected within is @xmath135)\ , { d^\prime}^2\,{\rm d}d^\prime \\ & = & \frac{a}{\rm radians}\ , \left[\alpha_i d_i\right]^3 \int_{\frac{\alpha_i d_i}{d_{max_i}}}^\infty \frac{\epsilon(\mu_i,\alpha^\prime)}{{\alpha^\prime}^4}\ , { \rm d}\alpha^\prime,\end{aligned}\ ] ] including the change in detection efficiency with the apparent angular size ; because the efficiency changes with distance , the survey volume associated with a given galaxy is not a simple conical section , but tapers off as @xmath136 approaches @xmath133 .
there is no minimum distance at which an lsb can be detected , because we imposed no upper limit on the angular size of galaxies within our survey .
there is an implicit upper limit set by the width of the survey fields , but the limit is so much larger than the angular scales of the galaxies in the survey that the effective minimum distance is zero . with equations
[ dmax ] and [ v ] , the total number density of lsbs with @xmath125 and intrinsic scale lengths @xmath137 between @xmath138 and @xmath139 is @xmath140 with 90% confidence , assuming the probability distributions derived in appendix a. as with the measurement of the surface density , the uncertainty is entirely dominated by the uncertainty due to the small number of galaxies in the final lsb subsample .
the value of @xmath141 changes negligibly other reasonable redshifts for r-26 - 1 are used . in order to compare the relative contribution of lsbs and normal galaxies , figure [ numdenfig ] shows the integrated number density of lsbs with central surface brightness fainter than @xmath72 in @xmath1 , as a function of limiting surface brightness .
superimposed on the integrated lsb number density are comparable measurements of the integrated number densities of normal galaxies , for several different determinations of the local luminosity function . to make `` comparable '' integrated number densities of normal galaxies we have restricted the integration to either ( 1 ) the range of absolute magnitude covered by the lsbs in our sample ( @xmath142 ) or ( 2 ) the range of intrinsic exponential disk scale lengths of our sample ( @xmath143 ) ( assuming that a normal galaxy s scale length can be determined from its luminosity , if it has a freeman central surface brightness of @xmath144 in @xmath14 and @xmath145 ) .
the corresponding number densities are labelled in figure [ numdenfig ] as `` same m '' and `` same @xmath7 '' , respectively .
as figure [ numdenfig ] shows , the integrated number density of lsbs with central surface brightness fainter than @xmath72 in @xmath1 is _ greater _ than the number density of normal galaxies with either similar luminosities or similar scale lengths .
we consider this result to be extremely robust .
the errors are dominated by the well - understood poisson statistics of the small number of lsbs , and thus , only an egregious , systematic error in the measured distances or in the integrated efficiencies ( eqn [ v ] ) could produce a large enough effect to change our conclusions . in spite of the small numbers of lsbs in the final sample , the large number density of lsbs would persist even if significant fractions of the sample were to be removed .
for example , even between @xmath72 and @xmath146 , the number density of lsbs remains comparable or greater than normal galaxies .
our conclusions hold as well if the confidence intervals are increased to 99% .
we have also considered the degree to which photometric errors could effect our measurement of @xmath141 , by examining @xmath147 as a function of central surface brightness @xmath6 .
brighter than @xmath148 , the effect of photometric errors is very small , such that an error of 1 magnitude in the zero point changes the derived number density by less than a factor of two , due to variations in the integral over the efficiency .
the effect is stronger for fainter central surface brightnesses , where the overall efficiency is lower and more sensitive to @xmath6 .
even so , the required zero point error would need to be larger than the observations admit .
for example , to reduce the number density for @xmath128 by a factor of 10 ( which would still leave @xmath141 comparable to the normal galaxy number density ) , the zero point would have to be systematically 0.9 magnitudes too faint in the fields containing the faintest lsbs . from our photometric calibration
, we believe that our mean zero point is good to at least 0.1 magnitudes , so there would have to be a temporary fluctuation in the zero point of the strip scan while the lsbs were passing overhead .
such a shift would have produced nearly a factor of 3 drop in the mean number of objects per field , based upon the number counts of postman et al .
( 1996 ) and would probably have been accompanied by a brightening of the sky level .
neither of these signatures is seen , given that the data was originally restricted to the parts of the strip scan which were stable in both sky brightness and number of detected objects .
the number of objects never varies by more than 25% in either the rms or the mean , suggesting that the zero point does not fluctuate by more than 0.25 magnitudes , based upon the number counts .
this is a strong upper limit , given that simple poisson statistics contributes 10% to the rms component and that the lost detection area due to bright stars probably makes a similar contribution .
the maximum zero point uncertainty suggests that , for @xmath128 , the most we could have possibly overestimated an individual galaxy s contribution to the number density is a factor of 2 , which is still contained within the 90% confidence interval .
thus , the total measurement of @xmath141 could be high by at most a factor of 2 , and only in the unlikely event that every lsb in the sample had the largest possible zero point error , in exactly the same direction .
we have reasons to believe that , if anything , our measurement of @xmath141 is an underestimate .
first , our choice of a circularly symmetric smoothing filter greatly reduces our sensitivity to edge - on lsbs , and thus we are only sensitive to the fraction of galaxies which are seen relatively close to face - on .
this limitation is evident in the fairly circular morphologies of all of the galaxies found in the survey .
second , we have been extremely conservative in our decision to classify detections as possible lsbs .
we have consciously excluded possible lsbs which we feel can not be adequately distinguished from clusters or tidal extensions off bright galaxies .
thus lsbs with small hii regions or central bulges are not included in our sample , and we are restricted to finding only lsbs which have classical i m morphologies .
there is a tendency for higher luminosity , large scale length lsbs to be earlier hubble types ( mcgaugh , private communication ) , and thus we are less sensitive to the most luminous lsbs .
this bias will lead us to underestimate the luminosity density even more severely than we are underestimating the number density .
note also that the galaxies which we have identified have very regular , smooth morphologies , allowing us to accurately simulate their detection and classification efficiencies . in figure [ numdenfig ] ,
the overdensity of lsbs is smaller when compared to the autofib survey luminosity function ( ellis et al.1996 ) , than when compared to the stromlo - apm ( loveday et al .
1992 ) or dars ( peterson et al .
1986 ) surveys , which both have limiting magnitudes of @xmath149
. however , the autofib survey combines the shallower dars survey with much deeper pencil beam surveys which reach limiting isophotal magnitudes of @xmath150 .
as discussed by mcgaugh ( 1994 ) , this implies that the autofib survey has a fainter surface brightness limit than the shallower stromlo - apm or dars surveys .
thus , the autofib survey is likely to have cataloged far more intrinsically low surface brightness galaxies than the shallow nearby surveys , increasing the integrated number density implied by the autofib luminosity function over what is measured in shallower surveys ; this is clear from the integrated number densities plotted in figure [ numdenfig ] .
( note also that the autofib luminosity function has the steepest faint - end slope of the three surveys . ) the different surface brightness ranges of the surveys , therefore , changes the apparent relative importance of the lsb and normal galaxy population .
the integrated luminosity density of lsb galaxies is presented in figure [ lumdenfig ] , along with the integrated luminosity density in normal galaxies . when restricted to the same range in absolute magnitude
, normal galaxies and lsb with @xmath151 and @xmath143 make a comparable contribution to the total luminosity density of the universe .
however , when restricted to the same range in scale length , the higher luminosity , high surface brightness disks dominate the luminosity density by factors of 2 - 6 over the lsb component , as might be expected by the drastic difference in intrinsic luminosity .
there is also is very little evidence for a major contribution to the luminosity density from galaxies whose disks are fainter than @xmath148 in @xmath1 ; even if there is possible evidence for a rise in number density with decreasing surface brightness in figure [ numdenfig ] , it is not enough to compensate for the increasing faintness of the disk .
we should note however , that these results apply only for a limited range of galaxy properties , and should not be applied more generally . because of our bias against lsbs with bulges , we may be systematically excluding the most luminous galaxies from our sample , and thus our conclusions only hold for the range of disk scale lengths and absolute magnitudes covered by our sample .
our calculated contribution of lsbs with @xmath151 to the luminosity denisity is comparable to that calculated in the brighter apm lsb survey ( sprayberry et al .
1997 ) .
we may compare the measured relative number densities and luminosity densities for lsbs and normal galaxies to those which are predicted by the formalism presented in dalcanton et al .
we do so by integrating the predicted galaxy number density for central surface brightnesses between @xmath73 and @xmath152 in @xmath14 ( assuming @xmath153 for the lsbs in our sample ) , and intrinsic disk scale lengths between @xmath143 , and comparing it to the integrated predicted number density for comparable normal galaxies , with `` comparable '' defined as in figure [ numdenfig ] , and assuming @xmath154 for the galaxies which are typically used to determine the local luminosity function ; this limit is most appropriate for the dars and stromlo - apm surveys .
we make a similar calculation of the luminosity density . for the theoretical models
, we have assumed the values of the values of @xmath155 for the mass - to - light ratio of baryons , @xmath156 for the baryonic mass fraction , and @xmath157 , and @xmath158 for the schechter function describing the distribution of galaxy masses .
we find that the predicted ratio of the number density in lsbs to the number density in normal galaxies with similar scale lengths is 2 ; this ratio varies somewhat with variations in the limiting surface brightness for normal galaxy surveys between @xmath159 and @xmath72 .
the predicted ratio is therefore in agreement with what is measured in figure [ numdenfig ] . restricting the normal galaxy sample to have similar absolute magnitudes to the lsb sample ,
the predicted ratio becomes 0.3 .
the agreement between the measured and predicted relative number density is not as good in this case , with there being even more lsbs measured than predicted .
however , comparing the same range of absolute magnitude is much more sensitive to the faint - end slope of the luminosity function at very faint magnitudes .
because of the rapid rise in number density with decreasing luminosity , the integrated number density depends strongly on the number density of the faintest galaxies ( @xmath160 in this case ) ; unfortunately , this limiting absolute magnitude is where the local luminosity function is most poorly determined , and may be extremely incomplete ( driver & phillipps 1996 ) .
the discrepancy may also result from the parameters of the theoretical model .
we have assumed that the theoretical mass function of galaxies has a steep faint - end slope of @xmath161 , which greatly inflates the number of low - mass galaxies relative to high - mass galaxies ; this accounts for the large difference seen in the predicted ratio of number densities when the normal galaxies are restricted to similar absolute magnitudes rather than similar scale lengths .
we have also assumed that a galaxy s mass is uncorrelated with its angular momentum , whereas in the bbks formalism ( bardeen et al.1996 ) , there should be a correlation wherein low mass galaxies tend have higher angular momenta ( catelan & theuns 1996 ) .
this correlation would increase the expected ratio of lbss to normal galaxies in a fixed magnitude range .
we see a similar level of agreement between the predicted and measured relative luminosity densities .
the predicted ratio of luminosity densities of normal galaxies to lsbs with the same range of scale lengths is @xmath162 , which also agrees reasonably well with the observations ( figure [ lumdenfig ] . when restricted to the same range of absolute magnitude the predicted ratio of normal galaxy luminosity density to lsb
luminosity density is 4 , whereas they are observed to be comparable .
this suggests that there is more luminosity density in lsbs than predicted . again , uncertainty in the measured faint - end of the local luminosity function , or a correlation between mass and angular momentum could bring these into agreement .
the importance of the low surface brightness galaxy population has been an outstanding question in astronomy .
claims have ranged from lsbs being a negligible component of the local universe ( roukema & peterson 1994 ) to outnumbering @xmath163 galaxies by orders of magnitude ( @xmath164 for scale lengths greater than @xmath165 ; schwartzenberg et al .
1995 ) . to address this recent debate ,
we have presented the first well - quantified measurement of the absolute number density and luminosity density of low surface brightness galaxies , using a large area ccd survey with well understood selection criteria , quantifiable detection efficiencies and follow - up spectroscopy of all lsb candidates .
our survey has selected lsbs which are comparable to the galaxies in the ngc catalog , in both size and distance , allowing a direct , meaningful comparison with local field galaxy surveys .
we find that the number density of lsb galaxies with @xmath166 is indeed greater than the number density of normal galaxies , when the normal galaxies are restricted to the same range in either absolute magnitude ( @xmath142 ) or in exponential disk scale length ( @xmath143 ) .
the exact ratio depends upon which local luminosity function is used for comparison , but lsbs typically outnumber comparable normal galaxies by factors of 2 or more .
the measured ratios are formally consistent with theoretical expectations based on assuming that high angular momentum and/or low mass protogalaxies collapse to form low surface brightness galaxies ( dalcanton et al .
1997 ) .
however , we also find that lsbs with central surface brightnesses fainter than @xmath167 and with scale lengths between @xmath168 make a less significant contribution to the luminosity density of the universe . in our survey , galaxies fainter than @xmath169 make almost no contribution to the luminosity density , and galaxies fainter than @xmath167 contribute 15 - 50% of the luminosity density contributed by comparable normal galaxies .
thus , a significant ( but not overwhelming ) fraction of the stars , and perhaps the mass , of the universe is in lsbs . the inferred number densities and luminosities are likely lower limits as the survey may have excluded significant number of lsbs due to selection effects .
by design , our survey does not include galaxies which are intermediate in surface brightness between @xmath167 and the characteristic freeman surface brightness ; these galaxies are intrinsically more luminous , and thus could make a much larger contribution to the luminosity density .
furthermore , we have restricted our selection to sm / im morphologies , whereas the brightest , largest scale length lsbs tend to have both bulges and spiral arms ( mcgaugh ; private communication )
. this bias against earlier type morphologies may well have lead us to underestimate the lsb luminosity density , and , less severely , the lsb number density .
our survey also includes a strong bias against edge - on lsbs ( due to using a circularly symmetric filter in the detection algorithm ) , which may again lead us to underestimate the lsb density .
acknowledgements tony tyson and ian smail are warmly thanked for advice on the photometric calibration , as is lori lubin for her assistance with data acquisition at the kpno 0.9 m and jim deveny for his help with the cryocam at the kpno 4 m .
tony also generously donated huge quantities of disk space , without which this project could not possibly have been done .
the referee is also thanked for very helpful commentary and suggestions .
jjd gratefully acknowledges useful discussions with rebecca bernstein and dan rosenthal .
support for jjd was provided by nasa through hubble fellowship grant # 2 - 6649 awarded by the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, for nasa under contract nas 5 - 26555 . partial support for dps provided by national science foundation grant ast-95 - 09919 .
ms was supported by nsf grants ast91 - 08834 and ast94 - 15574 .
* appendix a * probability distributions for @xmath170 and @xmath141 for small numbers of galaxies suppose we have a sample of @xmath171 galaxies , the @xmath134th which has angular scale length @xmath130 and central surface brightness @xmath131 .
assume that there is a unique overall detection efficiency @xmath172 associated with each scale length and surface brightness , and that there is a fixed rms uncertainty in @xmath173 of @xmath174 .
assuming that the underlying galaxy distribution in surface brightness and apparent scale length is @xmath175 , the total number @xmath176 of galaxies expected in a survey of area @xmath177 with central surface brightnesses between @xmath178 and @xmath179 and apparent angular scale lengths greater than @xmath180 is : @xmath181 if the survey had perfect efficiency , @xmath182 , the survey would recover @xmath183 galaxies .
thus , the total survey efficiency is @xmath184 .
suppose then , that an individual survey measures @xmath171 galaxies over the survey area .
the surface density becomes @xmath185 as shown in appendix b , for local volumes where cosmological curvature in negligible , @xmath186 for any distribution of intrinsic galaxy scale lengths .
we have assumed that @xmath187 and @xmath7 are independent variables , and have the freedom to choose any distribution of @xmath188 .
we take @xmath188 to be constant to first order , as is consistent with the data presented in figure [ numdenfig ] .
to calculate the probability distribution of the surface density @xmath170 , we first consider the effect of having a small number of galaxies in the sample .
we calculate this poisson uncertainty by adopting bayes theorem .
we assume that , given a single trial which measured @xmath171 events , the probability that infinite number of similar trials would measure a mean of @xmath189 events is @xmath190 . because we have no prior reason to prefer any value of @xmath189
, we assume that the prior probability @xmath191 is uniform , and thus that @xmath192 , which is simply the poisson probability of observing @xmath171 events for a system which would give a mean of @xmath189 events .
there is also a contribution to the uncertainty in @xmath170 from the uncertainty in the efficiency @xmath173 .
however , the integral in equation [ sigappendix ] averages out much of the uncertainty in the efficiency @xmath173 . for the particular example of our survey , while the pointwise estimation of @xmath193 and @xmath194 has an rms uncertainty of roughly @xmath195 , the integral over the efficiency in equation [ nbarappendix ] is much more tightly constrained .
the uncertainty contributed to the integral by @xmath193 is negligible ( @xmath196% ) , and the uncertainty contributed by @xmath194 is likewise small ( @xmath197% ) for central surface brightnesses fainter than @xmath73 . at brighter surface brightnesses , our approximation for @xmath194 overestimates the integral volume by @xmath198% .
however , we only have one galaxy in this surface brightness interval , and the direction of the error is to underestimate the number density , thus we expect this to contribute very little uncertainty to the total number density
. given that the poisson uncertainty is of the order of @xmath199 for our sample , we will neglect the contribution of @xmath200 to the uncertainty in @xmath170 .
thus , the final probability distribution for the surface density @xmath170 is @xmath201 we may also solve for the effective number density of the survey , given the distance @xmath129 of each galaxy , and the maximum distance at which each galaxy could be detected , @xmath202 ( eqn .
[ dmax ] ) .
the effective volume associated with each galaxy , @xmath203 , is an integral over distance , accounting for the changing efficiency with changing angular size : @xmath204)\ , { d^\prime}^2\,{\rm d}d^\prime \\ & = & \frac{a}{\rm radians}\ , \left[\alpha_i d_i\right]^3 \int_{\frac{\alpha_i d_i}{d_{max_i}}}^\infty \frac{\epsilon(\mu_i,\alpha^\prime)}{{\alpha^\prime}^4}\ , { \rm d}\alpha^\prime,\end{aligned}\ ] ] assuming that the galaxies are close enough that @xmath28 cosmological dimming of @xmath187 can be neglected . with the exception of the integral over surface brightness , this integral is identical to the one equation [ nbarappendix ] .
the uncertainties in each volume term @xmath203 are dominated by the poisson uncertainties which also dominate the calculation of @xmath170 ( here , @xmath205 ) .
the other possible contributions are from the uncertainty in the measurement of @xmath130 , which is @xmath206% , and in @xmath129 , which is less than @xmath207 .
these are also negligible compared to the poisson uncertainties ; in the limit of gaussian statistics , including a 10% uncertainty from the detection efficiency and a 10% efficiency from the measurement of @xmath130 would increase a 40% poisson uncertainty to only 42% .
the number density , @xmath141 , measured for the sample of @xmath171 galaxies is @xmath208 with our assumption that poisson statistics dominate uncertainties in @xmath203 , the probability for the individual terms @xmath209 is @xmath210 .
the final probability distribution of @xmath141 is then @xmath211 which can be solved with monte carlo integration by repeatedly drawing samples of galaxies from each of the @xmath212 .
* appendix b * recovering the number density and scale length distribution of lsbs + from the observed surface density in this appendix we investigate the relationship between the intrinsic number density of objects at redshift @xmath213 with surface brightness @xmath6 and physical scale length @xmath214 , @xmath215 , and the observed surface density of objects with apparent surface brightness @xmath187 and angular scale length @xmath7 , @xmath216 .
we also calculate the underlying distribution of lsb scale lengths from existing catalogs .
first , consider a volume element @xmath217 at an angular diameter distance @xmath218 , subtending one steradian of the sky .
the number density of objects in the shell that appear with angular scale length @xmath7 is @xmath219 where @xmath7 is measured in radians .
if we allow objects to be at large enough distances that surface brightness diminution plays a role , then the observed surface brightness scales as @xmath16 and the number density becomes @xmath220 the number of objects per steradian per @xmath221 is @xmath222 if a sample is close enough that all distances can be treated as euclidean , the volume element is @xmath223 , the surface brightness of a galaxy does not change with distance , and evolution in @xmath224 may be safely ignored .
therefore , for a uniform distribution of galaxies , equation [ surfden ] becomes @xmath225 note that @xmath216 will always be proportional to @xmath127 , independent of the distribution of intrinsic scale lengths .
any size galaxy can be seen to a larger distance if it is observed at a smaller limiting angular size ; smaller values of @xmath7 always probe larger volumes of the universe and always lead to larger surface densities .
if we make the further assumption that @xmath226 is separable , and can be expressed as @xmath227 , then @xmath228 for an assumed form of @xmath229 , the fraction of objects with physical scale length @xmath214 , equation [ surfdenlocalsep ] may be inverted to recover @xmath230 from the observed distribution of surface densities and angular scale lengths . however ,
if @xmath226 is not separable , than the integral in equation [ surfdenlocalsep ] is not independent of @xmath6 and the relative proportion of galaxies in terms of surface density @xmath216 does not reflect the relative proportion of galaxies in real space .
the integral over @xmath214 diverges unless @xmath229 falls off more quickly than @xmath231 for large @xmath214 ; this seems to be the case , as there are no known galaxies to date which have scale lengths larger than @xmath232 .
we may uncover some confirmation for the cutoff at large @xmath214 from the poss - ii survey for lsbs , a diameter - limited sample of lsbs ( schombert et al . 1992 ) which includes several extremely large galaxies similar to the giant lsb malin - i ( bothun et al .
hi observations of the `` v '' subsample of lsbs ( @xmath233 ) have a 70% detection rate , and imply a @xmath234 statistic compatible with a uniform redshift distribution . if @xmath235 is the number of galaxies in a diameter limited sample with physical radii at the isophotal diameter between @xmath236 and @xmath237 , then the number density of galaxies with physical radii between @xmath236 and @xmath238 is @xmath239 where @xmath240^{-1/3}$ ] and @xmath241 and @xmath242 are the diameter limits of the survey .
the luminosity density is @xmath243 assuming @xmath244 is the mean surface brightness within @xmath236 .
these distributions are plotted in figure [ schombertvfig ] . although the completeness and selection biases of the schombert et al . catalog
are poorly understood , the strong drop - off in luminosity density around @xmath245 seems quite robust ; including the galaxies that were not detected in hi could push the flat luminosity density out to @xmath246 , but no further . while there are some galaxies with rather large sizes , the evidence suggests that they are neither an important contributor to the number density or to the luminosity density .
mcgaugh & bothun ( 1994 ) have done follow up observations of roughly a dozen of the galaxies from the poss - ii survey which suggest that the exponential scale length is typically one - quarter to one - half of the isophotal radius .
this allows us to estimate the exponential scale length @xmath214 given the isophotal radius @xmath236 .
the schombert catalog data plotted in figure [ schombertvfig](b ) suggests that @xmath229 is best described by a power law @xmath247 between @xmath248 and @xmath249 , where here @xmath214 is the radius of the galaxy at the limiting isophote of the poss - ii survey .
llrrccclcl pg 0849 + 4748 & r-27 - 1 & 8 49 28.4 & + 47 48 37 & 23.1 & 4.3 & 3.6 & -18.3 & 8809 & h@xmath250,o[iii],h@xmath7pg 0847 + 4747 & r-26 - 1 & 8 47 44.9 & + 47 47 48 & 23.2 & 2.6 & - & - & - & pg 1521 + 4632 & m-232 - 1 & 15 21 32.4 & + 46 32 58 & 23.6 & 3.2 & 1.7 & -16.1 & 5600 : & break?pg 1136 + 4750 & q-129 - 2 & 11 36 46.1 & + 47 50 41 & 23.7 & 4.5 & 1.9 & -16.2 & 4270 &
h@xmath250,o[iii],h@xmath7pg 0914 + 4744 & q-42 - 1 & 9 14 24.5 & + 47 44 48 & 23.9 & 6.1 & 3.0 & -17.1 & 5400 & break , h@xmath250?pg 1133 + 4755 & r-127 - 1 & 11 33 35.4 & + 47 55 26 & 24.7 & 6.9 & 3.6 & -16.7 & 5250 & breakpg 1327 + 4637 & m-161 - 1 & 13 27 46.2 & + 46 37 47 & 24.8 & 6.2 & 3.5 & -16.5 & 5600 & break | we present results of a large area ccd survey for low surface brightness galaxies ( lsbs ) that reaches central surface brightnesses of @xmath0 in @xmath1 .
we have analyzed @xmath2 of transit scan data , and identified a statistical subset of 7 pure disk lsb s with central surface brightnesses fainter than @xmath3 and with angular exponential scale lengths larger than @xmath4 .
the lsb detection is entirely automated , and the selection efficiency of the survey is well - quantified . after correcting for the selection efficiency , we find a surface density of @xmath5 for lsbs in the considered range of @xmath6 and @xmath7 ( 90% confidence levels ) , with the largest correction being due to the area lost behind bright stars , and the difficulty in detecting lsbs with small angular sizes .
we have measured redshifts to the final sample of lsbs , and find them to be at distances comparable to those probed by large galaxy catalogs , and to have intrinsic scale lengths of @xmath8 , also comparable to normal galaxies .
we use the redshifts and the selection efficiency to calculate the number density in lsbs with @xmath9 and find @xmath10 , with 90% confidence .
the measurement of the absolute number density of lsbs probably represents a lower limit , due to very strong biases against lsbs with bulges or edge - on lsbs in our sample . comparing the lsb number density to the number density of normal galaxies with either similar scale lengths or similar luminosities
, we find that the number density of lsbs with @xmath9 is comparable to or greater than the number density of normal galaxies .
the luminosity density in lsbs is comparable to the luminosity density of normal galaxies with similar luminosities , but is a factor of 3 - 10 smaller than the luminosity density of normal galaxies with similar scale lengths .
the relative lsb number density and luminosity density agree well with the theoretical predictions of dalcanton et al .
( 1997 ) .
the redshift - space distribution of the lsbs suggests that the trend for low surface brightness galaxies to have weak small - scale correlations may continue to the fainter surface brightnesses covered in this survey .
= cmcsc10 _astr . ap .
_ , # 1to 0pt#1 |
the relationship between energy and information has been investigated extensively in the context of computation energy cost including a modern analysis of maxwell s demon @xcite-@xcite . in this letter , we show a new energy - information relation from a different point of view .
recently , it has been reported that energy can be transported by local operations and classical communication while retaining local energy conservation and without breaking causality @xcite-@xcite .
such protocols are called quantum energy teleportation ( qet ) and are based on ground - state entanglement of many - body quantum systems including spin chains @xcite , cold trapped ions @xcite and quantum fields @xcite . by performing a local measurement on a subsystem a of a many - body system in the ground state , information about the quantum fluctuation of a can be extracted . because the post - measurement state is not the ground state in general , some amount of energy is infused into a as qet energy input during this measurement , and the ground - state entanglement gets partially broken .
next , the measurement result is announced to another subsystem b of the many - body system at a speed much faster than the diffusion velocity of the energy infused by the measurement . soon after the information arrives at b
, energy can be extracted from b as qet energy output by performing a local operation on b dependent on the announced measurement data .
the root of the protocols is a correlation between the measurement information of a and the quantum fluctuation of b via the ground - state entanglement .
due to the correlation , we are able to estimate the quantum fluctuation of b based on the announced information from a and devise a strategy to control the fluctuation of b. by the above - mentioned selected local operation on b , the fluctuation of b can be more suppressed than that of the ground state , yielding negative energy density around b in the many - body system .
the concept of negative energy density has been investigated in quantum field theory for a long time @xcite .
quantum interference among total energy eigenstates can produce various states containing regions of negative energy density , although the total energy remains nonnegative .
the regions of negative energy density can appear in general many - body quantum systems by fixing the origin of the energy density such that the expectational value vanishes for the ground state . in spite of the emergence of negative energy density
, the total energy also remains nonnegative for the general cases . in the qet protocols , during the generation of negative energy density at b , surplus positive energy
is transferred from b to external systems and can be harnessed as the qet output energy . here
it should be emphasized that this output energy existed not at a but at b _ even _ _ before _ the start of the protocol and was hidden inside the zero - point fluctuation of b. of course , this zero - point energy is not available by usual local operations for b. however , by using a local operation dependent on a s information , it becomes possible to dig out b s zero - point energy by pair creation of the positive output energy from b and the negative energy of b. hence , we do not need to hire any physical carrier of energy from a to b like electric currents and photons , at least , during short - time qet processes .
needless to say , after the completion of qet process , the positive energy of a compensates for the negative energy of b during late - time free evolution of the many - body system .
the amount of output energy from b is upper bounded by the amount of input energy to a. using the usual protocols of quantum teleportation , quantum states can be transported from one party to another by the consumption of shared entanglement between the two parties @xcite . as is well known @xcite , transfer of a large number of quantum states
requires a large amount of consumption of shared entanglement as a physical resource . taking into account the fact
, it seems natural for the qet protocols to expect that a large amount of teleported energy also requests a large amount of consumption of the ground - state entanglement between a and b. if such a non - trivial relation exists between teleported energy and breaking of ground - state entanglement by measurement , the relation may shed new light on the interplay between quantum physics and quantum information theory . in this letter , the first example of the energy - entanglement relation for a minimal qet model
is presented .
the minimal qet model is the smallest physical system for which non - trivial qet can be implemented ; this model consists of two qubits with an interaction of the ising spin chain in the presence of a transverse magnetic field .
we explicitly show that for the minimal model , the consumption of entanglement between a and b during the measurement of a is lower bounded by a positive value that is proportional to the maximum amount of energy teleported from a to b. in addition , we obtain another inequality in which the maximum amount of energy teleported from a to b is lower bounded by a different positive value that is proportional to the amount of entanglement breaking between a and b by the measurement of a. these energy - entanglement inequalities are of importance because they help in gaining a profound understanding of entanglement itself as a physical resource by relating entanglement to energy as an evident physical resource .
first of all , we introduce the minimal qet model .
the system consists of two qubits a and b. its hamiltonian is the same as that of the ising spin chain in the presence of a transverse magnetic field as follows : @xmath0 , where each contribution is given by @xmath1 and @xmath2 and@xmath3 are positive constants with energy dimensions , @xmath4 is the x - component of the pauli operators for the qubit a ( b ) , and @xmath5 is the z - component for the qubit a ( b ) .
the constant terms in eqs .
( [ 1])-([3 ] ) are added in order to make the expectational value of each operator zero for the ground state @xmath6 : @xmath7 because the lowest eigenvalue of the total hamiltonian @xmath8 is zero , @xmath8 is a nonnegative operator:@xmath9 .
meanwhile , it should be noticed that @xmath10 has a negative eigenvalue , which can yield negative energy density .
the ground state is given by @xmath11 where @xmath12 is the eigenstate of @xmath5 with eigenvalue @xmath13 .
when local operations are performed to this system , we connect the system with a local appratus and ( or ) add local external forces like magnetic field .
however , time interval of the operations is assumed to be quite short , as often argued in quantum information theory .
thus interaction terms between the system and the external apparatus vanish in the hamiltonian after the short - time operation .
let @xmath14 denote a set of povm measurements @xcite for a which measurement operators @xmath15 with measurement output @xmath16 commute with the interaction hamiltonian @xmath17 .
the measurement operator @xmath15 takes the form of @xmath18 the coefficients @xmath19 , @xmath20 , @xmath21 and @xmath22 are real constants which satisfy @xmath23 the povm corresponding to @xmath15 is defined by @xmath24 which satisfies the completeness relation , @xmath25 by introducing the emergence probability @xmath26 of output @xmath16 for the ground state and a real parameter @xmath27 , the povm is written as follows : @xmath28 by taking suitable values of @xmath19 , @xmath20 , and @xmath21 , all values of @xmath26 and @xmath27 are permissible as long as they satisfy @xmath29 , @xmath30 and @xmath31 .
the post - measurement state of the two qubits with output @xmath16 is given by @xmath32 this measurement excites the system and the average post - measurement state has a positive expectational value @xmath33 of @xmath8 , which energy distribution is localized at a. in fact , the value defined by @xmath34 is computed as @xmath35 this infused energy @xmath33 is regarded as the qet energy input via the measurement of a. during the measurement , @xmath33 is transfered from external systems including the measurement device with a battery respecting local energy conservation .
the qet energy conservation law during local measurements to a qubit of a spin chain has been discussed in @xcite .
the key feature of this model is that any measurement of @xmath14 does not increase the average energy of b at all . by explicit calculations , the average values of the hamiltonian contributions @xmath10 and
@xmath17 are found to remain zero after the measurement and are the same as those of the ground state .
this measurement does not yield instantaneous change of @xmath17 .
therefore we have no direct force of a affecting b after the measurement .
thus , we can not extract energy from b by the standard way soon after the measurement .
in fact , if any local unitary operation @xmath36 independent of a s measurement result is performed to b , the post - operation state @xmath37 is given by @xmath38 the energy difference after the operation is calculated as @xmath39 -e_{a}=\langle g|w_{b}^{\dag } \left ( h_{b}+v\right ) w_{b}|g\rangle,\label{e01}\ ] ] where we have used @xmath40@xmath41 = 0,\ ] ] and the completeness relation of the povms:@xmath42 from eq .
( [ e01 ] ) , it is proven that the energy difference is nonnegative : @xmath39 -e_{a}=\langle g|w_{b}^{\dag}hw_{b}|g\rangle\geq0,\ ] ] because of a relation such that @xmath43 and the nonnegativity of @xmath8 . therefore , as a natural result , no local operation to b independent of @xmath16 extracts energy from the system . after a while ,
the infused energy @xmath33 diffuses to b. the time evolution of the expectational values @xmath10 and @xmath17 of the average post - measurement state is calculated as@xmath44 , \ ] ] and @xmath45 . therefore , energy can be extracted from b after a diffusion time scale of @xmath46 ; this is just a usual energy transportation from a to b. amazingly , the qet protocol can transport energy from a to b in a time scale much shorter than that of the usual transportation . in the protocol , the measurement output @xmath16 is announced to b. because the model is non - relativistic , the propagation speed of the announced output can be much faster than the diffusion speed of the infused energy and can be approximated as infinity .
soon after the arrival of the output @xmath16 , we perform a local operation @xmath47 on b dependent on @xmath16 .
then , the average state after the operation is given by @xmath48 in figure 1 , a schematic diagram of this qet model is presented
. the expectational value of the total energy after the operation is given by @xmath49 = \sum_{\mu}\langle g|m_{a}(\mu ) ^{\dag}u_{b}(\mu)^{\dag}hu_{b}(\mu)m_{a}(\mu)|g\rangle.\ ] ] on the basis of the fact that @xmath47 commutes with @xmath50 and eq .
( [ 4 ] ) , @xmath51 is computed as @xmath52 = -\operatorname*{tr}\left [ \rho\left ( h_{b}+v\right ) \right ] .\ ] ] further , on the basis of the fact that @xmath15 commutes with @xmath53 , @xmath10 and @xmath17 , the energy can be written as @xmath54 and @xmath55 . here
, let us write the general form of @xmath47 as follows : @xmath56 where @xmath57 is a real parameter , @xmath58is a three - dimensional unit real vector and @xmath59 is the pauli spin vector operator of b. then , an explicit evaluation of @xmath51 becomes possible
. the result is expressed as @xmath60 where @xmath61 is given by @xmath62 where @xmath63 is defined by @xmath64 -3hkq_{a}(\mu)n_{x\mu}n_{z\mu}.\ ] ] in order to maximize the teleported energy @xmath51 for a given povm measurement of a , let us first maximize @xmath61 in eq .
( [ 6 ] ) by changing the parameter @xmath65 .
this maximum value is calculated as @xmath66 ^{2}}-x(\mu ) .
\label{13}\ ] ] next , let us introduce a parametrization of @xmath67 and @xmath68 as @xmath69 and @xmath70 for fixed @xmath71 which runs over @xmath72 $ ] , where @xmath73 is a real parameter .
it is observed that @xmath74 in eq .
( [ 13 ] ) is a monotonically decreasing function of @xmath63 .
thus , we must find the minimum value of @xmath63 in terms of @xmath73 . by using the parametrization
, we can minimize @xmath63 as @xmath75 therefore , the maximum value of @xmath74 in terms of @xmath73 is obtained as follows : @xmath76 next , in order to maximize @xmath77 in terms of @xmath78 , let us write it as a function @xmath79 of @xmath78 : @xmath80 where @xmath81 , @xmath82 and @xmath83 are positive constants given by@xmath84 the derivative of @xmath79 can be calculated as @xmath85 where @xmath86 is a function given by @xmath87 it can be verified that @xmath86 and @xmath88 are nonpositive for @xmath89 $ ] .
this verification can be done as follows .
let us first consider an equation @xmath90 .
it turns out that , in the transformation of this equation for solving @xmath91 , the dependence of @xmath91 gets lost and we get just a constraint condition on @xmath26 and @xmath27 such that @xmath92 .
thus , if @xmath93 or @xmath94 , the equation @xmath95 holds for all @xmath89 $ ] .
if @xmath96 and @xmath97 , the solution @xmath91 does not exist and @xmath86 has a definite sign for @xmath98 $ ] . in order to check the sign ,
let us substitute @xmath99 into @xmath86 .
then , when @xmath100 , we get @xmath101 , and when @xmath102 , @xmath103 . thus , it is verified that @xmath86 and @xmath88 are nonpositive . therefore , @xmath79 takes the maximum value at @xmath104 .
this implies that @xmath61 can be maximized as @xmath105 .
this leads to our final expression of the maximum teleported energy for the measurement , which is clearly nonnegative , as follows : @xmath106 .
\label{21}\ ] ] the operation @xmath107 which attains the maximum of teleported energy is given by @xmath108 where @xmath109 is a real constant which satisfies @xmath110 besides , the teleported energy can be maximized among povm measurements of @xmath14 .
this is achieved when each povms are proportional to projective operators and given by@xmath111 .\ ] ]
next , we analyze entanglement breaking by the povm measurement of a and show two inequalities between the maximum teleported energy and the entanglement breaking . we adopt entropy of entanglement as a quantitative measure of entanglement .
the entropy of a pure state @xmath112 of a and b is defined as @xmath113 \ln\operatorname*{tr}_{a}\left [ measurement , the total system is prepared to be in the ground state @xmath6 . the reduced state of b is given by @xmath114 .\ ] ] after the povm measurement outputting @xmath16 , the state is transferred into a pure state @xmath115 .
the reduced post - measurement state of b is calculated as @xmath116 .\ ] ] the entropy of entanglement of the ground state is given by @xmath117\ ] ] and that of the post - measurement state with output @xmath16 is given by @xmath118 .\ ] ] by using these results , we define the consumption of ground - state entanglement by the measurement as the difference between the ground - state entanglement and the averaged post - measurement - state entanglement : @xmath119 interestingly , this quantity is tied to the quantum mutual information between the measurement result of a and the post - measurement state of b. let us introduce a hilbert space for a measurement pointer system @xmath120 of the povm measurement , which is spanned by orthonormal states @xmath121 corresponding to the output @xmath16 satisfying @xmath122 . then , the average state of @xmath120 and b after the measurement is given by @xmath123 by using the reduced operators @xmath124 $ ] and @xmath125 $ ] , the mutual information @xmath126 is defined as@xmath127 -\operatorname*{tr}_{b}\left [ \phi_{b}\ln\phi_{b}\right ] + \operatorname*{tr}_{\bar{a}b}\left [ \phi_{\bar{a}b}\ln\phi_{\bar{a}b}\right ] .\nonumber\ ] ] by using @xmath128 = \sum_{\mu } p_{a}(\mu)|\mu_{\bar{a}}\rangle\langle\mu_{\bar{a}}|$ ] and @xmath129 = \sum_{\mu}p_{a}(\mu)\rho_{b}(\mu)=\rho_{b}$ ] , it can be straightforwardly proven that @xmath126 is equal to @xmath130 .
this relation provides another physical interpretation of @xmath130 .
next , let us calculate @xmath130 explicitly .
all the eigenvalues of @xmath131 are given by @xmath132 , \label{29}\ ] ] where @xmath133 is a real constant which satisfies @xmath134 the eigenvalues of @xmath135 are obtained by substituting @xmath136 into eq .
( [ 29 ] ) . by using @xmath137 ,
@xmath130 can be evaluated as @xmath138 where @xmath139 is a monotonically increasing function of @xmath140 $ ] and is defined by@xmath141 it is worth noting that the optimal teleported energy @xmath142 in eq .
( [ 21 ] ) takes a form similar to eq .
( [ 30 ] ) as @xmath143 where @xmath144 is a monotonically increasing function of @xmath140 $ ] and is defined by @xmath145 .\ ] ] expanding both @xmath139 and @xmath144 around @xmath146 yields @xmath147 by deleting @xmath148 in the above two equations , we obtain the following relation for weak measurements with infinitesimally small @xmath27 : @xmath149 it is of great significance that this relation can be extended as the following inequality for general measurements of @xmath150 @xmath151 this inequality is one of the main results of this letter and implies that a large amount of teleported energy really requests a large amount of consumption of the ground - state entanglement between a and b. the proof of eq .
( [ 32 ] ) is as follows .
let us introduce two rescaled functions as follows .
@xmath152 it can be easily shown that @xmath153 is a convex function : @xmath154 from this convexity and @xmath155 , the function @xmath153 satisfies @xmath156 for @xmath157 .
on the other hand , it is observed that @xmath158 is a concave function , as shown below .
the derivative of @xmath158 is computed as @xmath159 where @xmath160 and @xmath161 is a positive function of @xmath162 defined as @xmath163 it should be noted that @xmath164 is a monotonically increasing function of @xmath148 .
the derivative of @xmath161 is calculated as @xmath165 where @xmath166 is a function of @xmath162 given by @xmath167 and satisfies a boundary condition as @xmath168 .
it is also easy to show that the derivative of @xmath166 is positive for @xmath169 : @xmath170 thus , @xmath166 and @xmath171 are positive for @xmath172 $ ] . from these results
, it has been proven that @xmath158 is concave for @xmath173 $ ] : @xmath174 . from this concavity and @xmath175
, it is shown that @xmath158 satisfies the relation @xmath176 because of eqs .
( [ 46 ] ) and ( [ 45 ] ) , we obtain the following inequality : @xmath177 this implies that the energy - entanglement inequality in eq .
( [ 32 ] ) holds .
in addition , we can prove another inequality between energy and entanglement breaking .
because the convex function @xmath153 and the concave function @xmath158 are monotonically increasing functions of @xmath173 $ ] which satisfy @xmath178 @xmath179 , we have the following relation:@xmath180 consequently , the following inequality , which is another main result of this letter , is obtained for all measurements of @xmath14 : @xmath181 } { \left ( 1+\cos\varsigma\right ) \ln\left ( \frac{2}{1+\cos\varsigma}\right ) + \left ( 1-\cos\varsigma\right ) \ln\left ( \frac{2}{1-\cos\varsigma}\right ) } \delta s_{ab}. \label{770}\ ] ] this ensures that if we have consumption of ground - state entanglement @xmath182 for a measurement of @xmath14 , we can in principle teleport energy from a to b , where the energy amount is greater than the value of the right - hand - side term of eq .
( [ 770 ] ) .
this bound is achieved for non - zero energy transfer by measurements with @xmath183 .
the inequalities in eq .
( [ 32 ] ) and eq .
( [ 770 ] ) help us to gain a deeper understanding of entanglement as a physical resource because they show that the entanglement decrease by the measurement of a is directly related to the increase of the available energy at b as an evident physical resource .
i would like to thank yasusada nambu for a comment .
this research has been partially supported by the global coe program of mext , japan , and the ministry of education , science , sports and culture , japan , under grant no .
21244007 . figure 1 : schematic diagram of the minimal qet model .
a povm measurement is performed on a with infusion of energy @xmath33 .
the measurement result @xmath16 is announced to b through a classical channel .
after the arrival of @xmath16 , a unitary operation dependent on @xmath16 is performed on b with extraction of energy @xmath51 . | protocols of quantum energy teleportation ( qet ) , while retaining causality and local energy conservation , enable the transportation of energy from a subsystem of a many - body quantum system to a distant subsystem by local operations and classical communication through ground - state entanglement . we prove two energy - entanglement inequalities for a minimal qet model .
these relations help us to gain a profound understanding of entanglement itself as a physical resource by relating entanglement to energy as an evident physical resource .
+ + + + + _ department of physics , faculty of science , tohoku university , + sendai 980 - 8578 , japan + [email protected]_ |
there is the classical weierstrass representation for minimal surfaces in euclidean @xmath0-space @xmath3 , which gives a local conformal parametrization for any minimal surface .
it involves choosing two holomorphic functions ( or perhaps meromorphic functions when considering the surfaces more globally ) on a riemann surface .
if one restricts to isothermic parametrizations , that is , conformal parametrizations that are also curvature lines coordinate systems , then the representation reduces to the choice of just one holomorphic function .
since every minimal surface has local isothermic coordinates ( away from umbilics ) , this reduction does not involve any loss of generality beyond avoiding umbilic points .
once one restricts to isothermic parametrizations , it becomes possible to give a definition for discrete analogs of minimal surfaces @xcite .
these analogs are `` discrete isothermic '' meshes ( a definition of this is given later in this paper ) .
they are comprized of planar quadrilaterals , which in particular have concircular vertices . by a transformation called the lawson correspondence or @xmath4-transformation or calapso transformation @xcite
, one can produce all constant mean curvature ( cmc ) @xmath2 surfaces in hyperbolic @xmath0-space @xmath1 from minimal surfaces in @xmath3 .
there is a corresponding holomorphic representation for those surfaces as well , first given by bryant @xcite .
correspondingly , without loss of generality beyond avoiding umbilics , one can restrict to isothermic coordinates in this case also , and one has a discrete analog of cmc @xmath2 surfaces in @xmath1 , first found by hertrich - jeromin @xcite . in the case of smooth surfaces
there is also a holomorphic representation for flat ( i.e. intrinsic curvature zero ) surfaces in @xmath1 @xcite and this also ties in to the above - mentioned bryant representation , as there are deformations from cmc @xmath2 surfaces in @xmath1 to flat surfaces via a family of linear weingarten surfaces in @xmath1 @xcite .
these do not include all linear weingarten surfaces , but rather a certain special subclass called linear weingarten surfaces of _ bryant type _
@xcite @xcite , so named because they have bryant - type representations .
thus it is natural to wonder if flat surfaces also have a discrete analog , and we will see here that they do .
once this discrete analog is found , a new question about `` singularities on discrete flat surfaces '' naturally presents itself , in this way : unlike the smooth isothermic minimal surfaces in @xmath3 and cmc @xmath2 surfaces in @xmath1 , smooth flat fronts have certain types of singularities , such as cuspidal edges and swallowtails ( in fact , indirectly , this is what the naming `` fronts '' rather than `` surfaces '' indicates ) .
the means for recognizing where the singularities are on smooth flat fronts are clear , and one can make classifications of those surfaces most generic types of singularities just from looking at the choices of holomorphic functions used in their representation @xcite .
however , in the case of discrete flat surfaces , it is not apriori clear where the singularities are , nor even what such a notion would mean . since one does not have first and second fundamental forms at one s disposal in the discrete case
, one must find an alternate way of defining singularities .
we aim towards this by defining and using a discrete analog of caustics , also called focal surfaces , for smooth flat fronts . for a smooth flat front
, the caustic is the surface comprised of all the singular points on all parallel surfaces of that flat front .
( the parallel surfaces are also flat . )
thus the singular set of the flat front can be retrieved by taking its intersection with its caustic . in the case of a smooth flat front ,
the caustic is again a flat surface , but this will not quite be the case for discrete flat surfaces . we will also present a number of examples of these discrete flat surfaces .
in addition to the rather simple examples of discrete cylinders and discrete surfaces of revolution , we will also discuss a discrete flat surface based on the airy equation .
this example exhibits swallowtail singularities and a stokes phenomenon , similar to that of the analogous surface in the smooth case , as shown by two of the authors in @xcite .
this last example hints at existence of a robust collection of discrete flat surfaces with interesting geometric properties yet to be explored .
thus , the purpose of this paper is to : 1 . provide a definition for discrete flat surfaces and discrete linear weingarten surfaces of bryant type in hyperbolic @xmath0-space @xmath1 ; 2 . give properties of these surfaces that justify our choice of definitions ( in particular , as smooth flat fronts have extrinsic curvature @xmath2
, we identify notions of discrete extrinsic curvature of discrete flat surfaces which do indeed attain the value @xmath2 ) ; 3 .
show that these surfaces have concircular quadrilaterals ; 4 .
study examples of these surfaces , and in particular look at swallowtail singularities and global properties of an example related to the airy equation ; 5 .
give a definition of discrete caustics for discrete flat surfaces ; 6 .
show that the caustics also have concircular quadrilaterals and that they provide a means for identifying a notion of singularities on discrete flat surfaces . in section [ section1 ]
we describe smooth and discrete minimal surfaces in euclidean @xmath0-space @xmath3 , to help motivate later definitions , and we also give the definition of a discrete holomorphic function , which will be essential to everything that follows . in section [ section2 ]
we describe smooth cmc @xmath2 surfaces , and flat surfaces and linear weingarten surfaces of bryant type in @xmath1 , again as motivational material for the definitions of the corresponding discrete surfaces in section [ section3 ] .
we prove in section [ section3 ] that discrete flat surfaces and linear weingarten surfaces of bryant type have concircular quadrilaterals .
also , section [ section3 ] provides a natural representation for discrete flat surfaces which gives the mapping of the surfaces as products of @xmath5 by @xmath5 matrices times their conjugate transposes , and we show that this representation applies to the case of discrete cmc @xmath2 surfaces as well . the definition for discrete cmc @xmath2 surfaces is already known @xcite , but the representation here for those surfaces is new . in section [ section4 ]
we look at a specific discrete example whose smooth analog is equivalent to solutions of the airy equation , and we look at the asymptotic behavior of that surface , which exhibits swallowtail singularities and a stokes phenomenon . in section [ section5 ] , we look at normal lines to discrete flat surfaces . with this
we can do several things . for example , we look at parallel surfaces ( which are also discrete flat ) and show that the area of corresponding quadrilaterals of the normal map equals the area of the quadrilaterals of the surface itself , as should be expected , since in some sense the extrinsic curvature of the surface is identically equal to @xmath2 ( note that the analogous statement is true for smooth surfaces with extrinsic curvature @xmath2 , infinitesimally ) . then , using distances from the surface s vertices to the intersection points of the normal lines , we consider a discrete analog of the extrinsic curvature and see that it is @xmath2 in the discrete case as well .
furthermore , those intersections give us a means to define discrete caustics , and , as mentioned above , we use those caustics to study the nature of `` singularities '' on discrete flat surfaces , in the final section [ section6 ] .
the authors thank udo hertrich - jeromin for fruitful discussions and valuable comments .
a useful choice of coordinates for a surface is isothermic coordinates .
not all surfaces have such coordinates , but cmc surfaces in space forms such as @xmath3 and @xmath1 do have them , away from umbilic points .
isothermic coordinates will be of central importance in this paper .
another useful tool in the study of surfaces in space forms is the hopf differential , which is defined as @xmath6 , where the surface @xmath7 is a map from points @xmath8 in a portion of the complex plane @xmath9 , @xmath10 is the bilinear extension of the metric for the ambient space form to complex vectors , and @xmath11 is the unit normal to the surface . when the coordinate @xmath8 is conformal and the surface is cmc , then @xmath12 will be holomorphic in @xmath8 .
umbilic points of the surface occur precisely at the zeros of the hopf differential . locally , away from umbilics
, we can always take a smooth minimal immersion @xmath13 into @xmath3 to have isothermic coordinates @xmath14 in a domain of @xmath15 .
let @xmath11 denote the unit normal vector to @xmath7 . then , setting @xmath16 , the hopf differential becomes @xmath17 for some real constant @xmath18 , and rescaling the coordinate @xmath8 , we may assume @xmath19 .
let @xmath20 be the stereographic projection of the gauss map @xmath11 to the complex plane , and set @xmath21 .
as we are only concerned with the local behavior of the surface , and we are allowed to replace the surface with any rigid motion of it , we may ignore the possibility that @xmath20 has poles or other singularities , and so the map @xmath22 is holomorphic . because we avoid umbilic points of @xmath7
, we also know that @xmath23 is never zero . thus the weierstrass representation is ( with @xmath24 regarded as lying in the complex plane @xmath9 ) @xmath25 associating @xmath26 , @xmath27 and @xmath28 with the quaternions @xmath29 , @xmath30 and @xmath31 , respectively , we have @xmath32 we have converted to a formulation using quaternions here , because this type of formulation has been used to define discrete minimal surfaces in @xmath3 and discrete cmc @xmath2 surfaces in @xmath1 , and we wish to make comparisons to those formulations .
note that by restricting to isothermic coordinates , we can then determine minimal surfaces by choosing just one holomorphic function @xmath20 . to define discrete minimal surfaces , we use discrete holomorphic functions @xmath33 , where @xmath34 is the square integer lattice @xmath35 , or a subdomain of it .
discrete holomorphic functions are defined as follows : defining the cross ratio of @xmath20 to be @xmath36 we say that @xmath20 is _ discrete holomorphic _ if there exists a discrete mapping @xmath37 to @xmath38 such that @xmath39 with @xmath40 and @xmath41 for all quadrilaterals ( squares with edge length @xmath2 and vertices in @xmath34 . ) see @xcite .
we call the discrete map @xmath37 a _ cross ratio factorizing function _ for @xmath20 .
note that @xmath37 is defined on edges of @xmath34 , not vertices .
note also that @xmath37 is symmetric , that is , @xmath42 and @xmath43 .
there is a freedom of a single real factor in the choice of these @xmath44 and @xmath45 , since we could replace all of them with @xmath46 and @xmath47 for any nonzero real constant @xmath48 , and all relevant properties would still hold . throughout this paper
we use @xmath48 to denote that free factor . in the above definition of the cross ratio
, we have a product of four terms .
since @xmath49 , these terms all commute , and so we could have written this cross ratio simply as a product of two fractions . however , when we later consider the cross ratio for quaternionic - valued objects or matrix - valued objects , commutativity no longer holds and the order of the product in the cross ratio becomes vital .
so , for later reference , we have chosen to write the cross ratio in the somewhat cumbersome way above .
the definition above for discrete holomorphic functions is in the `` broad '' sense .
the definition in the `` narrow '' sense would be that @xmath50 is identically @xmath51 on @xmath34 ( see @xcite , @xcite ) .
furthermore , note that , unlike the case of smooth holomorphic functions , the discrete derivative or discrete integral of a discrete holomorphic function is generally not another discrete holomorphic function .
let us exhibit some examples of discrete holomorphic functions : 1 .
let @xmath52 , and set @xmath53 for @xmath54 a complex constant .
2 . let @xmath55 , and set @xmath56 for @xmath54 a real or pure imaginary constant .
one could also take the function @xmath57 for choices of real constants @xmath58 and @xmath59 so that the cross ratio is identically @xmath51 , giving a discrete holomorphic function in the narrow sense .
[ item5 ] in section [ section4 ] we will describe a discrete flat surface based on a discrete version of the power function @xmath60 ( @xmath61 ) , which we define here . this function is discrete holomorphic in the narrow sense .
it is defined by the recursion @xmath62 we start with @xmath63 . for @xmath64
, the initial conditions should be @xmath65 we can then use to propagate along the positive axes @xmath66 and @xmath67 with @xmath68 and @xmath69 , respectively .
we can then compute general @xmath70 ( for both @xmath71 and @xmath72 ) by using that the cross ratio is always @xmath51 .
the @xmath70 will then automatically satisfy the recursion relation .
this definition of the discrete power function can be found in bobenko @xcite .
( it is also found in a recently published textbook @xcite . )
agafonov @xcite showed that these discrete power functions are embedded in wedges ( see figure [ fct43 ] ) , and are schramm circle packings ( see @xcite ) .
note that , for @xmath73 and @xmath74 , @xmath75 where @xmath76 denotes the pochhammer symbol , and a closed expression for general @xmath70 is still unknown .
we explore this difference equation in more detail in appendix [ appendix ] at the end of this paper . [ cols="^,^ " , ]
in this and the next section ( and in appendix [ appendix9 ] ) , since we will consider flat surfaces and their caustics exclusively , we will abbreviate the notation `` @xmath77 '' to `` @xmath7 '' .
let @xmath78 be the collection of focal points , i.e. the _ focal surface _ , also called the _ caustic _ , of a smooth flat surface @xmath7 in @xmath1 ( note that we should assume @xmath7 is a flat front , so that the caustic will exist ) . if @xmath79 is the lift of @xmath80 ( determined from @xmath20 ) , then ( see @xcite ) @xmath81 is a lift of @xmath82 .
although we have just described the caustic in terms of weierstrass data , it is independent of the choice of that data .
we know that @xmath78 is a flat surface ( see @xcite , @xcite ) , because its lift @xmath83 satisfies the following equation : @xmath84 for the case of a discrete flat surface @xmath7 in @xmath1 with discrete lift @xmath85 , we must first consider how to define the caustic @xmath78 .
we can define the normal @xmath86 as in at each vertex @xmath87 of @xmath7 , so we have normal geodesics emanating from each vertex , and we can consider when normal geodesics of adjacent vertices will intersect .
once we have those intersection points , we will see that we can consider them as vertices of @xmath78 , giving us a definition for @xmath78 .
[ single - caustic - lemma ] let @xmath7 be a discrete flat surface in @xmath1 with lift @xmath85 , as in , constructed using the discrete holomorphic function @xmath20 .
let @xmath88 be a cross ratio factorizing function for @xmath20 .
then the normal geodesics in @xmath1 emanating from two adjacent vertices @xmath87 and @xmath89 will intersect if and only if @xmath90 , in which case the intersection point is unique and is equidistant from @xmath87 and @xmath89 .
furthermore , even when the two normal geodesics do not intersect , they still lie in a single common geodesic plane .
consider one edge @xmath91 . applying an isometry of @xmath1 if necessary
, we may assume without loss of generality that @xmath92 then , with @xmath93 , @xmath94 thus @xmath95 the condition for the two normal geodesics to intersect is that there exist reals @xmath96 and @xmath97 so that @xmath98 in other words ( we now abbreviate @xmath99 to @xmath100 , and @xmath88 to @xmath37 ) , @xmath101 @xmath102 there exists a @xmath97 so that this last sum on the right - hand side is a diagonal matrix if and only if latexmath:[\[\left| \frac{\lambda \alpha - |dg|^2}{\lambda \alpha + @xmath1 if and only if @xmath104 this is equivalent to @xmath105 and then @xmath97 satisfies @xmath106 now , to get the diagonal terms in the above matrix equation to match , we want @xmath96 such that @xmath107\ ; , \ ] ] @xmath108\ ; .\ ] ] such a @xmath96 does exist , and in fact a computation shows that this @xmath96 is equal to @xmath97
. now the sign of @xmath109 determines which side of the quadrilateral the intersection lies on , and @xmath110 if and only if @xmath111 .
note that because @xmath112 , the distance from the intersection point of the normal lines to either of @xmath87 and @xmath89 is the same . by a further isometry of @xmath1 that preserves @xmath87 and @xmath86 , we may change @xmath20 to @xmath113 for some constant @xmath114 .
thus without loss of generality we may assume @xmath115 .
it is then clear from the above equations that the two geodesics emanating in the normal directions from @xmath87 and @xmath89 both lie in the geodesic plane @xmath116 of @xmath1 , with @xmath1 represented as in section [ section3pt1 ] .
this proves the last claim of the lemma .
[ single - caustic - lemma - but - in - r31 ] by an argument similar to that of the proof of lemma [ single - caustic - lemma ] , but simpler , we also have the following statements : let @xmath7 be a discrete flat surface in @xmath1 with lift @xmath85 , as in , constructed using the discrete holomorphic function @xmath20 .
let @xmath88 be a cross ratio factorizing function for @xmath20 .
let @xmath86 denote the normal as in at @xmath87 .
then the lines in @xmath117 ( not @xmath1 ) emanating from two adjacent vertices @xmath87 and @xmath89 in the directions of @xmath86 and @xmath118 ( respectively ) will either be parallel or will intersect at a unique point that is equidistant from @xmath87 and @xmath89 . the distance from either @xmath87 or @xmath89 to that intersection point
is @xmath119 this is true on each edge @xmath91 , regardless of the sign of @xmath120 .
furthermore , these two lines stemming from @xmath87 and @xmath89 will not be parallel if @xmath121 now , for all that follows , we introduce the following assumption .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ * assumption : * for @xmath20 the discrete holomorphic map used to construct a discrete flat surface , assume that all quadrilaterals in the image of @xmath20 in the complex plane are properly embedded .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ this implies the following properties : 1 .
@xmath99 is never zero for any edge @xmath122 between adjacent vertices @xmath123 and @xmath124 in the domain @xmath34 .
note that this was already assumed in section [ sect:2pt3 ] .
2 . @xmath99 and @xmath125 are never parallel for any square of edge - length @xmath2 with vertices @xmath126 in @xmath34 .
( if @xmath99 and @xmath125 are parallel , then @xmath127 all lie in one line , which implies that the interior of the quadrilateral in the complex plane is a half - plane , which is not properly embedded . ) because the cross ratio for @xmath20 is always negative , the term @xmath120 is negative on exactly all of the horizontal edges in the domain @xmath128 of @xmath7 , or exactly all of the vertical edges . applying a @xmath129 degree rotation to the domain if necessary
, we can , and now do , assume without loss of generality that @xmath120 is negative if and only if the edge @xmath91 is vertical in @xmath34 .
thus every vertical edge @xmath91 provides a unique intersection point of the normal geodesics , which we denote by @xmath130 .
the domain @xmath131 of this new mesh is now the collection of all vertical edges of @xmath34 .
if we let the vertical edges be represented by their midpoints , then we can consider @xmath78 to also be defined on a square grid , but now a shifted one , i.e. @xmath132 .
we call the discrete surface @xmath133 the _ caustic _ , or _ focal surface _ , of @xmath7 .
this new discrete surface @xmath134 is generally not a discrete flat surface , as the vertices in the quadrilaterals of the caustic will generally not be concircular .
however , it has a number of interesting properties closely related to discrete flat surfaces , which we will describe .
the first property is stated in this lemma : [ lem6pt6 ] the quadrilaterals of @xmath130 lie in geodesic planes of @xmath1 .
this is clear from the geometry of the construction , and from the fact that adjacent normal geodesics always lie in a common geodesic plane of @xmath1 , by lemma [ single - caustic - lemma ] .
now that we have seen the proof of lemma [ single - caustic - lemma ] , we are able to give one geometric justification for why we can say the discrete surfaces in section [ section4pt3 ] are `` flat '' . we will give an argument here showing the `` discrete extrinsic curvature '' is identically @xmath2 .
for a smooth flat surface given by a frame @xmath85 solving , we find that the two functions @xmath135 give the principle curvatures of the surface , and also give the inverses of the distances to the focal points in @xmath117 of the curves in the surface along which either @xmath136 or @xmath137 is constant . here
we are considering the focal points found in @xmath117 ( not @xmath1 ) , but we are finding those focal points with respect to the normal directions to the surface in @xmath1 given by the normal vectors , which are actually tangent vectors to @xmath1 itself .
since the extrinsic curvature @xmath138 is exactly @xmath2 , if @xmath139 , we have @xmath140 and @xmath141 , and we have that @xmath142 this last right - hand equality , of course , is clear from the fact that @xmath143
. however , the right - hand equality encodes that the extrinsic curvature is exactly @xmath2 in a way that can be applied to the discrete case , as follows : the corresponding equation in the discrete case is given by the corresponding summation about the four edges ( assume @xmath144 and @xmath145 the other case can be handled similarly ) latexmath:[\[\label{artanhguydiscrete } \text{arctanh}\left(\frac{-\lambda \alpha_{pq}+ + |g_q - g_p|^2}\right ) - \text{arctanh}\left(\frac{\lambda \alpha_{qr}+ + |g_s - g_r|^2}\right ) - \text{arctanh}\left(\frac{\lambda \alpha_{sp}+ quadrilateral with vertices associated to @xmath126 ( given in counterclockwise order about the quadrilateral in @xmath34 ) in the discrete surface .
the analogous geometric meaning of @xmath147 is preserved in the discrete case , as it is the inverse of the ( oriented ) distance from either @xmath87 or @xmath89 to the intersection point of the geodesics in @xmath117 stemming off of @xmath87 and @xmath89 in the directions of @xmath86 and @xmath118 , respectively ( see remark [ single - caustic - lemma - but - in - r31 ] ) .
furthermore , equation follows immediately from the definition of the cross ratio factorizing function @xmath37 . in this sense
, we can say that the `` discrete extrinsic curvature '' is identically @xmath2 .
we now consider a second approach to discrete extrinsic curvature .
for a smooth surface of constant extrinsic curvature @xmath2 , the infinitesimal ratio of the area of the gauss map to the area of the surface is exactly @xmath2 .
so another way to give a notion that the `` discrete extrinsic curvature '' be identically @xmath2 for a discrete flat surface is to show the analogous property in the discrete case . that is the purpose of the following lemma .
let @xmath7 be a discrete flat surface in @xmath1 with normal map @xmath11 .
consider the vertices @xmath148 of one quadrilateral ( in @xmath1 ) of the surface associated with the quadrilateral with vertices @xmath123 , @xmath124 , @xmath18 , @xmath149 ( given in counterclockwise order ) in @xmath34 .
these four vertices @xmath148 also determine another quadrilateral @xmath150 , now in @xmath117 , again with vertices @xmath148 , but now with geodesic edges @xmath151 , @xmath152 , @xmath153 , @xmath154 in @xmath117 , and which is planar in @xmath117 .
likewise , the normals @xmath155 determine a planar quadrilateral @xmath156 in @xmath117 with vertices @xmath155 and with geodesic edges @xmath157 , @xmath158 , @xmath159 , @xmath160 in @xmath117 .
these two quadrilaterals @xmath150 and @xmath156 lie in parallel spacelike planes of @xmath117 and have the same area .
because @xmath148 lie in a circle @xmath161 in @xmath1 , there exists a planar quadrilateral @xmath150 in @xmath117 with edges that are geodesics in @xmath117 , and with vertices @xmath148 . since @xmath86 and @xmath118 have reflective symmetry with respect to the edge of @xmath150 from @xmath87 and @xmath89 ( see remark [ single - caustic - lemma - but - in - r31 ] ) , and since similar symmetry holds on the other three edges of @xmath150 , we know that @xmath162 are the vertices of a planar quadrilateral @xmath163 with geodesic edges in @xmath117 .
it follows that @xmath155 are then the vertices of a planar quadrilateral @xmath156 with geodesic edges in @xmath117 .
in fact , @xmath150 , @xmath163 and @xmath156 all lie in parallel spacelike planes .
the goal is to show that @xmath150 and @xmath156 have the same area .
since @xmath150 and @xmath156 are parallel , it is allowable to replace the metric of @xmath117 with the standard positive - definite euclidean metric for @xmath164 and simply prove that @xmath150 and @xmath156 have the same area with respect to that metric .
the advantage of this is that it allows us to use known computational methods involving mixed areas .
see @xcite , for example , for an explanation of mixed areas .
noting that @xmath165 lies in the @xmath0-dimensional light cone of @xmath117 , i.e. @xmath166 , we define the lightlike vectors @xmath167 ( in fact , the normal geodesic at each vertex of @xmath7 in @xmath1 is asymptotic to the lines in the light cone determined by @xmath168 and @xmath169 . )
the two concircular sets @xmath170 and @xmath171 have the same real cross ratio , so the two quadrilaterals in @xmath117 that they determine are either congruent or dual to each other .
( in fact , it is generally the second of these two cases that occurs . )
it follows that the mixed area of @xmath168 and @xmath169 is zero @xcite . we then have , with `` @xmath172 '' denoting area and `` @xmath173 '' denoting mixed area , @xmath174 @xmath175 hence @xmath176 .
the proof of lemma [ single - caustic - lemma ] shows that for a vertical edge @xmath122 ( of length @xmath2 ) of @xmath34 , the geodesic edge @xmath151 , the geodesic through @xmath87 in the direction of @xmath86 and the geodesic through @xmath89 in the direction of @xmath118 form the boundary of a planar equilateral triangle in @xmath1 . in particular , it follows that @xmath87 , @xmath89 , @xmath177 and @xmath178 are concircular , for any value of @xmath179 .
one can also show that @xmath87 , @xmath89 , @xmath177 and @xmath178 are concircular even when @xmath122 is a horizontal edge in @xmath34 .
this shows that the quadrilaterals formed by the two points of an edge and the two points of the corresponding edge of a parallel flat surface are always concircular .
this in turn implies that a quadrilateral of the surface and the corresponding quadrilateral on a parallel surface have a total of eight vertices all lying on a common sphere - forming a cubical object with concircular sides .
this gives ( see @xcite ) a discrete version of a triply orthogonal system , that is , a map from @xmath180 or a subdomain of @xmath180 to @xmath3 where all quadrilaterals are concircular .
in lemma [ lem6pt6 ] we gave one property of caustics that is closely related to discrete flat surfaces .
here we give a second such type of property , as seen in theorem [ anotherthm - forcaustic ] below .
the equation for the lift @xmath85 of @xmath7 is @xmath181 however , since the formula for the surface has a mitigating scalar factor @xmath182 , we can change the equation above so that the potential matrix has determinant one , without changing the resulting surface , so let us instead use : @xmath183 @xmath184 we also now assume that @xmath48 is sufficiently close to zero so that @xmath185 for all edges @xmath91 .
we now define , for each vertical edge @xmath91 , @xmath186{\lambda \alpha_{pq } } } & 0 \\ 0 & \frac{\sqrt[4]{\lambda \alpha_{pq}}}{\sqrt{dg_{pq } } } \end{pmatrix } \cdot p \ ; , \ ] ] where @xmath187 and @xmath188 are any choice of nonnegative reals such that @xmath189 ( recall the definition of @xmath190 in ) .
this is a natural discretization of the @xmath83 for the case of smooth surfaces ( see ) , where we now must take a weighted average of @xmath191 and @xmath192 , and we allow any choice of weighting @xmath193 . for the smooth case in equation ,
the fourth root of the upper right term of @xmath194 divided by the lower left term gives the @xmath195 appearing in equation .
for the discrete case in equation , the fourth root of the upper right term of @xmath196 divided by the lower left term gives the @xmath197{\lambda \alpha_{pq}}$ ] appearing in equation here .
this explains why we insert the @xmath198{\lambda \alpha_{pq}}$ ] factors here .
note that @xmath198{\lambda \alpha_{pq}}$ ] is not real , because @xmath90 .
[ anotherthm - forcaustic ] the formula @xmath199 for the discrete caustic holds for all vertical edges @xmath91 , and this formula does not depend on the choice of @xmath187 and @xmath200 .
a computation gives @xmath201@xmath202 the scalar factor @xmath203 is the only part of @xmath204 that depends on @xmath187 and @xmath188 , but this scalar factor is irrelevant in the formula , so we see independence from the choice of @xmath187 and @xmath188 .
the result now follows from the proof of lemma [ single - caustic - lemma ] .
however , the choice of normal direction at the vertices of @xmath134 does depend on the choice of @xmath187 and @xmath188 , as the following equation shows : @xmath205 where @xmath206 is @xmath207 equation implies that , for both horizontal and vertical edges , the adjacent normal geodesics of the caustic typically do not intersect , for any generic choice of @xmath187 and @xmath188 .
the purpose of the next results is to show that the discrete caustics in section [ section5 ] have properties similar to the caustics in the smooth case . in what follows
, we will regard both the discrete flat surface @xmath7 and its caustic @xmath134 as discrete surfaces that have edges and faces in @xmath1 ( not just vertices ) . since there is a unique geodesic line segment between any two points in @xmath1 ,
the edge between any two adjacent vertices of @xmath7 is uniquely determined .
the same is true of @xmath134 .
then , since the image of the four vertices of any given fundamental quadrilateral in @xmath34 ( resp . in @xmath131 ) under @xmath7 ( resp .
@xmath134 ) has image lying in a single geodesic plane ( see theorem [ thm : concirc - quads ] and lemma [ lem6pt6 ] ) , the image of the fundamental quadrilateral in @xmath34 ( resp .
@xmath131 ) can be regarded as a quadrilateral in a geodesic plane of @xmath1 bounded by four edges of @xmath7 ( resp .
@xmath134 ) .
this is the setting for the results given in this section . in the case of a smooth flat surface ( front )
@xmath7 in @xmath1 and its smooth flat caustic @xmath78 , every point in @xmath78 is a point in the singular set of one of the parallel flat surfaces of @xmath7 . in lemma
[ finalprop1 ] , we are stating that every point in the discrete caustic @xmath78 of a discrete flat surface @xmath7 is a point in the edge set of some parallel flat surface of @xmath7 .
this , in conjuction with theorem [ finalprop2 ] , suggests a natural candidate for the definition of the singular set of a discrete flat surface .
the proof of lemma [ finalprop1 ] is immediate from the definitions of parallel surfaces and caustics .
[ finalprop1 ] let @xmath7 be a discrete flat surface defined on a domain @xmath208 determined by a discrete holomorphic function @xmath209 with properly embedded quadrilaterals , and let @xmath78 be its caustic .
let @xmath210 be any point in @xmath78 , so @xmath190 lies in the quadrilateral @xmath211 of @xmath78 that is determined by two adjacent vertices @xmath87 , @xmath89 of @xmath7 and the normal geodesics ( which contain two opposite edges of @xmath211 ) in the directions @xmath86 , @xmath118 at @xmath87 , @xmath89 , respectively .
( thus @xmath122 will be a horizontal edge of @xmath34 . )
@xmath190 can lie in either the interior of @xmath211 , or an edge of @xmath211 , or could be a vertex of @xmath211 . then @xmath190 lies in the edge @xmath212 of some parallel surface @xmath213 of @xmath7 .
the next proposition will be used in the proof of theorem [ finalprop2 ] .
[ finalprop3 ] let @xmath7 be a discrete flat surface with normal @xmath11 produced from a discrete holomorphic function @xmath20 with properly embedded quadrilaterals .
then for all vertices @xmath123 , @xmath86 is not tangent to any quadrilateral of @xmath7 having vertex @xmath87 .
take a quadrilateral with vertices @xmath87 , @xmath89 , @xmath214 and @xmath215 of @xmath7 so that @xmath122 is a horizontal edge of @xmath34 .
we may make all of the assumptions in the proof of lemma [ single - caustic - lemma ] , including the assumption that @xmath99 is real .
of course , the normal @xmath216 lies in the hyperplane @xmath217 of @xmath117 ( here we regard points of @xmath117 as hermitean matrices as in section [ section3pt1 ] ) , and so does the point @xmath89 , since @xmath115 . furthermore , @xmath122 is a horizontal edge of @xmath34 , so @xmath218 , which implies that @xmath89 does not lie in the geodesic in @xmath1 containing @xmath87 and tangent to @xmath86 .
however , @xmath20 has properly embedded quadrilaterals , so both @xmath125 and @xmath219 will not lie in @xmath38 , and thus @xmath215 will not lie in the hyperplane @xmath217 .
it follows that @xmath86 will not be parallel to the geodesic plane in @xmath1 containing the two geodesics from @xmath87 to @xmath89 and from @xmath87 to @xmath215 .
for a smooth flat surface ( front ) @xmath7 and its caustic @xmath78 , a parallel flat surface to @xmath7 , including @xmath7 itself , will meet @xmath78 along its singular set , and that singular set is generally a graph in the combinatorial sense ( whose edges consist of immersable curves ) .
furthermore , all vertices of that combinatorial graph have valence at least two . for example
, cuspidal edges form the edges of this graph , and swallowtails give vertices of this graph with valence two .
in particular , no cuspidal edge can simply stop at some point without continuing on to at least one other cuspidal edge ( as this would give a vertex of valence one ) .
the following theorem shows that an analogous property holds in the discrete case .
note that when we are speaking of the vertices of this combinatorial graph in the theorem below , these vertices are _ not _ the same as the vertices of the discrete flat surface , nor its discrete caustic , in general .
[ finalprop2 ] let @xmath7 be a discrete flat surface defined on a domain @xmath208 determined by a discrete holomorphic function @xmath209 with properly embedded quadrilaterals , and let @xmath78 be its caustic .
let @xmath213 be a parallel surface , and let @xmath220 be the set of all @xmath210 as in lemma [ finalprop1 ] , for that value of @xmath179 , and for any adjacent endpoints @xmath123 and @xmath124 of a horizontal edge of @xmath34 .
assume that no two adjacent vertices of @xmath213 are ever equal , and that the faces of @xmath78 are embedded .
then @xmath220 is a graph ( in the combinatorial sense ) with edges composed of geodesic segments lying in the image in @xmath1 of the horizontal edges of @xmath34 under @xmath213 , and with all vertices of @xmath220 having valence at least two .
the snowman shown on the right - hand side of figure [ uglyfigure8 ] ( see example [ example4pt12 ] ) provides an example to which theorem [ finalprop2 ] applies .
the airy example in section [ section4 ] also satisfies the conclusion of this theorem ( see figure [ uglyfigure6 ] ) , although it does not actually satisfy the condition in the theorem that the faces of the caustic be embedded .
at least one of the assumptions in theorem [ finalprop2 ] that the quadrilaterals of @xmath78 are embedded and that no two adjacent vertices of @xmath213 coincide is necessary . without them , the discrete hourglass , as seen in figure [ uglyfigure8 ] ( see example [ example4pt12 ] ) , would provide a counterexample to the result . however , it is still an open question whether both of those conditions are really needed .
there are reasons why it is not obvious that we can remove one of those two conditions .
we explore those reasons in appendix [ appendix9 ] . .
] we must show that all vertices of @xmath220 have valence at least two .
there are essentially only two situations for which we need to show this , one obvious and one not obvious .
the obvious case is when an entire edge @xmath221 lies in @xmath220 , and then the result is clear .
the non - obvious case that we now describe , where only a part of @xmath221 lies in @xmath220 , is essentially only one situtation , since any other non - obvious situation can be reformulated in terms of the notation given below .
without loss of generality , replacing @xmath213 by @xmath7 if necessary , we may assume that @xmath222 .
let @xmath34 be a domain in @xmath35 containing @xmath223 , @xmath224 , @xmath225 , @xmath226 , @xmath227 , @xmath228 , @xmath229 , @xmath230 , and let @xmath20 be a discrete holomorphic function defined on @xmath34 .
let @xmath231 be the resulting discrete flat surface .
we have a normal direction defined at each vertex of @xmath7 , which determines normal geodesics @xmath232 , @xmath233 at the vertices @xmath234 , @xmath235 , respectively .
note that @xmath232 and @xmath233 never intersect ( and thus @xmath234 and @xmath235 are never equal ) , although they do lie in the same geodesic plane , and that @xmath232 and @xmath236 ( resp . @xmath233 and @xmath237 ) intersect at a single point that we call @xmath238 ( resp .
@xmath239 ) , by lemma [ single - caustic - lemma ] . the point @xmath238 ( resp .
@xmath239 ) is equidistant from the two vertices @xmath234 and @xmath240 ( resp . @xmath235 and @xmath241 ) , as in lemma [ single - caustic - lemma ] .
now the caustic @xmath134 has two quadrilaterals , described here by listing their vertices in order about each quadrilateral : @xmath242 let @xmath137 be a point in the geodesic edge @xmath243 so that @xmath137 also lies in the edge @xmath244 .
since no two adjacent vertices of @xmath7 are ever equal , it follows that @xmath137 lies strictly in the interior of @xmath243 .
( see the left - hand side of figure [ uglyfigure1 ] . ) thus , since the face @xmath245 of the caustic is embedded , there exists a half - open interval @xmath246 $ ] or @xmath247 contained entirely in the interior of @xmath248 so that @xmath249 lies in @xmath245 . thus @xmath137 can become a vertex of the graph @xmath250 .
we wish to show that the valence at @xmath137 is at least two .
this would mean that the visual representation is more like in the right - hand side of figure [ uglyfigure1 ] , where the quadrilateral of @xmath7 with vertices @xmath251 is nonembedded .
it suffices to show that there exists a half - open interval @xmath252 $ ] or @xmath253 contained entirely in the interior of the geodesic edge @xmath254 so that : 1 .
@xmath255 lies in the face @xmath256 of the caustic , and 2 .
@xmath257 . because @xmath137 lies in both
the geodesic plane determined by @xmath251 and the geodesic plane determined by @xmath258 , and because proposition [ finalprop3 ] implies these two geodesic planes are not equal , @xmath137 must also lie in the line determined by @xmath259 . then , because @xmath137 lies in the embedded face @xmath245 , and because both @xmath244 and @xmath260 lie in the geodesic planar region between @xmath261 and @xmath262 , @xmath137 must lie in the edge @xmath259 itself .
( note that we have now proven that the quadrilateral with vertices @xmath263 , @xmath264 , @xmath265 and @xmath266 is not embedded , so in fact the visual representation must be more like in the right - hand side of figure [ uglyfigure1 ] . ) keeping in mind that @xmath137 also lies in the edge @xmath267 , then since both edges @xmath259 and @xmath267 lie in the plane determined by @xmath256 , and since @xmath256 is embedded , we conclude existence of such an interval @xmath255 with the required properties . the above lemma [ finalprop1 ] and theorem [ finalprop2 ] suggest that @xmath220 has many of the right properties to make it a natural candidate for the singular set of any discrete surface @xmath213 in the parallel family of @xmath7 .
we have chosen to consider the set @xmath220 in the image @xmath268 of a discrete flat surface , rather than in the domain @xmath34 itself ( as is usually done for the singular set in the smooth case ) , because @xmath220 becomes a collection of connected curves in the image , while in the domain @xmath34 it would jump discontinuously between points in the lower and upper horizontal edges of quadrilaterals of @xmath34 ( as we have seen in the above proof ) .
however , one could remedy this by inserting vertical lines between those lower and upper edge points in quadrilaterals of @xmath34 , and then consider the set in the domain .
in figure [ discairycmc ] , we have drawn some of the discrete surfaces in the linear weingarten family associated with the discretization of the airy equation . for making these graphics
, we used the discrete holomorphic power function .
we explain here how to solve the difference equation for determining the discrete power function , as follows : @xmath269 with the initial conditions @xmath270 once we know @xmath271 and @xmath272 , the full solution is given by solving the first equation for the cross ratio .
we fix @xmath273 and set @xmath274 then , it is seen from the cross ratio condition that @xmath275 if we set @xmath276 then @xmath277 satisfies @xmath278 we define the recurrence relations : @xmath279 so that @xmath280 the initial conditions are @xmath281 the relation is written in the form @xmath282 and the eigenvalues of the @xmath283 matrix just above are @xmath284 . the difference equation satisfied by @xmath285 is @xmath286 where @xmath287 and @xmath288 . once we have determined @xmath289 , then @xmath290 determine @xmath291 then , using equation , we can determine @xmath70 for all nonegative @xmath292 and @xmath273 .
we have the following fact , which was also stated in @xcite : [ lem - app8 ] suppose that @xmath70 is the discrete holomorphic function solving and for one choice of @xmath293 , and suppose @xmath294 is the same , but with @xmath293 replaced by @xmath295 .
then @xmath70 and @xmath296 satisfy equation with @xmath297 ( resp .
@xmath298 ) on horizontal ( resp .
vertical ) edges . that holds on the edges @xmath299 and @xmath300 can be easily confirmed from equation .
then an induction argument proves the result on all other edges as well .
if @xmath301 is a smooth nonconstant holomorphic function with respect to the usual complex coordinate @xmath8 for @xmath9 , then @xmath302 is a harmonic function , and the maximum principle for harmonic functions tells us that @xmath302 can not have a local finite minimum at an interior point of the domain .
thus , if @xmath303 has a local minimum at an interior point @xmath304 , it must be that @xmath305 .
there are various ways to discretize the notions of holomorphicity and harmonicity .
see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , to name just a few of the possible references however , the history of this topic goes back much further than just the references mentioned here .
these ways do provide for discrete versions of the maximum principle .
the definition we have chosen here for discrete holomorphic functions based on cross ratios , however , does not satisfy a particular simple - minded discrete version of the maximum principle , as we can see by the first explicit example below . a more sophisticated consideration is needed to produce a proper discrete version of the maximum principle , but we do not discuss that here , as the simplest questions are what are relevant to theorem [ finalprop2 ] .
[ exa9pt1 ] set @xmath306 .
then , with @xmath307 , set @xmath308@xmath309@xmath310 and extend @xmath20 to all of @xmath34 so that all cross ratios of @xmath20 on @xmath34 are @xmath51 .
that is , take @xmath20 so that all @xmath311 and all @xmath312
. then we have the following two properties : 1 . amongst all vertices of @xmath34
, @xmath303 has a strict minimum of @xmath313 at the interior vertex @xmath314 .
2 . amongst all edges of @xmath34 between
adjacent vertices @xmath123 and @xmath124 ( both horizontal and vertical ) , @xmath315 has a strict minimum of @xmath2 at the interior edge from @xmath314 to @xmath316 . which meet @xmath213 for some @xmath179 at a single point , with @xmath20 taken as in examples [ exa9pt1 ] and [ exa9pt2 ] .
the caustics are shown in red , and the @xmath213 are shown in blue.,title="fig : " ] which meet @xmath213 for some @xmath179 at a single point , with @xmath20 taken as in examples [ exa9pt1 ] and [ exa9pt2 ] .
the caustics are shown in red , and the @xmath213 are shown in blue.,title="fig : " ] taking @xmath20 as in the previous example , and taking @xmath317 , we can produce the flat surfaces @xmath213 and the caustic @xmath78 .
it turns out that the surface @xmath213 and the caustic @xmath78 intersect at just one point if @xmath318 is chosen correctly .
see the left - hand side of figure [ fig - appendix9 ] . in this case , two adjacent vertices ( coming from @xmath314 and @xmath316 ) of @xmath213 do coincide .
however , the quadrilaterals of the caustic @xmath78 are not embedded in this case , so this example does not suffice to show that both assumptions at the end of the first paragraph of theorem [ finalprop2 ] are truly needed . in light of the equations in the proof of lemma [ single - caustic - lemma ] , a natural next step toward understanding the role of coincidence of vertices of @xmath213 in theorem [ finalprop2 ] ( and thus toward understanding which assumptions the theorem really needs ) is to consider a discrete holomorphic function with the properties as in the next explicit example .
[ exa9pt2 ] taking the same domain @xmath34 as in the previous example , we now set @xmath319@xmath320@xmath321 and , like in the previous example , we extend @xmath20 to all of @xmath34 so that all @xmath311 and all @xmath312
. then we have the following two properties : 1 . amongst all vertical edges of @xmath34 between adjacent vertices @xmath123 and @xmath124 , @xmath315 has a strict minimum of @xmath2 at the interior edge from @xmath314 to @xmath316 .
2 . amongst any three vertical edges @xmath322 and @xmath323 and @xmath324 at the same height in @xmath34 ,
@xmath315 is ( strictly ) minimized at the central edge @xmath300 . using the function @xmath20 in example [ exa9pt2 ] , one might hope that the resulting surface @xmath213 would show the necessity of the assumption in theorem [ finalprop2 ] that the adjacent vertices of @xmath213 do not coincide .
however , it turns out that the quadrilaterals of @xmath78 are not embedded in this case as well .
see the right - hand side of figure [ fig - appendix9 ] , where again @xmath325 and @xmath179 ( @xmath326 ) is taken so that the two vertices of @xmath213 coming from @xmath314 and @xmath316 coincide .
a. i. bobenko , c. mercat and y. b. suris , _ linear and nonlinear theories of discrete analytic functions .
integrable structure and isomonodromic green s function _ , j. reine und angew .
math . 583 ( 2005 ) , 117 - 161 . | we define discrete flat surfaces in hyperbolic @xmath0-space @xmath1 from the perspective of discrete integrable systems and prove properties that justify the definition .
we show how these surfaces correspond to previously defined discrete constant mean curvature @xmath2 surfaces in @xmath1 , and we also describe discrete focal surfaces ( discrete caustics ) that can be used to define singularities on discrete flat surfaces . along the way
, we also examine discrete linear weingarten surfaces of bryant type in @xmath1 , and consider an example of a discrete flat surface related to the airy equation that exhibits swallowtail singularities and a stokes phenomenon . |
the orbital evolution of the magellanic clouds ( mcs ) with respect to the galaxy has been considered to be one of key parameters that control dynamical and chemical evolution of the lmc and the small magellanic cloud ( smc ) and the formation processes of the magellanic stream ( ms ) and its leading arms ( e.g. , murai & fujimoto 1980 , mf80 ; gardiner & noguchi 1996 , gn96 ; bekki & chiba 2005 , bc05 ; mastropietro et al . 2005 ) .
recent observational studies on the cmpm for the lmc using the high resolution camera ( hrc ) of the advanced camera for surveys ( acs ) on the @xmath6 have derived an accurate present 3d motion of the lmc around the galaxy ( e.g. , kallivayalil et al .
2006 , k06a ; piatek et al .
2008 , p08 ) .
one of remarkable results from these studies is that the lmc can possibly have a velocity with respect to the galaxy ( @xmath7 ) being @xmath8 km s@xmath2 ( kallivayalil et al .
2006 ) , which is significantly larger than that ( @xmath9 km s@xmath2 ) predicted by one of promising theoretical models for the formation of the ms ( e.g. , gn96 ) .
the observational results thus have profound implications on the past orbits of the mcs ( e.g. , besla et al . 2007 ;
kallivayalil et al .
2006 , k06b ) , a possible common halo of the mcs ( bekki 2008 ) , and the formation processes of the ms ( e.g. , diaz & bekki 2011 , db11 ; ruzicka et al .
2010 , r10 ) .
the previous work by k06a considered that the observed pm for each field ( pm(field ) ) is a combination of the pm of the center - of - mass ( cm ) for the lmc ( pm(cm ) ) and the field - dependent residual ( pm@xmath10(field ) ) as follows : @xmath11 in estimating the pm of the lmc cm ( i.e. , pm@xmath12(cm ) ) for each of the selected high - quality 13 fields , k06a very carefully considered how the internal rotation of the lmc ( `` rotation effect '' ) and the viewing angle ( `` perspective effect '' ) influence pm@xmath10(field ) and thereby made an average of the 13 pm@xmath12(cm ) to derive the cmpm . since the average pm is not a _
simple _ average of the observed pm of the 13 fields ( i.e. , not the average of pm(field ) ) , the observed cmpm can be pretty close to the true one , if all stars have almost circular motion and if the lmc has a very thin disk .
however , the lmc has a thick disk with a bar ( e.g.,van der marel et al .
2002 , vdm02 ) , which is indicative of larger local random motion both in radial and vertical directions ( i.e. , deviation from circular motion ) .
therefore , pm@xmath12(cm ) for each field can significantly deviate from the true pm of the lmc and the average pm@xmath12(cm ) can also deviate from the true cmpm if the number of the sampled field is small .
the purpose of this letter is to show how significantly the observationally derived cmpm can deviate from the true one by using a dynamical ( n - body ) model for the lmc with a known true cmpm . in the present study
, we first pick up randomly stellar particles with the particle number ( @xmath13 ) of @xmath14 in a n - body model for the lmc with a given structure and kinematics and thereby derive the cmpm of the lmc in the same way as done in previous observational studies .
we then compare the derived cmpm with the true one so that we can discuss the possible difference between the two cmpms .
this investigation is quite important , because the possible difference between the observed and true cmpms can not be observationally discussed owing to the lack of detailed information of the 3d positions and velocities of stars in each field .
recent different observational studies on the cmpm of the lmc have revealed different cmpm and maximum circular velocity ( @xmath15 ) of the lmc ( e.g. , k06a , p08 , and costa et al .
2009 ; c09 ) , and different pm studies using almost the same data set and adopting a similar pm measurement method have derived different cmpms and @xmath15 of the lmc ( k06a and p08 ) : this is yet to be explained .
furthermore p08 already suggested that a significant scatter ( @xmath16 km s@xmath2 ) in the derived pm of the sampled 21 lmc fields is due to significant departures from circular motion .
thus it is crucial to investigate how random motion in the lmc can affect the observational estimation of the cmpm in a quantitative way .
the present lmc model is consistent with a high - mass model in bc05 in terms of the disk structure and the dark matter density profile , but it is slightly different from bc05 in the dark matter fraction and the inner rotation curve profile ( also the absence of gas ) . the modeled lmc is consistent with the observed radial structure of the disk ( e.g. , bothun & thompson 1988 ) , the total mass ( e.g. , westerlund 1997 ; p08 ) , structure and kinematics of the thick disk ( vdm02 ) , and dark matter content ( vdm02 ) .
the lmc is composed of a dark matter halo and a stellar disk with the total masses being @xmath17 and @xmath18 , respectively .
following the observational results by vdm02 showing @xmath19 within 9kpc of the lmc , we assume that a reasonable mass fraction of the dark matter halo ( @xmath20 ) ) is 0.50 - 0.67 within the adopted lmc size .
we adopted an nfw halo density distribution ( navarro , frenk & white 1996 ) suggested from cdm simulations and the `` c''-parameter is set to be 12 .
the dark matter halo is truncated at the observationally suggested tidal radius of the lmc ( @xmath21 kpc ; vdm02 ) .
we mainly investigate the `` fiducial '' lmc model with the total mass ( @xmath22 ) of @xmath23 , @xmath24 , and @xmath25 km s@xmath2 .
the radial ( @xmath26 ) and vertical ( @xmath27 ) density profiles of the disk ( with the size @xmath28 of 7.5 kpc ) were assumed to be proportional to @xmath29 , with scale length @xmath30 = 0.2@xmath28 , and @xmath31 , with scale length @xmath32 = 0.06@xmath28 , respectively : the stellar disk has the radial and vertical scale length of 1.5 kpc and 0.45 kpc , respectively .
in addition to the rotational velocity caused both by the gravitational fields of the disk and dark halo components , the initial radial and azimuthal velocity dispersions were assigned to the disk component according to the epicyclic theory with toomre s parameter @xmath33 ( binney & tremaine 1987 ) ranging from 0 to 3 in the present study . by investigating models with different @xmath33
, we can discuss how local random motion ( due to non - zero velocity dispersions ) can introduce differences between the observed and true cmpms .
we run the lmc disk model for 20 dynamical time scales so that we can construct a `` barred model '' for our investigation .
there is no significant differences in stellar kinematics between the above relaxed and unrelaxed models . the line - of - sight velocity dispersions in the @xmath34- , @xmath35- , and @xmath36-components of stellar velocities ( @xmath37 , @xmath38 , and @xmath39 , respectively ) reach their maximum values of 55 km s@xmath2 , 55 km s@xmath2 , and 43 km s@xmath2 , respectively , at @xmath40 in the standard lmc model with @xmath41 .
we investigate possible differences between observed and true 3d velocities of the cm of the lmc rather than the cmpm differences , because we can more clearly show the differences without considering the coordinate transformation and the present location of the lmc with respect to the galaxy .
the cm of the lmc is assumed to be located at the center of the coordinate ( i.e , ( @xmath34 , @xmath35 , @xmath36 ) = ( 0 , 0 , 0 ) ) in all models .
the spin of the lmc disk is specified by two angles @xmath42 and @xmath43 in units of degrees , where @xmath42 is the angle between the @xmath36-axis and the vector of the angular momentum of the disk , and @xmath43 is the azimuthal angle measured from @xmath34 axis to the projection of the angular momentum vector of the disk onto the @xmath34-@xmath35 plane .
the lmc in a model is assumed to be moving with a `
true' 3d cm velocity of * v * = ( @xmath44 , @xmath45 , @xmath46 ) and therefore the initial velocity of each @xmath47-th stellar particle ( @xmath48 ) within the lmc is described as : @xmath49 where * v*@xmath50 is the 3d motion of the particle with respect to the center of mass of the lmc and depends on the location of the particle and the adopted dynamical model of the lmc ( e.g. , @xmath15 ) .
we first randomly pick up stellar particles with the total number of @xmath13 from 200000 particles in the lmc model .
each selected stellar particle is not literally a star but represents a field with local kinematics .
then we derive the 3d velocity of the cm ( * v*@xmath51 ) from that of each field as follows : @xmath52 where * v*@xmath53 is the 3d circular velocity vector at the position of the field and thus depends on inclination angles ( i.e. , @xmath42 and @xmath43 ) of the lmc and dynamical properties of the lmc ( e.g. , @xmath15 ) : thus we here consider both rotation and perspective effects properly
. this derivation of * v*@xmath51 is done in the same way as done in the previous _ hst _ pm studies ( e.g. , k06a ) . unlike observations , there is no uncertainty in * v*@xmath53 , because both the radial dependence of the rotation curve and inclination angles ( corresponding to viewing angles in observations ) in the lmc are precisely considered .
thus the simulated cm 3d velocity ( * v*@xmath54 ) is described as follows : @xmath55 the above * v*@xmath51 can deviate significantly from the true * v * because of velocity dispersions of the lmc .
we estimate the deviation for the sampled @xmath13 fields in a quantitative way as follows : @xmath56 we also derive the difference between the simulated and true 3d velocities ( @xmath57 ) for each field as follows : @xmath58 we estimate @xmath59 ( = ( @xmath60 , @xmath61 , @xmath62 ) ) and the absolute magnitude of the projected velocity ( @xmath63 ) for each sample . the method used for deriving * v*@xmath54 by the above equations ( 3)-(6 )
is referred to as `` the standard '' method .
we also adopt a `` simple method '' in which @xmath64 is derived without considering rotation and perspective effects : we estimate @xmath64 simply by making an average of @xmath65 ( i.e. , ignoring @xmath66 in the equation ( 3 ) ) . by comparing the results derived by the simple and standard methods
, we can demonstrate that it is important to consider rotation and perspective effects in deriving the cmpm of the lmc .
the present results do not depend on * v * at all , and thus we show the results for the models with * v*=(0 , 0 , 0 ) for clarity .
we mainly show the results for the standard lmc model with @xmath67 and @xmath68 in the present study , because the results of other models are essentially the same as those in this standard model .
1 shows that the difference between the velocity of each field and true cm velocity can be quite large and become up to @xmath69 km s@xmath2 in each velocity component for some fields , if the simple method ( i.e. , @xmath70 ) is applied for this fiducial lmc model with @xmath71 km s@xmath2 .
also the simulated cm velocity for the 13 fields ( * v*@xmath54 ) can not be close to the true one owing to the local random motion ( @xmath72 km s@xmath2 ) .
on the other hand , although the difference between the velocity of each field and the true cm velocity can be large ( up to @xmath1 km s@xmath2 ) in the standard method , the simulated lmc cm velocity ( * v*@xmath54 ) can be pretty close to the true one ( @xmath73 km s@xmath2 ) .
this clearly demonstrates that if rotation and perspective effects are taken into account , the simulated cm velocity can be closer to the true one .
2 shows how the deviation from the simulated lmc cm velocity ( * v*@xmath54 ) from the true cm one depends on @xmath33 and presence or absence of the central stellar bar in the lmc disk .
the simulated cm velocities are derived for 1000 different data samples each of which has @xmath13 fields .
the difference between the simulated and true velocities can be as large as 50 km s@xmath2 in the non - barred model with @xmath41 , though the average @xmath74 for the 1000 samples is only @xmath75 km s@xmath2 .
the origin of the significant deviation in some samples is due to local random motion in the lmc disk .
it is also clear that the simulated cm velocity can deviate more significantly from the true one in the model with a higher @xmath33 owing to the larger degree of local random motion : the average @xmath74 is @xmath76 km s@xmath2 for the non - barred model with @xmath77 .
the average @xmath74 is @xmath78 km s@xmath2 for the barred model , which means that there is no remarkable difference between non - barred and barred models fig .
3 shows the degree of deviation of the simulated lmc cm velocity from the true one depends strongly on the number of fields ( @xmath13 ) such that it can be higher in models with smaller @xmath13 : the average @xmath74 for @xmath79 , 30 , and 300 are 40 km s@xmath2 , 12 km s@xmath2 , and 4 km s@xmath2 , respectively .
it should be stressed here that even if 30 fields are used , the deviation can be still as large as 30 km s@xmath2 in each velocity component .
3 also shows that the deviation of the simulated cm velocity can be quite small ( @xmath80 km s@xmath2 ) , if 300 field are used .
this result implies that at least @xmath81 fields are necessary to make a very accurate cmpm of the lmc if we estimate the cmpm using the method by k06a .
4 demonstrates that even if we adopt the simple method , the difference between the simulated and true cm velocities can be well less than 10 km s@xmath2 for @xmath82 .
this result implies that the simple average of the observed pms for a large number of fields ( or stars ) can give a quite accurate cmpm of the lmc .
the present results depend on model parameters as follows : the deviation of the simulated lmc cm velocity from the true one can be smaller in the low - mass lmc model with @xmath83 , @xmath24 , and @xmath84 km s@xmath2 : the mean @xmath85 is 16 km s@xmath2 . the present results do not depend strongly on @xmath86 : the mean @xmath85 for @xmath24 and 0.67 are 22 km s@xmath2 and 18 km s@xmath2 , respectively
. the deviation of the simulated cm velocity of the lmc from the true one can be larger in the models with higher inclination angles ; for example , the mean @xmath85 is 34 km s@xmath2 for @xmath87 and @xmath88 .
the present study has first demonstrated that the lmc cmpm derived from @xmath89 fields can possibly deviate significantly from the true one ( up to @xmath90 km s@xmath2 ) even if rotation and perspective effects are carefully taken into account .
also the deviation of the observed cmpm from the true one depends on structural and kinematical properties ( e.g. , @xmath15 ) of the lmc .
the present study however can not claim that the 3d motion of the lmc by k06a deviates significantly from the true one , but it strongly suggests that there is a possibility that the observed 3d motion using the pms for a small number of the lmc fields _ can _ be significantly different from the true one owing to local random motions of stars in the lmc .
the possible deviation is very significant in terms of modeling the ms , because only a @xmath0 km s@xmath2 difference in a velocity component can cause a significant difference in the orbital evolution of the mcs ( e.g. , the longevity of the mcs binary status ) that is crucial for the formation process of the ms ( e.g. , mf80 ) .
our study also has pointed out that @xmath81 fields of the lmc are necessary to derive a quite accurate cmpm for the lmc , because the effect of local random motion in deriving the cmpm can become negligible for such a large number of the fields .
the present study suggests that the cmpm of the smc derived by k06b can be less accurate than that of the lmc , firstly because only a small number of the smc fields ( 5 ) are used for the pm measurement , and secondly because the stellar component can be better modeled as a system dynamically supported by velocity dispersion ( e.g. , bekki & chiba 2009 ) .
the latest pm study of the lmc by vieira et al .
( 2010 , v10 ) has derived the cmpm by simply making an average of the observed pms for 3822 stars and thereby has shown that the cmpm is different from those derived in previous studies ( e.g. , k06a , b ; p08 ) ; v10 have also shown that the lmc is currently orbiting the galaxy with @xmath91 km s@xmath2 , which is significantly smaller than @xmath92 km s@xmath2 derived by k06a using the @xmath6 proper motion results .
our results suggest that the simple average for 3822 stars in v10 can give an accurate cmpm , if observation - related errors for the pm measurement are really small for each field ( star ) .
this suggests that the current 3d velocity of the lmc by v10 should be also properly included in modeling the mcs , in particular , the formation of the ms .
if @xmath7 by v10 is really close to the true one , then the results by v10 would have a profound implication for the dynamical modeling of the ms formation .
recent dynamical models that are consistent with the cmpm by k06a ( i.e. , higher @xmath7 ) have failed to reproduce the bifurcated structure of the ms and the elongated leading arms self - consistently ( e.g. , besla et al .
2010 ; r10 ) .
db11 have successfully reproduced the above two fundamental observations in a model with bound orbits of the mcs ( with respect to the galaxy ) yet a rather high circular velocity ( 250 km s@xmath2 ) of the galaxy .
they thus suggested that bound orbits of the mcs are necessary to reproduce the two observations .
if the present 3d motion of the lmc by v10 is pretty close to the true one , then the lmc is bound to the galaxy with the maximum circular velocity of @xmath93 km s@xmath2 .
this means that the ms can be formed as a result of the lmc - smc - galaxy interaction , as shown by previous ms formation models with lower @xmath7 ( e.g. , gn96 ) .
given different values of the present 3d velocities of the lmc derived by different observations ( e.g. , k06a , p08 , c09 , and v10 ) , future theoretical studies for the ms formation would need to investigate at least all of the three representative models with @xmath7 @xmath94 km s@xmath2 ( c09 ) , @xmath95 km s@xmath2 ( v10 ) , and @xmath96 km s@xmath2 ( k06a ) .
in this situation , it would be fair to claim that a theoretical ms formation model that can best reproduce fundamental properties of the ms and the leading arms would predict the lmc cmpm that is closest to the true one .
besla , g. , kallivayalil , n. , hernquist , l. , robertson , b. , cox , t. j. , van der marel , r. p. , alcock , c. 2007 , apj , 668 , 949 besla , g. , kallivayalil , n. , hernquist , l. , van der marel , r. p. , cox , t. j. , keres , d. 2010 , apj , 721 , l97 | recent observational studies using the _ hubble space telescope ( hst ) _ have derived the center - of - mass proper motion ( cmpm ) of the large magellanic cloud ( lmc ) .
although these studies carefully treated both rotation and perspective effects in deriving the proper motion for each of the sampled fields , they did not consider the effects of local random motion in the derivation .
this means that the average pm of the fields ( i.e. , the observed cmpm ) could significantly deviate from the true cmpm , because the effect of local random motion can not be close to zero in making the average pm for the small number of the fields ( @xmath0 ) .
we discuss how significantly the observationally derived cmpm can deviate from the true cmpm by applying the same method as used in the observations for a dynamical model of the lmc with a known true cmpm .
we find that the deviation can be as large as @xmath1 km s@xmath2 ( @xmath3 mas yr@xmath2 ) , if the lmc has a thick disk and a maximum circular velocity of @xmath4 km s@xmath2 .
we also find that the deviation depends both on the total number of the sampled fields and on structure and kinematics of the lmc .
we therefore suggest that there is a possibility that the observed cmpm of the lmc deviates from the true one to some extent .
we also show that a simple mean of pm for a large number of the lmc fields ( @xmath5 ) can be much closer to the true cmpm . |
normal liquids exhibit homogeneous behavior in their dynamical properties over length scales larger than the correlation length of density fluctuations .
for example , the stokes
einstein relation that relates the self diffusion constant @xmath1 , viscosity @xmath9 , and temperature @xmath10 , d , [ se ] is usually accurate @xcite .
this relation is essentially a mean field result for the effects of a viscous environment on a tagged particle . in recent experimental studies
, it has been reported that the stokes
einstein relation breaks down as the glass transition is approached in supercooled liquid systems @xcite .
translational diffusion shows an enhancement by orders of magnitude from what would be expected from eq .
( [ se ] ) @xcite .
here , we show that this breakdown is due to fluctuation dominance in the dynamics of low temperature glass formers . these pertinent fluctuations are dynamic heterogeneities @xcite .
thus , the stokes
einstein breakdown is one further example of the intrinsic role of dynamic heterogeneity in structural glass formers @xcite . in the treatment
we apply , dynamic heterogeneity is a manifestation of excitation lines in space
time @xcite .
this picture leads to the prediction of dynamic scaling in supercooled liquids , @xmath11 . here , @xmath12 is the structural relaxation time for processes occurring at length scale @xmath13 , and @xmath14 is a dynamic exponent for which specific results have been established @xcite .
this picture and its predicted scaling results differ markedly from those derived with the view that glass formation is a static or thermodynamic phenomenon @xcite .
it also differs from mode coupling theory which predicts singular behavior at non zero temperature @xcite .
this paper is organized as follows . in sec .
[ sec2 ] , we introduce our model for a supercooled liquid with a probe molecule immersed in the liquid .
simulation results are given in secs .
[ sec3 ] and [ sec4 ] .
section [ sec4 ] also provides analytical analysis of the diffusion coefficient and the stokes einstein violation , and explains the origin of the decoupling of transport coefficients based on the excitation line picture of trajectory space .
comparison of our theory with recent experimental results is carried out in sec .
we conclude in sec .
[ sec6 ] with a discussion .
we imagine coarse graining a real molecular liquid over a microscopic time scale ( e.g. , larger than the molecular vibrational time scale ) , and also over a microscopic length scale ( e.g. , larger than the equilibrium correlation length ) . in its simplest form ,
we assume this coarse graining leads to a kinetically constrained model @xcite with the dimensionless hamiltonian , h=_i=1^n n_i , ( n_i=0 , 1 ) .
[ hamil ] here , @xmath15 coincides with lattice site @xmath16 being a spatially unjammed region , while @xmath17 coincides with it being a jammed region .
we call @xmath18 the `` mobility field ''
. the number of sites , @xmath19 , specifies the size of the system . from eq .
( [ hamil ] ) , thermodynamics is trivial , and the equilibrium concentration of defects or excitations is c = n_i= 11 + ( 1/ ) , where @xmath20 is a reduced temperature . we make explicit connection of @xmath20 with absolute temperature later when comparing our theory with experimental results .
the dynamics of these models obey detailed balance and local dynamical rules .
namely , @xmath21 { } \\ \end{array } n_i=1 , \ ] ] where the rate constants for site @xmath16 , @xmath22 and @xmath23 , depend on the configurations of nearest neighbors .
for example , in dimension @xmath5 , @xmath24 where @xmath25 reflects the type of dynamical facilitation . in the fredrickson andersen ( fa ) model @xcite , a state change is allowed when it is next to at least one defect . the facilitation function in this case
is given by , @xmath26 in the east model @xcite , dynamical facilitation has directional persistence .
the facilitation function in this case is @xmath27 in order to study translational diffusion in supercooled liquids , we extend the concept of dynamic facilitation to include a probe molecule .
the dynamics of a probe will depend on the local state of the background liquid .
when and where there is no mobility , the diffusive motion of the probe will be hindered . when and where there is mobility , the probe molecule will undergo diffusion easily . as such , in a coarse grained picture
, the probe molecule is allowed to jump from lattice site @xmath16 to a nearest neighbor site when site @xmath16 coincides with a mobile region , @xmath15 . in order to satisfy detailed balance
, we further assume that the probe molecule can move only to a mobile region , i.e. , x(t+t ) = x(t ) x n_x n_xx , where @xmath28 denotes the position of the probe at time @xmath29 .
units of time and length scales are set equal to a monte carlo sweep and a lattice spacing , respectively .
using the rules described in sec .
[ sec2 ] , we have performed monte carlo simulations of diffusion of a probe molecule in the fa and east models for various temperatures . for the purpose of numerical efficiency ,
we have used the continuous time monte carlo algorithm @xcite . in the all systems , @xmath19 was chosen as @xmath30 , and the simulations were performed for total times @xmath31 , with @xmath2 being the relaxation time of the model .
averages were performed over @xmath32 to @xmath33 independent trajectories . in fig .
[ fig1 ] , we show typical trajectories of probe molecules in the fa and east models . in the high temperature case ,
trajectory space is dense with mobile regions and there are no significant patterns in space time . as such ,
the dynamics is mean field like .
it is for this reason that the relaxation time in this case is inversely proportional to the equilibrium probability of excitation , @xmath34 ( see , for example , ref .
the probe molecule executes diffusive motion , without being trapped in immobile regions for any significant period of time .
the low temperature dynamics is different .
mobility is sparse , defects tend to be spatially isolated at a given time , and trajectory space exhibits space time patterns .
see fig .
[ fig1](b ) and ( c ) .
because of the facilitation constraint , an immobile region needs a nearest mobile region to become mobile at a later time .
the excitations therefore form continuous lines and bubble like structures in trajectory space . while inside a bubble , the probe molecule will be immobilized .
see , for example , the segment of the trajectory of a probe molecule for @xmath35 in fig .
[ fig1](b ) . due to exchanges between mobile and immobile regions ,
an immobile region can become mobile after a period of time . at that stage
the probe molecule can perform a random walk until it is again in an immobile region .
the motion of a probe molecule will manifest diffusive behavior over a time long enough for many dynamical exchanges to occur . in the east model at low temperatures such as pictured in fig .
[ fig1](c ) , the bubbles in space
time form hierarchical structures @xcite .
figure [ fig2 ] plots mean square displacements of probe molecules for the fa and east models for three different cases pictured in fig . [ fig1 ] . in the high temperature case
, the mean square displacement reaches its diffusive linear regime after a very short transient time . in the low temperature case , the probe molecule in the east model case reaches the diffusive regime after a longer time and over a larger length scale than that in the fa model with the same reduced temperature .
figure [ fig3 ] plots the diffusion coefficient of a probe molecule for the fa and the east models .
the diffusion coefficient is determined from the mean square displacement , d=_t ^2 t , where @xmath36 .
error estimates for our simulations are no larger than the size of the symbols . in the fa model ,
the diffusion coefficient exhibits arrhenius behavior for @xmath37 .
this behavior reflects the fact that relaxation dynamics in the fa model is similar to that of a strong liquid . in this regime , over more than 4 orders of magnitude in @xmath1 , the slope of @xmath38 vs
@xmath39 is close to 2 .
this result is consistent with the expected low temperature scaling ,
d_fa ~c^2 ~(-2/ ) , [ dfa ] as discussed in the next subsection . in the east model case , also pictured in fig .
[ fig3 ] , the diffusion coefficient decreases more quickly than arrhenius .
this super arrhenius behavior is due to the hierarchical nature of dynamics in the east model @xcite .
comparing the diffusion coefficients with the relaxation times of the background liquids demonstrates stokes
einstein violation in both models .
the relaxation times , @xmath2 , of the fa and the east models at different temperatures have been determined in prior work @xcite . when the stokes
einstein relation is satisfied , @xmath40 .
this behavior occurs in the fa and east models when @xmath41 , but fig .
[ fig4 ] shows that @xmath42 is enhanced from that behavior by 2 or 3 orders of magnitude when @xmath43 .
bear in mind , these deviations from stokes
einstein are @xmath5 results
. the appropriate generalization of the fa model to @xmath7 does not exhibit such large deviations . on the other hand
, we expect that generalizations of the east model , which is hierarchical and therefore fragile , will have weak dimensional dependence and continue to exhibit large deviations for @xmath7 .
we turn to the arguments that explain these claims now . for high temperatures , the local mobility field will tend to be close to its mean value , @xmath34 .
as such , both the relaxation mechanism of the material and the diffusional motion of the probe molecule make use of the same local mobility fields . for this reason , the diffusion coefficient and the relaxation time scale
are strongly coupled in this regime , leading to the stokes
einstein relation . at low temperatures , however ,
the dynamics of the system is not so simply related to the mean mobility field . here , the fluctuations of bubble structures dominate .
the relaxation time of the background liquid will approximately scale as the longest temporal extension of bubbles .
the _ persistence time _ of an individual lattice site , @xmath44 , is the time for which that site makes its first change in state .
its typical size will be intimately tied to the structural relaxation time of the liquid . for the fa model in @xmath5 , ~t_pers ~c^-3 .
[ taufa ] see , for example , refs .
@xcite .
this result is consistent with a simple argument concerning diffusive motions of excitation lines in the low temperature fa model @xcite . in particular , the structural relaxation times in the fa model is given by the time in which a typical bubble structure looses its identity through wandering motions of excitation lines .
the excitation line has a local diffusivity of @xmath45 .
( we use caligraphic @xmath46 to distinguish this diffusion constant for excitations from that for particles , @xmath1 . ) in order to form a bubble , an excitation line needs to wander distance of the order of the typical length between defects , @xmath47 .
therefore , the mean relaxation time is given by @xmath48 .
when the probe molecule is at the boundary of a bubble , it may not need to wait until the bubble closes in order to undergo diffusion ; rather , it can remain within mobile cells and diffuse around the boundary of the bubbles . in this way
, translational diffusion will be more facilitated than structural relaxation , leading to an enhanced diffusion in the fluctuation dominated low temperature region .
specifically , consider the dynamical _ exchange times _ , i.e. , the times between flipping events for a given lattice site .
see fig .
@xmath49 is such a time duration for an @xmath17 state and @xmath50 is such a time duration for an @xmath51 state .
the probe molecule can move only while in a mobile region .
further , the mean square displacement of the probe will be proportional to the number of diffusive steps that a probe molecule will take during the trajectory , @xmath52 , ^2~ ~tt_0 + t_1 .
[ msqd - fa ] here , @xmath53 is the length of a long trajectory in the fa model .
the average duration of the defect state , @xmath54 , is inversely proportional to the probability of a lattice site being mobile , @xmath34 , times the flip rate , @xmath23 . since @xmath55
, we have , t_1 ~c^-1 .
[ t1 ] from detailed balance , therefore , t_0 ~c^-2 . [ t0 ] since @xmath56 in the low temperature region , eqs .
( [ msqd - fa])([t0 ] ) give d_fa ~(x)^2 ~1t_0 ~c^2 .
this result explains eq .
( [ dfa ] ) . together with eq .
( [ taufa ] ) , it leads to d_fa ~^- , with @xmath57 in the @xmath5 fa model case .
this scaling is to be contrasted with the stokes
einstein result , @xmath58 . numerical simulation @xcite and
renormalization group analysis @xcite of higher dimension generalizations of the fa model indicate that for @xmath7 , @xmath59 .
however , the scaling @xmath60 remains true for all dimensions as it is based solely on detailed balance .
thus , for @xmath7 , @xmath61 . in other words
, there is only a weak breakdown in the stokes
einstein relation for strong liquids in @xmath7 . in the east model case ,
both the diffusion coefficient and the relaxation time show super - arrhenius behavior .
the hierarchical , fractal structure of pattern development in trajectory space for the east model does not allow a simple scaling analysis of the diffusion coefficient , and it is not obvious whether temperature independent scaling exists .
one can define a temperature dependent scaling exponents , @xmath62 and @xmath63 , @xmath64 so that d_east ~^-()/z ( ) .
interestingly , our numerical results indicate that @xmath65 is independent of temperature as shown in fig .
this exponent , @xmath66 for the @xmath5 east model , is very close to what many experiments and simulations have found for three dimensional glass forming liquids .
for example , a recent experiment finds that @xmath67 in the self diffusion of _
tris_-naphthylbenzene(tnb ) @xcite .
it was found that @xmath68 in a molecular dynamics simulation of lennard jones binary mixture @xcite and a recent detailed scaling analysis of numerical results shows @xmath69 @xcite .
presumably , such good agreement of scaling relation between the @xmath5 east model and higher dimension systems arises due to directional persistence of facilitation in the fragile liquid @xcite .
this persistence in higher dimensions causes motion to be effectively one dimensional @xcite .
therefore , dimensionality is not very significant for fragile glass formers . as such , for fragile systems , we expect that the scaling relation of the stokes
einstein violation will be reasonably well described by the @xmath5 east model .
based on this expectation , we further purse the comparison between theory and experiment .
swallen _ et al . _
@xcite measured the self translational diffusion coefficient of tnb near the glass transition temperature .
they observed an increase of @xmath70 from its high temperature limit by a factor of 400 near the glass transition temperature . in order to compare our results with these experiments ,
we need to determine the excitation concentration , @xmath34 , as function of temperature .
since tnb behaves as a fragile liquid , we determine the excitation concentration as a function of temperature by fitting the viscosity data of tnb @xcite with the generalization of the east model formulas to higher dimensions @xcite .
namely , ^2 , [ lntau ] where @xmath71 is the number of equally likely persistence directions on a cubic lattice , and ( c)= ( c_r)-j(1t-1t_r ) .
[ lnc ] the parameter @xmath72 is the energy scale associated with creating a mobile region from an immobile region , and @xmath73 is an appropriate reference temperature .
details on the fitting can be found in ref .
@xcite . taking @xmath74 ( the cubic lattice value ) and @xmath73 as the temperature
at which @xmath75 is half the value of @xmath76 , we determine that @xmath77 , and @xmath78 . the reduced temperature , @xmath20 of the east model is related to absolute temperature by = j(1t-1t_r ) + ( g / c_r ) .
once we have determined the excitation concentration as a function of the temperature , we can compare experimental data with our computed results for the stokes
einstein violation in the east model case . based on the argument that the scaling relation of the stokes
einstein violation ( @xmath79 ) remains robust in higher dimensions and from the dimensional dependence of eq .
( [ lntau ] ) , we expect .
[ east3d ] in fig .
[ fig7 ] we use this relationship to compare the extent of the stokes
einstein violation of the experimental system with our east model results .
the agreement between the two appears excellent .
there have been previous theoretical studies on the violation of the stokes
einstein relation in supercooled liquid systems . for example , kivelson and tarjus have argued that the stokes
einstein violation can be understood from their `` frustration limited domain '' model for supercooled liquids @xcite . assuming a distribution of local relaxation times associated with domain structures
, this model describes the translational diffusion and viscosity as corresponding to different averaging process of such a distribution .
their idea contrasts to ours in that the domain structure in their work is purely static , and the exchange between different domains are not considered .
hodgdon and stillinger have proposed a fluidized domain model @xcite . in their work
, it is assumed that the system consists of a sparse collection of fluid like domains in a background of more viscous media , and fluid like domains appear and disappear with a finite life time and rate .
relaxation times are determined by the rate of appearance of the fluid like domains , while translation diffusion also depends on the life
time of the domains . to the extent that these domains refer to space time and not simply space , this picture is not inconsistent with ours .
xia and wolynes have applied the so called `` random first order transition theory '' @xcite to the hodgdon
stillinger model @xcite . in this case , the picture is both mean field and static and decidedly contrary to our fluctuation dominated and dynamic view . from the perspective that stokes
einstein violation is a manifestation of fluctuation dominated dynamics , one expects that similar decoupling behavior occurs between other kinds of transport properties near the glass transition .
the extent to which such decoupling can appear depends upon microscopic details in the specific transport properties and materials under study .
for example , molecular rotations of a probe will be coupled to the mobility field , but less so than translations .
indeed , single molecule experiments indicate that rotations persist in both mobile and immobile regions of a glass former @xcite .
rotational motions can therefore average the effects of dynamic heterogeneity to a greater extent than translational motions .
as such , decoupling of rotational relaxation from structural relaxation can be more difficult to detect than violations of the stokes einstein relation .
precisely how such effects might be detected seems worthy of further theoretical analysis .
we are grateful to m.d .
ediger , d.a .
vandenbout , l. berthier and s. whitelam for discussions .
this work was supported at berkeley by the miller research fellowship ( yj ) and by the us department of energy grant no .
de - fg03 - 87er13793 ( dc ) , at oxford by epsrc grant no.gr/r83712/01 and the glasstone fund ( jpg ) , and at nottingham by epsrc grant no .
gr / s54074/01 ( jpg ) . | by applying the concept of dynamical facilitation and analyzing the excitation lines that result from this facilitation , we investigate the origin of decoupling of transport coefficients in supercooled liquids .
we illustrate our approach with two classes of models .
one depicts diffusion in a strong glass former , and the other in a fragile glass former . at low temperatures , both models exhibit violation of the stokes - einstein relation , @xmath0 , where @xmath1 is the self diffusion constant and @xmath2
is the structural relaxation time . in the strong case , the violation is sensitive to dimensionality @xmath3 , going as @xmath4 for @xmath5 , and as @xmath6 for @xmath7 . in the fragile case , however , we argue that dimensionality dependence is weak , and show that for @xmath5 , @xmath8 .
this scaling for the fragile case compares favorably with the results of a recent experimental study for a three - dimensional fragile glass former . |
the experimental investigation of the @xmath0-decay of @xmath1xe has been started more then 20 years ago . but
both two neutrino and neutrinoless modes of this process for @xmath1xe was not observed .
the results of last experiments are presented in table [ tab1 ] .
the theoretical estimations of half lifetime for @xmath3-decay are presented in table [ tab2 ] .
it is necessary to mention that in @xcite only one spectrum ( measured with enriched @xmath1xe ) was obtained . to calculate their limit
it was assumed that at any effect / background ratio in the energy range under investigation the effect did not exceed the actually measured background increased by a systematic error given in @xmath5 units ( @xmath5 is a standart deviation ) . in our work
the measurements were performed with both enriched xenon and natural xenon simultaneously . to evaluate the @xmath6-effect the energy spectra in region 2312@xmath72646 kev were analyzed .
the data for the analysis @xmath8-mode have been taken for 17280 hours of measurements .
this energy region is determined from calculated energy resolution for 2479 kev electrons ( r=7.0@xmath9 , @xmath10 kev ) and systematic error in definition of peak position ( @xmath11 kev ) .
number of events in the energy region @xmath12 kev registered by cpc _ _ n__1 and cpc _ _ n__2 in each run of measurements are presented in table [ tab4 ] . using recommendation given in @xcite and assuming that mean background is 10 events and measured one is 8 events , we obtain : @xmath13 where @xmath14 h @xmath15 yr and @xmath16 .
r. bernabei et al . , phys . lett . *
b 546 * , 23 ( 2002 ) .
r. luescher et al .
* b 434 * , 407 ( 1998 ) .
gavriljuk at al .
phys . of atomic nucl . v.*69 * , 12 , 2129 ( 2006 ) .
r. bernabei et al . , phys . lett .
* b 546 * , 23 ( 2002 ) e. caurier , f. nowacki , a. poves , j. retamosa , nucl . phys . * a 654 * , 973c ( 1999 ) .
rumyantsev , m.g .
urin , jetp lett * 61 * , 361 ( 1995 ) .
a. staudt , k. muto , h. klapdor - kleingrothaus , europhys . lett . * 13 * , 31 ( 1990 ) .
p. vogel and m.r .
zirnbauer , phys .
lett . * 57 * , 3148 ( 1986 ) v.n .
gavrin et al , `` intensivnost muonov kosmicheskikh luchei v laboratorii glubokogo zalozheniya ggnt '' , preprint inr ras , * p-698 * ( 1991 ) moskva .
gavriljuk et al . phys . of atomic nucl .
v.*67 * , 11 , 2039 ( 2004 ) .
feldman and r.d .
cousins , phys.rev.*d 57 * , 7 , 3873 ( 1998 ) . | search for @xmath0-decay of @xmath1xe with two high pressure proportional counters is carried out in baksan neutrino observatory .
the experiment is based on comparison of spectra measured with natural and enriched xenon .
the measured half life is equal to t@xmath2 yr ( 67% c.l . ) for @xmath3 decay mode .
no evidence has been found for neutrinoless @xmath0-decay .
the decay half life limit based on data measured during 17280 h is t@xmath4 yr ( 90% c.l . ) . |
bernstein polynomial basis plays an important role in computer graphics for geometric modeling , curve and surface approximation .
some interesting features have been investigated for this basis in the last decades ; for instance , it is proven to be an optimal stable basis among nonnegative bases in a sense described in @xcite .
also , it provides some optimal shape preserving features @xcite .
we refer to @xcite for detailed properties and applications in computer aided geometric design ( cagd ) .
bernstein basis has also been used for the numerical solution of differential , integral , integro - differential and fractional differential equations , see e.g. @xcite and the references therein .
however , it is not orthogonal and so leads to dense linear systems when using in numerical methods .
some numerical approaches implement the orthogonalized bernstein basis .
however , as we will see in the next section , it fails to keep some interesting properties of the bernstein basis .
another approach uses the dual bernstein polynomials ( dbp ) introduced by juttler in 1998 @xcite . to the best of our knowledge
, the dbp basis has been only discussed from the cagd point of view ( see the works of lewanowicz and wozny e.g. @xcite ) .
so it is of interest to explore some new aspects of this basis in order to facilitate the numerical methods for differential equations that are based on bernstein polynomials and to present a method for time fractional diffusion equation in two dimensions .
fractional partial differential equations ( fpdes ) have been widely used for the description of some important physical phenomena in many applied fields including viscoelastic materials , control systems , polymer , electrical circuits , continuum and statistical mechanics , etc .
the subdiffusion equation is a fpde describing the behavior of anomalous diffusive systems with the probability density of particles diffusing proportional to the mean square displacement @xmath0 with @xmath1 @xcite .
anomalous diffusion equations have been used for modeling transport dynamics , especially the continuous time random walk , the contamination in porous media , viscoelastic diffusion , etc @xcite . for the numerical solution of the one - dimensional problem
, we refer to @xcite and the references therein .
some classic numerical methods for pdes have been developed for the simulation of two - dimensional subdiffusion equation , for example the finite difference schemes @xcite , meshless methods @xcite , finite element method @xcite , alternating direction implicit methods @xcite , etc . in this paper , deriving some new aspects of dbps , we present suitable combinations of these functions in order to develop a dual - petrov - galerkin method for solving the following 2d subdiffusion equation @xcite @xmath2,\label{eq : main}\end{aligned}\ ] ] with the following initial and boundary conditions @xmath3,\label{bvs}\end{aligned}\ ] ] where @xmath4 , @xmath5 is the laplacian operator , @xmath6 , @xmath7 is the diffusion coefficient and @xmath8 is the source function . here
, @xmath9 denotes the caputo fractional derivative of order @xmath10 @xmath1 , with respect to @xmath11 defined as @xmath12 the main contribution of our work is the development of an accurate bernstein dual - petrov - galerkin method and the application for the numerical simulation of the 2d subdiffusion equation .
it is shown the method leads to sparse linear systems . to give a matrix approach of the method , we present some results concerning the dbps including a recurrence formula for the derivative , constructing a new basis using dbps , deriving the operational matrix of differentiation and also providing the transformation matrices between the dbps and the new basis .
the paper is organized as follows : section [ sec : bern=000026dual ] presents some new aspects of dbps and provides modal basis functions and the associated transformation matrices between the bases .
section [ sec : variationalformulation ] is devoted to the bernstein - spectral formulation of the subdiffusion problem ( [ eq : main])-([bvs ] ) and the stability and convergence results are discussed in section [ sec : error - estimation ] .
numerical examples are provided in section [ sec : numerical - examples ] .
the paper ends with some concluding remarks in section [ sec : con ] .
the bernstein polynomials with degree @xmath13 on the unit interval are defined by @xmath14 the set @xmath15 forms a basis for @xmath16 , the space of polynomials of degree not exceeding @xmath13 .
these polynomials possess end - point interpolation property , i.e. , @xmath17 also , the summation is one and the integral over the unit interval is constant , namely @xmath18 the derivative enjoys the three - term recurrence relation @xcite @xmath19 where we adopt the convention that @xmath20 for @xmath21 and @xmath22 . as we mentioned in the preceding section ,
the bernstein basis is not orthogonal .
the corresponding orthogonalized basis , obtained e.g. , by the gram - schmidt process fails to keep some interesting aspects of the original basis .
we will not consider this basis in the present work .
instead we turn to the dual basis .
the dbps are defined as @xmath23 with the coefficients given by @xmath24 it is verified that they satisfying the biorthogonal system [ @xcite , theorem 3 ] @xmath25 it is worth noting that less than a quarter of the entries of transformation matrix between the bernstein and dual bernstein basis @xmath26 $ ] , are to be computed directly by ( [ eq : dualcoefficients ] ) ; for it is bisymmetric , i.e. , symmetric about both of the main diagonal and antidiagonal .
another property which is used later is that the sum of the entries for each row ( column ) is equal to the order of the matrix , i.e. , @xmath27 in the next lemma , we present some properties of the dbps . [ dual properties]let @xmath13 be a fixed nonnegative integer .
the following statements hold .
\(i ) for all @xmath28 $ ] , @xmath29 .
\(ii ) for all @xmath28 $ ] , @xmath30 \(iii ) the basis functions have the same definite integral , i.e.,@xmath31 .
the first statement is an immediate consequence of the similar formula for bernstein polynomials , i.e. , @xmath32 from ( [ eq : dual ] ) , ( [ eq : sumofrowisconstant ] ) and ( [ eq : sumandintofbern ] ) , we have @xmath33 statement ( iii ) is also verified similarly .
the property ( i ) implies that @xmath34 , for @xmath35 + 1\leq i\leq n,$ ] need not to be computed by ( [ eq : dual])-([eq : dualcoefficients ] ) .
it especially gives @xmath36 .
one may use compact combinations of orthogonal polynomials as a basis in the galerkin methods for boundary value problems which leads to sparse linear systems in some problems ( see e.g. , @xcite ) . here
, we use this idea for the non - orthogonal bernstein polynomials to present a simple and accurate dual - petrov - galerkin spectral method for two - dimensional subdiffusion equation .
compact combinations of the basis functions are referred to as the modal basis functions ( see [ @xcite , section 1.3 ] ) .
vanishpsi]let @xmath37 be an integer , @xmath38 be the dual bernstein basis and @xmath39 set @xmath40 for @xmath41 where @xmath42 then , the polynomials @xmath43 vanish at 0 and 1 , so the set @xmath44 forms a basis for @xmath45 . by ( [ eq : dual ] ) and ( [ eq : bernbound ] ) , we have @xmath46 from lemma [ dual properties ]
, we infer @xmath47 by ( [ eq : dualvalue0 ] ) and ( [ eq : dualvalue1 ] ) , we obtain @xmath48 for @xmath49 it is easy to see @xmath44 is linearly independent . since @xmath50 , this set is a basis for @xmath45 .
this completes the proof .
figure [ fig : dual - bern ] . illustrates the dbps and the modal basis functions for @xmath51 .
it is seen that the modal basis functions have less values than the corresponding dual functions on the unit interval , expecting less round - off errors .
graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig:"]graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig:"]graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig : " ] graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig:"]graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig:"]graphs of dbps @xmath52 ( top ) and the modal basis functions @xmath53 ( bottom).,title="fig : " ] for @xmath37 , consider the @xmath54vector @xmath55 and the @xmath56vector @xmath57 consisting of dual functions given by ( [ eq : dual ] ) and the modal basis functions given by ( [ eq : compactbasis ] ) , respectively : @xmath58^{t},\label{eq : dualbasisvector}\\ & & \mathbf{\psi}(\cdot)=[\psi_{i}\left(\cdot\right):\,0\leq i\leq n-2]^{t}.\label{eq : modalbasisvector}\end{aligned}\ ] ] for simplicity , we ignore the dependence of the vectors on variable .
first , note that @xmath59 where @xmath60 $ ] is an @xmath61 matrix with three diagonals as @xmath62 to derive a formula for the derivative of the modal basis functions , we first prove the following result . the operational matrix for derivative of the dbps , @xmath63 satisfies @xmath64 where the matrix @xmath65 $ ] is given by @xmath66 the dbps @xmath55 is a basis for @xmath67 , so we expand @xmath68 for @xmath69 as @xmath70 integration by parts and ( [ eq : bernderiv ] ) imply that @xmath71 the biorthogonality ( [ eqbiorthogonality ] ) gives @xmath72 now , the result is proved by considering ( [ eq : dualvalue0 ] ) and ( [ eq : dualvalue1 ] ) .
the matrix @xmath63 is a sparse matrix of order @xmath73 with @xmath74 for @xmath75 @xmath76 ; for instance , see the matrix given below .
set @xmath77 for @xmath78 then , from ( [ eq : p sait to sait ] ) , we infer the following five - term recurrence relation is deduced @xmath79 where we set @xmath80 for @xmath21 and @xmath81 we derive the transformation matrices that map the bernstein and chebyshev coefficients now we derive the transformation matrix that maps the derivative of modal basis functions to dbps .
this facilitates the use of galerkin method in the next section . in the following
, @xmath82 matrix stands for a matrix with @xmath83 and @xmath84 nonzero diagonals below and above the main diagonal , respectively .
[ lemma1forq]let the vectors @xmath57 and @xmath55 be defined as in ( [ eq : dualbasisvector ] ) and ( [ eq : modalbasisvector ] ) , respectively .
then , @xmath85 where @xmath86 is an @xmath87 @xmath88 matrix given by @xmath89 combining ( [ eq : saitosaitilde(g ) ] ) with ( [ eq : p sait to sait ] ) , implies @xmath89 to prove that @xmath86 is a @xmath88 matrix , it is sufficient to show that @xmath90 for @xmath91 and @xmath92 for @xmath93 @xmath94 and for @xmath93 by ( [ dual properties ] ) @xmath95 note that @xmath96 s vanish at the boundary values according to proposition [ prop 1 .
vanishpsi ] .
the proof is complete . to see the sparsity of the transformation matrices , @xmath63 , @xmath97 and @xmath86 for @xmath98
are shown in the following .
in this section , at first the problem ( [ eq : main])-([bvs ] ) is discretized in time . then we develop a bernstein dual - petrov - galerkin method using ? .
the matrix formulation and the error estimations are also provided in this section . consider the subdiffusion equation ( [ eq : main ] ) at @xmath99 as @xmath100 let @xmath101 be an approximation of @xmath102 at @xmath103 for @xmath104 where @xmath105 is the time step length . the time fractional derivative can be approximated by definition ( [ eq : fracdef ] ) and using forward difference for the derivative inside as @xmath106 where @xmath107 and @xmath108 for @xmath109 and @xmath110 .
the error is bounded by @xmath111 where the coefficient @xmath112 depends only on @xmath102 @xcite .
the time discretization ( [ eq:4.1 - 1 ] ) is referred to as l1 approximation ( see e.g. @xcite ) . substituting from ( [ eq:4.1 - 1 ] ) into ( [ eq : attimek+1 ] ) and multiplying both sides by @xmath113 and dropping @xmath114 , the following time - discrete scheme is obtained @xmath115 with @xmath116 and @xmath117 is given by the initial condition ( [ iv ] ) with the error @xmath118 for @xmath119 it reads as @xmath120 the boundary conditions for the semidiscrete problem is @xmath121 on @xmath122 . consider the problem ( [ eq : semi - discrete ] ) with @xmath123 and the homogeneous dirichlet boundary conditions @xmath124 .
we seek an approximate solution in the sobolev space @xmath125 .
a weak formulation of the problem ( [ eq : semi - discrete ] ) is to find @xmath126 such that @xmath127 : @xmath128 let @xmath16 be the space of polynomials over @xmath129 with degree not exceeding @xmath13 and @xmath130 .
the galerkin formulation of the ( [ eq : weakform ] ) is to find @xmath131 such that @xmath132 : @xmath133 with @xmath134 being the standard @xmath135-norm and @xmath136 an interpolation operator .
since @xmath137 , and due to ( [ eq : bernbound ] ) , we choose a basis for it by removing the first and last bernstein polynomials of degree @xmath13 , i.e. , @xmath138^{t}.\label{eq : bernbasisvector}\end{aligned}\ ] ] using ( [ eq : bernderiv ] ) , it is easy to verify @xmath139 where @xmath140 is a tridiagonal matrix of order @xmath141 and @xmath142^{t}$ ] is an @xmath56vector .
assuming @xmath143 we use the tensor product of the basis functions of @xmath144 as a basis for two dimensional case , @xmath145 and consider an approximate solution of ( [ eq : semi - discrete ] ) as @xmath146 where @xmath147^{t}$ ] and @xmath148 $ ] .
let us use the following notations . @xmath149,\nonumber \\ & & b_{i , j}=\int_{i}{\phi_{j}(x)\psi_{i}(x)dx},\quad\mathbf{b}=[b_{i , j}],\nonumber \\ & & f_{i , j}^{k+1}=\int_{\omega}{i_{n}f^{k+1}(x , y)\psi_{j}(x)\psi_{i}(y)d\omega},\label{eq : integralsoff}\end{aligned}\ ] ] for @xmath150 taking the test functions of ( [ eq : galerkin ] ) as @xmath151 for @xmath152 it is seen that the spectral form ( [ eq : galerkin ] ) is equivalent to the following linear system : @xmath153 that can be equivalently written as a sylvester equation but it requires computing the inverse of * @xmath154*. although @xmath154 has only three nonzero diagonals , it can be shown that its inverse is a dense matrix and so we avoid transforming to sylvester equation .
instead we use the equivalent tensor product form @xmath155 with @xmath156^{t},\,q = n-1 $ ] .
it is worth to note that the coefficient matrix of the above system is the same for all time steps and to be evaluated just once for all @xmath157 in terms of the trial vector ( [ eq : bernbasisvector ] ) , and test vector ( [ eq : modalbasisvector ] ) , we may write @xmath158 to facilitate the computations , in what follows , these matrices are related to the transformation matrices introduced in section [ subsec : transformation - matrices - and ] .
first , note that by the biorthogonality ( [ eqbiorthogonality ] ) , we have @xmath159=:\mathbf{\tilde{i}}.\label{eq : itilde}\end{aligned}\ ] ] now from ( [ eq : saitosaitilde(g ) ] ) , and writing @xmath97 as @xmath160 $ ] , we get @xmath161.\end{aligned}\ ] ] so @xmath154 is a tridiagonal matrix whose entries are given by @xmath162 where @xmath163 s and @xmath164 s are easily computed by ( [ eq : a=00005bi=00005d and b=00005bi=00005d ] ) . on the other hand , from lemma [ lemma1forq ] , the bernstein operational matrix of differentiation ( [ eq : berndermatrix ] ) and ( [ eq : itilde ] ) , we obtain @xmath165\nonumber \\ & = & [ \mathbf{q}_{1},\dots,\mathbf{q}_{n-1}]\mathbf{d}^{t}+n[\mathbf{q}_{0},\mathbf{0},\dots,\mathbf{0},\mathbf{q}_{n}],\label{eq : amatrix}\end{aligned}\ ] ] where @xmath166 $ ] is a @xmath88 matrix introduced lemma [ lemma1forq ] .
hence , @xmath167 is a pentadiagonal matrix and @xmath168 is the product of a pentadiagonal and a tridiagonal matrix plus a sparse matrix . from lemma [ lemma1forq ] and ( [ eq : amatrix ] )
, it is seen that @xmath168 is a seven - diagonal matrix .
notice that the solution of linear system ( [ eq : tensorprodsys ] ) requires the matrices @xmath168 and @xmath169 @xmath168 is obtained by a sparse matrix - matrix multiplication ( [ eq : amatrix ] ) and entries of @xmath154 are given by ( [ eq : bentries ] ) . since the coefficient matrix of the linear system ( [ eq : tensorprodsys ] ) remains intact for a fixed @xmath170 , only the rhs vector to be computed for different time steps , @xmath171 up to a desired time .
so it is efficient to use a band - lu factorization for solving the system .
it is remarkable that for a @xmath172band matrix , the lu - factorization can be done with @xmath173 flops and backward substitutions require @xmath174 flops [ @xcite , section 4.3 ] .
for the error analysis , we assume the problem ( [ eq : main ] ) to be homogeneous , @xmath175 . for @xmath176
the bilinear form @xmath177 in ( [ eq : galerkin ] ) is continuous and coercive in @xmath178 .
the existence and uniqueness of the solution for both the weak form ( [ eq : galerkin ] ) and the galerkin form ( [ eq : galerkin ] ) is guarantied by the well - known lax - milgram lemma .
we define the following inner product and the associated energy norm on @xmath179 : @xmath180 [ th : stabilityforthe - weak - semidiscrete]the weak form ( [ eq : weakform ] ) is unconditionally stable : @xmath181 let @xmath182 in ( [ eq : weakform ] ) . then , @xmath183 giving ( [ eq : stabilityofsemidiscrete ] ) for @xmath184 by the definition ( [ eq : energynorm ] ) , the schwarz inequality and the inequality @xmath185 .
by mathematical induction , assume ( [ eq : stabilityofsemidiscrete ] ) holds for @xmath186 let @xmath187 in ( [ eq : weakform ] ) , i.e. , @xmath188 it is easy to see that the rhs coefficients in ( [ eq : semi - discrete ] ) are positive .
so we obtain @xmath189 so the proof is done .
[ th : convergenceforsemidiscrete]let @xmath102 be the solution of the equation ( [ eq : main ] ) with conditions ( [ iv])-([bvs ] ) and @xmath101 be the solution of the the semidiscrete problem ( [ eq : semi - discrete ] ) .
then , @xmath190 the idea of the proof comes from @xcite .
we first prove @xmath191 by ( [ eq : main ] ) and ( [ eq : semi - discfork=00003d0 ] ) , we have @xmath192 in which @xmath193 for @xmath194 and by using @xmath195 , @xmath185 and ( [ eq : truncationerrormulti ] ) , we get @xmath196 i.e. , ( [ eq : komaki ] ) holds for @xmath197 . by induction ,
assume ( [ eq : komaki ] ) holds for @xmath198 . using ( [ eq : main ] ) and ( [ eq : semi - discrete ] ) , we get @xmath199 for @xmath200 , it reads as @xmath201 proving ( [ eq : komaki ] ) for @xmath202 that completes the proof of ( [ eq : komaki ] ) .
consider @xmath203 , then there exists a@xmath204 @xmath205 such that @xmath206 which gives @xmath207 now using this along with ( [ eq : komaki ] ) proves ( [ eq : error u(t_k)-u^k ] ) . in order to derive ( [ eq : error u - u^k alpha->1 ] ) , we first prove @xmath208 by ( [ eq : forj=00003d1 ] ) , the inequality ( [ eq : komaki2 ] ) holds for @xmath197 .
assume ( [ eq : komaki2 ] ) holds for @xmath209 .
then , from ( [ eq : main ] ) , ( [ eq : semi - discrete ] ) and ( [ eq : truncationerrormulti ] ) , we obtain @xmath210 so ( [ eq : komaki2 ] ) holds for @xmath202 . from @xmath211 and ( [ eq : komaki2 ] ) , we get ( [ eq : error u - u^k alpha->1 ] ) .
let @xmath212 be the @xmath213-orthogonal projection operator from @xmath179 into @xmath214 associated with the energy norm @xmath215 defined in ( [ eq : energynorm ] ) .
due to the equivalence of this norm with the standard @xmath213 norm , we have the following error estimation [ @xcite ; relation ( 4.3 ) ] @xmath216 the idea of the proof for the following result comes from the paper @xcite .
let @xmath217 be the solution of the variational formulation ( [ eq : weakform ] ) and @xmath218 be the solution of the scheme ( [ eq : galerkin ] ) , assuming @xmath219 and @xmath220 for some @xmath221 then , @xmath222 for @xmath223 , where @xmath224 depends only on @xmath225 .
we have @xmath226 by the projection operator . by definition of the norm ( [ eq : energynorm ] ) , we get @xmath227 by the weak form ( [ eq : weakform ] ) , the rhs of the above equation is replaced as @xmath228 subtracting ( [ eq:11 ] ) from ( [ eq : galerkin ] ) , we have @xmath229 where @xmath230 and @xmath231 .
let @xmath232 , then @xmath233 with @xmath234 , we obtain @xmath235 as in the proof of theorem [ th : convergenceforsemidiscrete ] , it is first proved by induction that : @xmath236 for @xmath237 then , by using ( [ eq : boundfor b_k ] ) and the projection error ( [ eq : projectionerror ] ) the desired result is derived .
the following theorem is obtained by the triangle inequality @xmath238 along with the inequalities ( [ eq : error u(t_k)-u^k ] ) and ( [ eq : error u^k-(u_n)^k ] ) .
let @xmath102 be the solution of the problem ( [ eq : main ] ) with the initial and boundary conditions given by ( [ iv])-([bvs ] ) and @xmath218 be the solution of the scheme ( [ eq : galerkin ] ) . then , assuming @xmath239 and @xmath240 , we have @xmath241 the constants @xmath242 and @xmath224 are independent of @xmath170 , @xmath243 , @xmath13 .
it is seen that the method has the so - called spectral convergence in space and the order of convergence @xmath244 in time .
here , some numerical experiments are provided to show the accuracy and efficiency of the proposed method . for the computations , we use maple 18 . on a laptop with
cpu core i3 1.9 ghz and ram 4 running windows 8.1 platform . to compute the errors
, we use the discrete @xmath135 and @xmath245 errors defined as @xmath246 respectively , where @xmath102 is the exact solution of the problem ( [ eq : main])-([bvs ] ) , @xmath247 is the approximation solution ( [ eq : approxsol ] ) at @xmath248 @xmath249 and @xmath250 .
also , the convergence rates in space and time are respectively computed by @xmath251 where @xmath252 is the error with @xmath13 stands for the dimension of basis and @xmath170 is the time - step size .
however , as it is common in the literature , we will show the spectral convergence of the proposed method by logarithmic scaled error plots .
[ exa:1 sin(pi*x)sin(pi*y)t^2]consider the problem ( [ eq : main ] ) with @xmath253 and the exact solution @xmath254 .
table [ tab : spatialrateexample1 ] shows the convergence of the method for @xmath255 for some fractional orders .
also figure also , figure [ fig : ex1 ] demonstrate the logarithmic error plot for @xmath256 .
table [ tab : timerate ] illustrates the temporal rate of convergence with @xmath257 .
convergence of the spectral method in terms of some norms with @xmath258 and @xmath256 ] [ exa : ex2nosourse]to see the method works for the case in which there is no source term , consider the problem ( [ eq : main])-([bvs ] ) with the initial condition @xmath260 , @xmath253 and no source term @xcite .
the errors and also the rate of convergence are provided in table [ tab : spatialexample2 ] where the solution with @xmath257 is treated as the exact solution .
the errors are reported at @xmath261 we have used the eight point gauss - legendre quadrature rule to perform the integrals ( [ eq : integralsoff ] ) in the right hand side of the linear system ( [ eq : tensorprodsys ] ) .
numerical results confirm the convergence and the accuracy of the method .
in this paper , some new aspects of dual bernstein polynomials have been discussed .
a suitable compact combinations of these polynomials has been derived for developing a dual - petrov - galerkin variational formulation for the numerical simulation of two - dimensional subdiffusion equation .
it was shown that the method leads to sparse linear systems .
the illustrated numerical examples have been provided to show the accuracy of the method .
it is important to note that the transformation matrices and the operational matrix for differentiation of dual bernstein polynomials that have been obtained in this work can be used similarly for developing bernstein - based dual - petrov - galerkin galerkin methods for other fractional partial differential equations on bounded domains .
gao , g.h . ,
sun , h.w . ,
sun , z.z .
: stability and convergence of finite difference schemes for a class of time - fractional sub - diffusion equations based on certain superconvergence .
j. comput .
phys 280 , 510 - 528 ( 2015 ) jani , m. , babolian , e. , javadi , s. , bhatta , d. : banded operational matrices for bernstein polynomials and application to the fractional advection - dispersion equation .
numer algor .
doi : 10.1007/s11075 - 016 - 0229 - 1 maleknejad , k. , basirat , b. , hashemizadeh , e. : a bernstein operational matrix approach for solving a system of high order linear volterrafredholm integro - differential equations .
modelling 55 , 13631372 ( 2012 ) zhao , y. , zhang , y. , shi , d. , liu , f. , turner , i. : superconvergence analysis of nonconforming finite element method for two - dimensional time fractional diffusion equations . appl .
lett 59 , 38 - 47 ( 2016 ) | in this paper , we develop a bernstein dual - petrov - galerkin method for the numerical simulation of two - dimensional subdiffusion equation .
the equation is first discretized in time using the l1 approximation .
then , a spectral discretization is applied by introducing suitable combinations of dual bernstein polynomials as the test functions and the bernstein polynomials as the trial ones .
we derive the exact sparse operational matrix of differentiation for the dual bernstein basis which provides a matrix based approach for the spatial discretization .
it is also shown that the proposed method leads to banded linear systems that can be solved efficiently .
finally , the stability and convergence of the proposed method is discussed theoretically .
some numerical examples are provided to support the theoretical claims and to show the accuracy and efficiency of the method . |
the axion is a hypothetical elementary particle , which emerged originally from a proposal by peccei and quinn , intended to solve the strong cp problem @xcite in theoretical physics .
the axion is neutral , only interacts weakly with matter , has a low mass ( @xmath0 ) , spin zero , and a natural decay constant ( to 2 photons ) in the order of @xmath1 years .
the axion belongs to the family of weakly interacting sub - ev particles ( wisp ) .
another wisp , closely related to the axion is the paraphoton or hidden photon .
the existence of these wisps could not be confirmed yet and all experimental efforts to date have so far produced only exclusion results .
nevertheless there is strong motivation to advance the experimental `` low energy frontier '' as the axion is the most popular solution for the strong cp - problem
. many wisps are also excellent candidates for dark matter and explain numerous astrophysical phenomena .
wisps can be probed in the laboratory by `` light shining through the wall '' ( lsw ) experiments .
they exploit the very weak coupling to photons , allowing an indirect proof of the otherwise hidden particles without relying on any cosmological assumptions .
previous lsw experiments have been carried out with optical laser light at desy ( alps ) , cern ( osqar ) and fermilab ( gammev ) .
the concept of an optical lsw experiment can be adapted to microwaves @xcite .
a block diagram of the setup is shown in fig . [
fig : ovrblock ] , it consists of two identical low loss microwave cavities with a diameter of 140 mm , a height of 120 mm and a spacing between them of 150 mm .
one serves as wisp emitter and is excited by an external microwave source .
it develops a strong electromagnetic ( em ) field , which corresponds to a large amount of microwave photons @xmath2 .
theory predicts that some of these photons convert to paraphotons @xmath3 by kinetic mixing ( similar to neutrino oscillations ) or if the cavities are placed in a strong static magnetic field to axion - like particles by the primakoff effect @xcite .
both particles only interact very weakly with matter ( similar to neutrinos in this respect ) and thereby , in contrast to the photons , can traverse the cavity walls .
some wisps propagate towards the detection cavity , which is connected to a very sensitive microwave receiver .
the reciprocal conversion process transforms wisps to microwave photons , which can be observed as an excitation of the seemingly empty and well shielded detection cavity . since there is no energy loss associated with the wisp conversion process
, the regenerated photons in the detecting cavity have exactly the same energy as the photons in the emitting cavity .
thus , the signal which is coupled out from the detection cavity has the same frequency as the one which is generated on the emitting side , making a narrowband receiving concept feasible .
this paper will focus on the latest exclusion results for * paraphotons * from the microwave wisp search at cern . in a future upgrade
, an additional magnet will allow the search for axions .
considering current exclusion limits , it takes @xmath4 photons on the emitting side to generate one photon on the detection side , making this the most challenging aspect of an lsw experiment . the expected output power ( or photon flux ) from the detecting cavity towards the microwave receiver due to paraphotons is given by eq .
[ equ : power ] , @xmath5 where @xmath6 and @xmath7 are the loaded q factors of emitting and detection cavity , @xmath8 is the frequency where the experiment is carried out ( and to which the cavities are tuned ) , @xmath9 is planck s constant and @xmath10 is a dimensionless geometric form factor in the order of 1 , describing the position , shape and resonating mode of the cavities @xcite .
the rest mass of hidden photons is a priori unknown and given by @xmath11 .
the kinetic mixing parameter @xmath12 describes the likeliness of paraphoton - photon oscillations . a previous examination of coloumb s law indicates that @xmath13 in this energy range .
if there is no significant signal detected , an exclusion result can be produced by determining @xmath12 from the other known values .
this provides a convenient way to compare the achieved sensitivity to other experiments .
the parameters of the paraphoton experiment as it has been set up and carried out at cern in march 2012 , are summarized in table [ tbl : param ] . as no paraphotons
were observed , the corresponding exclusion limit in comparison to other experiments is shown in fig .
[ fig : exclplot ] .
.parameters of the paraphoton run in march 2012 [ cols="^ " , ]
on the left side of fig .
[ fig : ovrblock ] , a commercial microwave source is shown , which generates a signal on @xmath8 ( see table [ tbl : param ] ) which is amplified up to 50 w and drives the emitting cavity on its resonant frequency .
power is coupled in and out of each cavity with a small inductive coupling loop , adjusted for critical coupling to the te@xmath14 mode .
this mode has been chosen for its high q - factor and reasonable coupling to paraphotons compared to other modes .
the loaded q - factor of the silver coated brass cavities has been determined by a network analyzer , their 3 db bandwidth is @xmath15 khz
. a tuning stub perturbing the h - field allows to compensate manufacturing tolerances within a bandwidth of @xmath16 mhz . shielding is required around the detecting cavity and the microwave receiver to eliminate ambient electromagnetic interference ( emi ) and to mitigate coupling to the emitting cavity by simple em leakage
this would generate false positive results as a signal originating from leakage can not be distinguished from a signal propagating by wisp conversion . within 15 cm
, the field strength must be reduced by at least a factor of @xmath17 to get meaningful results .
the shielding box has been built from a straight piece of wr-2300 waveguide with the inside dimensions 58x29x100 cm .
feeding all rf signals over optical fibres by analog transceivers prevents the propagation of emi trough ordinary transmission lines .
an optical ethernet link is used to remote control the signal analyser .
shielding effectiveness has been measured by comparing the electric field strength in- and outside the wall with a calibrated electric field probe and a spectrum analyzer .
the shielding enclosure provides @xmath18 db and each of the cavities provides an additional @xmath19 110 db of shielding .
the combined em attenuation is @xmath20 db , making thermal noise the limiting factor for the minimum detectable signal . for an exclusion result
it is necessary to prove the detector is working and actually able to pick up any wisp related signals .
detection sensitivity will be limited if the resonant frequency of one or both cavities does not equal the system frequency @xmath8 .
this is especially delicate as the cavities are sensitive to temperature variations ; their resonant frequency is inversely proportional to the thermal expansion coefficient of their wall material . for brass ,
a temperature change of @xmath21 k leads to a detuning of @xmath22 khz and a reflection of around half the input power back towards the amplifier .
thermal drift is more critical for the emitting cavity as it has to dissipate up to 50 w of heat by forced air cooling without any external temperature stabilization . before data taking
, the cavity was operated at full power and its resonant frequency was kept constant manually .
after around 1 h , the cavity reached thermal equilibrium and no further tuning was necessary .
the reflected power @xmath23 indicates how far the emitting cavity is off tune , and is minimized for @xmath24 .
the relation is given in eq .
[ equ : cavreflfreq ] , @xmath25 where @xmath26 is the constant ( @xmath27 ) and known incident rf power at the frequency @xmath8 .
critical coupling has been assumed , with coupling losses and impedance mismatch not considered .
@xmath23 is measured on a directional coupler , placed between power amplifier and cavity .
tuning the detecting cavity once before the measurement run is sufficient , as it does not dissipate any power and changes in the laboratory s ambient temperature are small enough .
its resonant frequency was estimated by evaluating the spectral noise power density @xmath28 at the output of the low noise amplifier ( lna ) , given by eq .
[ eq : thnoise ] @xmath29\end{aligned}\ ] ] where @xmath30 db is the lna s gain and @xmath31 is the boltzmann constant .
three noise temperature terms have to be considered , @xmath32 describes the noise from the detecting cavity itself .
its noise temperature is frequency dependent by the reflection coefficient @xmath33 .
the maximum occurs at @xmath34 and is equal to ambient temperature .
k describes the intrinsic noise temperature of the amplifier .
@xmath36 describes the noise temperature of the amplifier input , which emits a noise wave towards the cavity where it is reflected back . as the transmission line between amplifier and cavity is short ( @xmath19 1 cm ) and
as the noise temperature of the cavity walls is significantly higher than the noise wave transmitted from the amplifier s input , a good estimate of the resonant frequency can be determined from the maximum of the noise power spectrum .
this has been done before and after the actual experimental run , the results are shown in fig .
[ fig : diagtune ] and indicate no significant ( bigger than @xmath37 ) drift of the emitting cavity s resonant frequency . , scaledwidth=50.0% ] the output signal of the detecting cavity is coupled out , amplified by the lna ( @xmath30 db ) and then further processed by an agilent exa n9010a signal analyzer .
the center frequency was set to @xmath38 to avoid internal spurious signals appearing at interesting parts of the spectrum .
the center frequency is shifted to baseband and the complex iq signal is digitized with 20 hz bandwidth .
the 11.5 h long time record is stored for further processing . as the capture memory of the analyzer limits the maximum number of continuously acquired samples to @xmath39 , a trade - off between recorded bandwidth and recording length had to be found .
for this reason the signal analyzer will be replaced by a dedicated processing chain without this limitation for the future . for offline data processing the spectral power of the recorded noise like signals
is estimated by welch s method implemented in a python script .
the time record is read and divided into segments overlapping by @xmath40 .
each segment is multiplied by a hann window and the complex spectra are calculated by a fast fourier transform . averaging these spectra trades resolution bandwidth for less variance in the noise floor . for a 11.5 h long time trace ,
resolution bandwidths as narrow as @xmath41hz can be achieved .
the average noise floor is determined by @xmath42 .
the frequency error of rf - source and signal analyzer needs to be within @xmath43 during the experimental run , otherwise the signal power will spread out over several bins in the spectrum , degrading the signal to noise ratio . while absolute frequency drifts are unavoidable , phase - locking rf - source and signal analyzer to a common 10 mhz frequency reference allows to achieve a good relative frequency stability .
this has been explained and successfully demonstrated down to @xmath44hz in @xcite .
immediately before the actual measurement , a test run was conducted where the shielding box was left open to provoke em leakage between the cavities .
the resulting power spectrum contained a single peak , clearly above the noise floor , spanning only one single bin .
its absolute position on the frequency axis was offset by @xmath45 mhz due to the finite resolution of the rf source . as no parameters were changed after the test run and only the shielding box
was closed , the wisp related signal is expected to appear at the same position in the spectrum .
the peaks within a window of @xmath46 mhz around the expected signal do not exceed the peaks in other parts of the spectrum , as shown in fig .
[ fig : spectra ] .
therefore an exclusion result is produced by setting the minimum detectable power ( @xmath47 dbm ) to the maximum peak within the frequency window .
no paraphotons were observed in the first measurement - run of the microwave wisp search at cern , improving the existing exclusion limits . several technical challenges , like @xmath48 db
em shielding between the cavities , keeping them on tune during the 11.5 h measurement run and filtering the signal with a bandwidth of @xmath41hz , had to be overcome .
+ we are grateful for the practical hints and assistance from m. gasior and m. thumm . thanks to r. jones , e. jensen and the be department management for support .
m. betz et al . , _ `` status report of the cern light shining through the wall experiment with microwave axions and related aspects '' _ , to appear in proc .
7th patras axion workshop , mykonos 2011 , arxiv:1204.4370[http://arxiv.org / abs / arxiv:1204.4370 ] | for the microwave equivalent of `` light shining through the wall '' ( lsw ) experiments , a sensitive microwave detector and very high electromagnetic shielding is required .
the screening attenuation between the axion generating cavity and the nearby detection cavity should be greater than 300 db , in order to improve over presently existing exclusion limits . to achieve these goals in practice , a box in a box
" concept was utilized for shielding the detection cavity , while a vector signal analyzer was used as a microwave receiver with a very narrow resolution bandwidth in the order of a few micro - hz .
this contribution will present the experimental layout and the results to date . |
in recent years much effort has been spent on the search for quantum two - level systems , qubits , that can be coherently controlled long enough for a sequence of controlled unitary operations to be performed on them .
the ultimate goal is to build a quantum computer out of these qubits .
proposals for qubits based on a variety of physical systems exist , each with its pros and cons . here
we consider solid state charge qubits based on josephson junctions ( jj ) arrays .
these have the advantage of being relatively easy to manipulate and the prospects for large scale manufacturing are comparatively good .
however , they suffer from severe decoherence effects .
the single cooper - pair box ( scb ) is the simplest proposal for a charge based jj qubit.@xcite for the scb , superposition of charge states was observed by bouchiat _
et al._@xcite and coherent evolution was demonstrated by nakamura _
et al_.@xcite a generalisation of the scb , which we call the circular array ( ca ) , was introduced by schllmann _
et al_.@xcite this circuit consists of an array of tunable jjs in a circular geometry .
the ca is similar to the scb and many results can be taken over _
mutatis mutandis_. the main difference is that the two charge states of the qubit are coupled through a higher order tunneling process . turning the coupling off
then allows the tunneling rate to be made very small leading to slow relaxation , and a long time to perform the measurement .
this is the key element of the quantum sample and hold ( quash ) measurement strategy.@xcite in this article we perform a more detailed study of the circular array . in particular
, we consider the effect of voltage fluctuations in the circuit ( ohmic noise ) as well as @xmath0 noise , believed to be caused by background charge fluctuations , and calculate the relaxation and dephasing times for these types of noise extending the previous treatment.@xcite we also study the measurement of the qubit s charge by a single - electron transistor ( set ) coupled capacitively to the ca .
this is done by numerically determining the time development of the density matrix following the treatment of makhlin _
et al._@xcite for the scb .
the circular array consists of two arrays with @xmath1 identical jjs each
these arrays are connected in series and separated by a capacitor @xmath2 , thus forming a circular geometry .
each jj , which is a small squid , has capacitance @xmath3 and a josephson energy @xmath4 which can be tuned by altering the magnetic flux @xmath5 through the squid loop .
the lead connecting the two arrays is grounded to allow charge to tunnel in and out of the circuit .
there are @xmath6 small islands , @xmath7 , each characterized by the number of excess cooper pairs @xmath8 and the phase of the superconducting order parameter @xmath9 ; these are quantum mechanically conjugate variables : @xmath10 = i \delta_{ij}$ ] .
each island charge is externally controlled by a gate voltage @xmath11 , applied via a small capacitor @xmath12 .
[ ca_circ ] shows the circular array together with the set that measures the charge on one of the islands next to @xmath13 . .
the box symbols ( without cross ) stand for normal junctions .
island @xmath1 on the ca and the island of the set is marked with dotted boxes.,scaledwidth=70.0% ] a qubit should have two states separated by a large gap , @xmath14 , from higher energy states and be weekly coupled to the environment to avoid rapid decoherence .
the ca fulfills this if @xmath15 . the energy scales present in the system is the charging energy for a cooper pair @xmath16 , the josephson energy @xmath17 , the superconducting gap @xmath18 and the temperature @xmath19 . in order to avoid quasiparticles in the system at low temperatures ,
the qubit is constructed so that @xmath18 is the largest energy in the problem .
furthermore , we choose the qubit to be in the charge regime @xmath20 , and impose @xmath21 to avoid thermal excitation of higher energy charge states , thus : @xmath22 the hamiltonian of the ca is @xmath23 where @xmath24 is the effective charge on island @xmath25 here @xmath26 is the gate charge on the island .
@xmath27 is the capacitance matrix
its nonzero elements are : @xmath28 , @xmath29 , @xmath30 and @xmath31 , where @xmath32 .
the matrix is symmetric and @xmath33 ; the inverse matrix @xmath34 has the same symmetries .
the sum over the josephson terms in ( [ h ] ) is taken over all pairs of islands connected by tunnel junctions . since we are studying a charge qubit it is convenient to write the hamiltonian in the charge basis @xmath35 where @xmath36 , and using that @xmath37 , which holds since @xmath8 and @xmath9 are conjugate , the josephson term becomes @xmath38 , where we have indicated the dependence on the external control parameters @xmath39 and @xmath5 . the two - level system _ ie _ the qubit we consider consists of states @xmath40 , @xmath41 with one excess cooper pair on either of the islands @xmath42 neighbouring @xmath2 : @xmath43 and @xmath44 , where the ones are for island @xmath1 and @xmath45 respectively . for @xmath46 these two states
are degenerate if @xmath47 and @xmath48 for @xmath49 .
if , in addition , @xmath50 , then the energy gap to higher charge states is @xmath51 . restricting ourselves to a finite charge space , the hamiltonian @xmath52 can be diagonalised numerically . in fig .
[ ca_energies ] we show the energy spectrum for the @xmath53 ca as a function of @xmath54 .
the other parameters are @xmath55 , @xmath56 , @xmath57 and @xmath58 ( this restriction gives a negligible error ) . for @xmath59 , the two lowest energy levels ( which are linear combinations of the states @xmath40 , @xmath60 with an excess cooper pair on island @xmath1 or @xmath45 respectively ) form a two - level system with a large gap to the higher energy states .
ca as functions of @xmath54 .
the other parameters are @xmath61 , @xmath62 and @xmath63 .
( charge states with @xmath64 are included in the diagonalization.),scaledwidth=47.0% ] we conclude that when @xmath65 , @xmath66 and @xmath67 , it is a good approximation to restrict the hilbert space of the hamiltonian in eq .
( [ h ] ) to the states @xmath40 and @xmath41 defined above .
we write the hamiltonian of this two - level system in spin-1/2 notation @xmath68 where @xmath69 are the pauli matrices in the basis @xmath70 , @xmath71 .
this hamiltonian controls the qubit
unitary operations can be performed on the qubit by tuning @xmath72 and @xmath73 via the external parameters @xmath74 and @xmath5 .
writing @xmath75 where @xmath76 , we find @xmath77 , where @xmath78 .
@xmath79 can be calculated numerically for given capacitance matrix , however , we can also perform an expansion in @xmath80 which is valid as long as @xmath1 is not too large . using cramer s rule for the elements in @xmath34
, we express @xmath79 in terms of cofactors and expand .
assuming , for simplicity , @xmath81 , this gives @xmath82 \times e_c \delta n_g \ \ \ .\ ] ] ( for @xmath53 and to leading order in @xmath12 , this reproduces a previous result.@xcite ) @xmath73 gives the cotunneling rate of a cooper pair from island @xmath1 to @xmath45 via the @xmath6 junctions and and its leading contribution is obtained by @xmath83 order perturbation theory , hence @xmath84 .
the exact numerical factor is not very illuminating it can however be determined for not too large @xmath1 . for the @xmath53 ca , with parameters as in fig .
[ ca_energies ] , we have @xmath85 e_c \delta n_g = 0.796 e_c \delta n_g$ ] and @xmath86 @xcite . ( a numerical diagonalisation gives @xmath87 and @xmath88 in reasonable agreement with the expansions taking into account that the ignored terms are of order @xmath89 and @xmath90 respectively . ) diagonalizing ( [ h2lvl ] ) gives the eigenvalues @xmath91 where @xmath92 , with corresponding eigenvectors @xmath93 and @xmath94 , where @xmath95 .
if @xmath96 denotes the pauli matrices in the energy eigenbasis , then @xmath97 can be written in the compact form @xmath98
the practical usefulness of a circuit like the ca as a qubit is ultimately limited by the coupling to external degrees of freedom .
these lead to decoherence of the qubit state and hence loss of quantum information to the environment .
the generic behaviour of the evolution of the qubit depends on the strength of the coupling and one identifies two regimes : the `` hamiltonian - dominated '' , where the coupling to the environment is weak enough for the time evolution of the qubit to be governed by the qubit hamiltonian @xmath97 , and the `` environment - dominated '' , where the coupling to the environment is so strong that it determines the dynamics of the qubit . in this article we consider only the hamiltonian - dominated regime .
note , however , that even if the coupling to the environment is weak under normal operation of the qubit , it becomes environment - dominated if @xmath99 , which may happen during the qubit manipulations . in the hamiltonian - dominated regime the evolution of the qubit
is conveniently described in the energy eigenbasis ( @xmath100 , @xmath101 ) .
the interaction with the environment leads to a decay of the off - diagonal elements in the qubit s density matrix with a charecteristic time @xmath102 , the dephasing time , @xmath103 whereas the diagonal elements of the density matrix decay to their thermal equilibrium values with a characteristic time @xmath104 , the relaxation time , @xmath105 where the thermal equilibrium value is @xmath106 .
a josephson junction charge qubit is sensitive to various electromagnetic fluctuations in the circuit ; we follow standard practice and model these with the `` spin - boson '' model @xcite with an ohmic spectrum .
in addition to the noise caused by these fluctuations one observes @xmath0 noise , which is believed to be due to background charge fluctuations in the substrate .
following shnirman _ et al_,@xcite we model this phenomenologically using again the spin - boson model but now with a @xmath0-spectrum .
the spin - boson model describing the qubit interacting with the environment has one independent bath for each island in the ca : @xmath107 here , @xmath97 is the qubit hamiltonian ( [ h2lvl ] ) , @xmath108 is a bath of harmonic oscillators with coordinates @xmath109 , momenta @xmath110 , masses @xmath111 and frequencies @xmath112 . the baths lead to voltage fluctuations @xmath113 coupling to @xmath114 .
( @xmath115 is the strength of the coupling between the qubit and the ath oscillator in bath @xmath25 . )
the effect of the environment is completely characterized by a spectral function , which for the spin - boson model in eq .
( [ hsb ] ) has the form @xmath116 to obtain the second equality one assumes that @xmath117 can be written as a power of @xmath118 up to some cut - off frequency @xmath119 which is assumed to be large compared to all other frequencies in the problem .
the parameter @xmath120 reflects the qualitative nature of the environment , and @xmath121 is a dimensionless measure of the strength of the coupling . to maintain @xmath122 dimensionless for all @xmath120 ,
an additional frequency scale @xmath123 enters for @xmath124 .
we model the voltage fluctuations @xmath125 on island @xmath25 by adding an impedance @xmath126 in series with @xmath127 , see , _ eg _ , ingold and nazarov@xcite or makhlin _
this impedance then has johnson - nyquist fluctuations @xmath125 between its terminals that in the spin - boson formalism correspond to the spectral function @xmath128 $ ] , where @xmath129 is the total impedance between the terminals of @xmath130 . from fig .
[ ca_circ ] one finds @xmath131^{-1}$ ] where @xmath132 , with @xmath133 ( assuming @xmath134 ) . following standard practice , we assume that the noise is purely resistive , @xmath135 , and if furthermore @xmath136 ( which holds for realistic @xmath137 and @xmath12 since @xmath138 ) , we obtain @xmath139 , which is linear in @xmath118 and hence corresponds to @xmath140 in eq .
( [ jw ] ) . using this
we can obtain the total spectral function @xmath117 for the circuit .
from eq .
( [ hsb ] ) we identify @xmath141 , where @xmath142 \big|_{n^g_i=\frac{c_g}{2e}\delta v_i}$ ] . simplifying this expression
yields @xmath143 , with @xmath144 .
the fluctuations @xmath125 are expressed in the oscillator coordinates , @xmath145 , with the spectral function in the form of eq .
( [ jw ] ) , @xmath146 .
this gives @xmath147 and hence , @xmath148 ^ 2 r_i$ ] . comparing this to eq .
( [ jw ] ) , with @xmath140 , we find @xmath149 where @xmath150 is the quantum of resistance . assuming the islands to be nearly identical , it is reasonable that @xmath137 are approximately the same for all islands . setting @xmath151 , @xmath152 ,
we are left with the factor @xmath153 , which we can calculate numerically for given capacitance matrix or as an expansion in @xmath80 .
the expansion is analogous to the one for @xmath154 in eq .
( [ bz ] ) assuming for simplicity @xmath155 we obtain @xmath156 \times \frac{r}{r_k } \bigg ( \frac{c_g}{c_0 } \bigg ) ^2.\ ] ] ( for @xmath53 and to leading order in @xmath12 , eq . ( [ av ] ) gives @xmath157 as derived previous.@xcite ) fig .
[ an ] shows @xmath158 as a function of @xmath1 for different @xmath80 .
the numerical result is shown as crosses , and the expansion to order @xmath159 as squares . as a function of @xmath1 for different @xmath80 .
the numerical result is shown as crosses and the analytic expansion as squares .
@xmath160 is given in units of @xmath161 and the scale on the @xmath162-axis is logarithmic ; @xmath151 for all islands , and @xmath81.,scaledwidth=50.0% ] for @xmath163 and @xmath164 , the expansion agrees very well with the numerical result ( at least when @xmath165 ) . for @xmath166 and @xmath167 ,
the expansion becomes negative at @xmath168 and @xmath169 respectively .
therefore no squares are seen in this region . in a typical circuit
we may have @xmath170 , yielding @xmath171 .
for @xmath172 , @xmath173 for all realistic @xmath1 .
the characteristic times @xmath104 and @xmath102 can be calculated within the spin - boson model using perturbation theory or path integral methods.@xcite for @xmath140 they are @xmath174 where @xmath175_+\rangle_\omega = 2 \hbar j(\omega ) \coth \frac{\hbar \omega}{2 k_b t}\ ] ] is the fourier transform of the symmetrized correlation function at thermal equilibrium .
these results can be applied to the ca simply by substituting @xmath158 from eq .
( [ av_exact ] ) .
one defines the pure dephasing rate @xmath176 , as @xmath177 .
the hamiltonian - dominated regime is realized when @xmath178 ( at least for @xmath179 @xcite ) .
for ohmic damping , @xmath180 , and the condition becomes @xmath181 .
noise with a power spectrum proportional to the inverse of the frequency is observed in many physical systems .
this @xmath0 noise is seen also in jj circuits @xcite , where it is believed to be caused by background charge fluctuations.@xcite here we follow shnirman
_ et al_,@xcite and model this noise with the spin - boson model . for @xmath182 and @xmath183 , ( [ sw ] ) gives the wanted power spectrum : @xmath184 with @xmath185 ( here @xmath186 is an adjustable parameter ) .
sub - ohmic environments ( _ ie _ those for which @xmath187 ) have not been much studied as it was believed that they rapidly localized the system in one of the @xmath114-eigenstates for any strength of the damping .
however , it has been realized that this is true only for large damping , whereas for weak damping the system behaves coherently@xcite . the times
@xmath104 and @xmath102 have been calculated for @xmath0 noise for the single bath spin - boson model in the hamiltonian - dominated regime@xcite .
the relaxation rate is @xmath188 the dephasing rates are only known for @xmath189 ; they are @xmath190 and @xmath191 both with logarithmic accuracy in @xmath192 @xcite . here
, @xmath193 is an infrared cut - off frequency which can be experimentally determined . to determine the times we need to determine @xmath192 . ) with @xmath184 , whereas eq .
( [ t_dep_h ] ) can not be used for the dephasing rates since @xmath194 for @xmath182 . ]
the state of the qubit is inferred by performing a quantum measurement of the charge using a set coupled capacitively to island @xmath1 of the circular array , see fig .
[ ca_circ ] .
we simulate this quantum measurement by studying the time development of the density matrix describing the ca and the set we follow closely makhlin _
et al_,@xcite where the corresponding problem is treated for the scb .
the set is a circuit with one normal island surrounded by two junctions connected to normal electrodes and a capacitor @xmath195 , see fig .
[ ca_circ ] . during manipulations of the qubit ,
no current flows through the set the set is turned off this is achieved by setting the transport voltage to zero , @xmath196 , and tuning the gate voltage @xmath197 away from a degeneracy point so that the coulomb blockade suppresses the tunneling through the set . when the manipulations of the qubit are done and one wants to read out the result , then the coulomb blockade is turned off by tuning the gate voltage to a degeneracy point and a transport voltage is turned on leading to a tunneling current through the set .
this current depends on the state of the qubit through the charge on the capacitor @xmath198 .
this leads to a measurement of the charge of the qubit .
the density matrix for the ca and the set can be written as @xmath199 after one has traced out the microscopic degrees of freedom in the left and right electrodes and in the island of the set . here
, @xmath25 labels the state of the qubit , @xmath200 is the number of ( excess ) electrons on the island in the set and @xmath201 is the number of electrons that have tunneled through the set .
it is possible to derive a master equation for @xmath202 as an expansion in the set tunneling terms.@xcite for low temperature and small @xmath203 , only transitions between two adjacent charge states of the set need to be taken into account ( these states are assumed to be @xmath204 below ) .
if one furthermore assumes that the tunneling is instantaneous then one obtains a set of simple equations for the diagonal matrix elements @xmath205 . in terms of the fourier transformed quantity @xmath206
, the final form is a system of eight coupled differential equations ( for each @xmath207 ) @xmath208 \choose i[h_{ctrl } + \delta h_{int},\hat{\rho}_1 ] } = { -\check{\gamma}_l \quad e^{-ik}\check{\gamma}_r \choose \check{\gamma}_l \quad -\check{\gamma}_r } { \hat{\rho}_0 \choose \hat{\rho}_1 } \ , \ ] ] where @xmath209 is short for the @xmath210 matrix @xmath211
. @xmath97 is the qubit hamiltonian with a renormalised capacitance matrix ( @xmath212 ) due to the presence of the set , @xmath213 is the coupling energy where @xmath214 is determined by the capacitances .
the tunneling rates @xmath215 are @xmath216_+ \ , \nonumber \\
\check{\gamma}_r \hat{\rho}_1 \equiv \gamma_r \hat{\rho}_1 - \pi \alpha_r [ \delta h_{int},\hat{\rho}_1]_+ \ , \end{aligned}\ ] ] where @xmath217 and @xmath215 is the tunneling conductance and the tunneling rate for the left / right junction of the set .
we want to study the current through the set as a function of time this is obtained from @xmath218 which can be interpreted as the probability that @xmath201 electrons have tunneled through the set during time @xmath219 .
to compute @xmath220 , we solve the differential equation ( [ me ] ) with suitable initial conditions .
we assume that the qubit and the set are disentangled initially , @xmath221 , and that the qubit is prepared in some state @xmath222 , @xmath223 . at time
@xmath224 , no electrons have tunneled through the set , thus @xmath225 and @xmath226 for the fourier transform . from the definition ( [ pmt ] ) , we find @xmath227 and we may choose @xmath228 . to calculate @xmath220 for the @xmath53 ca we need values for the parameters @xmath214 , @xmath229 , @xmath230 , @xmath231 and @xmath232 for the set , and @xmath72 and @xmath73 for the qubit .
we choose @xmath233 ff , @xmath234 , @xmath235ev and @xmath236ev ( the qualitative result will not be too sensitive to this choice ) .
hence @xmath237ev , @xmath238ev and @xmath239ev .
at the start of the measurement @xmath240 is turned off , @xmath241 .
@xmath72 is kept close to the degeneracy point , although @xmath242 to avoid the environment - dominated regime .
[ meas_p_1 ] shows examples of @xmath220 . in
( a ) a measurement in the off - state , @xmath243 , is shown , the other parameters are @xmath244ev ( corresponding to @xmath245 ) and @xmath246 . for comparison , fig .
[ meas_p_1](b ) shows a mesaurement where @xmath240 is on ; the parameters are in this case @xmath247ev ( which is the maximum value of @xmath73 assuming @xmath248 ) , @xmath249ev ( corresponding to @xmath250 ) and @xmath251 . as a function of @xmath201 for ten different times @xmath219 . in ( a ) , @xmath252 , @xmath244ev and @xmath246 , whereas in ( b ) , @xmath247ev , @xmath249ev and @xmath251.,title="fig:",scaledwidth=49.0% ] as a function of @xmath201 for ten different times @xmath219 . in ( a ) ,
@xmath252 , @xmath244ev and @xmath246 , whereas in ( b ) , @xmath247ev , @xmath249ev and @xmath251.,title="fig:",scaledwidth=49.0% ] we see that after a while , @xmath253 develops a two peak structure .
this is interpreted as follows .
at a given time , the total current is a superposition of two different currents flowing in the set , with weights given by the amplitudes , @xmath254 and @xmath255 , of the charge states in the qubit state . for increasing @xmath219 ,
the peaks move towards higher @xmath201 values since more electrons have tunneled through the set . during this process ,
the peaks widen , their magnitude decreases and the distance ( in @xmath201 ) between the peaks increases .
the ratio between the two peak values are @xmath256 in fig .
[ meas_p_1](a ) and @xmath257 ( the value decreases with time ) in fig .
[ meas_p_1](b ) to be compared to the ratios @xmath258 and @xmath167 respectively .
[ meas_p_1](a ) corresponds to our proposed measurement situation the off - state and we see that here the peaks are well separated after a while and the ratio of their peak values stay close to @xmath259 for a very long time . thus allowing for a slow accurate measurement of the charge . in fig .
[ meas_p_1](b ) on the other hand , the peaks are less well separated and the ratio agrees less well with @xmath259 . eventually the valley between the peaks will fill out and one broad peak forms due to mixing between the charge states a plateau indicating this process is clearly seen in ( b ) . to perform a measurement of the charge , the peaks must have moved apart to become well separated and not yet started to merge due to mixing .
this gives a window in time for measuring the charge in the off - state in fig .
[ meas_p_1](a ) this window is very large the lower bound is @xmath260 ns . the upper bound is not seen , however , we expect it to be considerably larger .
we here estimate the decoherence times for the circular array . to be specific we assume @xmath261 , @xmath262ev and @xmath263 mk @xmath264ev and , initially @xmath53 higher @xmath1 are commented on below .
calculations are performed by tuning @xmath265 in time . during this process
the set is off , and we are close to the degeneracy point . assuming @xmath266 , then @xmath267ev in addition we have @xmath268ev .
there is also a lower bound on @xmath269 , since the system is assumed to be in the hamiltonian - dominated regime .
the typical time per operation of the qubit is @xmath270 ps assuming @xmath271ev . for @xmath0 noise
, @xmath278 is only known for @xmath279 and @xmath280 .
we use these two cases to estimate @xmath102 , assuming that this gives the correct order of magnitude for general @xmath281 .
@xcite measured the factor @xmath282 for the scb and obtained @xmath283 .
if the @xmath0-noise is caused by background charge fluctuations then it is reasonable to assume that for the ca @xmath284 , since the ca has @xmath6 islands instead of the single island in the scb .
this gives @xmath285 and , for the parameters above , we find @xmath286 . from nakamura _ et al . _
@xcite we also take @xmath287 hz . for @xmath288 we find @xmath289 and @xmath290 ns if @xmath291ev and for @xmath280 , we find @xmath292 ns and @xmath293 ns if @xmath294ev .
this shows that the hamiltonian - dominated regime is realised if @xmath295ev .
after the calculations , the set is turned on and the measurement is started by tuning @xmath73 to @xmath296 .
the relaxation is now very slow : @xmath297 when @xmath298 for both ohmic and @xmath0 noise .
this gives ample time to measure @xmath299 without using an ultra - fast detector , _
[ meas_p_1 ] .
the dephasing is very rapid , thus the quantum state of the qubit is destroyed in a short time but it is still possible to measure the charge .
this is good enough as a read - out mesurement but not as part of an error correction protocol .
we conclude that it is the @xmath0 noise that limits the operation of the ca as a qubit it leads to the decoherence time @xmath301 ns .
the decoherence due to ohmic noise is much slower .
hence , in practice , the ohmic noise seems to be of little concern .
if the typical time for a quantum operation is @xmath302 ps , then the @xmath0 noise restricts one to @xmath303 operations , assuming @xmath73 and @xmath72 are restricted to values that realise the hamiltonian - dominated regime .
this is a severe restriction and once again underscores that @xmath0 noise is a serious problem for josephson junction charge qubits .
one possible solution to this might be the echo - technique demonstrated by nakamura _ et al .
_ @xcite .
we wish to thank peter gren , jochen walter , silvia corlevi , janik kailasvuori and hans hansson for useful discussions and sergei isakov for his computer skills .
anders karlhede was supported by the swedish science research council .
+ a. shnirman , g. schn and z. hernon , _ phys .
_ * 79 * , 2371 ( 1997 ) .
y. nakamura , y. a. pashkin and j. s. tsai , nature * 398 * , 786 ( 1999 ) .
v. bouchiat , d. vion , p. joyez , d. esteve and m. h. devoret , _ physica scripta _
vol t76 , 165 - 170 ( 1998 ) .
v. schllmann , p. gren , d. b. haviland , t. h. hansson and a. karlhede , _ phys . rev .
b _ * 65 * , 020505(r ) ( 2002 ) . y. makhlin , g. schn and a. shnirman , _ nature _ * 386 * , 305 ( 1999 ) .
y. makhlin , g. schn and a. shnirman , _ rev .
_ , * 73 * , 357 ( 2001 )
. a. o. caldeira and a. j. leggett , _ ann .
_ * 149 * , 374 ( 1983 ) .
a. shnirman , y. makhlin and g. schn , cond - mat/0202518 ( 2002 ) .
ingold and y. nazarov , _ single charge tunneling _ , nato asi series , vol .
b 294 , edited by h. grabert and m. h. devoret ( plenum press , new york , 1992 ) .
a. j. leggett _
_ , _ rev .
* 59 * , 1 ( 1987 )
. y. nakamura , yu .
a. pashkin , t. yamamoto and j. s. tsai , _ phys .
lett . _ * 88 * , 047901 ( 2002 ) .
u. weiss , _ quantum dissipative systems _ , world scientific , singapore ( 1992 ) .
a. b. zorin _
_ , _ phys .
b _ * 53 * , 13682 ( 1996 ) .
s. kehrein and a. mielke , _ phys .
a _ * 219 * , 313 ( 1996 ) . h. schoeller and g. schn , _ phys .
b _ * 50 * , 18436 ( 1994 ) . | we discuss a charge qubit consisting of a circular array of josephson junctions .
the two - level system we consider couples the two charge states through a higher order tunneling process thus making it possible to achieve a long relaxation time . using the spin - boson hamiltonian , we estimate decoherence due to ohmic as well as @xmath0 noise . we simulate the quantum mechanical measurement process by studing
the density matrix of the qubit and a capacitively coupled single - electron transistor that measures the charge . 0.1pc |
in anisotropic superconductors , local rotations of the lattice modulate the order parameter .
the most striking manifestation of this effect is the influence of grain boundaries in high - t@xmath2 superconductors@xcite .
less pronounced effects are also to be expected from lattice defects such as dislocations , as they twist the lattice axes in their vicinity .
transversely polarized phonons can , in principle , also couple to the superconducting order parameter through the local rotations that they induce .
if , along a given direction , there are strains which change the orientation of the lattice axes , the equilibrium order parameter must follow that distortion .
for instance , in a two dimensional square lattice , a parameter with @xmath3 symmetry , @xmath4 , must change into @xmath5 after a lattice rotation by an angle @xmath6 .
we express the local order parameter in terms of its components in a fixed frame of reference which is independent of the orientation of the lattice .
we expand the local order parameter in terms of the functions @xmath7 , defined in this external frame . in principle
, the symmetry of the square lattice allows for mixing of a pure @xmath8 function with higher spherical harmonics . under a rotation ,
each of these harmonics behaves differently , and requires a specific description in the external frame of reference mentioned earlier . for simplicity
, we will first ignore these higher order components .
their inclusion will be discussed later , as well as extensions to more complex order parameters .
we will now set up an effective ginzburg landau description of the situation described above .
let us first consider a model superconductor with very small coherence length . at each position
, the order parameter will match perfectly the orientation of the lattice axes .
there is no cost in free energy due to misorientations in the order parameter .
we can ignore gradient terms , and use an expression which contains only quartic and quadratic terms , provided that we insert in them the value of order parameter with respect to the local axes .
the coupling between neighboring regions is described by gradient terms .
there is a free energy cost associated to the inhomegeneities of the order parameter . in the present case
, the degree of inhomegeneity must be referred to the fixed , external frame of reference .
there is an increase in the free energy of the system when different areas of the lattice are misoriented , even if the local order parameter is aligned with the respective axes .
a particular example of this situation is a grain boundary , as described in the introduction .
the difference between the two ways of describing the order parameter , with respect to an external frame or to the local one , is determined by the degree of rotation of the lattice from one region to another .
let us define the rotation which takes the axes from one point to another by the angle @xmath6 . as mentioned previously ,
an order parameter which follows the lattice will look , in the external frame , like @xmath4 in one point and @xmath9 in the other .
the transformation which takes one expression into the other can be accomplished by modifying the two component vector @xmath7 by means of an operator of the type @xmath10 , such that : @xmath11 and : @xmath12 the integral is to be taken along the path in real space along which the lattice is rotated .
as @xmath6 is the total rotation along the path considered , we can write , for the example considered here : @xmath13 where @xmath14 is the gradient operator in real space , @xmath15 specifies the local rotation of the lattice axes , and @xmath16 is a pauli matrix .
the rotation @xmath15 can be written as@xcite : @xmath17 where @xmath18 are the local deviations of the lattice node at @xmath19 from equilibrium .
( [ rotation ] ) is valid for small deviations ( strains ) .
the expressions above allow us to relate the changes in the order parameter as seen in the two frames , and the rotations of the lattice , in terms of the local strains . the gradient term in the ginzburg landau description , which should be expressed with respect to the external frame , looks , in terms of the order parameter with respect to the local axes : @xmath20 where @xmath21 is the free energy .
@xmath22 is a number with dimensions of temperature , and of order @xmath23 .
finally , @xmath24 are proportionality constants which play no role in the analysis presented below .
we take the critical temperatures of the @xmath25 order parameters as different , as appropiate for a square lattice .
equation ( [ gl ] ) and the definitions ( [ gauge ] ) and ( [ rotation ] ) suffice to study the phenomenology of layered superconductors whose order parameter is well approximated by @xmath26 $ ] .
it is easy to see that , when @xmath4 and @xmath27 have the same critical temperature , and @xmath28 , the effects of the field can be made to vanish by performing a local rotation of the order parameter@xcite . in this case , the system has isotropic superconducting properties .
hence , rotations of the underlying crystal lattice leave the ( degenerate ) order parameter unaffected .
the formalism described here can be expressed in terms of a connexion , which depends on the lattice orientation .
the definition of parallel transport of a vector , in our case @xmath29 , needs to be modified by the rotation of the lattice .
the usual derivative is changed into a covariant derivative@xcite .
this technique lies at the basis of the extensive work done in topological phases@xcite , and has found use in many fields in condensed matter physics@xcite .
in addition to the effects described by ( [ gl ] ) , there may be an explicit coupling of the order parameter to the lattice strains .
the simplest coupling is proportional to @xmath30 , where @xmath31 defines the local compression , or expansion , in terms of the symmetric part of the strain tensor .
this term can also exist in a @xmath32 wave superonductor .
a coupling of this type is not negligible in the high - t@xmath2 compounds , due to the strong dependence of t@xmath2 on pressure@xcite .
the previous analysis can be generalized to three dimensional superconductors with arbitrary order parameters .
local rotations of the lattice can be expressed in terms of the parameter : @xmath33 where @xmath34 is the fully antisymmetric tensor . in order to determine the transformations induced by these rotations in the order parameter
, we must first decompose it into spherical harmonics with well defined transformation properties .
then , a gauge field which performs rotations in order parameter space is added to each component . finally , as in ( [ gl ] ) , all terms allowed by the lattice symmetry are included in the free energy .
the procedure is cumbersome , but straightforward .
for instance , we can add to ( [ gl ] ) higher harmonics which transform like @xmath35 , \sin [ ( 4 n + 2 ) \theta ] $ ] .
the gauge field associated to these components is of the type @xmath36 , where @xmath37 transforms one of these parameters into the other .
the gauge field induced by the higher harmonics of the order parameter are also proportional to the gradient of the lattice rotations , but with a larger constant of proportionality .
the term with @xmath38 in the free energy induces a reduction of these components of the order parameter .
this reduction will be stronger for higher harmonics than for lower ones .
we now consider specific lattice distortions within the layers .
the simplest static defect is a dislocation .
the strains in the vicinity of a dislocation can be calculated from the theory of elasticity . if the burgers vector is @xmath39 , then@xcite , @xmath40 where we are neglecting a possible anisotropy in the elastic constants of the lattice .
we now assume that the @xmath27 component is negligible , which is justified if , for instance , the parameter @xmath41 in ( [ gauge ] ) is less than zero .
then , the coupling of the lattice rotations to @xmath4 is due to terms quadratic in the gauge potential , and we obtain : @xmath42 and latexmath:[$| { \bf \vec{a } } |^2 = \frac { term is to reduce the effective critical temperature near the dislocation .
it suppresses a possible local enhancement of t@xmath2 due to the explicit dependence of the critical temperature on pressure@xcite . in the presence of a magnetic field
, the normal cores of vortices will be attracted towards the dislocations , to take adventage of the supression of t@xmath2 , leading to pinning .
near the core of the dislocation , @xmath44 , we can estimate the reduction of t@xmath2 ( see below ) as @xmath45 .
this effect can be significant in a material with a short coherence length , such as the copper oxides .
a grain boundary can be modelled as an array of dislocations .
the gauge fields induced by each of them can be added , to obtain the effect of the whole boundary . after some algebra , we find that , if the rotation of the axes across the boundary is not large : @xmath46 ^ 2 + \sin^2 \left ( \frac{\pi y}{\bar{a } } \right)}{\left [ \cosh \left ( \frac{\pi x}{\bar{a } } \right ) - \cos \left ( \frac{\pi y}{\bar{a } } \right ) \right]^4 } \nonumber \\ & \sim & \frac{\pi^2 \sin^4 ( \theta ) } { 4 a^2 } e^{- \frac { \pi x}{\bar{a } } } , \ , \ , \ , \ , \ , x \rightarrow \infty \label{grainb}\end{aligned}\ ] ] where @xmath47 is the lattice constant , @xmath6 is the angle between the axes at both sides of the boundary , and @xmath48 . from ( [ grainb ] ) and ( [ gl ] ) , we can estimate the reduction of the critical temperature at the boundary .
we assume that @xmath24 , and @xmath49 .
then , @xmath50 , where @xmath51 .
this calculation is rigorously valid when @xmath52 , that is , when strains change slowly .
a significant reduction of t@xmath2 can be expected near a twin boundary , which will act in a way similar to a weak link .
it would be interesting to check this experimentally@xcite .
the orientation of the lattice can also be modified dynamically , by transversely polarized phonons .
a longitudinal phonon induces local compressions and dilatations of the lattice .
a transverse phonon does not change the volume of the unit cell , but induces a shear deformation which changes from position to position .
this deformation can be viewed as a local rotation of the axes . in a material with a complex unit cell , there are also optical phonons which rotate locally the axes
. the rotations can be analyzed by means of the formalism discussed before .
the only difference with a static distortion is that the gauge field which accounts for the lattice rotations depends on dynamic displacements .
the time scales associated to the superconducting response are of the order of the inverse plasma frequency .
this frequency is much higher than typical phonon frequencies , so that the changes in t@xmath53 induced by phonons can be studied as if the lattice distortions were static .
transversely polarized acoustic phonons give rise to strains proportional to the gradient of their amplitudes .
hence , we expect that the fourier components of @xmath15 are given by : @xmath54 where @xmath55 is the amplitude of a phonon of momentum @xmath56 , and @xmath57 is the component of @xmath56 perpendicular to @xmath55 . for transversely polarized phonons , @xmath56 is always perpendicular to @xmath55 , and we do not need to take this restriction into account .
the gauge potential in ( [ gauge ] ) has a finite average , related to the thermal ( or quantum ) average value of the phonon amplitudes .
it is given by : @xmath58 we are analyzing materials which show strong differences in the superconducting properties along the in plane and out of plane directions .
phonons , however , need not be so anisotropic .
we assume that they are described by an isotropic sound velocity , @xmath59 , and a debye frequency @xmath60 , where @xmath47 is the lattice constant , which we take to be of the same order of magnitude in all directions . the three dimensional integral implicit in ( [ gaugeph ] )
is extended to those modes such that @xmath61 and @xmath62 and @xmath63 .
note that the ginzburg - landau description used here is valid at length scales greater than @xmath64@xcite . for simplicity , we take the bose occupancy of a mode of frequency @xmath65 to be @xmath66
. none of these approximations will alter the order of magnitude of the effect .
we can distinguish three regimes : i ) if @xmath67 , the upper cutoff in ( [ gaugeph ] ) is due to the bose factor .
ii ) when @xmath68 , the upper cutoff is @xmath69 and @xmath70 and iii ) if @xmath71 , the maximum value of @xmath72 is @xmath73 .
we can now estimate the reduction in the ( average ) value of the critical temperature in the same way as in the previous section .
we find : @xmath74 where @xmath75 is the density .
this reduction of the critical temperature should not be negligible in the high - t@xmath2 compounds , as @xmath64 is not very different from @xmath47 , the lattice constant .
in addition to a reduction in the average value of t@xmath2 , we can also analyze the fluctuations in this value .
we are treating phonons as a source of static disorder , which reduces t@xmath2 .
the value of @xmath76 fluctuates , changing the local values of the critical temperature .
if these fluctuations take place on scales larger than @xmath64 , a distribution of t@xmath2 s will be observed . as the changes in t@xmath2 go as : @xmath77 we find that : @xmath78 and : @xmath79 by comparing ( [ tcphonons ] ) and this expression
, one can see that fluctuations can not be neglected . finally , the typical length scale for fluctuations inside the planes is @xmath80 where @xmath81 is the electronic bandwidth .
hence , we expect large fluctuations in small scale domains .
transport measurements will be unable to resolve this structure , and the average @xmath82 will be observed .
transverse optical phonons of frequency @xmath83 .
give an additional reduction of t@xmath2 , @xmath84 where @xmath85 denotes the mean square displacement of the phonon .
it is , typically , a fraction , @xmath86 of @xmath87 . as in the previous case , we assume that the superconducting order parameter has a short coherence length in the out of plane direction . there is a contribution like ( [ phononopt ] ) for each optical phonon transversely polarized in the plane .
the frustration of the order parameter in anisotropic superconductors by dynamical vibrations implies that @xmath88 can not be much larger than the frequency of the phonons responsible of the effect .
the ginzburg landau expression ( [ gl ] ) contains terms which are linear in @xmath76 , which couple one component of the order parameter and the spatial derivatives of the other ( note that @xmath76 is an operator which rotates one component into the other ) .
these derivatives are different from zero in the presence of a supercurrent , because of the variation in the phase . in the case described by ( [ gl ] ) , however , these terms do not contribute , as @xmath89 .
terms of this type play a role in materials where the superconducting properties differ significantly along the lattice axes .
let us consider an orthorhombic superconductor , where the order parameter is @xmath1 .
in addition to the terms in ( [ gl ] ) we also have : @xmath90 where @xmath91 is the @xmath32 component of the order parameter . for the sake of simplicity , we ignore terms induced by the orthorhombic symmetry which are quadratic in the gradients@xcite .
these terms , needed for a complete description of the anisotropy in the penetration depth , do not modify qualitatively the picture discussed in the previous sections . in the absence of internal lattice rotations , @xmath92 .
the first term in ( [ ortho ] ) leads to a difference in the penetration depth along the two lattice axes .
local rotations of the lattice can be incorporated by adding the gauge field ( [ gauge ] ) to the derivatives of the @xmath0 components of the order parameter in ( [ ortho ] ) .
then , assuming that the equilibrium order parameter has both @xmath4 and @xmath91 components , we obtain terms of the type : @xmath93 this expression gives a contribution if @xmath94 , and , in particular , in the presence of a current .
the changes in a current carrying state induced by lattice rotations are best studied within london s framework , which is valid if the magnitude of the order parameter does not change .
we define the london tensor , with the same symmetry properties of the perfect lattice , such that @xmath95 , where @xmath96 is the electromagnetic vector potential . for the study of currents within the layers
, we have : @xmath97 where @xmath98 are the penetration depths along the two axes , and @xmath59 is the velocity of light .
london s expression for the free energy is@xcite : @xmath99 the deviation of tetragonal symmetry is given by @xmath100 .
taking into account that the gauge field related to the lattice distortions , @xmath101 , is itself a derivative , we can integrate by parts in eq.([orthoder ] ) , to obtain : @xmath102 this equation can be interpreted as a rotation of the london tensor , ( [ londont ] ) , induced by the distortion of the lattice .
the currents adjust to the directions of the local principal axes . at a grain boundary ,
for instance , @xmath96 must be continuous .
the component of the current parallel to the boundary must be discontinous , leading to a kink in the direction of the total current , as the london tensor which gives the current at both sides is discontinuous@xcite .
if , on the other hand , the boundary conditions prevent the current from acquiring a transverse component , the vector potential shows a kink , and a perpendicular magnetic field is generated at the boundary , as sketched in fig([twin]@xcite .
this picture , derived from london s theory , is valid at distances comparable to the penetration depth .
the discontinuity at the boundary is rounded off at shorter scales .
the present analysis implies that currents can also be deflected by other types of static defects , and by phonons .
the currents flowing near a dislocation , and the instantaneous current in the presence of phonons , are sketched in fig.([currents ] ) .
magnetic fields will be induced .
it is interesting to note that phonons can modulate the currents and give rise to electromagnetic fields .
so far , we have only considered the role of in plane distortions in layered superconductors .
if we include rotations around in plane axes , we must allow for components like @xmath103 in the order parameter . because of the strong anisotropy of these compounds , components which transform like higher powers of @xmath104 are also to be expected . in analogy with ( [ london ] ) , these components lead to contributions of the type @xmath105 , where the direction @xmath106 lies in the planes .
note that this effect should also be present in layered superconductors with @xmath32 symmetry within the layers .
the magnitude of the @xmath107 coupling is proportional to @xmath108 . a particular defect which generates rotations in the out of the plane direction is a screw dislocation , which seems to be common in sputtered films@xcite . in cylindrical coordinates
, we have that only @xmath109 is different from zero@xcite .
hence , the free energy of the superconductor has a term like @xmath110 .
a current flowing around the dislocation induces a current parallel to the axis of the dislocation .
this situation may arise if a vortex is pinned by a screw dislocation .
a similar effect can be realized in a monocrystalline whisker subject to shear strains .
then , @xmath111 , leading again to a @xmath110 coupling .
a current flowing along the axis of the whisker induces a tangential component , giving rise to a magnetic field in its interior .
a schematic picture of both situations is sketched in fig.([dislocation ] ) .
we have presented a formalism to study the influence of lattice rotations on anisotropic superconductors .
the analysis has been particularized to d - wave layered systems , but it is general enough to be applicable to more complex lattices and order parameters .
we expect the scheme outlined here to be also useful for the study of heavy fermions and other materials which show an anisotropic order parameter . for the systems considered here , we find that \i ) lattice distortions couple directly to an anisotropic order parameter , irrespective of other mechanisms , like the dependence of t@xmath2 on pressure .
\ii ) higher harmonics of the order parameter are more strongly frustrated by local rotations@xcite .
\iii ) lattice defects , such as dislocations , reduce locally t@xmath53 . in the presence of a magnetic field ,
the vortex cores will be attracted to these regions , giving rise to pinning .
\iv ) transversely polarized phonons also reduce t@xmath2 .
the value of the critical temperature can not be much larger than the typical frequency of these phonons .
\v ) local rotations deflect supercurrents in superconductors with inequivalent lattice axes .
in particular , the current through a grain boundary is deflected .
this work was done with support from cicyt ( spain ) through grant pb96 - 0875 .
valuable discussions with k. maki , and hospitality at the istituto eduardo caianello ( vietri , italy ) , are acknowledged .
c. c. tsuei , j. r. kirtley , c. c. chi , l. s. yu - jahnes , a. gupta , t. s. shaw , j. z. sun and m. b. ketchen , phys .
lett . * 73 * , 595 ( 1994 ) .
j. r. kirtley , c. c. tsuei , j. z. sun , c. c. hi , l. s. yu - jahnes , a. gupta , m. rupp and m. b. ketchen , nature ( london ) * 373 * , 225 ( 1995 ) .
m. sigrist and t. m. rice , rev .
phys . * 67 * , 503 ( 1995 ) .
the curl of @xmath112 is non zero only in the presence of a disclination .
such a defect induces a finite curvature in the two dimensional layer considered so far .
its description requires the inclusion of other harmonics , related to out of plane rotations of the @xmath4 order parameter .
m. nakahara , * geometry , topology and physics * , iop publishing , london , ( 1990 ) .
a. shapere and f. wilczek , * geometric phases in physics * , world scientific , singapore ( 1988 ) .
a. m. tsvelik , * quantum field theory in condensed matter physics * , cambridge u. p. , cambridge ( 1995 ) .
l. d. landau and e. f. lifschitz , * theory of elasticity * pergamon press , london ( 1959 ) . f. guinea , europhys .
* 7 * , 549 ( 1988 ) .
a. b. kuklov , phys .
b * 52 * , r7002 ( 1995 ) .
the present scheme can not be used to analyze the pair breaking due of short wavelength phonons .
it will be studied elsewhere .
k. a. kuznetsov , a. g. sun , b. chen , a. s. katz , s. r. bahcall , john clarke , r. c. dynes , d. a. gajewsky , s. h. han , m. b. maple , j. giapintzakis , j .-
kim and d. m. ginsberg , phys .
79 * , 3050 ( 1997 ) .
p. g. de gennes , * superconductivity of metals and alloys * , addison - wesley , new york ( 1966 ) . f. guinea , phys . rev .
b * 40 * , 9362 ( 1989 ) .
f. guinea , phys .
b * 41 * , 4733 ( 1990 ) .
in addition to the deflection discussed here , currents can be induced at a boundary , when the quartic terms in the ginzburg - landau expression are taken into account .
see d. l. feder , a. beardsall , a. j. berlinsky and c. kallin , preprint ( cond - mat/9705139 ) .
these terms are discussed in detail in@xcite .
currents flowing through a twin with a similar geometry have been studied by j. r. kirtley , k. a. moler and d. j. scalapino , phys .
b * 54 * , 886 ( 1997 ) and d. f. agterberg and m. sigrist , preprint ( cond - mat/9711157 )
. g. gerber , d. anselmetti , j. g. bednorz , j. mannhart and d. g. schlom , nature ( london ) * 350 * , 279 ( 1991 ) .
this effect may lead to the suppression of the kohn luttinger effect , which shows that an electron liquid is always unstable towards the formation of cooper pairs of high angular momenta .
local rotations of the lattice will greatly reduce the ( already small ) estimated value of the critical temprature .
w. kohn and j. m. luttinger , phys .
* 15 * , 524 ( 1965 ) . | lattice distortions in which the axes are locally rotated provide an intrinsic source of frustration in anisotropic superconductors . a general framework to study this effect
is presented .
the influence of lattice defects and phonons in @xmath0 and @xmath1 layered superconductors is studied . |
geometrically frustrated interactions and quantum fluctuation inhibit the stabilization of classical orderings . they sometimes become a trigger for the emergence of several exotic orders such as , quantum spin ice on the pyrochlore lattice@xcite , spin liquid state on the honeycomb lattice@xcite , and multiple magnetization plateaus@xcite . therefore ,
quantum spin systems with frustrated interactions have attracted great interest from both theoretical and experimental approaches . the @xmath1 antiferromagnetic heisenberg spin model on the shastry - sutherland lattice ( ssl)@xcite is one of such systems . the hamiltonian can be expressed by the nearest neighbor ( intradimer ) @xmath3 and the next nearest neighbor ( interdimer ) @xmath4 couplings : @xmath5
experimentally , @xmath6@xcite has been studied extensively for its realization of the model . in this compound ,
@xmath7 layers stack along the @xmath8-axis direction and each magnetic layer consists of @xmath9 ions carrying @xmath1 spins arranged in an orthogonal dimer structure that is topologically equivalent to the ssl . from several experimental observations ,
it was confirmed that the field dependence of the magnetization exhibits multiple magnetization plateaus . in these magnetization plateau states , wigner crystals of spin - triplet dimers@xcite are realized reflecting the strong competition between the kinetic energy gain and mutual repulsion of the dimers .
theoretically , this model has been studied in great detail@xcite and several states , such as the plaquette singlet state at zero field@xcite and a spin supersolid state in the magnetic fields@xcite were predicted . in a recent paper by sebastian et al.@xcite ,
the possibility of fractional magnetic plateaus was discussed in analogy to the quantum hall effect .
fractional magnetic plateaus have recently been discovered at low temperatures in rare - earth tetraborides @xmath10 [ r is a rare - earth element]@xcite .
the magnetic moment carrying @xmath11 ions in these compounds are arranged in a ssl in the @xmath12-plane . in @xmath2 , an extended magnetization plateau at @xmath0 was confirmed for @xmath13[t]@xmath14 3.6[t ] when a magnetic field is applied along the @xmath8-axis@xcite . here
@xmath15 is the normalized value by the saturation magnetization , @xmath16 .
in contrast to @xmath6 a strong anisotropy along the @xmath8-axis is expected owing to the crystal fields . from specific heat measurements@xcite
, it has been suggested that the degeneracy of the @xmath3=@xmath17 multiplet of @xmath18 is lifted - the lowest energy state for a single ion is the non - kramers doublet with @xmath19=@xmath20 and there exists a large energy gap to the first excited doublet . by restricting the local hilbert space to the lowest energy doublet
, the low - energy magnetic properties of the material can be described by a s=1/2 xxz model with ising - like exchange anisotropy and a ferromagnetic transverse coupling as discussed in the next section .
the magnetization curves for the effective hamiltonian have already been calculated@xcite .
the effective hamiltonian is the @xmath1 ising - like xxz model on the ssl and it is described by @xmath21 where @xmath22 denotes the ising anisotropy and @xmath23 . in the ising limit @xmath24 , it has been established by several approaches - such as monte carlo simulations @xcite and tensor renormalization - group analysis@xcite - that only @xmath25 plateau is stabilized .
the presence of the @xmath0 plateau has been confirmed when quantum spin fluctuation is included@xcite , and the ground - state phase diagram for @xmath26 has been calculated . however , even for finite @xmath22 , the @xmath25 plateau phase extends over a wider range of applied fields than the @xmath0 plateau phase .
since the @xmath25 plateau is not observed in @xmath27 , the above hamiltonian is insufficient to explain the experimental observations in @xmath2 . in a previous letter@xcite
, we argued that ferromagnetic @xmath28 and antiferromagnetic @xmath29 couplings ( see fig . [ model ] )
are necessary to explain the stabilization of an extended @xmath0 plateau in the absence of the @xmath25 plateau .
we also investigated the finite - temperature phase transition to the @xmath0 plateau state .
the results of finite - size scaling analysis indicated that a two - step second - order transition takes place - the difference between two critical temperatures is within 0.5% of @xmath3 . the universality class at both critical points is explained by the critical exponents of the two - dimensional ising model . for the finite - temperature transition
, a nontrivial question remains unanswered . in the @xmath0 plateau phase ,
the lowest energy state is four - fold degenerate .
therefore , it is naively expected that the universality class should be the same as that of the four - state potts model . as we discuss in this paper , the critical behavior
can indeed be explained by the four - state potts universality in the ising limit , while the quantum spin model shows a two - step transition with both transitions belonging to the two - dimensional ising universality class .
hence the phase diagrams for the thermal phase transitions for the two are different although both models possess the same symmetry .
the low energy behavior of the @xmath0 plateau is shown to be described by the generalized four - state chiral clock model .
as far as we know , the finite - temperature transition of the generalized chiral four - state clock model has not been studied precisely . therefore , clarification of the finite - temperature transition to the @xmath0 plateau phase is also valuable from the view point of statistical mechanics . in the present paper ,
we discuss the properties of the model introduced in our previous letter in greater details , and clarify the nature of the phase transitions to the @xmath0 plateau state .
, nearest - neighbor coupling @xmath4 , and additional couplings @xmath29 and @xmath28 . ] this paper is organized as follows . in section
ii , we reproduce the derivation of an effective hamiltonian that captures the unique magnetization features in @xmath2 .
an approach from the ising limit of the effective hamiltonian shows that @xmath28 couplings are important to stabilize the @xmath0 plateau state .
in section iii , we focus on the finite - temperature properties of the effective model and re - investigate the universality class of the transition to the @xmath0 plateau phase by extensive quantum monte carlo simulations .
we show that the critical properties of the quantum model is different from those in the ising limit . in section iv , a classical model , the generalized chiral four - state clock model ,
is proposed in order to understand the critical behavior of the original model . from monte carlo simulations
, we find that the topology of the phase diagram for the classical model is the same as that of the original model . in comparison with the phase diagram for the classical model
, we discuss the effects of quantum exchange on the finite - temperature transition . in section v , we discuss the magnetic properties of @xmath2 and predict the critical properties of the finite - temperature transition to the @xmath0 plateau phase .
vi is devoted to a summary of the results .
in this section , we derive the effective model for the low energy part of magnetic properties of @xmath2 again . as discussed in the introduction ,
the ground state of the single @xmath18 ions is a @xmath30 doublet with a large energy gap between the first excited doublet . at low temperatures ( compared to this energy gap )
, one can ignore the contributions from the higher multiplets and the low energy properties of the material are well described by an effective two - state model consisting of the @xmath30 doublet .
the fact that the interactions between the moments are derived from itinerant electrons enables us to expect isotropic type ( heisenberg type ) interactions .
therefore , we start from the hamiltonian , @xmath31 , where @xmath32 , @xmath33 , and @xmath34 , where @xmath35 is the coupling constant between two moments on the site @xmath36 and @xmath37 , and @xmath38 is the easy - axis anisotropy from the crystal fields . by treating @xmath39 as the biggest term , we find that there are four degenerate states , namely @xmath40 , @xmath41 , @xmath42 , and @xmath43 , in the lowest energy level .
we apply perturbation theory for @xmath44 and calculate the matrix elements among these four states .
the hamiltonian for arbitrary moment pairs can be described by the @xmath1 ising - like xxz hamiltonian , @xmath45 , where @xmath46 .
most importantly , the matrix elements for the transverse coupling ( @xmath47 and @xmath48 ) are proportional to @xmath49 . consequently
, if the magnetic moment @xmath3 is even , the transverse coupling always becomes ferromagnetic .
( note that the longitudinal component of the interaction becomes ferromagnetic or antiferromagnetic depending on the sign of @xmath35 . )
the interaction @xmath35 is expected to be the rkky type , because this compound is a metal .
therefore , the effect of the long - range interactions is significant when the magnetic properties of @xmath2 are considered .
the principal interactions necessary to reproduce the dominant features of magnetization in @xmath2 - in addition to the conventional ssl model with @xmath3 and @xmath4 couplings - are the @xmath29 and @xmath28 couplings , shown in fig . [ model ]
there is another coupling with shorter range than that of @xmath29 and @xmath28 - the next - nearest neighbor coupling @xmath50 that is orthogonal to @xmath3 in the plaquettes with the diagonal interactions in the original ssl model .
however , as we show in the following analysis , @xmath50 is not efficient in stabilizing the @xmath0 plateau state , because it stabilizes the @xmath25 plateau at the same time .
the hamiltonian considered here is described by @xmath51 where @xmath52 , @xmath53 , @xmath54 , and @xmath55 denote sums over all pairs on the bonds with the @xmath3 , @xmath4 , @xmath29 , and @xmath28 couplings , respectively .
the positive ( negative ) sign of each coupling denotes antiferromagnetic ( ferromagnetic ) interaction .
the original ssl model interactions are always assumed to be antiferromagnetic , i.e. , @xmath56 and @xmath57 . in the following
, we set @xmath3 as the unit of energy and express all the parameters of the model in units of @xmath3 .
we studied the above model on square lattices of the form @xmath58 with periodic boundary conditions .
when @xmath59 , the magnetic properties of the hamiltonian ( [ ham ] ) are qualitatively explained by considering the ising limit ( @xmath24 ) . at @xmath60 , there are two candidates of the ground states at @xmath61 , where @xmath62 and @xmath63 . for @xmath64 , each spin pair located on the diagonal bond @xmath3 forms an antiferromagnetic @xmath65 dimer ( acd ) ( fig .
[ spinconfig ] ( a ) ) , but macroscopic degeneracy remains due to cancellations among interactions on the coupling @xmath4 .
the local energy of such state can be estimated as @xmath66 . in the other limit @xmath67
, the system approaches the antiferromagnetic ising model on the square lattice , and the nel state becomes the ground state ( fig .
[ spinconfig ] ( b ) ) .
the local energy of the nel state is calculated as @xmath68 . as is expected , the energy levels of the two states cross at @xmath69 and @xmath70 , and the system shows a phase transition at the point .
the local energies of the @xmath71 and @xmath72 plateau state are estimated in the same manner .
the spin configurations in fig .
[ spinconfig ] ( c ) and ( d ) are realized in the @xmath72 and @xmath71 plateau phase , respectively .
( these configurations were also confirmed from the snapshot of spin configuration in the monte carlo simulations in our previous study@xcite . ) from the spin configurations , we obtain @xmath73 and @xmath74 , where @xmath75 and @xmath76 denote the local energy of the @xmath72 and @xmath71 plateau state . since the energy of the fully polarized state is given by @xmath77 , the magnetization curve shows a jump from the @xmath25 to @xmath78 at @xmath79 .
we show the field dependence of local energy for these five states in fig .
[ isingmodel ] .
the energies of the @xmath0 and @xmath25 plateaus , and the fully polarized state are always degenerate at the saturation field @xmath80 .
consequently , the degeneracy between these three states should be lifted when the additional couplings @xmath29 and @xmath28 are included . by estimating the energy gains due to @xmath29 and @xmath28
, we find that the ferromagnetic @xmath28 coupling is one of the most efficient ways to stabilize the @xmath71 plateau @xcite .
this analysis further demonstrates that the coupling @xmath50 , which is perpendicular to the diagonal coupling @xmath3 , does not stabilize the @xmath0 plateau , because the number of antiparallel spin pairs on the @xmath50 couplings equals that of the parallel ones .
hence this coupling was not included in the effective hamiltonian .
-symmetric phase ( see text ) .
solid ( open ) ellipses mean spin pairs having the total magnetization @xmath81 ( @xmath82 ) . ]
the qualitative discussion of the previous section is supplemented by large scale numerical simulation of the underlying model .
we developed a variant of the standard stochastic - series - expansion quantum - monte - carlo method based on the modified directed - loop algorithm to treat the longer - range interactions efficiently .
@xcite using this new algorithm , simulations of the hamiltonian ( [ ham ] ) were performed on square lattices @xmath58 . note that the negative sign problem does not appear in the quantum monte carlo computations for the hamiltonian ( [ ham ] ) , because the transverse coupling is ferromagnetic as discussed in the previous section . fig .
[ magnetization ] shows the magnetization curves up to @xmath83 , for different values of the ratio @xmath84 , and how the curves evolve upon the introduction of ferro- and antiferro - magnetic @xmath28 . in the computations , we fixed the @xmath28 coupling constant at @xmath85 and the ising anisotropy @xmath86 at @xmath87 . at @xmath60 ,
two magnetization plateaus appear at @xmath0 and 1/3 , consistent with previous studies @xcite .
when the longitudinal component of @xmath28 coupling is ferromagnetic , the system shows a strong tendency towards the stabilizing an extended plateau at @xmath0 .
snapshots of the monte carlo simulations in the @xmath0 plateau phase confirm the realization of spin configuration shown in fig .
[ spinconfig ] ( d ) .
as long as there is the finite ferromagnetic @xmath28 coupling , the @xmath25 plateau region shrinks , for all values of the ratio @xmath84 .
this follows naturally from the analysis of the ising spin model discussed above . in the ising limit
, the field extent of the @xmath0 plateau _ expands _ by an amount proportional to 4@xmath88 , while that of the @xmath25 plateau _ contracts _ by an amount proportional to 6@xmath88 . when the longitudinal component of @xmath28 coupling is antiferromagnetic both @xmath0 and @xmath72 plateaus disappear . for antiferromagnetic @xmath28 ,
no conclusive evidence of any plateaus has been obtained in our calculations except for @xmath69 .
a feature appears in the magnetization curve around @xmath89 for @xmath69 .
[ 2/9plateau ] ( a ) shows the expansion of the magnetization curve around @xmath90 for @xmath69 . for @xmath91 ,
the magnetization curve becomes flat at @xmath92 .
significantly , the @xmath92 plateau was also obtained for the parent shastry - sutherland model and observed in @xmath93.@xcite to identify the spin structure , we calculated the spin - spin and the bond - spin correlation functions .
( as for the definition of bond spins , please see eq .
( [ bond - spin ] ) ) the results are shown in fig .
[ 2/9plateau ] ( b ) and ( c ) , respectively . in this plateau region ,
the @xmath94 structure accompanying @xmath95 rotational symmetry breaking is clearly stabilized .
this is the same symmetry breaking as the @xmath0 plateau state .
this plateau state has a characteristic feature : the periodicity of the @xmath92 plateau state is longer than the distance of @xmath28 couplings .
furthermore , the periodicity of this plateau is the same as that discussed in ref .
@xcite , but the precise spin configuration is a little bit different from their results shown in figure 5 in ref .
@xcite because the magnitude of the moments on the qlaquette without the diagonal bonds is lager than that on the diagonal bonds having total @xmath96 .
( color online ) and @xmath87 .
circles indicate the results at @xmath97 case .
triangles and squares are the results of the ferromagnetic and antiferromagnetic longitudinal @xmath28 couplings , respectively .
all symbols accompany error bars , which are smaller than the symbol size ( here and the following figures ) . ] and the antiferromagnetic @xmath98 .
( b ) and ( c ) are results for spin - spin and bond - spin correlations at @xmath99 .
colored ( open ) circles correspond to the positive ( negative ) amplitude .
the area of circles is proportional to the amplitude .
error bars are smaller than the symbol sizes .
( d ) schematic spin configuration in the @xmath92 plateau state .
solid circles denote up spins .
yellow circles surrounded by up spins are occupied by up or down spins with almost equal probability . ] in the following , we focus on the finite - temperature transition to the @xmath0 plateau state .
the snapshot of the spin configuration of the @xmath0 plateau phase leads us to expect the @xmath95 ( @xmath100 ) rotation symmetry breaking around the center of the plaquette without diagonal coupling @xmath3 . in context of the bare spin language
, the universality class of the finite - temperature transition is expected to be the four - state potts universality class because the lowest energy state in the @xmath0 plateau is four - fold degenerate .
however , the other scenario is also possible for @xmath101 because the group @xmath100 has a subgroup @xmath102 . by introducing the @xmath95 lattice rotation `` @xmath103 '' around the center
, we can express the symmetry group in the paramagnetic phase as @xmath104 .
if the system exhibits a two - step phase transition , the symmetry breaks down to @xmath105 at the higher critical temperature @xmath106 , and the remaining symmetry breaks down to the trivial group @xmath107 at the lower critical temperature @xmath108 . to investigate the symmetry breaking at the critical point ,
we introduce two order parameters .
the order parameter of @xmath95 rotation symmetry breaking can be expressed by @xmath109 where @xmath110 is position of the diagonal bond , @xmath111 is a product of longitudinal component of two spins on the bond @xmath110 , and @xmath112 takes @xmath113 depending on the position of the diagonal coupling ( fig .
[ orderparameter ] ( a ) ) . that of @xmath114 rotation symmetry breaking is also given by @xmath115 where the suffixes of longitudinal spin operators represent the site indexes shown in fig .
[ orderparameter ] ( b ) , @xmath116 is a positional vector for plaquette . for @xmath117 ,
the system is in the paramagnetic phase , and @xmath118 and @xmath119 are satisfied . for @xmath120 ,
the system retains the @xmath114 ( @xmath102 ) rotation symmetry , and then @xmath121 and @xmath119 . below @xmath108 ,
the lowest energy state is characterized by @xmath121 and @xmath122 .
is defined by inner product of pair spins on each diagonal bond ( bold line ) and @xmath123 is the staggered magnetization of @xmath124 .
the bold black and gray lines indicate the sign of the factor @xmath112 .
( b ) definition of sublattice . ]
first , we show the results of the finite - temperature transition in the ising limit ( @xmath125 ) .
the monte carlo simulations up to @xmath126 were performed at @xmath127 and @xmath128 .
[ ising ] ( a ) displays the temperature dependence of the correlation ratio of the bare spins at @xmath129 and @xmath130 for different system sizes .
the curves cross at the critical temperature @xmath131 , reflecting the size invariance of the correlation ratio at the critical temperature .
the spin fluctuations freeze at this temperature . fig . [ ising ]
( b ) and ( c ) show the temperature dependence of the binder ratios @xmath132 for @xmath123 and @xmath133 for @xmath134 , respectively .
the system size dependence of both binder ratios disappears at the same critical temperature .
the crossing of the curves for @xmath132 and @xmath133 also indicates that the @xmath100 symmetry breaks down to the trivial group at @xmath131 .
the obtained results suggest that the transition belongs to the universality class of the four - state potts model .
this is confirmed by finite - size scaling analysis .
for the fourth order cumulant of @xmath123 and @xmath134 , we assume the scaling forms @xmath135 and @xmath136 , where @xmath137 is a scaling function and @xmath138 and @xmath139 denote the fourth order cumulant of @xmath123 and @xmath134 , respectively . for the static structure factor of the bond spins ,
@xmath140 is assumed , where @xmath141}$ ] . in the analysis ,
the critical exponents and the critical temperature are fixed at @xmath142 , @xmath143@xcite and @xmath144 , the results are shown in fig .
[ isingfss ] .
the excellent data collapse confirms that the transition in the ising limit belongs to the universality class of the four state potts model . in the ising limit case .
( a ) correlation ratio of the bare spin correlation .
( b ) the binder ratio @xmath132 of the bond - spin operator @xmath145 .
( c ) the binder ratio @xmath133 of @xmath134 . ] , @xmath146 , and @xmath139 in the ising limit case .
( a ) and ( c ) are the results for the fourth - order cumulant @xmath138 and @xmath139 .
( b ) scaling analysis for the static structure factor @xmath146 of bond spins . ]
next , we show the results of the quantum spin model with @xmath147 .
we performed computations up to @xmath148 .
[ fss_second ] shows the temperature dependence of the same order parameters as those defined in the above .
the obtained results clearly indicate that the transition is a two - step process . the curves for @xmath132 cross at @xmath149 .
this means that nel order of the bond spins @xmath124 is realized for @xmath150 .
since a quotient group @xmath151 , which is isomorphic to the group @xmath102 , breaks down at @xmath106 , it is naively expected that the transition belongs to the two - dimensional ising model universality class .
accordingly we performed finite - size scaling analysis assuming the scaling form @xmath135 and @xmath152 .
we emphasize here that the data collapse shown in fig .
[ fss_second ] ( c ) and ( d ) was obtained without any adjustable parameter , confirming that the transition at @xmath106 indeed belongs to the two - dimensional ising universality class .
[ fss_second ] ( b ) shows the temperature dependence of the binder ratio @xmath133 for @xmath134 , and it provides an evidence of the other phase transition .
the curves for @xmath133 for different system sizes intersect at @xmath153 and the difference @xmath154 is approximately @xmath155 , slightly larger than the previously obtained value@xcite .
the universality class of the phase transition at @xmath108 should be the same as that of the two - dimensional ising model because the remaining symmetry @xmath102 is also isomorphic to @xmath156 .
this is confirmed by the finite - size scaling results presented in fig .
[ fss_second ] ( e ) where we have used the same scaling function and critical exponent as for @xmath132 , viz . @xmath136 and @xmath152 . consequently , we conclude that there @xmath157 an intermediate phase with @xmath102-rotation symmetry phase that can be characterized by the bond nel order accompanying the internal antiferromagnetic bare - spin fluctuation .
the schematic spin configuration in this @xmath102-rotation symmetry phase is shown in fig .
[ spinconfig ] ( e ) .
if the antiferromagnetic fluctuation on the bond @xmath3 is relevant to the stabilization of the intermediate phase , the temperature range of the @xmath102-rotation symmetric phase expands upon increasing @xmath22 . in fig .
[ phasediagram ] , @xmath22 dependence of the critical temperatures is presented for fixed @xmath127 , @xmath158 and @xmath159 .
the results show that the critical temperature associated with the breaking of the @xmath114 rotational symmetry shifts to the lower temperature as @xmath86 increases .
this suggests that the dimerization of antiparallel spin on the diagonal bonds plays a key role in the transition .
as discussed in the followings , the phase diagram is qualitatively understood from a comparison with that of the generalized chiral four - state clock model . plateau phase .
( a ) and ( b ) are temperature dependence of the binder ratio of @xmath132 and @xmath133 , respectively .
( c ) and ( d ) are the results of the finite - size scaling analysis for fourth - order cumulant @xmath138 of staggered magnetization @xmath123 and the static structure factor @xmath146 of bond spins .
( e ) and ( f ) correspond to ( c ) and ( d ) for @xmath134 . in ( f ) , the static structure factor of @xmath134 is calculated from @xmath160 $ ] ] and @xmath161 .
i.m . in the phase diagram means the @xmath102-rotational symmetric phase .
schismatic spin configurations in the spin ordered phase and i.m .
are shown in fig .
[ spinconfig ] ( d ) and ( e ) , respectively .
open and solid circles indicate the second order transition with the two - dimensional ising universality , and an open square is that with the four - state potts universality .
an open triangle is the kosterslitz - thouless transition .
all lines are guide to the eyes . ]
in the previous section , we have investigated the finite - temperature phase transition to the @xmath0 plateau phase for the classical and quantum spin models .
since the symmetry of both models is same and the finite - temperature phase transition is focused on , it is expected that the schematic phase diagram should be the same
. however , the results obtained are different from this expectation . in order to understand the discrepancy
, we consider the generalized chiral four - state clock model .
as we show below , this simplified classical model captures the characteristics of phase diagram for the original model .
we begin by introducing eight - spin unit clusters ( windmill clusters ) in the @xmath0 plateau state as shown in fig .
[ orderparameter](b ) . in this phase ,
a down spin occupies one of four sites on a plaquette without the diagonal bond and up spins occupy the remaining three sites . if a down spin occupies the site labeled by ` 2 ' in fig .
[ orderparameter](b ) , the nearest down spin prefers to locate at the third - neighbor distance from it as long as the ferromagnetic @xmath28 coupling .
this indicates that the down spins in the @xmath0 plateau state occupies the same edge of diagonal bond connecting each plaquette . since the ferromagnetic @xmath28 coupling is essential for stabilizing the @xmath0 plateau state ( as discussed in sec .
ii ) , it is reasonable to describe the @xmath0 plateau state by an arrangement of such clusters .
this description becomes exact in the limit @xmath162 and @xmath125 . in the ising limit @xmath125 ,
the spin configuration in the @xmath0 plateau state can be expressed by the arrangement of the windmill clusters shown in fig .
[ atmodel ] ( a ) .
there are four kinds of clusters corresponding to the @xmath95-rotation symmetry breaking in the @xmath0 plateau .
we assign clock spins pointing at an angle to each clusters . here
we ignore clusters having @xmath163 for simplicity .
thus we obtain the simplified classical hamiltonian that can be regarded as the generalized four - state clock model on a square lattice with the nearest and next - nearest neighbor interactions , @xmath164 where @xmath165 denotes an angle of clock spins and takes @xmath166 , or @xmath167 .
the value of @xmath165 denotes the position of a down spin in the windmill clusters ( see fig .
[ atmodel ] ( a ) ) . @xmath168 and @xmath169 are interactions between two clock spins , and their values are listed in table i. the interactions , @xmath168 and @xmath169 , are calculated from the sums of coupling energy between spins located in different clusters . when @xmath125 , the values , a , b , c , and d , in @xmath168 and @xmath169 ( see table i ) are evaluated as a=4@xmath28 , b=2@xmath4 , c=0 , d=4@xmath4 , respectively .
the resulting classical hamiltonian retains @xmath100 rotational symmetry conditionally : the system requires a lattice rotation when a global angle shift of all clock spins is executed
. such requirement arises from the geometric characteristics of the ssl .
since the symmetry of the hamiltonian @xmath170 is lower than that of @xmath171 , we focus on the critical properties of @xmath170 in the followings .
[ table1 ] lcccc + @xmath172 & 0 & @xmath173 & @xmath174 & @xmath167 + 0 & a & b & c & b + @xmath173 & b & a & b & d + @xmath174 & c & b & a & b + @xmath167 & b & -d & b & a + lcccc + @xmath172 & 0 & @xmath173 & @xmath174 & @xmath167 + 0 & a & b & d & b + @xmath173 & b & a & b & c + @xmath174 & -d & b & a & b + @xmath167 & b & c & b & a + lcccc + @xmath172 & 0 & @xmath173 & @xmath174 & @xmath167 + 0 & a & c & a & c + @xmath173 & c & a & c & -a + @xmath174 & a & c & a & c + @xmath167 & c & -a & c & a + lcccc + @xmath172 & 0 & @xmath173 & @xmath174 & @xmath167 + 0 & a & c & -a & c + @xmath173 & c & a & c & a + @xmath174 & -a & c & a & c + @xmath167 & c & a & c & a + the simplified hamiltonian @xmath170 is identical to the conventional four - state potts model in the limit b = c = d=0 .
therefore , it is trivial that the finite - temperature phase transition at this point is a single second - order transition belonging to the four - state potts universality .
however , the universality class becomes nontrivial , when b@xmath175d@xmath1750 .
performing monte carlo simulations , we numerically investigated the finite - temperature transition of the classical hamiltonian @xmath170 with d=2b and a=-4 up to @xmath176 .
we summarize the obtained critical temperatures in fig .
[ tcs ] ( a ) .
these critical temperatures were estimated from the finite - size scaling analysis for the binder ratio and correlation ratio of the order parameters , @xmath177 and @xmath178 . and @xmath169 ( @xmath179 and @xmath180 ) .
] and @xmath178 , respectively .
red and blue symbols correspond to the cross sections on red solid and dashed lines drawn in ( b ) and ( c ) , respectively .
( b ) and ( c ) are expected phase diagrams of the present clock model . in ( b ) and ( c ) , an area painted by green indicates the region where the two - step phase transition with two - dimensional ising universality takes place .
a blue quarter circle in ( c ) means the area where the system undergoes a single phase transition . ] in the following , we focus on the results when c / a=0 and b / a are varied ( along the y - axis in fig . [ tcs ] ( b ) and ( c ) ) . for c / a=0 , the system undergoes a single second - order transition when the coupling ratio satisfies @xmath181 . to identify the universality class , we performed finite - size scaling analysis and present the results in fig .
[ clock40 ] .
( the scaling forms assumed here are the same as those shown in the previous section . ) in the analysis , we estimated @xmath182 from the crossing of the curves for the correlation ratio , @xmath183 , for different system sizes , where @xmath184 and @xmath185 . for @xmath186 , excellent data collapse is obtained with @xmath142 and @xmath143 .
this strongly suggests that the transition belongs to universality class of the four - state potts model@xcite . for @xmath187 , the critical exponent increases from @xmath142 to @xmath152 with the fixed ratio @xmath188 as b / a decreases . for @xmath189 ,
we confirm a clear evidence of two - step phase transition with @xmath152 at both critical temperatures .
therefore , the critical properties for @xmath189 belong to the two - dimensional ising universality class .
next we discuss the results when the parameters are varied along the dashed lines in fig .
[ tcs ] ( b ) and ( c ) , where b / a=-0.05 and @xmath190 . for @xmath191 ,
the system clearly undergoes a two - step phase transition with both transitions belonging to the two - dimensional ising universality class .
[ clock41 ] presents the results of the finite - size scaling analysis at b / a=-0.05 and c / a=0.5 . when the two - step phase transition takes place , the @xmath95 rotation symmetry breaks at the higher critical temperature and the @xmath114 rotation symmetry survives in the intermediate phase .
the lower critical temperature decreases as the difference between a and c decreases . for @xmath192 ,
the system seems to undergo a single phase transition and the @xmath114 rotation symmetry survives for any finite temperature .
( the static structure factor of @xmath193 ) when c / a=0 . in these analysis
, we estimated all critical temperatures from crossing points of the correlation ratios @xmath194 and @xmath195 .
the data collapse in ( a1 ) and ( b1 ) ( ( a4 ) and ( b4 ) ) was obtained by using the fixed critical exponents @xmath142 and @xmath143 ( @xmath152 and @xmath196 ) .
the critical exponents in ( a2 ) , ( b2 ) , ( a3 ) , and ( b3 ) were estimated to minimize the discrepancies among data . ] ( a ) and @xmath197 ( b ) at b / a=-0.1 and c / a=0.5 .
( c ) and ( d ) are the same analysis for the structure factor of each order parameter . in these analysis
, each critical temperature is estimated from a cross of the correlation ratios @xmath198 and @xmath199 .
all results are obtained by using the fixed critical exponents @xmath152 and @xmath196 . ]
the obtained results allow us to consider two scenarios of the phase diagram for the present four - state clock model .
one is that the system shows a single phase transition only at b = c=0 and the two - step phase transition takes place for @xmath200 and @xmath201 as shown in fig .
[ tcs ] ( b ) .
the other is that there exists a region having the four - state potts universality class around @xmath202 , denoted by the area in fig .
[ tcs ] ( c ) .
the critical exponents may show a crossover behavior at the boundary of the area .
further studies of the critical behavior of the extended four - state clock model are currently underway .
we compare the phase diagrams shown in fig .
[ phasediagram ] and fig .
[ tcs ] here .
the topology of the phase diagram for the simplified classical model is the same as that of the ising - like xxz model on the ssl . in both models ,
the region of the intermediate phase , where only the @xmath95 rotational symmetry is broken , shrinks rapidly as the system approaches the four - state - potts universality point .
when we consider the quantum effects in the original hamiltonian ( [ ham ] ) , fluctuation of antiferromagnetic spin pairs on the diagonal bond is enhanced when @xmath22 increases .
then , the sublattice magnetization of antiparallel spin pair closes to zero . in the case , where the antiparallel spin pair becomes a classical or quantum - triplet dimer with @xmath203 , the clock spins pointing towards @xmath204 and @xmath174 ( or @xmath173 and @xmath167 ) direction can not be distinguished from each other .
such dimerization effect is captured as a reduction in the difference between a and c. consequently , the lower transition temperature @xmath108 associated with the @xmath114 rotational symmetry breaking shifts toward @xmath70 as @xmath22 increases . by considering the parameter set associated with the ising limit of the original hamiltonian , we can estimate a parameter set of @xmath170 as a@xmath205 , b@xmath206 , c@xmath207 , and d@xmath208 . for @xmath127 and @xmath209 ,
@xmath170 shows a single phase transition with the critical exponent @xmath210 , in clear disagreement with that of the original model .
this is because we ignored the effect of the next - nearest neighbor couplings and the other states excluded from the four states treated here . a more quantitative estimation of the parameter set ( a
, b , c , d ) is required by adding such effects for a quantitative mapping of the original hamiltonian over the entire parameter range , but this is beyond the scope of this paper .
finally , we discuss the magnetization properties of @xmath2 . the @xmath0 plateau state is observed for @xmath2111.9[t]@[email protected][t ] at 2[k]@xcite . the coupling ratio @xmath214 and a strong ising anisotropy @xmath215
are expected from the crystal structure of @xmath2 and specific heat measurements@xcite .
these experimental results help us estimate the other parameters for the present model .
we obtain @xmath216 and @xmath217 at @xmath218 from the local energy estimation in the ising limit .
the magnetization curves at fixed @xmath219 , @xmath220 , @xmath218 and @xmath221 are shown in fig .
[ magcurve ] ( a ) .
it is found that the magnetization jumps appear at @xmath222 and @xmath223 not only in the ising limit but also for @xmath224 . from these critical fields
, we roughly estimate @xmath225/@xmath226 .
this value is in good agreement with the experimental value .
therefore our estimation for the coupling ratio is quantitatively consistent with the critical fields of the magnetization jumps observed in @xmath2 .
the @xmath227 plateau observed in the experiments @xcite appears to depend on the history of the system ; @xmath227 was observed around @xmath228[t ] , when the fields were decreased from the saturation fields , whereas the magnetization remained vanishingly small around @xmath228[t ] during the upsweep . in other words ,
a hysteresis loop was observed around @xmath228[t ] .
the inset of fig .
[ magcurve ] ( a ) is the magnetization curves at @xmath229 when we perform field sweeps .
we started with the nel state ( the fully - polarized state ) and increased ( decreased ) the applied field from @xmath230 ( @xmath231 ) .
a hysteresis behavior similar to the experiments can be observed in short - time simulations .
although the monte carlo dynamics is not directly comparable to the real dynamics , this similarity is suggestive .
the magnetization curves in the inset suggest a presence of shoulder around @xmath232 for @xmath233 . in this field region , we have confirmed the existence of a mixed state comprised of the nel order and domain walls constructed of fully - polarized spin chains from the snapshot of spin configuration .
the period of the domain walls is fluctuating and the average of the period changes continuously as the field decreases .
thus , the shoulder around @xmath232 seems to be a transient process in successive plateaus having the magnetization @xmath234 .
since the magnetization value of the shoulder shifts to the lower magnetization value as the system size increases , it seems to be smeared out in the thermodynamic limit .
more quantitative discussion of the @xmath2351/8 shoulder is desirable via comparison with the other experimental observations .
, @xmath219 @xmath220 .
red solid ( open ) symbols represent the results of the ising limit case ( the quantum spin model case ) when they have been obtained by the field sweeps from @xmath231 and the slow cooling .
blue ones are the results of the field sweeps from @xmath230 .
inset of ( a ) is the results of short time monte carlo calculations when the field sweep is performed .
( b ) and ( c ) show a finite - size scaling analysis for the fourth - order cumulant of @xmath132 and @xmath133 , respectively . ]
we comment the finite - temperature transition to the @xmath0 plateau phase . using the above parameters , we have studied it for the ising spin model and the quantum spin model numerically .
[ magcurve ] ( b ) and ( c ) are the results of the scaling analysis , when the quantum spin model is considered .
the results indicate that a single second - order transition takes place and the critical exponent equals @xmath236 .
the obtained results indicate that the critical property of the finite - temperature transition to the @xmath0 plateau phase can be explained by that of the four - state potts model .
this critical behavior was also confirmed in the ising limit case .
therefore , if our estimation for the coupling constants is correct , we expect the experimentally observed critical exponents in @xmath2 to belong to the four - state potts universality class .
the @xmath0 plateau has been observed in the other ssl compound @xmath237@xcite .
the magnetic moments of @xmath237 are derived from @xmath238 ions and the magnitude equals @xmath239 . in this compound ,
the strong ising anisotropy perpendicular to the ssl plains has been suggested .
when we focus on the magnetic properties at a very low temperature , the effective hamiltonian of the magnetic properties can be also described by the present model in the ising limit .
therefore we believe that our results help understand the magnetic order of the @xmath0 plateau observed in @xmath237 .
in summary , we have studied the magnetic properties of a model of interacting spins with ising - like exchange anisotropy and longer range interactions on the ssl that captures the low - temperature magnetic properties of the rare - earth tetraboride , @xmath2 .
we have focused on the finite - temperature transition to the @xmath0 plateau phase and investigated the universality class of the transition . in the ising limit
, the system shows a single second - order transition .
the critical behavior is well explained by that of the four - state potts model .
when there exists quantum exchange interactions , it has been confirmed that there is the two - step phase transition with an intervening intermediate phase .
both the transitions are belonging to the two - dimensional ising universality class . from the finite - size scaling analysis
, we have ascertained that the intermediate phase can be characterized by the `` @xmath114-rotation symmetric '' state retaining @xmath174-rotation symmetry on individual triplets with @xmath240 .
since the symmetry of the quantum spin model is the same as that of the classical spin version , it is naively expected that the critical behavior of both models will also be the same .
however , the obtained results have indicated that the phase diagrams for the finite - temperature transition are different . to understand the critical behavior of both the models , we have proposed the simplified classical model , namely the generalized four - state chiral clock model . by performing the monte carlo computations for the simplified classical model
, we have found that the universality class at the critical temperatures and the topology of the phase diagram are in agreement with those of the original models on the ssl .
finally , we have studied the magnetic properties of @xmath2 . from the ising limit analysis
, we have estimated the parameters that can explain ( qualitatively ) the magnetization curves observed in experiments . from the short - time monte carlo simulations ,
we have suggested that the @xmath227 plateau seems to be a metastable state .
we have also investigated the finite - temperature transition to the @xmath0 plateau state .
if our estimation for the coupling constants is correct , measurements of the critical exponents belonging to the universality class of the four - state potts model are expected in experiment .
the present research subject was suggested by c. d. batista and we would like to thank him .
the computation in the present work is executed on computers at the supercomputer center , institute for solid state physics , university of tokyo . the present work is financially supported by grant - in - aid for young scientists ( b ) ( 21740245 ) , grant - in - aid for scientific research ( b ) ( 19340109 ) , grant - in - aid for scientific research on priority areas `` novel states of matter induced by frustration '' ( 19052004 ) , and by next generation supercomputing project , nanoscience program , mext , japan .
s. onoda and y. tanaka , phys .
lett . * 105 * , 047201 ( 2010 ) .
z. y. meng , t. c. lang , s. wessel , f. f. assaad , and a. muramatsu , nature * 464 * , 847 ( 2010 ) .
sebastian , n. harrison , p. sengupta , c. d. batista , s. francoual , e. palm , t. murphy , h. a. dabkowska , and b. d. gaulin , pnas * 105 * , 20157 ( 2008 ) .
b. s. shastry and b. sutherland , physica b * 108 * , 1069 ( 1981 ) .
h. kageyama , k. oknizuka , y. ueda , n. v. mushnikov , t. goto , k. yoshimura , a. kosuge , j. phys .
* 67 * , 4304 ( 1998 )
. h. kageyama , k. yoshimura , r. stern , n. v. mushnikov , k. onizuka , m. kato , k. kosuge , c. p. slichter , t. goto , and y. ueda , phys .
* 82 * , 3168 ( 1999 ) .
m. takigawa , t. waki , m. horvatic , and c. berthier , j. phys .
79 , 11005 ( 2010 ) .
s. miyahara and k. ueda , j. phys . : condens .
matt . * 15 * , r327 ( 2003 ) .
a. koga and n. kawakami , phys , rev , lett , * 84 * , 4461 ( 2000 ) .
a. lauchli , s. wessel , and m. sigrist , phys .
b * 66 * , 014401 ( 2002 ) .
t. momoi and k. totsuka , phys .
b * 62 * , 15067 ( 2000 ) .
s. yoshii , t. yamamoto , m. hagiwara , t. takeuchi , a. shigekawa , s. michimura , f. iga , t. takabatake , and k. kindo , j. magn .
magn . mater . *
310 * , 1282 - 1284 ( 2007 ) .
s. michimura , a. shigekawa , f. iga , m. sera , t. takabatake , k. ohyama , and y. okabe , physica b * 378 - 380 * , 596 - 597 ( 2006 ) .
s. yoshii , t. yamamoto , m. hagiwara , a. shigekawa , s.michimura , f. iga , t. takabatake , and k. kindo , j. phys .
ser . * 51 * , 59 - 62 ( 2006 ) .
k. siemensmeyer , e. wylf , h .- j .
mikeska , k. flachbart , s. gabni , s. ma@xmath241a@xmath242 , p. priputen , a. evdokimova , and n. shitsevalova , arxiv:0712.1537 ( 2007 ) .
f. iga , a. shigekawa , y. hasegawa , s. michimura , t. takabatake , s. yoshii , t. yamamoto , m. hagiwara , k. kindo , j. magn .
mater , * 310 * , e443-e445 ( 2007 ) .
s. gabni , s. matas , p. priputen , k. flachbart , k. siemensmeyer , e. wulf , a. evdokimova , and n. shitsevalova , acta .
phys . pol .
a * 116 * , 227 ( 2008 ) .
k. siemensmeyer , e. wulf , h .- j .
mikeska , k. flachbart , s. gabni , s. matas , p. priputen , a. efdokimova , and n. shitsevalova , phys .
* 101 * , 177201 ( 2008 ) .
z. y. meng and s. wessel , phys .
b * 78 * , 224416 ( 2008 ) .
f. liu and s. sachdev , arxiv:0904.3018 .
c . chang and m .- f .
yang , phys . rev .
b 79 , 104411 ( 2009 ) . t. suzuki , y. tomita , and n. kawashima , phys . rev .
b 80 , 180405(r ) ( 2009 ) . y. kato and n. kawashima , phys .
e * 79 * , 021104 ( 2009 ) .
j. dorier , k. p. schmidt , and f. mila , phys .
lett . 101 , 250402 ( 2008 ) .
k. fukui and s. todo , j. comp
* 28 * , 2629 ( 2009 ) .
b. niehuis , in _ phase transition and critical phenomena _ , edited by c. domb and j. l. lebowitz ( adademic press , new york , 1987 ) , vol .
the energy gains of the @xmath72 plateau , @xmath71 plateau and the fully polarized state are @xmath243 , @xmath244 , and @xmath245 . | we study the finite - temperature transition to the @xmath0 magnetization plateau in a model of interacting @xmath1 spins with longer range interactions and strong exchange anisotropy on the geometrically frustrated shastry - sutherland lattice .
this model was shown to capture the qualitative features of the field - induced magnetization plateaus in the rare - earth tetraboride , @xmath2 .
our results show that the transition to the plateau state occurs via two successive transitions with the two - dimensional ising universality class , when the quantum exchange interactions are finite , whereas a single phase transition takes place in the purely ising limit . to better understand these behaviors , we perform monte carlo simulations of the classical generalized four - state chiral clock model and compare the phase diagrams of the two models .
finally , we estimate a parameter set that can explain the magnetization curves observed in @xmath2 .
the magnetic properties and critical behavior of the finite - temperature transition to the @xmath0 plateau state are also discussed . |
in a previous paper @xcite , we presented a comprehensive analysis on the lhc signatures of the type ii seesaw model of neutrino masses in the nondegenerate case of the triplet scalars . in this companion paper , another important signature the pair and associated production of the neutral scalars
is explored in great detail .
this is correlated to the pair production of the standard model ( sm ) higgs boson , @xmath15 , which has attracted lots of theoretical and experimental interest @xcite since its discovery @xcite , because the pair production can be used to gain information on the electroweak symmetry breaking sector @xcite .
since any new ingredients in the scalar sector can potentially alter the production and decay properties of the higgs boson , a thorough examination of the properties offers a diagnostic tool to physics effects beyond the sm .
the higgs boson pair production has been well studied for collider phenomenology in the framework of the sm and beyond @xcite , and extensively studied in various new physics models @xcite , as well as in the effective field theory approach of anomalous couplings @xcite and effective operators @xcite .
the pair production of the sm higgs boson proceeds dominantly through the gluon fusion process @xcite , and has a cross section at the @xmath16 lhc ( lhc14 ) of about @xmath17 at leading order @xcite . at next - to - leading order @xcite and to @xmath18 at next - to - next - to - leading order @xcite . ]
it can be utilized to measure the higgs trilinear coupling .
a series of studies have surveyed its observability in the @xmath3 , @xmath4 , @xmath19 , @xmath20 , and @xmath21 signal channels @xcite . for the theoretical and experimental status of the higgs trilinear coupling and pair production at the lhc , see refs .
@xcite . in summary , at the @xmath16 lhc with an integrated luminosity of @xmath6 ( lhc14@3000 ) , the trilinear coupling could be measured at an accuracy of @xmath22 @xcite , and thus leaves potential space for new physics . as we pointed out in ref .
@xcite , in the negative scenario of the type ii seesaw model where the doubly charged scalars @xmath23 are the heaviest and the neutral ones @xmath0 the lightest , i.e. , @xmath1 , the associated @xmath11 production gives the same signals as the sm higgs pair production while enjoying a larger cross section .
the leading production channel is the drell - yan process @xmath24 , with a typical cross section @xmath25-@xmath26 in the mass region @xmath27-@xmath28 .
additionally , there exists a sizable enhancement from the cascade decays of the heavier charged scalars , which also gives some indirect evidence for these particles .
the purpose of this paper is to examine the importance of the @xmath11 production with an emphasis on the contribution from cascade decays and to explore their observability .
the paper is organized as follows . in sec .
[ decay ] , we summarize the relevant part of the type ii seesaw and explore the decay properties of @xmath29 in the negative scenario .
sections [ eh ] and [ signal ] contain our systematical analysis of the impact of cascade decays on the @xmath0 production in the three signal channels , @xmath3 , @xmath4 , and @xmath5 .
we discuss the observability of the signals and estimate the required integrated luminosity for a certain mass reach and significance . discussions and conclusions are presented in sec .
[ dis ] . in most cases
, we will follow the notations and conventions in ref . @xcite .
the type ii seesaw and its various experimental constraints have been reviewed in our previous work @xcite .
here we recall the most relevant content that is necessary for our study of the decay properties of the scalars in this section and of their detection at the lhc in later sections .
the type ii seesaw model introduces an extra scalar triplet @xmath30 of hypercharge two @xcite on top of the sm higgs doublet @xmath31 of hypercharge unity .
writing @xmath30 in matrix form , the most general scalar potential is @xmath32 as in the sm , @xmath33 is assumed to trigger spontaneous symmetry breaking , while @xmath34 sets the mass scale of the new scalars .
the vacuum expectation value ( vev ) @xmath35 of @xmath31 then induces via the @xmath36 term a vev @xmath37 for @xmath30 .
the components of equal charge ( and also of identical @xmath38 in the case of neutral components ) in @xmath30 and @xmath31 then mix into physical scalars @xmath39 ; @xmath40 ; @xmath41 and would - be goldstone bosons @xmath42 , with the mixing angles specified by ( see , for instance , refs .
@xcite ) @xmath43 where an auxiliary parameter is introduced for convenience , @xmath44 to a good approximation , the sm - like higgs boson @xmath15 has the mass @xmath45 , the new neutral scalars @xmath29 have an equal mass @xmath46 , and the new scalars of various charges are equidistant in squared masses : @xmath47 there are thus two scenarios of spectra , positive or negative , according to the sign of @xmath48 . for convenience ,
we define @xmath49 . and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] and @xmath23 versus @xmath50 at some benchmark points of @xmath51 and @xmath52 : @xmath53 , from the upper to the lower panels .
[ brhp],title="fig : " ] in the rest of this section , we discuss the decay properties of the new scalars in the negative scenario with an emphasis on @xmath54 and @xmath40 .
the explicit expressions for the relevant decay widths can be found in refs .
it has been shown that @xmath0 decays dominantly into neutrinos for @xmath55 @xcite , resulting in totally invisible final states .
we will restrict ourselves to @xmath56 in this work , where @xmath0 dominantly decays into visible particles .
before we detail their decay properties , we give a brief account of the cascade decays of the charged scalars .
the branching ratios of the cascade decays are controlled by the three parameters , @xmath52 , @xmath51 , and @xmath50 .
the cascade decays dominate in the moderate region of @xmath52 and for @xmath51 not too small , where a minimum value of @xmath57 appears around @xmath58 @xcite . in fig .
[ brhp ] , the branching ratios of @xmath59 and @xmath23 are shown as a function of @xmath50 at some benchmark points of @xmath52 and @xmath51 . basically speaking , in the mass region @xmath60-@xmath28 ,
the cascade decays are dominant for a relatively large mass splitting @xmath51 ( as shown in the middle panel of fig .
[ brhp ] ) or a relatively small @xmath52 ( in the lower panel ) . at tree level
, @xmath54 can decay to @xmath61 , @xmath62 , @xmath63 , @xmath64 , and @xmath65 .
it can also decay to @xmath66 , @xmath67 , and @xmath68 through radiative effects .
similarly , @xmath69 , @xmath62 , @xmath70 at tree level , and it has the same decay modes as @xmath54 at the loop level . since we have chosen @xmath56 , the neutrino mode can be safely neglected for both @xmath54 and @xmath40 .
previous work usually concentrated on the decoupling region where the neutral scalars @xmath0 are much heaver than the light @xmath38-even higgs @xmath15 and the scalar self - couplings @xmath71 are taken to be zero for simplicity @xcite . in this case , the mixing angle @xmath72 , and the @xmath73 coupling [ being proportional to @xmath74 tends to vanish .
as a consequence , the @xmath75-pair mode is absent and the dominant channels are @xmath76 , @xmath64 for a heavy @xmath54 . in contrast , we take into account the effect of scalar self - interactions and focus on the nondecoupling regime , i.e. , @xmath0 are not much heavier than @xmath15 . for illustration
, we choose the benchmark values @xmath77 , @xmath78 ; then , @xmath48 is determined by eq .
( [ massrelation ] ) upon specifying @xmath79 . in the range
@xmath80-@xmath81 would not change the branching ratios significantly . ] to investigate the effect of the scalar self - interactions , we note the following features in the decays of @xmath54 .
1 ) the decay widths of @xmath82 differ from those of @xmath15 only by a factor of @xmath83 , which leads to similar behavior for @xmath54 and @xmath15 .
2 ) the only free parameter for the mixing between @xmath54 and @xmath15 is @xmath84 , because [ as shown in eq .
( [ mixangles ] ) ] the impact of @xmath85 is suppressed by a small @xmath52 and a relatively large mass difference between @xmath50 and @xmath86 while @xmath87 is fixed by @xmath86 .
3 ) @xmath84 enters the @xmath73 and @xmath88 couplings and thus affects the decays @xmath89 .
4 ) the @xmath90 coupling simplifies for @xmath91 such that the only free parameter in the decay @xmath92 is again @xmath84 . as a consequence of these features
, we shall choose @xmath84 as a free parameter and vary it in the range @xmath93 $ ] , and fix the couplings @xmath94 which are involved in loop - induced decays . and @xmath95 as a function of @xmath96 for various values of @xmath84 .
[ brh0tt],title="fig : " ] and @xmath95 as a function of @xmath96 for various values of @xmath84 .
[ brh0tt],title="fig : " ] we first examine the branching ratios of @xmath97 .
br(@xmath98 ) and br(@xmath95 ) are plotted in fig .
[ brh0tt ] for different mass regions of @xmath54 . for light fermions @xmath99 and gluons is similar , so we only present br(@xmath98 ) in fig .
[ brh0tt ] . ]
it is clear that the variation of br(@xmath98 ) is more dramatic for @xmath100 .
the maximum of br(@xmath98 ) appears at @xmath101 .
obviously , br(@xmath98 ) is a nonmonotonic function of @xmath84 , while br(@xmath95 ) monotonically increases with @xmath84 . as will be discussed later ,
this different behavior in the two mass regions is due mainly to a zero in the @xmath88 coupling . as a function of @xmath96 in the mass region @xmath27-@xmath28 .
right : branching ratios of @xmath102 as a function of @xmath96 in the mass region @xmath103-@xmath104 .
[ brh0ww],title="fig : " ] as a function of @xmath96 in the mass region @xmath27-@xmath28 .
right : branching ratios of @xmath102 as a function of @xmath96 in the mass region @xmath103-@xmath104 .
[ brh0ww],title="fig : " ] as a function of @xmath96 in the mass region @xmath27-@xmath28 .
right : branching ratios of @xmath102 as a function of @xmath96 in the mass region @xmath103-@xmath104 .
[ brh0ww],title="fig : " ] as a function of @xmath96 in the mass region @xmath27-@xmath28 .
right : branching ratios of @xmath102 as a function of @xmath96 in the mass region @xmath103-@xmath104 .
[ brh0ww],title="fig : " ] now we study the bosonic decays @xmath105 . in the left panel of fig .
[ brh0ww ] , we present the branching ratios of @xmath89 in the mass region @xmath27-@xmath28 . for most values of @xmath84 ,
br(@xmath106 ) increases with @xmath96 when @xmath107 , and varying @xmath84 for @xmath100 changes it considerably .
@xmath84 has a strong impact on br(@xmath106 ) in the mass region @xmath108 where the decay channel dominates overwhelmingly for @xmath109 but becomes negligible for @xmath84 approaching about @xmath110 .
however , once the @xmath92 channel is opened , @xmath106 is suppressed significantly independent of @xmath84 . the decay @xmath111 can not dominate when @xmath107 . in the mass region @xmath108 , it is complementary with the @xmath112 channel , so their behavior is just opposite .
more interestingly , there is a zero point for the @xmath88 coupling , which is proportional to @xmath113 . according to eq .
( [ mixangles ] ) , one obtains the corresponding @xmath50 at the zero : @xmath114 note that the above relation only holds for @xmath115 , since we are working in the scenario where @xmath116 .
the existence of the zero coupling explains the presence of the nodes in br(@xmath111 ) for @xmath117 . as a function of @xmath96 for various sets of @xmath118 values .
[ brh0aa],title="fig : " ] as a function of @xmath96 for various sets of @xmath118 values .
[ brh0aa],title="fig : " ] in the right panel of fig .
[ brh0ww ] , br(@xmath119 ) are shown in the mass region @xmath103-@xmath104 .
when @xmath120 , the dependence on @xmath84 is simple : a larger @xmath84 corresponds to a smaller br(@xmath92 ) and a larger br(@xmath111 ) .
it is clear that @xmath84 has a more significant impact in the mass region @xmath121 , and varying @xmath84 could change br(@xmath111 ) from @xmath122 to @xmath123 .
once @xmath96 exceeds @xmath124 , the evolution of br(@xmath119 ) becomes smooth with the increase of @xmath96 .
there also exists a zero point for the @xmath90 coupling , which can be obtained as for the @xmath64 channel : @xmath125 which is valid for @xmath126 .
finally , we investigate the loop - induced decays , @xmath127 .
in addition to the usual contributions from the top quark and @xmath75 boson , the new charged scalars @xmath59 and @xmath23 also contribute to the decays .
these new terms involve the @xmath128 and @xmath129 couplings , which are proportional to @xmath130 , \nonumber \\
h^0h^{++}h^{--}&:&(\lambda_2\sin\alpha\cos\theta_0-\lambda_4\cos\alpha\sin\theta_0).\end{aligned}\ ] ] one therefore has to consider the scalar self - couplings @xmath85 . for simplicity , we set @xmath131 and vary them from @xmath132 to @xmath133 . in fig .
[ brh0aa ] , we display br(@xmath134 ) and br(@xmath135 ) versus @xmath96 for some typical sets of @xmath118 values .
the evolution of both branching ratios crosses 3 orders of magnitude in this parameter region .
the resulting enhancement compared with @xmath136 in the sm looks significant : the maximal enhancement can be achieved at the level of @xmath137 for the @xmath134 channel at @xmath138 , and of @xmath139 for the @xmath135 channel at @xmath140 . as a function of @xmath141 for various values of @xmath84 .
[ bra0tt],title="fig : " ] as a function of @xmath141 for various values of @xmath84 .
[ bra0tt],title="fig : " ] similar to @xmath54 , the decay widths of @xmath142 differ from those of @xmath15 by a factor of @xmath143 with @xmath144 being given in eq .
( [ mixangles ] ) .
the only vertex which involves @xmath71 is the @xmath145 coupling proportional to @xmath146 . as a consequence
, one can only choose @xmath84 as a free parameter to illustrate the influence of scalar interactions .
in this section , we also vary @xmath84 from @xmath147 to @xmath148 and take the same benchmark values for @xmath52 and @xmath51 as for the @xmath54 decays . in the left panel of fig .
[ bra0tt ] , we present br(@xmath149 ) as a function of @xmath141 . on the @xmath150 channels
is similar to the @xmath151 mode . ] for a fixed value of @xmath84 , br(@xmath149 ) decreases as @xmath141 increases .
the dependence of br(@xmath149 ) on @xmath84 is simple : the larger @xmath84 is , the larger br(@xmath149 ) is . and
br(@xmath149 ) can be dominant with @xmath152 as long as @xmath153 is not fully opened .
the right panel of fig .
[ bra0tt ] shows br(@xmath154 ) , which is very similar to br(@xmath95 ) . as a function of @xmath141 for various values of @xmath84 .. [ bra0zh],title="fig : "
] as a function of @xmath141 for various values of @xmath84 .. [ bra0zh],title="fig : " ] and @xmath155 as a function of @xmath141 for various values of @xmath84 .
[ bra0aa],title="fig : " ] and @xmath155 as a function of @xmath141 for various values of @xmath84 .
[ bra0aa],title="fig : " ] we then study the most important decay @xmath153 . in fig .
[ bra0zh ] , we present br(@xmath153 ) as a function of @xmath141 in the low - mass region ( @xmath27-@xmath28 ) and high - mass region ( @xmath156-@xmath104 ) , respectively . the evolution of br(@xmath153 ) with @xmath141 and @xmath84 is just opposite to that of @xmath157 in the low- ( high- ) mass region .
the variation of br(@xmath153 ) with @xmath84 is dramatic below the @xmath70 threshold . in particular , near the @xmath70 threshold br(@xmath158 for @xmath159 , while br(@xmath153 ) tends to vanish for @xmath152 , which corresponds to the zero point of the @xmath145 coupling : @xmath160 with @xmath161 .
br(@xmath153 ) is totally dominant in the mass region between the @xmath70 and @xmath162 thresholds , and becomes comparable to br(@xmath154 ) when @xmath163 . at last , we study the one - loop - induced decays , @xmath164 .
these two channels can only be induced by the top quark in the loop since the @xmath165 , @xmath166 , and @xmath167 couplings are absent in the @xmath38-conserving case . in fig .
[ bra0aa ] , both br(@xmath168 ) and br(@xmath155 ) are displayed . for @xmath141 below the @xmath70 threshold , the variation in @xmath84 of br(@xmath168 )
increases as @xmath141 increases .
br(@xmath169 ) could reach @xmath170 for @xmath171 and @xmath152 , which is much smaller than the maximum of br(@xmath172 ) .
the variation in @xmath84 of br(@xmath155 ) is slightly steeper , with a maximum of @xmath173 at @xmath174 and @xmath152 . as a function of @xmath175 at the benchmark point in eq .
( [ bp ] ) .
[ brh0],title="fig : " ] as a function of @xmath175 at the benchmark point in eq .
( [ bp ] ) .
[ brh0],title="fig : " ] in the above , we have discussed the decay channels of @xmath54 and @xmath40 separately .
we have shown that the scalar self - interactions have a large impact on their branching ratios . in sec .
[ signal ] , we will explore their lhc signatures .
for this purpose , we choose the following benchmark values : @xmath176 the reason that we set relatively small values of @xmath52 and @xmath51 is to obtain large cascade decays of charged scalars as well as a large enhancement of neutral scalar production . in fig .
[ brh0 ] , we display all relevant branching ratios versus @xmath175 for this benchmark model , which is to be simulated in sec .
[ signal ] for the lhc in the @xmath3 , @xmath4 , and @xmath19 signal channels .
we pointed out in ref . @xcite the importance of the associated @xmath11 production in the nondegenerate case . to estimate the number of signal events , we simulated the signal channel @xmath4 at @xmath177 .
we found that , with a much higher production cross section than the sm higgs pair ( @xmath65 ) production , a @xmath178 excess in that signal channel is achievable for lhc14@300 . in the present work , we are interested in the observability of the associated @xmath11 production in the nondecoupling mass regime @xmath179-@xmath180 . in fig .
[ cs ] we first show the production cross sections for a pair of various scalars at lhc14 versus @xmath50 with a degenerate spectrum .
as before , we incorporate the next - to - leading - order ( nlo ) qcd effects by multiplying a @xmath181-factor of @xmath182 in all @xmath183 production channels @xcite . the @xmath65 production through gluon - gluon fusion at nlo ( @xmath184 )
is also indicated ( black dashed line ) for comparison .
one can see that the cross section for @xmath11 is about @xmath25-@xmath26 in the mass region @xmath27-@xmath28 , which is much larger than the @xmath65 production for most of the mass region and thus leads to great discovery potential . for a degenerate spectrum .
the black dashed line is for the sm @xmath65 production .
[ cs],title="fig : " ] for a degenerate spectrum .
the black dashed line is for the sm @xmath65 production .
[ cs],title="fig : " ] in general , the new scalars are nondegenerate for a nonzero @xmath48 . in the positive scenario where @xmath23 are the lightest
, the cascade decays of @xmath39 and @xmath0 can strengthen the observability of @xmath23 @xcite .
for the same reason , in the negative scenario where @xmath0 are the lightest , the charged scalars contribute instead to the production of @xmath0 through the cascade decays like @xmath185 . in this work ,
we study these contributions in the same way as was done for the positive scenario in refs .
@xcite .
we define the reference cross section @xmath186 for the standard drell - yan process @xmath187 which is independent of the cascade decay parameters @xmath52 and @xmath51 . a detailed study on the @xmath4 signal for this process with @xmath188 can be found in ref . @xcite . besides the above direct production , neutral scalars can also be produced from cascade decays of charged scalars .
these extra production channels include @xmath189 , @xmath190 , @xmath191 , and @xmath192 followed by cascade decays of charged scalars .
we consider first the associated @xmath189 production followed by cascade decays of @xmath39 , @xmath193 resulting in three final states classified by a pair of neutral scalars : @xmath194 , @xmath195 , and @xmath196 . noting that the last two originate only from cascade decays , any detection of such production channels would be a hint of charged scalars being involved .
using the fact that @xmath197 as well as the narrow width approximation , we calculate the production cross sections for these three final states : @xmath198\times \mbox{br}(h^\pm \to a^0 w^ * ) , \\
w^-\to h^- h^0)]\times \mbox{br}(h^\pm \to h^0 w^ * ) , \\
a^0a^0:z_1&=&[\sigma(pp\to w^+\to h^+ a^0)+\sigma(pp\to w^-\to h^- a^0)]\times \mbox{br}(h^\pm \to a^0 w^*).\end{aligned}\ ] ] the factor 2 in @xmath199 accounts for the equal contribution from the process with @xmath54 and @xmath40 interchanged
. the relations @xmath200 actually hold true for all of the four production channels , since for a given channel the same branching ratios ( such as for @xmath185 ) are involved , @xmath201 where @xmath202 , and @xmath203 refer to the cross sections for @xmath11 , @xmath195 , and @xmath196 production with the subscript @xmath204 denoting the production channels @xmath189 , @xmath205 , @xmath191 , and @xmath192 , respectively .
the relations imply that we may concentrate on the cross section of @xmath11 production .
naively , one would expect the next important channel to be @xmath190 since it only involves two cascade decays : @xmath206 but as already mentioned in ref .
@xcite , a smaller coupling and destructive interference between the @xmath207 and @xmath208 exchange make the cross section of @xmath190 production an order of magnitude smaller than that of @xmath11 even for a degenerate spectrum . considering further suppression due to cascade decays , @xmath209 is not important for the enhancement of @xmath11 production and can be safely neglected in the numerical analysis . at lhc14 and with @xmath78 , @xmath210 .
left : the red solid ( dashed ) line corresponds to @xmath186 ( @xmath211 ) .
right : the red line corresponds to @xmath212 from cascade decays @xmath213 , and the green line to @xmath11 from cascade decays ( @xmath214 ) .
the shaded regions are filled by scanning over @xmath51 and @xmath37 .
[ csx],title="fig : " ] at lhc14 and with @xmath78 , @xmath210 . left : the red solid ( dashed ) line corresponds to @xmath186 ( @xmath211 ) .
right : the red line corresponds to @xmath212 from cascade decays @xmath213 , and the green line to @xmath11 from cascade decays ( @xmath214 ) .
the shaded regions are filled by scanning over @xmath51 and @xmath37 .
[ csx],title="fig : " ] the contribution from @xmath191 is more important despite the fact that it involves three cascade decays : @xmath215\times\\ \nonumber & & \mbox{br}(h^{\pm\pm}\to h^\pm w^*)\mbox{br}(h^{\pm}\to h^0 w^*)\mbox{br}(h^{\pm}\to a^0 w^*).\end{aligned}\ ] ] as shown in fig .
[ cs ] , @xmath216 is the largest for a degenerate mass spectrum .
when cascade decays are dominant , the phase - space suppression of heavy charged scalars will be important .
so we expect that the @xmath11 production receives considerable enhancement from @xmath191 when the mass splitting is small and cascade decays are dominant .
finally , the last mechanism is @xmath192 , which involves four cascade decays : @xmath217 this mechanism is also promising since the cross section of @xmath192 production is slightly larger than @xmath218 production for a degenerate mass spectrum .
the phase - space suppression of @xmath219 is more severe than that of @xmath220 , because a pair of the heaviest @xmath23 are produced .
summing over all four of the above channels yields the contribution to the @xmath11 production from cascade decays , @xmath221 and the total production cross section of @xmath11 is then @xmath222 .
using eq .
( [ xxx ] ) , the total cross sections for the pair production @xmath212 , @xmath223 , @xmath224 , are given by @xmath225 since the enhancement from cascade decays depends on a not severely suppressed phase space and a larger branching ratio of cascade decays , we choose to work with a relatively smaller mass splitting and triplet vev as shown in eq .
( [ bp ] ) .
figure [ csx ] displays the cross sections of the @xmath11 , @xmath195 , and @xmath196 production as a function of @xmath50 .
as can be seen from the figure , the production of @xmath11 can be enhanced by a factor of 3 , while the @xmath212 production at the maximal enhancement can reach the level of @xmath186 .
this could make the detection of neutral scalar pair productions very promising in the negative scenario .
in this section we investigate the signatures of neutral scalar production at the lhc . from previous studies on the sm @xmath65 production , we already know that the most promising signal is @xmath3 , and @xmath4 is next to it , while both semileptonic and dileptonic decays of @xmath75 s in the @xmath19 channel are challenging . in this work
we analyze all three of the signals@xmath3 , @xmath4 , and @xmath226 ( @xmath227 for collider identification)as well as their backgrounds based on the benchmark model presented in eq .
( [ bp ] ) . in sec .
[ eh ] we discussed the drell - yan production of @xmath11 and the enhanced pair and associated production of neutral scalars @xmath0 due to cascade decays of charged scalars @xmath228 .
we are now ready to incorporate the branching ratios of @xmath0 decays for a specific signal channel .
for instance , the cross sections for the @xmath3 signal channel can be written as @xmath229,\\ s(b\bar{b}\gamma\gamma ) & = & x\times\left[\mbox{br}(h^0\to b\bar{b})\mbox{br}(a^0\to\gamma\gamma ) + \mbox{br}(h^0\to\gamma\gamma)\mbox{br}(a^0\to b\bar{b})\right]\\\nonumber & & + 2y\hspace{-0.25em}\times\mbox{br}(h^0\to b\bar{b})\mbox{br}(h^0\to\gamma\gamma ) + 2z\hspace{-0.35em}\times\mbox{br}(a^0\to b\bar{b})\mbox{br}(a^0\to\gamma\gamma).\end{aligned}\ ] ] here @xmath230 denotes the signal from the direct production @xmath24 alone , and @xmath231 includes contributions from cascade decays .
@xmath232 has a similar expression as @xmath233 , while @xmath234 is simpler since the decay mode @xmath235 is absent . , @xmath4 , and @xmath236 signal channels at lhc14 .
the red solid ( dashed ) line corresponds to the signal from @xmath186 ( @xmath211 ) , the green ( blue ) solid line corresponds to the signal from @xmath237 , and the purple dashed line shows the total cross section @xmath231 for the signal . the sm @xmath65 cross section is shown for comparison .
the lower right panel shows the enhancement factor @xmath238 in the three signal channels .
[ sgn],title="fig : " ] , @xmath4 , and @xmath236 signal channels at lhc14 .
the red solid ( dashed ) line corresponds to the signal from @xmath186 ( @xmath211 ) , the green ( blue ) solid line corresponds to the signal from @xmath237 , and the purple dashed line shows the total cross section @xmath231 for the signal .
the sm @xmath65 cross section is shown for comparison .
the lower right panel shows the enhancement factor @xmath238 in the three signal channels .
[ sgn],title="fig : " ] , @xmath4 , and @xmath236 signal channels at lhc14 .
the red solid ( dashed ) line corresponds to the signal from @xmath186 ( @xmath211 ) , the green ( blue ) solid line corresponds to the signal from @xmath237 , and the purple dashed line shows the total cross section @xmath231 for the signal .
the sm @xmath65 cross section is shown for comparison .
the lower right panel shows the enhancement factor @xmath238 in the three signal channels .
[ sgn],title="fig : " ] , @xmath4 , and @xmath236 signal channels at lhc14 .
the red solid ( dashed ) line corresponds to the signal from @xmath186 ( @xmath211 ) , the green ( blue ) solid line corresponds to the signal from @xmath237 , and the purple dashed line shows the total cross section @xmath231 for the signal .
the sm @xmath65 cross section is shown for comparison .
the lower right panel shows the enhancement factor @xmath238 in the three signal channels .
[ sgn],title="fig : " ] the theoretical cross sections for the @xmath3 , @xmath4 , and @xmath236 signal channels are plotted in fig .
[ sgn ] . the cross section @xmath239 is larger than that of the sm @xmath65 production until @xmath240 ; taking into account cascade enhancement pushes the corresponding @xmath50 further to @xmath241 .
@xmath242 is always larger than that of @xmath65 in the mass region @xmath27-@xmath14 , and interestingly , it keeps about the same value when @xmath243 .
the signal from @xmath195 is comparable with @xmath230 in these three channels only for @xmath243 , while in contrast the signal from @xmath196 becomes dominant for the @xmath3 and @xmath4 channels when @xmath244 .
therefore , we have a chance to probe the @xmath196 pair production in these two channels . also shown in fig .
[ sgn ] is the enhancement factor @xmath238 for the three signal channels at the benchmark point ( [ bp ] ) as a function of @xmath79 , which will help us understand the simulation results . in our simulation ,
the parton - level signal and background events are generated with * madgraph5 * @xcite .
we perform parton shower and fast detector simulations with * pythia * @xcite and * delphes3 * @xcite .
finally , * madanalysis5 * @xcite is responsible for data analysis and plotting . we take a flat @xmath245-tagging efficiency of 70% , and mistagging rates of 10% for @xmath246 jets and 1% for light - flavor jets , respectively .
jet reconstruction is done using the anti-@xmath247 algorithm with a radius parameter of @xmath248 .
we further assume a photon identification efficiency of @xmath249 and a jet - faking - photon rate of @xmath250 @xcite .
the main sm backgrounds to the signal are as follows : @xmath251 among them , @xmath3 and @xmath70 are irreducible , while @xmath252 is reducible and can be reduced by vetoing the additional @xmath253 s with @xmath254 and @xmath255 .
in addition , there exist many reducible sources of fake @xmath3 : @xmath256 where @xmath257 stands for a final - state @xmath258 misidentified as @xmath259 .
the remaining fake sources are subdominant and are thus not included in our simulation .
the qcd corrections to the backgrounds are included by a multiplicative @xmath181-factor of 1.10 and 1.33 for the leading cross sections of @xmath252 and @xmath70 at lhc14 @xcite , respectively .
the cross section of the @xmath3 background has been normalized to include fake sources and does not take nlo corrections into account . , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] , and @xmath260 for the signal @xmath3 and its backgrounds before applying any cuts at lhc14 .
[ fig : bbaa],title="fig : " ] the distributions of some kinematical variables before applying any cuts are shown in fig . [
fig : bbaa ] , where we assume @xmath261 . in our analysis , we require that the final states include exactly one @xmath245-jet pair and one @xmath262 pair and satisfy the following basic cuts : @xmath263 where @xmath264 is the particle separation , with @xmath265 and @xmath266 being the separation in the azimuthal angle and rapidity , respectively . here
we employ a tighter @xmath267 cut than is usually applied to suppress the qcd - electroweak @xmath3 background . the @xmath245-jet pair and @xmath262 pair
are then required to fall in the following windows on the invariant masses and fulfill the @xmath268 cut criteria : @xmath269 [ ! htbp ]
.evolution of signal and background cross sections ( in @xmath270 ) at lhc14 for the @xmath3 signal channel upon imposing the cuts one by one . for the cascade - enhanced signal
only the cross section passing all the cuts is shown .
the last two columns assume an integrated luminosity of @xmath6 .
[ tab : bbaacut ] [ cols="^,^,^,^,^,^,^",options="header " , ] with both @xmath75 s decaying leptonically , the final state appears as @xmath5 .
the dominant sm backgrounds are as follows : @xmath271 as before , the qcd correction is included by a multiplicative @xmath181-factor of 1.35 for the @xmath272 production @xcite .
we pick up the events that include exactly one @xmath245-jet pair and one opposite - sign lepton pair and filter them with the basic cuts : @xmath273 the separation and invariant mass of the @xmath245-jet pair are required to fulfill @xmath274 for the lepton pair , we reconstruct the transverse cluster mass @xmath275 : @xmath276 the distributions of @xmath275 , @xmath277 , and @xmath278 are shown in fig .
[ fig : bbll2v ] .
the peak of @xmath275 is always lower than @xmath50 by about @xmath279-@xmath280 , and the lepton separation @xmath277 in the signal is much smaller than in the @xmath272 background .
accordingly , we set a wide window on @xmath281 while tightening up the cuts on @xmath277 and @xmath278 : @xmath282 we find that @xmath281 is least efficient around @xmath283 , where the peak of @xmath281 for the @xmath272 background is around @xmath284 .
the very tight cuts on @xmath277 and @xmath278 are sufficient to suppress the background by 1 or 2 orders of magnitude , while keeping the number of signal events as large as possible .
we further combine the @xmath245-jet pair and the lepton pair into a cluster and construct the transverse cluster mass : @xmath285 of the @xmath3 channel versus @xmath50 reachable at lhc14@300 ( red region ) and lhc14@3000 ( green ) .
right : required luminosity to reach a @xmath286 ( red region ) and @xmath7 ( green ) significance in the @xmath3 channel versus @xmath50 at lhc14 .
the solid line corresponds to the signal from @xmath186 alone , and the dashed line corresponds to the total signal including cascade enhancement .
[ bbaa_sen],title="fig : " ] of the @xmath3 channel versus @xmath50 reachable at lhc14@300 ( red region ) and lhc14@3000 ( green ) .
right : required luminosity to reach a @xmath286 ( red region ) and @xmath7 ( green ) significance in the @xmath3 channel versus @xmath50 at lhc14 .
the solid line corresponds to the signal from @xmath186 alone , and the dashed line corresponds to the total signal including cascade enhancement .
[ bbaa_sen],title="fig : " ] which is an analog of @xmath287 in the previous subsection .
the distribution of @xmath288 is displayed in fig .
[ fig : bbll2v ] , which is very similar to that of @xmath287 in the @xmath3 channel .
although it looks from the @xmath288 distributions ( before any cuts are made ) that the @xmath272 background has a large overlap with the signal , the cuts on @xmath275 , @xmath277 , and @xmath278 actually modify them remarkably , so that a further cut on @xmath288 could improve the significance efficiently .
we apply a cut on @xmath288 as we did with @xmath287 , as well as one on @xmath260 : @xmath289 the results following the cutflow are summarized in table [ tab : bbllcut ] . for @xmath188 ,
the final significance is 4.41 ( 17.7 ) without ( with ) cascade enhancement . with cascade enhancement
this should be enough to discover the neutral scalars .
the signal channel is more promising for @xmath290 due to a slightly larger cross section and higher cut efficiencies .
the final significance is 7.78 ( 23.2 ) , which is also better than the @xmath3 and @xmath4 channels with the same mass .
finally , for @xmath291 , the significance becomes 3.56 ( 10.1 ) .
therefore , for our benchmark model , the only promising signal for such heavy neutral scalars ( @xmath292 ) comes from the @xmath19 channel . , but for the @xmath4 channel .
[ bbtata_sen],title="fig : " ] , but for the @xmath4 channel .
[ bbtata_sen],title="fig : " ] based on our elaborate analysis of signal channels in secs . [ sec : bbgg][sec : bbww ] , we examine the observability of the neutral scalars @xmath293 in the mass region @xmath294 by adopting essentially the same cuts as before . in the left panel of figs
. [ bbaa_sen ] , [ bbtata_sen ] , and [ bbll2v_sen ] we present the significance @xmath295 as a function of @xmath50 in the three signal channels @xmath3 , @xmath4 , and @xmath5 that is reachable for lhc14@300 and lhc14@3000 , respectively . the required luminosity to achieve a @xmath286 and @xmath7 significance
is displayed in the right panel of the figures . as was done in our previous analysis
, the effect of cascade enhancement is included by a factor @xmath238 in the final results .
as shown in figs .
[ bbaa_sen ] and [ bbtata_sen ] , both the @xmath3 and @xmath4 channel are typically sensitive to the low - mass region ( @xmath296 ) . in the absence of cascade enhancement , the @xmath286 significance
would never be reached for @xmath297 in the @xmath3 ( @xmath4 ) channel for lhc14@300 . however
, a cascade enhancement of @xmath298 ( as can be seen from fig . [ sgn ] ) in this mass region can greatly improve the observability , pushing the @xmath286 mass limit up to @xmath299 in the @xmath3 ( @xmath4 ) channel .
moreover , with cascade enhancement , one has a good chance to reach a @xmath7 significance if @xmath300 . in other words ,
the cascade enhancement significantly reduces the required luminosity .
for instance , to achieve a @xmath286 and @xmath7 significance in the @xmath3 ( @xmath4 ) channel with @xmath188 , the required luminosity is as low as @xmath301 and @xmath302 at lhc14 , respectively .
the @xmath4 channel is more promising , thanks to a relatively larger production rate . at the future
lhc14 with @xmath6 data , the heavier mass region can also be probed . with a maximal cascade enhancement , the @xmath286 and @xmath7 mass reach
is pushed to @xmath13 and @xmath303 , respectively , in the @xmath3 channel , which should be compared to @xmath304 and @xmath305 in the absence of enhancement .
for the @xmath4 channel , the enhancement factor @xmath238 can reach about @xmath306 above the @xmath75-pair threshold , upshifting the @xmath286 and @xmath7 mass reach to @xmath307 and @xmath308 , respectively , from @xmath309 and @xmath284 without the enhancement . , but for the @xmath5 channel .
[ bbll2v_sen],title="fig : " ] , but for the @xmath5 channel .
[ bbll2v_sen],title="fig : " ] the @xmath5 channel shown in fig .
[ bbll2v_sen ] is more special , compared with @xmath3 and @xmath4 .
it is relatively more sensitive to a higher mass between @xmath9-@xmath10 , where the decay mode @xmath106 dominates , while its observability deteriorates for @xmath310 due to phase - space suppression in the decay .
the cascade enhancement @xmath238 at our benchmark point ( [ bp ] ) is typically @xmath311-@xmath312 in the mass region @xmath27-@xmath14 , and decreases as @xmath50 increases . for lhc14@300 , the @xmath286 and @xmath7 mass reach is , respectively , @xmath313 and @xmath314 with maximal cascade enhancement .
these limits would just increase by @xmath315-@xmath316 for lhc14@3000 if there were no cascade enhancement , while with cascade enhancement the @xmath7 limit , for instance , is pushed up to @xmath14 .
finally , a @xmath286 or @xmath7 reach in the mass region @xmath9-@xmath10 requires an integrated luminosity of @xmath317 ( @xmath318 ) or @xmath319 ( @xmath320 ) with ( without ) cascade enhancement .
in this paper , we have systematically investigated the lhc phenomenology of neutral scalar pair production in the negative scenario of the type ii seesaw model . to achieve this goal
, we first examined the decay properties of the neutral scalars @xmath321 and found that the scalar self - couplings @xmath71 have a great impact on the branching ratios of @xmath0 . the coupling @xmath84 is important for tree - level decays of @xmath54 and @xmath40 , while one - loop - induced decays of @xmath54 further depend on @xmath322 and @xmath323 .
we found that the decay @xmath106 could dominate for @xmath324 with @xmath109 , while it can be neglected once @xmath96 is above the light scalar pair threshold @xmath325 .
moreover , the branching ratios of the decays @xmath127 can cross 3 orders of magnitude when varying the couplings @xmath71 , and there exist zero points for the @xmath88 , @xmath90 , and @xmath145 couplings .
the cross section of the drell - yan process @xmath24 for @xmath326 is much larger than that of the sm higgs pair production driven by gluon fusion . in this paper
, we studied the contributions to @xmath0 production from cascade decays of the charged scalars @xmath59 and @xmath23 .
there are actually three different states for the neutral scalar pair : @xmath11 , @xmath195 , and @xmath196 . here
, @xmath195 and @xmath196 can only arise from cascade decays of charged scalars , and their production rates always stay the same to a good approximation .
further , for a fixed value of @xmath50 , cascade enhancement is determined by the variables @xmath52 and @xmath51 . by tuning these two variables
, the associated production rate of @xmath11 can be maximally enhanced by about a factor of 3 , while those of the @xmath195 and @xmath196 pair production can reach the value of @xmath11 production through the pure drell - yan process .
we implemented detailed collider simulations of the associated @xmath11 production for three typical signal channels ( @xmath3 , @xmath4 , and @xmath19 with both @xmath75 s decaying leptonically ) .
the enhancement from cascade decays of charged scalars is quantified by a multiplicative factor @xmath238 . due mainly to a larger production rate , all three channels are more promising than the sm higgs pair case .
if there were no cascade enhancement , the @xmath7 mass reach of the @xmath3 , @xmath4 , and @xmath5 channels would be , respectively , @xmath8 , @xmath9 , and @xmath10 for lhc14@3000 .
the cascade enhancement pushes these limits up to @xmath12 , @xmath13 , and @xmath14 .
the @xmath3 and @xmath4 channels are more promising in the mass region below about @xmath284 , and the required luminosities for @xmath7 significance are @xmath327 and @xmath328 , respectively , at our benchmark point .
compared with these two channels , the @xmath5 channel is more advantageous in the relatively higher mass region @xmath9-@xmath14 , and the required luminosity for @xmath7 significance is about @xmath319 with maximal cascade enhancement .
needless to say , for the purpose of a full investigation on the impact of heavy neutral scalars on the sm higgs pair production , more sophisticated simulations are necessary .
we hope that this work may shed some light on further studies in both the phenomenological and experimental communities .
this work was supported in part by the grants no .
nsfc-11025525 , no .
nsfc-11575089 and by the cas center for excellence in particle physics ( ccepp )
. 000 z. l. han , r. ding , and y. liao , phys .
d * 91 * , 093006 ( 2015 ) [ arxiv:1502.05242 [ hep - ph ] ] .
g. aad _ et al . _ [ atlas collaboration ] , phys .
b * 726 * , 88 ( 2013 ) ; phys .
b * 734 * , 406 ( 2014 ) [ arxiv:1307.1427 [ hep - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , jhep * 1306 * , 081 ( 2013 ) [ arxiv:1303.4571 [ hep - ex ] ] .
g. aad _ et al . _ [ atlas collaboration ] , phys .
b * 716 * , 1 ( 2012 ) [ arxiv:1207.7214 [ hep - ex ] ] .
s. chatrchyan _ et al .
_ [ cms collaboration ] , phys .
b * 716 * , 30 ( 2012 ) [ arxiv:1207.7235 [ hep - ex ] ] .
t. plehn , m. spira , and p. m. zerwas , nucl .
b * 479 * , 46 ( 1996 ) [ erratum - ibid .
b * 531 * , 655 ( 1998 ) ] [ hep - ph/9603205 ] .
s. dawson , s. dittmaier , and m. spira , phys .
d * 58 * , 115012 ( 1998 ) [ hep - ph/9805244 ] .
a. djouadi , w. kilian , m. muhlleitner , and p. m. zerwas , eur .
j. c * 10 * , 45 ( 1999 ) [ hep - ph/9904287 ] .
u. baur , t. plehn , and d. l. rainwater , phys .
d * 67 * , 033003 ( 2003 ) [ hep - ph/0211224 ] .
e. asakawa , d. harada , s. kanemura , y. okada , and k. tsumura , phys .
d * 82 * , 115002 ( 2010 ) [ arxiv:1009.4670 [ hep - ph ] ] .
m. j. dolan , c. englert , and m. spannowsky , jhep * 1210 * , 112 ( 2012 ) [ arxiv:1206.5001 [ hep - ph ] ] .
a. papaefstathiou , l. l. yang , and j. zurita , phys .
d * 87 * , no . 1 , 011301 ( 2013 ) [ arxiv:1209.1489 [ hep - ph ] ] . f. goertz , a. papaefstathiou , l. l. yang , and j. zurita , jhep * 1306 * , 016 ( 2013 ) [ arxiv:1301.3492 [ hep - ph ] ] .
r. s. gupta , h. rzehak , and j. d. wells , phys .
d * 88 * , 055024 ( 2013 ) [ arxiv:1305.6397 [ hep - ph ] ] .
a. j. barr , m. j. dolan , c. englert , and m. spannowsky , phys .
b * 728 * , 308 ( 2014 ) [ arxiv:1309.6318 [ hep - ph ] ] .
d. de florian and j. mazzitelli , phys . rev .
* 111 * , 201801 ( 2013 ) [ arxiv:1309.6594 [ hep - ph ] ] .
m. j. dolan , c. englert , n. greiner , and m. spannowsky , phys .
lett . * 112 * , 101802 ( 2014 ) [ arxiv:1310.1084 [ hep - ph ] ] .
v. barger , l. l. everett , c. b. jackson , and g. shaughnessy , phys .
b * 728 * , 433 ( 2014 ) [ arxiv:1311.2931 [ hep - ph ] ] . c. englert , f. krauss , m. spannowsky , and j. thompson , phys .
b * 743 * , 93 ( 2015 ) [ arxiv:1409.8074 [ hep - ph ] ] .
t. liu and h. zhang , arxiv:1410.1855 [ hep - ph ] .
d. e. ferreira de lima , a. papaefstathiou , and m. spannowsky , jhep * 1408 * , 030 ( 2014 ) [ arxiv:1404.7139 [ hep - ph ] ] .
a. j. barr , m. j. dolan , c. englert , d. e. f. de lima , and m. spannowsky , arxiv:1412.7154 [ hep - ph ] . m. j. dolan , c. englert , and m. spannowsky , phys .
d * 87 * , no . 5 , 055002 ( 2013 ) [ arxiv:1210.8166 [ hep - ph ] ] .
a. arhrib , r. benbrik , c. -h .
chen , r. guedes , and r. santos , jhep * 0908 * , 035 ( 2009 ) [ arxiv:0906.0387 [ hep - ph ] ] .
n. craig , j. galloway , and s. thomas , arxiv:1305.2424 [ hep - ph ] .
b. hespel , d. lopez - val , and e. vryonidou , jhep * 1409 * , 124 ( 2014 ) [ arxiv:1407.0281 [ hep - ph ] ] .
g. d. kribs and a. martin , phys .
d * 86 * , 095023 ( 2012 ) [ arxiv:1207.4496 [ hep - ph ] ] .
j. cao , z. heng , l. shang , p. wan , and j. m. yang , jhep * 1304 * , 134 ( 2013 ) [ arxiv:1301.6437 [ hep - ph ] ] .
d. t. nhung , m. muhlleitner , j. streicher , and k. walz , jhep * 1311 * , 181 ( 2013 ) [ arxiv:1306.3926 [ hep - ph ] ] .
u. ellwanger , jhep * 1308 * , 077 ( 2013 ) [ arxiv:1306.5541 [ hep - ph ] ] .
b. bhattacherjee and a. choudhury , phys .
d * 91 * , 073015 ( 2015 ) [ arxiv:1407.6866 [ hep - ph ] ] .
n. d. christensen , t. han , and t. li , phys .
d * 86 * , 074003 ( 2012 ) [ arxiv:1206.5816 [ hep - ph ] ] .
l. wu , j. m. yang , c. p. yuan , and m. zhang , phys .
b * 747 * , 378 ( 2015 ) [ arxiv:1504.06932 [ hep - ph ] ] .
j. cao , d. li , l. shang , p. wu , and y. zhang , jhep * 1412 * , 026 ( 2014 ) [ arxiv:1409.8431 [ hep - ph ] ] . c. han , x. ji , l. wu , p. wu , and j. m. yang , jhep * 1404 * , 003 ( 2014 ) [ arxiv:1307.3790 [ hep - ph ] ] .
m. gouzevitch , a. oliveira , j. rojo , r. rosenfeld , g. p. salam , and v. sanz , jhep * 1307 * , 148 ( 2013 ) [ arxiv:1303.6636 [ hep - ph ] ] . j. m. no and m. ramsey - musolf , phys .
d * 89 * , 095031 ( 2014 ) [ arxiv:1310.6035 [ hep - ph ] ] .
r. grober and m. muhlleitner , jhep * 1106 * , 020 ( 2011 ) [ arxiv:1012.1562 [ hep - ph ] ] .
m. gillioz , r. grober , c. grojean , m. muhlleitner , and e. salvioni , jhep * 1210 * , 004 ( 2012 ) [ arxiv:1206.7120 [ hep - ph ] ] .
j. liu , x. p. wang , and s. h. zhu , arxiv:1310.3634 [ hep - ph ] .
a. arhrib , r. benbrik , r. b. guedes , and r. santos , phys .
d * 78 * , 075002 ( 2008 ) [ arxiv:0805.1603 [ hep - ph ] ] .
z. heng , l. shang , y. zhang , and j. zhu , jhep * 1402 * , 083 ( 2014 ) [ arxiv:1312.4260 [ hep - ph ] ] .
s. dawson , e. furlan , and i. lewis , phys .
d * 87 * , 014007 ( 2013 ) [ arxiv:1210.6663 [ hep - ph ] ] . c. y. chen , s. dawson , and i. m. lewis , phys .
d * 90 * , 035016 ( 2014 ) [ arxiv:1406.3349 [ hep - ph ] ] . c. o. dib , r. rosenfeld , and a. zerwekh , jhep * 0605 * , 074 ( 2006 ) [ hep - ph/0509179 ] .
b. yang , z. liu , n. liu , and j. han , eur .
j. c * 74 * , 3203 ( 2014 ) [ arxiv:1408.4295 [ hep - ph ] ] . c. y. chen , s. dawson , and i. m. lewis , phys .
d * 91 * , 035015 ( 2015 ) [ arxiv:1410.5488 [ hep - ph ] ] .
r. contino , m. ghezzi , m. moretti , g. panico , f. piccinini , and a. wulzer , jhep * 1208 * , 154 ( 2012 ) [ arxiv:1205.5444 [ hep - ph ] ] .
k. nishiwaki , s. niyogi , and a. shivaji , jhep * 1404 * , 011 ( 2014 ) [ arxiv:1309.6907 [ hep - ph ] ] .
n. liu , s. hu , b. yang , and j. han , jhep * 1501 * , 008 ( 2015 ) [ arxiv:1408.4191 [ hep - ph ] ] .
s. dawson , a. ismail , and i. low , phys .
d * 91 * , 115008 ( 2015 ) [ arxiv:1504.05596 [ hep - ph ] ] .
a. azatov , r. contino , g. panico , and m. son , phys .
d * 92 * , 035001 ( 2015 ) [ arxiv:1502.00539 [ hep - ph ] ] .
f. goertz , a. papaefstathiou , l. l. yang , and j. zurita , jhep * 1504 * , 167 ( 2015 ) [ arxiv:1410.3471 [ hep - ph ] ] .
a. pierce , j. thaler , and l. -t .
wang , jhep * 0705 * , 070 ( 2007 ) [ hep - ph/0609049 ] .
z. kang , p. ko , and j. li , arxiv:1504.04128 [ hep - ph ] . h. j. he , j. ren , and w. yao , arxiv:1506.03302 [ hep - ph ] .
j. baglio , a. djouadi , r. grber , m. m. mhlleitner , j. quevillon , and m. spira , jhep * 1304 * , 151 ( 2013 ) [ arxiv:1212.5581 [ hep - ph ] ] . s. dawson _
_ , arxiv:1310.8361 [ hep - ex ] .
t. p. cheng and l. f. li , phys .
d * 22 * , 2860 ( 1980 ) ; j. schechter and j. w. f. valle , phys .
d * 22 * , 2227 ( 1980 ) ; g. lazarides , q. shafi , and c. wetterich , nucl .
b * 181 * , 287 ( 1981 ) ; r. n. mohapatra and g. senjanovic , phys .
d * 23 * , 165 ( 1981 ) ; m. magg and c. wetterich , phys .
b * 94 * , 61 ( 1980 ) .
a. arhrib _ et al .
_ , phys .
d * 84 * , 095005 ( 2011 ) [ arxiv:1105.1925 [ hep - ph ] ] .
m. aoki , s. kanemura , m. kikuchi , and k. yagyu , phys .
d * 87 * , 015012 ( 2013 ) [ arxiv:1211.6029 [ hep - ph ] ] ; s. kanemura and k. yagyu , phys .
d * 85 * , 115009 ( 2012 ) [ arxiv:1201.6287 [ hep - ph ] ] .
m. aoki , s. kanemura , and k. yagyu , phys .
d * 85 * , 055007 ( 2012 ) [ arxiv:1110.4625 [ hep - ph ] ] .
m. chabab , m. c. peyranere , and l. rahili , phys .
d * 90 * , 035026 ( 2014 ) [ arxiv:1407.1797 [ hep - ph ] ] .
a. djouadi , phys . rep .
* 459 * , 1 ( 2008 ) [ hep - ph/0503173 ] .
fileviez perez , t. han , g. -y .
huang , t. li , and k. wang , phys .
d * 78 * , 015018 ( 2008 ) [ arxiv:0805.3536 [ hep - ph ] ] .
a. melfo , m. nemevsek , f. nesti , g. senjanovic , and y. zhang , phys .
d * 85 * , 055018 ( 2012 ) [ arxiv:1108.4416 [ hep - ph ] ] .
e. j. chun and p. sharma , phys .
b * 728 * , 256 ( 2014 ) [ arxiv:1309.6888 [ hep - ph ] ] .
a. g. akeroyd and h. sugiyama , phys .
d * 84 * , 035010 ( 2011 ) [ arxiv:1105.2209 [ hep - ph ] ] . j. alwall , m. herquet , f. maltoni , o. mattelaer , and t. stelzer , jhep * 1106 * , 128 ( 2011 ) [ arxiv:1106.0522 [ hep - ph ] ] ; j. alwall _ et al .
_ , jhep * 1407 * , 079 ( 2014 ) [ arxiv:1405.0301 [ hep - ph ] ] . t. sjostrand , s. mrenna , and p. z. skands , jhep * 0605 * , 026 ( 2006 ) [ hep - ph/0603175 ] . s. ovyn , x. rouby , and v. lemaitre , arxiv:0903.2225 [ hep - ph ] ; j. de favereau _ et al .
_ [ delphes 3 collaboration ] , jhep * 1402 * , 057 ( 2014 ) [ arxiv:1307.6346 [ hep - ex ] ] .
e. conte , b. fuks , and g. serret , comput .
commun . * 184 * , 222 ( 2013 ) [ arxiv:1206.1599 [ hep - ph ] ] .
g. aad _ et al . _ [ atlas collaboration ] , arxiv:0901.0512 [ hep - ex ] . s. dittmaier _ et al . _
[ lhc higgs cross section working group collaboration ] , arxiv:1101.0593 [ hep - ph ] .
g. cowan , k. cranmer , e. gross , and o. vitells , eur .
j. c * 71 * , 1554 ( 2011 ) [ erratum - ibid .
c * 73 * , 2501 ( 2013 ) ] [ arxiv:1007.1727 [ physics.data-an ] ] .
j. m. campbell and r. k. ellis , phys .
d * 62 * , 114012 ( 2000 ) [ hep - ph/0006304 ]
. m. cacciari , s. frixione , m. l. mangano , p. nason , and g. ridolfi , jhep * 0809 * , 127 ( 2008 ) [ arxiv:0804.2800 [ hep - ph ] ] ; p. brnreuther , m. czakon , and a. mitov , phys .
lett . * 109 * , 132001 ( 2012 ) [ arxiv:1204.5201 [ hep - ph ] ] . | this is a sequel to our previous work on lhc phenomenology of the type ii seesaw model in the nondegenerate case . in this work
, we further study the pair and associated production of the neutral scalars @xmath0 .
we restrict ourselves to the so - called negative scenario characterized by the mass order @xmath1 , in which the @xmath0 production receives significant enhancement from cascade decays of the charged scalars @xmath2 .
we consider three important signal channels@xmath3 , @xmath4 , @xmath5and perform detailed simulations .
we find that at the 14 tev lhc with an integrated luminosity of @xmath6 , a @xmath7 mass reach of @xmath8 , @xmath9 , and @xmath10 , respectively , is possible in the three channels from the pure drell - yan @xmath11 production , while the cascade - decay - enhanced @xmath0 production can push the mass limit further to @xmath12 , @xmath13 , and @xmath14 .
the neutral scalars in the negative scenario are thus accessible at lhc run ii . |
tunneling processes are very interesting nonperturbative phenomena .
one can find the example of such process already in quantum mechanics , where it causes the energy splitting .
there are tunneling processes in the field theory as well , for instance , pair production in external electromagnetic fields @xcite , decay of the false vacuum @xcite . in some spontaneously broken gauge theories
there are magnetic monopole and dyon solutions .
it is supposed that they can be produced in strong enough external electromagnetic fields . in the weak coupling regime
their masses are huge , and sizes are of order @xmath0 .
the probability of magnetic monopole pair production in constant magnetic field was calculated in the work of affleck and manton @xcite using the instanton method . in the work of bachas and porrati
@xcite the rate of pair production of open bosonic and supersymmetric strings in a constant electric field was calculated exactly . in the work of gorsky , saraikin and selivanov @xcite stringy deformed probability of monopole and w - boson
pair production was obtained quasiclassically .
it is possible for particles like monopole , dyon or w - boson to decay nonperturbatively in external fields @xcite .
monopole can also decay nonperturbatively in the external 2-form field @xcite .
in this paper we consider the process of magnetic monopole decay into electron and dyon , and w - boson decay into dyon and monopole using instanton method .
euclidean configuration corresponding to the monopole decay is represented on the fig.[classical traject ] .
exponential factor in the probability is given by the minimum of the electron , dyon and monopole total effective action ( see ( [ eff action ] ) ) .
this leads to ( [ classical action ] ) for the classical action . when monopole mass is equal to zero , one gets well - known result for exponential dependence of probability for pair production in external field
dyon is not pointlike particle , so to apply this method for calculation one must imply that the size of dyon is much smaller then the size of electron - dyon loop .
so , there is some condition imposed on the external field .
the approximation used in this case is analogous to the thin wall approximation in the problem of the false vacuum decay .
there is also condition of the dyon stability @xmath1 , where @xmath2 , @xmath3 , @xmath4 are masses of electron , dyon and monopole respectively .
contrary to spontaneous pair production , the process of the particle decay does nt occur for arbitrary masses .
it is shown that for the case when the relation ( [ condition on masses ] ) is fulfilled there are two negative eigenmodes , so , there is no decay at all .
coulomb corrections are taken into account similar to the work @xcite in the limit @xmath5 .
one can obtain schwingers @xcite result for probability of @xmath6 pair production summing the diagrams for vacuum amplitude similar to one represented on fig.[diagram ] : @xmath7 the probability connected with the amplitude is of the form @xmath8 where @xmath9 is the probability of pair production per unit time per unit volume . @xmath10 using the representation @xmath11 one obtains @xmath12 so , the probability for the process of @xmath6 pair production looks as follows @xmath13 the result obtained in the previous section is valid only for small coupling constant , since it does nt take into account self interaction of electron loop ( see fig.[photon interchange ] ) .
authors of @xcite have derived the expression for probability of monopole - antimonopole pair creation in external magnetic field using instanton method .
as we know the rate of decay follows from the imaginary part of the ground state energy @xmath14 where the energy @xmath15 is obtained from @xmath16}.\ ] ] using the wkb approximation one can do the integral and find the necessary probability per unit time per unit volume @xmath17 where @xmath18 is the classical action calculated on the instanton solution , and @xmath19 is one loop factor arising from the second variation of action .
we should note that operator of the second variation must have one and only one negative eigenvalue .
affleck and manton have calculated @xmath19 and @xmath18 with the following result @xmath20 where @xmath21 , @xmath22 are the mass and the charge of the monopole and @xmath23 is the strength of the external magnetic field .
+ the exponential factor in the probability can be immediately obtained by minimizing the effective action @xmath24 where l is the length of classical monopole trajectory in the magnetic field , q is the area restricted by this trajectory . note that recently instanton approach has been used for calculation of the probability of the pair production in the nonhomogeneous fields @xcite and in gravitational background @xcite .
now let us turn to the calculation of probability of the monopole decay in the external electric field .
it was argued @xcite that particles like monopole ( w - boson ) can decay in the external electric ( magnetic ) field into electron and dyon ( dyon and antimonopole ) , the junction such as ( dem ) naturally appears in the string theory . to find the probability we have to calculate the correction to green s function in the presence of electron and dyon .
green s function of free heavy monopole in external electro - magnetic field in euclidean time can be written as : @xmath25 taking into account one bounce correction we have @xmath26{rl } g(t,0;0,0)_{bounce}\sim&\int\mathcal{d}x\mathcal{d}z\exp\bigl(-m\int \sqrt{\dot{x}^{2}}dt - m_{d}\int \sqrt{\dot{z}^{2}}dt - m_{m}(t - h)\bigr . \\ \\
-&ie\int ( a_{\mu}(x)+a_{\mu}^{ext}(x))dx_{\mu}+ie\int ( a_{\mu}(z)+a_{\mu}^{ext}(z))dz_{\mu } \\ \\
\bigl.-&\int\frac{1}{4}f_{\mu\nu}^{2}d^{4}x\bigr ) , \end{array}\ ] ] where expression in the exponent is the well - known action for relativistic particles interacting with electromagnetic field .
this correction is the first term of expansion of full green s function @xmath27 , and as we know , twice the imaginary part of the mass yields the probability of the decay .
we consider the external field in minkowski space as constant electric field aligned in the @xmath28 direction .
the euclidean version of the field is of the form @xmath29 euclidean action for small coupling constant @xmath30 can be written as @xmath31 where @xmath32 are the lengths of the trajectories and @xmath33 is the area of the region restricted by trajectories .
we ll do path integral by steepest descent approximation .
at first we find solution to the classical equation of motion .
for the electron we have @xmath34 while for the dyon @xmath35 where we neglect interaction between particles , i.e. coulomb effects .
solutions to these equations are arcs of circles with radii @xmath36 , @xmath37 . introducing the angles that define the lengths of the arcs @xmath38 for electron and dyon respectively ( fig.[segment ] ) we have the action : @xmath26{rl } s_{cl}= & 2mr\theta_{1}+2m_{d}r\theta_{2}-2m_{m}r\sin\theta_{1}- \\ \\ - & ee\theta_{1}r^{2}+\frac{1}{2}eer^{2}\sin 2\theta_{1 }
-ee\theta_{2}r^{2}+\frac{1}{2}eer^{2}\sin 2\theta_{2}. \end{array}\ ] ] in order to find @xmath39 @xmath40 , @xmath41 , @xmath42 we minimize action taking into account that @xmath43 adding to the action term @xmath44 with the lagrange multiplier @xmath45 one gets @xmath46{rcl } 2mr-2m_{m}r\cos\theta_{1}+\lambda r\cos\theta_{1}-eer^{2}+eer^{2}\cos 2\theta_{1}&=&0 , \\ 2m_{d}r-\lambda r\cos\theta_{2}-eer^{2}+eer^{2}\cos 2\theta_{2}&=&0 , \\ r\sin\theta_{1}-r\sin\theta_{2}&=&0 , \\ 2m\theta_{1}-2m_{m}\sin\theta_{1}+\lambda\sin\theta_{1}-2ee\theta_{1}r+eer\sin 2\theta_{1}&=&0 , \\
2m_{d}\theta_{1}-\lambda\sin\theta_{2}-2ee\theta_{2}r+eer\sin 2\theta_{2}&=&0 .
\end{array } \right.\ ] ] solution to this system is @xmath46{rcl } r&=&\frac{m}{ee } , \\
r&=&\frac{m_{d}}{ee } , \\
\theta_{1}&=&\arccos\frac{m_{m}^{2}+m^{2}-m_{d}^{2}}{2mm_{m } } , \\
\theta_{2}&=&\arcsin\left(\frac{m}{m_{d}}\sin\theta_{1}\right ) , \end{array } \right . \label{rel btw ang}\ ] ] which amounts to the action @xmath26{rl } s_{cl}= & \frac{m^{2}}{ee}\arccos\frac{m_{m}^{2}+m^{2}-m_{d}^{2}}{2mm_{m}}+ \frac{m_{d}^{2}}{ee}\arccos\frac{m_{m}^{2}-m^{2}+m_{d}^{2}}{2m_{d}m_{m}}- \\ \\ - & \frac{mm_{m}}{ee}\sqrt{1-\left ( \frac{m_{m}^{2}+m^{2}-m_{d}^{2}}{2mm_{m}}\right)^{2}}. \end{array } \label{classical action}\ ] ] this result , of course , could be obtained in the way similar to that described in @xcite for the problem of induced false vacuum decay .
classical trajectories of electron , dyon and monopole are determined from the equilibrium condition in vertex via mechanical analogy ( @xmath47 are surface tensions and @xmath4 external force ) @xmath46{c } m\sin\theta_{1}=m_{d}\sin\theta_{2 } , \\ m_{m}-m\cos\theta_{1}-m_{d}\cos\theta_{1}=0 . \end{array } \right.\ ] ] it follows from these equations that if @xmath4 equals to zero then @xmath2 and @xmath3 will be equal to each other , i.e. creation of particles with different masses is accompanied by decay of some massive particle .
also we should note that such action for dyon and monopole implies that these particles have no size .
however , since these particles are not pointlike the approximation used works only for small enough external field .
the size of dyon ( @xmath48 ) should be much smaller then the size of electron - dyon loop ( @xmath49 ) , so the field should obey the condition @xmath50 also we should make some comment on the limit @xmath51 , which implies the condition @xmath52 . intuitively it seems to coincide with the result of circular symmetrical case of pair production .
the action for this case , as one can see from ( [ classical action ] ) , becomes @xmath53 which is the same as in the circular symmetric case .
let us make a short digression on the w - boson decay in the magnetic field .
it was argued that w - boson can decay nonperturbatively in external magnetic field into monopole and dyon as magnetic monopole in external electric field @xcite into electron and dyon .
however , since strong enough magnetic fields occur in cosmology more often then strong electric fields the formula for w - boson decay could be even more useful .
the problem of w - boson decay is identical to one of monopole decay since there is dual symmetry between electric and magnetic fields .
so , the result for the probability of w - boson decay reads as @xmath54 where @xmath55 , @xmath4 , @xmath3 are masses of w - boson , monopole and dyon respectively , @xmath23 is external magnetic field and @xmath22 is magnetic charge of monopole .
now we return to the monopole decay .
it was mentioned before that the classical trajectories of electron and dyon are the arcs of the circles . for the electron we have @xmath46{rcl } x_{0}^{cl}&=&r\sin ( 2\theta_{1}t-\theta_{1 } ) , \\ x_{1}^{cl}&=&-r\cos ( 2\theta_{1}t-\theta_{1})+r\cos\theta_{1}. \end{array } \right.\ ] ] operator of the second variation looks as follows @xmath56 where @xmath57 let us define @xmath58 then we get the equations for eigenfunctions and eigenvalues @xmath46{rcl } -\frac{1}{4\theta_{1}}f_{0}''+\frac{1}{4\theta_{1}}\cos2af_{0}''-\sin2af_{0}'+\frac{1}{4\theta_{1}}\sin2af_{1 } '' + \cos2af_{1}'-f_{1}'&=&\frac{\lambda}{ee } f_{0 } , \\ \\
\frac{1}{4\theta_{1}}\sin2af_{0}''+\cos2af_{0}'+f_{0}'-\frac{1}{4\theta_{1}}f_{1 } '' -\frac{1}{4\theta_{1}}\cos2af_{1}''+\sin2af_{1}'&=&\frac{\lambda}{ee } f_{1 } , \\ \\
-\frac{1}{2\theta_{1}}f_{2}''&=&\frac{\lambda}{ee}f_{2 } , \\
\\-\frac{1}{2\theta_{1}}f_{3}''&=&\frac{\lambda}{ee } f_{3}. \end{array } \right .
\label{equation second var}\ ] ] to solve these equations denote @xmath26{rcl } f = f_{0}+if_{1 } ,
\\ \bar{f}=f_{0}-if_{1 } , \end{array}\ ] ] and multiplying the second equation in ( [ equation second var ] ) by @xmath59 and adding to the first equation we get @xmath60 one can obtain complex conjugated equation by multiplying the second equation by @xmath59 and subtracting it from the first one . the shape of the equation tells us the form of the solution @xmath46{rcl } f&=&\mathrm{e}^{-ia}g , \\ g&=&g_{1}+ig_{2}. \end{array } \right.\ ] ] upon this substitution we have @xmath46{rcl } -\frac{1}{2\theta_{1}}g_{2}''-2\theta_{1}g_{2}&=&\frac{\lambda}{ee } g_{2 } , \\
\lambda g_{1}&=&0 , \end{array } \right.\ ] ] and @xmath46{rcl } f_{0}^{e}(t)&=&\left(a_{1}\cos\omega_{1 } t+b_{1}\sin\omega_{1 } t\right)\sin(2\theta_{1}t-\theta_{1 } ) , \\ f_{1}^{e}(t)&=&-\left(a_{1}\cos\omega_{1 } t+b_{1}\sin\omega_{1 } t\right)\cos(2\theta_{1}t-\theta_{1 } ) , \end{array } \right.\ ] ] @xmath61 similar manipulations for the dyon yield @xmath46{rcl } f_{0}^{d}(t)&= & \left(a_{2}\cos\omega_{2 } t+b_{2}\sin\omega_{2 } t\right)\sin(2\theta_{2}t-\theta_{2 } ) , \\ f_{1}^{d}(t)&=&\left(a_{2}\cos\omega_{2 } t+b_{2}\sin\omega_{2 } t\right)\cos(2\theta_{2}t-\theta_{2 } ) . \end{array } \right.\ ] ] @xmath62 there are zero eigenmodes for every particle corresponding to the translations .
the mode corresponding to the translation in @xmath28 direction is @xmath46{rcl } f_{0}&=&0 , \\
\end{array } \right.\ ] ] however , these perturbations for electron and dyon are not independent .
since worldline of monopole is straight with @xmath63 ( see fig.([allowed configuration ] ) ) , the unification of distorted trajectories should be closed .
the resulting electron - dyon loop should have vertexes with at @xmath63 . to satisfy the last condition one can require @xmath64 and using zero mode translate perturbed trajectory to @xmath63 . to satisfy the former condition one must require @xmath65 let us find solutions consistent with these restrictions .
for the solution with factor @xmath66 we get @xmath67 so , @xmath68 . for the solution with factor
@xmath69 we get @xmath70 that is @xmath71 .
finally we have @xmath46{rcl } f_{0}^{e , d}(t)&=&b_{1,2}\sin\pi t\sin(2\theta_{1,2}t-\theta_{1,2 } ) , \\
f_{1}^{e , d}(t)&=&\mp b_{1,2}\sin\pi t\cos(2\theta_{1,2}t-\theta_{1,2 } ) , \end{array } \right .
\label{sin solution}\ ] ] with eigenvalues @xmath72 and @xmath46{rcl } f_{0}^{e , d}(t)&=&a_{1,2}\cos 2\pi n t\sin(2\theta_{1,2}t-\theta_{1,2 } ) , \\
f_{1}^{e , d}(t)&=&\mp a_{1,2}\cos 2\pi n t\cos(2\theta_{1,2}t-\theta_{1,2 } ) , \end{array } \right . \label{cos solution}\ ] ] with eigenvalues @xmath73 it is obvious that first set of solutions already fulfills ( [ equal time diff ] ) , so , for this set we have independent perturbations for the electron and for the dyon .
the second set do not fulfills ( [ equal time diff ] ) for arbitrary @xmath74 and @xmath75 .
they must satisfy the condition @xmath76 thus @xmath77 it is quite difficult to find full quantitative expression for the determinant , but we can investigate it qualitatively .
we are interested in the presence of negative eigenmodes .
it is easy to see from ( [ cos eigenvalue ] ) that if @xmath78 , some eigenvalue becomes negative .
corresponding eigenfunctions are proportional to the classical solutions @xmath46{rcl } f_{0}^{e}&=&a m\sin(2\theta_{1}t-\theta_{1 } ) , \\ f_{1}^{e}&=&-a m\cos(2\theta_{1}t-\theta_{1 } ) , \end{array } \right.\ ] ] @xmath46{rcl } f_{0}^{d}&= & a m_{d}\sin(2\theta_{2}t-\theta_{2 } ) , \\ f_{1}^{d}&= & a m_{d}\cos(2\theta_{2}t-\theta_{2 } ) , \end{array } \right.\ ] ] and they are responsible for simultaneous increasing of the radii of the trajectories ( see fig.[radii increasing ] ) , where @xmath79 is the corresponding parameter .
similar negative eigenmode exists in the case of spontaneous pair production ( see for example @xcite ) .
one can calculate the action on perturbed configuration , expand it in power series by small deviation from classical radius and find that the quadratic correction is negative .
hence this perturbation is connected to the negative eigenmode .
indeed , correction to the action proportional to @xmath80 reads as @xmath81 since @xmath82 .
+ we can also consider the first set of solutions ( [ sin solution ] ) .
the solution corresponding to the particle with smaller mass , i.e. to the electron , for @xmath83 is @xmath46{rcl } f_{0}^{e}&=&b_{1}\sin\pi t\sin(2\theta_{1}t-\theta_{1 } ) , \\ f_{1}^{e}&=&-b_{1}\sin\pi t\cos(2\theta_{1}t-\theta_{1 } ) , \end{array } \right . \label{specific eigenmode}\ ] ] and @xmath84 it is obvious that @xmath85 when @xmath86 ( see fig.[pert of the electrons traject ] ) that is @xmath87 and the condition of dyon stability is @xmath88 one can also calculate correction to the action @xmath89 it is negative provided that condition on masses ( [ condition on masses ] ) is fulfilled . since the presence of the negative eigenmode connected to the changing of configuration s size do not depend on this condition
, we have two negative eigemodes in this case .
hence , there is no decay at all provided that condition ( [ condition on masses ] ) is true . in this case
the action on the euclidean configuration provides the nonperturbative renormalization of monopole s mass .
now we ll calculate the quantum corrections due to photon exchange between electron and dyon ( see fig.[coulomb ] ) . we know that @xmath90 and therefore using the relation above we obtain the contribution to the path integral from electromagnetic field @xmath91 we have to do the integral over the path which consists of trajectories of particles , schematically @xmath92 it is difficult to calculate this integral in general case hence we ll do it for the case @xmath5 , so we can consider trajectory of electron as semi circle and trajectory of dyon as a straight line .
using the dimensional regularization we get @xmath93 integration for dyon contribution gives @xmath94 and for electron - dyon part @xmath95 finally , collecting all together we find for the case @xmath96 the probability of monopole decay is of the form @xmath26{rl } \gamma\sim&\exp \biggl(\frac{m^{2}}{ee}\arccos\frac{m_{m}^{2}+m^{2}-m_{d}^{2}}{2mm_{m}}+ \frac{m_{d}^{2}}{ee}\arccos\frac{m_{m}^{2}-m^{2}+m_{d}^{2}}{2m_{d}m_{m}}- \\ & -\frac{mm_{m}}{ee}\sqrt{1-\left ( \frac{m_{m}^{2}+m^{2}-m_{d}^{2}}{2mm_{m}}\right)^{2}}+ \frac{3e^{2}}{16}\biggr ) , \label{exp for prob } \end{array}\ ] ] while for the case @xmath97 similar exponential factor corresponds to the monoples mass renormalization .
@xmath98 for the case @xmath99 , i.e. for bps case the negative eigenmode ( [ specific eigenmode ] ) becomes zero eigenmode and the probability for the decay reads as @xmath26{rl } \gamma\sim&\exp \biggl(\frac{\pi m^{2}}{2 ee}+ \frac{(m_{m}+m)^{2}}{ee}\arccos\left(\frac{1}{2}\left(1+\frac{m^{2}}{m_{m}^{2}}\right)^{-\frac{1}{2}}\right ) -\frac{mm_{m}}{ee}+ \frac{3e^{2}}{16}\biggr ) .
\label{exp for prob } \end{array}\ ] ]
in this paper the process of monopole decay in external electric field was investigated .
the probability of monopole decay and w - boson decay up to the exponential accuracy has been found .
it was shown that the result obtained is in agreement with one for the probability of pair production .
the the coulomb corrections were calculated for the process of monopole decay in external electric field .
the determinant was investigated qualitatively , calculation of full expression for the determinant will be made elsewhere .
it was found that monopole decay does nt occur for arbitrary masses of particles involved .
the relation between masses for the decay to exist was found ( [ condition on masses ] ) .
note that the well - known process of induced decay of the false vacuum @xcite satisfies this condition .
it is obvious that the probability of such processes is exponentially suppressed , but they could have some applications in cosmology .
the natural generalization of the problem would involve the temperature effects and the stringy corrections .
10 j. schwinger , phys .
82 , 664 ( 1951 ) .
affleck , o. alvarez , n.s .
manton , nucl .
b197 , 509 , ( 1982 ) .
affleck , n.s .
manton , nucl .
b194 , 38 , ( 1982 ) .
voloshin , i. yu .
kobzarev , l.b .
okun , yad .
20 , 1229 ( 1974 ) c.g .
callan , s. coleman , phys .
d16 , 1762 , ( 1977 ) s. coleman , phys .
d15 , 2929 ( 1977 ) a.s .
gorsky , k.a .
saraikin and k.g .
selivanov , nucl.phys . ,
b628 , ( 2002 ) , 270 - 294 , hep - th/0110178 . k. selivanov and m. voloshin , zhetp lett,42 ( 1985 ) 422 .
c. bachas and m. porrati , phys .
b 296 , 77 ( 1992 ) , hep - th/9209032 .
gerald v. dunne , c. schubert `` worldline instantons and pair production in inhomogeneous fields '' , hep - th/0507174 a.s .
gorsky , k.g .
selivanov `` junction and the fate of branes in external fields '' , nucl.phys .
b571 ( 2000 ) 120 - 134 , hep - th/9904041 b. pioline , j. troost `` schwinger pair production in @xmath100 '' , jhep 0503 ( 2005 ) 043 , hep - th/0501169 | the possibility of the magnetic monopole decay in the constant electric field is investigated and the exponential factor in the probability is obtained . corrections due to coulomb interaction are calculated . the relation between masses of particles for the process to exist
is obtained .
itep - th-54/05 1 cm a.k .
monin _ institute of theoretical and experimental physics , + b.cheremushkinskaya 25 , moscow , russia _ _
moscow state university , physics department , + vorobyevy gory 1 , moscow , russia _ e - mail : [email protected] 1.3 cm |
the viability of the bayesian approach to inverse problems was established in the pioneering text @xcite which , in particular , demonstrated the potential for markov chain monte carlo ( mcmc ) methods in this context . nonetheless , the high dimensional nature of the unknown , often found from discretizing a field , leads to difficult problems in the design of proposals which are cheap to implement , yet which mix efficiently .
one recent approach to tackle these problems has been the development of algorithms with mesh - free mixing times , such as those highlighted in @xcite ; these non - standard mcmc algorithms avoid the unnecessary penalties incurred by naive proposals related to exploration of the part of the parameter space dominated by the prior . nonetheless , in the large dataset or small
observational noise regimes , one is still confronted with an inference problem in high dimensions which , whilst of smaller order than the dimension of the pde solver , exhibits wide variations in scales in different coordinates of the parameterizations , leading to substantial challenges for algorithmic tuning .
a different approach , which we will adopt here , involves smc samplers @xcite .
these are particle methods which , in the context of bayesian inverse problems , build an approximation to a sequence of measures which interpolate from the prior to the posterior ; the sequential nature of the approximation allows for adaptation of the particle distribution and weights from the ( typically simple ) prior to the ( potentially very complex ) posterior .
recent work in the context of inverse problems @xcite has shown how , by using the aforementioned dimension - independent mcmc methods within smc , it is possible to construct algorithms which combine the desirable dimension - independent aspects of novel mcmc algorithms with the desirable self - adaptation of particle methods .
this combination is beneficial for complex posteriors such as those arising in the large dataset or small noise regimes ; in particular the computational results in @xcite demonstrate an order of magnitude speed - up of these new smc methods over the mcmc methods highlighted in @xcite , within the context of the inverse problem for the initial condition of the navier - stokes equation .
furthermore , recent works @xcite have shown that important aspects of this smc algorithm for inverse problems , such as adaptation , tempering and parallelization , have the potential to provide effective methods even for high - dimensional inverse problems
. the contributions of this article are three - fold : 1 . a computational study of smc methods for a class of bayesian inverse problems which arise in applications such as hydrology @xcite , and are more challenging to fit , in comparison to the navier - stokes inverse problem which was the focus of the development of novel smc methods in @xcite ; furthermore , with modification of the measurement set - up , the inverse problems considered also find application in medical imaging problems such as eit @xcite .
an enhancement of the class of smc methods introduced in @xcite which leads to greater efficiency and , in particular , the ability to efficiently solve the elliptic inverse problems which are the focus of this paper .
3 . a proof of the fact that these smc algorithms have monte carlo convergence rates which are mesh - independent and
, indeed converge in the function space setting .
this complements related theoretical work @xcite which establishes mesh - independence mixing for the novel mcmc methods which are used as proposal kernels within the smc approach of @xcite which we build upon here .
this article is structured as follows . in section [ 3dinv ]
we describe the bayesian model and associated pde . in section [ sec
: smc ] our computational procedure is outlined , along with our theoretical results . in section [ sec : numerics ] we present our numerical results .
the article is concluded in section [ sec : summ ] with a discussion of areas for future work .
consider two normed linear spaces , @xmath3 and @xmath4 , corresponding to the state space of the parameters ( @xmath5 ) and observations ( @xmath6 ) respectively .
we will observe data at spatial locations @xmath7 and we denote the observation at location @xmath8 as @xmath9 .
let @xmath10 and , for each @xmath11 , let @xmath12 be a random variable of zero mean ; then we will be concerned with models of the form : @xmath13 here @xmath14 is an underlying system behaviour for a given parameter @xmath15 , and @xmath16 expresses measurement ( and indeed sometimes model error ) at location @xmath8 . in our context , @xmath17 is associated to the solution of a pde , with parameter @xmath15 .
we are interested in drawing inference on @xmath15 , given a prior distribution on @xmath5 , conditional upon observing realizations of @xmath9 for a set of points @xmath18 , with @xmath19 .
this is the framework of our bayesian inverse problem . in subsection [ ssec :
fm ] we define the forward model , and in subsection [ ssec : modelk ] we describe prior modelling on our unknown .
subsection [ ssec : bayesian ] shows that the posterior distribution is well - defined and states a key property of the log - likelihood , used in what follows . in this paper , we focus on the general scenario where the forward map @xmath17 is described by an elliptic pde .
in particular , we work with a problem of central significance in hydrology , namely the estimation of subsurface flow from measurements of the pressure ( hydraulic head ) at certain locations @xmath8 in the domain of interest . the pressure and velocity are linked by darcy s law in which the subsurface permeability appears as a parameter ; estimating it is thus a key step in predicting the subsurface flow . in this subsection
we define the forward map from permability to pressure space . in detail , we consider the @xmath20dimensional cube @xmath21^d$ ] as our domain , in both the cases @xmath22 define a mapping @xmath23 , denoting pressure ( or hydraulic head ) , @xmath24 , denoting a quantity proportional to velocity , and @xmath25 , denoting permeability ( or hydraulic conductivity ) of soil @xcite .
the behaviour of the system is described through the elliptic pde : [ eq2 ] @xmath26 equation ( [ eq2]a ) is darcy s law and contains the permeability @xmath15 ( for us the key parameter ) ; equation ( [ eq2]b ) expresses continuity of mass and here @xmath27 is assumed known and characterizes the source / sink configuration ; in equation ( [ eq2]c ) @xmath28 is the boundary of the domain , and we are thus assuming a homogeneous boundary condition on the boundary pressure other boundary conditions , specifying the flux , are also possible .
together equations define an elliptic pde for pressure @xmath29 .
if @xmath15 is in @xmath30 and lower bounded by a positive constant @xmath31 a.e . in @xmath32 then , for every @xmath33 , there is a unique solution @xmath34 to the pde satisfying @xmath35 see @xcite and the references therein . in this setting ,
the forward map @xmath36 is well - defined and thus corresponds to solution of the elliptic pde for a given permeability field @xmath15 .
a typical choice of the source / sink function @xmath37 is @xmath38 the set of points @xmath39 denote the known position of sources or sinks , and the signs of each @xmath40 determine whether one has a source or sink at @xmath41 @xcite .
we note that the cleanest setting for the mathematical formulation of the problem requires @xmath33 and , in theory , will require mollification of the dirac s at each @xmath41 ; in practice this modification makes little difference to the inference .
we describe the modelling of @xmath15 in three dimensions ; simplification to the two dimensional setting is straightforward .
we begin by expressing the unknown model parameter as a fourier series : @xmath42 here we have scaled fourier coefficients @xmath43 and the real coefficients @xmath44 , complex coefficients @xmath45 ( satisfying @xmath46 ) and real - valued @xmath47 function @xmath48 will be chosen to enforce the mathematically necessary ( and physically sensible ) positivity restriction @xmath49 the use of fourier series in principle enables the representation of arbitrary functions in @xmath50 by use of periodic extensions
. however we will impose a rate of decay on the @xmath44 , in order to work in the setting of inversion for this problem , as developed in @xcite ; this rate of decay will imply a certain degree of smoothness in the function @xmath51 noting that the functions @xmath52 have @xmath30 norm equal to one , we can place ourselves in the setting of @xcite by assuming that , for some @xmath53 , @xmath54 , @xmath55 we choose @xmath56 of the form @xmath57 and then impose @xmath58 in dimension @xmath59 or @xmath60 in dimension @xmath61 .
given this set - up , we need to find a suitable prior for @xmath15 , so that the forward model @xmath14 is almost - surely well - defined , as well as reflecting any prior statistical information we may have .
there are several widely adopted approaches in the literature for prior parameterization of the permeability , the most common being the log - normal choice ( see @xcite for details and , for example , @xcite for a recent application ) , widely adopted by geophysicists , and the uniform case @xcite which has been succesfully adopted in the computational mathematics literature , building on earlier work of schwab in uncertainty quantification @xcite .
we work with the uniform priors popularized by schwab : we choose @xmath62}$ ] in the representation for @xmath15 , resulting in a pushforward measure @xmath63 on @xmath15 as in @xcite .
we let @xmath3 denote the separable banach space found from the closure , with respect to the @xmath64 norm , of the set of functions used in the representation of @xmath65 .
then @xmath63 is viewed as a measure on @xmath3 ; see @xcite for further details .
once the parameters @xmath56 are chosen to satisfy , the mean function @xmath66 can be chosen to ensure that there is @xmath67 such that @xmath65 is in @xmath30 and satisfies almost surely with respect to the prior @xmath63 on function @xmath15 .
we observe the pressure at certain locations , the set of which is denoted as @xmath68 .
we will suppose that for each @xmath69 and independently , @xmath70 , where @xmath71 is the normal distribution of mean 0 and known variance @xmath72 .
then the log - likelihood is , up to an irrelevant additive constant , given by @xmath73 along with the prior modelling in subsection [ ssec : modelk ] , this defines a scenario so that the forward model @xmath74 is , almost - surely , well - defined and , in fact , lipschitz . as in @xcite
we may then define a posterior @xmath75 on @xmath15 which has density with respect to @xmath63 given by : @xmath76 exploring the posterior distribution @xmath75 is the objective of the paper . in doing so , the following fact will be relevant ; it is easily established by using the fact that holds almost surely for @xmath77 , together with the bound on the solution of the elliptic pde given in .
[ lem : note ] there is a constant @xmath78 such that @xmath79 almost surely for @xmath80 we finish by noting that , in algorithmic practice , it is typically necessary ( see , however , @xcite in the context of mcmc ) to apply a spectral truncation : @xmath81 where @xmath82 is a truncation parameter . having defined the desired parameterization of @xmath15 , we consider the truncated vector of fourier coefficients @xmath83 as the object of inference in practice .
in this section we describe the application of smc methods to bayesian inversion .
subsection [ ssec : smc ] contains an explanation of the basic methodology and statement of a basic ( non - adaptive ) algorithm .
subsection [ ssec : t ] contains statement and proof of a convergence theorem for the basic form of the algorithm , notable because it applies in infinite dimensional spaces . in subsection
[ ssec : a ] we describe an adaptive version of the smc algorithm , which we use in practice .
let @xmath84 denote a measure space and @xmath63 a probability measure on that space .
we wish to sample from a target probability measure @xmath85 on @xmath84 , which has density with respect to @xmath63 known up to a normalizing constant : @xmath86 we introduce a sequence of `` bridging '' densities which enable us to connect @xmath63 to @xmath85 : @xmath87 where @xmath88 we refer to the @xmath89 as _ temperatures_. we let @xmath90 denote the probability measure with density proportional to @xmath91 with respect to @xmath63 . assuming that @xmath92 is finite @xmath63 almost surely we obtain @xmath93 we note that the assumption on @xmath94 being finite is satisfied for our elliptic inverse problem ; see lemma [ lem : note ] .
although @xmath95 may be far from @xmath63 , careful choice of the @xmath96 can ensure that @xmath97 is close to @xmath98 allowing gradual evolution of approximation of @xmath63 into approximation of @xmath85 .
other choices of bridging densities are possible and are discussed in e.g. @xcite .
let @xmath99 denote the sequence of ( nonlinear ) maps on measures found by applying bayes s theorem with likelihood proportional to @xmath100 and let @xmath101 be a sequence of markov kernels ( and equivalently , for notational convenience , the resulting linear maps on measures ) with invariant measure @xmath102 .
we define @xmath103 to be the nonlinear maps on measures found as @xmath104 .
explicitly we have , for each @xmath105 and any probability measure @xmath106 on @xmath107 : @xmath108 where we use the notation @xmath109 and @xmath110 .
it then follows that @xmath111 the standard smc algorithm is described in figure [ tab : smc ] .
it involves a population of @xmath112 particles evolving with @xmath113 . with no resampling ,
the algorithm coincides with annealed importance sampling as in @xcite .
with resampling at every step ( i.e. the case @xmath114 , where @xmath115 denotes the cut - off point for the effective sample size ( ees ) ) we define the empirical approximating measures by the iteration @xmath116 here @xmath117 ' '' '' _ _ 1 .
sample @xmath118 i.i.d . from @xmath63 and define the weights @xmath119 for @xmath120 set @xmath121 and @xmath122 .
2 . for each @xmath123 set @xmath124 and sample @xmath125 from @xmath126 calculate the normalized weights @xmath127 3 .
calculate the effective sample size ( ess ) : @xmath128 if @xmath129 : + resample @xmath130 according to the normalized weights @xmath131 ; + re - initialise the weights by setting @xmath132 for @xmath133 ; + let @xmath130 now denote the resampled particles .
if @xmath134 set @xmath135 and return to step 1 ; otherwise stop . ' '' '' the issue of dimensionality in smc methods has attracted substantial attention in the literature @xcite . in this section , using a simple approach for the analysis of particle filters which is clearly exposed in @xcite , we show that for our smc method it is possible to prove dimension - free error bounds . whilst the theoretical result in this subsection is not entirely new ( similar results follow from the work in @xcite ) , a direct and simple proof is included for non - specialists in smc methods , to highight the utility of smc methods in inverse problems , and to connect with related recent results in dimension - independent mcmc , such as @xcite , which are far harder to establish .
we will consider the algorithm in figure [ tab : smc ] with @xmath136 , so one resamples at every time step ( and this is multinomially ) .
note that then , for @xmath137 , at the end of each step of the algorithm the approximation to @xmath90 is given by @xmath138 which follows from the algorithm in figure [ tab : smc ] with @xmath136 or , equivalently , . throughout
, we will assume that there exists a @xmath139 such that for each @xmath137 and any @xmath140 @xmath141 we note that this holds for the elliptic inverse problem from the previous section , when the uniform prior @xmath63 is employed ; see lemma [ lem : note ] .
let @xmath142 denote the collection of all probability measures on @xmath107 .
let @xmath143 and @xmath144 denote two possibly random elements in @xmath142 , and @xmath145 expectation w.r.t .
@xmath146 we define the distance between @xmath147 by @xmath148 where the supremum is over all @xmath149 with @xmath150 this definition of distance is indeed a metric on the space of random probability measures ; in particular it satisfies the triangle inequality . in the context of smc the randomness underlying the approximations comes from the various sampling operations within the algorithm .
we have the following convergence result for the smc algorithm . [ th39 ]
assume and consider the smc algorithm with @xmath114 .
then , for any @xmath137 , @xmath151 for @xmath152 the result holds via lemma [ lem : tech_lem2 ] . for @xmath153 , we have , by the triangle inequality , lemma [ lem : tech_lem ] and lemma [ lem : tech_lem2 ] ( which may be used by the conditional independence structure of the algorithm )
, @xmath154 iterating gives the desired result .
[ rem : fil ] this theorem shows that the sequential particle filter actually reproduces the true posterior distribution @xmath155 , in the limit @xmath156 .
we make some comments about this .
* the measure @xmath155 is well - approximated by @xmath157 in the sense that , as the number of particles @xmath156 , the approximating measure converges to the true measure .
the result holds in the infinite dimensional setting . as a consequence
the algorithm as stated is robust to finite dimensional approximation . * in principle the theory applies even if the markov kernel @xmath158 is simply the identity mapping on probability measures .
however , moving the particles according to a non - trivial @xmath90-invariant measure is absolutely essential for the methodology to work in practice .
this can be seen by noting that if @xmath158 is indeed taken to be the identity map on measures then the particle positions will be unchanged as @xmath113 changes , meaning that the measure @xmath155 is approximated by weighted samples ( almost ) from the prior , clearly undesirable in general .
* the mcmc methods in @xcite provide explicit examples of markov kernels with the desired property of preserving the measures @xmath90 , including the infinite dimensional setting .
* in fact , if the markov kernel @xmath158 has some ergodicity properties then it is sometimes possible to obtain bounds which are _ uniform _ in @xmath29 ; see @xcite .
[ lem : tech_lem ] assume .
then , for any @xmath105 and any @xmath159 , @xmath160 for any measurable @xmath161 we have @xmath162(f ) & = \frac{1}{\mu(\ell_{n-1})}[\mu-\nu](\ell_{n-1}k_n(f))\\ & \quad\quad\quad\quad\quad + \frac{\nu(\ell_{n-1}k_n(f))}{\mu(\ell_{n-1})\nu(\ell_{n-1 } ) } [ \nu-\mu](\ell_{n-1}).\end{aligned}\ ] ] so we have , by minkowski , @xmath163(f)|^2]^{1/2 } & \leq \mathbb{e}^{\omega}\big[\big|\frac{1}{\mu(\ell_{n-1})}[\mu-\nu](\ell_{n-1}k_n(f))\big|^2\big]^{1/2}\\ & \quad\quad\quad\quad\quad+ \mathbb{e}^{\omega}\big[\big|\frac{\nu(\ell_{n-1}k_n(f))}{\mu(\ell_{n-1})\nu(\ell_{n-1 } ) } [ \nu-\mu](\ell_{n-1})]\big|^2\big]^{1/2}.\end{aligned}\ ] ] note that the ratio @xmath164 is an expectation of @xmath37 and is hence bounded by @xmath165 in modulus , if @xmath166 then using the fact that @xmath167 and @xmath168 ( see ) we deduce that @xmath163(f)|^2]^{1/2 } & \leq \frac{1}{\kappa^2}\mathbb{e}^{\omega}\big[\big|[\mu-\nu](\ell_{n-1}k_n(f)\kappa)\big|^2\big]^{1/2}\\ & \quad\quad\quad\quad\quad+ \frac{1}{\kappa^2}\mathbb{e}^{\omega}\big[\big|[\nu-\mu](\ell_{n-1}\kappa)]\big|^2\big]^{1/2}.\end{aligned}\ ] ] using the fact that @xmath167 and @xmath169 , with the first following from together with the markov property for @xmath158 , taking suprema over @xmath37 completes the proof .
[ lem : tech_lem2 ] the sampling operator satisfies @xmath170 let @xmath85 be an element of @xmath171 and @xmath172 a set of i.i.d
. samples with @xmath173 ; the randomness entering the probability measures is through these samples , expectation with respect to which we denote by @xmath174 in what follows .
then @xmath175 and , defining @xmath176 , we deduce that @xmath177 it is straightforward to see that @xmath178 furthermore , for @xmath179 , @xmath180 it follows that , for @xmath179 , @xmath181 since the result is independent of @xmath85 we may take the supremum over all probability measures and obtain the desired result . in practice
, the smc samplers algorithm requires the specification of @xmath182 as well as any parameters in the mcmc kernels .
as demonstrated in @xcite , the theoretical validity of which is established in @xcite , these parameters may be set on the fly .
first , we focus on the specification of the sequence of distributions . given step @xmath183 and @xmath184
, we select the next target density by adapting the temperatures to a required value for the effective sample size ( ess ) statistic as in @xcite ( see also @xcite for an alternative procedure ) .
so , for a user - specified threshold @xmath185 , we choose @xmath186 as the solution of @xmath187 .
one can use an inexpensive bisection method to obtain @xmath96 .
second , we turn to the specification of the mutation kernels @xmath158 .
several options are available here , but we will use reflective random walk metropolis proposals on each univariate component , conditionally independently .
we will adapt the random - move proposal scales , @xmath188 , with @xmath189 the co - ordinate and @xmath113 the time index .
a simple choice would be to tune @xmath188 to the marginal variance along the @xmath189-th co - ordinate ; since this is analytically unavailable we opt for the smc estimate at the previous time - step .
thus , we set @xmath190 where @xmath191 is a global scaling parameter . for @xmath191 itself , we propose to modify it based on the previous average acceptance rate over the population of particles ( denoted @xmath192 ) , to try to have average acceptance rates in a neighbourhood of 0.2 ( see e.g. @xcite and the references therein for a justification ) .
our adaptive strategy works as follows ; @xmath193 thus , we scale @xmath194 upwards ( downwards ) if the last average acceptance rate went above ( below ) a predetermined neighbourhood of 0.2 .
this approach is different to the work in @xcite .
in addition , one can synthesize a number , say @xmath195 , of baseline mcmc kernels , to obtain an overall effective one with good mixing ; this is a new contribution relative to @xcite . to adapt @xmath195 ,
we follow the following heuristic ; we propose to select @xmath195 using @xmath196 , with @xmath123 being a global parameter .
the intuition is that for random - walk - type transitions of increment with small standard deviation @xmath197 , one needs @xmath198 steps to travel distance @xmath199 in the state - space .
a final modification for practical computational reasons is that we force @xmath195 steps to lie within a predetermined bounded set , i.e. @xmath200 $ ] .
the adaptive smc algorithm works as in figure [ tab : smc ] , except in step 1 , before simulation from @xmath158 is undertaken , our adaptive procedure is implemented .
then one may resample ( or not ) and then move the samples according to @xmath158 .
in addition , the algorithm will run for a random number of time steps and terminate when @xmath201 ( which will happen in a finite number of steps almost surely ) .
in this section , we describe the details of our implementation ( section [ ssec : i ] ) , describe the objects of inference ( section [ ssec : oii ] ) and give our results in 2d ( section [ ssec:2d ] ) and 3d ( section [ ssec:3d ] ) .
the software used in our experiments has been implemented in c++ for the gnu@xmath202linux platform .
we used the libmesh library for finite elements computation @xcite , we used the fast fourier transform for rapid evaluation of the sum in @xmath65 at pre - determined grid - points in @xmath203 and we exploited parallel computation wherever possible , for which we used the mpi libraries .
our experiments were run on a computer server with 23 `` intel(r ) xeon(r)cpu x7460 @2.66ghz '' processors , each with 2 cores ; 50 gb memory and running `` redhat linux version 2.6.18-194.el5 '' operating system .
the experiments discussed in this paper used 20 processors .
all the colour plots of random fields ( e.g. permeability fields ) have been prepared using the rainbow color scheme from the r programming language / environment .
the scheme quantizes the hue quantity of hsv ( hue saturation value ) triplet of a pixel .
our level of quantization is selected to be 256 ( 8 bits ) , with the hue range of @xmath204 $ ] , hence we normalize the random fields to this range and quantize to 8 bits to get the hue value for a pixel .
saturation and value were taken to be 1 .
all images were computed using @xmath205 equi - spaced point evaluations from the respective random fields .
the work in @xcite investigates the performance of the bayesian approach for our elliptic inverse problem and gives sufficient conditions under which posterior consistency holds .
posterior consistency is concerned with `` ball probabilities '' of type @xmath206 where @xmath207 and @xmath208 is the @xmath209 neighbourhood of the true value of @xmath15 .
one way to check such a result numerically is to use the posterior estimates obtained via our method .
the estimated ball probabilities are computed as follows : @xmath210 although not all the conditions in @xcite required for posterior consistency to hold are fulfilled , we will nonetheless empirically investigate such a consistency property .
this also provides a severe test for the smc method since it implies posterior measures in the large dataset limit .
we consider equation in dimension @xmath61 and with source and sinks as specified in . our goal is to construct a sequence of posterior estimates , corresponding to increasing number of observations in order to numerically illustrate posterior consistency .
table [ tab_2d_param ] shows the parameters used in our experiments .
.parameter values of used for the 2d experiments . between 5 and 1000 steps are allowed for the iterates of the mcmc kernels .
the frequency cutoff determines the level of discretization of the permeability field .
finite elements d.o.f .
denotes the number of finite elements used in the numerical solution of the elliptic pde , higher values indicate better approximation at the expense of computational resources . for @xmath211
see . [ cols="<,^",options="header " , ] in figure [ smc_performance ] , we consider the performance of our smc algorithm in this very challenging scenario . in figure [ smc_performance ] ( a )
, we can see the average acceptance rates of the mcmc moves , over the time parameter of the smc algorithm .
we can observe that these acceptance rates do not collapse to zero and are not too far from 0.2 .
this indicates that the step - sizes are chosen quite reasonably by the adaptive smc algorithm and the mcmc kernels have some mixing ability . in figure [ smc_performance ] ( b ) , we can see the number of mcmc iterations that are used per - particle over the time parameter of the smc algorithm .
we can observe , as one might expect , that as the target distribution becomes more challenging , the number of mcmc steps required grows .
figure [ smc_performance ] indicates reasonable performance of our smc algorithm . in terms of inference ,
the posterior density estimates are shown in figure [ density ] .
recall that the priors are uniform .
these estimates indicate a clear deviation from the prior specification , illustrating that the data influence our inference significantly .
this is not obvious , and establishes that one can hope to use this bayesian model in real applications . 2.9 in 2.5 in
in this article we have presented an smc method for bayesian inverse problems and applied it to a particular elliptic pde inversion ; the methodology , however , is transferable to other pde inverse problems .
simulations demonstrated both the feasability of the smc method for challenging infinite dimensional inversion , as well as the property of posterior contraction to the truth .
in addition to simulations , we have provided a straightforward proof of the fact that smc methods are robust to the dimension of the problem .
there are several avenues for future research .
firstly , our error bounds explode w.r.t . the time parameter .
it is of interest to find realistic conditions for which this is not the case ( for instance the bounds in @xcite have assumptions which either do not hold or are hard to verify ) .
secondly , a further algorithmic innovation is to use multi - level monte carlo method as in @xcite , within the smc context ; this is being considered in @xcite . and
finally it is of interest to consider the use of these methods to solve other bayesian inference problems . | in this article we consider a bayesian inverse problem associated to elliptic partial differential equations ( pdes ) in two and three dimensions .
this class of inverse problems is important in applications such as hydrology , but the complexity of the link function between unknown field and measurements can make it difficult to draw inference from the associated posterior . we prove that for this inverse problem a basic smc method has a monte carlo rate of convergence with constants which are independent of the dimension of the discretization of the problem ; indeed convergence of the smc method is established in a function space setting .
we also develop an enhancement of the sequential monte carlo ( smc ) methods for inverse problems which were introduced in @xcite ; the enhancement is designed to deal with the additional complexity of this elliptic inverse problem . the efficacy of the methodology , and its desirable theoretical properties , are demonstrated on numerical examples in both two and three dimensions . + * keywords * : inverse problems , elliptic pdes , groundwater flow , adaptive smc , markov chain monte carlo . * sequential monte carlo methods for bayesian elliptic inverse problems * by alexandros beskos@xmath0 , ajay jasra@xmath1 , ege a. muzaffer@xmath1 & andrew m. stuart@xmath2 @xmath0department of statistical science , university college london , london , wc1e 7hb , uk .
+ e-mail:`[email protected] ` + @xmath1department of statistics & applied probability , national university of singapore , singapore , 117546 , sg .
+ e-mail:`[email protected] , [email protected] ` + @xmath2department of mathematics , university of warwick , coventry , cv4 7al , uk .
+ e-mail:`[email protected] ` |
for hamiltonian systems , the persistence of invariant tori influences the global properties of the dynamics .
the study of the break - up of invariant tori is thus an important issue to understand the onset of chaos . for two degrees of freedom ,
there are several numerical methods to determine the threshold of the break - up of invariant tori : for instance , greene s criterion @xcite , obstruction method @xcite , converse kam @xcite , frequency map analysis @xcite , or renormalization - group methods @xcite .
+ in this article , we propose to compute this threshold for a one - parameter family of hamiltonians with three degrees of freedom and for a specific frequency vector , by two techniques : by frequency map analysis and by renormalization .
the frequency map analysis is valid for any dimension , and has been applied to systems with a large number of degrees of freedom @xcite .
the set - up of renormalization - group transformations is also possible for any dimensions in the framework of ref .
@xcite , but only systems with two degrees of freedom have been investigated numerically .
+ we describe the renormalization - group transformation and we implement it numerically for the spiral mean torus .
the result is that the values of the critical coupling given by the renormalization coincide up to numerical precision with the thresholds of the break - up of the spiral mean torus ( of dimension 3 ) given by frequency map analysis .
the two methods we compare are completely independent , both conceptually and in their practical realizations .
the frequency map analysis is based on the analysis of trajectories , while the renormalization is based on a criterion of convergence of a sequence of canonical transformations .
+ we conjecture , on the basis of this numerical result , that the renormalization - group transformation converges up to the critical surface ( the set of hamiltonians where the torus of the given frequency is critical , i.e. at the threshold of its break - up ) , at least in a region of the critical surface of the hamiltonian space where critical couplings are small enough ( in order that the elimination procedure is well - defined @xcite ) .
+ we consider a class of hamiltonians with three degrees of freedom written in terms of actions @xmath0 and angles @xmath1 ( the 3-dimensional torus parametrized , e.g. , by @xmath2 ^ 3 $ ] ) @xmath3 where @xmath4 denotes the coupling parameter . in this article
, we consider the particular class of models for which the integrable part @xmath5 is given by @xmath6 where @xmath7 is the frequency vector of the considered invariant torus , and @xmath8 is another constant vector non - parallel to @xmath7 .
we suppose that @xmath7 is incommensurate , i.e. there is no nonzero integer vector @xmath9 such that @xmath10 .
+ since the quantity @xmath11 is conserved ( where @xmath12 denotes a vector orthogonal to @xmath8 and to @xmath7 ) , one can show ( even if @xmath13 is not a function on the three - dimensional torus ) that this model ( [ eqn : ham])-([eqn : h0 ] ) is intermediate between two and three degrees of freedom ; in appropriate coordinates it can be interpreted as one degree of freedom driven by a multi - periodic force with incommensurate frequencies @xmath7 . in particular , invariant tori in this intermediate model act as barriers in phase space ( limiting the diffusion of trajectories ) in a similar way as for two degrees of freedom hamiltonian systems .
we analyze in this article the break - up of invariant tori with spiral mean frequencies for this particular type of models , by choosing a special form of the perturbation ( see section [ sect : result ] ) , such that the model is equivalent to a pendulum driven by two periodic forces with incommensurate frequencies .
the method is however applicable to any perturbation and to the case of full three degrees of freedom @xcite .
+ we are interested in the stability of the torus with frequency vector @xmath7 . for the unperturbed hamiltonian @xmath5 ,
this torus is located at @xmath14 .
kolmogorov - arnold - moser ( kam ) theorems were proven for hamiltonians ( [ eqn : ham ] ) provided that @xmath7 satisfies a diophantine condition @xcite .
this theorem shows the existence of the torus with frequency vector @xmath15 for a sufficiently small and smooth perturbation @xmath16 .
the invariant torus is a small deformation of the unperturbed one .
the existence of the torus outside the perturbative regime is still an open question even if efforts have been made to increase lower bounds for specific models ( for a two dimensional model , see ref .
conversely , for sufficiently large values of the coupling parameter , it has been shown that the torus does no longer exist @xcite .
the aim of this paper is to determine @xmath17 such that @xmath18 has a smooth invariant torus of the given frequency for @xmath19 , and does not have this invariant torus for @xmath20 .
+ the invariant torus we study ( named the _ spiral mean _ torus ) has the frequency vector @xmath21 where @xmath22 is the spiral mean , i.e. the real root of @xmath23 ( @xmath24 ) . from some of its properties
, @xmath22 plays a similar role as the golden mean in the two degrees of freedom case @xcite .
the analogy comes from the fact that one can generate rational approximants by iterating a _ single _ unimodular matrix @xmath25 . in
what follows , we call _ resonance _ an element of the sequence @xmath26 where @xmath27 and @xmath28 the word
_ resonance _ refers to the fact that the small denominators @xmath29 appearing in the perturbation series or in the kam iteration , tend to zero geometrically as @xmath30 increases ( @xmath31 ) .
we notice that @xmath7 is an eigenvector of @xmath32 , where @xmath32 denotes the transposed matrix of @xmath25
. one can prove @xcite that @xmath7 satisfies a diophantine condition of the form : @xmath33 where @xmath34 , and @xmath35 .
the renormalization transformations are defined for a _ fixed _ frequency vector @xmath7 , and contain a partial elimination of the irrelevant modes ( the non - resonant part ) of the perturbation , and a rescaling of phase space .
the elimination of irrelevant modes is performed by iterating a change of coordinates as in kam theory .
we remark that other perturbative techniques can be used instead , leading to similar results .
the rescaling of phase space combines a shift of the resonances , a rescaling of time , and a rescaling of the actions .
the aim is to change the scale of the actions to a smaller one ( and a longer time scale ) .
this renormalization can be thought as a microscope in phase space .
the non - resonant modes are the ones which affect the motion at short time scales , and can be dealt with averaging methods .
we define the non - resonant modes to be the ones that satisfy the inequality @xmath36 this set of modes , denoted @xmath37 , is the interior of a cone around the @xmath7-direction in the space of 3-dimensional vectors , with angle @xmath38 .
we define the resonant modes as the fourier modes which do not satisfy the condition ( [ eqn : cond ] ) , i.e. this set , denoted @xmath39 , is the complement of @xmath37 in @xmath40 . since @xmath41 does not satisfy eq .
( [ eqn : cond ] ) for @xmath42 , @xmath39 contains the resonances that produce small denominators in the perturbation series or in the kam theory .
the `` frequency cut - off '' ( between resonant and non - resonant modes ) restricts the fourier modes that can be eliminated in one renormalization step , without running into small denominator problems ( the non - resonant modes ) . as it is common with cut - offs ,
there is not a single `` natural '' choice .
more generally , other choices in the splitting of @xmath43 into resonant and non - resonant modes should lead to the same results provided , e.g. , that the ratio @xmath44 is bounded on @xmath37 , and that the shift of the resonances contracts non - zero vectors @xmath45 in @xmath39 and maps them into @xmath37 after a finite number of iterations of the transformation @xcite .
+ the transformation , acting on a hamiltonian @xmath46 of the form @xmath47 where @xmath5 is given by eq . ( [ eqn : h0 ] ) , combines four steps : + * ( 1 ) * we shift the resonances @xmath48 : we require that the new angles @xmath49 satisfy @xmath50 for @xmath42 .
this is performed by the linear canonical transformation @xmath51 since @xmath25 is an integer matrix with determinant one , this transformation preserves the @xmath52-structure of the angles .
we notice that the resonance @xmath53 is changed into @xmath54 which satisfies the condition ( [ eqn : cond ] ) , i.e. it is a non - resonant mode : some of the resonant modes are turned into non - resonant ones by this linear transformation . + this step changes the frequency @xmath15 into @xmath55 ( since @xmath15 is an eigenvector of @xmath32 by construction ) , and the vector @xmath8 into @xmath56 . in order to keep a unit norm , we define the image of @xmath8 by @xmath57 * ( 2 ) * we rescale the energy ( or equivalently time ) by a factor @xmath22 ( i.e. we multiply the hamiltonian by @xmath22 ) , in order to keep the frequency fixed at @xmath15 . + * ( 3 ) * we rescale the actions : @xmath58 such that the mean - value of the coefficient of the quadratic term in @xmath59 is equal to @xmath60 .
this normalization condition is essential for the convergence of the transformation .
after steps 1 , 2 and 3 , the hamiltonian expressed in the new variables is @xmath61 for @xmath46 given by eq .
( [ eqn : hrg3d ] ) , this expression becomes @xmath62 thus the choice of the rescaling in the actions ( step 3 ) is @xmath63 where @xmath64 denotes the coefficient of the mean - value of the quadratic part , in the representation : @xmath65 * ( 4 ) * we perform a near - identity canonical transformation @xmath66 that eliminates completely the non - resonant part of the perturbation in @xmath59 .
this transformation satisfies the following equation : @xmath67 where @xmath68 denotes the projection operator on the non - resonant modes acting on a hamiltonian @xmath46 as @xmath69 equation ( [ eq : uh ] ) is solved by a newton method .
we iterate a change of coordinates as in kam theory that reduces the non - resonant modes of the perturbation from @xmath4 to @xmath70 .
one step of the elimination is performed by a lie transformation @xmath71 , generated by a function @xmath72 .
the expression of a hamiltonian @xmath46 in the new coordinates is given by @xmath73 where @xmath74 is the poisson bracket between two scalar functions of the actions and angles : @xmath75 and the operator @xmath76 is defined as @xmath77 .
the generating function @xmath78 is chosen such that the order @xmath4 of the non - resonant part of the perturbation vanishes .
we construct recursively a series of hamiltonians @xmath79 , starting with @xmath80 , such that the limit @xmath81 is canonically conjugate to @xmath59 but does not contain non - resonant modes , i.e. @xmath82 .
one step of this elimination procedure , @xmath83 , is done by applying a change of coordinates @xmath84 such that the order of the non - resonant modes of @xmath85 is @xmath86 , where @xmath87 denotes the order of the non - resonant modes of @xmath79 . at the @xmath88-th step ,
the order of the non - resonant modes of @xmath79 is @xmath89 , where @xmath90 is the order of the non - resonant modes of @xmath59 .
if this procedure converges , it defines a canonical transformation @xmath91 , such that the final hamiltonian @xmath92 does not contain any non - resonant mode . + the specific implementation of one step of this elimination procedure will be described more explicitly in the next sections .
we will discuss two versions of this transformation : the first one is a renormalization for hamiltonians in power series in the actions , and the second one is a slightly different version which eliminates only the non - resonant modes of the constant and linear part in the actions of the rescaled hamiltonian @xmath59 , and which allows us to define a renormalization within a space of quadratic hamiltonians in the actions , following thirring s approach of the kam theorem @xcite .
+ it has been proven in ref .
@xcite that , for a sufficiently small non - resonant part of the perturbation , the transformation @xmath66 is a well - defined canonical transformation such that a hamiltonian expressed in these new coordinates does not have non - resonant modes .
the domain of definition of @xmath66 has been extended to some non - perturbative domain in ref .
+ concerning the quadratic case , we lack at this moment a theoretical background to prove an analogous theorem .
the convergence of the elimination procedure in both cases outside the perturbative regime ( @xmath4 small ) is observed numerically .
+ in summary , the renormalization - group transformations we define act as follows : first , some of the resonant modes of the perturbation are turned into non - resonant modes by a frequency shift and a rescaling .
then , a kam - type iteration eliminates these non - resonant modes , while slightly changing the resonant modes .
we define in this section , one step of the elimination of the non - resonant modes of the rescaled hamiltonian @xmath59 .
the renormalization transformation acts on the following family of hamiltonians @xmath93 with @xmath94 .
we suppose that @xmath95 is nonzero ( in order to have a twist direction in the actions ) .
the approximations involved in this transformation are of two types : we truncate the fourier series of the functions @xmath96 , i.e. we approximate a scalar function @xmath97 of the angles by @xmath98}(\u{\phi})=\sum_{\u{\nu}\in{\mathcal c}_l } f_{\u{\nu } } e^{i\u{\nu}\cdot\u{\phi}},\ ] ] where @xmath99 , and we also neglect all the terms of order @xmath100 that are produced by the transformation . + one step of the elimination procedure @xmath101 is constructed as follows : we consider that @xmath96 depends on a small parameter @xmath4 , such that @xmath102 is of order @xmath4 .
we define @xmath5 , the integrable part of @xmath46 as @xmath103 and the perturbation of @xmath5 is denoted @xmath104 . in order to eliminate the non - resonant modes of @xmath96 to the first order in @xmath4
, we perform a lie transformation @xmath105 generated by a function @xmath78 of order @xmath4 and of the form @xmath106 the first terms of @xmath107 are @xmath108 .
the function @xmath78 is determined by imposing that the order @xmath4 vanishes : @xmath109 this condition determines the non - resonant modes of @xmath78 . for the resonant ones , we choose @xmath110 .
thus we notice that the mean value of @xmath111 is zero . the constant
@xmath112 eliminates the linear term in the @xmath113-variable , @xmath114 , by requiring that @xmath115 : @xmath116 we solve eq .
( [ eq : dets ] ) by a newton method with an initial condition which satisfies @xmath117 , since @xmath118 is expected in general to be small
. then we compute @xmath119 by calculating recursively the poisson brackets @xmath120 , for @xmath42 .
denoting @xmath121 , @xmath107 becomes @xmath122 we consider the following family of quadratic hamiltonians in the actions and described by three scalar functions of the angles @xmath123 the kam transformations are constructed such that the iteration stays within the space of hamiltonians quadratic in the actions @xcite . in order to prove the existence of a torus with frequency vector @xmath15 for hamiltonian systems described by eq .
( [ eqn : hquadra ] ) , it is not necessary to eliminate @xmath124 , but only @xmath125 and @xmath97 ( the main point is that the torus with frequency vector @xmath7 is located at @xmath14 for any @xmath46 with @xmath126 , even if @xmath46 is not globally integrable ) .
the elimination of @xmath97 and @xmath125 can be achieved with canonical transformations with generating functions that are linear in the action variables , and thus map the family of hamiltonians ( [ eqn : hquadra ] ) into itself .
this is very convenient numerically , as one only works with three scalar functions @xmath124 , @xmath125 and @xmath97 .
the only approximation involved in the numerical implementation of the transformation is a truncation of the fourier series of these functions , according to eq .
( [ eqn : coeff ] ) .
+ in this section , we describe one step of the elimination of the non - resonant modes @xmath127 .
we assume that @xmath125 and @xmath97 depend on a ( small ) parameter @xmath4 , in such a way that @xmath128 and @xmath129 are of order @xmath4 .
the idea is to eliminate the non - resonant modes of @xmath125 and @xmath97 to first order in @xmath4 , at the expense of adding terms that are of order @xmath130 in the resonant modes and of order @xmath131 in the non - resonant modes .
this is performed by a lie transformation @xmath132 generated by a function @xmath78 of order @xmath4 linear in the action variables , of the form @xmath133 characterized by two scalar functions @xmath134 , @xmath135 , and a constant @xmath112 .
the expression of the hamiltonian in the new variables is obtained by eq .
( [ eqn : exp ] ) .
a consequence of the linearity of @xmath78 in @xmath136 is that the hamiltonian @xmath107 is again quadratic in the actions , and of the form @xmath137 this can be seen by this simple argument : given a quadratic function in the @xmath113-variable and a linear function @xmath78 , then @xmath138 is again quadratic in the @xmath113-variable .
the derivatives @xmath139 and @xmath140 are linear in the @xmath113-variable , and @xmath141 is quadratic , while @xmath142 is constant in this variable .
therefore , @xmath138 given by eq .
( [ eqn : poiss ] ) is quadratic .
consequently , iterating this argument , @xmath143 is also quadratic .
we notice that the vector @xmath8 remains unchanged during each step of the elimination .
+ the functions @xmath134 , @xmath135 , and the constant @xmath112 are chosen in such a way that @xmath144 and @xmath145 vanish to order @xmath4 .
the constant @xmath112 corresponds to a translation in the actions , which has the purpose of eliminating the mean value of the linear term in the @xmath113-variable .
then we express @xmath107 by calculating recursively the poisson brackets @xmath120 , for @xmath146 , like in the previous section .
+ we notice that for some purposes it is more convenient to eliminate also the non - resonant part of @xmath124 , together with the one of @xmath125 and @xmath97 ( see the remarks in refs @xcite ) . but such elimination procedure generates arbitrary orders in the @xmath147-variable and this leads to the first version of the transformation .
the advantage to work with this second version of the elimination procedure is that the hamiltonians ( [ eqn : hquadra ] ) are described by only three scalar functions of the angles .
thus it is numerically more efficient since the renormalization map is of lower dimension . concerning the renormalization transformation for hamiltonians in power series in the actions , we truncate the fourier series of each scalar function of the angles with a cut - off parameter @xmath148 , and the taylor series with a cut - off parameter @xmath149 . for fixed @xmath148 and @xmath149 , the dimension of the renormalization map is equal to @xmath150 . concerning the renormalization defined for quadratic hamiltonians , the renormalization map is of dimension @xmath151 .
+ _ remark : _ another advantage to work the thirring s version of the renormalization is that it can be generalized more easily to non - degenerate hamiltonians @xmath152 with @xmath153 .
( see ref .
we consider the following quasi - periodically driven pendulum model : @xmath154 with the frequencies @xmath155 , @xmath156 , and @xmath157 a real parameter .
this model can be mapped into the following degenerate hamiltonian system with three degrees of freedom : @xmath158 where @xmath159 and the perturbation @xmath97 is given by @xmath160 this can be seen by considering the three angles @xmath161 , @xmath162 , and @xmath163 .
the hamiltonian ( [ eq : fp ] ) becomes : @xmath164 where we added @xmath165 to the hamiltonian in order to satisfy the equations of motion for the new variables @xmath166 and @xmath167 when @xmath168 at time @xmath169 .
the linear canonical transformation @xmath170 with @xmath171 maps hamiltonian ( [ eq : hamint ] ) into hamiltonian ( [ eqn : param ] ) .
we are looking at the break - up of the invariant torus with frequency vector @xmath7 which is located at @xmath172 for hamiltonian ( [ eqn : param ] ) with @xmath173 .
we notice that this torus is located at @xmath174 for hamiltonian ( [ eq : fp ] ) and its frequency is equal to @xmath175 .
the three main resonances ( of order @xmath4 ) are located at @xmath176 , @xmath177 , and @xmath178 , and thus the torus is located in between the resonances @xmath177 and @xmath176 .
we first discuss two rough estimates of the critical coupling obtained by some drastic simplifications : @xmath179 applying chirikov s criterion @xcite gives the following approximate value for the critical coupling @xmath180 .
@xmath181 if we neglect the effect of the resonance located at @xmath178 , e.g. , by setting @xmath182 in the perturbation of eq .
( [ eq : fp ] ) , we can apply the renormalization procedure described in ref .
@xcite for hamiltonians with two degrees of freedom . since hamiltonian ( [ eq : hamint ] ) does not depend on @xmath167 in this case
, @xmath183 is constant and the problem reduces to the study of the break - up of the invariant torus with frequency vector @xmath184 for the following hamiltonian system with two degrees of freedom : @xmath185 this method gives @xmath186 .
+ we perform the complete renormalization procedure described in the preceding sections for hamiltonian system ( [ eqn : param ] ) for a given value of @xmath157 .
we fix the cut - off parameters @xmath148 and @xmath149 of the renormalization transformation , and we take successively larger couplings @xmath4 in order to determine whether the renormalization transformation @xmath187 converges to an integrable hamiltonian , or whether it diverges . by a bisection procedure , we determine the critical coupling @xmath188 such that as @xmath88 tends to @xmath189 : @xmath190 table 1 gives the values of @xmath17 for @xmath191 as a function of @xmath148 and @xmath149 computed by the two renormalization transformations .
we notice that the values @xmath192 converge to @xmath193 as @xmath148 and @xmath149 grow .
+ figure [ fig : mu ] shows the values of @xmath194 for @xmath195 $ ] determined by the renormalization transformation for quadratic hamiltonians with cut - off parameter @xmath196 .
frequency map analysis@xcite allows to study the destruction of kam tori by looking at the regularity of the frequency map defined from the action like variables to the numerically determined frequencies . in the present case ( [ eq : fp ] ) , the frequency map is very simple as the system is equivalent to a one degree of freedom system with a quasiperiodic perturbation @xcite .
the angle variable @xmath197 can then be fixed to @xmath198 , and we are left to a one dimensional map @xmath199 , @xmath200 , where @xmath201 is determined numerically from the output of the numerical integration of @xmath202 over a time interval of length @xmath203 , starting with initial conditions @xmath204 @xcite . on
the set of kam tori @xmath205 is regular , or more precisely , it can be extended to a smooth function .
thus , when @xmath205 appears to be non regular , this is an indication that all tori are destroyed in the corresponding interval .
this allows to obtain a global vision of the dynamics of the system , as illustrated by figure [ fig : fmafig1 ] for hamiltonian ( [ eq : fp ] ) with @xmath191 , where @xmath206 is plotted versus @xmath207 for @xmath208 and @xmath209 , and for @xmath210 $ ] for a moderate precision ( @xmath211 ) .
it seems clear on these figures that in the region between the two resonances @xmath212 and @xmath213 , there are no tori left for @xmath214 , while many of them remain in the region @xmath215 $ ] for @xmath208 . in order to have a more detailed view for the destruction of the tori with frequency @xmath216 , we have extended the time interval to @xmath217 , and reduced very much the stepsize in @xmath207 ( figure [ fig : fmafig2 ] ) . with these settings and for @xmath191
, it appears clearly that the torus with frequency @xmath216 is destroyed for @xmath218 ( b ) , while the behavior of the frequency map appears to be very regular for @xmath219 ( a ) .
it should be noted that the frequency map analysis provides a criterion for the destruction of tori .
the fact that the frequency curve appears to be irregular provides an evidence that the tori are destroyed , but when the curve is regular , a higher accuracy ( which means a longer time span @xmath203 ) could reveal the destruction of additional tori . for @xmath218 and for @xmath191 , all tori in the vicinity of @xmath216
are destroyed , which is in agreement with the value @xmath220 found with the renormalization technique , as this is the largest value of @xmath221 for which the renormalization converges .
+ we have performed computations for other values of the parameter @xmath157 .
figure [ fig : mu ] shows the agreement between the couplings obtained by frequency map analysis and the ones obtained by the renormalization transformation .
a better accuracy can be obtained by taking a larger cut - off parameter @xmath148 for the renormalization computations .
it appears clearly in figure [ fig : fmafig1 ] that the torus with frequency @xmath216 is not the last torus to be destroyed in the interval of frequencies @xmath222 $ ] for @xmath191 , and that many tori still survive for @xmath223 .
we thus have searched for the value of the parameter @xmath221 for which the last torus disappears .
this is done by increasing the value of the parameter @xmath221 until all tori disappear . in this procedure , at each stage we identify the resonant islands and chaotic regions , and then decrease the stepsize in @xmath207 .
this allows one to obtain very refined details as one gets close to the critical value of @xmath221 .
the frequencies of the last invariant torus can be investigated by frequency map analysis .
this method gives a critical threshold at about @xmath224 ( see figure [ fig : fmafig3 ] ) .
for hamiltonian ( [ eq : fp ] ) with @xmath191 , the last invariant torus is located between the resonances @xmath176 and @xmath177 at @xmath225 .
the corresponding frequencies for the last three - dimensional torus for the family of hamiltonians ( [ eqn : param ] ) are @xmath226 , @xmath227 and @xmath228 .
a nearby invariant torus has the frequencies @xmath229 , @xmath230 and @xmath228 .
the critical coupling for the break - up of this invariant torus can be computed by the renormalization transformation defined in sec .
[ sec.2 ] , mainly by defining first a unimodular transformation that maps the frequencies of the torus into the frequencies @xmath231 , @xmath22 and 1 ( the perturbation is then different from eq . ( [ eq : pertu ] ) ) .
the renormalization gives @xmath224 .
+ if we neglect the resonance located at @xmath178 , renormalization methods and frequency map analysis show that there are two last kam tori , one located at @xmath232 and the other one at @xmath233 , with a critical threshold of @xmath234 .
thus the effect of the third resonance at @xmath178 is to destabilize the motion in the region between @xmath176 and @xmath177 closer to @xmath177 ( since the third resonance is closer to @xmath177 ) .
consequently the last kam torus is located nearer @xmath176 than in the system without the third resonance , and the value of the threshold of global stochasticity ( break - up of the last kam surface ) is smaller .
we acknowledge useful discussions with g. gallavotti , h. koch , and r.s .
support from ec contract no .
erbchrxct94 - 0460 for the project `` stability and universality in classical mechanics '' is acknowledged .
cc thanks support from the carnot foundation .
j. laskar , introduction to frequency map analysis , in : c. sim ( ed . ) , hamiltonian systems with three or more degrees of freedom , nato asi series , kluwer academic publishers , dordrecht , 1999 , pp .
134 150 .
.critical coupling @xmath17 for hamiltonian ( [ eq : fp ] ) with @xmath191 as a function of the cut - off parameters @xmath148 and @xmath149 , computed with the transformation rg1 defined in sec . [
sec : general ] , and with the transformation rg2 defined in sec .
[ sec : quadra ] .
frequency map analysis indicates that the spiral mean torus is broken for couplings larger than 0.0895 . [ cols="^,^,^,^,^,^,^",options="header " , ] ) with @xmath191 . for each selected value of the parameter @xmath221 , the set of frequencies @xmath201 for which tori are not destroyed is plotted versus @xmath221 .
the critical value of @xmath221 for which no tori will survive is thus @xmath242 . ] | we consider a class of hamiltonians with three degrees of freedom that can be mapped into quasi - periodically driven pendulums .
the purpose of this paper is to determine the threshold of the break - up of invariant tori with a specific frequency vector .
we apply two techniques : the frequency map analysis and renormalization - group methods .
the renormalization transformation acting on a hamiltonian is a canonical change of coordinates which is a combination of a partial elimination of the irrelevant modes of the hamiltonian and a rescaling of phase space around the considered torus .
we give numerical evidence that the critical coupling at which the renormalization transformation starts to diverge is the same as the value given by the frequency map analysis for the break - up of invariant tori .
furthermore , we obtain by these methods numerical values of the threshold of the break - up of the last invariant torus
. u invariant tori , renormalization , hamiltonian systems 05.45.ac , 05.10.cc , 45.20.jj |
the main technical contribution of this paper is an extension of matouek s range emptiness and reporting data structures @xcite ( see also @xcite for a dynamic version of the problem ) to the case of general semi - algebraic ranges . [ [ ray - shooting - amid - balls . ] ] ray shooting amid balls .
+ + + + + + + + + + + + + + + + + + + + + + + + a motivating application of this study is ray shooting amid balls in @xmath6 , where we want to construct a data structure of linear size with near - linear preprocessing , which supports ray shooting queries in sublinear time .
typically , in problems of this sort , the bound on the query time is some fractional power of @xmath1 , the number of objects , and the goal is to make the exponent as small as possible . for example ,
ray shooting amid a collection of @xmath1 arbitrary triangles can be performed in @xmath14 time ( with linear storage ) @xcite .
better solutions are known for various special cases .
for example , the authors have shown @xcite that the query time can be improved to @xmath13 , when the triangles are all _ fat _ , or are all stabbed by a common line . at the other end of the spectrum ,
one is interested in ray shooting algorithms and data structures where a ray shooting query can be performed in logarithmic or polylogarithmic time ( or even @xmath29 time , for any @xmath10 ; this is @xmath21 in our shorthand notation ) . in this case
, the goal is to reduce the storage ( and preprocessing ) requirements as much as possible .
for example , for arbitrary triangles ( and even for the special case of fat triangles ) , the best known bound for the storage requirement ( with logarithmic query time ) is @xmath30 @xcite . for balls , mohaban and sharir @xcite , gave an algorithm with @xmath31 storage and @xmath21 query time .
however , when only linear storage is used , the previously best known query time ( for balls ) is @xmath14 ( as in the case of general triangles ) . in this paper
we show , as an application of our general range emptiness machinery , that this can be improved to @xmath13 time .
when answering a ray - shooting query for a set @xmath32 of input objects , one generally reduces the problem to that of answering _ segment emptiness _
queries , following the parametric searching scheme proposed by agarwal and matouek @xcite ( see also megiddo @xcite for the original underlying technique ) .
a standard way of performing the latter kind of queries is to switch to a dual parametric space , where each object in the input set is represented by a _
point_. a segment @xmath33 in @xmath6 is mapped to a surface @xmath34 , which is the locus of all the points representing the objects that @xmath33 touches ( without penetrating into their interior ) .
usually , @xmath34 partitions the dual space into two portions , one , @xmath35 , consisting of points representing objects whose interior is intersected by @xmath33 , and the other , @xmath36 , consisting of points representing objects that @xmath33 avoids .
the segment - emptiness problem thus transforms into a range - emptiness query : does @xmath35 contain any point representing an input object ? [ [ range - reporting - and - emptiness - searching . ] ] range reporting and emptiness searching .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + range - emptiness queries of this kind have been studied by matouek @xcite ( see also agarwal and matouek @xcite ) , but only for the case where the ranges are halfspaces bounded by hyperplanes . for this case
, matouek has established a so - called _ shallow - cutting lemma _ , that shows the existence of a @xmath37-cutting of the @xmath1 given surfaces ( hyperplanes in this case ) .
see below for more details . ] that covers the complement of the union of any @xmath38 given halfspace ranges , whose size is significantly smaller than the size of a @xmath37-cutting that covers the entire space .
this lemma provides the basic tool for partitioning a point set @xmath0 , in the style of @xcite , so that _ shallow _ hyperplanes ( those containing at most @xmath39 points of @xmath0 below them , say , for some given parameter @xmath40 ) cross only a small number of cells of the partition ( see below for more details ) .
this in turn yields a data structure , known as a _ shallow partition tree _ , that stores a recursive partitioning of @xmath0 , which enables us to answer more efficiently halfspace range _ reporting _
queries for shallow hyperplanes , and thus also halfspace range emptiness queries . using this approach
, the query time ( for emptiness ) improves from the general halfspace range searching query cost of @xmath41 to @xmath42 .
reporting takes @xmath43 , where @xmath27 is the output size .
consequently , one way of applying this machinery for more general semi - algebraic ranges is to `` lift '' the set of points and the ranges into a higher - dimensional space by means of an appropriate _ linearization _ , as in @xcite , and then apply the above machinery .
( for this , one needs to assume that the given ranges have _ constant description complexity _ , meaning that each range is a boolean combination of a constant number of polynomial equalities and inequalities of constant maximum degree .
however , if the space in which the ranges are linearized has high dimension , the resulting range reporting or emptiness queries become significantly less efficient .
moreover , in many applications , the ranges are boolean combinations of polynomial ( equalities and ) inequalities , which creates additional difficulties in linearizing the ranges , resulting in even worse running time .
an alternative technique is to give up linearization , and instead work in the original space .
as follows from the machinery of @xcite ( and further elaborated later in this paper ) , this requires , as a major tool , the ( existence and ) construction of a decomposition of the _ complement of the union _ of @xmath38 given ranges ( in the case of segment emptiness , these are the ranges @xmath35 , for an appropriate collection of segments @xmath33 ) , into a small number of `` elementary cells '' ( in the terminology of @xcite see also below ) . here
we face , especially in higher dimensions , a scarcity of sharp bounds on the complexity of the union itself , to begin with , and then on the complexity of a decomposition of its complement .
often , the best one can do is to decompose the entire arrangement of the given ranges , which results in too many elementary cells , and consequently in an algorithm with poor performance . to recap , in the key technical step in answering general semi - algebraic range reporting or emptiness queries , the best current approaches are either to construct a cutting of the _ entire _ arrangement of the range - bounding surfaces in the original space , or to construct a shallow cutting in another higher - dimensional space into which the ranges can be linearized .
for many natural problems ( including the segment - emptiness problem ) , both approaches yield relatively poor performance . as we will shortly note , in handling general semi - algebraic ranges , we face another major technical issue , having to do with the construction of efficient _ test sets _ of ranges ( in the terminology of @xcite , elaborated below ) . addressing
this issue is a major component of the analysis in this paper , and is discussed in detail later on .
[ [ our - results . ] ] our results .
+ + + + + + + + + + + + we propose a variant of the shallow - cutting machinery of @xcite for the case of semi - algebraic ranges , which avoids the need for linearization , and works in the original space ( which , for the case of ray shooting amid balls , is a 4-dimensional parametric space in which the balls are represented as points ) . while the machinery used by our variant is similar in principle to that in @xcite ,
there are several significant technical difficulties which require more careful treatment .
matouek s technique @xcite , as well as ours , considers a finite set @xmath44 of shallow ranges ( called a _
test set _ ) , and builds a data structure which caters only for ranges in @xmath44 .
matouek shows how to build , for any given parameter @xmath40 , a set of halfspaces of size polynomial in @xmath40 , which represents well _ all _ @xmath45-shallow ranges , in the following sense : for any _ simplicial partition _ into @xmath46 subsets of roughly equal size , each enclosed by some simplex ( in the linear case ) or some elementary cell ( in the general semi - algebraic case ) ; see @xcite and section [ section : range_emptiness ] below .
] @xmath47 with parameter @xmath40 , let @xmath48 denote the maximal number of cells of @xmath47 crossed by a halfspace in @xmath44 .
then each @xmath45-shallow halfspace crosses at most @xmath49 cells of @xmath47 , where @xmath25 is a constant that depends on the dimension .
unfortunately ( for the present analysis ) , the linear nature of the ranges is crucially needed for the proof , which therefore fails for non - linear ranges .
being a good representative of all shallow ranges , in the above sense , is only one of the requirements from a good test set @xmath44 .
the other requirements are that @xmath44 be small , so that , in particular , it can be constructed efficiently , and that the ( decomposition of the ) complement of the union of any subset of @xmath44 have small complexity .
all these properties hold for the case of halfspaces bounded by hyperplanes , studied in @xcite .
[ testset_for_empty_only ] as it turns out , and hinted above , obtaining a `` good '' test set @xmath44 for general semi - algebraic ranges , with the above properties , is not an easy task .
we give a simple general recipe for constructing such a set @xmath44 , but it consists of more complex ranges than those in the original setup . a major problem with this recipe is that since the members of @xmath44 have a more complex shape , it becomes harder to establish good bounds on the complexity of ( the decomposition of ) the complement of the union of any subset of these generalized ranges .
nevertheless , once a good test set has been shown to exist , and to be efficiently computable , it leads to a construction of an efficient elementary - cell partition with a small crossing number for any empty or shallow original range . using this construction recursively ,
one obtains a _ partition tree _ , of linear size , so that any shallow original range @xmath50 visits only a small number of its nodes ( where @xmath50 visits a node if it _ crosses _ the elementary cell enclosing the subset of that node , meaning that it intersects this cell but does not fully contain it ) , which in turn leads to an efficient range reporting or emptiness - testing procedure .
this part , of constructing and searching the tree , is almost identical to its counterparts in the earlier works @xcite , and we will not elaborate on it here , focusing only on the technicalities in the construction of a single `` shallow '' elementary - cell partition . developing all this machinery , and then putting it into action , we obtain efficient data structures for the following applications , improving previous results or obtaining the first nontrivial solutions .
these instances are : [ [ ray - shooting - amid - balls - in-3-space . ] ] ray shooting amid balls in 3-space .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + given a set @xmath32 of @xmath1 balls in @xmath6 , we construct , in @xmath8 time , a data structure of @xmath7 size , which can determine , for a given query segment @xmath33 , whether @xmath33 is empty ( avoids all balls ) , in @xmath13 time . plugging this data structure into the parametric searching technique of agarwal and matouek @xcite , we obtain a data structure for answering ray shooting queries amid the balls of @xmath32 , which has similar performance bounds .
we represent balls in 3-space as points in @xmath51 , where a ball with center @xmath52 and radius @xmath40 is mapped to the point @xmath53 , and each object @xmath54 is mapped to the surface @xmath55 , which is the locus of all ( points representing ) balls tangent to @xmath56 ( i.e. , balls that touch @xmath56 , but do not penetrate into its interior ) . in this case
, the range of an object @xmath56 is the upper halfspace @xmath57 consisting of all points lying above @xmath55 ( representing balls that intersect @xmath56 ) .
the complement of the union of a subfamily of these ranges is the region below the lower envelope of the corresponding surfaces which are considerably more complex than just lines or segments , but are nevertheless still of constant description complexity . ]
the _ minimization diagram _ of this envelope is the 3-dimensional euclidean voronoi diagram of the corresponding set of objects .
thus we reveal ( what we regard as ) a somewhat surprising connection between the problem of ray shooting amid balls and the problem of analyzing the complexity of euclidean voronoi diagrams of ( simply - shaped ) objects in 3-space . [
[ farthest - point - from - a - line - or - from - any - convex - set - in - reals3 . ] ] farthest point from a line ( or from any convex set ) in @xmath6 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath0 be a set of @xmath1 points in @xmath6 .
we wish to preprocess @xmath0 into a data structure of size @xmath7 , so that , for any query line @xmath15 , we can efficiently find the point of @xmath0 farthest from @xmath15 .
this is a useful routine for approximating polygonal paths in three dimensions ; see @xcite . as in the ray shooting problem
, we can reduce such a query to a range emptiness query of the form : given a cylinder @xmath58 , does it contain all the points of @xmath0 ? ( that is , is the complement of the cylinder empty ? ) we prefer to regard this as an instance of the complementary _ range fullness _ problem , which seeks to determine whether a query range is _ full _
( i.e. , contains all the input points ) .
our machinery can handle this problem .
in fact , we can solve the range fullness problem for any family of _ convex _ ranges in 3-space , of constant description complexity .
our solution requires @xmath7 storage and near linear preprocessing , and answers a range fullness query in @xmath16 time , improving the query time @xmath13 given by agarwal and matouek @xcite .
we then apply this result to solve the problem of finding the largest - area triangle spanned by a set of @xmath1 points in 3-space .
the resulting algorithm requires @xmath17 time , which improves a previous bound of @xmath59 due to daescu and serfling @xcite .
we also adapt our machinery to compute efficiently the largest - perimeter triangle and the largest - height triangle spanned by such a point set . in both this , and the preceding ray - shooting applications
, we use the general , more abstract recipe for constructing good test sets .
[ [ fat - triangle - and - circular - cap - range - emptiness - searching - and - reporting . ] ] fat triangle and circular cap range emptiness searching and reporting . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + finally , we consider two planar instances of the range emptiness and reporting problems , in which we are given a planar set @xmath0 of @xmath1 points , and the ranges are either _
@xmath19-fat triangles _ or sufficiently large _ circular caps _ ( say , larger than a semidisk ) . the general technique of agarwal and matouek @xcite yields , for any class of planar ranges with constant description complexity , a data structure with near linear preprocessing and linear storage , which answers such queries in time @xmath16 ( for emptiness ) or @xmath60 ( for reporting ) .
we improve the query time to @xmath21 and @xmath23 , respectively , in both cases . in these planar applications , we abandon the general recipe , and construct good test sets in an ad - hoc ( and simpler ) manner . for @xmath19-fat triangles ( i.e. , triangles with the property that each of their angles is at least @xmath19 , which is some fixed positive constant ) ,
the test set consists of `` canonical '' @xmath61-fat triangles , and the fast query performance is a consequence of the fact that the complement of the union of @xmath38 @xmath62-fat triangles is @xmath63 , for any constant @xmath64 @xcite .
it is quite likely that our machinery can also be applied to other classes of fat objects in the plane , for which near - linear bounds on the complexity of their union are known @xcite .
however , constructing a good test set for each of these classes is not an obvious step .
we leave these extensions as open problems for further research . for circular caps ,
the motivation for range emptiness searching comes from the problem of finding , for a query consisting of a point @xmath65 and a line @xmath15 , the point of @xmath0 which lies above @xmath15 and is nearest to @xmath65 ( we only consider the case where @xmath65 lies on or above @xmath15 ) .
such a procedure was considered in @xcite .
using parametric searching , the latter problem can be reduced to that of testing for emptiness of a circular cap centered at @xmath65 and bounded by @xmath15 ( the assumption on the location of @xmath65 ensures that this cap is at least a semidisk ) . here too we manage to construct a test set which consists of ( possibly slightly smaller ) circular caps , and we exploit the fact that the complexity of the union of @xmath38 such caps is @xmath66 , as long as the caps are not too small ( relative to their bounding circles ) , to obtain the fast performance stated above . [ [ approximate - range - counting . ] ] approximate range counting .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + adapting the recent techniques of @xcite , we can turn our solutions into efficient algorithms for approximate range counting ( with small relative error ) for the cases mentioned above .
that is , for a specified @xmath10 , we can preprocess the input point set @xmath0 into a data structure which can efficiently compute , for any query range @xmath50 , an approximate count @xmath67 , satisfying @xmath68 .
the performance of the resulting algorithms is detailed in section [ sec : apx ] .
as observed in the papers just cited , approximate range counting is closely related to the range emptiness problem , which in fact is a special case of the former problem .
the algorithm in @xcite performs approximate range counting by a randomized binary search over @xmath69 , where the search is guided by repeated calls to an emptiness testing routine on various random samples of @xmath0 .
this algorithm uses emptiness searching as a black box , so , plugging our solutions for this latter problem into their algorithm , we obtain efficient approximate range counting algorithms for the ranges considered in this paper .
see section [ sec : apx ] for details .
[ [ related - work . ] ] related work .
+ + + + + + + + + + + + + our study was originally motivated by work by daescu and others @xcite on path approximations and related problems . in these applications
one needs to compute efficiently the vertex of a subpath which is farthest from a given segment ( connecting the two endpoints of the subpath ) .
these works used the standard range searching machinery of @xcite , and motivated us to look for faster implementations .
the general range emptiness ( or reporting ) problem was studied by the authors a few years ago @xcite . in this earlier version ,
we did not manage to handle properly the issue of constructing a good test set , so the results presented there are somewhat incomplete .
the present paper builds upon the previous one , but provides a thorough analysis of this aspect of the problem , and consequently obtains a complete and efficient solution to the problems listed above , and lays down the foundation for obtaining efficient solutions to many other similar problems
we believe indeed that the applications given here only scratch the surface of the wealth of potential future applications of this sort .
we begin with a brief review of the main concepts and notations used in our analysis .
[ [ range - spaces . ] ] range spaces .
+ + + + + + + + + + + + + a range space is a pair @xmath70 , where @xmath71 is a set and @xmath72 is a collection of subsets of @xmath71 , called _
ranges_. in our applications , @xmath73 , and @xmath74 is a collection of semi - algebraic sets of some specific type , each having _ constant description complexity_. that is , each set in @xmath74 is given as a boolean combination of a constant number of polynomial equalities and inequalities of constant maximum degree . to simplify the analysis , we assume , as in @xcite , that all the ranges in @xmath74 are defined by a single boolean combination , so that each polynomial @xmath75 in this combination is @xmath76-variate , and each range @xmath50 has @xmath77 degrees of freedom , so that if we substitute the values of these @xmath77 parameters into the last @xmath77 variables of each @xmath75 , the resulting boolean combination defines the range @xmath50 .
this allows us to represent the ranges of @xmath74 as points in an appropriate @xmath77-dimensional parametric space . under these special assumptions ,
the range space @xmath70 has _ finite vc - dimension _
, a property formally defined in @xcite .
informally , it ensures that , for any finite subset @xmath0 of @xmath71 , the number of distinct ranges of @xmath0 is @xmath78 , where @xmath79 is the vc - dimension . as a matter of fact , we will consider range spaces of the form @xmath80 , where @xmath81 is a finite point set , and each range in @xmath82 is the intersection of @xmath0 with a range in @xmath74 . [ [ cuttings . ] ] cuttings .
+ + + + + + + + + given a finite collection @xmath74 of @xmath1 semi - algebraic ranges in @xmath2 , as above , and a parameter @xmath83 , a _
@xmath84-cutting _ for @xmath74 is a partition @xmath85 of @xmath2 ( or of some portion of @xmath2 ) into a finite number of relatively open cells of dimensions @xmath86 , so that each cell is _ crossed _ by at most @xmath39 ranges of @xmath74 , where a range @xmath87 is said to cross a cell @xmath3 if @xmath88 , but @xmath50 does not fully contain @xmath3 .
we will also need to consider _ weighted _ @xmath84-cuttings , where each range @xmath87 has a positive weight @xmath89 , and each cell of @xmath85 is crossed by ranges whose total weight is at most @xmath90 , where @xmath91 is the overall weight of all the ranges in @xmath74 .
[ [ shallow - ranges . ] ] shallow ranges .
+ + + + + + + + + + + + + + + a range @xmath87 is called _
@xmath27-shallow _ with respect to a set @xmath0 of points in @xmath2 , if @xmath92 .
[ [ elementary - cells . ] ] elementary cells .
+ + + + + + + + + + + + + + + + + define , as in @xcite , an _ elementary cell _ in @xmath2 to be a connected relatively open semi - algebraic set of some dimension @xmath93 , which is homeomorphic to a ball and has constant description complexity . as above , we assume , for simplicity , that the elementary cells are defined by a single boolean combination involving @xmath77 free variables , and each cell is determined by fixing the values of these @xmath77 parameters . [ [ elementary - cell - partition . ] ] elementary cell partition .
+ + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath0 be a set of @xmath1 points in @xmath2 .
an _ elementary cell partition _ of @xmath0 is a collection @xmath94 , for some integer @xmath38 , such that ( i ) @xmath95 is a partition of @xmath0 ( into pairwise disjoint subsets ) , and ( ii ) each @xmath96 is an elementary cell that contains the respective subset @xmath97 . in general , the cells @xmath96 need not be disjoint . usually , one also specifies a parameter @xmath98 , and requires that @xmath99 for each @xmath100 , so @xmath101 .
[ [ the - function - zetar . ] ] the function @xmath102 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in lemma [ lemma : shallow_cutting ] and theorem [ theo : part ] , we use a function @xmath102 that bounds the number of elementary cells in a decomposition of the complement of the union of any @xmath40 ranges of @xmath74 .
we assume that @xmath102 is `` well behaved '' , in the sense that for each @xmath103 there exists @xmath104 such that @xmath105 for every @xmath40 .
we also assume that @xmath106 .
[ [ nualpha - samples - and - shallow - eps - nets . ] ] @xmath107-samples and shallow @xmath12-nets .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we recall the result of li et al .
@xcite , and adapt it , similar to the recent observations in @xcite , to obtain a useful extension of the notion of @xmath12-nets .
let @xmath108 be a range space of finite vc - dimension @xmath79 , and let @xmath109 be two given parameters . consider the distance function @xmath110 a subset @xmath111 is called a _ @xmath107-sample _ if for each @xmath112 we have @xmath113 [ lls ] a random sample @xmath114 of @xmath115 elements of @xmath71 is a @xmath107-sample with probability at least @xmath116 .
har - peled and sharir @xcite show that , by appropriately choosing @xmath19 and @xmath117 , various standard constructs , such as @xmath12-nets and @xmath12-approximations , are special cases of @xmath107-samples .
here we follow a similar approach , and show the existence of small - size _ shallow @xmath12-nets _ , a new notation introduced in this paper .
let us first define this notion . let @xmath108 be a range space of finite vc - dimension @xmath79 , and let @xmath118 be a given parameter .
a subset @xmath111 is a shallow @xmath12-net if it satisfies the following two properties , for some absolute constant @xmath25 .
\(i ) for each @xmath119 and for any parameter @xmath120 , if @xmath121 then @xmath122 .
\(ii ) for each @xmath119 and for any parameter @xmath120 , if @xmath123 then @xmath124 .
note the difference between shallow and standard @xmath12-nets : property ( i ) ( with @xmath125 ) implies that a shallow @xmath12-net is also a standard @xmath12-net ( possibly with a recalibration of @xmath12 ) . property ( ii )
has no parallel in the case of standard @xmath12-nets there is no guarantee how a standard net interacts with small ranges . [ epsshallow ]
a random sample @xmath114 of @xmath126 elements of @xmath71 is a shallow @xmath12-net with probability at least @xmath116 .
* proof : * take @xmath127 , say , and calibrate the constants in the size of @xmath114 to guarantee , with probability @xmath116 , that @xmath114 is an @xmath128-sample .
assume that this is indeed the case .
for a range @xmath112 , put @xmath129 and @xmath130 .
we have @xmath131 that is , @xmath132 or @xmath133 this is easily seen to imply properties ( i ) and ( ii ) . for ( i ) , let @xmath134 be a range for which @xmath121 ; that is , @xmath135 , for some absolute constant @xmath136 ( proportional to the vc - dimension )
. then @xmath137 for ( ii ) , let @xmath134 be a range for which @xmath123 ; that is , @xmath138 .
then @xmath139 for another absolute constant @xmath50 ( again , proportional to the vc - dimension ) . @xmath140
[ [ shallow - cutting - in - the - semi - algebraic - case . ] ] shallow cutting in the semi - algebraic case .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we begin by extending the shallow cutting lemma of matouek @xcite to the more general setting of semi - algebraic ranges .
this extension is fairly straightforward , although it involves several technical steps that deserve to be highlighted .
[ app : not ] [ lemma : shallow_cutting ]
let @xmath74 be a collection of @xmath1 semi - algebraic ranges in @xmath2 .
assume that the complement of the union of any subset of @xmath38 ranges in @xmath74 can be decomposed into at most @xmath141 elementary cells , for a well - behaved function @xmath142 as above .
then , for any @xmath98 , there exists a ( @xmath143)-cutting @xmath85 with the following properties : + ( i ) the union of the cells of @xmath85 contains the complement of the union of @xmath74 .
+ ( ii ) @xmath85 consists of @xmath144 elementary cells . + (
iii ) the complement of the union of the cells of @xmath85 is contained in a union of @xmath46 ranges in @xmath74 .
see figure [ fig : shallow - lemma ] for an illustration .
_ the proof is a fairly routine adaptation of the proof in @xcite .
we employ a variant of the method of chazelle and friedman @xcite for constructing the cutting .
let @xmath145 be a random sample of @xmath46 ranges of @xmath74 , and let @xmath146 denote the complement of the union of @xmath145 .
by assumption , @xmath146 can be decomposed into at most @xmath144 elementary cells .
the resulting collection @xmath85 of these cells is such that their union clearly contains the complement of the union of @xmath74 .
moreover , the complement of the union of @xmath85 is the union of the @xmath46 ranges of @xmath145 .
hence , @xmath85 satisfies all three conditions ( i)(iii ) , but it may fail to be a @xmath84-cutting .
shallow_lemma.pstex_t this latter property is enforced as in @xcite , by further decomposing each cell @xmath147 of @xmath85 that is crossed by more than @xmath39 ranges of @xmath74 , using additional subsamples from the surfaces that cross @xmath147 .
specifically , for each cell @xmath147 of @xmath85 , let @xmath148 denote the subset of those ranges in @xmath74 that cross @xmath147 , and put @xmath149 .
if @xmath150 , we sample @xmath151 ranges from @xmath148 , construct the complement of the union of these ranges , decompose it into at most @xmath152 elementary cells , and clip them to within @xmath147 .
the resulting collection @xmath153 of subcells , over all cells @xmath147 of the original @xmath85 , clearly satisfies ( i ) .
the analysis of @xcite ( see also @xcite ) establishes an exponential decay property on the number of cells of @xmath85 that are crossed by more than @xmath154 ranges , as a function of @xmath155 .
specifically , as in @xcite , the expected number of such cells is @xmath156 , where @xmath157 is another random sample of @xmath74 , where each member of @xmath74 is chosen with probability @xmath158 .
this property implies , as usual @xcite , that @xmath153 is ( with high probability ) a @xmath84-cutting , and it also implies that the size of @xmath153 is still @xmath144 , assuming @xmath142 to be well behaved . since we have only refined the original cells of @xmath85 , the number of ranges that cover the complement of the union of the final cells is still @xmath46 .
@xmath140 a special case that arises frequently is where each range in @xmath74 is an upper ( or lower ) halfspace bounded by the graph of some continuous @xmath159-variate function . in this case
the complement @xmath56 of the union of @xmath40 ranges is the portion of space that lies below the _ lower envelope _ of the bounding graphs . in this case
, it suffices to decompose the graph of the lower envelope itself into at most @xmath102 elementary cells .
indeed , having done that , we can extend each cell @xmath147 within the envelope into the cell @xmath160 consisting of all points that lie vertically below @xmath147 .
the new cells decompose @xmath56 and are also elementary . as already discussed in the introduction , obtaining tight or nearly tight bounds for @xmath102
is still a major open problem for many instances of the above setup .
for example , decomposing an upper envelope of @xmath40 @xmath159-variate functions of constant description complexity into @xmath161 elementary cells is still open for any @xmath162 .
( this bound is best possible in the worst case , since it is the worst - case tight bound on the complexity of such an undecomposed envelope @xcite . )
the cases @xmath163 ( upper envelope of curves in the plane ) and @xmath164 ( upper envelope of 2-dimensional surfaces in 3-space ) are easy . in these cases
@xmath102 is proportional to the complexity of the envelope , which in the worst case is near - linear for @xmath163 and near - quadratic for @xmath164 @xcite . in higher dimensions ,
the only general - purpose bound known to date is the upper bound obtained by computing the _ vertical decomposition _ of the _ entire _ arrangement of the given surfaces , and extracting from it the relevant cells that lie on or above the envelope . in particular , for @xmath165 the bound is @xmath166 , as follows from the results of @xcite .
this leaves a gap of about a factor of @xmath40 between this bound and the bound @xmath167 on the complexity of the undecomposed envelope .
of course , in certain special cases , most notably the case of hyperplanes , as studied in @xcite , both the envelope and its decomposition have ( considerably ) smaller complexity . the situation with the complexity of the union of geometric objects is even worse . while considerable progress was recently made on many special cases in two and three dimensions ( see @xcite for a recent comprehensive survey ) , there are only very few sharp bounds on the complexity of unions in higher dimensions .
worse still , even when a sharp bound on the complexity of the union is known , obtaining comparable bounds on the complexity of a decomposition of the complement of the union is a much harder problem ( in @xmath168 dimensions ) . as an example
, the union of @xmath1 congruent infinite cylinders in 3-space is known to have near - quadratic complexity @xcite , but it is still an open problem whether its complement can be decomposed into a near - quadratic number of elementary cells .
[ [ partition - theorem - for - shallow - semi - algebraic - ranges . ] ] partition theorem for shallow semi - algebraic ranges .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we next apply the new shallow cutting lemma to construct an elementary cell partition of a given input point set @xmath0 , with respect to a specific set @xmath44 of ranges .
this is done in a fairly similar way to that in @xcite ( see also @xcite ) .
a major difference in handling the semi - algebraic case is the construction of a set @xmath44 of ranges that will be ( a ) small enough , and ( b ) representative of all shallow ( or empty ) ranges , in a sense discussed in detail below .
the method given in @xcite does not work in the general semi - algebraic case , and different , sometimes ad - hoc approaches need to be taken .
the following theorem summarizes the main part of the construction ( except for the construction of @xmath44 ) .
[ theo : part ] let @xmath0 be a set of @xmath1 points in @xmath2 , let @xmath74 be a family of semi - algebraic ranges of constant description complexity , and let @xmath40 be fixed .
let @xmath44 be another finite collection ( not necessarily a subset of @xmath74 ) of semi - algebraic ranges of constant description complexity with the following properties : ( i ) the ranges in @xmath44 are all @xmath45-shallow .
( ii ) the complement of the union of any @xmath38 ranges of @xmath44 can be decomposed into at most @xmath141 elementary cells , for any @xmath38 .
( iii ) any @xmath45-shallow range @xmath169 can be covered by the union of at most @xmath79 ranges of @xmath44 , where @xmath79 is a constant .
+ then there exists an elementary cell partition @xmath47 of @xmath0 , of size @xmath46 , into subsets of size roughly @xmath39 , such that the crossing number of any @xmath45-shallow range in @xmath74 with the cells of @xmath47 is either @xmath170 , if @xmath171 , for any fixed @xmath10 , or @xmath172 , otherwise . the proof , which , again , is similar to those in @xcite , proceeds through the following steps .
we first have : [ lemma : half_partition ] let @xmath0 be a set of @xmath1 points in @xmath2 , and @xmath173 a parameter .
let @xmath44 be a set of @xmath45-shallow ranges , with the property that the complement of the union of any subset of @xmath38 ranges of @xmath44 can be decomposed into at most @xmath141 elementary cells , for any @xmath38 .
then there exists a subset @xmath174 of at least @xmath175 points and an elementary cell partition @xmath94 for @xmath176 with @xmath177 for all @xmath100 , such that each range of @xmath44 crosses at most @xmath178 cells @xmath96 of @xmath47 .
_ we will inductively construct disjoint sets @xmath179 of size @xmath39 and elementary cells @xmath180 such that @xmath181 for each @xmath100 .
the construction terminates when @xmath182 .
suppose that @xmath183 have already been constructed , and set @xmath184 .
we construct @xmath97 as follows : for a range @xmath185 , let @xmath186 denote the number of cells among @xmath187 crossed by @xmath3 .
we define a weighted collection @xmath188 of ranges , so that each range @xmath185 appears with weight ( or multiplicity ) @xmath189 .
we put @xmath190 . by lemma [ lemma : shallow_cutting ] and by our assumption
that the function @xmath102 is well behaved , there exists a @xmath191-cutting @xmath192 for the weighted collection @xmath188 of size at most @xmath193 , for an appropriate choice of @xmath194 , with the following properties : the union of @xmath192 contains the complement of the union of @xmath44 , and the complement of the union of @xmath192 is contained in the union of @xmath195 ranges of @xmath44 .
since all these ranges are @xmath45-shallow , the number of points of @xmath0 not in the union of @xmath192 is at most @xmath196 , and our assumptions on @xmath102 imply that this is smaller than @xmath197 , if we choose @xmath77 appropriately . since we assume that @xmath198 , it follows that at least @xmath197 points of @xmath199 lie in the union of the at most @xmath193 cells of @xmath192 . by the pigeonhole principle ,
there is a cell @xmath96 of @xmath192 containing at least @xmath39 points of @xmath199 .
we take @xmath97 to be some subset of @xmath200 of size exactly @xmath39 , and make @xmath96 the cell in the partition which contains @xmath97 .
we next establish the asserted bound on the crossing numbers between the ranges of @xmath44 and the elementary cells @xmath201 , in the following standard manner .
the final weight @xmath202 of a range @xmath185 with crossing number @xmath48 ( with respect to the final partition ) is @xmath203 . on the other hand , each newly added cell @xmath96
is crossed by ranges of @xmath44 of total weight @xmath204 , because @xmath96 is an elementary cell of the corresponding weighted @xmath191-cutting @xmath192 .
the weight of each of these crossing ranges is doubled at the @xmath100-th step , and the weight of all the other ranges remains unchanged .
thus @xmath205 .
hence , for each range @xmath206 we have @xmath207 .
@xmath140 [ [ discussion . ] ] discussion .
+ + + + + + + + + + + the limitation of lemma [ lemma : half_partition ] is that the bound that it derives ( a ) applies only to ranges in @xmath44 , and ( b ) includes the term @xmath208 .
an ingenious component of the analysis in @xcite overcomes both problems , by choosing a test set @xmath44 of ranges whose size is only polynomial in @xmath40 ( and , in particular , is independent of @xmath1 ) , which is nevertheless sufficiently representative of all shallow ranges , in the sense that the crossing number of any @xmath45-shallow range is @xmath209 .
this implies that lemma [ lemma : half_partition ] holds for _ all _ shallow ranges , with the stronger bound which does not involve @xmath208 .
unfortunately , the technique of @xcite does not extend to the case of semi - algebraic ranges , as it crucially relies on the linearity of the ranges .
the following lemma gives a sufficient condition for a test set @xmath44 to be representative of the relevant shallow ranges , in the sense that @xmath44 satisfies the assumptions made in theorem [ theo : part ] .
that is : [ lemma : const_cover ] let @xmath0 be a set of @xmath1 points in @xmath2 , and let @xmath74 be a family of semi - algebraic ranges with constant description complexity . consider an elementary - cell partition @xmath210 of @xmath0 such that @xmath211 for each @xmath100
let @xmath44 be a finite set of @xmath45-shallow ranges ( not necessarily ranges of @xmath74 ) , so that the maximal crossing number of a range @xmath212 with respect to @xmath47 is @xmath48 .
then , for any range @xmath87 which is contained in the union of at most @xmath79 ranges of @xmath44 ( for some constant @xmath79 ) , the crossing number of @xmath50 is at most @xmath213 .
_ let @xmath87 be a range for which there exist @xmath79 ranges @xmath214 of @xmath44 such that @xmath215 . then
, if @xmath50 crosses a cell @xmath96 of @xmath47 , then at least one of the covering ranges @xmath216 must either cross @xmath96 or fully contain @xmath96 .
the number of cells of @xmath47 that can be crossed by any single @xmath216 is at most @xmath48 , and each @xmath216 can fully contain at most one cell of @xmath47 ( because @xmath216 is @xmath45-shallow ) . in the construction of the partition , we can even rule out the possibility that a range @xmath216 fully contains a cell of @xmath47 .
this however has no effect on the asymptotic bounds that the analysis derives . ] hence , the overall number of cells of @xmath47 that @xmath50 can cross is at most @xmath213 , as asserted .
@xmath140 [ [ proof - of - theoremtheopart . ] ] proof of theorem [ theo : part ] .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + apply lemma [ lemma : half_partition ] to the input set @xmath217 , with parameter @xmath218 .
this yields an elementary - cell partition @xmath219 for ( at least ) half of the points of @xmath220 , which satisfies the properties of that lemma .
let @xmath221 denote the set of the remaining points of @xmath220 , and set @xmath222 .
apply lemma [ lemma : half_partition ] again to @xmath221 with parameter @xmath223 , obtaining an elementary cell partition @xmath224 for ( at least ) half of the points of @xmath221 .
we iterate this process @xmath225 times , until the set @xmath226 has fewer than @xmath39 points .
we take @xmath47 to be the union of all the elementary - cell partitions @xmath227 formed so far , together with one large cell containing all the remaining points of @xmath226 .
the resulting elementary - cell partition of @xmath0 consists of at most @xmath228 subsets , each of size at most @xmath39 .
the crossing number of a range in @xmath44 is , by lemma [ lemma : half_partition ] , @xmath229 our assumptions on @xmath142 imply that if @xmath171 , for any fixed @xmath10 , the first terms add up to @xmath230 ; otherwise we can bound their sum by @xmath231 .
hence , by the properties of @xmath44 and by lemma [ lemma : const_cover ] , the crossing number of any empty range is also @xmath232 or @xmath233 , respectively .
@xmath140 [ [ partition - trees - and - reporting - or - emptiness - searching . ] ] partition trees and reporting or emptiness searching . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as in the classical works on range searching @xcite , we apply theorem [ theo : part ] recursively , and obtain a _ partition tree _
@xmath234 , where each node @xmath235 of @xmath234 stores a subset @xmath236 of @xmath0 and an elementary cell @xmath237 enclosing @xmath236 .
the children of a node @xmath235 are obtained from an elementary cell partition of @xmath236each of them stores one of the resulting subsets of @xmath236 and its enclosing cell . at the leaves ,
the size of the subset that is stored is @xmath46 . testing a range @xmath50 for emptiness
is done by searching with @xmath50 in @xmath234 . at each visited node @xmath235 , where @xmath238 , we test whether @xmath239 , in which case @xmath50 is not empty . otherwise , we find the children of @xmath235 whose cells are intersected by @xmath50 . if there are too many of them we know that @xmath50 is not empty .
otherwise , we recurse at each child . reporting is performed in a similar manner . if @xmath240 we output all of @xmath237 .
otherwise , we find the children of @xmath235 whose cells are intersected by @xmath50 . if there are too many of them we know that @xmath50 is not @xmath241-shallow ( with respect to @xmath236 ) , so , if @xmath40 is a constant , we can afford to check every element of @xmath236 for containment in @xmath50 , and output those points that do lie in @xmath50 .
if there are not too many children , we recurse in each of them .
the efficiency of the search depends on the function @xmath141 .
if @xmath242 then an emptiness query takes @xmath243 time , and a reporting query takes @xmath244 , where @xmath77 is the output size . thus making @xmath142 ( i.e. , @xmath27 ) small is the main challenge in this technique . [ [ a - general - recipe - for - constructing - good - test - sets . ] ] a general recipe for constructing good test sets .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath74 be the given collection of semi - algebraic ranges of constant description complexity .
as above , we assume that each range @xmath87 has @xmath77 degrees of freedom , for some constant parameter @xmath77 , so it can be represented as a point @xmath245 in a @xmath77-dimensional parametric space , which , for convenience , we denote as @xmath246 .
each input point @xmath247 is mapped to a region @xmath248 , which is the locus of all points representing ranges which contain @xmath75 .
we fix a parameter @xmath249 , and choose a random sample @xmath114 of @xmath250 points of @xmath0 , where @xmath251 is a sufficiently large constant .
we form the set @xmath252 , construct the arrangement @xmath253 , and let @xmath254 denote the region consisting of all points contained in at most @xmath27 ranges of @xmath255 , where @xmath256 and @xmath257 is an absolute constant that we will fix later .
we decompose @xmath258 into elementary cells , using , e.g. , vertical decomposition @xcite . in the worst case
, we get @xmath259 elementary cells @xcite .
is sufficient for our purpose . ]
let @xmath147 be one of these cells .
we associate with @xmath147 a generalized range @xmath260 in @xmath2 , which is the union @xmath261 . since @xmath147 has constant description complexity , as do the ranges of @xmath74 , it is easy to show that @xmath260 is also a semi - algebraic set of constant description complexity ( see @xcite ) .
we define the test set @xmath44 to consist of all the generalized ranges @xmath260 , over all cells @xmath147 in the decomposition of @xmath258 , and claim that , with high probability ( and with an appropriate choice of @xmath257 ) , @xmath44 is a good test set , in the following three aspects .
\(i ) * compactness .
* @xmath262 ; that is , the size of @xmath44 is polynomial in @xmath40 and independent of @xmath1 .
\(ii ) * shallowness . *
each range @xmath260 in @xmath44 is @xmath263-shallow with respect to @xmath0 , for some constant parameter @xmath136 .
\(iii ) * containment .
* every _ @xmath45-shallow _ range @xmath87 is contained in a single range @xmath260 of @xmath44 .
property ( i ) is obvious .
consider the range space @xmath264 , where @xmath265 consists of all generalized ranges @xmath260 , over all elementary cells @xmath147 of the form arising in the above vertical decomposition .
it is a fairly easy exercise to show that @xmath264 also has finite vc - dimension .
see , e.g. , @xcite . by theorem [ epsshallow ] ,
if @xmath251 is a sufficiently large constant ( proportional to the vc - dimension of @xmath264 ) then @xmath114 is a shallow @xmath84-net for both range spaces @xmath266 and @xmath264 , with high probability , so we assume that @xmath114 is indeed such a shallow @xmath84-net .
let @xmath267 .
note that any point @xmath268 in @xmath260 lies in a range @xmath87 with @xmath269 . by definition
, @xmath245 also belongs to @xmath248 , and so @xmath248 crosses or fully contains @xmath147 . since @xmath147 is @xmath270-shallow in @xmath253
, it is fully contained in at most @xmath271 regions @xmath248 , for @xmath272 ( and is not crossed by any such region ) .
hence @xmath273 , so , since @xmath114 is a shallow @xmath84-net for @xmath264 , we have @xmath274 , so @xmath260 is @xmath275-shallow , which establishes ( ii ) . for ( iii ) , let @xmath169 be an @xmath45-shallow range . since @xmath114 is a shallow @xmath84-net for @xmath266 , and @xmath276 , we have @xmath277 . hence , with @xmath278 , @xmath279
, so there is a cell @xmath147 of the decomposition which contains @xmath50 , which , by construction , implies that @xmath280 , thus establishing ( iii ) . to make @xmath44 a really good test set
, we also need the following fourth property : \(iv ) * efficiency .
* there exists a good bound on the associated function @xmath141 , bounding the size of a decomposition of the complement of the union of any @xmath38 ranges of @xmath44 .
the potentially rather complex shape of these generalized ranges makes it harder to obtain , in general , a good bound on @xmath142 . in what follows
, we manage to use this general recipe in two of our four applications ( ray shooting amid balls and range fullness searching ) , with good bounds on the corresponding functions @xmath281 . in two other planar applications ( range emptiness searching with fat triangles and with circular caps ) , we abandon the general technique , and construct ad hoc good test sets .
[ [ remark ] ] remark : + + + + + + + in the preceding construction , we wanted to make sure that every @xmath45-shallow range @xmath169 is covered by a range of @xmath44 .
if we only need this property for _ empty _ ranges @xmath50 ( which is the case for emptiness testing ) , it suffices to consider only the @xmath282-level of @xmath253 , i.e. , the complement of the union of @xmath255 .
other than this simplification , the construction proceeds as above .
let @xmath0 be a set of @xmath1 points in 3-space , and let @xmath74 be a set of convex ranges of constant description complexity .
we wish to preprocess @xmath0 in near - linear time into a data structure of linear size , so that , given a query range @xmath169 , we can efficiently determine whether @xmath50 contains all the points of @xmath0 .
alternatively , we want to report all the points of @xmath0 that lie outside @xmath50 .
( this is clearly a special case of range emptiness searching or range reporting , if one considers the complements of the ranges in @xmath74 . ) for simplicity , we only focus on the range fullness problem ; the extension to reporting `` outliers '' is similar to the standard treatment of reporting queries , as discussed earlier .
we present a solution to this problem , with @xmath16 query time , thereby improving over the best known general bound of @xmath13 , given in @xcite , which applies to any range searching ( e.g. , range counting ) with semi - algebraic sets ( of constant description complexity ) in @xmath6 . to apply our technique to this problem we first need to build a good test set . since fullness searching is complementary to emptiness searching , we need a property complementary to that assumed in theorem [ theo : part ] ( see also lemma [ lemma : const_cover ] ) .
in fact , we will enforce the property that every full range @xmath50 fully contains a single range of @xmath44 , which is `` almost full '' ( contains at least @xmath283 points of @xmath0 ) .
as above , assuming the ranges of @xmath74 to have @xmath77 degrees of freedom , we map each range @xmath87 to a point @xmath245 in @xmath246 .
a point @xmath284 is mapped to a region @xmath248 which is the locus of all the points @xmath245 that correspond to ranges @xmath50 which contain @xmath75 .
we fix @xmath83 , take a random sample @xmath114 of @xmath285 points of @xmath0 ( with a sufficiently large constant of proportionality ) , construct the _ intersection _ @xmath286 , and decompose it into elementary cells . for each resulting cell @xmath3 , let @xmath287 denote the _ intersection _ @xmath288 . as above , since @xmath3 has constant description complexity , @xmath287 is a semi - algebraic set of constant description complexity .
note that , since the ranges in @xmath74 are convex , each range @xmath287 is also convex ( albeit of potentially more complex shape than that of the original ranges ) .
define the test set @xmath44 to consist of all the generalized ranges @xmath287 , over all cells @xmath3 in the decomposition of @xmath289 .
we argue that @xmath44 satisfies all four properties required from a good test set : ( i ) compactness : as above , the size of @xmath44 is polynomial in @xmath40 ( it is at most @xmath259 ) .
( ii ) shallowness ( or , rather , `` almost fullness '' ) : for each cell @xmath3 and any @xmath87 with @xmath290 , @xmath245 lies in all the sets @xmath248 , for @xmath291 , and thus @xmath292 . by construction , we also have @xmath293 .
apply the @xmath12-net theory @xcite to the range space @xmath294 , where the ranges of @xmath295 are complements of ranges of the same form as the ranges @xmath287 . since @xmath296 for each cell @xmath3 in the decomposition
, we have , with high probability , the property that for each cell @xmath3 of @xmath289 , @xmath287 contains at least @xmath283 points of @xmath0 , so it is an `` almost full '' range .
( iii ) containment : let @xmath87 be a full range . then , in particular , @xmath292 . then @xmath297 , and let @xmath3 be the cell of @xmath289 containing @xmath245 .
then , by construction , @xmath298 .
( iv ) efficiency : finally , we show that the complexity of a decomposition of the intersection of any @xmath38 ranges in @xmath44 , is @xmath299 , so @xmath300 . [
claim : decomposition_of_convex ] let @xmath44 be a set of convex `` almost full '' ranges , each containing at least @xmath301 points of @xmath0 . the intersection , @xmath56 , of any @xmath38 ranges @xmath302
can be decomposed into @xmath299 elementary cells .
_ since all ranges in @xmath44 are convex , @xmath56 is a convex set too .
assume , for simplicity of presentation , that @xmath56 is nonempty and has nonempty interior , and fix a point @xmath303 in that interior .
we can regard the boundary of each @xmath304 as the graph of a bivariate function @xmath305 in spherical coordinates about @xmath303 .
then @xmath306 is the graph of the lower envelopes of these functions .
since the @xmath304 s have constant description complexity , ( the graph of ) each @xmath307 is also a semi - algebraic set of constant description complexity .
hence the combinatorial complexity of @xmath306 is @xmath299 @xcite .
moreover , since @xmath306 is 2-dimensional , we can partition it into @xmath299 trapezoidal - like elementary cells , using a variant of the vertical decomposition technique , and then extend each such cell @xmath308 to a 3-dimensional cone - like cell @xmath147 , which is the union of all segments connecting @xmath303 to the points of @xmath308 .
the resulting cells @xmath147 constitute a decomposition of @xmath56 into @xmath299 elementary cells , as claimed .
@xmath140 using the machinery developed in the preceding section , we therefore obtain the following result .
[ thm : full ] let @xmath0 be a set of @xmath1 points in @xmath6 , and let @xmath74 be a family of convex ranges of constant description complexity
. then one can construct , in near linear time , a data structure of linear size so that , for any range @xmath87 , it can determine , in @xmath16 time , whether @xmath50 is full . [ [ reporting - outliers . ] ] reporting outliers .
+ + + + + + + + + + + + + + + + + + + to extend the above approach to the problem of reporting outliers , we apply a construction similar to that in the `` general recipe '' presented above .
that is , we take the @xmath271 deepest levels of @xmath309 , for an appropriate constant @xmath257 , decompose them into elementary cells , and construct a generalized range @xmath287 for each of these cells @xmath3 .
the general machinery given above implies the following result : [ thm : full_report ]
let @xmath0 be a set of @xmath1 points in @xmath6 , and let @xmath74 be a family of convex ranges of constant description complexity .
then one can construct , in near linear time , a data structure of linear size so that , for any range @xmath87 , it can report the points of @xmath0 in the complement of @xmath50 , in @xmath60 time , where @xmath27 is the query output size . a useful application of the data structure of theorem [ thm : full ] is to _ farthest point queries_. in such a problem we are given a set @xmath0 of @xmath1 points in @xmath6 , and wish to preprocess it , in near - linear time , into a data structure of linear size , so that , given a convex query object @xmath303 ( from some fixed class of objects with constant description complexity ) , we can efficiently find the point of @xmath0 farthest from @xmath303 .
we solve this problem using parametric searching @xcite .
the corresponding decision problem is : given the query object @xmath303 and a distance @xmath310 , determine whether the minkowski sum @xmath311 is full , where @xmath312 is the ball of radius @xmath310 centered at the origin .
the smallest @xmath310 with this property is the distance to the farthest point from @xmath303 . with an appropriate small - depth parallel implementation of this decision problem ,
the parametric searching also takes time @xmath16 . reporting the @xmath27 farthest points from @xmath303 , for any parameter @xmath27 ,
can be done in @xmath60 time , using a simple variant of this technique .
let @xmath0 be a set of @xmath1 points in @xmath2 .
we wish to find the triangle whose vertices belong to @xmath0 and whose area ( respectively , perimeter , height ) is maximal .
this problem is a useful subroutine in path approximation algorithms ; see @xcite .
daescu and serfling @xcite gave an @xmath59-algorithm for the 3-dimensional largest - area triangle .
in @xmath313 dimensions , the running time is @xmath314 . in @xmath6 , our technique , without any additional enhancements , yields the improved bound @xmath315 , using the following straightforward procedure .
for each pair of points @xmath316 , we find the farthest point @xmath317 from the line @xmath318 , compute the area of @xmath319 , and output the largest - area triangle among those triangles .
the procedure performs farthest - point queries from @xmath320 lines , for a total cost of @xmath315 , as claimed .
we can improve this solution , using the following standard decomposition technique , to an algorithm with running time @xmath17 .
first , the approach just described performs @xmath321 farthest - point queries on a set of @xmath114 points in time @xmath322 , where the second term is the preprocessing cost of preparing the data structure . before continuing , we note the following technical issue .
recall that we find the farthest point from a query line @xmath15 by drawing a cylinder @xmath323 around @xmath15 , whose radius @xmath310 is the smallest ( unknown ) radius for which @xmath323 contains @xmath0 .
the concrete value of @xmath310 is found using parametric searching . in the approach that we follow now
, we will execute in parallel @xmath320 different queries , each with its own @xmath310 , so care has to be taken when running the parametric search with this multitude of different unknown values of @xmath310 .
while there are several alternative solutions to this problem , we use the following one , which seems the cleanest .
let @xmath324 be a fixed parameter .
for each pair @xmath325 of distinct points of @xmath0 , let @xmath326 denote the cylinder whose axis passes through @xmath327 and @xmath328 and whose radius is @xmath329 . in the decision procedure ,
we specify the value of @xmath330 , and perform @xmath320 range fullness queries with the cylinders @xmath326 .
if all of them are found to be full , then @xmath331 , where @xmath332 is the ( unknown ) maximal area of a triangle spanned by @xmath0 ; otherwise @xmath333 .
( with a somewhat finer implementation , we can also distinguish between the cases @xmath334 and @xmath335 ; we omit the details of this refinement . ) to implement the decision procedure , we apply a duality transform , where each cylinder @xmath58 in 3-space is mapped to a point @xmath336 , where @xmath337 is some parametrization of the axis of @xmath58 and @xmath310 is its radius . in this dual parametric 5-space , a point @xmath284 is mapped to a surface @xmath338 , which is the locus of all ( points representing ) cylinders which contain @xmath75 on their boundary . note that the portion of space below ( resp . , above ) @xmath338 , in the @xmath310-direction , consists of points dual to cylinders which do not contain ( resp . , contain ) @xmath75 .
let @xmath339 .
fix some sufficiently large but constant parameter @xmath340 , and construct a @xmath341-cutting @xmath85 of @xmath342 , using the vertical decomposition of a random sample of @xmath343 surfaces of @xmath344 ( see , e.g. , @xcite ) . as follows from @xcite , the combinatorial complexity of @xmath85
is @xmath345 .
we distribute the @xmath320 points dual to the query cylinders among the cells of @xmath85 , in brute force , and also find , in equally brute force , for each cell @xmath147 the subset @xmath346 of surfaces which cross @xmath147 .
we ignore cells which fully lie below some surface of @xmath344 , because cylinders whose dual points fall in such a cell can not be full ( the decision algorithm stops as soon as such a point ( cylinder ) is detected ) . for each of the remaining cells @xmath147 , we repeat this procedure with the subset of the points dual to the surfaces in @xmath346 and with the subset of cylinders whose
dual points lie in @xmath147 .
we keep iterating in this manner until we reach cuttings whose cells are crossed by at most @xmath39 dual surfaces , where @xmath40 is some ( non - constant ) parameter that we will shortly fix .
as is easily checked , the overall number of cells in these cuttings is @xmath347 .
we then run the preceding weaker procedure on each of the resulting cells @xmath147 , with the set @xmath348 of points dual to the surfaces which cross @xmath147 and with the set @xmath349 of cylinders whose dual points lie in @xmath147 . letting @xmath350 denote the number of these cylinders , the overall cost of the second phase of the procedure is @xmath351 since @xmath340 is a constant , the overall cost of the first phase is easily seen to be proportional to the overall size of the resulting subproblems , which is @xmath352 .
overall , the cost is thus @xmath353 choosing @xmath354 , this becomes @xmath17 . running a generic version of this decision procedure in parallel
is fairly straightforward .
the cuttings themselves depend only on the dual surfaces , which do not depend on @xmath332 , so we can construct them in a concrete , non - parametric fashion . locating the points dual to the query cylinders can be done in parallel , and , since @xmath340 is a constant , this takes constant parallel depth for each of the logarithmically many levels of cuttings
. the second phase can also be executed in parallel in an obvious manner . omitting the further easy details ,
we conclude that the overall algorithm also takes @xmath17 time .
[ [ largest - perimeter - triangle . ] ] largest - perimeter triangle .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + the above technique can be adapted to yield efficient solutions of several problems of a similar flavor .
for example , consider the problem of computing the largest - perimeter triangle among those spanned by a set @xmath0 of @xmath1 points in @xmath6 .
here , for each pair @xmath325 of points of @xmath0 , and for a specified perimeter @xmath355 , we construct the ellipsoid of revolution @xmath356 , whose boundary is the locus of all points @xmath65 satisfying @xmath357 .
( here , of course , we only consider pairs @xmath325 with @xmath358 . )
we now run @xmath320 range fullness queries with these ellipsoids , and report that @xmath359 if at least one of these ellipsoids in not full , or @xmath360 otherwise , where @xmath361 is the largest perimeter .
the efficient implementation of this procedure is carried out similar to the preceding algorithm , except that here the dual representation of our ellipsoids require six degrees of freedom , to specify the foci @xmath327 and @xmath328 . unlike the previous case , the dual surfaces @xmath338 do depend on @xmath355 , so , in the generic implementation of the decision procedure we also need to construct the various @xmath341-cuttings in a generic , parallel manner . if we add @xmath355 as a seventh degree of freedom , but then the overall performance of the algorithm deteriorates .
] however , since @xmath340 is a constant , this is easy to do in constant parallel depth per cutting .
a @xmath341-cutting in @xmath362 has complexity @xmath363 @xcite .
a modified version of the preceding analysis then yields : the largest - perimeter triangle among those spanned by a set of @xmath1 points in @xmath6 can be computed in @xmath364 time .
[ [ largest - height - triangle . ] ] largest - height triangle .
+ + + + + + + + + + + + + + + + + + + + + + + + in this variant , we wish to compute the triangle with largest height among those determined by a set @xmath0 of @xmath1 points in @xmath6 . here , for each pair @xmath325 of points of @xmath0 , and for a specified height @xmath365 , we construct the cylinder @xmath366 , whose axis passes through @xmath327 and @xmath328 and whose radius is @xmath365 .
we run @xmath320 range fullness queries with these cylinders , and report that @xmath367 if at least one of these cylinders in not full , or @xmath368 otherwise , where @xmath369 is the desired largest height .
the efficient implementation of this procedure is carried out as above , except that here the dual representation of these cylinders require only four degrees of freedom , once @xmath365 is specified . as in the preceding case , here too the surfaces of @xmath344 also depend on @xmath365 , so we need a generic parallel procedure for constructing @xmath341-cuttings for these surfaces , which however is not difficult to achieve , since @xmath340 is a constant .
we omit the simple routine details .
since a @xmath341-cutting in @xmath51 has complexity @xmath370 @xcite , a modified version of the preceding analysis then yields : the largest - height triangle among those spanned by a set of @xmath1 points in @xmath6 can be computed in @xmath371 time .
[ [ further - extensions . ] ] further extensions .
+ + + + + + + + + + + + + + + + + + + we can extend this machinery to higher dimensions , although its performance deteriorates as the dimension grows .
the range fullness problem in @xmath2 , for @xmath162 , can be handled in much the same way as in the 3-dimensional case .
when extending claim [ claim : decomposition_of_convex ] , we have an intersection of @xmath38 convex sets of constant description complexity in @xmath2 , and we can regard the boundary of the intersection as the lower envelope of @xmath38 @xmath159-variate functions of constant description complexity , each representing the boundary of one of the input convex sets , in spherical coordinates about some fixed point in the intersection . the complexity of the lower envelope is @xmath372 @xcite .
however , we need to decompose the region below the envelope into elementary cells , and , as already noted , the only known general - purpose technique for doing so is to decompose the entire arrangement of the graphs of the @xmath38 boundary functions , and select the cells below the lower envelope .
the complexity of such a decomposition is @xmath373 @xcite .
this implies that @xmath374 .
the rest of the analysis , including the construction of a good test set , is done in essentially the same manner .
hence , using the machinery of the previous section , we obtain : [ fulld ] let @xmath0 be a set of @xmath1 points in @xmath2 , for @xmath162 , and let @xmath74 be a family of convex ranges of constant description complexity
. then one can construct , in near linear time , a data structure of linear size so that , for any range @xmath87 , it can determine , in @xmath375 , whether @xmath50 is full .
[ [ finding - the - largest - area - triangle - in - realsd . ] ] finding the largest - area triangle in @xmath2 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath0 be a set of @xmath1 points in @xmath2 , for @xmath162 , and consider the problem of finding the largest - area triangle spanned by @xmath0 .
we apply the same method as in the 3-dimensional case , whose main component is a decision procedure which tests @xmath320 cylinders for fullness .
a cylinder ( with a line as an axis ) in @xmath2 has @xmath376 degrees of freedom , so the dual representation of our @xmath320 cylinders is as points in @xmath377 .
the best known bound on the complexity of a @xmath84-cutting in this space is @xmath378 . applying this bound and the bound in theorem [ fulld ] ,
the overall cost of the decision procedure is @xmath379 optimizing the value of @xmath40 , and applying parametric searching , we get an algorithm for the maximum - area triangle in @xmath2 with running time @xmath380 we can extend the other problems ( largest - perimeter or largest - height triangles ) in a similar manner , and can also obtain algorithms for solving higher - dimensional variants , such as computing the largest - volume tetrahedron or higher - dimensional simplices .
we omit the straightforward but tedious analysis , and the resulting cumbersome - looking bounds .
let @xmath381 be a set of @xmath1 balls in 3-space . we show how to preprocess @xmath381 in near - linear time into a data structure of linear size , so that , given a query ray @xmath310 , the first ball that @xmath310 hits can be computed in @xmath13 time , improving the general bound @xmath14 mentioned in the introduction .
as already noted , we use the parametric - searching technique of agarwal and matouek @xcite , which reduces the problem to that of efficiently testing whether a query segment @xmath382 intersects any ball in @xmath381 , where @xmath65 is the origin of @xmath310 and @xmath383 is a parametric point along @xmath310 .
[ [ parametric - representation - of - balls - and - segments . ] ] parametric representation of balls and segments .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we move to a parametric 4-dimensional space , in which balls in 3-space are represented by points , so that a ball with center at @xmath52 and radius @xmath40 is mapped to the point @xmath53 . a segment @xmath33 , or for that matter , any closed nonempty set @xmath384 of constant description complexity , is mapped to a surface @xmath55 , which is the locus of all points representing balls that touch @xmath56 but are openly disjoint from @xmath56 . by construction
, @xmath55 is the graph of a totally defined continuous trivariate function @xmath385 , which is semi - algebraic of constant description complexity .
moreover , points below ( resp . , above ) @xmath55 represent balls which are disjoint from @xmath56 ( resp . , intersect @xmath56 ) .
moreover , for any such set @xmath56 , @xmath386 is , by definition , the ( euclidean ) distance of @xmath65 from @xmath56 .
hence , given a collection @xmath387 of @xmath38 sets , the minimization diagram of the surfaces @xmath388 ( that is , the projection onto the 3-space @xmath389 of the lower envelope of these surfaces ) is the nearest - neighbor voronoi diagram of @xmath390 .
we use this property later on , in deriving a sharp bound on the resulting function @xmath281 .
[ [ building - a - test - set - for - segment - emptiness . ] ] building a test set for segment emptiness .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + here we use the general recipe for constructing good test sets , which covers each empty segment @xmath33 by a fairly complex `` canonical '' empty region @xmath56 , which has nonetheless constant description complexity . in parametric 4-space , each such region @xmath56
is mapped to the upper halfspace above the corresponding surface @xmath55 ; this is the set of all balls that intersect @xmath56 .
the complement of the union of @xmath38 such ranges is the portion of 4-space below the lower envelope of the corresponding surfaces @xmath391 . using the connection between this envelope and the voronoi diagram of the @xmath392 s
, we are able to decompose ( the diagram and thus ) the complement of the union into @xmath393 elementary cells . in more detail
, the construction proceeds as follows .
we start by choosing a random sample @xmath114 of @xmath285 balls of @xmath381 , to construct a test set @xmath44 for empty segment ranges , with respect to @xmath114 . while we do not have a clean , explicit geometric definition of these ranges , they will satisfy , as above , all the four requirements from a good test set .
also , we spell out the adaptation of the general recipe to the present scenario , to help the reader see through one concrete application of the general recipe .
specifically , we move to a dual space , in which segments in 3-space are represented as points .
segments in 3-space have six degrees of freedom ; for example , we can represent a segment by the coordinates of its two endpoints . the dual space is therefore 6-dimensional .
each ball @xmath394 is mapped to a surface @xmath395 , which is the locus of all points representing segments which touch @xmath396 but do not penetrate into its interior ; that is , either they are tangent to @xmath397 , at a point in their relative interior , or they have an endpoint on @xmath396 but are openly disjoint from @xmath397 .
let @xmath255 denote the collection of the surfaces dual to the balls of @xmath114 .
construct a @xmath84-cutting of @xmath253 , which consists of @xmath398 elementary cells @xcite .
each cell @xmath147 has the property that all points in @xmath147 represent segments which meet a fixed set of balls from among the balls in @xmath114 and avoid all other balls of @xmath114 ; the set depends only on @xmath147 . for each cell
@xmath147 whose corresponding set of balls is _ empty _ , we define @xmath399 to be the union , in 3-space , of all segments @xmath33 whose dual points lie in @xmath147 . since @xmath147 is an elementary cell , @xmath399 is a semi - algebraic set of constant description complexity ( see , e.g. , @xcite ) .
moreover , @xmath399 is an _
@xmath114-empty _ region , in the sense that it is openly disjoint from all the balls in @xmath114 . since we have to work in parametric 4-space , we map each region @xmath399 of the above kind into a range @xmath260 in 4-space , which is the locus of all ( points representing ) balls which intersect @xmath399 . as discussed above , @xmath260 is the upper halfspace bounded by the graph @xmath400 of the distance function from points in @xmath6 to @xmath399 .
we define the desired test set @xmath44 to consist of all the ranges @xmath260 , as just defined , and argue that @xmath44 indeed satisfies all four properties required from a good test set : ( a ) compactness : @xmath401 , so its size is small .
( b ) shallowness : with high probability , each range in @xmath44 is @xmath45-shallow , since it does not contain any point representing a ball in @xmath114 ( and we assume that the sample @xmath114 does indeed have this property , which makes all the ranges in @xmath44 @xmath45-shallow ) .
( c ) containment : each empty segment is also @xmath114-empty , so its dual point lies in some cell @xmath147 of the cutting , whose associated subset of balls is empty . by construction , we have @xmath402 .
that is , any ball intersecting @xmath33 also intersects @xmath399 , so the range in 4-space that @xmath33 defines is contained in @xmath260 , i.e. , in a single range of @xmath44 .
( d ) efficiency : the complement of the union of any @xmath38 ranges in @xmath44 can be decomposed into @xmath403 elementary cells .
the proof of ( d ) proceeds as follows .
the complement of the union of @xmath38 ranges , @xmath404 , is the region below the lower envelope of the corresponding surfaces @xmath405 . to decompose this region
, it suffices to produce a decomposition of the 3-dimensional minimization diagram of these surfaces , and extend each of the resulting cells into a semi - unbounded vertical prism , whose `` ceiling '' lies on the envelope .
the combinatorial complexity of the minimization diagram of a collection @xmath406 of @xmath38 trivariate functions of constant description complexity is are shaped , nor how `` wildly '' they can intersect one another , as long as each of them has constant description complexity . ] @xmath403 @xcite . moreover , as noted above , the minimization diagram is the euclidean nearest - neighbor voronoi diagram of @xmath390 .
we can decompose each cell @xmath407 of the diagram ( or , more precisely , the portion of the cell outside the union of the @xmath408 s ) using its _ star - shapedness _ with respect to its `` site '' @xmath408 ; that is , for any point @xmath409 , the segment connecting @xmath75 to its nearest point on @xmath408 is fully contained in @xmath410 . as is easy to verify
, this property holds regardless of the shape , or intersection pattern , of the regions in @xmath390 .
we first decompose the 2-dimensional faces bounding @xmath411 into elementary cells , using , e.g. , an appropriate variant of 2-dimensional vertical decomposition , and then take each such cell @xmath412 and extend it to a cell @xmath413 , which is the union of all segments , each connecting a point in @xmath412 to its nearest point on @xmath408 .
the resulting cells , obtained by applying this decomposition to all cells of the diagram , form a decomposition of the portion of the diagram outside the union of the @xmath408 s , into a total of @xmath403 elementary cells , as desired .
the union of the @xmath408 s themselves , being a subcollection of cells of a 3-dimensional arrangement of @xmath38 regions of constant description complexity , can also be decomposed into @xmath403 cells , using standard results on vertical decomposition in three dimensions @xcite .
using lemma [ lemma : const_cover ] and the machinery of section [ section : range_emptiness ] , in conjunction with the parametric searching technique of @xcite , we thus obtain the following theorem . [
main_balls ] ray shooting amid @xmath1 balls in 3-space can be performed in @xmath13 time , using a data structure of @xmath7 size , which can be constructed in @xmath8 time .
[ [ remark-1 ] ] remark : + + + + + + + in the preceding description , we only considered _ empty _ ranges .
if desired , we can extend the analysis to obtain a data structure which also supports `` reporting queries '' , in which we want to report the first @xmath27 balls hit by a query ray .
we omit the details of this straightforward extension .
[ [ fat - triangle - reporting - and - emptiness - searching . ] ] fat triangle reporting and emptiness searching .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath414 be a fixed constant , and let @xmath0 be a set of @xmath1 points in the plane .
we wish to preprocess @xmath0 , in @xmath8 time , into a data structure of size @xmath7 , which , given an @xmath19-fat query triangle @xmath50 ( which , as we recall , is a triangle all of whose angles are at least @xmath19 ) , can determine in @xmath21 time whether @xmath415 , or report in @xmath23 time the points of @xmath0 in @xmath50 , where @xmath416 .
to do so , we need to construct a good test set @xmath44 .
we use the following `` canonization '' process ( an ad - hoc process , not following the general recipe of section [ section : range_emptiness ] ) . as above , we apply the construction to a random sample @xmath114 of @xmath285 points of @xmath0 . for simplicity ,
we first show how to canonize _ empty _ triangles , and then extend the construction to shallow triangles .
( as before , the first part suffices for emptiness searching , whereas the second part is needed for reporting queries . )
let @xmath20 be an @xmath19-fat empty triangle , which is then also @xmath114-empty .
we expand @xmath20 homothetically , keeping one vertex fixed and translating the opposite side away , until it hits a point @xmath417 of @xmath114 .
we then expand the new triangle homothetically from a second vertex , until the opposite side hits a second point @xmath418 of @xmath114 , and then apply a similar expansion from the third vertex , making the third edge of the triangle touch a third point @xmath419 of @xmath114 .
we end up with an @xmath114-empty triangle @xmath420 , homothetic to , and containing , @xmath20 , each of whose sides passes through one of the points @xmath421 .
see figure [ fig : tricanon1 ] .
( it is possible that some of these expansions never hit a point of @xmath114 , so we may end up with an unbounded wedge or halfplane instead of a triangle .
also , the points @xmath422 need not be distinct . )
tricanon1.pstex_t let @xmath423 be the set of orientations @xmath424 .
we turn the side containing @xmath417 clockwise and counterclockwise about @xmath417 , keeping its endpoints on the lines containing the other two sides , until we reach an orientation in @xmath423 , or until we hit another point of @xmath114 ( which could also be one of the points @xmath425 ) , whichever comes first .
each of the new sides forms , with the two lines containing the two other sides , a new ( openly ) @xmath114-empty @xmath426-fat triangle ; the union of these two triangles covers @xmath420 .
see figure [ fig : tricanon2 ] .
tricanon2.pstex_t for each of the two new triangles , @xmath427 , we apply the same construction , by rotating the side containing @xmath418 clockwise and counterclockwise , thereby obtaining two new triangles whose union covers @xmath427 .
we then apply the same construction to each of the four new triangles , this time rotating about @xmath419 .
overall , we get up to eight new triangles whose union covers @xmath20 .
each of these new triangles is @xmath61-fat , openly @xmath114-empty , and each of its sides either passes through two points of @xmath114 , or passes through one point of @xmath114 and has orientation in @xmath423 . since @xmath428 , it follows that the overall number of these canonical covering triangles is @xmath429 .
( we omit the easy extensions of this step to handle unbounded wedges or halfplanes , or the cases where the points @xmath304 , or some of the newly encountered points , lie at vertices of the respective triangles . )
we take @xmath44 to be the collection of these canonical triangles , and argue that @xmath44 indeed satisfies the properties of a good test set : ( a ) compactness : @xmath430 , so its size is small .
( b ) shallowness : with high probability , each range in @xmath44 is @xmath45-shallow ( and , as usual , we assume that this property does indeed hold ) .
( c ) containment : by construction , each @xmath19-fat empty triangle is contained in the union of at most eight triangles in @xmath44 .
( d ) efficiency : being @xmath61-fat , the union of any @xmath38 triangles in @xmath44 has complexity @xmath63 @xcite , so the associated function @xmath142 satisfies @xmath431 .
this , combined with lemma [ lemma : const_cover ] and the machinery of section [ section : range_emptiness ] , lead to the following theorem .
[ theorem : fat_triangles_emptiness ] one can preprocess a set @xmath0 of @xmath1 points in the plane , in near - linear time , into a data structure of linear size , so that , for any query @xmath19-fat triangle @xmath20 , one can determine , in @xmath21 time , whether @xmath432 . [ [ reporting - points - in - fat - triangles . ] ] reporting points in fat triangles .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we can extend the technique given above to the problem of reporting the points of @xmath0 that lie inside any query fat triangle .
for this , we need to construct a test set that will be good for shallow ranges and not just for empty ones . using theorem [ epsshallow ] ,
we construct ( by random sampling ) a shallow @xmath84-net @xmath433 of size @xmath285 .
we next canonize every @xmath45-shallow @xmath19-fat triangle @xmath20 , by the same canonization process used above , with respect to the set @xmath114 .
note that each of the resulting canonical triangles contains ( in its interior ) the same subset of @xmath114 as @xmath20 does . by the properties of shallow @xmath84-nets , since @xmath434 ,
we have @xmath435 , so all the resulting canonical triangles are @xmath436-shallow with respect to @xmath114 , for some absolute constant @xmath25 . again , since @xmath114 is a shallow @xmath84-net , all the canonical triangles are @xmath437-shallow with respect to @xmath0 , for another absolute constant @xmath438 .
hence , the resulting collection @xmath44 of canonical triangles is a good test set for all shallow fat triangles , and we can apply the machinery of section [ section : range_emptiness ] to obtain a data structure of linear size , which can be constructed in near - linear time , and which can perform reporting queries in fat triangles in time @xmath23 , where @xmath27 is the output size of the query . [ [ range - emptiness - searching - with - semidisks - and - circular - caps . ] ] range emptiness searching with semidisks and circular caps .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the motivation for studying this problem comes from the following problem , addressed in @xcite .
we are given a set @xmath0 of @xmath1 points in the plane , and wish to preprocess it into a data structure of linear size , so that , given a query point @xmath65 and a query line @xmath15 , one can quickly find the point of @xmath0 closest to @xmath65 and lying above @xmath15 .
in the original problem , as formulated in @xcite , one also assumes that @xmath65 lies on @xmath15 , but we will consider , and solve , the more general version of the problem , where @xmath65 also lies above @xmath15 . the standard approach ( e.g. , as in @xcite ) yields a solution with linear storage and near - linear preprocessing , and query time @xmath16 .
we present a solution with query time @xmath21 . using parametric searching @xcite
, the problem reduces to that of testing whether the intersection of a disk of radius @xmath310 centered at @xmath65 with the halfplane @xmath439 above @xmath15 is @xmath0-empty .
the resulting range is a circular cap larger than a semidisk ( or exactly a semidisk if @xmath65 lies on @xmath15 ) .
again , the main task is to construct a good test set @xmath44 for such ranges , which we do by using an ad - hoc canonization process , which covers each empty circular cap by @xmath440 canonical caps , which satisfy the properties of a good test set ; in particular , we will have @xmath441 .
( as before , we consider here only the case of emptiness detection , and will consider the reporting problem later . ) to construct a test set @xmath44 we choose a random sample @xmath114 of @xmath285 points of @xmath0 and build a set of canonical empty ranges with respect to @xmath114 .
let @xmath442 be a given circular cap ( larger than a semidisk ) with center @xmath25 , radius @xmath310 , and chord supported by a line @xmath15 .
we first translate @xmath15 in the direction which enlarges the cap , until either its portion within the disk @xmath443 of the cap touches a point of @xmath114 , or @xmath15 leaves @xmath443 .
see figure [ fig : capcanon1](left ) . in the latter case , @xmath58
is contained in a complete @xmath114-empty disk , and it is fairly easy to show that such a disk is contained in the union of at most three canonical @xmath114-empty disks , each passing through three points of @xmath114 or through two diametrically opposite points of @xmath114 ; there are at most @xmath444 such disks .
capcanon1.pstex_t suppose then that the new chord ( we continue to denote its line as @xmath15 ) passes through a point @xmath65 of @xmath114 , as in the figure .
let @xmath423 be a set of @xmath440 canonical orientations , uniformly spaced and sufficiently dense along the unit circle , for some small constant value @xmath19 .
rotate @xmath15 about @xmath65 in both clockwise and counterclockwise directions , until we reach one of the two following events : ( i ) the orientation of @xmath15 belongs to @xmath423 ; or ( ii ) the portion of @xmath15 within @xmath443 touches another point of @xmath114 . in either case
, the two new lines , call them @xmath445 , become canonical there are only @xmath446 such possible lines .
note that our original cap @xmath58 is contained in the union @xmath447 , where @xmath448 and @xmath449 .
moreover , although the new caps need no longer be larger than a semidisk , they are not much smaller this is an easy exercise in elementary geometry .
see figure [ fig : capcanon1](right ) .
we next canonize the disk of @xmath58 ( which is also the disk of @xmath450 and @xmath451 ) .
fix one of the new caps , say @xmath450 .
expand @xmath450 from the center @xmath25 , keeping the line @xmath452 fixed , until we hit a point @xmath417 of @xmath114 ( lying in @xmath453 ) .
see figure [ fig : capcanon2](left ) .
if @xmath25 lies in @xmath453 then we move @xmath25 parallel to @xmath452 in both directions , again keeping @xmath452 itself fixed and keeping the circle pass through @xmath417 , until we obtain two circular caps , each passing through @xmath417 and through a second point of @xmath114 ( if we do not hit a second point , we reach a quadrant , bounded by @xmath15 and by the line orthogonal to @xmath15 through @xmath417 ) .
the union of the two new circular caps covers @xmath450 .
see figure [ fig : capcanon2](right ) .
capcanon2.pstex_t if @xmath25 lies in @xmath454 , we move it along the two rays connecting it to the endpoints @xmath455 of the chord defined by @xmath452 . as before ,
each of the motions stops when the circle hits another point of @xmath114 in @xmath453 , or when the motion reaches @xmath456 or @xmath457 .
we claim that @xmath450 is contained in the union of the two resulting caps .
indeed , let @xmath458 and @xmath235 denote the locations of the center at the two stopping placements .
we need to show that , for any point @xmath459 we have either @xmath460 or @xmath461 .
if both inequalities did not hold , then both @xmath458 and @xmath235 would have to lie on the side of the perpendicular bisector of @xmath462 containing @xmath417 .
this is easily seen to imply that @xmath25 must also lie on that side , which however is impossible ( because @xmath463 ) .
see figure [ fig : c1lemma ] .
c1lemma.pstex_t next , we take one of these latter caps , @xmath464 , whose bounding circle passes through @xmath417 and through a second point @xmath418 of @xmath465 , and move its center along the bisector of @xmath466 in both directions , keeping the bounding circle touch @xmath417 and @xmath418 , and still keeping the line @xmath452 supporting the chord fixed .
we stop when the first of these events takes place : ( i ) the center reaches @xmath452 , in which case the cap becomes a semidisk ( this can happen in only one of the moving directions ) .
( ii ) the center reaches the midpoint of @xmath466 .
( iii ) the bounding circle touches a third point of @xmath467 .
( iv ) the central angle of the chord along @xmath452 is equal to some fixed positive angle @xmath468 .
the union of the two new caps covers @xmath464 .
( it is possible that during the motion the moving circle becomes tangent to @xmath452 , and then leaves it , in which case the corresponding final cap is a full disk . ) similarly , if the center of @xmath464 lies on @xmath452 ( which can happen when the motion in the preceding canonization step reaches @xmath457 or @xmath456 ) , then we canonize its disk by translating the center to the left and to the right along @xmath469 until the bounding circle touches another point of @xmath467 , exactly as in the preceding case ( shown in figure [ fig : capcanon2](right ) ) .
let @xmath470 be one of the four new caps . in all cases
@xmath470 is canonical : for the first kind of caps , the stopping condition that defines @xmath470 is ( ii ) or ( iii ) then either the circle bounding @xmath470 passes through three points of @xmath114 or it passes through two diametrically opposite points of @xmath114 .
there are a total of @xmath444 such circles , and since @xmath470 is obtained ( in a unique manner ) by the interaction of one of these circles and one of the @xmath446 canonical chord - lines , there is a total of @xmath471 such caps .
if the stopping condition is ( i ) , the cap is also canonical , because the center of the containing disk is the intersection point of a bisector of two points of @xmath114 with one of the @xmath446 canonical chord - lines , so there is a total of @xmath472 such caps . in the case of condition ( iv )
, there are only @xmath446 such disks , for a total of @xmath472 caps .
similar reasoning shows that the caps resulting in the second case are also canonical , and their number is @xmath472 .
we take the test set @xmath44 to consist of all the caps of the final forms , and argue that it satisfies the properties of a good test set : ( a ) compactness : @xmath473 , so its size is small .
( b ) shallowness : with high probability , each range in @xmath44 is @xmath45-shallow ( and we assume that this property does indeed hold ) .
( c ) containment : each empty cap is also @xmath114-empty , so , by the above canonization process , it is contained in the union of @xmath440 caps of @xmath44 .
( d ) efficiency : each cap @xmath474 is _ @xmath475-covered _ , for appropriate fixed constants @xmath476 , in the terminology of @xcite , meaning that for each point @xmath477 there exists an @xmath19-fat triangle touching @xmath75 , contained in @xmath58 , and with diameter which is at least @xmath136 times the diameter of @xmath58 . in addition , the boundaries of any two ranges in @xmath44 ( or of any two circular caps , for that matter ) intersect in at most four points , as is easily checked . as follows from the recent analysis of de berg @xcite ,
the complexity of the union of any @xmath38 ranges in @xmath44 is @xmath478 .
hence , the complement of the union of any @xmath38 ranges in @xmath44 can be decomposed into @xmath66 elementary cells , making @xmath441 . in conclusion ,
we obtain : [ thm : caps ] let @xmath0 be a set of @xmath1 points in the plane .
we can preprocess @xmath0 , in near - linear time , into a data structure of linear size , so that , for any query circular cap @xmath58 , larger than a semidisk , we can test whether @xmath479 is empty , in @xmath21 time . combining theorem [ thm : caps ] with parametric searching ,
we obtain : [ cor : caps ] let @xmath0 be a set of @xmath1 points in the plane .
we can preprocess @xmath0 , in near - linear time , into a data structure of linear size , so that , for any query halfplane @xmath439 and point @xmath480 , we can find , in @xmath21 time , the point in @xmath481 nearest to @xmath65 .
* the machinery developed in this section also applies to smaller circular caps , as long as they are not too small .
formally , if the central angle of each cap is at least some fixed constant @xmath482 , the same technique holds , so we can test emptiness of such ranges in @xmath21 time , using a data structure which requires @xmath7 storage and @xmath8 preprocessing . thus theorem [ thm : caps ] carries over to this scenario , but corollary [ cor : caps ] does not ,
because we have no control over the `` fatness '' of the cap , as the disk shrinks or expands , when the center of the disk lies in @xmath483 , and once the canonical caps become too thin , the complexity of their union may become quadratic .
[ [ reporting - points - in - semidisks - and - circular - caps . ] ] reporting points in semidisks and circular caps .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as in the case of fat triangles , we can extend the technique to answer efficiently range reporting queries in semidisks or in sufficiently large circular caps .
we use the same canonization process , with respect to a random sample @xmath114 of size @xmath285 which is a shallow @xmath84-net , and argue , exactly as in the case of fat triangles , that the resulting collection of canonical caps is a good test set for shallow semidisk or larger cap ranges . applying the machinery of section [ section : range_emptiness ] ,
we obtain a data structure of linear size , which can be constructed in near - linear time , and which can perform reporting queries in semidisks or larger caps , in time @xmath23 , where @xmath27 is the output size of the query .
given a set @xmath0 of @xmath1 points in @xmath2 , a set @xmath74 of semi - algebraic ranges of constant description complexity , and a parameter @xmath484 , the _ approximate range counting _ problem is to preprocess @xmath0 into a data structure such that , for any query range @xmath169 , we can efficiently compute an approximate count @xmath67 which satisfies @xmath485 as in most of the rest of the paper , we will only consider the case where the size of the data structure is to be ( almost ) linear , and the goal is to find solutions with small query time .
the problem has been studied in several recent papers @xcite , for the special case where @xmath0 is a set of points in @xmath2 and @xmath74 is the collection of halfspaces ( bounded by hyperplanes ) .
a variety of solutions , with near - linear storage , were derived ; in all of them , the dependence of the query cost on @xmath1 is close to @xmath486 , which , as reviewed earlier , is roughly the same as the cost of halfspace range emptiness queries , or the overhead cost of halfspace range reporting queries @xcite .
the fact that the approximate range counting problem is closely related to range emptiness comes as no surprise , because , when @xmath487 , the approximate count @xmath77 must be @xmath282 , so range emptiness is a special case of approximate range counting .
the goal is therefore to derive solutions that are comparable , in their dependence on @xmath1 , with those that solve emptiness ( or reporting ) queries . as just noted , this has been accomplished for the case of halfspaces . in this section we extend this technique to the general semi - algebraic case .
the simplest solution is to adapt the technique of aronov and har - peled @xcite , which uses a procedure for answering range emptiness queries as a `` black box '' .
specifically , suppose we have a data structure , @xmath488 , for any set @xmath176 of @xmath489 points , which can be constructed in @xmath490 time , uses @xmath491 storage , and can determine whether a query range @xmath169 is empty in @xmath492 time .
using such a black box , aronov and har - peled show how to construct a data structure for @xmath1 points using @xmath493 storage and @xmath494 preprocessing , where @xmath495 is some constant for which @xmath496 and @xmath497 , for any @xmath498 . given a range @xmath169 , the data structure of @xcite returns , in @xmath499 time , an approximate count @xmath67 , satisfying @xmath500 .
the intuition behind this approach is that a range @xmath50 , containing @xmath38 points of @xmath0 , is expected to contain @xmath501 points in a random sample from @xmath0 of size @xmath40 , and no points in a sample of size smaller than @xmath502 .
the algorithm of @xcite then guesses the value of @xmath251 ( up to a factor of @xmath503 ) , sets @xmath40 to be an appropriate multiple of @xmath502 , and draws many ( specifically , @xmath504 random samples of size @xmath40 . if @xmath50 is empty ( resp . , nonempty ) for many of the samples then , with high probability , the guess for @xmath38 is too large ( resp .
, too small ) .
when we can not decide either way , we are at the correct value of @xmath38 ( up to a relative error of @xmath79 ) .
the actual details of the search are somewhat more contrived ; see @xcite for those details . plugging our emptiness data structures into the machinery of @xcite , we therefore obtain the following results . in all these applications we can take @xmath505 , so , in the terminology used above , the overall data structure uses @xmath506 storage and @xmath507 preprocessing .
let @xmath0 be a set of @xmath1 points in the plane , and let @xmath19 , @xmath79 be given positive parameters .
then we can preprocess @xmath0 into a data structure of size @xmath508 , in time @xmath509 , for any @xmath10 , such that , for any @xmath19-fat query triangle @xmath20 , we can compute , in @xmath510 time , for any @xmath10 , an approximate count @xmath511 satisfying @xmath512 .
let @xmath0 be a set of @xmath1 points in the plane , and let @xmath79 be a given positive parameter .
then we can preprocess @xmath0 into a data structure of size @xmath508 , in time @xmath509 , for any @xmath10 , such that , for any line @xmath15 , point @xmath75 on @xmath15 or above @xmath15 , and distance @xmath313 , we can compute , in @xmath510 time , for any @xmath10 , an approximate count @xmath513 of the exact number @xmath514 of the points of @xmath0 which lie above @xmath15 and at distance at most @xmath313 from @xmath75 , so that @xmath515 .
let @xmath0 be a set of @xmath1 points in @xmath6 , @xmath74 a collection of convex semi - algebraic ranges of constant description complexity , and @xmath79 a given positive parameter .
then we can preprocess @xmath0 into a data structure of size @xmath508 , in time @xmath509 , for any @xmath10 , such that , for any query range @xmath87 , we can compute , in @xmath516 time , for any @xmath10 , an approximate count @xmath67 of the number of points of @xmath0 outside @xmath50 , satisfying @xmath517 .
let @xmath381 be a set of @xmath1 balls in @xmath6 , and let @xmath79 be a given positive parameter .
then we can preprocess @xmath381 into a data structure of size @xmath508 , in time @xmath509 , for any @xmath10 , such that , for any query ray @xmath310 , we can compute , in @xmath518 time , for any @xmath10 , an approximate count @xmath519 of the exact number @xmath520 of the balls of @xmath381 intersected by @xmath310 , satisfying @xmath521 .
* remark : * another approach to approximate range counting has been presented in @xcite , in which , rather than using range emptiness searching as a black box , one modifies the partition tree of the range emptiness data structure , and augments each of its inner nodes with a so - called _ relative @xmath522 approximation _ sets , which are then used to obtain the approximate count of a range .
this approach too can be adapted to yield efficient approximate range counting algorithms for semialgebraic ranges , with a slightly improved dependence of their performance on @xmath79 .
we omit details of such an adaptation in this paper .
in this paper we have presented a general approach to efficient range emptiness searching with semi - algebraic ranges , and have applied it to several specific emptiness searching and ray shooting problems .
the present study resolves and overcomes the technical problems encountered in our earlier study @xcite , and presents more applications of the technique .
clearly , there are many other applications of the new machinery , and an obvious direction for further research is to `` dig them up '' . in each such problem ,
the main step would be to design a good test set , with associated function @xmath281 as small as possible , using either the general recipe or an appropriate ad - hoc analysis .
many specific instances of this step are likely to generate interesting ( and often hard ) combinatorial questions . for example , as already mentioned earlier , we still do not know whether the complement of the union of @xmath1 ( congruent ) cylinder in @xmath6 can be decomposed into @xmath523 elementary cells . p. k. agarwal and j. erickson , geometric range searching and its relatives , in : _ advances in discrete and computational geometry _ ( b. chazelle , j. e. goodman and r. pollack , eds . ) , ams press , providence , ri , 1998 , pp .
156 . p.
k. agarwal , j. pach and m. sharir , state of the union ( of geometric objects ) , in _ proc .
joint summer research conf . on discrete and computational geometry : 20 years later _ ,
452 , ams , 2008 , pp .
948 .
b. chazelle , h. edelsbrunner , l. j. guibas , and m. sharir , a singly exponential stratification scheme for real semi - algebraic varieties and its applications , _ theoret .
_ , 84 ( 1991 ) , 77105 . also in _ proc .
16th int . colloq . on automata , languages and programming _ ( 1989 ) , pp . | in a typical range emptiness searching ( resp . , reporting ) problem , we are given a set @xmath0 of @xmath1 points in @xmath2 , and wish to preprocess it into a data structure that supports efficient range emptiness ( resp .
, reporting ) queries , in which we specify a range @xmath3 , which , in general , is a semi - algebraic set in @xmath2 of constant description complexity , and wish to determine whether @xmath4 , or to report all the points in @xmath5 .
range emptiness searching and reporting arise in many applications , and have been treated by matouek @xcite in the special case where the ranges are halfspaces bounded by hyperplanes . as shown in @xcite ,
the two problems are closely related , and have solutions ( for the case of halfspaces ) with similar performance bounds . in this paper
we extend the analysis to general semi - algebraic ranges , and show how to adapt matouek s technique , without the need to _ linearize _ the ranges into a higher - dimensional space .
this yields more efficient solutions to several useful problems , and we demonstrate the new technique in four applications , with the following results : \(i ) an algorithm for ray shooting amid balls in @xmath6 , which uses @xmath7 storage and @xmath8 preprocessing , to mean an upper bound of the form @xmath9 , which holds for any @xmath10 , where @xmath11 is a constant that depends on @xmath12 . ] and answers a query in @xmath13 time , improving the previous bound of @xmath14 .
\(ii ) an algorithm that preprocesses , in @xmath8 time , a set @xmath0 of @xmath1 points in @xmath6 into a data structure with @xmath7 storage , so that , for any query line @xmath15 ( or , for that matter , any simply - shaped convex set ) , the point of @xmath0 farthest from @xmath15 can be computed in @xmath16 time . this in turn yields an algorithm that computes the largest - area triangle spanned by @xmath0 in time @xmath17 , as well as nontrivial algorithms for computing the largest - perimeter or largest - height triangle spanned by @xmath0 .
\(iii ) an algorithm that preprocesses , in @xmath8 time , a set @xmath0 of @xmath1 points in @xmath18 into a data structure with @xmath7 storage , so that , for any query @xmath19-fat triangle @xmath20 , we can determine , in @xmath21 time , whether @xmath22 is empty .
alternatively , we can report in @xmath23 time , the points of @xmath22 , where @xmath24 .
\(iv ) an algorithm that preprocesses , in @xmath8 time , a set @xmath0 of @xmath1 points in @xmath18 into a data structure with @xmath7 storage , so that , given any query semidisk @xmath25 , or a circular cap larger than a semidisk , we can determine , in @xmath21 time , whether @xmath26 is empty , or report the @xmath27 points in @xmath26 in @xmath28 time .
adapting the recent techniques of @xcite , we can turn our solutions into efficient algorithms for approximate range counting ( with small relative error ) for the cases mentioned above .
our technique is closely related to the notions of nearest- or farthest - neighbor generalized voronoi diagrams , and of the union or intersection of geometric objects , where sharper bounds on the combinatorial complexity of these structures yield faster range emptiness searching or reporting algorithms .
h
s 1v_1 # 1hayim says : # 1 # 1micha says : # 1 |
the physical mechanism by which galaxies acquire their angular momentum is an important problem that has been the subject of investigation for nearly sixty years ( hoyle 1949 ) .
this reflects the fundamental role played by angular momentum of galactic material in defining the size and shapes of galaxies ( e.g. fall & efstathiou 1981 ) . yet despite its physical significance , a precise and accurate understanding of the origin of galactic angular momentum remains one of the missing pieces in the galaxy formation puzzle .
a fundamental assumption in current galaxy formation models is that galaxies form when gas cools and condenses within the potential wells of dark matter halos ( white & rees 1978 ) .
consequently it is probable that the angular momentum of the galaxy will be linked to the angular momentum of its dark matter halo ( e.g. fall & efstathiou 1980 ; mo , mao & white 1998 ; zavala , okamoto & frenk 2007 ) . within the context of hierarchical structure formation models , the angular momentum growth of a dark matter proto - halo
is driven by gravitational tidal torquing during the early stages ( i.e. the linear regime ) of its assembly .
this `` tidal torque theory '' has been explored in detail ; it is a well - developed analytic theory ( e.g. peebles 1969 , doroshkevich 1979 , white 1984 ) and its predictions are in good agreement with the results of cosmological @xmath3-body simulations ( e.g. barnes & efstathiou 1987 ; warren et al .
1992 ; sugerman , summers & kamionkowski 2000 ; porciani , dekel & hoffman 2002 ) . however , once the proto - halo has passed through maximum expansion and the collapse has become non - linear , tidal torquing no longer provides an adequate description of the evolution of the angular momentum ( white 1984 ) , which tends to decrease with time . during this phase
it is likely that merger and accretion events play an increasingly important role in determining both the magnitude and direction of the angular momentum of a galaxy ( e.g. bailin & steinmetz 2005 ) .
indeed , a number of studies have argued that mergers and accretion events are the primary determinants of the angular momenta of galaxies at the present day ( gardner 2001 ; maller , dekel & somerville 2002 ; vitvitska et al . 2002 ) .
+ it is common practice to quantify the angular momentum of a dark matter halo by the dimensionless `` classical '' spin parameter ( peebles 1969 ) , @xmath4 where @xmath5 is the magnitude of the angular momentum of material within the virial radius , @xmath0 is the virial mass , and @xmath6 is the total energy of the system .
it has been shown that halos that have suffered a recent major merger will tend to have a higher spin parameter @xmath1 than the average ( e.g. hetznecker & burkert 2006 ; power , knebe & knollmann 2008 ) .
therefore one could argue that within the framework of hierarchical structure formation that higher mass halos should have larger spin parameters _ on average _ than less massive systems because they have assembled a larger fraction of their mass ( by merging ) more recently .
however , if we consider only halos in virial equilibrium , should we expect to see a correlation between halo mass and spin ?
one might navely expect that more massive systems will have had their maximum expansion more recently and so these systems will have been tidally torqued for longer than systems that had their maximum expansion at earlier times .
this suggests that spin should _ increase _ with timing of maximum expansion and therefore halo mass .
however , one finds at best a weak correlation between mass and spin for equilibrium halos at @xmath7=0 ( e.g. cole & lacey 1996 ; maccio et al . 2007 ;
bett et al .
2007 , hereafter b07 ) , and the correlation is for spin to _ decrease _ with increasing halo mass , contrary to our nave expectation . in this paper , we report on a ( weak ) correlation between spin and mass for equilibrium halos at redshift @xmath7=10 .
the trend is for higher - mass halos to have smaller spins , and is qualitatively similar to the one reported by b07 for the halo population at @xmath7=0 .
we present the main evidence in support of this correlation in section [ sec : correlation ] and we consider its implications for galaxy formation in section [ sec : conclusions ] .
for the simulations presented in this paper we have adopted the cosmology as given by spergel et al .
( 2003 ) ( @xmath8 , @xmath9 , @xmath10 , and @xmath11km / sec / mpc ) .
each run employed @xmath12 particles and differed in simulation box - size @xmath13 , which leads to the particle mass @xmath14 differing between runs
@xmath15 , where @xmath16 .
this allows us to probe a range of halo masses at redshift @xmath7=10 .
the primary parameters of these simulations are summarized in table [ tab : simu ] .
halos in all runs have been identified using the mpi parallelized version of the ` ahf ` halo finder ( ` amiga`s - halo - finder ) , which is based on the ` mhf ` halo finder of gill , knebe & gibson ( 2004 ) . for each halo
we compute the virial radius @xmath17 , defined as the radius at which the mean interior density is @xmath18 times the background density of the universe at that redshift .
this leads to the following definition for the virial mass @xmath0 : @xmath19 note that @xmath18 is a function of redshift and amounts to @xmath20 at redshift @xmath7=10 , @xmath21 at @xmath7=1 , and the `` usual '' @xmath22 at @xmath7=0 ( cf .
gross 1997 ) .
table [ tab : simu ] summarises the total number of halos ( @xmath23 ) recovered by ` ahf ` , while @xmath24 gives the redshift that the simulation has been evolved to .
we add that we ran five realisations of b20 to redshift @xmath7=10 in order to have a statistically significant sample of halos in that particular model .
however , we also note that the fitting parameters presented in the following section are robust in the sense that they do not depend on whether we stack the halos from those five runs or use them individually .
we show in a companion paper ( power , knebe & knollmann 2008 ) that a substantial fraction of the halo population at high redshift is not in virial equilibrium . because we wish to examine the spin distribution of equilibrium halos , it is important to account for unrelaxed systems when investigating correlations between spin and halo mass .
for example , it has been shown that the spin can increase sharply in the aftermath of mergers with mass ratios as modest as 5:1 ( e.g. hetznecker & burkert 2006 ) , and that the degree to which a halo is in dynamical relaxation is as important as recent merging history in its influence on spin ( donghia & navarro 2007 ) . to ensure that halos in our sample are in virial equilibrium , we compute the virial ratio for each halo , which we define as @xmath25 here @xmath26 represents the kinetic energy , @xmath27 the potential energy , and @xmath28 the surface pressure of a given halo of mass @xmath0 . by including @xmath28 ,
we can account for the effect of infalling material on the dynamical state of the halo .
each of these quantities are evaluated using all gravitationally bound particles , and we adopt the formula of shaw et al .
( 2006 ) for the surface pressure term @xmath28 ( cf .
equations.(4)-(6 ) in their study ) . in figure
[ fig : qspmass ] we show that the relation between halo mass and @xmath29 can vary with mass .
this is apparent at redshift @xmath7=1 , where we find a trend for more massive halos to be less virialised .
in contrast , high redshift halos are less virialised on average ( as indicated by the increased average @xmath30 ) , but we find no apparent trend with mass .
why is there a mass dependence at @xmath7=1 but not at @xmath7=10 ?
there are two factors .
the first is that high redshift halos `` see '' an effective slope of the initial power spectrum of @xmath31 , and so the time at which a particular mass scale starts to collapse is relatively insensitive to mass .
therefore we do not expect to find a strong correlation between virial state @xmath29 and mass .
we have checked this halo populations drawn from the simulations of scale - free cosmologies of knollmann , power & knebe ( 2008 ) and our interpretation is consistent with the correlations we find in these runs .
the second is that the typical collapsing mass @xmath32 at @xmath7=10 is small of order @xmath33 and because we resolve mass scales that have collapsed more or less simultaneously , we see a population that has yet to relax . at @xmath7=1 ,
the typical collapsing mass is much larger of order @xmath34 and so we resolve a population of halos whose mass accretion histories are more diverse .
the most massive systems tend to be ones that have formed most recently , and are therefore the least dynamically relaxed .
we have used the following relation between @xmath29 and @xmath0 to classify dynamically relaxed and unrelaxed systems : @xmath35 we allow the @xmath29 values of halos in our sample to deviate from these scaling relations by not more than @xmath36 with @xmath37 ( indicated by the dashed lines in fig .
[ fig : qspmass ] ) .
furthermore , we consider only halos that contain at least @xmath38 particles within their virial radius to ensure that we are not influenced by particle discreteness .
interestingly , when computing spin , the tighest restriction on particle number comes not from the calculation of angular momentum but from the calculation of the potential energy . by comparing analytic solutions with monte carlo realisations of navarro , frenk & white ( 1997 ) haloes ,
we find that at least 600 particles are required if the energy is to be computed to better than @xmath39 .
calculating the total energy @xmath6 of a halo is computationally expensive , and so computing @xmath1 using eq .
( [ eq : lambda ] ) is also expensive .
this prompted bullock et al .
( 2001 ) to introduce a modified spin parameter @xmath40 where @xmath41 measured the circular velocity at the virial radius @xmath17 and @xmath5 represents the absolute value of the angular momentum .
we follow bullock et al . and compute spin using equation ( [ eq : spintsp ] ) . in figure
[ fig : spinmass ] , we investigate the correlation between halo spin @xmath42 and mass @xmath0 .
we show only halos that fulfill our selection criteria ( individual dots ) and bin the data in five mass bins equally spaced in log - space between @xmath43 and @xmath44 of the considered halos at the respective redshift .
the values plotted as histograms thereby represent the weighted mean of all spin parameters in the respective mass range where the weight is inversely proportional to the error estimate @xmath45 for the spin parameter of a halo consisting of @xmath3 particles as derived in bullock et al .
( 2001 ) ( cf
. equation ( 7 ) in that study ) .
the error bars indicate the standard deviation of the spin parameter values in the bin from the weighted mean .
the best fitting power - laws to these histograms reveal that @xmath46 with @xmath47 this indicates that there is a _ weak _ correlation at high redshifts for spin to decrease with increasing mass , albeit stronger than the one at @xmath7=1 .
we compute spearman rank correlation coefficients at @xmath7=10 ( 1 ) and find @xmath48 . as an alternative approach
, we fit a lognormal function to each of our halo samples at @xmath7=10 and @xmath7=1 , @xmath49 the resulting curves are presented in fig .
[ fig : pspin ] whereas the best - fit parameters , median values for @xmath50 and median halo masses @xmath51 are given in table [ tab : lambdatsp ] .
inspection of the best - fit parameters confirm that the median spin declines as we move from less massive to more massive objects at high redshift .
because of the weak nature of the measured correlation it is vitally important to check its credibility by performing a statistical analysis . to this extent
we investigate the sensitivity of the logarithmic slope @xmath52 with respect to a number of parameters that enter into its determination , namely the number of bins @xmath53 used for the histograms ; the virialisation criterion parametrized via @xmath54 ; and the minimum number of particles @xmath55 within a halo s virial radius .
note that we vary one parameter at a time , keeping the others at their fixed `` standard '' values .
the results are presented in table [ tab : nbins ] , table [ tab : qlim ] , and [ tab : nmin ] .
we find that bin number has practically no effect on the slope .
similarly we find that varying the virialisation criterion @xmath54 has little effect on the slope of the relation between mass and spin , regardless of redshift ( table [ tab : qlim ] ) .
in contrast , we find that the minimum number of particles within a halo s virial radius has a strong and systematic effect on the result at @xmath7=1 as @xmath55 increases , we find that the logarithmic slope becomes shallower .
this does not appear to be true at high redshift , although as we go to earlier times we find that the number of massive haloes becomes progressively smaller and our determination of @xmath52 becomes increasingly unreliable .
these tests lead us to believe that our results are both stable and reliable , and our main result holds : _ the correlation between spin paramater @xmath42 and halo mass @xmath0 is one order of magnitude larger at redshift @xmath7=10 than at @xmath7=1 . _
+ as a further test of the credibility of the correlation we measure between halo mass and spin , we use the criteria of three other studies to select our halo sample .
these are : * maccio et al . ( 2007 ) criteria : * * @xmath56 * * @xmath57 * * @xmath58 * bett et al .
( 2007 ) criteria : * * @xmath59 * * @xmath60 * neto et al . ( 2007 ) criteria : * * @xmath38 * * @xmath61 * * @xmath62 here @xmath55 is again the minimum number of particles in a halo , @xmath54 the virial limit as defined in eq .
( [ eq : qlim ] ) , @xmath63 measures the distance between the most bound particle and the centre of mass in units of the virial radius , @xmath64 is an indicator of how well the density profile of the halo can be fitted by a navarro , frenk & white ( 1997 ) profile ( cf .
( 2 ) in maccio et al . 2007 ) , and @xmath65 is the fraction of mass in subhalos .
the resulting power law slopes @xmath52 ( cf .
( [ eq : logfit ] ) ) for @xmath66 and @xmath67 are presented in table [ tab : etal ] .
again we note that there appears to be a much stronger correlation between @xmath42 and @xmath0 at higher redshift . while the relation is consistent with zero at @xmath7=1 ( as confirmed by maccio et al .
2007 ) spin and mass are correlated at @xmath7=10 .
how this result relates to the bett et al .
( 2007 ) result who find a weak correlation at @xmath7=0 will be discussed in the following section .
we have performed a careful investigation of the relation between virial mass and dimensionless spin parameter for dark matter halos forming at high redshifts @xmath2 in a @xmath68cdm cosmology .
the result of our study , which is based on a series of cosmological @xmath3-body simulations in which box size was varied while keeping particle number fixed , indicates that there is a _ weak _ correlation between mass and spin at @xmath7=10 , such that the spin decreases with increasing mass .
if there is a correlation at @xmath7=1 , we argue that it is significantly weaker than the one we find at @xmath7=10 ; this is in qualitative agreement with the findings of previous studies that focused on lower redshifts ( maccio et al .
2007 , shaw et al .
2005 , lemson & kauffmann 1999 ) .
+ interestingly , b07 find a weak correlation between median spin and halo mass at @xmath7=0 in the millennium simulation ( springel et al . 2005 ) , in the same spirit as the one presented here for @xmath7=10 : lower mass halos tend to have higher spins .
however , as we show , the correlation between halo mass and spin is weaker at @xmath7=1 than at @xmath7=10 , whereas the correlation reported in b07 for halos at @xmath7=0 is much stronger than the one we find at @xmath7=1 .
this is not what one would expect , and so it is important to try and understand the source of the difference between our result at @xmath7=1 and the b07 result at @xmath7=0 .
b07 fitted a 3@xmath69-order polynomial to the median spins of halos in the mass range @xmath70 at @xmath7=0 .
the form of this polynomial is extremely sensitive to the precise values of the best - fit parameters ( bett , private communication ) and it is not straightforward to extrapolate its behaviour outside of the given mass range and redshift .
we derive our estimates of the power - law exponents from the spin distribution with respect to halo mass at @xmath7=1 .
our halos lie in the mass range @xmath71 .
b07 base their median spins upon @xmath72 million halos with correspondingly small errors , and note that the weak nature of the trend of spin with mass makes it hard to detect .
this suggests to us that the correlation between mass and spin at @xmath7=10 is remarkably strong rather than the correlation at @xmath7=1 being too weak ! when studying correlations between halo mass and spin , great care must be taken in defining the halo sample . in particular , we find that mass resolution ( i.e. the number of particles with which a halo is resolved ) and the degree of virialisation of a halo can have a significant effect on the strength of the correlation ( at least at @xmath7=1 , cf . table [ tab : nmin ] ) .
this at least is in good agreement with the findings of b07 .
+ we note that power & knebe ( 2006 ) demonstrated that the size of simulation box can lead to a suppression of angular momentum in smaller boxes , due to the absence of longer wavelength perturbations in the initial conditions .
this will lead to a bias in our estimate of @xmath1 ( approximately a @xmath73 effect ) but we have verified that the spin distributions we obtain from a b20 run truncated on scales larger than the longest wavelength perturbation modelled in the b1 run produces results that are consistent . indeed , we would expect the correlation to be strengthened if the b1 spins were corrected for box size effects . + it is interesting to speculate on the consequences of this correlation for galaxy formation at high redshifts and the galaxy population we observe today . in the standard picture of galaxy formation ,
gas cools on to dark matter halos and is shock heated to the virial temperature of the halo .
the angular momentum of the gas and the dark matter should ( initially ) be similar because they are subject to the same tidal field . as the innermost densest parts of the gaseous halo cool , they will settle into a gaseous disk with a scale length determined by the specific angular momentum of the gas , which we would expect to be related to the angular momentum of the halo ( e.g. zavala , okamoto & frenk 2007 ) .
if more massive halos at high redshifts show a tendency to have smaller spin parameters , the gas disks will have lower specific angular momenta and therefore will be more centrally concentrated .
if star formation rate correlates with surface density , then we might expect the star formation rate to be enhanced in more massive halos . because massive halos tend to form preferentially in high density , highly clustered environments in which the merger rate also tends to be enhanced
, then we might expect star formation to proceed more rapidly and at earlier times in these halos .
might this explain the effect of `` downsizing '' ( e.g. cowie et al . 1996 ) , the successive shifting of star formation from high- to low - mass galaxies with decreasing redshift ?
we shall pursue this in a more quantitative manner in future work .
ak would like to thank swinburne university for its hospitality where this work was initiated .
ak further acknowledges funding through the emmy noether programme of the dfg ( kn 755/1 ) .
cp acknowledges funding through the arc discovery projected funded `` commonwealth cosmology initiative '' , grant dp 0665574 .
the simulations presented in this paper were carried out on the beowulf cluster at the centre for astrophysics & supercomputing , swinburne university as well as the sanssouci cluster at the astrophysikalisches institut potsdam .
ccccccccc b01 & 1 & 4.9 @xmath74 & 10 & & 8780 & & 286 + b02 & 2.5 & 7.8 @xmath75 & 10 & & 7991 & & 201 + b05 & 5 & 6.2 @xmath76 & 1 & 16917 & 6532 & 832 & 109 + b10 & 10 & 4.9 @xmath77 & 1 & 18589 & 4360 & 949 & 37 + b20 & 20 & 4.0 @xmath78 & 1 & 20514 & 10947 & 995 & 27 + b100 & 100 & 4.9 @xmath79 & 1 & 24696 & & 978 & + ccccccccccc & & + & & b01 & 0.042 & 0.538 & 0.041 & 5.95e+06 + & & b02 & 0.040 & 0.539 & 0.037 & 8.19e+07 + b05 & 0.033 & 0.521 & 0.031 & 7.92e+08 & & b05 & 0.036 & 0.516 & 0.035 & 5.63e+08 + b10 & 0.037 & 0.527 & 0.034 & 6.15e+09 & & b10 & 0.033 & 0.540 & 0.029 & 4.14e+09 + b20 & 0.038 & 0.543 & 0.036 & 5.40e+10 & & b20 & 0.030 & 0.251 & 0.027 & 3.33e+10 + b100 & 0.037 & 0.535 & 0.035 & 5.25e+12 & & + ccccc 4 & -0.001 @xmath80 0.169 & -0.061 @xmath80 0.191 + 5 & -0.002 @xmath80 0.149 & -0.059 @xmath80 0.171 + 6 & -0.005 @xmath80 0.137 & -0.053 @xmath80 0.155 + 7 & -0.006 @xmath80 0.128 & -0.056 @xmath80 0.144 + 8 & -0.003 @xmath80 0.120 & -0.069 @xmath80 0.132 + ccccc 0.05 & -0.007 @xmath80 0.148 & -0.058 @xmath80 0.153 & 2123 & 264 + 0.10 & -0.004 @xmath80 0.148 & -0.062 @xmath80 0.168 & 3365 & 497 + 0.15 & -0.002 @xmath80 0.149 & -0.059 @xmath80 0.171 & 3811 & 660 + 0.20 & -0.005 @xmath80 0.151 & -0.052 @xmath80 0.173 & 3994 & 739 + 0.25 & -0.002 @xmath80 0.153 & -0.052 @xmath80 0.173 & 4165 & 774 + ccccc 100 & 0.003 @xmath80 0.129 & -0.037 @xmath80 0.158 & 19707 & 5478 + 200 & 0.002 @xmath80 0.137 & -0.038 @xmath80 0.167 & 10809 & 2534 + 300 & 0.002 @xmath80 0.142 & -0.035 @xmath80 0.174 & 7336 & 1526 + 600 & -0.002 @xmath80 0.149 & -0.059 @xmath80 0.171 & 3811 & 660 + 1000 & -0.004 @xmath80 0.155 & -0.058
@xmath80 0.196 & 2303 & 343 + 2000 & -0.011 @xmath80 0.164 & -0.048 @xmath80 0.186 & 1154 & 120 + ccccc neto et al .
( 2007 ) & 0.000 @xmath80 0.149 & -0.041 @xmath80 0.178 & 3486 & 429 + maccio et al . (
2007 ) & 0.001 @xmath80 0.146 & -0.040 @xmath80 0.184 & 4275 & 512 + bett et al .
( 2007 ) & 0.003 @xmath80 0.146 & -0.040 @xmath80 0.175 & 8582 & 1955 + | we report on a correlation between virial mass @xmath0 and spin parameter @xmath1 for dark matter halos forming at redshifts @xmath2 .
we find that the spin parameter decreases with increasing halo mass .
interestingly , our analysis indicates that halos forming at later times do not exhibit such a strong correlation , in agreement with the findings of previous studies .
we briefly discuss the implications of this correlation for galaxy formation at high redshifts and the galaxy population we observe today . |
event shapes belong to the most widely used variables to study qcd dynamics , especially at @xmath0 and @xmath1 colliders .
event - shape observables are defined in terms of the four - momenta of jets in the final state . recently a large set of new hadronic event - shape variables has been proposed in ref .
an important aspect of these variables is their normalization to the total transverse momentum or energy in the event .
therefore it is anticipated that energy scale uncertainties should cancel out to a large extent .
thus we believe that they represent an useful tool for very early measurements of the properties of qcd events at lhc and the tuning of monte carlo models .
analogously to the @xmath0 event shapes , one can define hadronic event shapes in the transverse plane , for example the central transverse thrust : @xmath2 where @xmath3 are the transverse momenta with respect to the beam axis @xmath4 .
the transverse axis , for which the maximum is obtained , is the thrust axis @xmath5 .
the variable which is typically used for perturbative calculations is @xmath6 .
the central thrust minor is a measure for the out - of - plane momentum : @xmath7 below the results of a first simulation study @xcite of these event - shapes variables at the compact muon solenoid ( cms ) are summarized .
@xmath8 is used to simulate proton - proton collisions with a centre of mass energy @xmath9 @xcite .
the events have been passed through a full geant based simulation of the cms detector .
events are preselected by requiring two or more calorimeter jets , corrected in their relative and absolute response , with a transverse energy @xmath10 within a region of @xmath11 .
if the two hardest jets of the event are not in this central region , the event is rejected .
only central corrected calorimeter jets with @xmath10 are used for the event - shape calculation .
the threshold on the transverse energy of the leading jet is set at @xmath12 .
[ sec : corrections ] the effect of jet energy corrections on the normalized event - shape distributions can be judged by comparing the corrected and uncorrected distributions with the corresponding generator level distribution .
a threshold of 30 @xmath13 on the transverse energy of uncorrected jets is applied , which corresponds to a generator level jet threshold of approximately 60 @xmath13 .
similarly , the threshold on the uncorrected leading jets is @xmath14 .
all three distributions agree well with deviations within 5 - 7 % over most of the range as fig [ fig : l2l3_thrust ] illustrates .
[ fig : l2l3_thrust ] often , the leading source of systematic errors in qcd data analysis is the limited knowledge of the jet energy scale ( jes ) and , to a lesser extent , the jet energy resolution . by definition
, event - shape variables are expected to be rather robust against both sources of systematic errors .
we assume a global uncertainty of 10% on the knowledge of the jet energy scale .
the resulting normalized event - shape distributions deviate by 5 - 10% from the reference distribution over the whole energy range as can be seen in fig .
[ fig : calo_thrust ] .
the effect of the jet energy resolution is studied by applying the jet energy resolution smearing function [ eq : smear_ass ] on generator level jets : @xmath15 the smeared event - shape distributions deviate by less than @xmath16 from the unsmeared distribution over most of the energy range ( fig .
[ fig : smear ] ) .
in order to demonstrate the sensitivity of hadronic event - shape distributions to different models of multi - jet production , we compare the central transverse thrust and thrust minor distributions to the generator level predictions as obtained from two generators that contain different models of qcd multi - jet production , @xmath8 and @xmath17 @xcite . the @xmath18 samples used in our study contains qcd processes from 2 up to 6 jets . in fig .
[ fig : alp_pyt_trthr ] the distributions of the central transverse thrust and central thrust minor can be seen .
these events are selected from a jet trigger , based on the calibrated transverse energy of the hardest jet @xmath19 with a prescale of 100 .
the error bars on the data points include the statistical uncertainties corresponding to @xmath20 of integrated luminosity and the systematic errors due to jet energy scale and jet energy resolution as discussed in the previous section .
the corrected calorimeter jets correspond to the samples , and they are found to be compatible with the generator level jets from . it can be seen that there is a significant difference with respect to the distribution , reflecting the different underlying matrix element calculations in the generators and the different parameter choices .
the result shows that hadronic event shapes can be powerful handles in comparing and tuning different models of multi - jet production . with the statistical and dominant systematic errors expected after @xmath20 of integrated luminosity .
the prescale of the trigger is assumed to be 100 .
the distributions are compared to the generator level distributions of @xmath21 and @xmath18 .
, scaledwidth=90.0% ] with the statistical and dominant systematic errors expected after @xmath20 of integrated luminosity .
the prescale of the trigger is assumed to be 100 .
the distributions are compared to the generator level distributions of @xmath21 and @xmath18 .
, scaledwidth=90.0% ]
in this note we demonstrate the use of hadronic event shapes at the lhc .
the event - shape variables are evaluated using calorimeter jet momenta as input .
they are shown to be not very dependent on the effect of jet energy corrections .
we present an estimate of the dominant systematic uncertainties at the startup , resulting from jet energy resolution effects and from the limited knowledge of the jet energy scale . using the examples of the central transverse thrust and central thrust minor
, we show that early measurements of event - shape variables allow to study differences in the modeling of qcd multi - jet production . the cms collaboration , `` study of hadronic event shape variables '' , + cms pas qcd-08 - 003 , _ http://cms-physics.web.cern.ch/cms-physics/public/qcd-08-003-pas.pdf _
t. sjostrand , s. mrenna and p. skands , `` pythia 6.4 physics and manual '' jhep * 05 * ( 2006 ) 026 [ hep - ph/0603175 ] . | in this note a study of hadronic event shapes in qcd events at the large hadron collider ( lhc ) is presented .
calorimetric jet momenta , determined by various jet clustering algorithms , are used as input for calculating various event - shape variables which probe the structure of the hadronic final state .
it is shown that the normalized event - shape distributions are robust under variations of the jet energy scale and resolution effects , which makes them particularly suitable for early data analysis and tuning of monte carlo models . |
lipi public cluster ( lpc ) is a cluster - based computing facility maintained by lembaga ilmu pengetahuan indonesia - lipi ( the indonesian institute of sciences ) @xcite .
although it is still a small scale cluster in the sense of number of nodes already installed , it has unique characteristics among existing clusters around the globe due to its openness . here
`` open '' means everyone can access and use it anonymously for free to execute any types of parallel programmings @xcite .
the development of lpc was initially motivated by real needs for high performance and advanced computing , especially in the research field of basic natural sciences . even in indonesia
, the needs are growing along with the advances of scientific researches . in the last decades , clustering low specs ( and low cost ) machines becomes popular to realize an advanced computing system comparable to , or in most cases better than the conventional mainframe - based system with significant cost reduction @xcite .
in general a cluster is designed to perform a single ( huge ) computational task at certain period .
this makes the cluster system is usually exclusive and not at the level of appropriate cost for most potential users , neither young beginners nor small research groups , especially in the developing countries like indonesia .
it is clear that the cluster is in that sense still costly , although there are certain needs to perform such advanced computings .
no need to say about educating young generations to be the future users familiar with parallel programmings .
this background motivates us to further develop an open and free cluster environment for public @xcite . according to its nature lpc is , in contrast with any conventional clusters , designed to accommodate multiple users with their own parallel programmings executed independently at the same period .
therefore an issue on resource allocation is crucial , not only in the sense of allocating hardwares to the appropriate users but also to prevent any interferences among them . in lpc
we have deployed a new algorithm to overcome this problem , namely the dependent @xcite and independent multi - block approaches @xcite . concerning its main objective as a training field to learn parallel programmings , the public cluster should be accessible and user - friendly for all users with various level of knowledges on parallel programming .
it also should have enough flexibility regarding various ways of accessing the system in any platforms as well .
this can be achieved by deploying web - based interfaces in all aspects .
presently we have resolved some main issues , such as security from anonymous users to prevent any kinds of interference among different tasks running simultaneously on multi blocks @xcite , algorithm for resource allocation management @xcite and the real - time monitoring and control over web for both administrators and end - users @xcite . in this paper
we first present briefly the concept of lpc including the work flow from the initial registration to the execution of computational works .
thereafter we discuss the main part of this paper , that is the architecture of web - interface in lpc .
finally we conclude with some comments and discussion .
in order to overcome the issues mentioned in the preceeding section , it is clear that we should develope and utilize an integrated web - based interface in lpc .
however , concerning real demands in indonesia presently , we assume that our potential users would be beginners in parallel programmings who are going to use it moreless for educational or self - learning purposes .
although the cluster is also going to be used by some experts and larger research groups to perform more serious advanced computings , we do not expect any anonymous users with heavy computational works .
the reason is mainly because of the limited resources , i.e. number of nodes and its specifications .
actually we rather prefer as many as people use our cluster as a training field to learn parallel programmings . in average
we plan to provide only @xmath0 nodes in a relatively short period , namely less than 3 days , for each anonymous user .
these characteristics completely differ with another existing clusters around the world . because they are usually used by certain people or groups bringing similar types of computational works which are appropriate for the cluster after particular procedures like submitting proposals and any kinds of letter of agreements . in our case , incoming users are completely anonymous and then there is no way in any means to know the type of computational works being executed in the cluster .
this fact consequently brings another problems as maintaining job executions owned by different users at same period simultaneously . under these assumptions and conditions
, the public cluster should fulfill very challenging requirements that might be irrelevant in any conventional clusters , that is : * security : + this is the main issue we should encounter from the first .
the users should have enough priviledges and freedom to optimize their opportunities to learn parallel programming , while their access must be limited at the maximum level for the sake of security of the whole system and another active users at the same period .
* flexibility : + it is impossible to provide the same level of flexibility for anonymous users as well - defined users with direct accesses through , etc , but we should allow as much as possible the users to execute their codes on lpc . also there should be a freedom on assigning the number of nodes for each user , since each user could require various number of nodes depending on the computational capacities they actually need .
* stability : + simultaneous executions by different users with various programmes in different blocks of cluster without any interferences among them requires new innovations on cluster management .
this problem includes some techniques to incorporate wide range of nodes with different specifications , that is ranging from intel 486 to the latest athlon based nodes . *
efficiency : + since the cluster is dynamically divided into several blocks with various number of nodes inside according to the users requests , each node should be able to be completely turned on or off partially without any interruptions to another working nodes . all of these require modifications on some existing tools , and also new developments on alternative softwares and hardwares as well @xcite .
regarding to the above - mentioned concept , we have defined the work flow in lpc as follows : 1 .
a new user should complete an initial registration by providing personal data , the content of job will be performed and the number of nodes requested for the job .
the application is reviewed and verified by the administrator .
after approval , the administrator will assign the nodes provided and its usage period .
3 . after reconfirmations by user , proving their agreements with the provided nodes and usage period
, the administrator will switch the nodes on that also activate all daemons automatically .
4 . the users should adjust their parallel programmes to fit the provided nodes .
the user uploads all necessary programmes and libraries if any . at this stage
the programmes can be executed immediately .
the administrator and automated system will monitor the usage of all running users .
. finished jobs and the results can be downloaded by owners
once the usage period is over , the nodes are turned off automatically .
again , these procedures can be done fully and remotely through web . we should mention that the types and number of nodes allocated for a newly assigned block is done automatically utilizing a decision making tool based on the extended genetic algorithm embedded in the web - interface @xcite .
now we are ready to discuss the architecture of web - interface in lpc .
as mentioned earlier , the web - interface takes over all aspects in lpc , from communication among the nodes till running related tools and softwares to perform computational works .
it is also the only way connecting the internal system with the rest of the world . in term of communication , there are two kinds of communication in lpc : a ) between users and gateway server through http protocol , and b ) between gateway server and nodes consisting blocks of cluster through or .
this architecture is intended to limit direct access to a block of cluster for security reason . on the other hand ,
this still allows the system to interact with another elements of cluster as io server and master nodes of each block . as depicted in fig .
[ fig : komunikasi ] , we put a gateway server as a common entrance for users which serves all related web - based applications .
the gateway server further sends and receives commands related to file or parallel programming operations using secure remote login / .
the web application in lpc should accommodate the needs of any parallel programmings which vary and change along the time .
then we must make it as flexible as possible to guarantee its compatibilities in the future .
also , any future modifications in the web application must not alter another software components .
this can be achieved by adopting three - tier software architecture @xcite .
this method is appropriate for distributed client / server applications which require high performances in terms of flexibility , maintainability , reusability and scalability , while at the same time it simplifies the complexities of users distributed processes .
this architecture is shown in fig .
[ fig:3layer ] . as can be seen in the figure , each layer
is completely separated which could then simplifies its developments and implementations .
presentation layer behaves as a front interface and interacts directly by receiving inputs from users .
the html files in this layer could contain either static or dynamic ( through application and logic layer ) contents .
for instance in the case of lpc they are generated dynamically by python scripts .
the application and logic layer handles all programming logics to serve user needs as uploading the jobs , executing codes and so forth .
further , the data access layer provides the access to data files from the database system , configuration files or generated by any commands in operating system .
next , the main task for web - interface in lpc is providing the users an appropriate environment for parallel programming . at time being , there are several parallel programming environments like mpi , pvm and so on @xcite .
also some modified versions of them as mpich ( version 1 and 2 ) and lam - mpi .
the web - interface should allow users to choose the preferred one , or to change from one to another . in order to enable dynamic change of environment according to user preferences
, we have deployed a mechanism based on the modules package @xcite .
this package enables environment switching by changing its path and variables through a single command .
further issue is how to keep the whole performance of cluster during operation .
this can be done by embedding `` real - time '' control and monitoring system and displaying the results through web .
we have developed a dedicated device for this purpose using microcontroller - based hardwares @xcite .
the control system manages the hardware aspect of each node , for instance turning them on or off .
the microcontrollers are interfaced by python and some external programmes in c++ .
the monitoring system takes the data of external physical observables such casing temperatures and humidities .
the physical parameters for each node can be retrieved easily through bios . both control and monitoring systems are integrated to realize an automated hardware control , i.e. shuting down a node if the temperature exceeds the predefined threshold , etc . beside these hardware - based management systems , we have also implemented the software - based management system to control all nodes more efficiently .
we have deployed a tool set called cluster command and control ( c3 ) @xcite and ganglia @xcite .
c3 bundles useful commands to enable automatic processes throughout multiple nodes , while ganglia is used to monitor the current condition of cluster through web - interface .
we have integrated these tools in our web - interface to create a more comprehensive web - based interface matches specific needs in lpc .
lastly , we should comment on the communication protocol among the nodes in lpc .
all commands are sent to nodes through secure shell login / passwordlessly using rsa / dsa keys .
this handy solution can be used for any commands , either with or without standard outputs .
the flow of commands between nodes and user through gateway server is diagrammatically given in fig .
[ fig : command ]
we have introduced the web - interface in lpc which integrates all aspects of its hardware and software .
although web - based interfaces in clusters are commonly known and actively developed by many groups , the web - interface for public clusters is quite unique and requires more careful developments .
moreover , in any existing conventional clusters , the web - interfaces are moreless complementary tools . while in public clusters like lpc
, the web - interface is very crucial and the only solution to make it open for public .
we would like to comment some points regarding the current status of lpc : * in some specific cases , the work - flow in lpc can be automated to reduce administration works .
for instance , it could be realised for a `` public '' cluster limited to a community with predefined users and types of computational works . however , in our case the present work flow is the simplest and most `` automated '' one regarding the security and its level of openness . *
related to this , the main issue in daily operation is to prevent in advanced malicious users who try to submit huge never - ending computational tasks .
actually , this is themain reason we keep current work - flow which is quite powerfull to deal with this potential problem .
* at time being , the computational results should be downloaded by users .
although it is also possible to make a real - time monitoring for on - going works , we are not going to implement such services in the future . because it is irrelevant in parallel programmings which could normally take few days computational time .
another reason is such services could exhaust the limited resources of the system , then it is not worthy compared to its importances .
* the usage time provided for an approved user is determined based on the their request since we do not confirm their code in detail . however , as our basic policy , we provide in most only several days usage time for each anonymous user in one new request .
* we have implemented our own resource allocation algorithm based on the extended genetic algorithm embedded in the web - interface @xcite .
this is different with the conventional cluster where the resource allocation should be performed during the computational period to achieve the best performance . in the public cluster like lpc ,
the conventional resource allocation is irrelevant , because it is done only at the first time of node assignment after the approval procedure .
once a block consisting of some resources has been allocated , it remains unchanged during the usage period .
of course , resource allocation within a block is still relevant , but it uses the existing tools chosen by users .
moreover , doing an appropriate load - balancing within a block used by user is an important aspect of learning parallel programming , so this task should be done by the users themself . *
the lpc so far enjoys numerous visits and many active users as can be seen in its web @xcite , ranging from high school to college students .
this fact has definitely proven the usability of the system by ( mostly ) beginners in parallel programmings . *
the current system is in principle portable to other existing clusters .
especially the concept could be an interesting alternative for some old clusters that might still be usefull for public educational purposes . as future work , we are going to investigate more comprehensively the overall performances , especially in the case of simultaneous heavy computational jobs of multiple users .
however , the current performances have been reported partially in our previous works @xcite . also following recent progress in high performance computing , we are in the step of connecting some blocks in lpc into grid computing .
this work is financially supported by the riset kompetitif lipi in fiscal year 2007 under contract no .
11.04/sk / kppi / ii/2007 and the indonesia toray science foundation research grant 2007 . 99 lipi public cluster , http://www.cluster.lipi.go.id .
handoko : `` public cluster : mesin paralel terbuka berbasis web '' , indonesian copyright , no .
b 268487 ( 2006 ) .
a. goscinski , m. hobbs and j. silcock : `` a cluster operating system supporting parallel computing '' , cluster computing , vol .
4 , pp . 145156 ( 2001 ) . z. akbar , slamet , b.i .
ajinagoro , g.j .
ohara , i. firmansyah , b. hermanto and l.t .
handoko : `` open and free cluster for public '' , proceeding of the international conference on rural information and communication technology 2007 , bandung , indonesia,2007 .
z. akbar , slamet , b.i .
ajinagoro , g.j .
ohara , i. firmansyah , b. hermanto and l.t .
handoko : `` public cluster : parallel machine with multi - block approach '' , proceeding of the international conference on electrical engineering and informatics , bandung , indonesia , 2007 .
z. akbar and l.t .
handoko : `` multi and independent block approach in public cluster '' , proceeding of the 3rd information and communication technology seminar , surabaya , indonesia , 2007 .
z. akbar and l.t .
handoko : `` resource allocation in public cluster with extended optimization algorithm '' , proceeding of the international conference on instrumentation , communication , and information technology , bandung , indonesia , 2007 .
i. firmansyah , b. hermanto , slamet , hadiyanto and l.t .
handoko : `` real - time control and monitoring system for lipi public cluster '' , proceeding of the international conference on instrumentation , communication , and information technology , bandung , indonesia , 2007 .
j. foreman , j. gross , r. rosenstein , d. fisher , k. brune : `` c4 software technology reference guide - a prototype '' , handbook ( cmu / sei-97-hb-001 ) , software engineering institute , carnegie mellon university , january 1997 . for examples , see : + message passing interface forum : `` mpi2 : a message passing interface standard '' . international journal of high performance computing applications , 12(12):1299 , 1998 .
john l. furlani : `` modules : providing a flexible user environment '' , proceedings of the fifth large installation systems administration conference ( lisa v ) , pp . 141 - 152 , san diego , ca , september 30 - october 3 , 1991 .
ray flanery , al geist , brian luethke and stephen l. scott : `` cluster command & control ( c3 ) tool suite '' , 3rd austrian - hungarian workship on distributed and parallel systems ( dapsys 2000 ) in conjunction with europvm / mpi 2000 , balatonfured , lake balaton , hungary , september 10 - 13 , 2000 .
http://www.csm.ornl.gov / torc / c3/. ganglia monitoring system , http://ganglia.sourceforge.net . | a web - based interface dedicated for cluster computer which is publicly accessible for free is introduced .
the interface plays an important role to enable secure public access , while providing user - friendly computational environment for end - users and easy maintainance for administrators as well . the whole architecture which integrates both aspects of hardware and software
is briefly explained .
it is argued that the public cluster is globally a unique approach , and could be a new kind of e - learning system especially for parallel programming communities . |
the total instantaneous radiated power@xcite of a particle carrying charge q is the larmor formula for non - relativistic motion @xmath0 and the linard formula for relativistic motion @xmath1\ ] ] in this paper , we use gaussian units with @xmath2 the speed of light , @xmath3 the usual lorentz factor @xmath4 , dots @xmath5 signifying time - derivatives and @xmath6 , where @xmath7 is the particle s velocity .
also , the neutrinos considered here are dirac neutrinos with magnetic dipole moments @xmath8 . in this paper , @xmath9 will always denote rest frame magnetic moments . using the same convention as @xcite
, the electromagnetic field notation will be : unprimed quantities such as @xmath10 denotes laboratory quantities ( here a laboratory magnetic field ) and the @xmath11 , such as @xmath12 , denotes rest frame quantities ( here a rest frame magnetic field ) .
we will find that the corresponding formulae for radiation power losses of magnetic dipoles are much more complicated than eq(1,2 ) of charged particles because @xmath13 @xmath14 in the early universe , once cosmological neutrinos decouple from matter , conventional wisdom posits that they would then lose energy adiabatically from the cosmic expansion @xcite .
however , neutrinos will still interact with early - universe magnetic fields @xmath10 .
such primordial fields are thought to be non - uniform ( fluctuating ) and large @xmath15 g @xcite .
we will show that these early - universe magnetic field conditions can allow efficient neutrino cooling after they decouple from matter .
reference @xcite gives the energy radiated per unit solid angle per unit frequency interval for a moving magnetic dipole moment ( @xmath16 is the look unit vector ) : @xmath17e^{i\omega ( t - \vec{n } \cdot \vec{r}(t)/c)}|^{2}\ ] ] integration by parts yields the time - dependent vector potential @xmath18 @xmath19 \}\ ] ] doing the differentiation , we get our first result , which is the non - relativistic angular distribution @xmath20|^{2}\ ] ] interestingly , the ` jerk ' @xmath21 contributes , which is not the case for an accelerated charge . altogether , there are 10 separate terms ( compare to eq(1 ) above ) for the instantaneous radiation of a non - relativistically moving magnetic dipole , obtained by integrating eq(7 ) @xmath22\ ] ] eq(8 ) will be applied to solar neutrons to see if magnetic dipole moment radiation is a factor in their detection in the inner parts of the solar system .
we are interested in solar neutrons associated with large flares .
a typical energy is 8 - 10 mev @xcite .
the neutrons can be modeled as originating near the solar limb and going through a high magnetic field @xmath23 10@xmath24 g ( i. e. about 1 tesla ) where such strength extends to about one solar diameter away .
the question is whether or not these neutrons radiate energy away before they leave the sun s environment .
because of the tenuous low density plasma that they transit and because of their low interaction nuclear cross section , solar neutrons can be considered to be a collisionless particle . however , the neutron will experience forces due to the solar magnetic field .
the force equation is @xmath25 where , as already noted , @xmath9 and @xmath12 are rest frame fields . working this out
gives @xmath26 we need the jerk , so we have to take the time derivative of eq(10 ) to obtain : @xmath27 the next vector equation is the torque equation ( no collisional relaxation term present and since we take @xmath28 ) @xmath29 where @xmath30 is the spin of the neutron .
the connection between the spin @xmath30 and the magnetic moment @xmath9 is @xmath31 where @xmath32 is a constant .
so the third vector equation is @xmath33 finally , we need @xmath34 given by taking the time derivative of eq(13 ) : @xmath35
the neutron magnetic moment @xcite is @xmath36 where @xmath37 is the nuclear magneton .
if we assume that the solar magnetic field has small variations and fluctuations , then only the first term in eq(8 ) contributes , with @xmath34 given by eq(14 ) , which has the maximum value @xmath38 the space ( ` labatory ' ) frame of one tesla for the sun s environment translates to a higher strength value in the neutron s rest frame by a @xmath3 .
since the neutrons are non - relativistic ( @xmath39 ) , we neglect any lorentz transformation involved . working this out for a nominal one tesla @xmath12
, we find @xmath40 even if we include a non - stationary solar magnetic field component ( therefore other terms in eq(8 ) come into play ) , neutron radiation losses are completely negligible for this problem .
from the last section , it is obvious that non - negligible magnetic moment radiation requires great magnetic field strengths and ultra - relativistic particle motion .
@xmath41 , eq(8 ) , in section 2 and the torque equation , eq(12 ) , and force equation , eq(9 ) , both in section 3 , all have to be generalized . in this section
, we present the generalization of the instantaneous power radiated .
the instantaneous power is a lorentz scalar , so time is replaced with proper time , a scalar .
lorentz tensors come into play , but they have to be contracted together to get a scalar .
the formula clearly has two separate quantities reflecting the two separate terms in the non - relativistic form .
the first term is the double time derivative of the magnetic moment .
magnetic moments are entries of the anti - symmetric dipole tensor @xmath42 @xcite @xmath43 where @xmath9 is the magnetic moment vector and @xmath44 is the electric dipole vector , both in the rest frame of the particle .
experimentally , no elementary particle has been found ( particle data group @xcite ) that carries an intrinsic ( rest frame ) non - zero @xmath44 so we set this term to zero .
the first term thus becomes @xmath45 \rightarrow \frac{2}{3 c^{3 } } [ \frac{1}{2 } \frac{d^{2 } d^{\mu\nu}}{d \tau^{2 } } \frac{d^{2 } d_{\mu\nu}}{d \tau^{2}}]\ ] ] using @xmath46 and @xmath47 we get @xmath45 \rightarrow \frac{2}{3 c^{3 } } [ \gamma^{8 } ( \vec{\beta}\cdot \dot{\vec{\beta}})^{2 } \dot{\vec{\mu}}^{2 } + 2 \gamma^{6}(\vec{\beta}\cdot \dot{\vec{\beta } } ) ( \dot{\vec{\mu}}\cdot \ddot{\vec{\mu } } ) + \gamma^{4}\ddot{\vec{\mu}}^{2}]\ ] ] the remaining is the piece @xmath48 so the second and final term in the instantaneous power emitted is @xmath49 \rightarrow \frac{2}{3 c^{3 } } [ \frac{1}{m^{2}c^{2 } } \frac{d^{2}(d^{\nu\mu}p_{\mu})}{d \tau^{2 } } \frac{d^{2}(d_{\nu\lambda}p^{\lambda})}{d \tau^{2}}]\ ] ] hence , the relativistic instantaneous power @xmath50 radiated from a magnetic dipole moment is @xmath51\ ] ] working out the algebra gives the final answer @xmath52\end{aligned}\ ] ] eq(25 ) is the relativistic magnetic dipole moment radiation loss equivalent to the linard formula eq(2 ) for charged particles .
eq(25 ) can be applied immediately to electron synchrotrons , using an example , the cornell electron storage ring @xcite .
electrons will radiate magnetic dipole radiation in conjunction with their usual charge radiation .
since the machine designers have not accounted for the former , it must be truely tiny , even for large accelerators . for circular machines ( call the loss @xmath53 ) ,
the energy change per revolution is small , so @xmath54 , and since @xmath55 , @xmath56 , the only term in eq(25 ) that is present is @xmath57^{2}\ ] ] now @xmath58 and assuming perpendicularity between @xmath21 and @xmath9 @xmath59 the acceleration @xmath60 where @xmath61 is the radius of the synchrotron .
thus @xmath62 . with @xmath63 , the radiative energy loss per revolution @xmath64 is @xmath65 expressing the energy loss per revolution in @xmath66 , @xmath67 , we find @xmath68^{4}}{r[meters]^{3}}\ ] ] where the beam energy @xmath69 $ ] is expressed in @xmath70 and the accelerator radius @xmath71 $ ]
is expressed in meters .
for the cornell machine , @xmath72 \simeq 12 $ ] and @xmath71 \simeq 122 $ ] giving @xmath73 which , indeed , is completely ignorable .
though the solar neutron problem and the electron synchrotron problem have vanishingly small radiative dipole power losses , we will see that cosmological neutrinos in the early universe may dramatically cool down from this mechanism . to do the calculation , we must have the generalization of the non - relativistic force equation and the non - relativistic torque equation .
of the two generalizations , the former is trivial and the latter complicated . introducing the velocity 4-vector @xmath74 @xmath75 and
the gradient 4-vector @xmath76 @xmath77 the generalization of eq(9 ) is @xmath78 where @xmath79 is the invariant rest mass and @xmath80\ ] ] we will need the expression @xmath81 so reducing eq(34 ) gives @xmath82
we have to generalize eq(12 ) to make it covariant . the resulting equation for charged particles@xcite
is the thomas@xcite - bargmann@xcite - michel - teledi equation . for want of a better name , the neutron / neutrino equation we need
will just be called the neutrino moment ( nm ) equation .
we follow @xcite .
the relativistic spin in the rest frame will be denoted by @xmath30 and the spin in the laboratory frame denoted by @xmath83 , where this relativistic spin is an axial 4-vector .
recall that spin is an intrinsic quantum operator where the value of @xmath84 in @xmath85 has a quantum origin .
in particular , the quantum 3-spin @xmath86 satisfies @xmath87 = i \epsilon^{klm}\sigma^{m}$ ] .
because the quantum origins of @xmath84 may preclude a direct evaluation from first principles ( e. g. the anomalous g value of the nucleons ) , the relativistic formalism finesses the possible unknown value of the g term , by making it a parameter .
the relationship between @xmath30 and @xmath88 is @xcite @xmath89 where @xmath83 satisfies @xmath90 as given in @xcite , the possible covariant terms for the time evolution of @xmath83 is @xmath91 multiplying eq(39 ) by @xmath92 and using eq(38 ) , we determine that @xmath93 and @xmath94 .
since eq(39 ) must reduce to eq(12 ) when @xmath95 , then @xmath96 and we obtain the covariant equation @xmath97 -\frac{u^{\alpha}}{c^{2}}(s_{\lambda}\frac{d u^{\lambda}}{d \tau})\ ] ] we process this equation , eventually obtaining an equation for the @xmath98 @xmath99 + \frac{\gamma}{1+\gamma}[\vec{s } \times ( \vec{\beta } \times \frac{d \vec{\beta}}{d \tau})]\ ] ] finally giving the nm equation @xmath100 + \frac{\gamma}{1+\gamma}[\vec{\mu } \times ( \vec{\beta } \times \frac{d \vec{\beta}}{d \tau})]\ ] ] where @xmath81 is given by eq(36 ) in conjunction with eq(35 ) . in eq(42 ) , @xmath32 is the constant in @xmath101 so @xmath102
we already know that non - negligible magnetic dipole radiation loss from individual elementary particles requires magnetic field strengths comparable to those found in neutron stars . early universe cosmological neutrinos that decouple from matter at temperatures @xmath103 2.6 mev @xcite have energy @xmath104 3.9 mev giving a @xmath105 ( since @xmath106 ev ) with fluctuating magnetic field strengths @xmath15 g @xcite . in this section
, we estimate their power loss after they decouple from matter .
the key is the existence of the @xmath107 terms of eq(25 ) . for decoupling neutrinos , with @xmath108 ,
these terms convey an enormous @xmath109 enhancement .
even still , there needs the presence of fluctuating magnetic fields of neutron star strength , to overcome the small neutrino magnetic moment @xmath110 , value given by the particle data group @xcite @xmath111 where @xmath112 is the bohr magneton .
the standard model allows a massive dirac neutrino to possess a tiny magnetic dipole moment @xcite @xmath113 where @xmath114 is the neutrino mass in ev .
however , most extensions of the standard model predict anomalous magnetic dipole moments that approach the experimental limit given by eq(44 ) . for example , reference @xcite gives an extension of the minimal supersymmetric standard model by including a vector - like leptonic generation which arises in many grand unfied theories and string models .
this leads to neutrino magnetic dipole moments many orders of magnitude larger than the standard model prediction and a clear signal for new physics . from eq(42 ) , as long as @xmath115 is not parallel to the cosmological magnetic field @xmath10 , then @xmath116 is dominated by the first term of eq(42 ) , for large fields @xmath117 characterizing the fluctuating cosmological field by its impact on the neutrino by @xmath118 and making the parameterizations @xmath119 with @xmath120^{2}$ ] then @xmath121 \nonumber \\ & \sim & 2.19 \ ; \omega^{2 } 10^{2a - 2x + 9 } \ ; { \rm mev / s } \end{aligned}\ ] ] there are certainly regions , fig . 1 , of parameter space
@xmath122 where the magnetic dipole moment radiation leads to fast neutrino cooling .
it can be argued successfully that any physics associated with magnetic moments is subtle : the electric charge is a lorentz scalar , whereas the magnetic dipole moment is a significant component of a second order anti - symmetric tensor , ensuring complicated dynamics .
it was well known that ordinary magnetic fields can not lead to measurable magnetic moment radiation losses , but the enhancement for relativistic particles was not anticipated .
because cosmological neutrinos and anti - neutrinos are not expected to mutually annihilate in the expansion of the universe as do the charged particles , there would be an almost unlimited number of neutrinos and anti - neutrinos surviving the big bang .
if these cosmological neutrinos and anti - neutrinos radiate energy after they decouple from matter , and if this loss is a significant fraction of their thermal energy , they can condense and be a component of dark matter .
condensed neutrino matter would then form the largest structures in the universe , dwarfing the visible galaxies in size and mass .
this paper outlines a viable cooling mechanism that allows such a scenario .
further work requires a detailed simulation of neutrino primordial density fluctuations in concert with a full electromagnetic treatment of early universe magnetic field generations in order to answer the question of neutrino condensation timelines .
99 j. d. jackson , _ classical electrodynamics _ , second edition , wiley , new york ( 1975 ) e. w. kolb and m. s. turner , _ the early universe _
, addison - wesley , reading , mas .
( 1990 ) d. grasso and h. r. rubinstein , phys . rep . * 348 * 163 ( 2001 ) jacques m. wagstaff , et al . , arxiv:1304.4723v1 [ astro-ph.co ] g. h. share et al , arxiv:1007.2349v1 [ astro-ph.sr ] j. beringer et al , phys . rev .
d * 86 * , 0100001 ( 2012 ) a. peletminskii and s. peletminskii , eur .
j. c * 42 * , 505 ( 2005 ) http://www.lns.cornell.edu/research/ap/cesr/ l. h. thomas , nature * 117 * , 514 ( 1926 ) ; philos . mag . * 3 * , 1 ( 1927 ) v. bargmann , l. michel and v. l. telegdi , phys .
* 2 * , 435 ( 1959 ) k. fujikawa and r. shrock , phys .
* 45 * 963 ( 1980 ) amin aboubrahim et al . , arxiv:1312.2505v2 [ hep - ph ] | we compute the power radiated per unit solid angle of a moving magnetic dipole moment , and its instantaneous radiated power , both non - relativistically and relativistically .
this is then applied to various interesting situations : solar neutrons , electron synchrotrons and cosmological dirac neutrinos .
concerning the latter , we show that hypothesized early - universe big bang conditions allow for neutrino radiation cooling and provide an energy loss - mechanism for subsequent neutrino condensation .
instantaneous power radiated from magnetic dipole moments p.d .
morley + system of systems analytics , inc .
+ 11250 waples mill road + suite 300 + fairfax , va 22030 - 7400 + and + d.j .
buettner + aerospace corporation + p. o. box 92957 + los angeles , ca 90009 - 2957 keywords : neutrinos , dipole radiation , condensation |
the advent of high - resolution optical imaging and follow - up spectroscopic surveys in the past two decades , has led to the recognition of a morphologically distinct class of stellar assemblies in and around external galaxies , the first few cases of which were found in the fornax cluster @xcite . these so - called ultra - compact dwarfs ( ucds ; @xcite ) manifest themselves as compact objects with typical effective radii of 10 @xmath5 pc and absolute v - band magnitudes of -14 @xmath6 -9mag , just intermediate between the classical globular clusters ( gcs ) and dwarf elliptical galaxies .
when spectroscopic information is available , ucds appear to harbor a predominantly old stellar population ( e.g. , @xcite ; janz et al .
2016 ) , in some cases with an extended star formation history ( e.g. , @xcite ) .
candidate ucds are now routinely found in dense environments such as galaxy clusters ( e.g. , fornax , @xcite ; virgo , @xcite ; centaurus , @xcite ; coma , @xcite ; perseus , @xcite ) and galaxy groups ( e.g. , hcg22 and hcg90 , @xcite ) , but also in relatively isolated galaxies such as the sombrero (= m104 ; @xcite ) , ngc3923 and ngc4546 @xcite .
the nature of ucds , however , is far less clear than their names might have indicated .
viable formation scenarios proposed for ucds include : ( i ) they are the residual cores of tidally - stripped nucleated dwarf galaxies in dense environments @xcite ; ( ii ) they are the end - product of the aggregation of young massive star clusters formed during violent gas - rich galaxy mergers @xcite ; and ( iii ) they are ultra - massive gcs , as expected by extension of the gc luminosity function @xcite . adding to the mystery of ucds is a closely - relevant , growing class of stellar systems , namely , extended stellar clusters ( escs ) . with visual luminosities largely overlapping the typical range of gcs ( -9 @xmath6 -5 ) , the escs earn their names for their characterisic sizes ( 10 @xmath7 pc ) that are much larger than the classical gcs .
a number of escs have also been found in galactic disks ( e.g. , ngc1023 and ngc3384 , @xcite ) , which are not easily explained by the scenario of stripped satellites .
recently , brns & kroupa ( 2012 ; hereafter bk12 ) compiled from the literature a catalog of 813 confirmed and candidate ucds and escs . in reality
, the known ucds and escs may represent a heterogeneous family of stellar systems that the above formation scenarios conspire to form .
studies of ucds and escs have been concentrated on optical observations and theoretical investigations .
up till now we know very little about their properties at other wavelengths , in particular the x - ray band .
for gcs , it has long been established that they exhibit an over - abundance of low - mass x - ray binaries ( lmxbs ) with respect to the field ( @xcite ; @xcite ) , which is attributed to the efficient dynamical formation of neutron star binaries in the dense core of gcs ( @xcite , @xcite , @xcite ) .
x - ray surveys of extragalactic gcs , mostly accomplished by the _ chandra x - ray observatory _ , have revealed that on average @xmath05% of gcs exhibit an x - ray counterpart ( presumably lmxbs ) at a limiting luminosity of @xmath8 , and that more massive gcs have a higher probability of hosting lmxbs @xcite .
if some ucds ( and escs ) are indeed giant versions of gcs , they are naturally expected to host lmxbs .
the first ucd reported to produce x - ray emission lies in m104 ( sucd1 ; @xcite ) , which has an x - ray luminosity of @xmath9 @xcite . in view of the rapidly growing inventory of ucds and escs
, it would be interesting to examine the incidence rate of x - ray sources in these dense stellar systems , which should provide clues about their internal structure .
on the other hand , many ucds found in galaxy clusters can well be the remnant of stripped nucleated galaxies .
the recent discovery of a candidate super - massive black hole ( with a mass of @xmath0@xmath10 ) in the ucd of m60 ( m60-ucd1 ; @xcite ) , based on spatially - resolved stellar kinematics , lends strong support to this hypothesis .
strader et al .
( 2013 ) identified a variable x - ray source with m60-ucd1 .
x - ray emission could be an important tracer of this and other putative massive black holes embedded in ucds .
in this work , we conduct a systematic survey of x - ray emission from ucds and escs using archival _
observations , in order to shed light on the origin and nature of these intriguing objects . in section 2
we describe the sample selection , data reduction and x - ray source detection . in section 3
we present the x - ray counterparts of the ucds and escs and analyze their properties in close comparison with the x - ray population found in gcs .
discussion and summary of our results are given in sections 4 and 5 , respectively .
quoted errors are at the 1@xmath11 confidence level throughout this work unless otherwise noted .
we adopt the bk12 catalog of ucds and escs as our primary sample , which results from an exhaustive search of the literature available at the time .
we note that only a ( growing ) fraction of the cataloged sources have been spectroscopically confirmed .
nevertheless , bk12 showed that the spectroscopically confirmed sources are statistically representative of the full sample , in particular in the @xmath12 plane .
the ucds and escs were collectively dubbed extended objects ( eos ) in bk12 to reflect the unclear physical distinction between the two groups , a naming convention we follow in this work .
we caution that the term eo " should only be treated as a technical , rather than physical , distinction from the more compact gcs .
the 813 cataloged sources were tentatively associated with 65 host galaxies ranging from dwarfs to milky way - like normal galaxies , and to giant ellipticals embedded in galaxy clusters .
we dropped sources without explicit celestial coordinates .
we also neglected sources associated with local group galaxies or with m87 , the cd galaxy of the virgo cluster .
sources within the local group are essentially well - studied escs , while m87 hosts a large and still growing population of ucds @xcite that deserve a joint optical / x - ray investigation elsewhere .
lastly , we excluded sources located in the perseus and coma clusters , chiefly to avoid potential bias introduced by the relatively poor x - ray source detection sensitivity , which is expected for the large distances and substantial local background caused by the hot intra - cluster medium in these two massive galaxy clusters .
we cross - correlated the remaining sources with the _ chandra _ archive , requiring that all sources of interest fall within 8 arcmin from the aimpoint of an acis - i or acis - s observation , to ensure good sensitivity and accurate source positioning . in this way we selected 449 sources from the bk12 catalog .
we then added to the bk12 sources 60 escs recently identified in ngc1023 from _ hst _ imaging ( @xcite ) , all of which have an effective radius @xmath13 10 pc , including an _
@xmath14cen_-like object with @xmath15 of 10 pc and @xmath16 of -8.9 mag ( hereafter ngc1023-eo1 ) . we have converted the @xmath17-band magnitudes given in forbes et al .
( 2014 ) into v - band magnitudes , assuming an underlying simple stellar population with age of 10 gyr , half - solar metallicity and a kroupa initial mass function ( imf ) .
we also added two recently discovered ucds that are among the most massive ucds ever found ( m60-ucd1 : @xcite , @xcite ; m59-ucd3 : @xcite , @xcite ) .
our final sample consists of 511 eos , which are distributed in 27 host galaxies . among them , 5 galaxies had only one observation and the other 22 with multiple exposures .
data of the queried observations are publicly available by june 2015 .
we reprocessed the _
chandra _ data with ciao v4.7 and the corresponding calibration files , following the standard procedure .
briefly , we produced count and exposure maps in the 0.5 - 2 ( @xmath18 ) , 2 - 8 ( @xmath19 ) , and 0.5 - 8 ( @xmath20 ) kev bands for each observation .
the exposure maps were weighted by an absorbed power - law spectrum , with a photon index of 1.7 and an absorption column density @xmath21 = @xmath22 . for galaxies with multiple observations , we calibrated their relative astrometry by matching the centroid of commonly detected point sources , using the ciao tool _
reproject_aspect_. the count and exposure maps of individual observations were then reprojected to a common tangential point , i.e. , the optical center of the putative host galaxy , to produce combined images of optimal sensitivity for source detection .
the energy - dependent effective area among the acis ccds was taken into account , assuming the above incident spectrum , so that the quoted count rates throughout this work refer to acis - s3 .
table 1 presents basic information of the host galaxies , including position , distance , effective exposure , and the number of eos within the _ chandra _ field - of - view ( fov ) . at distances of a few mpc and beyond , ucds and escs should remain unresolved even under the superb angular resolution afforded by _
chandra_. to form the basis of identifying x - ray counterpart of the ucds / escs , we ran source detection in the 0.5 - 2 ( @xmath18 ) , 2 - 8 ( @xmath19 ) and 0.5 - 8 ( @xmath20 ) kev bands for each host galaxy , following the procedure detailed in @xcite and @xcite .
the source detection algorithm has been successfully applied for detecting x - ray sources in m104 ( @xcite ) , which is also included in the present study . for each detected source , background - subtracted , exposure
map - corrected count rates in the individual bands are derived from within the 90% enclosed energy radius . according to the assumed incident spectrum ,
we have adopted a @xmath20-band count - rate - to - luminosity conversion factor of @xmath23 , where @xmath24 is the distance to a given galaxy ( table 1 ) .
the 0.5 - 8 kev detection limit for each galaxy field , ranging between a few @xmath25 to a few @xmath26 ( with a median of @xmath27 ) , is given in table 1 .
for the galaxies with multiple observations , we repeated the above procedure in individual exposures to probe long - term source variability .
we search for x - ray counterpart of the ucds and escs from the list of detected sources in each galaxy field , by adopting a matching radius of @xmath28 , which corresponds to a physical scale of @xmath0156 pc at a distance of 16.1 mpc , the median of our sample galaxies .
this choice allows for position uncertainty of sources found at relatively large off - axis angles 50 and off - axis angles of @xmath29-@xmath30 have 95% positional uncertainties of 05 - 18 . ] and the ( typically unknown ) relative astrometry between the x - ray and optical observations .
a total of 17 x - ray counterparts are thus identified , including 1 in ngc1023 , 1 in ngc1399 , 6 in ngc4365 , 1 in m89 , 1 in m59 , 1 in m60 , 2 in m104 ( sombrero ) , and 4 in ngc5128 ( cen a ) .
notably , all eight host galaxies are early - type galaxies , although the _ chandra _ pointed observations might have been biased against late - type galaxies . to assess the probability of random matches , we artificially shift the positions of all detected sources in each field by @xmath31 in ra and dec and average the number of coincident matches in the four directions .
we find essentially zero random matches in all the fields except for ngc4365 , in which the above exercise results in 3.25 random matches .
this can be understood , since ngc4365 has both a large population ( 216 ) of eos and a large number ( 369 ) of detected x - ray sources . by reducing the matching radius to @xmath32
, we find 11 x - ray counterparts , indicating that some of the eliminated matches could indeed be interlopers in particular , 4 eliminated cases are in ngc4365 .
nevertheless , we consider all 17 sources as true x - ray counterparts and provide their basic properties in table 2 , in which we also quote the bk12 catalog for their v - band absolute magnitude and effective radius . from the literature we find that at least 7 of the 17 sources ( ngc1399-eo12 , m104-eo01 , m59-ucd3 , m60-ucd1 , ngc5128-eo03 , ngc5128-eo05 , ngc5128-eo01 ) are confirmed ucds / escs , for which we quote the spectroscopically - derived metallicity in table 2 ; for the remaining sources , we have estimated the metallicity from their optical color , assuming a simple stellar population with age of 10 gyr and a kroupa imf .
roughly an equal number of x - ray counterparts are found in the metal - rich ( [ m / h ] @xmath13 0.3 ) and metal - poor ( [ m / h ] @xmath33 0.3 ) ucds / escs .
figures 1 and 2 show the _ chandra _ 0.5 - 8 kev images of the eight fields , with the positions of the identified x - ray sources marked . among the eight host galaxies , some have a well - documented population of gcs , which are valuable for a direct comparison with the ucds / escs in the observed x - ray properties .
for this purpose , we adopt the published gc catalogs of ngc1399 @xcite , ngc4365 @xcite , m104 @xcite and ngc5128 @xcite . again by adopting a matching radius of @xmath28 , we find 101 , 184 , 65 and 10 x - ray counterparts for the gcs in ngc1399 , ngc4365 , m104 and ngc5128 , respectively ( marked by green circles in figures 1 and 2 ) . the number of gcs within the individual fov ( 5140 in total ) and the number of their identified x - ray counterparts are listed in table 1
. the total incidence rate of x - ray sources in gcs ( hereafter xgcs ) is @xmath34% .
this is to be contrasted with the @xmath2% incidence rate of the x - ray - emitting ucds / escs ( hereafter collectively called xeos ) found in all _ chandra _ fields ( or @xmath35$]% , if only sources in the bk12 catalog are taken into account ) .
we caution that our parent sample of ucds and escs , resulted from literature compilation , is likely more heterogeneous than the gc sample , in terms of completeness . a comparison between our identifications and the x - ray identifications by pandya et al .
( 2016 ) is warranted . from their own literature compilation of ucd candidates , which are distributed primarily in galaxy clusters , pandya et al .
( 2016 ) identified 21 x - ray counterparts by adopting a matching radius of @xmath36 , among which six sources ( ngc1399-eo12 , m104-eo01 , m60-ucd1 ,
ngc5128-eo02 , ngc5128-eo05 and ngc5128-eo01 ) are in common with our xeos .
the remaining 15 sources are not included in our primary sample .
a close examination indicates that some of these sources , considered ucds by pandya et al .
( 2016 ) , have an effective radius of 3 - 8 pc ( i.e. , more typical of gcs ) , and thus would not have appeared in the bk12 catalog . on the other hand , the 11 xeos that are identified by us but not included in pandya et al .
( 2016 ) are essentially sources not in their parent sample . in figure
[ fig : hardness ] , we show the 0.5 - 8 kev intrinsic luminosity ( @xmath37 ) against hardness ratio of the 17 xeos .
the hardness ratio , defined as @xmath38 and listed in table 2 , is calculated from the observed counts in the @xmath18 ( 0.5 - 2 kev ) and @xmath19 ( 2 - 8 kev ) bands , using a bayesian approach @xcite . for comparison ,
we also plot in figure [ fig : hardness ] the 360 xgcs from four host galaxies ( section 3.1 ; table 2 ) , which are presumably lmxbs .
none of the xeos exhibits @xmath39 , i.e. , the regime of ultra - luminous x - ray sources , where black hole binary systems may be relevant .
the majority of xgcs also fall short of this threshold ; only four xgcs in ngc1399 have @xmath39 .
the xeos and xgcs are also similar in the distribution of their hardness ratios .
notably , the ultra - massive m59-ucd3 is the softest among all xeos .
this source is detected in the @xmath18-band but not in the @xmath19-band , thus having a hardness ratio of -1 .
visual inspection of the _ chandra _ image indicates that all 7 photons from m59-ucd3 have an energy below 1.8 kev .
several xeos have sufficient net counts for spectral analysis . for such sources ,
we extract their spectra from a 2@xmath40-radius circle , and the corresponding background spectra from a concentric ring with inner - to - outer radii of @xmath41-@xmath42 .
spectra extracted from multiple exposures of the same source are co - added .
all the spectra appear virtually featureless , and thus we fit them with an absorbed power - law model , requiring that the equivalent hydrogen column density is no less than the galactic foreground value ( kalberla et al . 2005 ) .
we obtain meaningful constraints on the photon - index for m104-eo01 ( i.e. , sucd1 ) , m60-ucd1 , ngc1399-eo12 , ngc4365-eo117 , ngc5128-eo1 and ngc5128-eo5 , with best - fit values of @xmath43 , @xmath44 , @xmath45 , @xmath46 , @xmath47 and @xmath48 , respectively .
these values are consistent with the typical range of lmxbs .
we also examine the long - term flux variability of xeos , based on the source count rates measured from individual observations .
following @xcite , we define source variability @xmath49 , where @xmath50 is the highest count rate among individual detections and @xmath51 the statistical upper limit of the lowest detected count rate .
the values of @xmath52 are listed in table 2 .
a value of @xmath53 is given for the two xeos with only one observation , while ngc4365-eo039 has an ill - defined @xmath52 because it is only detected in the combined image .
the remaining xeos show moderate ( @xmath54 ) to strong ( @xmath55 ) variability , with the strongest variability found in ngc5128-eo01 ( @xmath56 ) .
we note that one xeo , ngc4365-eo006 , has been identified in only one of the six observations available for ngc4365 ; its flux was apparently too low to be detected in the other five exposures as well as in the combined image .
the prevalence of flux variability in the xeos , just as in the xgcs ( @xcite ) , suggest that the bulk of the detected x - ray emission arises from a single source rather than superposition of multiple sources .
figure [ fig : reffmv ] shows the effective radius ( @xmath15 ) versus absolute v - band magnitude ( @xmath16 ) for the 511 ucds and escs studied in this work . by definition , all ucds and escs considered here have a size lower limit of 10 pc .
the apparent paucity of objects around @xmath57 might have arisen from selection effect due to the heterogeneous nature of the bk12 catalog , but otherwise can be viewed as a technical division between ucds and the less luminous escs ( see discussion in forbes et al .
bk12 noticed that the majority of eos in late - type galaxies have @xmath58 .
we highlight the xeos with red diamonds in figure [ fig : reffmv ] , which distribute rather evenly across the entire range of @xmath16 .
ten of the 17 xeos have @xmath59 , whereas six of the 9 most luminous eos ( with @xmath60 ) remain undetected in x - rays . for comparison ,
we show in figure [ fig : reffmv ] the gcs of m104 , among which those with an x - ray counterpart are further highlighted by blue squares .
gcs in the other three galaxies are not shown due to the lack of available v - band magnitudes , but we expect that the m104 gcs are representative .
we note that the great majority ( 94% ) of the gcs have @xmath61 4 pc .
the median v - band magnitude of the gcs is -8.20 mag , and 52 out of the 65 xgcs ( i.e. , 80% ) are found in the brighter half .
this clearly indicates that more massive gcs are more likely to host an lmxb , a familiar trend already noted by many previous work @xcite .
the great majority of eos in our sample remain individually undetected in x - rays , partly owing to the relatively high limiting luminosity for most galaxies , which , at face value , is a good fraction of the eddington luminosity of neutron star binaries .
we employ a stacking analysis for the undetected objects to shed light on their average x - ray properties .
to do so , we collect the 0.5 - 8 kev counts registered within a @xmath62 box around each source of interest .
these can be a subset of our total sample , e.g. , eos in a single galaxy .
an eo is excluded if it is located within @xmath63 from an already detected x - ray source , to minimize contamination from psf - scattered photons .
next , we measure signals from the stacked count image .
after several tests we choose to accumulate the total on - source counts within a @xmath28-radius circle , and estimate the background counts within a ring with inner - to - outer radii of @xmath41-@xmath42 , after scaling the enclosed area .
the signal - to - noise radio ( s / n ) is calculated accordingly .
we have also measured the cumulative exposure time in a similar fashion .
we examine the stacked signals of individual galaxies with at least 10 eos ( table 1 ) . however
, none of these galaxies alone gives a s / n @xmath13 3 .
we further stack all undetected eos from ngc1399 , ngc4365 , m104 and ngc5128 to enhance the s / n , which results in an average count rate of @xmath64 . defining a v - band luminosity - weighted mean distance , @xmath65^{-\frac{1}{2}}$ ]
, where @xmath66 and @xmath67 are the v - band luminosity and distance of the @xmath68th eo , the above count rate corresponds to an equivalent 0.5 - 8 kev luminosity of @xmath69 for a distance of @xmath70 mpc .
similarly , stacking all undetected gcs from the four galaxies , we obtain an average count rate of @xmath71 , or an equivalent 0.5 - 8 kev luminosity of @xmath72 per gc .
this suggests that the individually undetected eos and gcs have on - average comparable x - ray emission .
the similarity in the x - ray properties ( luminosity , spectra and variability ) of the xeos and xgcs ( section 3.2 ) strongly suggests that lmxbs also dominate the x - ray emission from eos , if the presumption that the xgcs are essentially lmxbs holds .
pandya et al .
( 2016 ) drew a similar conclusion .
the presence of lmxbs in the dense , predominately old stellar systems of eos has been naively expected . however , unlike the statistical behavior of the xgcs , the most luminous xeos do not show a clear tendency of hosting an x - ray source ( figure [ fig : reffmv ] ) .
the overall incidence rate of xeos is also substantially low than that of the xgcs ( @xmath03% vs. 7% ; section 3.1 ) .
these findings can be understood as follows .
it has long been recognized that the abundance ( i.e. , number per unit stellar mass ) of luminous lmxbs in gcs is about two orders of magnitude higher than that of the galactic field ( @xcite ; @xcite ) .
this over - abundance is widely accepted as the result of stellar dynamical interactions in the dense core of gcs , where an isolated neutron star ( ns ) can be captured by a main sequence star through tidal force ( @xcite ) , by a giant star through collision ( @xcite ) , or by a primordial binary through exchange scattering ( @xcite ) .
all these processes are governed by the so - called stellar encounter rate , @xmath73 , where @xmath74 is the core stellar density , @xmath75 is the core radius and @xmath11 the velocity dispersion
. the core stellar density of gcs can be as high as @xmath76-@xmath77 , compared to the typical mass density of 0.1 - 1@xmath78 in the field , and it is this crucial factor that leads to the high incidence rate of lmxbs in gcs .
the ucds and escs are also dense stellar systems , but are less so than the gcs .
this is demonstrated by the dashed lines in figure [ fig : reffmv ] , which mark equal values of the effective luminosity density , defined as @xmath79/@xmath80 .
approximately , one may take the effective luminosity density as a proxy of the stellar mass density .
most gcs show @xmath81 , incidentally a threshold above which no ucd / esc exists .
notably , all but three xgcs have @xmath81 , consistent with the scenario of lmxbs having formed from dynamical interaction in gcs . on the other hand ,
the xeos can now be divided into two groups : those with @xmath82 and those with @xmath83 .
members of the latter group all lie at @xmath84 mag , i.e. , they are practically ucds .
in particular , m59-ucd3 and m60-ucd1 belong to this group , and indeed they have the highest @xmath85 of all ucds , implying that the x - ray counterparts of these two ultra - massive ucds are also lmxbs .
we find that the incidence rate of the xeos with @xmath83 is @xmath86% , coming much closer to that of the xgcs .
the likely reason is that the square dependence of stellar encounter rate on @xmath85 is partially compensated by the larger size of the ucds ( a cubic dependence on @xmath87 ) . in the same regard ,
the presence of xeos with @xmath82 is rather surprising , because the stellar encounter rate in these objects would be a factor of @xmath88 further lower .
recall that some of the 6 xeos found in ngc4365 , typically with @xmath82 , might be interlopers rather than true associations ( section 3.1 ) . to investigate this issue
further , we repeat the above stacking analysis ( section 3.3 ) for two subgroups of the individually undetected eos , one with @xmath82 and the other with @xmath89 .
both subgroups show statistically significant signals , with average count rates of @xmath90 and @xmath91 @xmath92 per eo , and equivalent 0.5 - 8 kev luminosities of @xmath93 and @xmath94 , respectively . that the less dense subgroup has a lower average luminosity matches our anticipation , and also suggests that even the eos of the lower stellar densities can have a sizable population of dynamically - formed lmxbs .
indeed , if dynamical effects are irrelevant , as is the case in the field , the expected number of ( field ) lmxbs is @xmath95 , empirically derived from the lmxb populations in the milky way and nearby galaxies ( @xcite ) .
this relation predicts a negligibly small number of @xmath96 xeos , if we sum up the v - band light from all eos and assume a v - band mass - to - light ratio of 3.3 , appropriate for a simple stellar population with age of 10 gyr and half - solar metallicity . in the above discussion we have neglected the role of stellar - mass black holes ( bhs ) , which are usually thought to be absent in gcs due to their early segregation and subsequent mutual scattering at the core , although growing evidence now suggest that stellar - mass bhs do exist in some gcs ( @xcite ; @xcite ) .
the eos , in particular the massive ucds which might be the remnant of stripped galaxies ( e.g. , @xcite ) , can harbor stellar - mass bhs .
we note that the x - ray luminosities of all the xeos are compatible with ns binaries ( section 3.2 ) and do not seemingly require the presence of bh binaries .
potentially also relevant is the stellar velocity dispersion .
compared to gcs , the larger velocity dispersion of eos not only affects their stellar encounter rate , but also helps retain some of the otherwise escaping nss and bhs .
a more quantitative treatment of all these affects is premature at this stage .
the presence of lmxbs in eos implies that the underlying stellar population is an evolved one .
the dynamical formation timescale of an ns binary via tidal capture , following hut & verbunt ( 1983 ) , is @xmath97^{-1}{\rm gyr}$ ] , where @xmath98 and @xmath99 are the characteristic stellar mass and radius , and @xmath100 is the mass of ns .
this can be regarded as an independent evidence of eos being predominantly old stellar systems , consistent with existing optical spectroscopic studies .
one caveat of this conclusion is that our _ chandra _ sample is biased against eos found in disk , typically late - type galaxies ( section 3.1 ) .
such eos as possible descendants of recently aggregated massive star clusters ( e.g. , @xcite ) might not have sufficient time to form lmxbs , although they are also unlikely to harbor high - mass x - ray binaries unless with very recent star formation .
this latter case can be tested using optical observations . while we have shown that the observed properties of the xeos can be reasonably understood as them being lmxbs , the alternative possibility that some of the xeos trace the x - ray emission from an embedded massive black hole ( mbh ) should not be easily dismissed .
m60-ucd1 has been shown to harbor a mbh of @xmath10 ( seth et al .
2014 ) , and we find it to be an x - ray source with @xmath101 ( table 2 ) . while the current x - ray data does not unambiguously relate the detected x - ray emission to the mbh ( see also pandya et al .
2016 ) , we can infer an upper limit for its eddington ratio , @xmath102 , assuming that the x - ray band typically accounts for 10% of the mbh s bolometric luminosity .
likewise , the ultra - massive m59-ucd3 is detected with a rather moderate @xmath103 , but has an atypical soft x - ray spectrum , whose nature remains to be understood with enhanced s / n .
finally , we note that another ultra - massive ucd , m59co ( chilingarian & mamon 2008 ) , is undetected in the same _ chandra _ data of m59-ucd3 .
future high - sensitivity x - ray and optical observations should continue to provide important clues to the existence of mbhs as well as dynamical structure in the ucds and escs .
we have presented a systematic study of x - ray emission from ucds and escs based on archival chandra observations .
the main results in this paper are as follows : * a total of 17 x - ray counterparts are identified with 0.5 - 8 kev luminosities above @xmath0@xmath1 , which are distributed in eight early - type host galaxies . in the meantime , 360 x - ray counterparts of gcs are identified in four of the eight host galaxies . the incidence rate of x - ray sources in ucds / escs and gcs are @xmath2% and @xmath34% respectively .
* the spectral and temporal properties of the x - ray - detected ucds / escs are broadly similar to the x - ray - detected gcs , and are typical of lmxbs . *
a stacking analysis further shows that there is on - average substantial x - ray emission from the individually non - detected ucds and escs , which is again comparable to that from the individually undetected gcs . *
the x - ray properties of ucds / escs strongly suggest that they harbor a sizable population of lmxbs that have been formed from stellar dynamical interactions , consistent with the stellar populations in these dense systems being predominantly old .
* for the most massive ucds , there remains the possibility that a central massive black hole produces the detected x - ray emission .
future high - resolution , high - sensitivity x - ray and optical observations of a carefully selected sample of ucds and escs hold promise to solving their internal structure and dynamics . this work is supported by the national natural science foundation of china under grant 11133001 .
the authors wish to thank eric peng and yanmei chen for helpful comments .
m.h . is grateful to the hospitality of kiaa / pku during her summer visit .
z.l . acknowledges support from the recruitment program of global youth experts .
cccccccccc ngc247 & 11.785625 & -20.760389 & 3.6 & 10.0 ( 2 ) & 5.8@xmath104 & 2 + ngc891 & 35.639224 & 42.349146 & 10.0 & 171.6 ( 3 ) & 5.2@xmath104 & 6 + ngc1023 & 40.1000421 & 39.0632850 & 11.0 & 200.9 ( 5 ) & 5.5@xmath104 & 60 & 1 & + ngc1316 & 50.673750 & -37.208056 & 19.9 & 20.0 ( 1 ) & 6.3@xmath105 & 45 + ngc1380 & 54.113750 & -34.976028 & 18.3 & 41.6 ( 1 ) & 4.5@xmath105 & 13 + ngc1399 & 54.620941 & -35.450657 & 18.2 & 489.7 ( 14 ) & 3.7@xmath105 & 14 & 1 & 401 & 101 + m81 & 148.888221 & 69.065295 & 3.7 & 383.5 ( 25 ) & 5.1@xmath106 & 44 + ngc3115 & 151.308250 & -7.718583 & 9.8 & 1138.6 ( 11 ) & 1.4@xmath104 & 5 + ngc3311 & 159.175000 & -27.527500 & 53.7 & 31.9 ( 1 ) & 7.7@xmath107 & 19 + ngc3923 & 177.757059 & -28.806017 & 21.0 & 102.1 ( 2 ) & 3.0@xmath105 & 3 + ngc4278 & 185.028434 & 29.2807561 & 16.1 & 580.1 ( 9 ) & 5.8@xmath104 & 1 + ngc4365 & 186.117852 & 7.317673 & 21.4 & 195.8 ( 6 ) & 2.0@xmath105 & 216 & 6 & 3922 & 184 + m84 & 186.265597 & 12.886983 & 16.7 & 117.2 ( 4 ) & 2.0@xmath105 & 1 + ngc4382 & 186.350451 & 18.191487 & 15.2 & 49.9 ( 3 ) & 2.8@xmath105 & 4 + ngc4406 & 186.548928 & 12.946222 & 16.1 & 39.8 ( 3 ) & 9.5@xmath105 & 2 + ngc4449 & 187.046261 & 44.093630 & 3.8 & 100.9 ( 3 ) & 9.7@xmath106 & 7 + ngc4472 & 187.444841 & 8.000476 & 15.8 & 462.0 ( 10 ) & 1.4@xmath105 & 1 + m89 & 188.915864 & 12.5563414 & 16.0 & 201.4 ( 4 ) & 1.1@xmath105 & 2 & 1 & & + m104 & 189.997633 & -11.623054 & 11.1 & 194.2 ( 4 ) & 9.0@xmath104 & 10 & 2 & 659 & 65 + ic3652 & 190.243750 & 11.184556 & 15.2 & 5.1 ( 1 ) & 1.8@xmath107 & 1 + m59 & 190.509348 & 11.647027 & 15.5 & 30.1 ( 2 ) & 3.8@xmath105 & 1 & 1 & + m60 & 190.916564 & 11.552706 & 16.6 & 307.9 ( 6 ) & 1.3@xmath105 & 1 & 1 & + ngc4660 & 191.132917 & 11.190306 & 16.4 & 5.1 ( 1 ) & 2.1@xmath107 & 1 + ngc4696 & 192.205208 & -41.310833 & 37.6 & 779.3 ( 15 ) & 1.2@xmath107 & 2 + ngc5128 & 201.365063 & -43.019113 & 3.8 & 843.2 ( 24 ) & 6.6@xmath106 & 16 & 4 & 158 & 10 + m51 & 202.484200 & 47.230600 & 8.0 & 856.6 ( 14 ) & 1.3@xmath104 & 21 + ngc5846 & 226.622017 & 1.605625 & 26.9 & 149.9 ( 3 ) & 6.7@xmath105 & 13 + cccccccccccc ngc1023-eo1 & 40.1191 & 39.0608 & 40.1192 & 39.0610 & @xmath108 & @xmath109 & 2.1 & @xmath110 & 10.0 & -8.92 & 2.0 + ngc1399-eo12 & 54.82383 & -35.42506 & 54.82373 & -35.42503 & @xmath111 & @xmath112 & 33.0 & @xmath113 & 10.0 & -11.10 & 0.40 + ngc4365-eo019 & 186.06212 & 07.32031 & 186.06209 & 07.32052&@xmath114 & @xmath115 & 1.4 & @xmath116 & 17.4 & -8.23 & 0.3 + ngc4365-eo039 & 186.10850 & 07.36425 & 186.10835 & 07.36376&@xmath117 & @xmath118 & - & @xmath119 & 15.2 & -7.65 & 0.15 + ngc4365-eo006 & 186.11362 & 07.30964 & 186.11357 & 07.30957 & @xmath120 & @xmath121 & 10.9 & @xmath122 & 20.7 & -9.73 & 0.15 + ngc4365-eo117 & 186.11893 & 07.30381 & 186.11858 & 07.30391 & @xmath123 & @xmath124 & 8.1 & @xmath125 & 11.1 & -6.64 & 0.15 + ngc4365-eo206 & 186.15944 & 07.35225 & 186.15929 & 07.35184 & @xmath126 & @xmath127 & 10.9 & @xmath128 & 17.4&-5.53 & 0.3 + ngc4365-eo015 & 186.16258 & 07.28169 & 186.16245 & 07.28119 & @xmath129 & @xmath130 & 1.0 & @xmath131 & 21.2 & -8.43 & 0.02 + m89-eo1 & 188.90877 & 12.55032 & 188.90878 & 12.55081 & @xmath132 & @xmath133 & 1.4 & @xmath134 & 26.6 & -11.02 & 0.2 + m104-eo08 & 189.99429 & -11.63936 & 189.99474 & -11.63904 & @xmath135 & @xmath136 & 5.0 & @xmath137 & 15.6 & -6.15 & 0.1 + m104-eo01 & 190.01304 & -11.66786 & 190.01315 & -11.66781 & @xmath138 & @xmath139 & 9.3 & @xmath140 & 14.7 & -12.30 & 0.83 + m59-ucd3 & 190.54605 & 11.64479 & 190.54615 & 11.644546 & @xmath141 & @xmath142 & 1.0 & @xmath143 & 20 & -14.6 & 0.98 + m60-ucd1 & 190.89987 & 11.53464 & 190.89977 & 11.534675 & @xmath144 & @xmath145 & 5.6 & @xmath146 & 24 & -14.20 & 0.95 + ngc5128-eo03 & 201.24246 & -42.93619 & 201.24250 & -42.93621 & @xmath147 & @xmath148 & 1.8 & @xmath149 & 10.7 & -10.23 & 0.39 + ngc5128-eo02 & 201.27383 & -43.17519 & 201.27388 & -43.17508 & @xmath150 & @xmath151 & 4.6 & @xmath152 & 10.6 & -10.31 & - + ngc5128-eo05 & 201.37621 & -42.99300 & 201.37629 & -42.99297 & @xmath153 & @xmath154 & 9.3 & @xmath155 & 11.9 & -9.87 & 0.013 + ngc5128-eo01 & 201.38167 & -43.00078 & 201.38176 & -43.00079 & @xmath156 & @xmath157 & 49.1 & @xmath158 & 13.5 & -11.17 & 0.1 + bekki , k. , couch , w. j. , & drinkwater , m. j. , 2001 , , 522 , l105 blakeslee , j. p. , cho , h. , peng , e. w. , ferrarese , l. , et al .
2012 , , 746 , 88 blom , c. , spitler , l. r. , & forbes , d. a. , 2012 , , 420 , 37 brodie , j. p. , & larsen , s. s. , 2002 , , 124 , 1410 brns , r. c. , kroupa , p. , fellhauer , m. , metz , m. , & assmann , p. , 2011 ,
a&a , 529 , 138 brns , r. c. , & kroupa , p. , 2012
, , 547 , a65 ( bk12 ) caso , j. p. , bassino , l. p. , richtler , t. , smith castelli , a. v. , & faifer , f. r. , 2013 , , 430 , 1088 chattopadhyay , a. k. , chattopadhyay , t. , davoust , e. , et al .
2009 , , 705 , 1533 chiboucas , k. , tully , r. b. , marzke , r. o. , et al .
2011 , , 737 , 86 chilingarian , i. v. , & mamon , g. a. , 2008 , mnras , 385 , l83 clark , g. w. , 1975 , , 199 , l143 da rocha , c. , mieske , s. , georgiev , i. y. , hilker , m. , et al .
2011 , , 525 , 86 drinkwater , m. j. , jones , j. b. , gregg , m. d. , & phillipps , s. , 2000 , , 17 , 227 drinkwater , m. j. , gregg , m. d. , hilker , m. , et al .
2003 , , 423 , 519 drinkwater , m. j. , gregg , m. d. , couch , w. j. , et al .
2004 , , 21 , 375 fabbiano , g. , 2006 , , 44 , 323 fabian , a. c. , 1975 , , 173 , 161 fellhauer , m. , & kroupa , p. , 2002
, , 330 , 642 forbes , d.a . , pota , v. , usher , c. , et al .
2013 , mnras , 435 , l6 forbes , d. a. , almeida , a. , spitler , l. r. , & pota , v. , 2014 , , 442 , 1049 gilfanov , m. , 2004 , , 349 , 146 gregg , m. d. , drinkwater , m. j. , evstigneeva , e. , et al .
2009 , , 137 , 498 harris , g. l. h. , gmez , m. , harris , w. e. , johnston , k. , et al .
2012 , , 143 , 84 hasegan , m. , jordan , a. , cote , p. , et al .
2005 , , 627 , 203 hau , g. k. t. , spitler , l. r. , forbes , d. a. , et al .
2009 , , 394 , l97 hilker , m. , infante , l. , vieira , g. , kissler - p .
m. , & richtler , t. , 1999 , , 134 , 75 hills , j. g. , 1975 , , 175 , 1 janz , j. , norris , m.a . , forbes , d.a .
, et al . 2016 ,
mnras , 456 , 617 kalberla , p.m. , burton , w.b . , hartmann , d. , arnal , e.m .
, bajaja , e. , morras r. , pppel , w.g.l .
2005 , a&a , 440 , 775 katz , j. i. , 1975 , , 253 , 698 kim , m. , kim , d .- w . ,
wilkes , b. j. , et al .
, 169 , 401 li , z. , spitler , l. r. , jones , c. , et al .
, 721 , 1368 liu , c .- z . , peng , e. w. , toloba , e. , et al .
2015 , , 812 , l2 liu , c .- z . , peng , e. w. , ct , p. , et al .
2015 , , 812 , 34 maccarone , t. , kundu , a. , zepf , s. e. , rhode , k. l. 2007 , nature , 445 , 183 madrid , j. p. , graham , a. w. , harris , w. e. , et al .
2010 , , 722 , 1707 madrid , j. p. , & donzelli , c. j. , 2013 , , 770 , 158 mieske , s. , hilker , m. , & infante , l. , 2002 , , 383 , 823 mieske , s. , hilker , m. , jordn , a. , infante , l. , & kissler - p . , m. , 2007 , , 472 , 111 mieske , s. , hilker , m. , jordn , a. , infante , l. , et al .
2008 , , 487 , 921 mieske , s. , hilker , m. , & misgeld , i. , 2012 , , 537 , a3 misgeld , i. , mieske , s. , hilker , m. , richtler , t. , et al .
2011 , , 531 , a4 norris , m. a. , & kannappan , s. j. , 2011 , , 414 , 739 norris , m. a. , escudero , c. g. , faifer , f. r. , et al .
2015 , , 451 , 3615 pandya , v. , mulchaey , j. , greene , j.e .
2016 , arxiv:1601.01690 park , t .- y . ; kashyap , v. l. ; siemiginowska , a. , et al . 2006 , , 652 , 601 paudel , s. , lisker , t. , & janz , j. , 2010 , , 724 , l64 penny , s. j. , forbes , d. a. , strader , j. , et al .
2014 , , 439 , 3808 pfeffer , j. , & baumgardt , h. , 2013 , , 433 , 1997 phillipps , s. , drinkwater , m. j. , gregg , m. d. , & jones , j. b. , 2001 , , 560 , 201 sandoval , m. a. ; vo , r. p. ; romanowsky , a. j. , et al .
2015 , , 808l , 32 seth , a. c. , van den bosch , r. , mieske , s. , baumgardt , h. , brok , m. d. , et al .
2014 , , 513 , 398 spitler , l. r. , larsen , s. s. , strader , j. , brodie , j. p. , et al .
2006 , , 132 , 1593 strader , j. , chomiuk , l. , maccarone , t. j. , et al .
2012 , , 490 , 71 strader , j. , seth , a. c. , forbes , d. a. , et al .
2013 , , 775 , l6 sutantyo , w. , 1975 , , 44 , 227 wang , q.d . , 2004 , , 612 , 159 zhang , h .- x . , peng , e. w. , ct , p. , liu , c .- z . , et al .
2015 , , 802 , 30 | we have conducted a systematic study of x - ray emission from ultra - compact dwarf ( ucd ) galaxies and extended star clusters ( escs ) , based on archival _ chandra _ observations . among a sample of 511 ucds and escs complied from the literature ,
17 x - ray counterparts with 0.5 - 8 kev luminosities above @xmath0@xmath1 are identified , which are distributed in eight early - type host galaxies . to facilitate comparison
, we also identify x - ray counterparts of 360 globular clusters ( gcs ) distributed in four of the eight galaxies .
the x - ray properties of the ucds and escs are found to be broadly similar to those of the gcs .
the incidence rate of x - ray - detected ucds and escs , @xmath2% , while lower than that of the x - ray - detected gcs [ ( @xmath3% ] , is substantially higher than expected from the field populations of external galaxies . a stacking analysis of the individually undetected ucds / escs further reveals significant x - ray signals , which corresponds to an equivalent 0.5 - 8 kev luminosity of @xmath0@xmath4 per source . taken together ,
these provide strong evidence that the x - ray emission from ucds and escs is dominated by low - mass x - ray binaries having formed from stellar dynamical interactions , consistent with the stellar populations in these dense systems being predominantly old . for the most massive ucds
, there remains the possibility that a putative central massive black hole gives rise to the observed x - ray emission . |
in formal language theory , quotient is a basic and very important operation and plays a fundamental role in the construction of minimal deterministic finite automata ( dfa ) . given a formal language @xmath3 over an alphabet @xmath4 , the left quotient @xmath5 of @xmath3 by a word @xmath6 is defined as the language @xmath7 , where @xmath8 is the free monoid of words over @xmath4 .
the famous myhill - nerode theorem then states that @xmath3 is a regular language if and only if the number of different left quotients of @xmath3 ( also called the quotient complexity @xcite of @xmath3 ) is finite .
moreover , a minimal dfa which recognizes @xmath3 can be constructed in a natural way by using left quotients as states .
in particular , this means that the quotient complexity of @xmath3 is equal to the size of the minimal dfa which recognizes @xmath3 . the notion of left quotient of a formal language by a word can be extended to quotients by a formal language in two ways . given two formal languages @xmath9 , the left quotient of @xmath3 by @xmath10 , denoted by @xmath11 , is defined as the union of @xmath5 for all words @xmath6 in @xmath10 .
another extension is less well - known , if not undefined at all .
we define the _ left residual _ of @xmath3 by @xmath10 , denoted by @xmath12 , as the intersection of @xmath5 of all words in @xmath10 .
similarly we have @xmath13 , the right quotient of @xmath3 by @xmath10 , and @xmath14 , the right residual of @xmath3 by @xmath10 . regarding each left residual of @xmath3 as a state
, there is a natural way to define an automaton , which is called the
_ universal automaton _
@xcite of @xmath3
. the universal automaton of a formal language @xmath3 contains many interesting information ( e.g. factoraization ) of @xmath3 @xcite and plays a very important role in constructing the minimal nondeterministic finite automaton ( nfa ) of @xmath3 @xcite .
former power series are extensions of formal languages , which are used to describe the behaviour of weighted automata ( i.e. finite automata with weights ) .
weighted automata were introduced in 1961 by schtzenberger in his seminal paper @xcite .
a formal power series is a mapping from @xmath8 , the free monoid of words over @xmath4 , into a semiring @xmath15 .
depending on the choice of the semiring @xmath15 , formal power series can be viewed as weighted , multivalued or quantified languages where each word is assigned a weight , a number , or some quantity .
weighted automata have been used to describe quantitative properties in areas such as probabilistic systems , digital image compression , natural language processing .
we refer to @xcite for an detailed introduction of weighted automata and their applications . despite that
a very large amount of work has been devoted to the study of formal power series and weighted automata ( see e.g. @xcite for surveys ) , the important concept of quotient as well as universal automata has not been systematically investigated in this weighted context . the only exception seems to be @xcite , where the quotient of formal power series ( by word ) was discussed in pages 10 - 11 .
when the semiring is complete , it is straightforward to extend the definition of the quotient of a formal power series @xmath0 from words to series : we only need to take the weighted sum of all left quotients of @xmath0 by word in @xmath4 .
our attempt to characterize the residual of a formal power series @xmath0 by a formal power series as the weighted intersection of all left quotients of @xmath0 by word in @xmath4 is , however , unsuccessful .
several important and nice properties fail to hold anymore .
the aim of this paper is to introduce the quotient and residual operations in formal power series and study their application in the minimization of weighted automata . to overcome the above obstacle with residuals
, we require the semiring to be a complete c - semiring ( to be defined in section 2 ) , and then give a characterization of residuals in terms of quotients by word .
many nice properties and useful notions then follow in a natural way .
the remainder of this paper is organized as follows .
section 2 introduces basic notions and properties of semirings , formal power series , and weighted automata .
quotients of formal power series are introduced in section 3 , where we also show how to construct the minimal deterministic weighted automata effectively . in section 4
, we introduce the residuals and factorizations of formal power series . using the left residuals
, we define the universal weighted automaton @xmath16 for arbitrary formal power series @xmath0 in section 5 , and justify its universality in section 6 . an effective method for constructing the universal automaton
is described in section 7 , which is followed by a comparison of the quotient and the residual operations .
the last section concludes the paper .
we recall in this section the notions of semirings , formal power series , weighted automata , and weighted contex - free grammar .
interested readers are referred to @xcite for more information .
a 5-tuple @xmath17 is called a _ semiring _
if @xmath15 is a set containing at least two different elements @xmath18 and @xmath19 , and @xmath20 and @xmath21 are two binary operations on @xmath15 such that * @xmath20 is associative and is commutative and has identity @xmath18 ; * @xmath21 is associative and has identity @xmath19 and null element @xmath18 ( i.e. , @xmath22 for all @xmath23 ) ; and * @xmath21 distributes over @xmath20 , i.e. , for all @xmath24 , @xmath25 and @xmath26 . intuitively , a semiring is a ring ( with unity ) without subtraction . all rings ( with unity ) , as well as all fields ,
are semirings , e.g. , the integers @xmath27 , rationals @xmath28 , reals @xmath29 , complex numbers @xmath30 .
lattices provide another important type of semirings .
recall that a partially ordered set @xmath31 is a _ lattice _ if for any two elements @xmath32 , the least upper bound @xmath33 and the greatest lower bound @xmath34 exist in @xmath35 .
a lattice @xmath31 is _ distributive _ , if @xmath36 for all @xmath37 ; and _ bounded _
, if @xmath3 contains a smallest element , denoted 0 , and a greatest element , denoted 1 .
let @xmath31 be any bounded distributive lattice .
then @xmath38 is a semiring .
because a distributive lattice @xmath3 also satisfies the dual distributive law @xmath39 for all @xmath37 , the structure @xmath40 is also a semiring .
other important examples of semirings include : * the boolean semiring @xmath41 ; * the semiring of the natural numbers @xmath42 with the usual addition and multiplication ; * the tropic semiring @xmath43 with min and @xmath44 extended to @xmath45 in a natural way ; * the min - sum semiring of nonnegative reals @xmath46 ; * the semiring of ( completely positive ) super - operators on a hilbert space @xmath47 @xmath48 .
we note in the last semiring the addition is not idempotent and the product is not commutative .
this semiring is recently used in model checking quantum markov chains @xcite and the study of finite automata with weights taken from this semiring just initiated .
let @xmath49 be an index set and let @xmath15 be a semiring .
an infinitary sum operation @xmath50 is an operation that associates with every family @xmath51 of elements of @xmath15 an element @xmath52 of @xmath15 .
a semiring @xmath15 is called _ complete _ if it has an infinitary sum operation @xmath53 for each index set @xmath49 and the following conditions are satisfied : * @xmath54 , @xmath55 , and @xmath56 for @xmath57 . * @xmath58 if @xmath59 and @xmath60 for @xmath57 . *
@xmath61 and @xmath62 .
this means that a semiring @xmath15 is complete if it is possible to define infinite sums ( i ) that are extensions of the finite sums , ( ii ) that are associative and commutative , and ( iii ) that satisfy the distributive laws .
a semiring @xmath15 is a _ c - semiring _
if @xmath20 is idempotent ( i.e. , @xmath63 for all @xmath23 ) , @xmath64 is commutative , and @xmath19 is the absorbing element of @xmath20 ( i.e. , @xmath65 for any @xmath23 ) . in general , for a semiring @xmath15 , we define a preorder @xmath66 over the set @xmath15 by @xmath67 if @xmath20 is idempotent , then @xmath66 is also a partial order .
suppose @xmath15 is a c - semiring .
for any @xmath68 , we have @xmath69 and @xmath70 ( the least upper bound of @xmath71 and @xmath72 ) in the poset @xmath73 .
if @xmath15 is clear from the context , then @xmath15 is omitted and we simply write @xmath74 for @xmath66 in the following . a semiring @xmath15 is called a _ complete c - semiring _ if @xmath15 is a complete semiring and a c - semring . in a complete c - semiring ,
the infinitary sum @xmath52 is exactly the least upper bound of @xmath75 ( @xmath76 ) in @xmath15 under the induced partial order @xmath66 . in this case , @xmath52 is also written as @xmath77 .
complete c - semiring is a special kind of the notion of _ quantale
_ @xcite , which is a complete lattice @xmath3 equipped with a multiplication operator @xmath21 such that @xmath78 is a semigroup satisfying the following distributive laws : @xmath79 since the infinite distributive laws ( eq . [ eq : quantale - dist ] ) holds , there are two _ adjunctions _ or _ residuals _ , denoted @xmath80 ( left residual ) and @xmath81 ( right residual ) , respectively , satisfying the following adjunction ( residual ) conditions , @xmath82 when the given quantale is commutative , i.e. , the operation @xmath21 is commutative , then the left residual is the same as the right residual . in this case , we call the left residual ( and the right residual ) the residual , denoted by @xmath83 .
then we have @xmath84 we have the following proposition .
let @xmath15 be a complete c - semiring .
then @xmath15 is a commutative quantale with unit @xmath19 as the largest element of @xmath15 .
let @xmath4 be an alphabet and @xmath15 a semiring .
write @xmath85 for the set of all finite strings ( or words ) over @xmath4 , and write @xmath86 for the empty string
. then @xmath85 is the free monoid generated by @xmath4 under the operation of concatenation .
we write @xmath87 for all finite non - empty strings over @xmath4 .
a _ formal power series _
@xmath0 is a mapping from @xmath85 into @xmath15 . for simplicity
, we also call a formal power series as a _ series _ , or an _ @xmath15-subset _ of @xmath85 .
the value of @xmath0 at a word @xmath88 is denoted @xmath89 or @xmath90 in this paper .
we write @xmath0 as a formal sum @xmath91 where the values @xmath89 are referred as the _ coefficients _ of @xmath0 .
the collection of all power series @xmath0 as defined above is denoted by @xmath92 . for a series @xmath0 , if @xmath93
, then @xmath0 is called _
proper_. for any series @xmath0 , @xmath0 can be written as the sum of a proper series and and non - proper series , i.e. , @xmath94 for s series @xmath0 on @xmath4 , the _ support _ of @xmath0 is defined as @xmath95 for two series @xmath0 and @xmath96 in @xmath92 , we define @xmath97 whenever @xmath98 for any @xmath88 .
weighted automata are an extension of the classical finite automata .
[ def : l - vfa ] let @xmath15 be a semiring . a _ weighted automaton
_ with weights in @xmath15 is a 5-tuple @xmath99 , where @xmath100 denotes a set of states , @xmath4 is an input alphabet , @xmath101 is a mapping from @xmath102 to @xmath15 , and @xmath49 and @xmath103 are two mappings from @xmath100 to @xmath15 .
we call @xmath104 a _ finite _ weighted automaton if both @xmath100 and @xmath4 are finite sets ; we call @xmath104 a _ deterministic weighted automaton _ ( dwa for short ) if @xmath101 is crisp and deterministic , i.e. , @xmath101 is a mapping from @xmath105 into @xmath100 , and @xmath106 .
the mapping @xmath101 is called the _ weighted ( state ) transition relation_. intuitively , for any @xmath107 and @xmath108 , @xmath109 stands for the weight that input @xmath110 causes state @xmath111 to become @xmath112 .
@xmath49 and @xmath103 represent the ( weighted ) initial and , respectively , final state . for each @xmath113
, @xmath114 indicates the weight that @xmath112 is an initial state , @xmath115 expresses the weight that @xmath112 is a final state .
our definition of deterministic weighted automaton is different from the one used in e.g. @xcite , where a weighted automaton @xmath99 is _ deterministic _ or _ sequential _ if there exists a unique state @xmath116 in @xmath100 such that @xmath117 and for all @xmath113 and all @xmath118 , there is at most one @xmath119 such that @xmath120
. these two definitions are not identical in general .
in fact , deterministic weighted automaton called in this paper is just the _ simple _ deterministic weighted automaton defined in @xcite , for the detail comparison , we refer to @xcite .
a simple deterministic weighted automaton is obviously a sequential weighted automaton defined in @xcite .
the following example shows that the converse does not hold in general .
let @xmath121 be the tropical semiring .
suppose @xmath122 and @xmath0 is the formal power series over @xmath4 defined by @xmath123 , where @xmath124 denotes the length of the string @xmath125
. then @xmath0 can not be accepted by any simple deterministic weighted automaton ( see proposition [ pro : dffa ] below ) .
however , @xmath0 can be accepted by a sequential weighted automaton @xmath126 , where @xmath127 if @xmath128 and @xmath129 otherwise .
if @xmath15 is _ locally finite _
, however , then the two definitions are equivalent in the sense that they accept the same class of former power series , where a semiring is locally finite if every sub - semiring generated by a finite set is also finite @xcite .
two weighted automata can be compared by a morphism .
[ de : morphism ] a _ homomorphism _ ( or _ morphism _ ) between two weighted automata @xmath99 and @xmath130 is a mapping @xmath131 , satisfying the following conditions : @xmath132 for any @xmath107 and @xmath118 .
a morphism @xmath133 is _ surjective _ if @xmath131 is onto and @xmath134 , @xmath135 , where @xmath136 in this case , @xmath137 is also called the _ morphic image _ of @xmath104 .
if @xmath138 is one - to - one , then we call @xmath104 a _ sub - automaton _ of @xmath137 .
a morphism @xmath133 is called a _ strong homomorphism _ if @xmath139 in case @xmath133 is an onto strong homomorphism , we call @xmath137 a _ quotient _ of @xmath104 .
we say @xmath133 is an _ isomorphism _ if it is bijective and its inverse @xmath140 is also a morphism . in this case , we say @xmath104 is isomorphic to @xmath137 .
the behaviour of a weighted automaton is characterized by the formal power series it recognizes . to introduce this formal power series ,
we extend the weighted transition function @xmath141 to a mapping @xmath142 as follows : * for all @xmath119 , set @xmath143 if @xmath128 , and @xmath144 otherwise ; * for all @xmath145 , define @xmath146 if @xmath104 is deterministic , the extension @xmath147 of transition function @xmath101 is defined similar as in the classical case .
it is easy to see that for any @xmath148 we have @xmath149.\ ] ] [ dfn : recognized - fps ] for a weighted automaton @xmath99 , the formal power series _ recognized _ or _ accepted _ by @xmath104 , written@xmath150 ,
is defined as follows : latexmath:[\[\label{eq : recognized - fps } if @xmath104 is deterministic , then the formal power series recognized by @xmath152 is defined as @xmath153 is a _ regular series _ or an _ @xmath15-regular language _ on @xmath4 if it is recognized by a finite weighted automaton ; and say @xmath0 is a _ dwa - regular series _ or a _ dwa - regular language _ on @xmath4 if it is recognized by a finite dwa .
it was proved by schtzenberger @xcite that regular series are precisely the rational formal power series for all semirings .
so we also say a regular series as a rational series in this paper .
we next recall the concept of weighted context - free grammar and weighted context - free series .
a _ weighted context - free grammar _ is in essence a classical context - free grammar together with a function mapping rules of the grammar to weights in a certain semiring ( @xcite ) .
let @xmath15 be a semiring .
a weighted context - free grammar ( wcfg ) is defined as a tuple @xmath154 , where @xmath4 ( the set of terminal symbols ) and @xmath155 ( the set of non - terminal symbols ) are two finite sets that are disjoint , @xmath156 ( the start or initial symbol ) is an element in @xmath155 , @xmath157 is a mapping from @xmath158 ( the set of productions or rules ) to @xmath15 .
similar to the classical case , we can define the induction of a weighted context - free grammar @xmath159 .
suppose @xmath160 is a weighted production , and @xmath161 are elements in @xmath162 .
we say @xmath163 is a _
direct induction _ of @xmath164 with weight @xmath165 , denoted by @xmath166 . for productions
@xmath167 in @xmath162 , if @xmath168 , then we say @xmath169 is an induction of @xmath170 with weight @xmath171 , denoted by @xmath172 . the formal power series @xmath173 generated by @xmath159 is defined as @xmath174 ( @xmath88 ) . a series @xmath0 is called _ context - free _ if there is a weighted context - free grammar @xmath159 such that @xmath175 . we recall several well - know operations of formal power series ( cf .
@xcite ) . for two formal power series @xmath176 , a value @xmath177 , and @xmath178
, we define @xmath179 where @xmath180 if @xmath181 .
we call @xmath182 and @xmath183 the _ sum _ and , respectively , the _ concatenation _ ( or _ cauchy product _ ) of @xmath0 and @xmath96 , and call @xmath184 , @xmath185 , @xmath186 , and @xmath187 the _ left scalar product _ ,
right scalar product _ , the _ kleene closure _ , and the _ reversal _ of @xmath0 , respectively . given a weighted automaton , @xmath188
, three other operations can be defined for the formal power series recognized by @xmath152 .
for any two states @xmath189 in @xmath100 , and any @xmath190 , we define @xmath191 the following result holds .
[ pro : factor - automata ] suppose @xmath188 is a weighted automaton . for any @xmath113 ,
we have @xmath192 for any @xmath193 , we have @xmath194 hence , @xmath195 .
in this section , we first introduce quotients of formal power series and then , upon this operation , introduce for each formal power series @xmath0 a canonical weighted automaton that recognizes @xmath0 .
properties of the quotient operation is also studied .
let @xmath196 be a formal power series , @xmath197 a word .
the _ left quotient _ of @xmath0 by @xmath6 , written @xmath198 , is the formal power series @xmath199 defined as : @xmath200 dually , the _
right quotient _ of @xmath0 by @xmath6 , written @xmath201 , is the formal power series @xmath202 defined as : @xmath203 for any @xmath204 .
the left quotient operation introduces an equivalent relation on @xmath85 .
[ dfn : non - distinguishable - relation ] suppose @xmath205 .
we say two words @xmath206 and @xmath207 are _ non - distinguishable _ in @xmath0 , written @xmath208 , if @xmath209 , i.e. , if @xmath210 for all @xmath197 .
it is straightforward to show that @xmath211 is an equivalent relation on @xmath85 . for each word @xmath197 ,
we write @xmath212_a$ ] for the equivalent class of @xmath211 that contains @xmath6 .
[ le : mini dwa ] suppose @xmath213 .
the mapping defined by @xmath212\mapsto u^{-1}a$ ] @xmath214 is a bijection from the set of equivalent classes of @xmath211 to the set of left quotients of @xmath0 .
we define a deterministic weighted automaton @xmath215 as follows : * @xmath216 is the quotient of @xmath85 modulo @xmath211 ; * @xmath217 is defined as @xmath218,\sigma)=[\theta\sigma ] , \hspace*{8 mm } ( \sigma\in{\sigma^{\ast}})\ ] ] * @xmath219 is the singleton state @xmath220 $ ] in @xmath221 ; * @xmath222 is defined by @xmath223)=a(\theta)$ ] for @xmath193 .
[ pro : minimal dwa ] suppose @xmath224
. then @xmath1 is the minimal dwa that recognizes @xmath0 .
it is easy to see that @xmath225 is well - defined and , hence , @xmath215 is a dwa .
it is straightforward to show that @xmath1 recognizes @xmath0 . in other words
, there is a dwa @xmath104 that accepts @xmath0 for any formal power series @xmath0 . in general
, @xmath1 is not finite ; but , when it is finite , @xmath1 is the minimal dwa that recognizes @xmath0 . by lemma [ le : mini dwa ] ,
an equivalent minimal dwa @xmath226 that recognizes @xmath0 can be constructed by using the left quotients of @xmath0 as follows * @xmath227 , the set of all left quotients of @xmath0 by a word ; * @xmath228 defined by @xmath229 * @xmath230 ; * @xmath231 is defined by @xmath232 for @xmath193 .
the following proposition presents a characterization of formal power series that can be recognized by a finite dwa .
[ pro : dffa ] suppose @xmath205 .
given @xmath177 , we write @xmath233 , @xmath234 , and @xmath235}=\ { \theta\in{\sigma^{\ast}}| a(\theta)=r\}$ ]
. then the following statements are equivalent .
* @xmath0 can be recognized by a finite dwa .
* @xmath236 is finite and each @xmath237 is a regular language for @xmath238 .
* @xmath236 is finite and each @xmath235}$ ] is a regular language for @xmath238 .
* there exist @xmath239 and @xmath240 regular languages @xmath241 over @xmath4 , which are pairwise disjoint , such that @xmath242 .
* @xmath211 has finite index , i.e. , @xmath221 is finite .
the proof is similar to that given for lattices in @xcite .
but _ when can a regular series be recognized by a finite dwa _ ?
this is closely related with the structure of semiring @xmath15 .
it can be shown that ( cf.@xcite ) , for any regular series @xmath0 , @xmath0 can be recognized by a finite dwa iff the monoid @xmath243 and @xmath244 are both locally finite , where a monoid @xmath245 is locally finite if every submonoid generated by a finite subset of @xmath246 is also finite . in particular , if @xmath15 is finite , regular series and dwa - regular series are the same .
the left quotient of a series @xmath0 by a word @xmath6 can be regarded as a left action of @xmath85 on @xmath92 , i.e. , a mapping @xmath247 .
the action satisfies the following conditions .
[ pro : quotient - word ] let @xmath15 be a semiring .
suppose @xmath205 , @xmath86 is the empty word in @xmath85 , @xmath248 are words in @xmath85 , and @xmath240 is a value in @xmath15 .
then we have * @xmath249 , @xmath250 , @xmath251 ; * @xmath252 , @xmath253 ; * @xmath254 , @xmath255 , @xmath256 , @xmath257 @xmath258 .
straightforward .
when the semiring @xmath15 is complete , the left quotient operation can be extended from words to series in a natural way .
let @xmath259 be two formal power series in @xmath92 .
we define the _ left quotient _ of @xmath0 by @xmath10 , denoted by @xmath260 , as @xmath261 similarly , we can define the _
right quotient _ of @xmath0 by @xmath262 for any @xmath263 , denoted by @xmath264 as @xmath265 it is easy to see that when the formal power series @xmath10 is a word @xmath6 , then @xmath266 and @xmath267 .
we summarize some algebraic properties of the quotient operation . [
pro : quotient - fps ] let @xmath15 be a complete semiring .
suppose @xmath268 .
then * @xmath269 , @xmath270 ; * @xmath271 , @xmath272 ; * @xmath273 , @xmath274 ; * @xmath275 , @xmath276 ; * @xmath277 , @xmath278 , @xmath279 , @xmath280 ; * if @xmath15 is commutative , then @xmath281 , @xmath282 .
if the semiring @xmath15 is commutative , the following lemma shows that the right quotient is dual to the left quotient , where the reversal operation @xmath283 is defined ( see eq .
[ eq : a^r ] ) as @xmath284 for any @xmath181 .
[ lemma : right - quotient ] suppose @xmath15 is a complete semiring .
for @xmath285 , if @xmath286 , @xmath287 for any @xmath288 , then @xmath289 in particular , if @xmath0 is a classical language , i.e. , @xmath290 or if @xmath15 is commutative , the above equalities hold .
straightforward . by the above result ,
if the semiring @xmath15 is commutative , then the right quotient is dual to the left quotient by the reversal operation .
properties of the right quotient can be dually obtained in this case . in this subsection
, we study the language properties of the quotient of series .
we first recall a preliminary result which will be used in the proof .
@xcite][le : proper ] suppose @xmath262 is a proper and regular series in @xmath92 .
then there exists a weighted automaton @xmath291 that recognizes @xmath262 such that @xmath292 and @xmath293 for any @xmath113 .
[ pro : quotient - closeness ] let @xmath15 be a complete semiring .
suppose @xmath294 are former power series in @xmath92 . then * if @xmath0 is regular , then @xmath260 and @xmath264 are regular .
* if @xmath0 is dwa - regular , then @xmath260 and @xmath264 are dwa - regular .
* for a commutative semiring @xmath15 , if @xmath0 is context - free , @xmath10 and @xmath262 are regular , then @xmath260 and @xmath264 are context - free . * if @xmath0 is context - free , @xmath10 and @xmath262 are dwa - regular , then @xmath260 and @xmath264 are context - free
. see appendix a. using the quotient of a series @xmath0 by a series , we have a related deterministic weighted automaton @xmath295 where @xmath296 , and @xmath297 . because the accessible part of @xmath298 is the minimal dwa @xmath299 ( cf .
( [ eq : m - a ] ) ) , hence @xmath300 .
recall in the classical case , if @xmath0 is a regular language over @xmath4 , then @xmath301 ( the set of all left quotients of @xmath0 by languages ) is a finite set .
this does not hold in general .
in fact , the finiteness of @xmath302 heavily depends on the finiteness of the semiring @xmath15 .
[ pro : quotient automaton ] let @xmath15 be a semiring . then @xmath15 is finite if and only if @xmath298 is finite for any dwa - regular series @xmath0 in @xmath92 , where @xmath303 is the weighted automaton that recognizes @xmath0 defined in ( [ eq : b_a ] ) .
suppose @xmath298 is finite for any dwa - regular series @xmath0 .
in particular , @xmath298 is finite for @xmath304 .
take @xmath305 , for any @xmath177 , let @xmath306 ( which is a series over @xmath4 ) . by a simple calculation ,
the left quotient @xmath307 is @xmath308 , i.e. , @xmath309 , where @xmath310 for any @xmath88 .
then , as a subset of @xmath311 , the set @xmath312 is finite . hence , @xmath15 is a finite set .
conversely , suppose @xmath313 is finite .
for any @xmath314 , let @xmath315 . then @xmath316 . since @xmath0 is regular , there exist finite regular languages @xmath241 such that @xmath317 . by proposition [ pro : quotient - fps ]
, it follows that @xmath318 .
since @xmath319 is regular , the set @xmath320 is finite .
it follows that the set @xmath321 is also finite . in the remainder of this paper
, we will consider the residual operations on formal power series .
in this section , we study the residuals and factorizations of formal power series . for @xmath322 , the _ left residual _ of @xmath0 by @xmath10 , written as @xmath323 , is defined as the largest @xmath324 such that @xmath325 .
similarly , we define the _ right residual _ of @xmath0 by @xmath10 , written as @xmath326 , as the largest @xmath324 such that @xmath327 .
we note that @xmath323 and @xmath326 need not exist for all @xmath328 .
the following proposition shows that residuals always exist when @xmath15 is a complete c - semiring .
we recall that each c - semiring is commutative .
suppose @xmath15 is a complete c - semiring .
for any @xmath329 , @xmath323 and @xmath330 do exist and we have @xmath331 because @xmath15 is a complete c - semiring , we know @xmath332 exists for any subset @xmath333 of @xmath92 .
moreover , the concatenation operation satisfies the following conditions : @xmath334 that is to say , @xmath92 is a quantale under the operations @xmath335 and concatenation .
it is then easy to see that @xmath336 and hence @xmath337 is the largest @xmath338 such that @xmath339 , i.e. , @xmath340 .
similarly , we have @xmath341 .
the following proposition shows that left residual and right residual are dual to each other .
[ prop : dual - residuals ] let @xmath15 be a complete c - semiring .
. then we have @xmath342 and @xmath343 @xmath344 .
straightforward . by the above result , in the following
, we often consider only one ( either left or right ) residual .
dual results can be applied to the other residual .
left residual has close connection with left quotient .
[ pro : quotient - word ] let @xmath15 be a complete c - semiring .
suppose @xmath0 is a formal power series in @xmath92 , @xmath6 is a word in @xmath85 .
then @xmath198 is the largest language @xmath262 such that @xmath345 , i.e. , @xmath346 .
note that @xmath347 if @xmath348 and @xmath18 otherwise .
it follows that @xmath349 . if @xmath345 , then @xmath350 for any @xmath204 . hence , @xmath351 .
therefore @xmath198 is the largest @xmath262 such that @xmath345 . in the classical case , given two languages @xmath328 over @xmath4
, we always have @xmath352 in the weighted case , residuals are usually not the intersection of a set of quotients by word .
but we can still represent residuals in terms of quotients by word . before
give the details , we summarize some basic algebraic properties of the residuals . [ pro : residual quotient ] let @xmath15 be a complete c - semiring . for @xmath329 ,
we have * @xmath353 , @xmath354 .
* the operations @xmath355 defined by @xmath356 and @xmath357 are two closure operators , where @xmath358 is a closure operator if @xmath359 , @xmath360 and @xmath361 .
* @xmath362 , @xmath363 .
* @xmath364 , @xmath365 .
* @xmath366 . *
@xmath367 , @xmath368 .
we give the proof of ( vi ) , the others are straightforward . by ( ii ) , @xmath369 , then it follows that @xmath370 . on the other hand , since @xmath371 , it follows that @xmath372 .
hence , @xmath367 .
similarly , we have @xmath368 .
we next give the characterization of residuals in terms of quotients by word , where @xmath373 is the residual operation in the quantale @xmath15 defined ( cf .
( [ a - rw - b ] ) ) . [ pro : characterization of residual quotient ] let @xmath15 be a complete c - semiring .
suppose @xmath374 .
then , for any @xmath375 , we have @xmath376 by proposition [ prop : dual - residuals ] , we need only prove eq.([eq : x|a ] ) . for @xmath204 ,
let @xmath377 .
then @xmath378 is a series .
we show @xmath379 , i.e. , @xmath378 is the largest series @xmath262 such that @xmath325 .
first , @xmath380 second , if @xmath325 , then , for any @xmath381 , @xmath382 .
it follows that @xmath383 for any @xmath197 , thus , @xmath384 , i.e. , @xmath385 .
this shows that @xmath386 .
the following proposition shows that dwa - regular series are closed under residual operations .
[ pro : implication - language ] suppose @xmath15 is a complete c - semiring , and @xmath387 . if @xmath0 is dwa - regular , then so are @xmath323 and @xmath330 . by proposition [ prop : dual - residuals ] , we need only consider right residuals .
suppose that @xmath388 is a dwa accepting @xmath0 .
then we have @xmath389 for any @xmath390 .
define another weighted automaton , @xmath391 , where @xmath392 then @xmath393 hence , @xmath330 is dwa - regular . in formal language theory , factorization is an important notion that is closely related to quotients and residuals @xcite .
this notion can be generalized to formal power series straightforwardly .
[ de : factor ] for any @xmath394 , if @xmath325 , then we call @xmath395 a sub - factorization of @xmath0 . furthermore , if the sub - factorization @xmath395 is maximal , i.e. , if @xmath396 , @xmath397 and @xmath398 , then @xmath399 and @xmath400 . in this case
we call @xmath395 a factorization of @xmath0 , and write @xmath401 for the set of all factorizations of @xmath0 .
the relationship between residuals and factorizations of a series is as follows .
[ pro : factor - language ] let @xmath15 be a complete c - semiring . for @xmath387 , we have * @xmath402 if and only if @xmath403 and @xmath404 . * if @xmath405 , then there exists @xmath402 such that @xmath406 and @xmath407 .
\(i ) suppose @xmath402 we show @xmath403 and @xmath404 .
for any @xmath288 , we note that @xmath408 if and only if @xmath383 .
hence , @xmath409 .
this shows that @xmath410 .
conversely , by @xmath411 and the maximality of the factorization , we know @xmath412 .
hence , @xmath404 .
similarly , we can show @xmath403 . on the other hand ,
suppose @xmath403 and @xmath413 . by definition ,
we know @xmath262 is the largest @xmath338 such that @xmath339 , and @xmath10 is the largest @xmath414 such that @xmath415 .
this shows that @xmath395 is a factorization of @xmath0 .
\(ii ) let @xmath416 , @xmath417 .
it is clear that @xmath406 . by prop .
[ pro : residual quotient ] ( ii ) and ( vi ) , we know @xmath418 and @xmath419 = a / y$ ] .
therefore @xmath402 and @xmath406 and @xmath407 . by the above proposition ,
a factorization of a formal power series @xmath0 is just a pair @xmath395 such that @xmath10 is the right residual of @xmath0 by @xmath262 and @xmath10 is the left residual of @xmath0 by @xmath10 . in this case
, we also call @xmath10 the _ left factor _ of @xmath0 by @xmath262 and @xmath262 the _ right factor _ of @xmath0 by @xmath10 , respectively .
since @xmath420 , by proposition [ pro : factor - language ] ( ii ) @xmath0 itself is both a left factor and a right factor .
we denote the corresponding right factor and left factor by @xmath421 and @xmath422 , respectively , where @xmath423 , and call @xmath424 and @xmath425 the _ initial _ and _ final _ factorization , respectively .
we write @xmath426 ( @xmath427 ) for the set of left ( right ) residuals of @xmath0 , i.e. , @xmath428 let @xmath429 be the mapping defined as @xmath430 . by proposition [ pro : residual quotient ] , it is easy to see that @xmath133 is a bijection .
moreover , we have @xmath431
in this section , we use the residuals of a formal power series @xmath0 to construct a weighted automaton which recognizes @xmath0 . to this end , we introduce the notion of the inclusion degree . [ def : inclusion regree ] suppose @xmath15 is a complete c - semiring . for two @xmath15-subsets @xmath432 ,
the _ inclusion degree _ of @xmath433 into @xmath434 , denoted by @xmath435 , is defined as @xmath436 the following lemma summarizes several useful properties of the inclusion degree operator . [ le : inclusion ] suppose @xmath15 is a complete c - semiring . for @xmath437 , any @xmath438 , and any @xmath439 , we have * @xmath440 iff @xmath441 . * if @xmath442 , then @xmath443 . *
@xmath444 .
[ def : universal automaton ] suppose @xmath15 is a complete c - semiring . for @xmath224 , the _ universal weighted automaton _ of @xmath0 , denoted by @xmath2 , is a weighted automaton @xmath445 , @xmath446 , where @xmath447 for any @xmath442 , @xmath108 .
[ pro : ufa1 ] suppose @xmath15 is a complete c - semiring . for @xmath205 , and @xmath448 and @xmath108
, we have @xmath449 since @xmath402 , @xmath450 similarly , we can prove the case for @xmath451 .
as for @xmath452 , it is sufficient to show that @xmath453 iff @xmath454 for any @xmath438 .
this is because , @xmath455 iff @xmath456 , iff @xmath457 , iff @xmath458 , iff @xmath454 .
hence , @xmath459 .
similarly , we have @xmath460 .
the extension of @xmath461 has the following form .
[ pro : extension ] let @xmath15 be a complete c - semiring and suppose @xmath224 . for @xmath462 , @xmath463 , and any @xmath464 , we have @xmath465 first , for any @xmath438 , @xmath466 iff @xmath467 , iff @xmath468 , iff @xmath469 , iff @xmath470 .
hence , @xmath471 .
similarly , we have @xmath472
. we shall show @xmath473 by induction on the length @xmath124 for @xmath464 . if @xmath474 , this is just the definition of @xmath452 .
given @xmath464 and @xmath118 , we show that ( [ eq : extension of eta ] ) holds for @xmath475 if ( [ eq : extension of eta ] ) holds for @xmath125 .
@xmath476 conversely , @xmath477 which implies @xmath478 hence , @xmath479
. therefore , the equality ( [ eq : extension of eta ] ) holds .
we give some properties of universal weighted automaton as follows .
[ pro : future - past ] let @xmath15 be a complete c - semiring and suppose @xmath224 . for @xmath402
, we have @xmath480 , @xmath481 . for any @xmath390 , @xmath482 on the other hand , @xmath483 hence , @xmath480 .
similarly , we have @xmath481 .
[ th : universal 1 ] let @xmath15 be a complete c - semiring and suppose @xmath224 .
then we have @xmath484 . for any @xmath390 , @xmath485 on the other hand , @xmath486 therefore , @xmath484 .
so far , we have defined two canonical weighted automata for each @xmath224 , viz .
the minimal dwa @xmath1 and the universal weighted automaton @xmath2 .
both automata recognize @xmath0 .
recall in the classical case , if @xmath0 is a regular language over @xmath4 , then @xmath1 and @xmath2 are both finite automata .
this does not hold in general for weighted automata . in
the following we discuss when @xmath2 is finite . for @xmath224 ,
write @xmath487 and let @xmath488 be the @xmath489-sublattice of @xmath15 generated by @xmath490
. then it is well - known that @xmath491 is finite iff @xmath490 is finite ( cf .
@xcite ) .
the following proposition shows the relationship between the finiteness of @xmath2 and the finiteness of @xmath1 .
[ pro : finiteness ] let @xmath15 be a complete c - semiring
. for @xmath224 , the following conditions are equivalent . * @xmath2 is finite , i.e. , @xmath401 is finite . *
@xmath0 can be accepted by a finite dwa , and @xmath490 in eq .
( [ eq : s - a ] ) is finite .
* @xmath1 is finite , i.e. , @xmath211 has finite index , and @xmath490 is finite . in particular , if @xmath15 is finite c - semiring or a linear order lattice , the above conditions are equivalent , and the condition `` @xmath490 is finite '' can be omitted . by proposition
[ pro : dffa ] , @xmath0 is recognized by a finite dwa iff @xmath211 has finite index .
it remains to show that @xmath0 has finite factorizations iff @xmath490 is finite .
assume that @xmath0 can be recognized by a finite dwa and @xmath490 is finite .
this implies that @xmath211 has finite index , i.e. , @xmath221 is a finite set . if @xmath492 , then @xmath493 for any @xmath204 .
it follows that @xmath494 thus @xmath10 induces a unique mapping from @xmath216 into @xmath491 .
since @xmath221 is finite and @xmath490 is finite , the latter implies that the set @xmath491 is also finite .
then the set of the mappings from @xmath221 into @xmath491 is also finite .
therefore , @xmath401 is finite .
on the other hand , suppose @xmath401 is finite .
we first show @xmath1 is a finite dwa .
note that by theorem [ th : universal 1 ] , @xmath0 is accepted by the finite universal weighted automaton @xmath2 . by proposition [ pro : m - minimal ] ( the proof of which is independent to this proposition ) , @xmath1 is a sub - automaton of @xmath2 .
therefore , @xmath0 can be accepted by a finite dwa . to end the proof , we show @xmath490 is finite .
we prove this by contradiction .
suppose @xmath490 is infinite . since @xmath236 is finite , there is @xmath495 such that the subset @xmath496 of @xmath490 is infinite .
assume that @xmath497 for @xmath197 .
for any @xmath438 , define a series @xmath498 as , @xmath499 if @xmath500 and @xmath501 otherwise .
consider the right residual @xmath502 , by a simple calculation , we have @xmath503 .
it follows that the set @xmath504 is infinite , then @xmath427 is infinite , and thus @xmath401 is infinite . a contradiction
therefore @xmath490 is finite .
we give two examples to illustrate the construction of the universal weighted automaton of a formal power series .
[ ex : finite ] assume @xmath505 , @xmath506 , where @xmath15 is a linear order lattice under the natural order of the integer numbers . consider the formal power series @xmath224 defined as follows , @xmath507 @xmath0 can be recognized by the dwa @xmath388 presented in figure 1 , where @xmath508 . ,
scaledwidth=50.0% ] the set of factorizations of @xmath0 is @xmath509 , where @xmath510 , @xmath511 , for all @xmath512 ; @xmath513 @xmath514 @xmath515 of the formal power series @xmath0,scaledwidth=50.0% ] by definition , the universal weighted automaton of @xmath0 is @xmath516 , @xmath517 , where for the value @xmath518 of @xmath519 in the given @xmath15-subset . ] * @xmath520 , * @xmath521 , * @xmath522 is either 2 or 1 , as shown in figure 2 .
[ ex : infinite ] assume @xmath122 , @xmath523 is the tropical semiring .
consider the formal power series @xmath224 defined as follows @xmath524 by proposition [ pro : dffa ] , @xmath0 can not be recognized by any finite dwa .
however , as a regular series , @xmath0 can be recognized by a finite weighted automaton @xmath525 , where @xmath526 , @xmath527 .
the universal weighted automaton @xmath2 has infinite states @xmath528 , where @xmath529 ; and for @xmath530 , @xmath531 @xmath532 , for any @xmath533 . for any @xmath534
, we have @xmath535 if @xmath536 and @xmath537 otherwise , @xmath538 and @xmath539 . in this case , @xmath2 is not a finite weighted automaton .
in this section , we show the weighted automaton @xmath2 defined in the previous section satisfies the following universal property .
[ dfn : universal - property ] suppose @xmath224 and @xmath540 is a weighted automaton that recognizes @xmath0 .
we say @xmath540 satisfies the _
universal property _ if there exists a morphism from @xmath541 to @xmath540 for any weighted automaton @xmath541 such that @xmath542 . to demonstrate the universality of @xmath2
, we introduce a canonical mapping for each weighted automaton @xmath541 that recognizes a subset of @xmath0 .
let @xmath15 be a complete c - semiring .
suppose @xmath224 and @xmath130 is a weighted automaton such that @xmath542 .
we define @xmath543 by @xmath544 for any @xmath545 , where @xmath546 we call @xmath547 the canonical mapping from @xmath541 to @xmath2 .
the following lemma shows that @xmath133 is a morphism .
[ lemma : universality ] let @xmath15 be a complete c - semiring .
suppose @xmath224 . if @xmath130 is a weighted automaton such that @xmath542 , then the canonical mapping @xmath547 is a morphism from @xmath137 into @xmath2 . if @xmath544 , then it is obvious that @xmath548 and @xmath549 .
it follows that @xmath550 and @xmath551 . for the remainder
part , notice that @xmath552 . then @xmath553 i.e. , @xmath554 . hence , @xmath555 .
then it follows that @xmath556 therefore , @xmath547 is a morphism from @xmath137 into @xmath2 .
the universality of @xmath2 follows immediately .
[ th : universality ] let @xmath15 be a complete c - semiring . for @xmath224 ,
the universal weighted automaton @xmath2 satisfies the universal property , i.e. there is a morphism to @xmath2 from any weighted automaton @xmath137 such that @xmath542 .
the following lemma establishes the connection between the formal power series accepted by two weighted automata connected by a morphism . [ lemma : morphism1 ] let @xmath15 be a complete c - semiring .
suppose @xmath99 and @xmath130 are two weighted automata .
if @xmath133 is a morphism from @xmath104 into @xmath137 , then we have @xmath557 for any @xmath113 , and thus , @xmath558 . if @xmath133 is a strong homomorphism , then @xmath559 . for any @xmath390 , @xmath560 hence , @xmath561 and @xmath562 .
then it follows that @xmath558 .
if @xmath133 is a strong homomorphism , then it can be easily verified that @xmath563 for any @xmath390 .
then it follows that @xmath564 hence , @xmath559 .
suppose @xmath99 is a weighted automaton .
we say two states @xmath111 and @xmath112 in @xmath100 are _ mergible _ in @xmath104 if there exist a weighted automaton @xmath130 that accepts the same language as @xmath104 and a surjective morphism @xmath565 such that @xmath566 .
[ pro : mergible ] let @xmath15 be a complete c - semiring .
suppose @xmath213 .
then there is no mergible states in the universal weighted automaton @xmath2 .
otherwise , there is a weighted automaton @xmath567 and a surjective morphism @xmath568 such that @xmath569 , and @xmath570 for two distinct states @xmath395 and @xmath571 in @xmath2 .
then we have , @xmath572 and thus , @xmath573 .
similarly , we have @xmath574 .
therefore , @xmath575 this contradicts with the maximality of the factorization @xmath395 and @xmath571 .
in fact , @xmath2 is the largest non - mergible weighted automaton .
[ co : largest ] let @xmath15 be a complete c - semiring .
suppose @xmath224 .
then @xmath2 is the largest weighted automaton among those that accept @xmath0 but have no mergible states .
a weighted automaton @xmath137 accepting @xmath0 that has strictly more states than @xmath2 is sent into @xmath2 by a morphism which is necessarily non - injective .
moreover , @xmath2 is the smallest ` universal ' weighted automaton .
[ pro : least ] let @xmath15 be a complete c - semiring .
suppose @xmath224 .
then @xmath2 is the smallest weighted automaton among those that accept @xmath0 and have the universal property .
suppose that @xmath567 has the universal property with respect to the @xmath0 . as @xmath2 accepts @xmath0 , there should be a morphism from @xmath2 into @xmath567 .
as @xmath2 has no mergible states , this morphism should be injective : @xmath567 has at least as many states as @xmath2 .
applying theorem [ th : universality ] to weighted automata that accepts @xmath0 , we obtain the following corollary .
[ co : m - minimal ] let @xmath15 be a complete c - semiring .
suppose @xmath224 and @xmath137 is a weighted automaton that accepts @xmath0 . if @xmath541 has no mergible states , then @xmath137 is a sub - automaton of @xmath2 . by theorem [ th : universality ]
, the canonical mapping @xmath133 from @xmath137 to @xmath2 is a morphism .
because @xmath137 has no mergible states , we know @xmath133 must be one - to - one .
therefore , @xmath137 is a sub - automaton of @xmath2 .
it is not difficult to show that any minimal weighted ( determinate or non - determinate ) automaton that accepts @xmath0 has no mergible states .
the above corollary then suggests a simple way for searching the minimal ( non - determinate ) weighted automaton that accepts @xmath0 : it suffices to check the sub - automaton of the universal automaton @xmath2 which accepts @xmath0 and has minimal states .
the following proposition shows that @xmath1 , the minimal dwa that accepts @xmath0 , is also a sub - automaton of @xmath2 . [
pro : m - minimal ] let @xmath15 be a complete c - semiring .
suppose @xmath205 .
then @xmath1 is a sub - automaton of @xmath2 . by corollary [ co : m - minimal ] , we need only show that @xmath1 has no mergible states .
recall the minimal dwa that accepts @xmath0 is @xmath576 , where * @xmath577 , * @xmath578 , * @xmath579 , * @xmath580 , * @xmath581 is @xmath582 . then for any state @xmath583 , @xmath584 .
we show that there are no mergible states in @xmath104 .
otherwise , there are two distinct states @xmath198 , @xmath585 in @xmath100 , but there exists another weighted automaton @xmath137 and a morphism @xmath133 from @xmath104 into @xmath137 such that @xmath137 is the morphic image of @xmath104 , @xmath559 and @xmath586 .
since @xmath587 , there exists @xmath88 such that @xmath588 , i.e. , @xmath589 .
note that @xmath590 then @xmath591 since @xmath592 , it follows that @xmath593 , a contradiction occurs .
therefore , @xmath104 has no mergible states .
in general , it is not effective to construct all factorizations of @xmath0 .
suppose @xmath388 is an arbitrary dwa accepting @xmath0 .
in this section , we give an effective method to construct @xmath2 by using the dwa @xmath594 .
let @xmath595 be the @xmath596-sublattice generated by @xmath491 as defined in eq.([eq : s - a ] ) , i.e. , @xmath597 it is well known that @xmath595 is finite iff @xmath491 is finite ( cf.@xcite ) iff @xmath490 is finite .
write @xmath598 . if @xmath104 is finite and @xmath490 is finite , then @xmath599 is also finite .
we construct a weighted automaton @xmath600 as follows : @xmath601 where @xmath602 is defined by @xmath603 .
we next define a mapping @xmath133 from @xmath604 to @xmath2 . to this end
, we first establish the correspondence between weighted states and factorizations of @xmath0 .
[ prop : xf&yf ] let @xmath15 be a complete c - semiring .
suppose @xmath224 and @xmath388 is an arbitrary dwa accepting @xmath0 .
then @xmath605 is a factorization of @xmath0 for any weighted state @xmath606 , where @xmath607 and , for any @xmath390 , @xmath608 therefore , the mapping @xmath133 defined by @xmath609 is a mapping from @xmath598 to @xmath401 . without loss of generality , we assume that @xmath104 is accessible .
since @xmath104 is a dwa , for any @xmath113 , there exists @xmath610 such that @xmath611 .
moreover , if @xmath612 for another @xmath375 , then @xmath593 for any @xmath439 . by this observation
, we have , for any @xmath204 , @xmath613 if we let @xmath614 for any @xmath610 , then we obtain a series @xmath615 such that @xmath616 .
the mapping @xmath617 is also onto .
[ pro : facor - state1 ] let @xmath15 be a complete c - semiring .
suppose @xmath224 and @xmath388 is a dwa that accepts @xmath0 . for any @xmath402 , there is a weighted state @xmath618 such that @xmath619 .
define a weighted state @xmath618 as , for any @xmath113 , @xmath620 by the proof of proposition [ pro : finiteness ] , @xmath10 induces a unique mapping from @xmath216 into @xmath491 , so @xmath433 is well - defined .
we show @xmath619 in the following . for any @xmath390
, we have @xmath621 hence , @xmath619 .
furthermore , we have [ pro : strong homo ] let @xmath15 be a complete c - semiring .
. the mapping @xmath133 defined in proposition [ prop : xf&yf ] is a strong homomorphism from weighted automaton @xmath600 _ onto _ the universal weighted automaton @xmath2 , and thus @xmath622 .
see appendix b. define an equivalence relation @xmath623 on @xmath599 as follows : @xmath624 it is clear that @xmath625 iff @xmath626 . using this equivalence relation ,
we obtain a quotient weighted automaton from @xmath604 , denoted by @xmath627 , which is isomorphic to @xmath2 .
[ co : construction ] let @xmath224 and @xmath604 be as in proposition [ pro : strong homo ] .
suppose @xmath627 is the quotient weighted automaton of @xmath604 modulo the equivalent relation @xmath623 on @xmath599 .
then @xmath628 is isomorphic to @xmath2 .
once the dwa @xmath104 is finite and @xmath490 is finite ( this condition can be guaranteed if @xmath15 is finite or @xmath15 is a linear - order lattice as declared in proposition [ pro : finiteness ] ) , the equivalence @xmath623 defined by eq.([eq : sim - relation ] ) can be effectively constructed .
this is because , @xmath629 takes at most @xmath630 states , i.e. , the set @xmath631 as a subset of @xmath100 has at most @xmath632 states , the corresponding @xmath633 has at most @xmath632 values .
therefore , it is sufficient to check these states in eq.([eq : sim - relation ] ) . in turn
, it is sufficient to check those @xmath390 with @xmath634 in eq.([eq : sim - relation ] ) .
hence , the equivalence relation @xmath623 is decidable and the weighted automaton @xmath627 can be effectively constructed .
we next give one example .
[ ex : finite - state ] consider the formal power series @xmath0 in example [ ex : finite ] , which is recognized by the finite dwa as shown in figure 1 .
then the weighted automaton @xmath635 , where @xmath636 with @xmath637 and @xmath638 for any @xmath113 , @xmath639 @xmath640 and @xmath641 and @xmath642 is either 2 or 1 ( see figure 3 ) . clearly , @xmath643 is isomorphic to @xmath2 .
, scaledwidth=50.0% ] we next give a detailed examination of the equivalent relation @xmath623 .
the correspondence @xmath133 between the weighted states and the factorizations of @xmath0 stated in proposition [ prop : xf&yf ] may be not one - to - one .
there may have more than one weighted states correspond to a given factorization .
however , the following proposition asserts that there exists a largest weighted states .
[ pro : largest state ] let @xmath15 be a complete c - semiring .
suppose @xmath205 and @xmath388 is a dwa that accepts @xmath0 . if @xmath644 for all weighted states @xmath645 ( @xmath76 ) , then @xmath646 .
therefore , for each right residual @xmath262 , there exists a largest @xmath647 such that @xmath648 . by the above propositions , we know for each right residual @xmath262 of @xmath0 there exists a ( unique ) largest weighted state @xmath647 such that @xmath648 . furthermore , suppose @xmath402 .
then @xmath649 is defined by @xmath650 as implied in the proofs of propositions [ prop : xf&yf ] and [ pro : facor - state1 ] . in general , for a weighted state @xmath618 , the structure of the largest weighted state @xmath649 such that @xmath651 is unclear .
but we have the following estimation .
[ pro : largest element1 ] let @xmath15 be a complete c - semiring .
suppose @xmath652 , @xmath653 is a weighted automaton , and @xmath618 a weighted state .
if @xmath654 and @xmath655 , then @xmath656 , where @xmath657 .
let @xmath658 .
then for each @xmath659 we have @xmath660 thus @xmath661 .
it follows that @xmath662 , i.e. , @xmath663 for any @xmath113 .
therefore , @xmath664 for any @xmath665 , hence @xmath656 . recall that if @xmath15 is the two elements boolean algebra @xmath666 , then @xmath667 forms the whole set of those largest elements ( cf.@xcite ) .
it is still unclear whether each largest weighted state can be represented as the intersection of @xmath668 for a set of @xmath669 .
for a formal power series @xmath0 , we have introduced the notions of ( left and right ) quotients and ( left and right ) residuals of @xmath0 .
as can be seen in the above discussion , these two kinds of operations have different behaviors when considering their algebraic and language properties .
especially , with the associated weighted automata , @xmath298 and @xmath2 have different structure , although they are equivalent as language recognizers .
for example , there exists series @xmath0 such that @xmath298 is infinite but @xmath2 is finite . in this section , we consider the order relation between the quotients and the residuals of a same series @xmath0 . because the duality between left and right quotients ( residuals ) , we need only compare the left quotient @xmath260 and the left residual @xmath323 of @xmath0 by a series @xmath10 .
our results show that all the four possibilities are possible .
first , if @xmath10 is a word @xmath197 or @xmath670 and @xmath671 , then it is obvious that @xmath672 .
second , let @xmath673 be the two element boolean algebra . then for any languages @xmath674 we have @xmath675 moreover , @xmath672 iff @xmath676 for any @xmath677 ; but if there exists @xmath677 such that @xmath678 , then the above inclusion is strict .
third , let @xmath15 be the tropical semiring @xmath679 , @xmath680 . for @xmath224 , @xmath197 , and @xmath681 , we have @xmath682 , while @xmath683 if @xmath684 and @xmath685 . therefore @xmath686 if @xmath684 and @xmath685 .
fourth , let @xmath15 be the tropical semiring @xmath687 . take @xmath505 .
consider the formal power series @xmath224 defined as follows @xmath688 let @xmath689 .
we show @xmath260 and @xmath323 are incomparable .
by @xmath690 we know @xmath691 , but @xmath692 . therefore @xmath260 and @xmath323 are incomparable .
it is still unclear when ( i.e. for what kind of @xmath15 and @xmath224 ) we will have an uniform order relation between @xmath260 and @xmath323 .
in this paper , we have defined the quotient and residual operations for formal power series .
the algebraic and closure properties under these operations were discussed .
our results show that most nice properties are kept in formal power series .
moreover , we introduced two canonical weighted automata @xmath1 and @xmath2 for each formal power series @xmath0 using the quotients and , respectively , residuals of @xmath0 .
it was shown that @xmath1 is the minimal dwa of @xmath0 , and @xmath2 is the universal weighted automaton of @xmath0 which contains as a sub - automaton any weighted automaton that accepts @xmath0 but has no mergible states .
in particular , any minimal weighted ( deterministic or non - deterministic ) automaton of @xmath0 is a sub - automaton of @xmath2 .
this suggests an efficient way to find ( approximations of ) the minimal weighted nfa of @xmath0 .
last but not least , we also showed , under a rather weak restriction , that @xmath2 is finite iff @xmath1 is finite , and that @xmath2 can be effectively constructed when we have a finite dwa that recognizes @xmath0 .
there are still several open problems left unsolved .
suppose @xmath10 and @xmath0 are two formal power series .
is @xmath323 regular ( context - free ) whenever @xmath0 is regular ( context - free ) ?
what is the precise relation between @xmath260 and @xmath323 ? when characterizing residuals in terms of quotients by word , we require the underlying semiring to be a complete c - semiring . it is easy to see that this requirement is not necessary .
another problem then is , to what extent can we develop the related theory in a weaker semiring structure , e.g. in quantale ?
another interesting question is to develop an abstraction scheme for formal power series and weighted automata based on semiring homomorphism . just like in soft constraint satisfaction @xcite
, we expect that semiring homomorphisms can play an important role in approximately computing the minimal nfa and the universal weighted automaton of a formal power series .
\(i ) suppose @xmath99 is a weighted automaton accepting @xmath0 .
then , for any @xmath390 , we have @xmath693 define another two weighted automata @xmath694 and @xmath695 , @xmath696 , where @xmath697 @xmath698 then @xmath699 @xmath700 hence , @xmath701 , and @xmath702 are regular .
\(ii ) replace weighted automaton @xmath104 in the proof of ( i ) by a dwa , we can prove that @xmath702 is dwa - regular .
since @xmath0 is dwa - regular , by proposition [ pro : dffa ] , there are @xmath703 and regular languages @xmath241 such that @xmath704 . by proposition [ pro : quotient - fps ] , @xmath705 . by lemma [ lemma : right - quotient ] ,
@xmath706 , it follows that @xmath707 is dwa - regular for any @xmath708 , and thus @xmath260 is dwa - regular by proposition [ pro : dffa ] again .
\(iii ) let @xmath709 .
then @xmath710 and @xmath711 is the proper part of @xmath262 .
it is obvious that @xmath262 is regular iff @xmath711 is regular . by proposition [ pro : quotient - fps ] , @xmath712 .
@xmath713 is context - free since context - free languages are closed under scalar operation . to show @xmath264 is context - free
, it suffices to show that @xmath714 is context - free . without loss of generality ,
we assume that @xmath262 is proper , i.e. , @xmath715 .
there exist a context free - grammar @xmath716 and a weighted automaton @xmath291 satisfying the conditions of lemma [ le : proper ] , such that @xmath717 and @xmath718 .
construct a new weighted context - free grammar @xmath719 , where , @xmath720 , @xmath721 , @xmath722 is constructed as follows , ( ii-1 ) @xmath723 for each @xmath724 ; ( ii-2 ) @xmath725 if @xmath726 and @xmath727 ; ( ii-3 ) @xmath728 if @xmath729 and @xmath730 .
let @xmath88 .
we note that @xmath731\}\\ ay^{-1}(w ) & = & \sum\{a(wu)y(u ) | u\in{\sigma^{\ast}}\ } \\ & = & \sum\{(r_{11}\otimes r_{12}\otimes \cdots\otimes r_{1l})\otimes ( \delta(q_0,\sigma_1,q_1)\otimes \cdots \otimes \delta(q_{m-1},\sigma_m , q_f))| \\ & & \hspace*{10 mm } ( \exists \alpha_1,\cdots,\alpha_{l-1 } ) ( \exists \sigma_1,\cdots,\sigma_m \in \sigma)(\exists q_0 , q_1,\cdots , q_{m-1}\in q ) \\ & & \hspace*{30 mm } \mbox{such that\ } s\overset{r_{11}}{\rightarrow } \alpha_1\overset{r_{12}}{\rightarrow}\cdots\overset{r_{1l}}{\rightarrow}w\sigma_1\cdots\sigma_m\}.\end{aligned}\ ] ] to show @xmath732 , it suffices to show that they have the same non - zero sum - terms in @xmath15 .
that is , each non - zero sum - term @xmath733 in the equality of @xmath734 also appears in the equality of @xmath735 as a sum - term , and vise verse .
on one hand , suppose there exists a sequence @xmath736 such that @xmath737 i.e. , @xmath738 is a non - zero term in @xmath734 .
notice that the semiring @xmath15 is assumed to be commutative and a production of type ( ii-3 ) commutes with one of the type ( ii-1 ) or ( ii-2 ) .
the above sequence of productions can be rearranged so that all the productions of type ( ii-3 ) precede those of type ( ii-1 ) and ( ii-2 ) .
hence we may assume that by type ( ii-3 ) productions @xmath739 and by type ( ii-1 ) and ( ii-2 ) productions @xmath740 since the induction ( [ eq:24 ] ) is by type ( ii-1 ) and ( ii-2 ) , each @xmath741 is in @xmath4 .
since every term in the induction ( [ eq:23 - 2 ] ) corresponds to a production of @xmath157 , it follows that there exists @xmath742 such that @xmath743 is an induction in @xmath159 .
furthermore , for each @xmath744 , we have either @xmath745 or @xmath746 for some @xmath747 .
let @xmath665 be the largest integer such that @xmath748 and let @xmath749 , @xmath750 . because @xmath293 for any @xmath113 , it follows that @xmath751 and @xmath752 for any @xmath753 . omitting the weight @xmath19 by using the type ( ii-1 ) productions , i.e. , @xmath754
, it follows that the leaving terms , after rearranging , is @xmath755 and @xmath756 this shows that the term @xmath757 is also a term of @xmath735 . on the other hand , assume that @xmath758 in @xmath159 for @xmath759 and @xmath760 , and there exists states @xmath761 in @xmath100 such that @xmath762 then @xmath763 is a term in the sum @xmath735 . since the induction ( [ eq:25 ] ) is in @xmath159 , by type ( ii-3 ) productions we have @xmath764 where @xmath765 is chosen as in ( [ eq:26 ] ) for @xmath766 . applying type ( ii-1 ) productions
to @xmath767 and type ( ii-2 ) productions to @xmath768 , @xmath769 we see that @xmath770 hence , @xmath771 is a term in the sum of @xmath734 . therefore , @xmath732 for any @xmath88 .
therefore , @xmath772 and @xmath264 is context - free .
since @xmath15 is commutative , by the duality between left and right quotients , it follows that @xmath260 is also context - free once @xmath10 is regular and @xmath0 is context - free .
\(iv ) by the proof of statement ( iii ) , we have the following simple observation : if @xmath104 is a dwa in the proof of statement ( iii ) , i.e. , @xmath104 is a classical deterministic finite automaton with a unique final state .
then the commutativity of @xmath15 is not necessary in the proof of the above proposition .
this is because , in this case , the weights in type ( ii-1 ) and ( ii-2 ) productions take values @xmath19 , and a production of type ( ii-3 ) commutes with one of type ( ii-1 ) or ( ii-2 ) . in this case , if we let @xmath773 , then @xmath264 is context - free . for a general finite dwa @xmath774 , we write @xmath775 for a finite dwa with unique final state @xmath112 for any state @xmath112 in @xmath100 .
if we let @xmath776 , then @xmath777 is context - free by the above observation . by a simple calculation
, we have @xmath778 .
then , by proposition [ pro : quotient - fps ] ( ii ) , we have @xmath779 since @xmath777 is context - free for any @xmath113 and the family of weighted context - free language is closed under scalar product and finite sum , it follows that @xmath264 is context - free .
similar to the proof of statement ( ii ) , @xmath260 is context - free if @xmath10 is dwa - regular and @xmath0 is context - free .
this shows that the statement ( iv ) holds .
let us first show the following equality holds , @xmath780 consider @xmath781 . noting that if @xmath6 does not have the form @xmath786 for any @xmath375 , then @xmath787 .
thus , @xmath788 next , let us prove the following equality @xmath789 on one hand , suppose @xmath790 for some @xmath791 and @xmath792
. then we have @xmath793 for any @xmath113 .
therefore , @xmath794 for any @xmath113 . hence , @xmath795 this further implies that @xmath796 we next show @xmath801 for some weighted state @xmath434 ( in fact , the largest weighted state @xmath434 such that @xmath802 ) , where @xmath434 is defined by , @xmath803 by proposition [ pro : facor - state1 ] , @xmath434 satisfies @xmath802 .
then @xmath804 and then eq.([eq:49 ] ) holds .
we next show that @xmath133 also satisfies the following two conditions : @xmath807 for eq.([eq : ja_xy ] ) , we have @xmath808 , and @xmath619 , then @xmath809 . if we let @xmath810 , and thus @xmath811 can take @xmath812 , this shows that eq.([eq : ja_xy ] ) holds . | quotient is a basic operation of formal languages , which plays a key role in the construction of minimal deterministic finite automata ( dfa ) and the universal automata . in this paper
, we extend this operation to formal power series and systemically investigate its implications in the study of weighted automata . in particular , we define two quotient operations for formal power series that coincide when calculated by a word .
we term the first operation as ( left or right ) _ quotient _ , and the second as ( left or right ) _ residual_. to support the definitions of quotients and residuals , the underlying semiring is restricted to complete semirings or complete c - semirings .
algebraical properties that are similar to the classical case are obtained in the formal power series case .
moreover , we show closure properties , under quotients and residuals , of regular series and weighted context - free series are similar as in formal languages . using these operations ,
we define for each formal power series @xmath0 two weighted automata @xmath1 and @xmath2 .
both weighted automata accepts @xmath0 , and @xmath1 is the minimal deterministic weighted automaton of @xmath0 .
the universality of @xmath2 is justified and , in particular , we show that @xmath1 is a sub - automaton of @xmath2 .
last but not least , an effective method to construct the universal automaton is also presented in this paper . formal power series ; weighted automaton ; complete c - semiring ; quotient ; residual ; universal automaton ; factorization |
carbon nanotubes ( cnts ) are ideal model systems for the exploration of electrical transport in low dimensions.@xcite two sorts of nanotubes ( nts ) exist : single - wall and multi - wall nts .
an ideal ( undoped ) swnt can be either metallic or semiconducting , but here , we only focus on metallic nts . in these tubes ,
the current is carried by two modes , modes , spin - degeneracy is not included . ]
leading to a conductance of @xmath4 , provided backscattering is absent .
recent experiments have shown that scattering within metallic swnt is weak . in a particular nice experiment
the electrostatic potential drop along a voltage - biased nt was measured by using the tip of an atomic - force microscope as a probe.@xcite for swnts , most of the potential drops at the contacts .
in contrast , for mwnts a considerable fraction of the potential drops along the tube , suggesting intrinsic scattering in mwnts .
a length dependent resistance was deduced before from electric resistance measurements on multiply contacted mwnts.@xcite the typical value for the resistance per unit length is .
we mention , that there is one conflicting results : frank _ et al._@xcite came to the conclusion that mwnts are ballistic conductors even at room temperature.d subbands is comparable to @xmath5 . ]
seemingly compelling evidence for diffusive transport in mwnts is provided by measurements of the magnetoresistance , both in parallel and perpendicular magnetic field.@xcite for example , the resistance modulation in parallel magnetic field can be described very well by the altshuler - aronov - spivak ( aas ) theory ( weak - localization in a cylindrical conductor ) , which relies on _ diffusive _
transport.@xcite these experiments did also show that the electrical current is preferentially carried by the outermost tube , at least at low temperatures .
hence , a single nanotube is probed , albeit one with a large diameter of , which is about ten times larger than that of prototype swnts .
as emphasized before , a metallic swnt is characterized by only @xmath6 @xmath2d - modes , a property that should be independent of the diameter .
how can we than reconcile the availability of only @xmath7 modes for an ideal nt with the observation of diffusive motion .
diffusive transport requires @xmath8 .
may it be that mwnts are doped to a such a degree that @xmath8 ? by using a new gating technique ( electrochemical gating ) , we have recently shown that mwnts are indeed ( hole- ) doped.@xcite the number of @xmath2d - modes is @xmath9 , but rather @xmath10 , see section 2 .
mwnts are not single - mode , but rather few mode quasi - one - dimensional wires .
whether they are @xmath2d diffusive , i.e. quasi - ballistic with a mean - free path @xmath11 exceeding the circumference @xmath12 , or @xmath13 diffusive is another question . taking ( @xmath14 is tube length ) , yields which is of order of the circumference .
this simple estimate is in good agreement with measurements of the energy - dependent tunneling dos @xmath15 .
@xmath15 is not structureless , as would be expected if , but shows features reminiscent of quantization into @xmath2d - modes , albeit with some broadening.@xcite recently , we have studied gate - induced conductance fluctuations in mwnts at low temperatures and tried to compare the measurements with ucf theory . in the regime of thermally - induced averaging ,
i.e. for tubes which are much longer than the phase - coherence length @xmath16 and/or the thermal length @xmath17 , the functional dependence is in agreement with theory .
these data allow to deduce @xmath16 , which follows nyquist - dephasing below .
however , when we approach the universal limit , i.e. if @xmath18 , the temperature dependence of conductance fluctuations markedly deviates from standard theories .
this has led us to study shorter tubes in more detail by measuring the differential conductance @xmath19 as a function of transport voltage @xmath20 and gate voltage @xmath21 in the fully coherent regime , i.e. for @xmath22 .
displaying @xmath23 in a greyscale plot helps to recognize the underlying physics .
this is in particular true for single - electron charging effects which might be present simultaneously to quantum interference effects , both modulating the equilibrium conductance .
single - electron charging effects ( single - electron tunneling = set ) , such as coulomb blockade and coulomb oscillations were observed in swnts from the beginning.@xcite however , in our own work on mwnts we have never observed clear evidence of coulomb blockade until now .
we have argued that this absence is due to the low - ohmic contacts in our experiments , which are always of order .@xcite in contrast , measurements on mwnts with high - ohmic contacts ( ) display the conventional features of single - electron charging effects.@xcite the evaporation of au over the nanotubes , the method we prefer for fabricating contacts , leads to contact resistances that can be as low as at room temperature .
this is low enough to suppress set . at cryogenic temperature , however , contact resistances usually increase , so that set may show up . in section 3
we present our first ( and still rare ) observation of set in transport through a mwnt with ` low - ohmic ' contacts .
since the coupling to the contacts is rather strong , the conductance is dominated by higher - order co - tunneling processes .
these new data allow to extract the single - particle level spacing @xmath24 of the mwnt quantum dot .
similar to the result from electrochemical gating , the measured @xmath24 suggests that @xmath25 modes are occupied .
in order to determine the degree of doping in mwnts ( or other nanotubes ) the position of the fermi energy need to be determined .
one possible approach is to measure the conductance as a function of a gate voltage @xmath21 , which shifts the fermi energy @xmath0 .
if the conductance @xmath26 increases ( decreases ) with increasing @xmath0 , the nt is n - doped ( p - doped ) , see fig .
1a . at , or in the vicinity of the charge - neutrality point ( cnp ) ,
the number of modes is minimal .
it is @xmath7 for a metallic tube and zero for a semiconducting one.d bandstructure in fig .
1a is considerably smeared , so that no abrupt transitions are expected if @xmath0 crosses a band onset . ]
hence , one expects to see a conductance minimum ( or a resistance maximum ) if @xmath0 crosses the cnp .
ideally , for an undoped nt , @xmath0 should lie at the cnp and the resistance maximum should appear at @xmath27 .
if this maximum is observed at @xmath28 , the fermi level is shifted due to some doping . in order to deduce @xmath0 from the measured gate voltage ,
the electrochemical capacitance between gate and nt needs to be known .
conventional back gates ( see fig .
1b ) have relatively low capacitance .
consequently fermi - level shifts are small even for large applied gate voltages . in this case , the resistance maximum can not be measured , and hence , the doping level can not be determined .
this problem is circumvented in our approach , in which the coupling capacitance is made very large .
we immerse the nanotube into an electrolyte through which gating is achieved , see fig 1c . for details , see ref .
very large gate capacitances are possible due to the large double - layer capacitance of the electrolyte nt interface .
this has an important implication : because the double - layer capacitance is much larger than the electrochemical capacitance of the nt , the total capacitance ( series connection of the two ) is determined by the nanotube alone .
hence , there is a one - to - one correspondence between gate - voltage change @xmath29 and fermi level shift @xmath30 , i.e. @xmath31 . electrochemical gating is studied at room temperature on single mwnts with lithographically defined au contacts evaporated over the nts .
the nanotube - contact structure is fabricated on degenerately doped si with a thick spacer layer .
the subtsrate can be used as a conventional backgate , fig 1b . if , as outlined in fig .
1c , a positive gate voltage @xmath21 is applied , the nt - electrolyte interface is polarized by the attraction of cations . consequently , the fermi level increases in the nt to maintain charge neutrality .
an example for the dependence of the nanotube resistance on ( electrochemical ) gate voltage is shown in fig . 2 .
a pronounced resistance maximum is seen at @xmath32 , which we identify as the cnp . in order to reach the cnp the fermi energy
must first be raised by .
this suggests that the nt is considerably hole - doped at @xmath27 .
there is , however , a time dependence ( not shown).@xcite after immersion , the resistance maximum is observed at a smaller voltage of .
the position of maximum @xmath33 then gradually moves to larger voltages to finally stabilize at in this example .
hence , the doping level increases in the electrolyte . typical fermi - level shifts in air ( before immersion ) amount to . based on measurements of the electrical resistance of swnts in a controlled environment of different gases
it was suggested that hole - doping is induced by by oxygen.@xcite our own experiments show that oxygen has an effect on the doping state of mwnts . however , the effect is weak , as much larger doping shifts are seen in ambient air .
an example is shown in fig .
, we compare the evolution of the resistance of a mwnt at zero ( back- ) gate voltage when exposed to oxygen and air . prior to exposure
, the nt was annealed at in vacuum . as a consequence of added hole doping the resistance
is seen to decrease .
the decrease is much more pronounced for air than for oxygen .
the resistance increases again in vacuum , nicely demonstrating that the oxidizing species are only weakly adsorbed .
the strongest source of doping is most likely due to water , as large fermi - level shifts can be induced by water vapour alone . for the interpretation of previous electrical measurements ,
the net doping concentration @xmath34 and the fermi - level shift for a ` virgin ' mwnt in air are important .
typical values for the latter are . comparing this with the average @xmath2d subband spacing ( for a diameter nt )
, we conclude that @xmath35 subbands may contribute to the conductance instead of @xmath7 for an ideal metallic nt .
a doping - induced corresponds to a doping concentration of , or expressed per surface area to giving approximately one elementary charge per @xmath36 carbon atoms .
as mentioned in the introduction single - electron tunneling ( set ) such as coulomb oscillations have been observed in mwnts and swnts , provided the contacts were high - ohmic , i.e. .@xcite in the opposite limit of relatively low - ohmic contacts of order charging effects have not yet been observed in transport through mwnts . though we have observed pronounced conductance oscillations as a function of gate voltage @xcite and also as a function of magnetic field in the latter type of samples @xcite ,
these can not readily be assigned to charging effects , as the oscillations were strongly aperiodic .
we have assigned these oscillations to quantum interference of non - interacting quasi - particles , with the interference being caused by the presence of randomly distributed scatterers .
it may be that this interpretation need to be reconsidered , since we have recently observed set effects in mwnts low - ohmic contacts ( ) . in order to entangle set from quantum interference
it is useful to measure @xmath37 as a function of gate voltage @xmath21 and transport voltage @xmath20 .
the obtained greyscale plots greatly helps to recognize the underlying physics . in small ( semi ) isolated objects , like our mwnt devices
, transport may be blocked by coulomb interaction ( coulomb blockade ) .
this is the case , if the energy for adding a single electron to the tube ( the single - electron charging energy @xmath38 ) is larger than the thermal energy @xmath5 and if the object is only weakly coupled to external leads ( contact resistance @xmath39 ) . coulomb blockade is suppressed if the energy of two charge states are degenerate .
for example , if @xmath40 , where @xmath41 denotes the total ground - state energy of the nt with @xmath42 electrons , an equilibrium current can flow via the sequence @xmath43 .
such degeneracies can be induced by the gate voltage @xmath21 and appear in the simplest possible model ( constant interaction model ) periodic in @xmath21 . in swnts coulomb blockade has been observed by several groups before.@xcite until now , it was believed that mwnts do not display coulomb blockade because of low contact resistances .
recently , we have observed clear signatures of single - electron charging in a long mwnt ( distance between the contacts ) with reasonably low - ohmic contacts ( i.e. contact resistance @xmath44 ) .
an example is shown in fig .
4 . in ( a ) , measured for @xmath45 , a regular sequence is seen consisting of a large ` diamond ' followed by three smaller ones .
this pattern suggests a four - fold degeneracy of the single - electron states which we assign to spin degeneracy and an additional twofold orbital degeneracy .
the latter may be caused by a peculiar degeneracy inherent to the graphite lattice , the so - called k , k@xmath46 degeneracy which originates from the presence of two identical carbon atoms per unit cell .
if this interpretation is correct , this is quite amusing , because this basic feature has up to now not been seen in swnts .
4b shows @xmath37 measured at .
now , an irregular sequence is observed .
it is well known that the k , k@xmath46 degeneracy is lifted in magnetic - field .
on the one hand , we do see clear signatures of coulomb blockade suggesting low transparent contacts . on the other hand ,
the measured two - terminal resistance is low , of order of @xmath47 , suggesting highly ( or intermediate ) transparent contacts .
therefore , higher - order tunneling processes can play an important role in the electrical transport .
a well known higher - order tunneling process is the kondo effect , i.e. the dynamic screening of the spin on the dot due to exchange - correlation effects between the dot state and the electrodes .
we observe the kondo effect in a mwnt for the first time , see the features in fig .
4 . , which are highlighted by arrows . at @xmath48 and within the region of coulomb blockade
there are two ` ridges ' of enhanced conductance . for the simplest filling scheme , i.e. @xmath49
, these ridges are expected to form in every second ` diamond ' where the spin is @xmath50 .
the kondo effect has been observed before in swnts.@xcite the data in fig . 4
allow to extract the single - particle level spacing @xmath24 .
the large diamonds are large because an electron has to be added to a new level .
the addition energy is therefore the sum of charging energy @xmath51 and @xmath24 . due to the fourfold degeneracy ,
the addition energy for the next three electrons is only @xmath51 .
comparing the energies leads to a mean level spacing of .
the @xmath52d level spacing is predicted to be for one mode , yielding ( ) . the number of modes participating is therefore .
again we see that there are more than @xmath7 modes occupied in mwnts .. ] since the measured two - terminal conductance @xmath53 , is a typical value for all our mwnt devices with contact separation @xmath54 nm . ]
the mean transmission probability @xmath55 per mode is only @xmath56 ( here , we have taken @xmath57 ) . if all modes would have @xmath58 , co - tunneling effects like the kondo effects could safely be neglected .
hence , there must be a distribution of transmission eigenvalues similar to that in many - mode diffusive wires .
for the latter , random - matrix shows that the distribution is bimodal .
although we are dealing with a very small ensemble ( only 10 modes ) , the distribution of transmission values is most likely bimodal , too .
though there are @xmath25 modes available to accommodate charge , charge transport to the contacts is determined by only one mode ( or a very small number of modes ) .
if the mode with highest transmission probability has the appropriate @xmath55 ( not too low , but also not too high ) , the kondo effect will show up.@xcite if @xmath55 is larger , the observed conductance pattern can no longer be assigned to co - tunneling set effects but rather reflects fabry - perot type interference.@xcite
multi - wall nanotubes in air are hole doped , most likely induced by the adsorption of water
. consequently , there are more than two @xmath2d - modes occupied .
a typical number is @xmath59 , not counting spin degeneracy .
estimates for the scattering mean - free path @xmath11 vary widely in the literature and also among our own experiments .
however , surprisingly many experiments suggests that the @xmath11 is of order of the circumference of the tube .
this may be a coincidence , but might also have a deeper reason : it was pointed out very early by white _
et al._@xcite that the mean - free path should grow with tube diameter @xmath60 according to @xmath61 , provided the number of scatterers per unit length is constant . on the other hand ,
if we assume that the source of scattering is a direct consequence of doping , the scattering density is proportional to the doping density which for a given environment can be taken to be constant .
therefore , the number of dopants per unit length ( or the number of scatterers ) grows linearly with @xmath60 .
the combination of both effects results in a diameter independent mean - free path .
another argument for a diameter - independent mean - free path in mwnts was put forward by egger and gogolin.@xcite at low temperature , single - electron tunneling ( set ) effects have been observed in mwnts with low - ohmic contacts , i.e. .
this has allowed to determine the single electron level spacing which appears to be consistent with the estimated number of occupied modes deduced from electrochemical gating .
the presence of co - tunneling points to high transmission at the contacts .
even stronger : the prevailing absence of set effects in mwnts with @xmath53 suggests that at least one highly transmissive channel is present at low temperature .
99 for a review , see physics world , vol . 13 , no . 6 , june 2000 .
a. bachtold _
et al . _ ,
lett . * 84 * , 6082 ( 2000 ) . c. schnenberger _
et al . _ ,
. phys . * a 69 * , 283 ( 1999 ) .
s. frank _ et al . _ ,
science * 280 * , 1744 ( 1998 ) .
a. bachtold _
et al . _ ,
nature * 397 * , 673 ( 1999 ) .
m. krger _
et al . _ ,
. lett . * 78 * , 1291 ( 2001 ) .
a. bachtold _
et al . _ , cond - mat/0012262 .
s. tans _ et al .
_ , nature * 386 * , 474 ( 1997 ) ; m. bockrath _ et al .
_ , science * 275 * , 1922 ( 1997 )
. l. roschier _
et al . _ ,
. lett . * 75 * , 728 ( 1999 ) ; m. ahlskog _
et al . _ ,
appl . phys
. lett . * 77 * , 4037 ( 2000 ) .
et al . _ ,
science * 287 * , 622 ( 2000 ) ; p. g. collins _ et al .
_ , science * 287 * , 1801 ( 2000 ) .
m. buitelaar _
, to be published .
j. nygard _
et al . _ ,
nature * 408 * , 342 ( 2000 ) .
j. nygard and d. cobden , cond - mat/0105289 .
et al . _ ,
nature * 411 * , 665 ( 2001 ) . c.t .
white and t.n .
todorov , nature * 393 * , 240 ( 1998 ) .
r. egger and a.o .
gogolin , cond - mat/0101246 . | employing strong electrostatic gating ( liquid - ion gating ) , the position of the fermi energy @xmath0 ( relative to the charge - neutrality point ) was determined in multi - wall carbon nanotubes ( mwnts ) .
@xmath0 is negative ( hole doping ) and amounts to for mwnts in air .
evidence that water , and not oxygen , is the main source of doping has been found . as a consequence ,
the number @xmath1 of occupied @xmath2d - modes ( not counting spin ) is @xmath3 , i.e. .
this is supported by the single - electron level spacing , deduced from observed single - electron charging effects ( set ) at low temperature .
the latter are dominated by co - tunneling processes , as we observe the kondo effect .
this provides evidence , that highly transmissive channels are present in mwnts at low temperature , despite the ` disorder ' observed in previous experiments . |
most of the accretion onto the supermassive black hole ( smbh ) found in the center of most massive galaxies is heavily obscured by the surrounding dust and gas ( e.g. , @xcite ) . in the local universe , @xmath075% of the seyfert 2
galaxies are heavily - obscured ( @xmath4@xmath110@xmath5 @xmath6 ; @xcite ) .
many of these , if at @xmath7@xmath81 , where most of the black hole growth occurs , would not be identified in x - rays even in very deep ( @xmath11 msec ) chandra or xmm / newton exposures @xcite .
locating and quantifying this heavily obscured smbh growth , in particular at high redshifts , is currently one of the fundamental problems in astrophysics .
because the energy absorbed at optical to x - ray wavelengths is later re - emitted in the mid - ir , it is expected that all active galactic nuclei ( agn ) , even the most obscured ones , should be very bright mid - ir sources ( e.g. , @xcite ) .
hence , it is not surprising that a large number of heavily obscured even compton - thick ( @xmath4@xmath110@xmath9@xmath6 ) agn have been found amongst the luminous and ultra - luminous infrared galaxies ( ( u)lirgs ; l@xmath10@xmath110@xmath2 and @xmath110@xmath11l@xmath3 respectively ) , both locally @xcite and at high redshift @xcite .
deep x - ray observations performed using the xmm - newton ( e.g. , @xcite ) , chandra @xcite and suzaku @xcite observatories have shown that most ulirgs are intrinsically faint x - ray sources , most likely due to the effects of obscuration , while their x - ray spectra show combined signatures of starburst and agn activity .
the key features observed in the x - ray spectra of ulirgs are a soft thermal component , typically associated with star formation , a heavily - obscured ( n@xmath12@xmath010@xmath9 @xmath6 ) power - law associated with the agn direct emission , and a prominent emission line at @xmath06.4 kev , identified with fluorescence emission from iron in the k@xmath13 ionization level , originating either in the accretion disk or in the surrounding material @xcite .
the presence of heavily - obscured agn among the most extreme ulirgs at @xmath7@xmath141 - 2 has recently been established from deep spitzer observations @xcite .
most of these sources have very high , quasar - like , intrinsic luminosities , and hence most likely do not constitute the bulk of the heavily - obscured agn population @xcite . establishing the fraction of ( u)lirgs
that host a lower luminosity agn is a more challenging task .
recent works based on x - ray stacking @xcite and using 70-@xmath15 m selected sources @xcite report a steep decrease in the fraction of agn with decreasing ir luminosity , going from @xmath0100% at l@xmath16=10@xmath17 l@xmath3 to @xmath1810% at l@xmath16=10@xmath19 l@xmath3 . in the local universe , @xcite found that the incidence of low - luminosity , seyfert - like , agn as a function of stellar mass is more complicated , and is influenced by other parameters .
for example , the dependence of agn fraction on stellar mass can be opposite if galaxy morphology is considered ( increases with decreasing mass in the early - type galaxy population ) . in this work
, we estimate the fraction of heavily - obscured agn in mid - ir - luminous and massive galaxies at high redshift , few of which are individually detected in x - rays . the main goal is to constrain the amount of obscured smbh accretion happening in distant galaxies .
this can be done thanks to the very deep x - ray observations available in the chandra deep fields and the very low and stable chandra background , which allows for the efficient stacking of individually undetected sources . throughout this letter ,
we assume a @xmath20cdm cosmology with @xmath21=0.7 , @xmath22=0.27 and @xmath23=0.73 , in agreement with the most recent cosmological observations @xcite .
by stacking individually - undetected sources selected at longer wavelengths , it is possible to detect very faint x - ray emitters using chandra observations . for example
, this technique was used successfully by @xcite in the chandra deep field north ( cdf - n ) to measure the average x - ray emission from a sample of lyman break galaxies at @xmath7@xmath1424 and by @xcite to detect x - rays from red galaxies at @xmath7@xmath02 .
more recently , samples of heavily - obscured agn candidates selected based on their mid - ir properties have been studied in x - rays via chandra stacking ( e.g. , @xcite ) .
the 4 msec chandra observations of the chandra deep field south ( cdf - s ) , are currently the deepest view of the x - ray sky . in addition , the cdf - s has been observed extensively at many wavelengths . the multiwavelength data available on the ( e)cdf - s were presented by @xcite . very relevant for this work are the deep spitzer observations available in this field , using both the infrared array camera ( irac ) and the multiband imaging photometer for spitzer ( mips ) , from 3.6 to 24 @xmath15 m .
also critical is the availability of good quality photometric redshifts ( @xmath24@xmath7/(1+@xmath7)=0.008 for @[email protected] ) obtained thanks to deep observations in 18 medium - band optical filters performed using subaru / suprime - cam @xcite .
we generated our sample starting with the 4959 spitzer / mips 24 @xmath15 m sources in the region covered by the chandra observations that have photometric redshift @[email protected] , and hence rest - frame e@xmath14 kev emission falling in the high - sensitivity chandra range . in addition , sources individually detected in x - rays and reported in the catalogs of @xcite , @xcite or @xcite were removed from our sample , as these sources will otherwise dominate the stacked spectrum .
we then inspected the remaining sources to eliminate individual detections in the 4 msec data not present in the 2 msec catalog of @xcite .
we further excluded 28 sources that meet the selection criteria of @xcite for heavily obscured agn , @xmath26/@xmath27@xmath11000 and @xmath25-@[email protected] ( vega ) , because we expect these sources to contain an intrinsically luminous agn ( quasar ) , while the aim of this work is to find additional hidden accretion in less luminous objects .
the median redshift of the sources in our final sample is 1.32 ( average @xmath7=1.5 ) with a standard deviation of 0.77 . in order to perform x - ray stacking in the rest - frame
, we started with the regenerated level 2 merged event files created by the chandra x - ray center .
for each source , we extracted all events in a circle of 30@xmath29 radius centered in the optical position .
the energy of each event was then converted to the rest frame using the photometric redshift of the source .
using standard ciao @xcite tools we then generated seven x - ray images for each source covering the energy range from 1 - 8 kev in the rest - frame with a fixed width of 1 kev .
images for individual sources were then co - added to measure the stacked signal .
total counts were measured in a fixed 5@xmath29 aperture , while the background was estimated by randomly placing apertures with same the area , 5@xmath29 to 30@xmath29 away from the center .
several groups have found ( e.g , @xcite and references therein ) that the fraction of galaxies containing an agn is a strong function of their ir luminosity .
hence , it is a natural choice to divide our sample in terms of total ir luminosity .
the infrared luminosity was estimated from the observed 24 @xmath15 m luminosity assuming the relation found by @xcite : @xmath30(l@xmath10)=1.02 + 0.972 @xmath30(l@xmath31 ) .
we further assumed that the @xmath32 correction between observed - frame 24 @xmath15 m and rest - frame 12 @xmath15 m luminosity for these sources is negligible , as shown by @xcite .
we then separated our sample in 4 overlapping bins : @xmath33@xmath110@xmath2@xmath34 , @xmath33@xmath15@xmath3510@xmath19@xmath34 , 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 and @xmath33@xmath110@xmath19@xmath34 and stacked them independently .
the number of sources in each sample is 670 , 1545 , 2342 and 3887 respectively . in fig .
[ obs_spec ] we present the stacked spectra as a function of rest - frame energy , both in total counts and normalized at 1 kev to highlight the difference in spectral shape among the different samples . at e@xmath85 kev ,
the spectra begin to diverge , where we expect the agn emission to dominate even for heavily - obscured sources .
there is a clear trend , with more high energy x - ray emission with increasing ir luminosity .
@xmath110@xmath2@xmath34 ( _ filled circles _ ) , @xmath33@xmath15@xmath3510@xmath19@xmath34 ( _ squares _ ) , 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 ( _ triangles _ ) and @xmath33@xmath110@xmath19@xmath34 ( _ open circles _ ) .
_ right panel _ : same as left panel but normalized at 1 kev in order to highlight the differences in spectral shape among the different samples .
the largest differences are at e@xmath85 kev , where there is a clear trend in the relative intensity as a function of ir luminosity , suggesting a larger fraction of agn in the most luminous ir sources.,title="fig:",scaledwidth=22.0% ] @xmath110@xmath2@xmath34 ( _ filled circles _ ) , @xmath33@xmath15@xmath3510@xmath19@xmath34 ( _ squares _ ) , 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 ( _ triangles _ ) and @xmath33@xmath110@xmath19@xmath34 ( _ open circles _ ) . _ right panel _ : same as left panel but normalized at 1 kev in order to highlight the differences in spectral shape among the different samples .
the largest differences are at e@xmath85 kev , where there is a clear trend in the relative intensity as a function of ir luminosity , suggesting a larger fraction of agn in the most luminous ir sources.,title="fig:",scaledwidth=22.0% ]
the spectra shown in fig .
[ obs_spec ] can not be directly interpreted , as the detector - plus - telescope response information is lost after the conversion to rest - frame energy and stacking .
hence , we perform simulations assuming different intrinsic x - ray spectra in order to constrain the nature of the sources dominating the co - added signal .
we use the xspec code @xcite to convolve several intrinsic input spectra with the latest response functions for the chandra acis - i camera used in the cdf - s observations .
we then compare these simulated spectra with the observations in our sample of ir - selected sources . the low energy spectrum of ( u)lirgs is dominated by a combination of a thermal plasma component with temperatures @[email protected] kev , particularly important at e@xmath183 kev , and the emission from high - mass x - ray binaries ( hmxbs ) at 1@xmath18e ( kev)@xmath1810 ( e.g. , @xcite ) .
for each source , we generated a simulated spectrum using a combination of these two components , taking into account the luminosity and redshift of the source .
for the hmxb population we assumed a power - law given by @xmath37=1.2 and cutoff energy @xmath38=20 kev , consistent with recent observations ( e.g. , @xcite ) .
this component was normalized assuming the relation between ir and x - ray luminosity in starburst galaxies found by @xcite .
for the thermal component , we assumed a black body with temperature @xmath36=0.7 kev .
the normalization of this component was then adjusted to match the observations at @xmath39@xmath183 kev . in order to compute the possible contribution from heavily - obscured agn to the stacked spectrum we assumed the observed x - ray spectrum of the nearby ulirg iras19254 - 7245 , as observed by suzaku @xcite .
in addition to the starburst emission described above , the x - ray spectrum is described by an absorbed , compton - thick , power - law with @xmath37=1.9 , @xmath4=10@xmath9 @xmath6 , and a possible scattered component , characterized by a power - law with @xmath37=1.9 , no absorption , and 1% of the intensity of the direct emission .
the resulting simulated spectral components and the comparison with the observed stacked spectrum for sources with the four samples defined above are shown in fig .
[ simul_spec ] .
it is not possible to explain the observed stacked spectral shape using only a plausible starburst spectrum without invoking an agn component , which dominates at e@xmath15 kev .
the average intrinsic rest - frame 2 - 10 kev agn luminosity needed to explain the observed spectrum , assuming that every source in the sample contains an agn of the same luminosity , is 6@xmath3510@xmath40 erg s@xmath41 for sources with @xmath33@xmath110@xmath2@xmath34 , 3@xmath3510@xmath40 erg s@xmath41 for sources with @xmath33@xmath15@xmath3510@xmath19@xmath34 , 5@xmath3510@xmath42 erg s@xmath41 in the sample with 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 and 7@xmath3510@xmath42 erg s@xmath41 for sources with @xmath33@xmath110@xmath19@xmath34 .
all of these are ( intrinsically ) very low - luminosity agn ; even if there is a range , it is extremely unlikely to include high - luminosity quasars like those discussed in previous stacking papers .
an alternative possibility is that the extra emission at e@xmath15 kev is due entirely to the fe k@xmath43 line , provided the errors in the photometric redshifts in these samples are significantly larger than the values reported by @xcite .
regardless of the template assumed for the agn emission , we obtain similar values for the average agn luminosity in each sample .
the median hard x - ray luminosity for the chandra sources with measured photometric redshifts in the catalog of @xcite is 4.1@xmath3510@xmath44 erg s@xmath41 for the sources in the @xmath33@xmath110@xmath2@xmath34 sample , 3.5@xmath3510@xmath44 erg s@xmath41 in the @xmath33@xmath15@xmath3510@xmath19@xmath34 group , 5.7@xmath3510@xmath40 erg s@xmath41 for sources with 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 and 1.6@xmath3510@xmath44 erg s@xmath41 in the @xmath33@xmath110@xmath19@xmath34 sample .
hence , if the heavily - obscured agn in our stacked samples have the same median intrinsic luminosity this would indicate that 15% ( 98 sources ) of the 670 galaxies with @xmath33@xmath110@xmath2@xmath34 contain a heavily - obscured agn .
this fraction is @xmath09% ( 132 and 205 sources respectively ) in the @xmath33@xmath15@xmath3510@xmath19@xmath34 and 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 samples . for sources with @xmath33@xmath110@xmath19@xmath34
this fraction is @xmath185% .
the integrated intrinsic x - ray emission in the rest - frame 2 - 10 kev band due to the heavily - obscured agn in this sample , obtained by multiplying the intrinsic x - ray luminosity by the number of sources and dividing by the studied area , is @xmath04.6@xmath3510@xmath45 erg @xmath6 s@xmath41 deg@xmath46 . for comparison ,
the total emission from all the x - ray detected agn in the cdf - s is 1.63@xmath3510@xmath47 erg @xmath6 s@xmath41 deg@xmath46 .
hence , this extra agn activity can account for @xmath022% of the total smbh accretion . adding this to the obscured smbh growth in x - ray detected agn @xcite
, we confirm that most smbh growth , @xmath070% , is significantly obscured and missed by even the deepest x - ray surveys @xcite . performing a similar study on the 28 sources with @xmath26/@xmath27@xmath11000 and @xmath25-@[email protected] that we previously excluded , we find a very hard x - ray spectrum , harder than that of the @xmath33@xmath110@xmath2@xmath34 sources .
this spectrum is consistent with a population of luminous agn with intrinsic rest - frame 2 - 10 kev luminosity @xmath02@xmath3510@xmath44 erg s@xmath41 and negligible contribution from the host galaxy , except at e@xmath182 kev where the thermal component is @xmath030% of the total emission .
this result justifies our choice of removing these sources from our study ( otherwise they would dominate the stacked signal ) , while at the same time it confirms the agn nature of the vast majority of these sources , in contrast to the suggestion that the extra ir emission could be due to star - formation processes @xcite .
a similar result for these high - luminosity sources was found by @xcite : in a sample of 99 mid - ir excess sources in the cosmos field he found a strong stacked signal at e@xmath06 kev , which he interpreted as due to the fe k@xmath43 line , a clear signature of agn emission and high obscuration ( see discussion below ) . by design
, none of the sources in our sample are individually detected in x - rays , nor do they satisfy the selection criteria of @xcite .
however , it is interesting to investigate if they present other agn signatures .
for example , 237 out of the 1545 sources with @xmath33@xmath15@xmath3510@xmath19@xmath34 in our sample ( 15% ) are found inside the agn irac color - color region defined by @xcite .
for comparison , in the sample of 2342 sources with 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 in which from the stacked hard x - ray signal we determined a negligible agn fraction there are 327 galaxies ( 14% ) in the @xcite region .
this suggests that the irac color - color diagram can not be used to identify heavily - obscured low - luminosity agn , because the near - ir emission in these sources is dominated by the host galaxy @xcite . at longer wavelengths ,
83 of the 1545 sources with @xmath33@xmath15@xmath3510@xmath19@xmath34 were detected in the deep vla observations of the cdf - s @xcite .
in contrast , only 33 sources in the 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 sample were detected in these observations . using the @xmath48 ratio between 1.4 ghz and 24 @xmath15 m flux densities ( e.g. , @xcite )
, we find that in the @xmath33@xmath15@xmath3510@xmath19@xmath34 sample , only 14 sources have @xmath48@xmath18 - 0.23 and can be considered `` radio - loud '' @xcite , and in the 5@xmath3510@xmath19@xmath34@xmath1@xmath33@xmath110@xmath19@xmath34 sample only 10 sources have @xmath48@xmath18 - 0.23 .
hence , we conclude that the fraction of bona fide radio - loud sources is negligible and that in most cases the radio emission is produced by star - formation processes . in order to investigate the fraction of heavily - obscured agn as a function of other galaxy parameters , we performed x - ray stacking of samples sorted by stellar mass .
stellar masses were taken from @xcite , who performed spectral fitting to the extensive optical and near - ir spectro - photometry using fast @xcite and the stellar templates of @xcite assuming the @xcite initial mass function and solar metallicity .
we further restricted our sample to sources with @[email protected] , for which photometric redshifts and stellar masses are very well determined ( @xmath24z/(1+@xmath7)=0.007 ) .
we then divided the sample into three mass bins : m@xmath110@xmath2m@xmath3 , 10@xmath2@xmath1 m ( m@xmath3)@xmath110@xmath19 and 10@xmath19@xmath1 m ( m@xmath3)@xmath110@xmath49 .
the resulting stacked x - ray spectra are shown in fig .
[ obs_spec_mass ] .
10@xmath2m@xmath3 ( _ red squares _ ) , 10@xmath19@xmath18m@xmath1810@xmath2m@xmath3 ( _ blue triangles _ ) and 10@xmath49@xmath18m@xmath1810@xmath19m@xmath3 ( _ black circles _ ) .
_ right panel : _ same but normalized at 1 kev . in the m@xmath110@xmath2m@xmath3 sample ,
the strong excess at e=6 - 7 kev , which we associate with the fe k@xmath43 line , is an indicator of agn activity .
similarly , for the sources with 10@xmath19@xmath18m@xmath1810@xmath2m@xmath3 there is a hard x - ray spectrum , also suggesting a significant agn fraction .
these preliminary results indicate that these heavily - obscured moderate - luminosity agn are predominantly present in the most massive galaxies.,title="fig : " ] 10@xmath2m@xmath3 ( _ red squares _ ) , 10@xmath19@xmath18m@xmath1810@xmath2m@xmath3 ( _ blue triangles _ ) and 10@xmath49@xmath18m@xmath1810@xmath19m@xmath3 ( _ black circles _ ) . _
right panel : _ same but normalized at 1 kev . in the m@xmath110@xmath2m@xmath3 sample ,
the strong excess at e=6 - 7 kev , which we associate with the fe k@xmath43 line , is an indicator of agn activity .
similarly , for the sources with 10@xmath19@xmath18m@xmath1810@xmath2m@xmath3 there is a hard x - ray spectrum , also suggesting a significant agn fraction .
these preliminary results indicate that these heavily - obscured moderate - luminosity agn are predominantly present in the most massive galaxies.,title="fig : " ] for sources with m@xmath110@xmath2m@xmath3 , there is a significant excess at 6 - 7 kev , above a spectrum that otherwise declines with increasing energy
. this might be due to the presence of the fe k@xmath43 line , a clear indicator of agn activity .
contrary to the case of stacking as a function of ir luminosity ( fig .
[ simul_spec ] ) , here we do not find evidence for an absorbed power - law the 6 - 7 kev feature is simply too sharply peaked . possibly the restriction to @[email protected] for the mass - binned stacking , where photometric redshifts are most accurate , reveals an emission line that is broadened by less accurate photometric redshifts in the full sample .
that is , the feature in the @xmath33-binned stack that we interpreted as a heavily absorbed power law may instead be an fe k@xmath43 line broadened artificially by bad photometric redshifts . in the 10@xmath2@xmath1 m ( m@xmath3)@xmath110@xmath19 sample we found a significant hardening of the x - ray spectrum ( fig .
[ obs_spec_mass ] ) , suggesting the presence of a significant fraction of agn .
in contrast , only a soft spectrum , consistent with star - formation emission , can be seen for sources with 10@xmath19@xmath1 m ( m@xmath3)@xmath110@xmath49 .
taken together , these results indicate that agn are predominantly present in the most massive galaxies , in agreement with the conclusions of @xcite and others .
this will be elaborated in a paper currently in preparation .
the fraction of compton - thick agn in the local universe is still heavily debated .
@xcite reported a fraction of @xmath08% in a flux - limited sample of sources detected in the swift / bat all - sky @xcite and international gamma - ray astrophysics laboratory ( integral ; @xcite ) surveys . from an integral volume - limited survey at @[email protected]
, @xcite found a higher fraction of 24% , suggesting that even surveys at e@xmath110 kev are potentially biased against the detection of compton - thick agn .
the fraction of moderate - luminosity compton - thick sources in our sample of sources with @xmath33@xmath15@xmath3510@xmath19@xmath34 , relative to all agn in the cdf - s , is @xmath025% ( 132/525 ) , assuming that compton - thick and compton - thin agn have similar median intrinsic luminosities .
this indicates that there is no major evolution in the number of moderate - luminosity heavily - obscured agn from @xmath7=0 to 2 .
in contrast , at higher luminosities , @xcite reported that the ratio of obscured to unobscured quasars increased from @xmath01 at @xmath7=0 to @xmath02 - 3 at @xmath7@xmath142 .
hence , although all these estimates are still uncertain , it appears that the evolution of compton - thick agn depends strongly on their luminosity .
we further speculate that this is indication that the triggering of low - luminosity agn is not related to the major merger of gas - rich galaxies as found by @xcite for high - luminosity quasars or that the time delay between galaxy interactions and black hole growth is long @xcite .
we thank the referee , fabrizio fiore , for very useful and constructive comments .
support for the work of et and ks was provided by the national aeronautics and space administration through chandra / einstein post - doctoral fellowship award numbers pf8 - 90055 and pf9 - 00069 respectively issued by the chandra x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 .
cmu and cc acknowledge support from nsf grants ast-0407295 , ast-0449678 , ast-0807570 and yale university . , v. , della ceca , r. , piconcelli , e. , severgnini , p. , bassani , l. , cappi , m. , franceschini , a. , iwasawa , k. , malaguti , g. , marziani , p. , palumbo , g. g. c. , persic , m. , risaliti , g. , & salvati , m. 2004 , , 420 , 79 , w. n. , alexander , d. m. , hornschemeier , a. e. , garmire , g. p. , schneider , d. p. , barger , a. j. , bauer , f. e. , broos , p. s. , cowie , l. l. , townsley , l. k. , burrows , d. n. , chartas , g. , feigelson , e. d. , griffiths , r. e. , nousek , j. a. , & sargent , w. l. w. 2001 , , 122 , 2810 , d. , eisenhardt , p. , gorjian , v. , kochanek , c. s. , caldwell , n. , eisenstein , d. , brodwin , m. , brown , m. j. i. , cool , r. , dey , a. , green , p. , jannuzi , b. t. , murray , s. s. , pahre , m. a. , & willner , s. p. 2005
, , 631 , 163 | we take advantage of the rich multi - wavelength data available in the chandra deep field south ( cdf - s ) , including the 4 msec chandra observations ( the deepest x - ray data to date ) , in order to search for heavily - obscured low - luminosity agn among infrared - luminous galaxies .
in particular , we obtained a stacked rest - frame x - ray spectrum for samples of galaxies binned in terms of their ir luminosity or stellar mass .
we detect a significant signal at e@xmath01 to 8 kev , which we interpret as originating from a combination of emission associated with star - formation processes at low energies combined with a heavily - obscured agn at e@xmath15 kev .
we further find that the relative strength of this agn signal decays with decreasing ir luminosity , indicating a higher agn fraction for more luminous ir sources .
together , these results strongly suggest the presence of a large number of obscured agn in ir - luminous galaxies . using samples binned in terms of stellar mass in the host galaxy , we find a significant excess at e=6 - 7 kev for sources with m@xmath110@xmath2m@xmath3 , consistent with a large obscured agn population in high mass galaxies .
in contrast , no strong evidence of agn activity was found for less - massive galaxies .
the integrated intensity at high energies indicates that a significant fraction of the total black hole growth , @xmath022% , occurs in heavily - obscured systems that are not individually detected in even the deepest x - ray observations .
there are also indications that the number of low - luminosity , heavily - obscured agn does not evolve significantly with redshift , in contrast to the strong evolution seen in higher luminosity sources . |
the density matrix approach was introduced to describe statistical concepts in quantum mechanics by landau @xcite , dirac @xcite , and von neumann @xcite . in several branches of physics like polarized spin assemblies or qudit systems , and cavity electrodynamics
the density matrix approach can be cast into a su(d ) description @xcite .
the bloch vector parametrization was used to describe the 2-level problem which later on was generalized to describe beams of particles with spin @xmath0 in terms of what are known as fano statistical tensors @xcite . in particular @xmath1 projectors defining the generators of a unitary algebra
have been introduced in @xcite to expand a density matrix of spin systems , even more , they established a procedure to reconstruct the density matrix by a finite number of magnetic dipole measurements with stern - gerlach analyzers and concluded that it was necessary to do at least @xmath2 measurements to reconstruct the density matrix of pure states while @xmath3 were required for mixed states @xcite . an experimental reconstruction of a cavity state for @xmath4 using this method is given in @xcite .
another approach uses the moore - penrose pseudoinverse to express the elements of the spin density matrix in terms of @xmath5 probabilities of spin projections @xcite . a method to reconstruct any pure state of spin in terms of coherent states
is provided in @xcite and by means of non orthogonal projectors on coherent states a reconstruction of mixed states can be done @xcite . a parametrization based on cholesky factorization @xcite
was first used to guarantee the positivity of the spin density matrices in @xcite , and more recently , a tomographic approach to reconstruct them @xcite . in the last twenty years
, a lot of work related with parametrization of the density matrices of @xmath6-level quantum systems has been done @xcite .
this is due to its applications to quantum computation and quantum information systems @xcite .
the decomposition of the density matrix into a symmetrized polynomial in lie algebra generators has been determined in @xcite . a novel tensorial representation for density matrices of spin states , based on weinberg s covariant matrices , may be another important generalization of the bloch sphere representation @xcite .
actually , there are several parametrizations of finite density matrices : generalizations of the bloch vector @xcite , the canonical coset decomposition of unitary matrices @xcite , the recursive procedures to describe @xmath7 unitary matrices in terms of those of @xmath8@xcite , by factorizing @xmath7 unitary matrices in terms of points on complex spheres @xcite , and by defining generalized euler angles @xcite . even in the case of composite systems
there are parametrizations of finite density matrices @xcite .
recently we have established a procedure to determine the extremal density matrices of a qudit system associated to the expectation value of any observable @xcite .
these matrices provide an extremal description of the mean values of the energy , and in the case of restricting them to pure states the energy spectrum is recovered .
so , apart from being an alternative tool to find the eigensystem one has information of mixed states which minimize its mean value .
the aim of this work is to give another option to compute extremal density matrices in a qudit space by means of an algebraic approach that leads to an underdetermined linear system in terms of the components of the bloch vector @xmath9 , the antisymmetric structure constants @xmath10 of a @xmath11 algebra , and the parameters of the hamiltonian operator @xmath12 .
their solution , in general , implies to get the bloch vector in terms of a known number of free components .
these are determined by establishing a system of equations associated to the characteristic polynomial of the density matrix .
finally , one arrives to the extremal density matrices of the expectation value of the hamiltonian , which for the pure case let us obtain the corresponding full spectrum or for the mixed case at most @xmath13 extremal mean value energies .
another goal is to bring and join different algebraic tools in the study of the behaviour of both the density matrix and hermitian operators .
any hermitian hilbert - schmidt operator acting on the @xmath6-dimensional hilbert space can be expressed in terms of the identity operator plus a set of hermitian traceless operators @xmath14 which are the generators of the @xmath11 algebra . in this basis , the hamiltonian operator @xmath15 and the density matrix @xmath16 are written as @xcite @xmath17 with the definitions @xmath18 and @xmath19 .
these generators are completely characterized by means of their commutation and anticommutation relations given by @xmath20 & = & 2 \ , i \sum_{q=1}^{d^2 - 1 } f_{jkq } \hat{\lambda}_{q } \ , , \label{s1:eq3}\\ \{\hat{\lambda}_{j},\hat{\lambda}_{k}\ } & = & \frac{4}{d}\delta_{jk}\hat{i } + 2 \sum_{q=1}^{d^2 - 1 } d_{jkq } \hat{\lambda}_{q } \ , , \label{s1:eq4}\end{aligned}\ ] ] where @xmath21 and @xmath22 are the symmetric and antisymmetric structure constants @xmath23 \hat{\lambda_q } ) \ , , \label{s1:eq6}\end{aligned}\ ] ] and consequently , it follows the multiplication law @xcite @xmath24 a realization of the generators can be given by the generalized gell - mann matrices @xcite , consisting in @xmath25 symmetric matrices @xmath26 plus @xmath27 antisymmetric matrices @xmath28 and @xmath29 diagonal ones @xmath30 where @xmath31 and @xmath32 are matrices with @xmath33 in the component @xmath34 and @xmath35 otherwise .
this type of realization belongs to the so called generalized bloch vector parametrization @xcite .
the fano statistical tensors @xcite , the multipole moments @xcite , the weyl matrices @xcite , and the generalized gell - mann matrices @xcite , belong to this group .
therefore a vector with @xmath36 real components define the so called generalized bloch vector @xcite , @xmath37 whose magnitude is bounded by @xcite @xmath38 where the equality specifies a necessary condition to represent a pure state .
in general , a @xmath39 unitary transformation acting on a hermitian matrix implies a rotation in its components , i.e. , @xmath40 where in the last equality , one has defined @xmath41 and @xmath42 are elements of an orthogonal matrix that belongs to the @xmath43 group , which provides the adjoint representation of @xmath39 @xcite .
the density matrix must satisfy the following three properties : ( a ) it is hermitian , ( b ) it has trace one , and ( c ) all its eigenvalues are positive semidefinite . while for dimension @xmath44 , the condition @xmath45 implies ( c ) , for @xmath46 that is not true .
the positivity conditions of the density matrix are established by the set @xmath47 of coefficients of its corresponding characteristic polynomial .
this set can be obtained by means of the recursive relation known as newton - girard formulas @xcite @xmath48 with the definitions @xmath49 , @xmath50 , and @xmath51 , for @xmath52 .
therefore the allowed density matrix must satisfy the following system of @xmath53 simultaneous polynomial equations @xmath54 where the constants @xmath55 fix the degree of mixing of the system .
thus , they must be in the region given by @xcite @xmath56 where @xmath57 denotes a binomial coefficient .
the upper bound defines the most mixed state and then it has maximum entropy , while the lower bound specifies pure states which have zero entropy . additionally , the @xmath58 for @xmath59 @xcite .
all of them are polynomial functions in terms of the invariants of the density matrix , i.e. , @xmath60 , for @xmath52 . in terms of @xmath61
, it is defined the symmetric matrix called bezoutian @xcite @xmath62 a polynomial with real coefficients has reals roots iff the bezoutian matrix is positive definite @xcite .
hence , the compatible region among the invariants is obtained with the intersection of the positivity conditions of the density matrix from ( [ s2:eq1 ] ) with the respective positivity conditions of the bezoutian ( see details in appendix [ b ] ) .
the rayleigh quotient @xmath63 of a hermitian hilbert - schmidt operator @xmath64 is @xmath65 where @xmath66 is a @xmath6-dimensional complex vector .
since the rayleigh quotient is invariant under scale transformations , in searching its maximum or minimum it suffices to confine the search on unit norm vectors , i.e. , when @xmath67 @xcite .
this leads to define the numerical range @xmath68 , which is the set of all possible rayleigh quotients @xmath69 over the unit vectors : @xmath70 the numerical range @xmath68 is a closed interval on the real axis , whose end points are the extreme eigenvalues of @xmath64 @xcite .
this result is a particular case of the courant - fischer theorem @xcite , which states that every eigenpair ( eigenvalue and eigenvector ) of @xmath64 is the solution of a optimization ( max - min problem ) of @xmath68 in some subspace of @xmath64 .
therefore , eigenvectors and eigenvalues of @xmath64 are the critical points and critical values , respectively , of the rayleigh quotient and @xmath68 is the convex hull of its eigenvalues . in the density matrix formalism , the numerical range of the hamiltonian ( or any hermitian operator ) can be identified with its mean value in an arbitrary state @xmath71 , i.e. , @xmath72 @xcite . in this scheme ,
a useful theorem is the following one .
* theorem 1 * @xcite .
_ let @xmath73 and @xmath71 be @xmath74 hermitian matrices with their eigenvalues @xmath75 and @xmath76 , respectively , arranged in descending order , v.g . , @xmath77 and @xmath78 .
thus , one has the inequality @xmath79 if either inequality is an equality , then @xmath15 and @xmath16 commute .
_ since the equality is easy to verify when @xmath15 and @xmath16 are diagonals ( diagonal frame ) , this theorem leads to the assumption that the density matrix can be adapted to get the spectrum of @xmath15 if they both commute . in that way ,
a related fact is the following
. * proposition 1 . * _ for an arbitrary @xmath80 commuting with @xmath15 , in any frame @xmath81 depends on at most @xmath53 variables parametrizing @xmath80 . _
_ proof_. since @xmath80 and @xmath15 commute , they are simultaneously diagonalizable therefore , in the diagonal frame , @xmath82 where we use the convention @xmath83 and @xmath84 . by applying the relation ( [ s1:eq13 ] )
one has @xmath85 and @xmath86 with @xmath87 as the orthogonal matrix from ( [ s1:eq14 ] ) . setting aside scalar matrices , since orthogonal transformations
preserve the dot product , @xmath88 is an invariant quantity and depends on at most @xmath53 variables of @xmath80 . _ q.e.d .
_ with the aim to provide further properties of the set of density matrices that commute with @xmath89 and propose an algorithm to compute the extremal mean values of the energy in the density matrix formalism , we consider from here on @xmath15 as a given non - scalar matrix and establish the proposition : * proposition 2 .
* _ let @xmath15 and @xmath71 be two finite @xmath90 hermitian matrices where @xmath71 represents an arbitrary density matrix .
for a fixed degree of mixture , the critical points of @xmath91 determine the extremal density matrices @xmath92 commuting with @xmath73 and @xmath93 , with @xmath94 .
_ proof_. suppose that @xmath16 is unitarily related to a density matrix @xmath80 which commute with @xmath15 then , @xmath95 , where @xmath96 define a @xmath39 unitary transformation .
therefore , with the mean value @xmath97 one can define the scalar function @xmath98 where the real constants @xmath99 and @xmath100 are the components of the expansion of @xmath15 and @xmath101 respectively , in a basis for the hilbert space of hermitian operators . otherwise , if @xmath102 is sufficiently close to the identity , by considering @xmath103 as infinitesimal parameters one can make a taylor series expansion of the function as follows @xmath104 where we have defined @xmath105 as the @xmath39 unitary transformation is infinitesimal , one has that @xmath106 + \frac{i^2}{2 } \sum_{q , p=1}^{d^2 - 1 } \theta_q \theta_p [ \hat{\lambda}_q \,,\ , [ \hat{\lambda}_p \,,\ , \hat{\rho}^c ] ] \ , .
\label{s3:eq12 } \end{aligned}\ ] ] substituting the last expression into ( [ s3:eq8 ] ) , comparing with ( [ s3:eq9 ] ) and by applying the cyclic property of the trace , eqs .
( [ s3:eq10 ] ) and ( [ s3:eq11 ] ) lead to @xmath107 \ , \hat{\lambda}_q \right ) \
, , \label{s3:eq13 } \\
\epsilon_{p , q } & = & i^2 { \rm tr } \left ( [ \hat{\lambda}_p \,,\ , \hat{\rho}^c ] \ , [ \hat{\lambda}_q \,,\ , \hat{h } ] \right ) \ , .
\label{s3:eq14}\end{aligned}\ ] ] the algebraic system which determines the critical points is given by equating eq .
( [ s3:eq13 ] ) to zero , for @xmath108 , whereby @xmath109 = \boldsymbol{0}$ ] .
hence , @xmath110 achieves its extreme values at @xmath80 . in that sense , any density matrix which commutes with @xmath15 and optimizes its mean value , is extremal .
even though the commutativity is satisfied by hypothesis , it implies that any state @xmath16 can be approximated at first - order by @xmath101 and @xmath110 has an error that vanishes to the second - order in @xmath111 , i.e. , @xmath112 thus , by means of the proposition @xmath33 , @xmath101 depends in general on @xmath53 variables that are fixed by establishing a degree of mixture through the expressions ( [ s2:eq3 ] ) .
finally , the highest degree of the polynomial @xmath113 in ( [ s2:eq3 ] ) is @xmath114 .
then , by bezout s theorem , the number of solutions for the polynomial system ( known as bezout bound or bezout number ) is at most the product of the degree of all the equations , i.e. , @xmath115 @xcite .
all this implies that the single critical matrix @xmath80 represents at most @xmath116 different critical density matrices @xmath92 , where @xmath117 .
_ by matching results , in the non degenerate case of @xmath118 , if all @xmath92 are pure states , they must correspond to one - dimensional eigenprojectors of @xmath118 .
in a previous work @xcite we proposed an approach to obtain information of the energy spectrum of a hamiltonian by considering its mean value together with @xmath53 constraints to guarantee the positivity of the density matrix .
this is achieved by defining the function @xmath119 which depends on @xmath120 real parameters @xmath121 and @xmath36 independent variables @xmath122 associated to the expansions ( [ s1:eq1 ] ) and ( [ s1:eq2 ] ) , respectively
. additionally , there are @xmath53 lagrange multipliers @xmath123 and @xmath53 positive real constants @xmath124 to fix the degree of purity of the density matrix ( see the bound ( [ s2:eq1 ] ) ) .
one can note that @xmath125 is a continuous function because is the sum of the rayleigh quotient @xmath126 , where @xmath127 with @xmath128 , and the positivity constraints which are polynomials in the variables of the density matrix .
therefore , in order to reach all the eigenvalues of @xmath89 and its numerical range , one must find the min - max sets of @xmath125 with respect to the variables @xmath122 and the lagrange multipliers @xmath123 .
their respective derivatives give @xmath129 algebraic equations , @xmath130 these sets of algebraic equations determine the extremal values of the density matrix , i.e. , @xmath131 and @xmath132 for which the expressions ( [ s3:eq6 ] ) and ( [ s3:eq7 ] ) are satisfied . by substituting @xmath133 into equation ( [ s1:eq2 ] )
one obtains the extremal density matrices .
if we restrict the solutions to pure states @xmath134 , we have shown explicitly that the energy spectrum of the hamiltonian is recovered for @xmath135 and @xmath136 @xcite .
extremal expressions for the mean value of the hamiltonian can be obtained with density matrices representing mixed quantum states , which determine also the corresponding mixture of eigenstates of the hamiltonian . from here on , we describe an alternative algebraic procedure to get the extremal density matrices which is simpler than the one mentioned above .
first of all , notice that propositions @xmath33 and @xmath137 in section [ ray ] are based on the assumption of a common basis , which it is always possible to find if @xmath138 and @xmath15 commute .
thus , in the following paragraphs , we propose for the pure case ( or mixed case ) a systematic approach to get information about the hamiltonian spectrum ( or mean value of the hamiltonian ) , i.e. , its numerical range ( interval of extremal mean values of @xmath15 ) , without making use of a diagonalization procedure .
first we replace into the commutator @xmath139=0 $ ] the expressions eqs .
( [ s1:eq1 ] ) and ( [ s1:eq2 ] ) and use the properties of the generators @xmath140 of the @xmath11 algebra . then the expression ( [ s3:eq13 ] ) gives rise to the @xmath36 dimensional homogeneous system of equations @xmath141 that determines the critical points and where @xmath142 is the bloch vector defined in ( [ s1:eq10 ] ) .
the matrix elements of the skew symmetric matrix @xmath143 of @xmath36 dimensions are given by @xmath144 where @xmath145 are the antisymmetric structure constants of the @xmath11 algebra .
a single solution of the homogeneous system ( [ s3:eq15 ] ) can be obtained through the gauss - jordan elimination method and it is identified as the critical bloch vector @xmath146 , with its @xmath147 free variables equal to the dimension of the null space of @xmath148 .
this implies that maximal mixed states are always critical for any observable because the null space always contains the zero vector .
on the other hand , notice that @xmath149 $ ] are hermitian vectors spanning the tangent space of the orbits associated with @xmath15 , with @xmath150 .
then by substituting the hamiltonian expression ( [ s1:eq1 ] ) one gets @xmath151 note that these vectors give the rows of @xmath148 , and the number of independent vectors @xmath152 is determined by the rank of the gram matrix @xmath153 therefore @xmath152 determines the dimension of the tangent space of the hamiltonian orbits and the rank of @xmath148 , i.e. , @xmath154 @xcite .
furthermore , the maximal dimension of the orbit occurs when the hamiltonian is non degenerate , i.e. , when @xmath155 and by comparing @xmath36 with @xmath152 one has that the system ( [ s3:eq15 ] ) is always underdetermined with @xmath156 free variables ( see table [ tab ] ) . to clarify the method
, we are going to discuss the non degenerate and degenerate cases of @xmath15 separately . in section [ examples ] we shall illustrate the method for quantum systems of dimensions @xmath157 and @xmath158 .
in this case , the rank of the matrix @xmath159 is given by @xmath160 and the @xmath161 free variables reach its minimum number , i.e. , @xmath162 .
therefore , @xmath146 is given by @xmath163 where the @xmath164 components @xmath165 are functions of the parameters of the hamiltonian , the antisymmetric structure constants and a set of @xmath53 independent free variables .
it is natural to choose this set from the diagonal generators ( [ s1:eq9 ] ) of @xmath11 .
the substitution of @xmath146 in ( [ s1:eq2 ] ) gives its associated critical density matrix denoted as @xmath80 .
therefore , the determination of the @xmath53 free variables is done by solving the system of @xmath53 polynomial equations ( [ s2:eq3 ] ) , which by proposition 2 has at most @xmath116 different solutions .
this is also in agreement with the analysis in the diagonal representation @xmath166 , wherein the action of the permutation group of @xmath161 elements produces @xmath167 matrices @xcite .
they satisfy the same polynomial system ( [ s2:eq3 ] ) but give different mean values of @xmath15 . therefore , the critical density matrices are given by @xmath168 with @xmath169 denoting the bloch vector solution , the variables @xmath170 are function only of the known quantities , i.e. , the structure constants , the parameters of the hamiltonian @xmath171 , and @xmath53 constants @xmath172 .
the number @xmath169 of solutions decreases up to @xmath6 when ( [ s3:eq18 ] ) represent pure states ( all @xmath173 ) and the extremal density matrices are one - dimensional orthogonal projectors . as to be expected
, the expectation value of the hamiltonian is in general given by @xmath174 for each critical ( or extremal ) density matrix @xmath92 , with @xmath175 .
for the pure case the expectation values yield the energy spectrum of the system and the extremal density matrices are orthogonal projectors .
in this case the rank of the matrix @xmath159 satisfies that @xmath176 , whose value is associated to the orbits of the hamiltonian ( see table [ tab ] ) . as the critical bloch vector @xmath146 has @xmath177 free variables with @xmath178 one then has that @xmath179 .
thus , if @xmath180 denotes the number of variables appearing in the expression for the mean value of the hamiltonian in the state @xmath80 , @xmath181 , there are two cases to consider : * when @xmath182 , one has to select @xmath183 components of the density matrix bloch vector to have @xmath53 free variables and then solve the polynomial system of equations ( [ s2:eq3 ] ) .
* when @xmath184 , one has only to pick up @xmath53 components of the density matrix bloch vector from the set of @xmath180 elements , and again to solve the mentioned polynomial equation . in both cases
@xmath53 free variables will be determined by the polynomial system ( [ s2:eq3 ] ) .
the remaining @xmath185 components can be taken equal to zero because they do not affect the commutator of @xmath15 with @xmath80 .
specifically , if we are interested in the eigensystem , i.e. , all the set of @xmath186 , one can apply the following method recursively : * make zero the @xmath187 components , to solve the polynomial system ( [ s2:eq3 ] ) , whose solution give at least two extremal density matrices . *
take the trace of the first set of solutions @xmath188 with the commuting general density matrix @xmath80 , this yields a system of algebraic equations by asking the orthogonality conditions , i.e. , @xmath189 , where @xmath114 is a label counting the number of solutions .
* substitute the solutions of the linear system , to express the new critical density matrix in terms of the free variables , where @xmath53 are fixed by means of the positivity conditions ( all @xmath186 ) and the rest of the components can be taken equal to zero . *
return to step 1 and repeat the procedure again , until one gets @xmath6 orthogonal projectors . of course
, one has in this case several solutions related with the degeneracy of the hamiltonian in similar form as in the standard diagonalization procedure of a finite hamiltonian matrix .
although this may seem arbitrary , ultimately it is related to the codimension conditions @xcite .
this topic will be addressed in a future contribution .
similarly to ( [ s3:eq19 ] ) , in all cases , the energy spectrum is given by @xmath190 for each critical density matrix @xmath92 , with @xmath191 .
for @xmath44 , one has that the generators @xmath193 can be realized in terms of the pauli matrices , i.e. , @xmath194 , @xmath195 and @xmath196 .
therefore the density and hamiltonian matrices can be written in terms of the bloch vectors @xmath197 and @xmath198 , @xmath199 where @xmath200 is also called the polarization vector .
thus by substituting the expressions ( [ s5:eq1 ] ) into ( [ s3:eq13 ] ) we obtain the condition that the bloch vectors of @xmath201 and the density matrix are parallel , @xmath202 its solution gives the critical bloch vector @xmath203 with a free variable @xmath204 , according with the dimensions of the orbits of the hamiltonian ( see table [ tab ] ) . from expressions ( [ s2:eq3 ] )
, one has a single positivity condition @xmath205 by substituting ( [ s5:eq5 ] ) into the above equation , we obtain @xmath206 where we define @xmath207 and @xmath208 . by means of ( [ s5:eq5 ] ) and ( [ s5:eq7 ] ) , we find the critical density matrices @xmath209 which correspond exactly to the solutions given in @xcite .
note that @xmath210 .
substituting them into ( [ s3:eq19 ] ) we get @xmath211 we can distinguish two types of solutions : * pure case ( when @xmath212 ) : one has @xmath213 , the eigenvalues @xmath214 of the hamiltonian , and from ( [ s5:eq8 ] ) , with @xmath213 , the corresponding orthogonal projectors . * mixed case ( when @xmath215 ) : the extremal density matrices for the expectation value of the hamiltonian are given in terms of the convex sum @xmath216 where @xmath217 indicates the probability of finding the system with eigenvalue @xmath218 while @xmath219 the corresponding probability of finding an energy @xmath220 .
for the qutrit case , the generators @xmath221 , with @xmath222 can be realized in terms of the gell - mann matrices @xcite .
thus , an arbitrary density matrix is given by @xmath223 while the matrix ( [ s3:eq16 ] ) takes the form @xmath224 which is a real skew symmetric matrix , and to simplify the matrix notation we define @xmath225 and @xmath226 .
we consider the following hamiltonian matrix @xmath227 where the parameters @xmath228 and @xmath229 are real parameters .
it represents a hamiltonian written in terms of the angular momentum @xmath230 with @xmath231 .
this hamiltonian has been used to describe a two mode bose - einstein condensate where the parameter @xmath228 represents the atom - atom interaction , and @xmath229 is related with the tunnelling parameter or a symmetric system of two interacting qubits @xcite .
the bloch vector for the hamiltonian is given by @xmath232 substituting the components of @xmath233 into ( [ s5:eq13 ] ) , one finds that the rank of @xmath143 is @xmath234 , implying that the hamiltonian is non degenerate . solving the system of equations ( [ s3:eq15 ] ) with the gauss - jordan elimination method , one obtains the extremal bloch vector for the density matrix @xmath235 thus the associated critical density matrix
is found by replacing the components of @xmath146 into ( [ s5:eq12 ] ) , which is denoted by @xmath236 .
for this case , to guarantee the positivity of the density matrix , one must consider @xmath237 newly one has two types of solutions : * pure case ( taking @xmath238 ) : solving the system of equations @xmath212 and @xmath239 , one arrives to @xmath136 different solutions for @xmath240 and @xmath241 , denoted by @xmath242 which yield three independent bloch vectors of the density matrix , namely @xmath243 whose norm is equal to @xmath244 and the scalar products between them are equal to @xmath245 .
thus it is straightforward to check that the corresponding extremal density matrices are orthogonal projectors associated to the energy eigenvalues of the hamiltonian @xmath246 , @xmath247 . * mixed case : for any other values for @xmath248 and @xmath249 in the region shown in fig .
[ region](a ) , one can solve the polynomial system given by ( [ s5:eq23 ] ) .
as an example we take @xmath250 and @xmath251 .
there are @xmath234 different solutions for @xmath240 and @xmath241 which give rise to @xmath234 bloch vectors , @xmath252 whose corresponding extremal expectation values of the hamiltonian are given by @xmath253 we find the expansion of the extremal density matrices for the mixed case in terms of the pure case described before , @xmath254 note that the expressions ( [ mixto3 ] ) can be checked by calculating the expectation value of the hamiltonian with the expansions given in the last expression .
\(a ) ( b ) now we consider the hamiltonian matrix given by @xmath255 in this case the bloch vector characterising the hamiltonian is given by @xmath256 with @xmath257
. replacing this values into the matrix ( [ s5:eq13 ] ) , the rank of @xmath258 is @xmath259 , which according to table [ tab ] the hamiltonian exhibits a double degeneracy .
thus , if @xmath260 the diagonal representation is @xmath261 , or in opposite way , if @xmath262 then @xmath263 . by applying the gauss - jordan method to ( [ s3:eq15 ] ) ,
it yields @xmath264 hence , the corresponding critical bloch vector ( [ s3:eq17 ] ) is given by @xmath265 with @xmath158 free parameters and its associated critical density matrix is denoted as @xmath266 .
now , in order to obtain the eigensystem of @xmath15 , we are going to use the procedure established before for the degenerated case : * thus we select the components @xmath267 , solve the polynomial condition ( [ s5:eq23 ] ) with @xmath238 , and we get the following bloch vectors of the density matrix @xmath268 these bloch vectors yield two density matrices @xmath269 , and @xmath270 which are not independent , both by taking the trace with the hamiltonian give an energy eigenvalue @xmath271 .
* we establish the algebraic system of equations , @xmath272 whose solution together with the positivity condition gives another bloch vector @xmath273 therefore we have obtained another extremal density matrix orthogonal to @xmath269 , and @xmath270 and the expectation value of the hamiltonian yields the eigenvalue @xmath274 .
until now we have obtained @xmath137 independent and orthogonal projectors , we chose @xmath269 , and @xmath275 . *
we repeat the procedure by establishing the algebraic system of equations @xmath276 whose solution give the bloch vector @xmath277 thus one gets another orthogonal projector @xmath278 and the expectation value of the hamiltonian is @xmath279 .
we have obtained the complete eigensystem of the degenerated hamiltonian .
for the eigenvalue @xmath271 , we indeed have a family of projectors yielding the same eigenvalue .
this family is associated to the standard problem , when there is degeneracy , of the diagonalization of a hamiltonian matrix , i.e. , we can take any linear combination of the corresponding independent eigenstates . for the states space of a quartit
, the density matrix is given by @xmath280 where we define @xmath281 .
we consider the hamiltonian matrix @xmath282 with @xmath283 , @xmath228 and @xmath284 as real parameters . in the basis of the generalized gell - mann matrices @xmath221 , with @xmath285 , the parameters bloch vector for the hamiltonian , @xmath286 ,
is given by @xmath287 with @xmath288 .
therefore , its associated matrix ( [ s3:eq16 ] ) is @xmath289 where to simplify the matrix notation we have defined @xmath290 , @xmath291 , @xmath292 , @xmath293 , and @xmath294 .
we are going to consider two illustrative instances to exemplify the non degenerate and degenerate cases . if @xmath295 , @xmath296 , and @xmath297 , thus the rank of @xmath258 equals to @xmath298 . in consequence , from table [ tab ] , for these values the hamiltonian ( [ sw : eq2 ] ) is non degenerate . by applying the gauss - jordan method to the system ( [ s3:eq15 ] ) , one gets the bloch vector of the density matrix , @xmath299 where @xmath300 the components @xmath301 are free variables , which are determined by establishing the system of polynomial equations ( [ s3:eq7 ] ) , where the constants @xmath302 , @xmath249 and @xmath303 must lie inside the allowed region exhibited in fig . [ region](b ) .
we consider two cases : ( i ) the pure case when one has @xmath304 , which has four independent solutions for the parameters @xmath305 .
the extremal density matrices are projectors defined by @xmath306 and @xmath307 .
they are functions of the parameter @xmath284 and the corresponding expectation values are plotted in fig [ mixto](b ) .
the levels are indicated by dotted lines , which indicates that for @xmath308 the hamiltonian system is degenerated by pairs .
\(a ) as a function of @xmath309 . ( a ) pure case with @xmath304 ; and ( b ) mixed case with @xmath310
are plotted with continuous lines .
the dotted lines represent the minimum and maximum eigenvalues of the pure case.,title="fig:",scaledwidth=40.0% ] ( b ) as a function of @xmath309 .
( a ) pure case with @xmath304 ; and ( b ) mixed case with @xmath310 are plotted with continuous lines .
the dotted lines represent the minimum and maximum eigenvalues of the pure case.,title="fig:",scaledwidth=40.0% ] \(ii ) the mixed case is established by taking from the region exhibited in fig .
[ region](c ) the values @xmath310 .
one has 24 different extremal expectation values of the hamiltonian , @xmath234 for each energy level of the pure case .
the results are shown also in fig .
[ mixto](b ) with continuous lines .
notice that the extremal expectation values are contained within the minimum and maximum eigenvalues of the hamiltonian . for @xmath311 in the hamiltonian ( [ sw : eq2 ] ) ,
the rank of @xmath258 is @xmath312 implying , from table [ tab ] , that @xmath73 is doubly degenerate and its diagonal representation is of the form @xmath313 . by applying the gauss - jordan method to the system ( [ s3:eq15 ] ) , one gets the bloch vector of the density matrix with seven free components @xmath314 ; the others can be written as @xmath315 the mean value of @xmath15 in the state @xmath266 depends only on the components @xmath316 and @xmath317 ; the remaining free variables can be chosen from the mentioned free set . by considering @xmath318 equal to zero , the components @xmath319 are obtained by solving the system of polynomial equations ( [ s3:eq7 ] ) in terms of the constants @xmath320 .
for the pure case , associated to @xmath304 , the set of solutions for this polynomial systems are @xmath321 where @xmath322 .
therefore , their respective critical density matrices are @xmath323 which correspond to orthogonal projectors of @xmath15 related to its degenerate eigenvalues , given respectively by @xmath324 in order to find the remaining projectors , one constructs the linear system of equations @xmath325 where the @xmath266 is written in terms of the general density matrix that commutes with the hamiltonian . by solving it for @xmath326 and @xmath316 in terms of @xmath327 and @xmath317
, one finds @xmath328 then by replacing this solution into @xmath80 , setting @xmath329 and @xmath330 equal to zero , and solving the polynomial system ( [ s3:eq7 ] ) for @xmath304 in terms of @xmath331 , one has @xmath332 which yield the one rank projectors @xmath333 the respective expectation values of the hamiltonian are @xmath334 finally , it is possible to corroborate that the set @xmath335 are a complete set of orthogonal one rank projectors because @xmath336 .
the main contribution of our work is to give an algebraic procedure to find extremal density matrices for a given hamiltonian .
our approach applies to both the degenerate and non degenerate cases of the hamiltonian .
the examples of the procedure are given for dimensions @xmath337 and show that the hamiltonian spectrum for the pure case is recovered . for the mixed case , we have verified that the extremal values of the expectation value of the hamiltonian is a convex sum of the corresponding results for the pure case .
we want to enhance that the method can be applied by replacing the hamiltonian for any observable acting on a qudit space .
we established that an extremal density matrix commutes with the hamiltonian operator and optimises its mean value .
we demonstrated that at most @xmath53 variables are necessary to find extremal density matrices with appropriate positivity conditions , for the non - degenerated case of the finite matrix hamiltonian . in the degenerate pure case
, one has more free components of the extremal density matrix which can be selected by asking orthogonality between the projectors , which allow us to obtain the energy spectrum . finally , in appendix [ b ] following the method given in @xcite
, we find also the compatible regions between the coefficients of the characteristic polynomial of the density matrix in terms of the positivity conditions of the bezoutian matrix in order to provide a self - contained approach .
this work was partially supported by conacyt - mxico ( under project no .
238494 ) and dgapa - unam ( under project no .
in110114 ) .
the authors would like to thank giuseppe marmo , margarita a. manko and vladimir i. manko for their valuable comments and also to conacyt - mxico for the ph.d .
scholarship to a.f .
for @xmath338 , the elements @xmath61 form an integrity basis for all polynomial @xmath339 invariants . in terms of them
, it is defined the symmetric matrix called bezoutian given in eq .
( [ b : eq1 ] ) . a polynomial with real coefficients has reals roots iff the bezoutian matrix is positive definite @xcite .
hence , the compatible region among the global invariants is obtained with the intersection of the positivity conditions of the density matrix from ( [ s2:eq1 ] ) with the respective positivity conditions of the bezoutian , mainly in its determinant @xmath340 @xcite . besides ,
due to @xmath341 is equal to the discriminant of the characteristic polynomial of @xmath71 , the degeneracy condition is obtained by the vanishing of @xmath341 @xcite . on the other hand , the relation between the constants @xmath342 with @xmath343
is established by @xmath344 with @xmath338 .
next we establish the allowed regions of the @xmath342 and @xmath345 for the matrix hamiltonians with dimensions @xmath346 .
consequently , the bezoutian matrix for @xmath44 is @xmath347 where from formula ( [ b : eq2 ] ) , it is obtained @xmath348 . then the condition @xmath349 gives the positivity condition @xmath350 , which corroborates the maximum value ( [ s2:eq1 ] ) . for the case @xmath351 ,
the positivity conditions ( [ s2:eq1 ] ) of the density matrix are @xmath352 while the bezoutian matrix is @xmath353 by applying the cayley - hamilton theorem and the formula ( [ b : eq2 ] ) one obtains @xmath354 similarly to the case of @xmath44 , @xmath355 is the only relevant positivity condition of @xmath356 .
expressed in terms of @xmath357 , it gives @xmath358 thus , the inequalities system formed by ( [ b : eq3 ] ) , ( [ b : eq4 ] ) and ( [ b : eq6 ] ) produces the compatible region between @xmath248 and @xmath359 .
this is shown in fig .
[ region](a ) .
the bottom line is associated with one eigenvalue zero ( yielding the condition on @xmath248 for the case @xmath44 ) while the other curves imply two equal eigenvalues for the density matrix .
the @xmath360 case is associated to density matrices of pure states and the highest is the maximal mixed state ( all the eigenvalues are equal ) . in the case of @xmath361 ,
the positivity conditions of the density matrix are given by @xmath362 and the respective bezoutian matrix is @xmath363 with @xmath364 in this case , @xmath365 is the main condition ; nevertheless , the remaining ones are crucial to avoid fake points in the compatible region for @xmath366 .
all these conditions are @xmath367 where the last one is @xmath365 .
hence , for the set @xmath366 , the region which satisfies the inequalities system formed by ( [ b : eq7 ] ) and ( [ b : eq9 ] ) , is shown in fig .
[ region](b ) .
notice that , by making zero @xmath368 , we obtain the @xmath351 result , while by making zero two eigenvalues of the density matrix the line associated to the case @xmath44 is obtained ( @xmath369 ) .
inside the solid figure ( orange color ) one has the solution for @xmath158 eigenvalues of the density matrix different from zero whereas the surfaces are associated to @xmath137 degenerated eigenvalues ( blue color ) .
the curve for the case with three equal eigenvalues and the other different is also shown ( green color ) .
20 landau l , 1927 z. phys . * 45 * 430 . + ter haar d , 1965 _ collected papers of l.d . landau _
pergamon press .
dirac p.a.m .
1929 proc .
. soc . * 25 * 62 .
von neumann j , 1932 _ mathematische grundlagen der quantenmechanik _ berlin : springer . + von neumann j. 1955 . _
mathematical foundations of quantum mechanics _ princeton university press .
mahler g , waberruss v a 1995 _ quantum networks : dynamics of open nanostructures . _ springer .
fano u , 1953 phys . rev . * 90 * 577 .
+ fano u , 1957 rev .
* 29 * 74 .
blum k , 2012 _ density matrix theory and applications _ springer .
newton r , young b l , 1968 annals of physics * 49 * 393 .
park j l , band w , 1971 founds . of physics
* 1 * 211 .
walser r , cirac j l , zoller p , 1996 .
. lett . * 77 * 2658 .
klose g , smith g , jessen p s , 2001 .
* 86 * 4721 .
amiet j p , weigert s , 1999 j. opt .
b : quantum semiclass . opt . * 1 * l5 .
amiet j p , weigert s , 2000 j. opt .
b : quantum semiclass . opt .
* 2 * 118 .
horn r , johnson c , 2013 _ matrix analysis _ cambridge university press .
chung s u , trueman t l , 1975 .
d * 11 * 633 .
dodonov v v , manko v i , 1997 phys .
a * 229 * 335 .
manko v i , manko o v , 1997 j. exp .
phys . * 85 * , 430 .
castaos o , lpez - pea r , manko m a , manko v i , 2003 , j. phys .
* 36 * 4677 .
castaos o , lpez - pea r , manko m a , manko v i , 2003 , j. j. opt .
b : semiclass . opt . * 5 * 227 .
bertlmann r. , krammer p , 2008 j. phys .
a : math . theor . * 41 * 235303 .
hioe f t , eberly j h , 1981 .
lett . * 47 * 838 . akhtarshenas s j , 2006 optics and spectroscopy * 103 * 411 . brning e , mkel h ,
messina a , petruccione f , 2012 j. mod .
opt . * 59 * 1 .
jarlskog c , 2006 j. math .
phys . * 47 * 013507 .
nielsen m a , chuang i l , 2010 _ quantum computation and quantum information _ cambridge university press .
46 * 082103 .
giraud o , braun d , baguette d , bastin t , martin j , 2015 .
lett . * 114 * 080401 .
fujii k , funahashi k , kobayashi t , 2006 int .
methods mod .
* 03 * 269 .
dita p , 2005 j. phys a : math .
38 * 2657 .
tilma t , sudarshan e c g , 2002 j. phys a : math . gen . * 35 * 10467 .
spengler c , huber m , hiesmayr b , 2010 j. phys . a : math
. theor . * 43 * 385306 . brning e , chruciski d , petruccione f. , 2008 open syst .
figueroa a. , lpez j. , castaos o. , lpez - pea r , manko m a , manko v i , + 2015 j. phys .
a : math . theor . * 48 * 065301 .
kimura g , 2003 phys .
a * 314 * 339 .
macfarlane a j , sudbery a , weisz p h 1968 .
* 11 * , 77 .
kimura g , kossakowski a , 2005 .
information dyn .
* 12 * 207 .
mallesh k s , mukunda n. 1997 pramana j. phys . * 49 * 371 .
bengtsson i , zyczkowski k , 2006 _ geometry of quantum states .
an introduction to quantum entanglement . _ cambridge university press .
seroul r , 2000 _ programming for mathematicians .
_ springer .
byrd m s , khaneja n , 2003 phys . rev .
a * 68 * 062322 .
deen s m , kabir p k , 1971 .
d * 4 * 1662 .
procesi c. 2007 _ lie groups : an approach through invariants and representations .
_ springer .
gerdt v p , khvedlidze a m , palii yu g , 2014 j. math .
sci . * 200 * 682 .
procesi c , schwarz g , 1985 invent . math . * 81 * 539 . lax p d , 2007 _ linear algebra and its applications . _
wiley - interscience .
bhatia r , 1997 _ matrix analysis .
_ springer .
gawron p , puchala z , miszczak j a , skowronek l , zyczkowski k. 2010 j. math . phys .
* 51 * , 102204 .
van der waerden b l , 1991 _ modern algebra ii .
_ springer .
cox d , little j , oshea d , 2005 _ using algebraic geometry .
_ springer .
gelfand i m , kapranov m m , zelevinsky a v , 2008 _ discriminants resultants and multidimensional determinants . _
sturmfels b , 2002 _ solving systems of polynomial equations .
_ number 97 , ams regional conference series helgason s , 1978 _ differential geometry , lie groups , and symmetric spaces .
_ academic press inc .
kus m , zyczkowski k , 2001 phys .
a * 63 * 032307 .
linden n , popescu s , sudbery a , 1999 phys .
lett . * 83 * 243 .
ercolessi , e. , marmo , g. , morandi , g. , 2001 int .
a * 16 * 5007 .
schirmer s g , zhang t , leahy j v , 2004 j. phys . a : math
37 * 1389 .
gantmacher f r , 1959 _ the theory of matrices vol 1 . _ ams chelsea publishing .
boya l , dixit k , 2008 phys .
a * 78 * 042108 .
keller j , 2008 linear algebra appl .
* 429 * 2209 caspers w j , 2008 j. of phys .
: conference series * 104 * 012032 .
prez - campos c , gonzlez - alonso j r , castaos o , lpez - pea r , 2010 ann .
( n.y . ) * 325 * 325 .
kurosch a.g .
1984 _ higher algebra .
_ mir publishers .
bhattacharya m , raman c , 2007 phys . rev .
a * 75 * 033405 .
+ bhattacharya m , 2007 am . j. phys . * 75 * 942 . | an algebraic procedure to find extremal density matrices for any hamiltonian of a qudit system is established .
the extremal density matrices for pure states provide a complete description of the system , that is , the energy spectra of the hamiltonian and their corresponding projectors . for extremal density matrices representing mixed states ,
one gets mean values of the energy in between the maximum and minimum energies associated to the pure case .
these extremal densities give also the corresponding mixture of eigenstates that yields the corresponding mean value of the energy .
we enhance that the method can be extended to any hermitian operator . |
discrete fractional transforms are the generalizations of the ordinary discrete transforms with one additional fractional parameter . in the past decades ,
various discrete fractional transforms including discrete fourier transform @xcite , @xcite , discrete fractional hartley transform @xcite , discrete fractional cosine transforms and discrete sine transform @xcite have been introduced and found wide applications in many scientific and technological areas including digital signal processing @xcite , image encryption @xcite , @xcite , @xcite and digital watermarking @xcite and others .
different fast algorithms for their implementations have been separately developed to minimize computational complexity and implementation costs . in @xcite
a discrete fractional hadamard transform for the vector of length @xmath3 was introduced , however a fast algorithm for the realization of this transform has not been proposed . in our previous paper @xcite
we describe a rationalized algorithm for dfrht possessing a reduced number of multiplications and additions .
analysis of the mentioned algorithm shows that not all of existing improvement possibilities have been realized . in this paper , we proposed a novel algorithm for the discrete fractional hadamard transform that require fewer total real additions and multiplications than our previously published solution .
a hadamard matrix of order @xmath0 is a @xmath4 symmetric matrix whose entries are either 1 or @xmath5 and whose rows are mutually orthogonal . in this paper
we will use the normalized form of this matrix and we will denote it by @xmath6 . for @xmath7 the hadamard matrices can be recursively obtained due to sylvester s construction @xcite : @xmath8\!,\ \mathbf{h}_{n}=\frac{1}{\sqrt{2 } } \left [ \begin{array}{cc } \mathbf{h}_{\frac{n}{2 } } & \mathbf{h}_{\frac{n}{2}}\\ \mathbf{h}_{\frac{n}{2 } } & -\mathbf{h}_{\frac{n}{2 } } \end{array}\right]\ ] ] for @xmath9 .
definition of the discrete fractional hadamard ( dfrht ) transform is based on an eigenvalue decomposition of the dht matrix . any real symmetric matrix ( including the hadamard matrix )
can be diagonalized , e.g. written as a product @xcite @xmath10 where @xmath11 is a diagonal matrix of order @xmath12 , whose diagonal entries are the eigenvalues of @xmath6 @xmath13\ ] ] + @xmath14 $ ] - the matrix whose columns are normalized mutually orthogonal eigenvectors of the matrix @xmath6 . the eigenvector @xmath15 is related to the eigenvalue @xmath16 .
a superscript @xmath17 denotes the matrix transposition .
the dfrht matrix of order @xmath7 with real parameter @xmath18 was first defined in @xcite .
this matrix can be regarded as a power of the dht matrix , where the exponent @xmath19 @xmath20 for @xmath21 the dfrht matrix is converted into the identity matrix , and for @xmath22 it is transformed into the ordinary dht matrix . generally the dfrht matrix is complex - valued . an essential operation , by obtaining the discrete fractional hadamard matrix , defined by ( [ eq : rozklad2 ] ) , is calculating the eigenvalues and the eigenvectors of the matrix @xmath6 .
the only eigenvalues of the unnormalized hadamard matrix of order @xmath23 are known to be @xmath24 and @xmath25 @xcite , hence the normalized hadamard matrix @xmath26 has only the eigenvalues 1 and @xmath5 .
a method for finding the eigenvectors of hadamard matrix was firstly presented in @xcite , but in @xcite a recursive method for calculation the eigenvectors of the hadamard matrix order @xmath27 based on the eigenvectors of the hadamard matrix of order @xmath12 has been proposed .
we will use this method to obtain the dfrht matrix . here
we will present it briefly . in @xcite
it was proven that if @xmath28 ( @xmath29 ) is an eigenvector of hadamard matrix of order @xmath23 associated with an eigenvalue @xmath30 , then vector @xmath31 \label{eq : wektorhat}\ ] ] will be an eigenvector of the matrix @xmath32 associated with the eigenvalue @xmath30 .
in @xcite it was proven that if @xmath28 is an eigenvector of @xmath6 associated with an eigenvalue @xmath30 , then the vector @xmath33 \label{eq : wektortilde}\ ] ] will be an eigenvector of the matrix @xmath32 associated with the eigenvalue @xmath34 .
these two results allow as to generate the eigenvectors of hadamard matrix of order @xmath27 from the eigenvectors of hadamard matrix of order @xmath12 . knowing the straightforward calculated eigenvectors of the matrix @xmath35 @xmath36\;\ ; \mathbf{v}_{2}^{(1)}=\left [ \begin{array}{c } 1-\sqrt{2}\\ 1\\ \end{array } \right ] \label{eq : wektoryh2}\ ] ] associated with eigenvalues 1 and @xmath5 respectively , the eigenvectors for hadamard matrix of arbitrary order @xmath23 can be recursively computed .
in @xcite it was also shown so this recursively computed eigenvectors of matrix @xmath26 will be orthogonal .
it should be noted that for any @xmath23 there are only two distinct eigenvalues of hadamard matrix , so for @xmath37 the eigenvalues are degenerated . because of this fact the set of eigenvectors proposed in @xcite and @xcite is not unique .
the igenvectors @xmath28 for @xmath38 , which are columns of the matrix @xmath39 ( after normalization ) , as well as their associated eigenvalues @xmath40 , can be however ordered in different ways . in @xcite
it has been also established a method of ordering the eigenvectors . in many cases ,
including the case of discrete fractional transforms is used so - called sequency ordering of the eigenvectors .
this means that the @xmath41-th eigenvector has @xmath41 sign - changes .
the discrete hermite - gaussians , eigenvectors of discrete fourier transform matrix are ordered this way as well @xcite .
we will show this method of ordering of the eigenvectors in example [ example1 ] .
[ example1 ] the number of sign - changes in eigenvectors @xmath42 and @xmath43 of matrix @xmath35 , determined by ( [ eq : wektoryh2 ] ) , is equal to 0 and 1 respectively . using expressions ( [ eq : wektorhat ] ) and
( [ eq : wektortilde ] ) we obtain the eigenvectors of matrix @xmath44 : @xmath45\;\ ; \mathbf{\tilde v}_{4}^{(0)}= \left [ \begin{array}{c } -b\\ -b^2 \\ 1\\ b\\ \end{array } \right]\;\ ; \mathbf{\hat v}_{4}^{(1)}= \left [ \begin{array}{c } -b\\ 1 \\
-b^2\\ b\\ \end{array } \right]\;\ ; \mathbf{\tilde v}_{4}^{(1)}=
\left [ \begin{array}{c } b^2\\ -b \\
-b\\ 1\\ \end{array } \right],\ ] ] where @xmath46 .
the numbers of sign - changes in the above vectors are 0 , 1 , 3 , 2 respectively ( @xmath47 ) .
therefore , a sequency ordered set of eigenvectors of matrix @xmath44 will be as follows : @xmath48 the corresponding eigenvalues will be equal to : @xmath49 the relations obtained in example [ example1 ] can be easily generalized as follows : @xmath50 \mathbf{v}_{2n}^{(4l+1)}=\mathbf{\tilde v}_{2n}^{(2l)}\\[6pt ] \mathbf{v}_{2n}^{(4l+2)}=\mathbf{\tilde v}_{2n}^{(2l+1)}\\[6pt ] \mathbf{v}_{2n}^{(4l+3)}=\mathbf{\hat v}_{2n}^{(2l+1 ) } \end{array } \right.\ ] ] for @xmath51 and @xmath52 for @xmath53 .
both the eigenvectors of the matrix @xmath35 and the eigenvectors obtained for higher order hadamard matrices are not normalized .
let the notation @xmath54 means the euclidean norm of vector @xmath28 .
in @xcite it was shown that for any @xmath7 we have the relationship @xmath55 where @xmath38 and @xmath56 .
if we take the designation @xmath57 , then normalized eigenvectors of hadamard matrix of order @xmath7 will take the form @xmath58 using the normalized and sequency ordered eigenvectors of the hadamard matrix , the eigenvalue decomposition ( [ eq : rozklad ] ) of the hadamard matrix can be written as follows : @xmath59 where @xmath11 is the diagonal matrix whose non - zero elements are @xmath60 for @xmath38 .
hence the definition ( [ eq : rozklad2 ] ) of dfrht matrix will take the form : @xmath61 where @xmath62 for @xmath38 .
our goal is to calculate the discrete fractional hadamard transform for an input signal @xmath63 in which the number of samples is equal to @xmath23 . by @xmath64
we denote an output signal which is calculated using the formula @xmath65 supposing that the matrix @xmath66 is given , to calculate the output signal it is necessary to perform @xmath1 complex multiplications and @xmath67 complex additions .
if the input signal is real , then the number of real multiplications will be equal to @xmath68 , and the number of real additions will be equal to @xmath69 .
if we use the decomposition ( [ eq : rozklad2n ] ) of the matrix @xmath66 by calculating ( [ eq : transformata ] ) and will perform the matrix - vector multiplication from the right side to the left , the most time - consuming operations are multiplications of matrices @xmath70 and @xmath71 by the vector , because those matrices are not diagonal .
if we interchange the columns of the matrix @xmath71 in the prescribed manner , we obtain a matrix @xmath72 of special structure , which can be generated recursively .
we will show it in example [ example2 ] . it will allow to reduce the number of arithmetical operations by calculating the products of matrices @xmath70 and @xmath71 by the vector .
[ example2 ] the matrices @xmath71 for @xmath73 are as follows : @xmath74,\;\ ; \mathbf{v}_4= \left [ \begin{array}{cccc } 1&-b&b^2&-b\\ b&-b^2&-b&1\\ b&1&-b&-b^2\\ b^2&b&1&b\\ \end{array } \right],\;\;\ ] ] @xmath75.\ ] ] the matrix @xmath76 has some specific structure .
now we consider the matrix @xmath77 .
if in the matrix @xmath77 the second and fourth columns will be interchange and then the third and fourth columns will be interchange too , we obtain the following matrix : @xmath78 = \left [ \begin{array}{cc } \mathbf{v}_2&-b\mathbf{v}_2\\ b\mathbf{v}_2&\mathbf{v}_2\\ \end{array } \right].\ ] ] the matrix @xmath77 differs from the matrix @xmath79 only in the order of the columns . therefore , the matrix @xmath77 can be obtained by post - multiplying the matrix @xmath79 by the permutation matrix @xmath80 : @xmath81 where @xmath82.\ ] ] now we consider the matrix @xmath83 .
if we perform the following permutation of columns of this matrix : @xmath84 as a result we obtain the following matrix : @xmath85= \left [ \begin{array}{cc } \mathbf{\overline{v}}_4&-b\mathbf{\overline{v}}_4\\ b\mathbf{\overline{v}}_4&\mathbf{\overline{v}}_4\\ \end{array } \right].\ ] ] as previously , we can write : @xmath86 where @xmath87.\ ] ] if we generalize the above considerations for @xmath7 we can write : @xmath88 for @xmath89 . for @xmath90 we can also write @xmath91 where @xmath92 is an identity matrix of order two @xmath93 the permutation matrix @xmath94 of order @xmath95 can be obtained recursively from the permutation matrix @xmath96 of order @xmath97 according to the following relation : @xmath98,\ \mathbf{p}_{n}=\mathbf{s}_{n } \left[\begin{array}{cc } \mathbf{p}_{\frac{n}{2}}\hspace{0.2 cm } & \mathbf{0}_{\frac{n}{2}}\\ \mathbf{0}_{\frac{n}{2}}\hspace{0.2 cm } & \mathbf{p}_{\frac{n}{2}}\mathbf{j}_{\frac{n}{2}}\\ \end{array}\right]\!\ ] ] for @xmath9 .
@xmath99 is the perfect shuffle permutation matrix of order @xmath12 , @xmath100 is the counter - identity matrix of order @xmath101 and @xmath102 is zero matrix .
the perfect shuffle permutation is the permutation that splits the set consisting of an even number of elements into two piles and interleaves them .
it can be written as follows : @xmath103 for example @xmath104 , \
\mathbf{j}_4= \left [ \begin{array}{cccc } 0&0&0&1\\ 0&0&1&0\\ 0&1&0&0\\ 1&0&0&0\\ \end{array } \right].\ ] ] if we write the matrix @xmath71 as a product @xmath105 the expression ( [ eq : rozklad2n ] ) will take the form : @xmath106 the product @xmath107 is a diagonal matrix , which has the same diagonal entries as the matrix @xmath108 but in different order and for a chosen parameter @xmath109 it may be prepared in advance . if we denote this product multiplied by a factor @xmath110 by @xmath111 : @xmath112 the dfrht algorithm ( [ eq : transformata ] ) will take the following form : @xmath113 where the matrix @xmath114 can be generated recursively : @xmath115\;\ ; \mathbf{\overline{v}}_{2k}= \left [ \begin{array}{cc } \mathbf{\overline{v}}_k&-b\mathbf{\overline{v}}_k\\ b\mathbf{\overline{v}}_k&\mathbf{\overline{v}}_k\\ \end{array } \right ] \label{eq : defvn}\ ] ] for @xmath116 .
the most time - consuming operations by calculating the dfrht transform according to ( [ eq : transformatan ] ) are multiplications of matrices @xmath117 and @xmath72 by the vector . since in the matrix @xmath72
occur only following powers of @xmath118 we can write this matrix as follows : @xmath119 in the figure [ fig : figure1 ] it was shown the way of calculating the matrix - vector product @xmath120 , using the expression ( [ eq : sum ] ) . in this paper ,
the graph - structural models and data flow diagrams are oriented from left to right .
straight lines in the figures denote the operation of data transfer .
we use the usual lines without arrows specifically so as not to clutter the picture .
note that the circles in this figure shows the operations of multiplication by a number inscribed inside a circle . in turn , the rectangles indicate the matrix - vector multiplications by matrices , scaledwidth=70.0% ] although it may seem strange , we will see that such an operation allows to reduce the number of multiplication and additions by multiplying the matrix @xmath121 by a vector .
it should be noted that because of the recursive relation ( [ eq : defvn ] ) between matrices @xmath122 and @xmath123 , the following recursive relation between the matrices @xmath124 , @xmath125 and @xmath126 occurs : @xmath127=\mathbf{i}_{n}\ ] ] @xmath128 \label{eq : defan}\ ] ] @xmath129\ ] ] for @xmath130 and @xmath131=log@xmath2 , where @xmath132=\mathbf{i}_{2},\;\;\ ; \mathbf{a}_{2}^{(1)}= \left [ \begin{array}{cc } 0&-1\\ 1&0\\ \end{array } \right].\ ] ] to clarify our idea we show the explicit form of expressions ( [ eq : sum ] ) and ( [ eq : defan ] ) for @xmath90 , @xmath133 and @xmath134 in an example [ example3 ] . [ example3 ] @xmath135 where the matrices @xmath136 and @xmath137 are presented above .
@xmath138 where @xmath139= \left [ \begin{array}{cc } \mathbf{a}_{2}^{(0)}&\mathbf{0}_{2}\\ \mathbf{0}_{2}&\mathbf{a}_{2}^{(0)}\\ \end{array } \right]= \mathbf{i}_4,\ ] ] @xmath140= \left [ \begin{array}{cc } \mathbf{a}_{2}^{(1)}&-\mathbf{a}_{2}^{(0)}\\ \mathbf{a}_{2}^{(0)}&\mathbf{a}_{2}^{(1)}\\ \end{array } \right],\ ] ] @xmath141= \left [ \begin{array}{cc } \mathbf{0}_{2}&-\mathbf{a}_{2}^{(1)}\\ \mathbf{a}_{2}^{(1)}&\mathbf{0}_{2}\\ \end{array } \right ] .\ ] ] @xmath142 where @xmath143=
\left [ \begin{array}{cc } \mathbf{a}_{4}^{(0)}&\mathbf{0}_{4}\\ \mathbf{0}_{4}&\mathbf{a}_{4}^{(0)}\\ \end{array } \right]= \mathbf{i}_8,\ ] ] @xmath144= \left [ \begin{array}{cc } \mathbf{a}_{4}^{(1)}&-\mathbf{a}_{4}^{(0)}\\ \mathbf{a}_{4}^{(0)}&\mathbf{a}_{4}^{(1)}\\ \end{array } \right],\ ] ] @xmath145= \left [ \begin{array}{cc } \mathbf{a}_{4}^{(2)}&-\mathbf{a}_{4}^{(1)}\\ \mathbf{a}_{4}^{(1)}&\mathbf{a}_{4}^{(2)}\\ \end{array } \right],\ ] ] @xmath146= \left [ \begin{array}{cc } \mathbf{0}_{4}&-\mathbf{a}_{4}^{(2)}\\ \mathbf{a}_{4}^{(2)}&\mathbf{0}_{4}\\ \end{array } \right].\ ] ] now we will evaluate the number of arithmetical operations , which are necessary to calculate the matrix - vector product @xmath147 .
we note , that in a general case such an operation requires @xmath1 multiplications and @xmath67 additions .
now we will calculate the numbers of multiplications and additions needed for this operation if we use the expression ( [ eq : sum ] ) for the matrix @xmath72 , i.e. @xmath148 since the non - zero entries of matrices @xmath149 , @xmath150 are only 1 and -1 , no multiplications are needed when calculating the matrix - vector products @xmath151 .
the only multiplications we have to perform are multiplications of the vectors @xmath152 by the powers of @xmath153 : @xmath154 by @xmath153 , @xmath155 by @xmath156 by @xmath157 . because the number @xmath153 is constant and known , its powers @xmath158 , @xmath159 may be prepared in advance .
thus , the number of multiplication by calculating the matrix - vector product @xmath160 is equal to @xmath161log@xmath0 .
let us examine the number of additions , we need to perform , when calculating the matrix - vector product @xmath160 .
the total number of additions consist of number of additions by calculating the matrix - vector products @xmath152 , and @xmath162 additions which are needed to calculate the sum of vectors : @xmath163 , @xmath164 , @xmath165 .
since , according to ( [ eq : defan ] ) , the matrices @xmath166 have specific structures , the products @xmath152 can be obtained by subtracting the products @xmath167 , @xmath168 of twice smaller size and summing the products @xmath169 , @xmath170 ( excluding products @xmath163 and @xmath171 which can be obtained even in a simpler way ) . by @xmath172^t$ ] we denote the first half of the input vector @xmath173 and by @xmath174^t$ ] - the second half of this vector , as it was shown , for @xmath134 , in the figure [ fig : figure2 ] . using the products of twice smaller size : @xmath175 , @xmath176 , @xmath177 , @xmath178 for @xmath179,scaledwidth=100.0% ] it
should be noted that the products @xmath167 and @xmath170 are used to calculate both products @xmath180 and @xmath181 .
for example the products @xmath182 and @xmath183 are used to calculate @xmath184 and @xmath185 .
it allows to reduce the number of additions , because the some products are used twice .
of course , this procedure can be repeated and each of products @xmath167 , @xmath168 , @xmath169 , @xmath170 can be calculated by summing ( subtracting ) products of twice smaller size and so on .
it can be continued until calculating products of matrices @xmath186 and @xmath187 by two - element sub - vectors of the vector @xmath63 .
the whole process of going down by calculating the product @xmath147 is presented , for @xmath134 , in the figure [ fig : figure3 ] . , scaledwidth=70.0% ] the expression ( [ eq : rozklad4 ] ) can be also written as the matrix - vector product , as follows : @xmath188 where @xmath189,\ ] ] @xmath190\otimes\mathbf{i}_n,\ ] ] @xmath191 the symbol @xmath192 denotes the kronecker product of matrices , and @xmath193 is the matrix ( row vector ) whose all entries are equal to 1 .
the matrix @xmath194 is responsible for multiplications of the matrices @xmath149 , @xmath195 , @xmath196 , @xmath197 by the input vector @xmath173 , the matrix @xmath198 - for multiplications of those products by the proper powers of @xmath153 , and the matrix @xmath199 - for aggregation of results .
the process of going down by calculating the product @xmath160 , which has been presented in figures [ fig : figure2 ] and [ fig : figure3 ] , can be also described in the therm of matrices product .
it means factorisation of the matrix @xmath194 into the product of @xmath131 matrices @xmath200 the matrices which occur on the right side of expression ( [ eq : rozklada ] ) have the following forms : @xmath201\ ] ] where @xmath202\otimes\mathbf{i}_{1}=\mathbf{a}_{2}^{(0)},\;\ ; \mathbf{\overline{a}}_{2\times 2}^{(1)}=\mathbf{a}_{2}^{(1)}\otimes\left [ 1\right]\otimes\mathbf{i}_{1}=\mathbf{a}_{2}^{(1)}.\ ] ] @xmath203,\ ] ] where @xmath204\otimes\mathbf{i}_{2},\;\ ; \mathbf{\overline{a}}_{4\times 8}^{(1)}=\mathbf{a}_{2}^{(1)}\otimes\left[0\;\ ; 1\right]\otimes\mathbf{i}_{2}\ ] ] and the matrix @xmath205 denotes the matrix @xmath206 which columns were circularly shifted by 2 positions to the right , and the matrix @xmath207 denotes the matrix @xmath206 which columns were circularly shifted by 2 positions to the left . the last matrix @xmath208 is defined as @xmath209\ ] ] where @xmath210\otimes\mathbf{i}_{n/2},\;\ ; \mathbf{\overline{a}}_{n\times nn}^{(1)}=\mathbf{a}_{2}^{(1)}\otimes\left[0\;\;0\;\;\ldots \;\;1\right]\otimes\mathbf{i}_{n/2}.\ ] ] using the expression ( [ eq : rozklada ] ) the algorithm ( [ eq : rozkladmac ] ) of calculating the product @xmath160 can be written as follows : @xmath211 the expression ( [ eq : productv ] ) allows for evaluating the total number of additions , which are needed to calculate the matrix - vector product @xmath160 .
we assume that the input vector @xmath173 is real - valued .
each of matrices @xmath212 , for @xmath213 , is the direct sum of @xmath214 identical blocks and the single block is the vertical concatenation of @xmath215 matrices .
the firs , indicated by @xmath216 , and the last , indicated by @xmath217 , do not need any additions or subtractions by multiplying them by a vector .
the @xmath218 others matrices , which are sums of @xmath216 and @xmath217 , after circularly shifting their columns , so multiplying each of them by a vector needs @xmath219 additions . to calculate the product @xmath220 we have to perform @xmath221 additions .
the product of the matrix @xmath198 by a vector do not need any additions and the product of the matrix @xmath222 by a vector needs @xmath162 additions .
thus the total number of additions by calculating the products @xmath160 , according to ( [ eq : productv ] ) , is equal to @xmath223 .
example [ example4 ] shows the explicit form of the algorithm ( [ eq : productv ] ) with all occurring in it matrices for @xmath134 .
[ example4 ] the algorithm ( [ eq : productv ] ) of calculating the product of matrix @xmath224 by the vector @xmath225 is as follows : @xmath226 where @xmath227= \left [ \begin{array}{cccccccc } 1&0&0&0&0&0&0&0\\ 0&1&0&0&0&0&0&0\\ 0&-1&0&0&0&0&0&0\\ 1&0&0&0&0&0&0&0\\ 0&0&1&0&0&0&0&0\\ 0&0&0&1&0&0&0&0\\ 0&0&0&-1&0&0&0&0\\ 0&0&1&0&0&0&0&0\\ 0&0&0&0&1&0&0&0\\ 0&0&0&0&0&1&0&0\\ 0&0&0&0&0&-1&0&0\\ 0&0&0&0&1&0&0&0\\ 0&0&0&0&0&0&1&0\\ 0&0&0&0&0&0&0&1\\
0&0&0&0&0&0&0&-1\\ 0&0&0&0&0&0&1&0\\ \end{array } \right],\ ] ] @xmath228=\ ] ] @xmath229,\ ] ] @xmath230=\ ] ] @xmath231,\ ] ] @xmath232\otimes \mathbf{i}_{8},\ ] ] @xmath233 it is easy to check that in this case the total number of addition is equal to 48 and the number of multiplications is equal to 24 ( we can see it also in the figure [ fig : figure3 ] ) .
now we return to the dfrht algorithm ( [ eq : transformatan ] ) . according to ( [ eq : sum ] ) the matrix @xmath121
can be written as the sum of the matrices @xmath149 , @xmath150 with coefficients @xmath234 .
the transposed matrix @xmath235 can be written as the sum of the transposed matrices @xmath236 ,
@xmath237 with the same coefficients @xmath234 : @xmath238 since the matrices with the even indexes @xmath149 , @xmath239 are symmetric and the matrices with the odd indexes are asymmetric the expression ( [ eq : sumt ] ) can be written as follows : @xmath240 according to ( [ eq : rozkladmac ] ) the matrix @xmath121 from expression ( [ eq : sum ] ) can be transformed into the product @xmath241 so the matrix @xmath235 may by also transformed from ( [ eq : sumt1 ] ) into the following product : @xmath242 where the matrix @xmath243 is defined as follows : @xmath244 and @xmath245.\ ] ] the others matrices in the expression ( [ eq : rozkladmac1 t ] ) are the same as that in the expression ( [ eq : rozkladmac ] ) .
taking into account the decompositions ( [ eq : rozkladmac1 ] ) and ( [ eq : rozkladmac1 t ] ) of matrices @xmath72 and @xmath117 respectively the dfrht algorithm ( [ eq : transformatan ] ) will take the form : @xmath246 where the matrix @xmath194 is decomposed according to ( [ eq : rozklada ] ) .
for example , for @xmath134 this algorithm will take the following form : @xmath247 figure [ fig : figure4 ] shows a data flow diagram of the algorithm for 8 point dfrht . ) for @xmath134,scaledwidth=100.0% ]
let us assess the computational complexity in term of numbers of multiplications and additions required for dfrht calculation .
calculation of the discrete fractional hadamard transform for a real - valued vector @xmath63 of length @xmath7 , assuming that the matrix @xmath66 defined by ( [ eq : rozklad2 ] ) is given , requires @xmath248 multiplications of a complex number by a real number and @xmath249 complex additions .
each multiplication of a complex number by a real number needs two real multiplications and each addition of two complex numbers requires two real additions .
hence the numbers of real multiplications and real additions required for computing the dfrht using the naive method are equal to @xmath250 and @xmath251 respectively .
let us now evaluate the computational complexity of the dfrht realization with the help of the procedure ( [ eq : transformata2 ] ) . as it was discussed in the section [ advantages
] , if we use the factorized representation of the matrices @xmath117 and @xmath72 , calculating the product of the real - valued matrix @xmath117 and the real - valued vector @xmath63 requires @xmath162 real multiplications and @xmath223 real additions . as a result ,
we again obtain the real - valued vector .
then there is computed the product of the complex - valued diagonal matrix @xmath252 and the real - valued vector calculated previously ( we assume that for a predetermined parameter @xmath109 , the diagonal elements of this matrix were calculated in advance ) .
the calculation of this product requires @xmath253 real multiplications .
the resulting complex - valued vector is then multiplied by the factorized matrix @xmath72 . this operation requires @xmath254 real multiplications and @xmath255 real additions . the total numbers of arithmetic operations to compute dfrht of size @xmath12 using our new algorithm are @xmath256 real multiplications and @xmath257 real additions .
it is easy to check that even for small @xmath131 the numbers of arithmetic operations required for realization of the proposed algorithm are several times less than in the naive method of computing .
tables [ tab:1 ] and [ tab:2 ] display the numbers of multiplications and additions required for the dfrht transform implementation of the real - valued input signal of the length @xmath7 .
these numbers were calculated for three methods of the transform implementation : the direct multiplication of the dfrht matrix by a vector of the input data , calculation according to authors algorithm described in the work @xcite , and according to the algorithm ( [ eq : transformata2 ] ) proposed in this article .
it is easy to check that for @xmath258 the number of arithmetic operations , required for dfrht transform realization according to the proposed algorithm , is smaller than in the other two methods of dfrht computing . [ !
h ] rrrr @xmath7 & direct method & method @xcite & proposed algorithm + 2 & 8 & 6 & 10 + 4 & 32 & 18 & 32 + 8 & 128 & 54 & 88 + 16 & 512 & 162 & 224 + 32 & 2048 & 486 & 544 + 64 & 8192 & 1458 & 1280 + 128 & 32768 & 4374 & 2944 + 256 & 131072 & 13122 & 6656 + 512 & 524288 & 39366 & 14848 + 1024 & 2097152 & 118098 & 32768 + [ !
h ] rrrr @xmath7 & direct method & method @xcite & proposed algorithm + 2 & 4 & 5 & 6 + 4 & 24 & 25 & 36 + 8 & 112 & 95 & 144 + 16 & 480 & 325 & 480 + 32 & 1984 & 1055 & 480 + 64 & 8064 & 3325 & 1440 + 128 & 32512 & 10295 & 4032 + 256 & 130560 & 31525 & 10752 + 512 & 523264 & 95855 & 69120 + 1024 & 2095104 & 290125 & 168960 +
the article presents the novel algorithm for the dfrht performing .
the algorithm has a much lower computational complexity than the direct way of the dfrht implementation .
the computational procedure for dfrht calculating is described in kronecker product notation .
the kronecker product algebra is a very compact and simple mathematical formalism suitable for parallel realization .
this notation enables us to represent adequately the space - time structures of an implemented computational process and directly maps these structures into the hardware realization space . for simplicity , we considered the synthesis of a fast algorithm for the dfrht calculation for @xmath259 .
however it is clear that the proposed procedure was developed for the arbitrary case when the order of the matrix is a power of two .
sylvester j.j . : thoughts on inverse orthogonal matrices , simultaneous sign successions , and tessellated pavements in two or more colours , with applications to newton s rule , ornamental tile - work , and the theory of numbers .
philos . mag .
34 , 461 - 475 ( 1867 ) | we present a novel algorithm for calculating the discrete fractional hadamard transform for data vectors whose size @xmath0 is a power of two .
a direct method for calculation of the discrete fractional hadamard transform requires @xmath1 multiplications , while in proposed algorithm the number of real multiplications is reduced to @xmath0log@xmath2 .
example.eps gsave newpath 20 20 moveto 20 220 lineto 220 220 lineto 220 20 lineto closepath 2 setlinewidth gsave .4 setgray fill grestore stroke grestore |
the social adoption of new technologies in networked cyberphysical systems relies heavily on the privacy preservation of individual information .
social networking , the power grid , and smart transportation are only but a few examples of domains in need of privacy - aware design of control and coordination strategies . in these scenarios ,
the ability of a networked system to fuse information , compute common estimates of unknown quantities , and agree on a common view of the world is critical . motivated by these observations , this paper studies the multi - agent average consensus problem , where a group of agents seek to agree on the average of their individual values by only interchanging information with their neighbors .
this problem has numerous applications in synchronization , network management , and distributed control / computation / optimization . in the context of privacy preservation , the notion of differential privacy has gained significant popularity due to its rigorous formulation and proven security properties , including resilience to post - processing and side information , and independence from the model of the adversary . roughly speaking ,
a strategy is differentially private if the information of an agent has no significant effect on the aggregate output of the algorithm , and hence its data can not be inferred by an adversary from its execution .
this paper is a contribution to the emerging body of research that studies privacy preservation in cooperative network systems , specifically focused on gaining insight into the achievable trade - offs between privacy and performance in multi - agent average consensus .
the problem of multi - agent average consensus has been a subject of extensive research in networked systems and it is impossible to survey here the vast amount of results in the literature .
we provide @xcite and the references therein as a starting point for the interested reader . in cyberphysical systems ,
privacy at the physical layer provides protection beyond the use of higher - level encryption - based techniques .
information - theoretic approaches to privacy at the physical layer have been actively pursued @xcite .
recently , these ideas have also been utilized in the context of control @xcite .
the paper @xcite also surveys the more recent game - theoretic approach to the topic . in computer science ,
the notion of differential privacy , first introduced in @xcite , and the design of differentially private mechanisms have been widely studied in the context of privacy preservation of databases .
the work @xcite provides a recent comprehensive treatment .
a well - known advantage of differential privacy over other notions of privacy is its immunity to post - processing and side information , which makes it particularly well - suited for multi - agent scenarios where agents do not fully trust each other and/or the communication channels are not fully secure . as a result , this notion has been adopted by recent works in a number of areas pertaining to networked systems , such as control @xcite , estimation @xcite , and optimization @xcite . of relevance to our present work , the paper @xcite studies the average consensus problem with differentially privacy guarantees and proposes an adjacency - based distributed algorithm with decaying laplace noise and mean - square convergence .
the algorithm preserves the differential privacy of the agents initial states but the expected value of its convergence point depends on the network topology and may not be the exact average , even in expectation .
by contrast , the algorithm proposed in this work enjoys almost sure convergence , asymptotic unbiasedness , and an explicit characterization of its convergence rate .
our results also allow individual agents to independently choose their level of privacy .
the problem of privacy - preserving average consensus has also been studied using other notions of privacy .
the work @xcite builds on @xcite to let agents have the option to add to their first set of transmitted messages a zero - sum noise sequence with finite random length , which in turn allows the coordination algorithm to converge to the exact average of their initial states .
as long as an adversary can not listen to the transmitted messages of an agent as well as all its neighbors , the privacy of that agent is preserved , in the sense that different initial conditions may produce the same transmitted messages .
this idea is further developed in @xcite , where agents add an infinitely - long exponentially - decaying zero - sum sequence of gaussian noise to their transmitted messages .
the algorithm has guaranteed mean - square convergence to the average of the agents initial states and preserves the privacy of the nodes whose messages and those of their neighbors are not listened to by the malicious nodes , in the sense that the maximum - likelihood estimate of their initial states has nonzero variance .
finally , @xcite considers the problem of privacy preserving maximum consensus .
we study the average consensus problem where a group of agents seek to compute and agree on the average of their local variables while seeking to keep them differentially private against an adversary with potential access to all group communications .
this privacy requirement also applies to the case where each agent wants to keep its initial state private against the rest of the group ( e.g. , due to the possibility of communication leakages ) .
the main contributions of this work are the characterization and optimization of the fundamental trade - offs between differential privacy and average consensus .
our first contribution is the formulation and formal proof of a general impossibility result .
we show that as long as a coordination algorithm is differentially private , it is impossible to guarantee the convergence of agents states to the average of their initial values , even in distribution .
this result automatically implies the same impossibility result for stronger notions of convergence . motivated by it , our second contribution is the design of a linear laplacian - based consensus algorithm that achieves average consensus in expectation the most that one can expect .
we prove the almost sure convergence and differential privacy of our algorithm and characterize its accuracy and convergence rate .
our final contribution is the computation of the optimal values of the design parameters to achieve the most accurate consensus possible .
letting the agents fix a ( local ) desired value of the privacy requirement , we minimize the variance of the algorithm convergence point as a function of the noise - to - state gain and the amplitude and decay rate of the noise .
we show that the minimum variance is achieved by the one - shot perturbation of the initial states by laplace noise .
this result reveals the optimality of one - shot perturbation for static average consensus , previously ( but implicitly ) shown in the sense of information - theoretic entropy .
various simulations illustrate our results .
this section introduces notations and basic concepts .
we denote the set of reals , positive reals , non - negative reals , positive integers , and nonnegative integers by @xmath0 , @xmath1 , @xmath2 , @xmath3 , and @xmath4 , respectively .
we denote by @xmath5 the euclidean norm .
we let @xmath6 denote the space of vector - valued sequences in @xmath7 . for @xmath8
, we use the shorthand notation @xmath9 and @xmath10 . @xmath11 and @xmath12 denote the identity matrix and the vector of ones , respectively . for @xmath13 , @xmath14 denotes the average of its components .
we let @xmath15 .
note that @xmath16 is diagonalizable , has one eigenvalue equal to @xmath17 with eigenspace @xmath18 and all other eigenvalues equal @xmath19 . for a vector space @xmath20 ,
we let @xmath21 denote the vector space orthogonal to @xmath22 .
a matrix @xmath23 is stable if all its eigenvalues have magnitude strictly less than 1 .
a function @xmath24 belongs to class @xmath25 if it is continuous and strictly increasing and @xmath26 .
similarly , a function @xmath27 belongs to class @xmath28 if @xmath29 belongs to class @xmath25 for any @xmath30 and @xmath31 is decreasing and @xmath32 for any @xmath33 . for @xmath34 , the euler function
is given by @xmath35 .
note that @xmath36 for a function @xmath37 and sets @xmath38 and @xmath39 , we use @xmath40 and @xmath41 . in general , @xmath42 . finally , for any topological space @xmath43 , we denote by @xmath44 the set of borel subsets of @xmath43 .
we present some useful notions on algebraic graph theory following @xcite .
let @xmath45 denote a weighted undirected graph with vertex set @xmath22 of cardinality @xmath46 , edge set @xmath47 , and symmetric adjacency matrix @xmath48 .
a path from @xmath49 to @xmath50 is a sequence of vertices starting from @xmath49 and ending in @xmath50 such that any pair of consecutive vertices is an edge .
the set of neighbors @xmath51 of @xmath49 is the set of nodes @xmath50 such that @xmath52 .
a graph is connected if for each node there exists a path to any other node .
the weighted degree matrix is the diagonal matrix @xmath53 with diagonal @xmath54 .
the laplacian is @xmath55 and has the following properties : * @xmath56 is symmetric and positive semi - definite ; * @xmath57 and @xmath58 , i.e. , @xmath19 is an eigenvalue of @xmath56 corresponding to the eigenspace @xmath59 ; * @xmath60 is connected if and only if @xmath61 , so @xmath19 is a simple eigenvalue of @xmath56 ; * all eigenvalues of @xmath56 belong to @xmath62 $ ] , where @xmath63 is the largest element of @xmath64 . for convenience ,
we define @xmath65 . here
we briefly review basic notions on probability following @xcite .
consider a probability space @xmath66 . if @xmath67 are two events with @xmath68 , then @xmath69 .
for simplicity , we may sometimes denote events of the type @xmath70 by @xmath71 , where @xmath72 is a logical statement on the elements of @xmath73 .
clearly , for two statements @xmath72 and @xmath74 , @xmath75 a random variable is a measurable function @xmath76 . for any @xmath77 and any random variable @xmath43 with
finite expected value @xmath78 and finite nonzero variance @xmath79 , chebyshev s inequality states that @xmath80 for a random variable @xmath43 , let @xmath81 $ ] and @xmath82 denote its expectation and cumulative distribution function , respectively . a sequence of random variables @xmath83 converges to a random variable @xmath43 * almost surely ( a.s . ) if @xmath84 ; * in mean square if @xmath85 , { \mathbb{e}}[x^2 ] < \infty$ ] for all @xmath86 and @xmath87 = 0 $ ] ; * in probability if @xmath88 for any @xmath89 ; * in distribution or weakly if @xmath90 for any @xmath91 at which @xmath82 is continuous .
almost sure convergence and convergence in mean square imply convergence in probability , which itself implies convergence in distribution .
moreover , if @xmath92 for all @xmath93 and some fixed random variable @xmath94 with @xmath95 < \infty$ ] , then convergence in probability implies mean square convergence , and if @xmath43 is a constant , then convergence in distribution implies convergence in probability .
a zero - mean random variable @xmath43 has laplace distribution with scale @xmath96 , denoted @xmath97 , if its pdf is given by @xmath98 for any @xmath99 .
it is easy to see that @xmath100 has exponential distribution with rate @xmath101 .
this section briefly describes notions of robustness for discrete - time systems following @xcite .
consider a discrete - time system of the form @xmath102 where @xmath103 is a disturbance input , @xmath104 is the state , and @xmath105 is a vector field satisfying @xmath106 .
the system is globally input - to - state stable ( iss ) if there exists a class @xmath28 function @xmath107 and a class @xmath25 function @xmath108 such that , for any bounded input @xmath109 , any initial condition @xmath110 , and all @xmath93 , @xmath111 where @xmath112 . the system has a @xmath25-asymptotic gain if there exists a class @xmath25 function @xmath113 such that , for any initial condition @xmath110 , @xmath114 if a system is iss , then it has a @xmath25-asymptotic gain .
furthermore , any lti system @xmath115 is iss if @xmath116 is stable .
consider a group of @xmath46 agents whose interaction topology is described by an undirected connected graph @xmath60 .
the group objective is to compute the average of the agents initial states while preserving the privacy of these values against potential adversaries eavesdropping on all the network communications .
note that this privacy requirement is the same as the case where each agent wants to keep its initial state private against the rest of the group due to the possibility of communication leakages .
we next generalize the exposition in @xcite to provide a formal presentation of this problem .
the state of each agent @xmath117 is represented by @xmath118 .
the message that agent @xmath49 shares with its neighbors about its current state is denoted by @xmath119 . for convenience , the aggregated network state and the vector of transmitted messages
are denoted by @xmath120 and @xmath121 , respectively .
agents update their states in discrete time according to some rule , @xmath122 with initial states @xmath123 , where the state - transition function @xmath124 is such that its @xmath49th element depends only on @xmath125 and @xmath126 .
the messages are calculated from @xmath127 where @xmath128 is such that its @xmath49th element depends only on @xmath125 and @xmath129 . for simplicity , we assume that @xmath130 and @xmath131 are continuous .
@xmath132 is a vector random variable , with @xmath133 being the noise generated by agent @xmath49 at time @xmath134 from an arbitrary distribution .
consequently , @xmath135 and @xmath136 are sequences of vector random variables on the total sample space @xmath137 whose elements are noise sequences @xmath138 .
although one could choose @xmath131 to only depend on @xmath139 , corrupting the messages by noise is necessary to preserve privacy . given an initial state @xmath140 , @xmath136 is uniquely determined by @xmath138 according to - .
therefore , the function @xmath141 such that @xmath142 is well defined .
we next introduce the notion of differential privacy . *
( differential privacy)*[def : privacy ] given @xmath143 , the initial network states @xmath144 and @xmath145 are @xmath146-adjacent if , for some @xmath147 , @xmath148 for @xmath149 . given @xmath150 , the dynamics - is @xmath151-differentially private if , for any pair @xmath144 and @xmath145 of @xmath146-adjacent initial states and any set @xmath152 , @xmath153 a final aspect to consider is that , because of the presence of noise , the agents states under might not converge exactly to their initial average @xmath154 , but to a neighborhood of it .
this is captured by the notion of accuracy . *
( accuracy)*[def : accuracy ] for @xmath155 $ ] and @xmath156 , the dynamics - is @xmath157-accurate if , from any initial state @xmath140 , the network state @xmath158 converges to @xmath159 as @xmath160 , with @xmath161 = { \operatorname{ave}(\theta_0 ) } { \mathbf{1}}_n$ ] and @xmath162 . in definition [ def : accuracy
] , the type of convergence of @xmath158 to @xmath163 can be any of the four classes described in section [ subsection : probability ] . furthermore
, for each notion of convergence , @xmath164-accuracy is equivalent to the convergence of @xmath158 to @xmath165 .
we are finally ready to formally state our problem . *
( differentially private average consensus)*[problem : main ] design the dynamics , the inter - agent messages , and the distribution of noise sequences @xmath138 such that asymptotic average consensus is achieved with @xmath157-accuracy while guaranteeing @xmath151-differential privacy for ( finite ) @xmath151 , @xmath166 , and @xmath167 as small as possible .
in this section we establish the impossibility of solving problem [ problem : main ] with @xmath164-accuracy , even if considering the weakest notion of convergence . [
prop : bound ] consider a group of agents executing a distributed algorithm of the form with messages generated according to .
then , for any @xmath168 , agents can not simultaneously converge to the average of their initial states in distribution and preserve @xmath151-differential privacy of their initial states .
we reason by contradiction .
assume there exists an algorithm that achieves convergence in distribution to the exact average of the network initial state and preserves @xmath151-differential privacy of it .
since the algorithm must preserve the privacy of _ any _ pair of @xmath146-adjacent initial conditions , consider a specific pair satisfying @xmath169 for some @xmath170 and @xmath171 for all @xmath172 . since @xmath154 is fixed ( i.e. , deterministic ) , the convergence of @xmath173 , @xmath174 to @xmath154 is also in probability .
thus , for any @xmath117 and any @xmath89 , we have @xmath175 , for @xmath176 .
therefore , for any @xmath177 , there exists @xmath86 such that for all @xmath117 , @xmath178 now , considering - , it is clear that , for any fixed initial state @xmath140 and any @xmath86 , @xmath179 is uniquely determined by @xmath180 and @xmath181 is uniquely determined by @xmath179 .
therefore , the functions @xmath182 such that @xmath183 are well defined and continuous ( due to continuity of @xmath130 and @xmath184 ) .
next , for @xmath185 , define @xmath186 , where @xmath187 and @xmath188 is the @xmath189-neighborhood of @xmath190 . by , we have @xmath191 note that @xmath192 is open as it is the continuous pre - image of an open set , so @xmath193 is borel . to reach a contradiction , we define @xmath194 and show that @xmath195 can be made arbitrarily small by showing that @xmath196 . to do this , note that by the definitions of @xmath197 , @xmath198 and @xmath192 , we have @xmath199 recall that in , @xmath130 is such that the next state of each agent only depends on its current state and the messages it receives .
hence , since for all @xmath172 , @xmath200 , we have from that @xmath201 where @xmath202 is the same as @xmath203 except that the requirement on @xmath204 ( to be close to @xmath205 ) is relaxed .
now , since @xmath206 and , by choosing @xmath207 , we get @xmath208 , we conclude that @xmath209 , which implies @xmath210 , so we get @xmath211 as desired . now , let @xmath212 .
for any initial condition @xmath140 , @xmath213 hence , since the algorithm is @xmath151-differentially private , @xmath214 thus , using and , we have for all @xmath177 , @xmath215 which is clearly a contradiction because @xmath151 is a finite number , completing the proof .
since convergence in distribution is the weakest notion of convergence , proposition [ prop : bound ] implies that a differentially private algorithm can not guarantee any type of convergence to the exact average .
therefore , in our forthcoming discussion , we relax the exact convergence requirement and allow for convergence to a random variable that is at least unbiased ( i.e. , centered at the true average ) .
here , we develop a solution to problem [ problem : main ] . consider the following linear distributed dynamics , @xmath216 for @xmath217 , where @xmath218 is the step size , @xmath219 is a diagonal matrix with diagonal @xmath220 and @xmath221 for each @xmath117 , and the messages are generated according to @xmath222 where the @xmath49th component of the noise vector @xmath223 has the laplace distribution @xmath224 at any time @xmath225 with @xmath226 note that is a special case of ( since @xmath227 ) and a special case of .
also note that without the term @xmath228 , the average of the agents initial states would be preserved throughout the evolution . *
( comparison with the literature ) * the proposed algorithm - has similarities and differences with the algorithm proposed in @xcite which can be expressed ( with a slight change of notation in using @xmath229 instead of @xmath230 ) as @xmath231 \theta(k ) + \big[s - s d^{-1 } l \big ] \eta(k ) .
\end{aligned}\ ] ] if each agent selects @xmath232 , then we recover - . as we show later , this particular choice results in an unbiased convergence point , while in general the expected value of the convergence point of the algorithm in @xcite depends on the graph structure and may not be the true average .
furthermore , this algorithm is only shown to exhibit mean square convergence of @xmath158 for @xmath233 , while here we provide an explicit expression for the convergence point and establish convergence in the stronger a.s .
sense for larger range of @xmath234 . as we show later , the inclusion of @xmath235 is critical , as it leads to identifying the optimal algorithm performance . on a different note ,
the algorithms in @xcite and @xcite add a noise sequence to the messages which is correlated over time
the latter using a different notion of privacy .
@xcite generate a single noise at time @xmath236 and add a scaled version of it to the messages at every time @xmath237 , leading to an effectively `` one - shot''-type of perturbation .
we show in section [ section : optimal - noise - selection ] that the one - shot approach is optimal for static average consensus while sequential perturbation is necessary for dynamic scenarios .
this section analyzes the asymptotic correctness of the algorithm - and characterizes its rate of convergence .
we start by establishing convergence . *
( asymptotic convergence)*[prop : convergence ] consider a network of @xmath46 agents executing the distributed dynamics - .
define the random variable @xmath163 as @xmath238 then , @xmath163 is well - defined a.s . , and the states of all agents converge to @xmath163 almost surely
. note that @xmath221 ensures that @xmath239 is not empty .
substituting into , the system dynamics is @xmath240 with @xmath241 and @xmath242 .
for any @xmath243 , let @xmath244 multiplying both sides of by @xmath245 on the left and using the fact that @xmath245 and @xmath56 commute , the dynamics of @xmath246 can be expressed as @xmath247 notice that @xmath248 is forward invariant under . therefore , considering @xmath248 as the state space for and noting that @xmath249 is stable on it , we deduce that is iss .
consequently , this dynamics has a @xmath25-asymptotic gain ( c.f .
section [ subsection : iss ] ) , i.e. , there exists @xmath250 such that @xmath251 therefore , @xmath252 implies @xmath253 . by definition , the latter means that there is @xmath89 such that for all @xmath254 there exists @xmath255 with @xmath256 .
in other words , there exists a subsequence @xmath257 such that @xmath258 for all @xmath259 .
this , in turn , implies that for all @xmath259 , @xmath260 , i.e. , @xmath261 according to , this chain of implications gives @xmath262 the last equality holds because @xmath263 .
therefore , we conclude @xmath264 from , we see that a.s .
convergence of @xmath139 requires a.s .
convergence of @xmath265 as well .
left multiplying by @xmath266 , we obtain for all @xmath225 , @xmath267 which in turn implies @xmath268 we next prove that @xmath269 converges almost surely to @xmath163 . in order for the latter to be well - defined over @xmath73
, we simply set @xmath270 when the series does not converge .
clearly , for any @xmath271 such that @xmath272 converges for all @xmath117 , we have @xmath273 .
therefore , using , @xmath274 note that , for each @xmath117 and any @xmath275 , if @xmath276 for all @xmath277 , then @xmath278 converges .
therefore , using and the definition of laplace distribution , we have for all @xmath275 , @xmath279 for each @xmath117 , because @xmath280 , there exists @xmath281 such that @xmath282 for @xmath283 .
therefore , using the euler function @xmath284 , @xmath285 for all @xmath275 , and hence , @xmath286 this , together with and , implies that @xmath287 , which completes the proof .
[ rem : ms ] from and the fact that @xmath288 , we have @xmath289 for all @xmath93 .
it is straightforward to show @xmath290 < \infty$ ] , so , using proposition [ prop : convergence ] , @xmath158 also converges to @xmath291 in mean square .
our next aim is to characterize the convergence rate of the distributed dynamics - . given the result in proposition [ prop : convergence ] , we define the exponential mean - square convergence rate of the dynamics - as @xmath292}{{\mathbb{e}}\big[(\theta(0 ) - \theta_\infty { \mathbf{1}}_n)^t ( \theta(0 ) - \theta_\infty { \mathbf{1}}_n)\big ] } \right)^\frac{1}{2k } \hspace*{-1ex}.\end{aligned}\ ] ] in the absence of noise ( @xmath293 ) , this definition coincides with the conventional exponential convergence rate of autonomous linear systems , see e.g. , @xcite . *
( convergence rate)*[prop : conv - rate ] under the hypotheses of proposition [ prop : convergence ] , the exponential mean - square convergence rate of the distributed dynamics - is @xmath294 where @xmath295 and @xmath296 is the spectral radius of @xmath297 . for convenience ,
we let @xmath298 denote the convergence error at @xmath93 and @xmath299 .
our first goal is to obtain an expression for @xmath300 $ ] . from and
the proof of proposition [ prop : convergence ] , we have @xmath301 almost surely .
then , from , we have almost surely for all @xmath93 , @xmath302 where we have used the fact that @xmath303 .
next , note that for all @xmath304 , @xmath305 where we have used the facts that @xmath16 is idempotent and @xmath306 for any @xmath307 .
let @xmath308 .
notice that @xmath309 has spectral radius @xmath310 and the same eigenvectors as @xmath56 .
then , using twice , we have almost surely for all @xmath304 , @xmath311 by the independence of @xmath312 over time , we have @xmath313 & = \theta_0^t { \mathcal{a}}^{2k } \theta_0 \\ \notag & \quad + \sum_{j = 0}^{k - 2 } { \mathbb{e}}[\eta(j)^t b { \mathcal{a}}^{2k - 2 - 2j } b \eta(j ) ] \\ \notag & \quad + { \mathbb{e}}[\eta(k - 1)^t b l_\text{cpt}^2 b \eta(k - 1 ) ] \\ & \quad + \sum_{j = k}^\infty { \mathbb{e}}[\eta(j)^t s \pi_n^2 s \eta(j ) ] , \end{aligned}\ ] ] for all @xmath304 .
next , we upper bound the exponential mean - square convergence rate @xmath78 .
let @xmath314 and note that for any @xmath315 and any @xmath307 , @xmath316 & = \sum_{i = 1}^n 2 b_i^2(j ) ( n^t n)_{ii } \\ & \le 2 \overline c^2 \overline q^{2j } { \text{tr}}(n^t n ) = 2 \overline c^2 \overline q^{2j } \|n\|_f^2 , \end{aligned}\ ] ] where @xmath317 denotes the frobenius norm . therefore , @xmath318 \le \theta_0^t { \mathcal{a}}^{2k } \theta_0 + 2 \overline
c^2 \sum_{j = 0}^{k - 2 } \overline q^{2j } \|{\mathcal{a}}^{k - 1 - j } b\|_f^2 \\ & + 2 \overline c^2 \overline q^{2(k - 1 ) } \|l_\text{cpt } b\|_f^2 + 2 \overline c^2 \sum_{j = k}^\infty \overline q^{2j } \|\pi_n s\|_f^2 .
\end{aligned}\ ] ] since the frobenius norm is submultiplicative , @xmath319 for any matrix @xmath320 , and @xmath321 , we have @xmath318 \le \theta_0^t { \mathcal{a}}^{2k } \theta_0 + c_1 \sum_{j = 0}^{k - 2 } \overline q^{2j } \overline \lambda^{2k - 4 - 2j } + c_2 \overline q^{2k } , \end{aligned}\ ] ] where @xmath322 and @xmath323 are constants .
note that for any @xmath324 , we have @xmath325 .
therefore , using the fact that the supremum of a sum is less than or equal to the sum of suprema , we have @xmath326}{{\mathbb{e}}\big[\hat \theta_0^t \hat \theta_0\big ] } \le \sup_{\theta_0 \in { \ensuremath{\mathbb{r}}}^n } \frac{\theta_0^t { \mathcal{a}}^{2k } \theta_0}{{\mathbb{e}}\big[\hat \theta_0^t \hat \theta_0\big ] } \\ & \qquad \qquad \qquad + \sup_{\theta_0 \in { \ensuremath{\mathbb{r}}}^n}\frac{c_3 ( k - 1 ) \max\{\overline q , \overline \lambda\}^{2k } + c_2 \overline q^{2k}}{{\mathbb{e}}\big[\hat \theta_0^t \hat \theta_0\big ] } , \end{aligned}\ ] ] where @xmath327 .
let @xmath328 be the initial disagreement vector .
it is straightforward to verify that @xmath329 and @xmath330 = \tilde \theta_0^t \tilde \theta_0 + \frac{1}{n } \sum_{i = 1}^n \frac{2 c_i^2 s_i^2}{1 - q_i^2 } \triangleq \tilde \theta_0^t \tilde \theta_0 + c_4 .
\end{aligned}\ ] ] therefore , @xmath331}{{\mathbb{e}}\big[\hat
\theta_0^t \hat \theta_0\big ] } & \le \sup_{\tilde \theta_0 \in ( { \ensuremath{\mathbb{r}}}{\mathbf{1}}_n)^\perp } \frac{\tilde \theta_0^t { \mathcal{a}}^{2k } \tilde \theta_0}{\tilde \theta_0^t \tilde \theta_0 + c_4 } \\ & + \frac{c_3 ( k - 1 ) \max\{\overline q , \overline \lambda\}^{2k } + c_2 \overline q^{2k}}{\inf_{\tilde \theta_0 \in ( { \ensuremath{\mathbb{r}}}{\mathbf{1}}_n)^\perp } ( \tilde \theta_0^t \tilde \theta_0 + c_4 ) } \\ & \hspace{-1in}= \overline \lambda^{2k } + c_3 c_4^{-1 } ( k - 1 ) \max\{\overline q , \overline \lambda\}^{2k } + c_2 c_4^{-1 } \overline q^{2k}. \end{aligned}\ ] ] by raising the right hand side of the above expression to the power @xmath332 and taking the limit as @xmath333 , the constant / polynomial factors converge to 1 and the terms containing @xmath334 dominate the sum , proving that @xmath335 .
similarly , we can lower bound @xmath78 as follows . from , we have for all @xmath304 , @xmath336 \ge \theta_0^t { \mathcal{a}}^{2k } \theta_0 \\ & \rightarrow \mu \ge \lim_{k \to \infty } \left(\sup_{\tilde \theta_0 \in ( { \ensuremath{\mathbb{r}}}{\mathbf{1}}_n)^\perp } \frac{\tilde \theta_0^t { \mathcal{a}}^{2k } \tilde \theta_0}{\tilde \theta_0^t \tilde \theta_0 + c_4 } \right)^{1/2k } = \overline \lambda , \end{aligned}\ ] ] and @xmath336 \ge { \mathbb{e}}[\eta(k)^t s \pi_n^2 s \eta(k ) ] = \sum_{i = 1}^n c_{5i } q_i^{2k } \\ & \rightarrow \mu \ge \lim_{k \to \infty } \left(\sup_{\tilde \theta_0 \in ( { \ensuremath{\mathbb{r}}}{\mathbf{1}}_n)^\perp } \frac{\sum_{i = 1}^n c_{5i } q_i^{2k}}{\tilde \theta_0^t \tilde \theta_0 + c_4 } \right)^{1/2k } = \overline q , \end{aligned}\ ] ] where @xmath337 for all @xmath338 . therefore , @xmath339 , completing the proof .
note that @xmath340 is the convergence rate of the noise - free ( and non - private ) laplacian - based average consensus algorithm , while @xmath341 is the worst - case decay rate of the noise sequence among the agents . from
, the convergence rate @xmath78 is the larger of these two values , confirming our intuition that the slower rate among them is the bottleneck for convergence speed .
also , note that @xmath340 depends on the network topology @xmath60 while @xmath341 is independent of it .
having established the convergence properties of the algorithm , this section characterizes the extent to which our design solves problem [ problem : main ] by providing guarantees on its accuracy and differential privacy .
the next result elaborates on the statistical properties of the agreement value of the algorithm .
[ cor : accuracy ] under the hypotheses of proposition [ prop : convergence ] , the convergence point @xmath163 is an unbiased estimate of @xmath154 with bounded dispersion , @xmath342 as a result , the algorithm - is @xmath343-accurate for any @xmath344 .
since noises are independent over time and among agents , we deduce from that for any @xmath93 , @xmath345 and @xmath346 which establishes unbiasedness and bounded dispersion for any time . as @xmath333 , we get @xmath347 and @xmath348 the @xmath349-accuracy follows directly by applying chebyshev s inequality for @xmath350 . * ( comparison with the literature
contd ) * proposition [ prop : convergence ] and corollary [ cor : accuracy ] establish almost sure convergence , with the expected value of convergence being the average of the agents initial states .
in contrast , the results in @xcite establish convergence in mean square , and the expected value of convergence depends on the network topology . in both cases , the accuracy radius @xmath166 decreases with the number of agents as @xmath351 .
the expression for @xmath349-accuracy in corollary [ cor : accuracy ] shows that one can not obtain the ideal case of @xmath352-accuracy , and that @xmath166 is a decreasing function of @xmath72 , with @xmath353 as @xmath354 .
this is an ( undesirable ) consequence of the lack of preservation of the average under due to the term @xmath355 . in turn
, the presence of this expression helps establish the differential privacy of the algorithm with bounded , asymptotically vanishing noise , as we show next .
[ prop : privacy ] under the hypotheses of proposition [ prop : convergence ] , let @xmath356 for each @xmath117 , where @xmath146 is the adjacency bound in . then , the algorithm preserves the @xmath357-differential privacy of agent @xmath49 s initial state for all @xmath117 .
consequently , the algorithm is @xmath151-differential private with @xmath358 .
consider any pair of @xmath146-adjacent initial conditions @xmath144 and @xmath145 and an arbitrary set @xmath359 .
for any @xmath93 , let @xmath360 where @xmath361 is the sample space up to time @xmath134 , @xmath362 is given in , and @xmath363 is the set composed by truncating the elements of @xmath364 to finite subsequences of length @xmath365 .
then , by the continuity of probability ( * ? ? ? * theorem 1.1.1.iv ) , @xmath366 for @xmath367 , where @xmath368 is the @xmath369-dimensional joint laplace pdf given by @xmath370 next , we define a bijection between @xmath192 and @xmath371 . without loss of generality ,
assume @xmath372 for some @xmath170 , where @xmath373 and @xmath374 for all @xmath172 .
then , for any @xmath375 , define @xmath376 by @xmath377 for @xmath378 .
it is not difficult to see that @xmath379 , so @xmath380 .
since the converse argument is also true , the above defines a bijection .
therefore , for any @xmath380 there exists a unique @xmath381 such that @xmath382 note that @xmath383 is fixed and does not depend on @xmath376 .
thus , we can use a change of variables to get @xmath384 comparing for @xmath385 with , we see that both integrals are over @xmath192 with different integrands . dividing the integrands for any @xmath375 yields , @xmath386 due to ,
the geometric series in the exponent of the multiplicative term is convergent .
therefore , integrating both sides over @xmath192 and letting @xmath333 , we have @xmath387 which establishes the @xmath388-differential privacy for agent @xmath389 .
the fact the @xmath389 can be any agent establishes , while the last statement follows from definition [ def : privacy ] . since the algorithm - converges almost surely ( cf .
proposition [ prop : convergence ] ) and is differentially private ( cf .
proposition [ prop : privacy ] ) , proposition [ prop : bound ] implies that it can not achieve @xmath164-accuracy , as noted above when discussing corollary [ cor : accuracy ] .
the explicit privacy - accuracy trade - off is given by the relation between @xmath390 and @xmath391 , i.e. , ( c.f .
, ) @xmath392 so @xmath393 increases as any @xmath357 is decreased and vice versa . we optimize this trade - off over @xmath394 in section [ section : optimal - noise - selection ] and depict the optimal trade - off curve for a test network in section [ section : simulations ] . even though the choice of laplacian noise in is not the only one that can be made to achieve differential privacy , it is predominant in the literature @xcite .
the work @xcite shows that laplacian noise is optimal ( among all possible distributions ) in the sense that it minimizes the entropy of the transmitted messages while preserving differential privacy . * ( comparison with the literature contd ) * proposition [ prop : privacy ] guarantees the @xmath357-differential privacy of agent @xmath49 s initial state independently of the noise levels chosen by other agents .
therefore , each agent can choose its own level of privacy , and even opt not to add any noise to its messages , without affecting the privacy of other agents .
in contrast , in @xcite , agents need to agree on the level of privacy before executing the algorithm . in both cases ,
privacy is achieved against an adversary that can hear everything , independently of how it processes the information .
in contrast , the algorithm in @xcite assumes the adversary uses maximum likelihood estimation and only preserves the privacy of those agents who are sufficiently `` far '' from it in the graph ( an agent is sufficiently far if the adversary can not listen to it and all of its neighbors ) . the latter work uses a different notion of privacy based on the covariance of the maximum likelihood estimate which allows for guaranteed exact convergence , in the mean - square sense , to the true average . in this section , we discuss the effect on the algorithm s performance of the free parameters present in our design . given the trade - off between accuracy and privacy , cf . , we fix the privacy levels @xmath395 constant and study the best achievable accuracy of the algorithm as a function of the remaining free parameters .
each agent @xmath396 gets to select the parameters @xmath229 , @xmath397 , @xmath398 determining the amount of noise introduced in the dynamics , with the constraint that @xmath399 , where @xmath400 given the characterization of accuracy in corollary [ cor : accuracy ] , we consider as cost function the variance of the agents convergence point , i.e. , @xmath163 , around @xmath154 , giving @xmath401 the next result characterizes its global minimization . *
( optimal parameters for variance minimization)*[prop : optimal ] for the adjacency bound @xmath402 and privacy levels @xmath403 fixed , the optimal value of the variance of the agents convergence point is @xmath404 the infimum is not attained over @xmath405 but approached as @xmath406 and @xmath407 for all @xmath117 . for each @xmath117 , with the privacy level fixed , the expression follows directly from . for convenience ,
we re - parameterize the noise decaying ratio @xmath398 as @xmath408 note that @xmath409 . substituting and into , we obtain ( with a slight abuse of notation , we also use @xmath410 to denote the resulting function ) , @xmath411}. \end{aligned}\ ] ] therefore , to minimize @xmath410 , each agent has to independently minimize the same function @xmath412 of its local parameters @xmath413 over @xmath414 .
figure [ fig : phi ] illustrates the graph of this function over @xmath64 .
since @xmath64 is not compact , the infimum might not be attained , and in fact , this is the case .
it is easy verify that @xmath415 .
now , for all @xmath416 , @xmath417 so @xmath418 if @xmath419 , then @xmath420 . if @xmath421 , then @xmath422
therefore , for all @xmath416 , @xmath423 , which completes the proof . given that differential privacy is resilient to post - processing , an alternative design strategy to preserve the differential privacy of agents initial states is to inject noise only at the initial time , @xmath424 . from
, the introduction of a one - shot noise by agent @xmath49 corresponds to @xmath425 which is not feasible if @xmath426 .
this can also be seen by rewriting as @xmath427 so if @xmath426 for any @xmath49 , @xmath173 directly ( not only through @xmath428 ) depend on @xmath429 . however ,
if @xmath235 , one can verify using a simplified version of the proof of proposition [ prop : privacy ] that @xmath425 also preserves @xmath357-differential privacy of @xmath429 with @xmath430 .
this results in a cost of @xmath431 showing that the optimal accuracy is also achieved by one - shot perturbation of the initial state at time @xmath236 and injection of no noise thereafter .
a similar conclusion ( that one - shot laplace perturbation minimizes the output entropy ) can be drawn from @xcite , albeit this is not explicitly mentioned therein . *
( dynamic average consensus ) * one - shot perturbation would no longer be optimal in dynamic average consensus scenarios @xcite , where agents seek to compute the average of individual exogenous , time - varying signals ( the `` static '' average consensus considered here would correspond to the exogenous signals being constant ) . in such scenarios , there is a recurrent flow of information at each node whose privacy can no longer be preserved with one - shot perturbation .
sequential perturbation as in - is then necessary and the variance of the noise sequence has to dynamically depend on the rate of information flow to each node .
although the detailed design of such algorithms is beyond the scope of this work , such an algorithm can be designed following the idea of the sequential perturbation design of this work and the proof of its privacy in proposition [ prop : privacy ] . to see this , note that ( for @xmath432 ) we `` tune '' the amount of noise injection @xmath133 so that the privacy of @xmath433 is preserved at each round @xmath237 , but @xmath434 is the amount of `` retained information '' of @xmath435 at round @xmath134 and plays the same role as @xmath436 in the dynamic average consensus problem .
in this section , we report simulation results of the distributed dynamics - on a network of @xmath437 agents .
figure [ fig : graph ] shows the random graph used throughout the section , where edge weights are i.i.d .
and each one equals a sum of two i.i.d .
bernoulli random variables with @xmath438 .
the agents initial states are also i.i.d . with distribution @xmath439 .
as can be seen from and , neither accuracy nor privacy depend on the initial values or the communication topology ( albeit according to the convergence rate depends on the latter ) . in all the simulations , @xmath440 and @xmath441 for all @xmath117
. figure [ fig : sweep ] depicts simulations with @xmath442 and @xmath443 while sweeping @xmath444 over @xmath445 $ ] with logarithmic step size . for each value of @xmath444
, we set @xmath446 with @xmath447 for each @xmath117 and repeat the simulation @xmath448 times .
for each run , to capture the statistical properties of the convergence point , the graph topology and initial conditions are the same and only noise realizations change .
figure [ fig : sweep](a ) shows the empirical ( sample ) standard deviation of the convergence point as a function of @xmath444 , verifying the optimality of one - shot perturbation . in particular ,
notice the sensitivity of the accuracy to @xmath444 close to @xmath449 .
figure [ fig : sweep](b ) shows the ` settling time ' , defined as the number of rounds until convergence ( measured by a tolerance of @xmath450 ) , as a function of @xmath444 .
the fastest convergence is achieved for @xmath449 , showing that one - shot noise is also optimal in the sense of convergence speed .
we have observed the same trends as in figure [ fig : sweep ] for different random choices of initial conditions and network topologies .
note that the settling time depends on both the convergence rate and the initial distance from the convergence point @xmath451 .
the former is constant at @xmath452 for @xmath453 $ ] .
the latter depends on @xmath454 , which in turn depend on @xmath444 by .
this explains the trend observed in figure [ fig : sweep](b ) .
figure [ fig : tradeoff ] depicts the privacy - accuracy trade - off for the proposed algorithm .
we have set @xmath455 , @xmath456 , and @xmath457 and then swept @xmath458 logarithmically over @xmath459 $ ] . in figure
[ fig : tradeoff](a ) , the algorithm is run 25 times for each value of the @xmath458 and the error @xmath460 for each run is plotted as a circle . in figure
[ fig : tradeoff](b ) , the sample variance of the convergence point @xmath163 is shown as a function of @xmath458 together with the theoretical value given in proposition [ prop : optimal ] .
in both plots , we see an inversely - proportional relationship between accuracy and privacy , as expected .
figure [ fig : hist ] shows the histogram of convergence points for @xmath461 runs of the algorithm with @xmath462 , @xmath455 and @xmath456 ( optimal accuracy ) .
the distribution of the convergence point is a bell - shaped curve with mean exactly at the true average , in accordance with corollary [ cor : accuracy ] .
although the distribution of @xmath163 is provably non - gaussian , the central limit theorem , see e.g. , @xcite , implies that it is very close to gaussian since the number of agents is large . finally , figure [ fig : mu ] illustrates the convergence rate of the algorithm . here , for @xmath463 , @xmath464 , @xmath465 , and the same topology as in the previous plots ,
the initial agents states are randomly selected and the whole algorithm is run 100 times with different noise realizations @xmath138 , each time until @xmath466 iterations . for each value of initial states and
each @xmath467 , we empirically approximate the quantity @xmath468}{{\mathbb{e}}\big[(\theta(0 ) - \theta_\infty { \mathbf{1}}_n)^t ( \theta(0 ) - \theta_\infty { \mathbf{1}}_n)\big ] } \right)^{1/2k}\end{aligned}\ ] ] by taking the sample mean instead of the expectation in the numerator and denominator .
we repeat this whole process 50 times for different random initial conditions and plot the result , together with the theoretical value of @xmath78 ( which in this case equals @xmath469 ) given by proposition [ prop : conv - rate ] .
as figure [ fig : mu ] shows , the supremum of the resulting curves converges to @xmath78 as @xmath333 , as expected .
we have studied the problem of multi - agent average consensus subject to the differential privacy of agents initial states .
we have showed that the requirement of differential privacy can not be satisfied if agents states weakly converge to the exact average of their initial states .
this result suggests that the most one can expect of a differentially private consensus algorithm is that the consensus value is unbiased , i.e. , its expected value is the true average , and the variance is minimized .
we have designed a linear consensus algorithm that meets this objective , and have carefully characterized its convergence , accuracy , and differential privacy properties .
future work will include the investigation of the limitations and advantages of differential privacy in multi - agent systems , the extension of the results to distributed optimization , filtering , and estimation , and the design of algorithms for privacy preservation of the network structure and other parameters such as edge weights and vertex degrees .
this work was supported by nsf award cns-1329619 .
f. bullo , j. corts , and s. martnez .
_ distributed control of robotic networks_. applied mathematics series . princeton university press , 2009 .
isbn 978 - 0 - 691 - 14195 - 4 .
electronically available at http://coordinationbook.info . c. dwork , f. mcsherry , k. nissim , and a. smith . calibrating noise to sensitivity in private data analysis . in _ proceedings of the 3rd theory of cryptography conference _ , pages 265284 , new york , ny , march 2006 .
s. han , u. topcu , and g. j. pappas .
differentially private convex optimization with piecewise affine objectives . in _
ieee conf . on decision and control _ , pages 21602166 , los angeles , ca , december 2014 .
z. huang , s. mitra , and g. dullerud .
differentially private iterative synchronous consensus . in _ proceedings of the 2012 acm workshop on privacy in the electronic society _ , pages 8190 , new york , ny , 2012 .
z. huang , y. wang , s. mitra , and g. e. dullerud .
on the cost of differential privacy in distributed control systems . in _ proceedings of the 3rd international conference on high confidence networked systems ( hicons ) _ , pages 105114 , berlin , germany , april 2014 .
z. huang , s. mitra , and n. vaidya . differentially private distributed optimization . in _ proceedings of the 2015 international conference on distributed computing and networking _ , pilani , india ,
january 2015 .
m. kefayati , m. s. talebi , b. h. khalaj , and h. r. rabiee .
secure consensus averaging in sensor networks using random offsets . in _
ieee intern .
conf . on telec . , and malaysia intern
. conf . on communications _
, pages 556560 , penang , may 2007 .
s. s. kia , j. corts , and s. martnez .
dynamic average consensus under limited control authority and privacy requirements .
_ international journal on robust and nonlinear control _ , 250 ( 13):0 19411966 , 2015 . a. mukherjee , s.a.a .
fakoorian , j. huang , and a.l .
principles of physical layer security in multiuser wireless networks : a survey . _ ieee communications surveys & tutorials _ , 160 ( 3):0 15501573 , 2014 .
y. wang , z. huang , s. mitra , and g. e. dullerud .
entropy - minimizing mechanism for differential privacy of discrete - time linear feedback systems . in _
ieee conf . on decision and control _ ,
pages 21302135 , los angeles , ca , december 2014 . | this paper studies the multi - agent average consensus problem under the requirement of differential privacy of the agents initial states against an adversary that has access to all the messages .
we first establish that a differentially private consensus algorithm can not guarantee convergence of the agents states to the exact average in distribution , which in turn implies the same impossibility for other stronger notions of convergence .
this result motivates our design of a novel differentially private laplacian consensus algorithm in which agents linearly perturb their state - transition and message - generating functions with exponentially decaying laplace noise .
we prove that our algorithm converges almost surely to an unbiased estimate of the average of agents initial states , compute the exponential mean - square rate of convergence , and formally characterize its differential privacy properties .
we show that the optimal choice of our design parameters ( with respect to the variance of the convergence point around the exact average ) corresponds to a one - shot perturbation of initial states and compare our design with various counterparts from the literature .
simulations illustrate our results .
average consensus , differential privacy , multi - agent systems , exponential mean - square convergence rate , networked control systems |
an incredible amount of our advances in understanding the universe over the past 100 years can be traced back to einstein s groundbreaking 1915 paper on general relativity .
general relativity is used to understand the precession of mercury s perihelion , the orbital decay of binary pulsars , and the expansion and evolution of the universe from the time of the big bang , among many other phenomena .
one of the most important uses of general relativity in observational astronomy is its ability to accurately describe how a massive object can act as a gravitational lens , bending and magnifying the light from a source located elsewhere in the universe .
while gravitational lensing is studied on the largest scales ( e.g. , galaxy clusters act as gravitational lenses for galaxies at higher redshifts ) , it also has observable effects at smaller scales , where gravitational lensing effects are used to detect the existence of extrasolar planets.@xcite this planet detection method relies on the fact that the light from a distant star may be lensed by objects in a closer stellar system .
depending on the relative alignment of earth , the distant star , and the closer stellar system , the light from the distant star may be lensed by a planet , star , or both in the stellar system . in all cases ,
the angular separation between the multiple images that are typically formed by these lensing events are too small ( on the order of 1 milliarcsecond ) to be resolved by telescopes on earth .
hence , we call this lensing effect gravitational microlensing , in order to distinguish it from macrolensing events , in which the multiple images can be resolved.@xcite despite the fact that the gravitationally lensed images can not be resolved , we can still determine that a microlensing event is taking place by observing the lensed star over an extended time period . the motion of the lensing stellar system with respect to the distant star gradually changes the alignment between the earth , stellar system , and distant star , which affects the amount of lensing that occurs . over tens of days , the light we detect from the distant star may appear to increase and then decrease in brightness as the star in the stellar system moves into and out of alignment with our line - of - sight to the distant star .
an extrasolar planet in the stellar system may cause a similar brightening and then dimming of the distant star , but on a time scale that may be as short as a few hours.@xcite using gravitational microlensing to detect extrasolar planets thus requires observations of a large patch of sky over an extended period of time .
liebes ( 1964)@xcite was the first to realize that gravitational lensing could be used to detect extrasolar planets , although the search for extrasolar planets using gravitational microlensing began with the work of mao and paczynski ( 1991).@xcite the first confirmed detection of an extrasolar planet by gravitational microlensing was by bond et al .
( 2004).@xcite to date , thirty - seven planets have been discovered using this technique.@xcite while this represents approximately only 2% of all known extrasolar planets , gravitational microlensing provides the ability to detect earth - mass planets around 1 au from their parent stars.@xcite this mass and distance combination is interesting since such extrasolar planets could hypothetically harbor life as we know it .
the detection and study of extrasolar planets lies at the cutting - edge of modern astrophysics@xcite , and it is also a topic that excites the imagination of general education students and the public at large .
consequently , we developed new active engagement curricular materials for teaching this topic in college - level general education introductory astronomy courses ( hereafter astro 101 ) .
these materials , informed by decades of work into how people learn , especially in stem disciplines@xcite , include new _ lecture - tutorials _ , think - pair - share activities , homework assignments , lecture slides , and assessment questions focused on the detection of extrasolar planets .
we created these active learning instructional activities and assessments to provide astro 101 instructors with an easy - to - implement , evidence - based curriculum that meaningfully engages students in intellectually and scientifically robust collaborative discussions about extrasolar planets .
this article focuses on one of these new _ lecture - tutorials _ , which addresses the detection of extrasolar planets via gravitational microlensing . while the design and development process for this new _ lecture - tutorial _ mirrored the process we used in the development of previous _
_ lecture - tutorials__@xcite , the topic of gravitational lensing posed an interesting challenge .
this topic is traditionally taught using representations ( written , mathematical , graphical , etc . )
that require a level of understanding of physics and mathematics appropriate for advanced undergraduates or graduate students .
how could we design an activity that engages astro 101 students , who do not possess the mathematical preparation necessary to interact with these representations , to do more than simply recall declarative knowledge ?
while over 35% of astro 101 students in a national survey@xcite listed their major as being in science , engineering , or architecture , many of these students are at the beginning of their undergraduate careers and are years away from being able to take a course in general relativity .
the other 65% of astro 101 students are majors in the arts , humanities , social sciences , education , and professional fields ( e.g. , business , nursing , etc.)@xcite ; instruction that is at too high or too low of a cognitive level may adversely affect their overall attitudes toward science , which has important ramifications given that they represent the nation s future journalists , historians , business leaders , teachers , politicians , parents , taxpayers , and voters . in order to create an activity that inspires and engages astro 101 students at an appropriate cognitive level , we had to judiciously develop a constellation of representations appropriate for the population of interest .
each representation we created has its own pedagogical value and set of constraints .
when working in concert together , these representations can be very effective at elevating students conceptual understandings and reasoning abilities related to the detection of extrasolar planets via gravitational microlensing .
our ultimate goal for the new _ lecture - tutorial _ is to enable astro 101 students to engage in more expert - like reasoning about a general relativistic phenomenon and develop a greater discipline fluency . in this article
, we describe how we selected and designed the representations we used in the _ lecture - tutorial _ detecting exoplanets with gravitational microlensing . "
we first delve into the mathematics underlying gravitational microlensing and illustrate why we needed to emphasize certain aspects of our representations and deemphasize others in order to make it possible for students to develop physically correct understandings .
we then use the theoretical framework of linder ( 2013)@xcite and fredlund et al .
( 2014)@xcite to analyze the affordances of the different representations and to explain how they work together to aid in the development of students discipline fluency . in this paper , the affordances " of a representation are defined as the elements of disciplinary content that students are able to access and reason about using the representation .
this article provides a unique example of how to present advanced topics , such as general relativity , to a novice audience .
additionally , this article details the difficult considerations that stem curriculum developers must work through as they try to select and design appropriate new pedagogical representations that engage students at deep and meaningful levels .
before discussing the details of gravitational microlensing , we wish to introduce a key concept in observational astronomy : the light curve .
a light curve is a graph of the observed brightness of an astronomical body as a function of ( in this case ) time .
light curves have many applications ; for example , the decay rate of a supernova s light curve is an important factor in identifying the type of supernova . as another example , periodic dips " in the light curve of a star can reveal the presence of planets transiting the star as they orbit , as studied by the recent kepler mission . because of their broad applicability as well as their relative approachability for a novice , we suggest that the light curve is an excellent representation for teaching about a variety of astrophysical phenomena .
when a single object , such as star , acts as a gravitational microlens for a distant star , we observe a time - symmetric light curve that shows how the image of the distant star brightens and then dims over the course of the microlensing event .
microlensing causes the solid angle subtended by the star to change .
this change in solid angle affects the number of photons we receive from the star , leading to the observed change in its apparent brightness .
when the lensing system includes both a star and a planet , the planet may also microlens the distant star , creating a perturbation to the light curve ; see han ( 2007 ) for a classification of the different types of planetary light curve perturbations.@xcite every representation has its limitations , and those limitations may be explicit or implicit . instructors and curriculum designers must be aware of these limitations or else they run the risk of encouraging students to draw physically inappropriate conclusions from a given representation .
for example , one might be tempted to reason as follows : more massive planets will cause greater distortions in the geometry of spacetime , so one should be able to detect more photons coming to earth and for a longer period of time than for cases with less massive planets .
this reasoning would then lead one to conclude that both the width and the height of the planetary perturbations in the light curves ( such as those in figures 6 - 9 ) are both correlated with the planet s mass . yet in the _ lecture - tutorial _ on this topic , we use only the width of perturbations in the light curve to reason about the masses of the objects causing them . to motivate this and other such pedagogical decisions
, we must employ the mathematics of general relativity to describe how a single object lenses the light emitted by a more distant source .
the purpose of the following mathematical treatment is to unpack for the reader the disciplinary content knowledge relevant to a full understanding of the physics underlying the phenomenon of microlensing .
we have chosen to use the symbolic mathematical representations below because the readership of the _ american journal of physics _ possesses sufficient mathematical sophistication to access their meaning .
this enables us to bring to light for the reader the origins of the quantitative relationships we are interested in teaching students to reason about in an expert - like manner .
we wish to emphasize that while the results derived here motivate the pedagogical choices we make as curriculum designers , we do not use the mathematical representation directly in our curriculum as astro 101 students do not have the prerequisite knowledge for unpacking this representation .
we used hartle s text , _ gravity : an introduction to einstein s general relativity _ , to inform our mathematical understanding of gravitational lensing.@xcite for the present discussion , we wish to derive the answer to the question : how do lens mass and lens alignment affect the light curve of the distant light source ?
the basic outline of this derivation is that we use the bending angle of light rays to determine the change in solid angle subtended by the image of a distant source when its light is lensed .
this change in the apparent area of the object will lead to a corresponding change in the apparent brightness .
the physical mechanism at work is that the gravitational field of an object between earth and the distant source will curve the trajectories of photons , which then arrive at earth from various angles .
this set of angles results in an image that is distorted from its original size and shape .
note that we use gravitational units ( @xmath0 ) , which are commonly used in general relativistic calculations .
this choice results in mass and length having the same units , allowing direct comparison between quantities of mass and distance . since we are interested in the detection of extrasolar planets by microlensing , we assume a pointlike lensing object of mass @xmath1 .
the deflection angle @xmath2 for a light ray , which is the difference in angle between its initial and final trajectories , passing by the mass with an impact parameter @xmath3 is @xmath4 where @xmath5 is the schwarzschild radius of the lensing object .
most of the bending of the light ray takes place over a distance of order @xmath5 .
because interstellar distances are vastly larger than this length scale , it is an excellent approximation to treat the trajectory of the light ray as being straight everywhere except at a single point , where the line will bend through the full angle @xmath2 .
this approximation is called the thin lens approximation .
we can now solve the problem using only euclidean trigonometry , following the diagram shown in figure 1 . in figure 1 , the angles are all extremely small , so @xmath6 for the angles @xmath2 , @xmath7 , @xmath8 .
these approximations give the horizontal distances labeled at the top of figure 1 , from which we write @xmath9 next , we substitute @xmath10 and @xmath11 , then rearrange to find : @xmath12 we now have an expression for the positions @xmath8 at which the images of the source appear in the sky .
both sides of this equation have units of radians squared , making it reasonable to define an angle @xmath13 : @xmath14 with this definition , we can solve for the two values of @xmath8 more easily and find that @xmath15 the physical meaning of @xmath13 can be found by considering the case of perfect alignment between source , lens , and observer .
substituting @xmath16 into our solution yields @xmath17 .
thus , the angle @xmath13 is the angular radius of the ring of images formed in the case of perfect alignment ( an einstein ring ) .
note that the radius of the einstein ring is directly proportional to the square root of the mass of the lens : @xmath18 .
this will be important in answering the original question of how the shape of the microlensing light curve depends on the lens mass .
the observed brightness of the image of the light source depends only on its intrinsic brightness and the solid angle it subtends .
therefore , the ratio of the brightness of the lensed image to the brightness of the unlensed image is just the ratio of the solid angles subtended by the images .
the intrinsic brightness of the object is not affected by the gravitational lensing , so it cancels out in this ratio .
this cancellation will be another important factor in our analysis of the light curve .
as such , @xmath19 where @xmath20 and @xmath21 are the lensed and unlensed brightnesses , and @xmath22 and @xmath23 are the solid angles subtended by the lensed and unlensed images , respectively . using the previous result for @xmath24we can calculate how the solid angle subtended by an image will change when it is lensed .
figure 2 shows the original image of an object , and then the two lensed images , as seen from the observer s point of view . since the angles @xmath7 and @xmath8 are very small , we will again use the approximation @xmath25 for these angles .
the formulas for solid angle then become @xmath26 @xmath27 we can substitute these expressions into the brightness ratio : @xmath28 notice that @xmath29 will cancel out the azimuthal angle subtended by the original image does not affect the ratio of the lensed and unlensed brightnesses . in the limit of small lensing effects ,
this simplifies further : @xmath30 the second term can be found by differentiating our solution for the angles @xmath24 at which the lensed images appear with respect to the source angle : @xmath31 the first term can also be expressed in terms of only @xmath7 and @xmath13 by substituting the expression for @xmath24 . combining these equations and simplifying yields : @xmath32 this shows that the outer image is brighter than the original , and the inner image is dimmer than the original . in the case of microlensing
, we can not distinguish the lensed images from each other .
instead , we simply detect the sum of the light from the images .
the ratio of the total microlensed brightness to the original brightness is therefore @xmath33 rewriting the two terms with a common denominator allows us to combine them into a single expression : @xmath34 one can instead write the equation in terms of a dimensionless parameter @xmath35 : @xmath36 now we address the original question : how do lens mass and lens alignment affect the light curve of the distant light source ? we need to consider both the width and the height of the peak .
this can be done by making @xmath7 ( or equivalently , @xmath37 ) a function of time in the above equations .
this function starts at a maximum , decreases to some minimum value , and then increases back to a maximum .
recall that the angular radius of the einstein ring is proportional to the square root of the lens s mass : @xmath18 .
the width ( or duration ) of the peak increases as the mass of the lensing object increases .
to first order , the duration of the lensing event is the time for which the distant source is within an angular distance @xmath13 of the lens , as seen from earth .
a more massive lens has a larger value for @xmath13 , and so it will take a longer amount of time for the relative motion of the source and lens to traverse this distance , for a given velocity .
we therefore say that a more massive lens causes a wider peak .
the height of the peak that is , the ratio of the maximum brightness to the unlensed brightness also increases as the mass of the lensing object increases .
this occurs when the lens and source are most closely aligned when seen from earth . in terms of the variables in our equations
, this is when @xmath38 ( or when @xmath39 ) . examining this in terms of either formulation
should give the same result . at this time , @xmath40 given this value for @xmath41 , increasing the mass of the lens will increase @xmath13 .
the einstein angle is a quadratic term in the numerator , and we know that @xmath18 , so the numerator is a linear function of the lens mass .
in contrast , the denominator grows less - than - linearly with the lens mass , since the term inside the square root function is quadratic in @xmath13 , and therefore linear in @xmath1 . as such , the value of the overall fraction should increase as @xmath1 increases if the other variables are held constant .
that is to say , all else being equal , more massive lenses should produce greater brightening effects . the same conclusion can be reached in the dimensionless formulation . for a detectable microlensing event , the distant source will be inside the einstein ring of the lens at the time of minimum separation , so @xmath42 .
consider the dimensionless equation for the maximum brightness : @xmath43 as @xmath44 decreases from 1 to 0 , this fraction grows monotonically toward infinity , demonstrating that more closely aligning the source and lens always increases the change in brightness . since @xmath45 , and @xmath13 grows as the square root of the lens mass , increasing the mass of the lens has the effect of decreasing @xmath37 for a given value of @xmath7 .
qualitatively , one can imagine that adding mass to the lens would expand its einstein ring , and that a source at a fixed location would therefore become relatively closer to being aligned with the center of the lens . in either formulation of the brightness equation
, we find that a greater lens mass causes a greater brightening than a smaller lens mass , assuming that the separation angle as a fraction of the einstein angle ( @xmath46 , aka @xmath37 ) is the same for both masses .
consequently , it is tempting to represent extrasolar planets of different masses with light curve perturbations that differ in both width and height . however , if we are comparing the planetary perturbations on two different light curves , we can not automatically assume that both perturbations correspond to the same value of @xmath37 .
if we want students to use the height of the planetary perturbations to reason about the relative masses of extrasolar planets , then we have to include the caveat that the perturbations always correspond to the same ratio between the separation angle and the einstein angle .
there is an incredible pedagogical risk associated with trying to bring this level of discipline knowledge to the population of students with which we are working . not only would this caveat intellectually over burden many astro 101 students with more novel variables than they have the capacity to simultaneously reason about , it is also not representative of commonly occurring scenarios investigated by active researchers in extrasolar planet detection .
so rather than unnecessarily limit the physical situations we offer learners to only a limited set of geometric configurations which do not fully represent what is found in the universe , we purposefully chose to use only the width of a planetary perturbation , and not its height , to convey information about the extrasolar planet s mass in the _ lecture - tutorial_. note that while some _ lecture - tutorials _ do require students to physically interpret mathematical equations , the detecting exoplanets with gravitational microlensing " _ lecture - tutorial _ does not . while the mathematical representation employed here has rich affordances for a mathematically expert population , astro 101 students do not have the expertise needed to engage with this representation meaningfully . confronting these students with this representation
would therefore not only be ineffective for promoting learning , but rather could lead to a belief that the topic is impenetrable to them and that there is no point in attempting to learn about it .
the purpose of the above derivation was to afford instructors a richer understanding of the effects of the physical variables on the microlensing processes , so that they can have a better understanding of how the relative positions and masses of lensing bodies affect observable astronomical outcomes .
we hope this understanding will enable readers to more effectively engage their learners in appropriate modes of cognition about discipline - specific ideas related to the detection of extrasolar planets via gravitational microlensing .
having said this , we pose to the reader the same challenge we encountered when designing curriculum on the topic .
having followed a mathematical route to deeply understanding gravitational microlensing , and knowing that astro 101 students can not follow the same pathway , how might one effectively teach students to engage in deep , expert - like reasoning about the relationships between astrophysical variables and observable effects in microlensing scenarios ? in the next section , we describe in detail the theoretical pedagogical framework we applied in addressing this challenge .
linder ( 2013)@xcite uses the term disciplinary affordance " to refer to the potential of a representation for providing access to disciplinary knowledge . different representations ( e.g. , written words , equations , diagrams , graphs , etc . )
have different disciplinary affordances , and no single representation by itself captures all aspects of the physical situation it is modeling .
however , multiple representations may work together to create a
collective disciplinary affordance , " providing a more holistic model of the physical situation and more potential access points for the development of disciplinary knowledge and fluency . from this perspective , learning involves recognizing the disciplinary affordances of multiple complementary representations and coordinating the information provided by those representations to reason about novel physical situations .
instructors must strive to create learning environments which enable students to better understand the affordances of multiple discipline representations .
astrophysicists who use gravitational microlensing to detect extrasolar planets use a wide variety of representations in their own work . for the purposes of this section
, we will discuss three representations .
two are now familiar to the reader : light curves , which constitute the fundamental observational evidence for microlensing events , and mathematical equations that astrophysicists use to relate information from these light curves to the physical parameters of the extrasolar planetary system .
the third is a reduced - dimensional pictorial representation of the curved spacetime that determines the paths of lensed light . in this representation ,
the task of depicting the curvature of four - dimensional spacetime is replaced by depicting a two - dimensional sheet " that is deformed and then projected onto the page .
since reasoning about and depicting higher - dimensional structures is very difficult cognitively and artistically , this reduced - dimensional representation ( hereafter rdr ) is an efficacious proxy that allows for appropriate qualitative reasoning about the curvature of spacetime .
these representations are powerful , but fredlund et al .
( 2014)@xcite raise an important caveat : they have evolved through a process of rationalization , " meaning they have become densely packed with knowledge that was developed over an extended period of time and through a series of preceding arguments .
the graphical , pictorial and mathematical representations used to describe gravitational microlensing constitute dense packages of information , including trigonometry , impact parameters , the physics of lenses , flux , and a conceptual model of gravitation as the curvature of spacetime , to name just a few .
this dense packaging of information presents a significant challenge to astro 101 students who are attempting to interpret these representations .
furthermore , students studying the detection of extrasolar planets via gravitational microlensing must reckon with the idea that we are detecting light that comes neither from the planet itself nor from the star it orbits , but rather from another , even more distant star .
all of these factors combined make the detection of extrasolar planets by gravitational microlensing an incredibly intellectually challenging topic for astro 101 students to reason about .
the solution , according to fredlund et al . , is to help students unpack " these representations so that students may
come to ` see ' the parts of intended meaning that are not directly discernible in the representation . "
note that unpacking a representation often involves utilizing or even creating simpler representations .
fredlund et al .
attribute the effectiveness of many per - validated instructional techniques to the fact that they unpack and disambiguate the disciplinary affordances of many physics representations.@xcite in practice , it is often necessary to set a foundation for students to unpack a complex representation by first teaching them simpler representations that afford them access to the content embedded in the more complex representation .
for example , the motion in two dimensions " tutorial from the university of washington s _ _ tutorials in introductory physics__@xcite ultimately guides students to unpack the mathematical representation of the centripetal acceleration vector .
however , the tutorial can only accomplish this by first getting students to represent kinematic quantities by drawing and combining vector arrows .
this representation is both approachable to novices yet still rich enough to be used in discourse between experts . in this example , vector arrows are a type of representation whose affordances are apparent to a large enough population of students that they can be productively used by novices to unpack more complex representations .
but it is not always the case that the set of existing representations typically used by experts within a discipline contains the complete set of representations needed to unpack a sophisticated discipline - specific representation for a novice .
instructors and curriculum developers may find they have to incorporate novel representations to help students unpack more advanced representations in the process of developing more expert - like discipline fluency . in our work to make advanced astrophysics topics more accessible to the astro 101 population , we have found it necessary to create many representations that would not typically be found in a textbook or used by two experts in the discipline who are engaged in discourse with one another on the topic .
rather , these novel representations are especially tailored for teaching and learning about a discipline - specific topic , and so we call them pedagogical discipline representations " ( pdrs ) . in some cases pdrs
are significantly simplified versions of representations experts might use in a colloquium talk when discussing a topic ; in others they can be highly contextualized representations with unique features that purposefully unpack specific discipline concepts , engage novice learners pre - existing mental models , and promote critical discourse . in the detecting exoplanets with gravitational microlensing " _
lecture - tutorial _
, we use a mix of pdrs and representations that would be familiar to microlensing experts .
we carefully chose each representation so that its affordances would be accessible to students and so that it would help them improve their abilities to reason about microlensing and extrasolar planet detection . a typical implementation of this _ lecture - tutorial _
, following the procedure detailed in prather et al .
( 2004)@xcite , is as follows : first , the instructor teaches a short lesson ( @xmath47 15 minutes ) on the topic , after which students work in small peer groups on the _ lecture - tutorial_. in completing the _ lecture - tutorial _ , students collaboratively discuss and reach consensus on a series of carefully sequenced and scaffolded conceptual questions .
students are able to seek help from the instructor and/or teaching assistants , who engage the student learning groups in socratic dialogue . afterward , the instructor debriefs the activity by eliciting students questions and engaging in a class - wide discussion of their answers . following the debrief , the instructor can check students understanding by administering a series of formative assessment questions designed for the specific _ lecture - tutorial_. figure 3 shows the first representation students encounter in the _ lecture - tutorial_. the arrows represent the paths of some photons emanating from a distant star . as figure 3 shows , those paths may be altered by something in a hidden region of spacetime between the distant star and earth .
this pdr affords students the idea that different regions of spacetime can alter the path light takes by different amounts . from this representation
, students can start to reason about how a distant star may appear brighter to an observer on earth when more photons are re - directed ( or lensed ) toward earth .
this establishes an important conceptual foundation students need in order to understand why the brightness of a distant star may change due to gravitational microlensing .
additionally , it helps students begin to understand that the light we detect comes from the distant star , and not from the objects in the stellar system responsible for lensing that distant star s light .
the first representation in the _ lecture - tutorial _ shows students how a region of spacetime ( hidden by the black boxes ) may deflect light by different amounts . ]
we deliberately hid the objects in this region of spacetime with a black box in the figure 3 pdr .
we could have included the representations of the lensing objects masses and the corresponding curvature of spacetime in these regions , which would have enabled the pdr in figure 3 to convey more information about the connection between mass , spacetime curvature , and the lensing of the light rays .
however , decades of research show that students are most likely to develop deep and long - lasting knowledge when they are actively engaged in the construction of that knowledge.@xcite restricting the amount of information displayed in figure 3 protects students from needing to simultaneously engage in two novel tasks , namely reasoning about the bending of light rays and processing the rdr .
this restriction therefore reduces the cognitive load experienced by students , enabling them to more effectively engage in constructing their own understanding about the relationship between the degree of bending of light rays and the amount of spacetime curvature ( and , hence , the mass present ) in a region of spacetime . after developing this understanding
, students can then apply their full cognitive resources to engaging with the _ lecture - tutorial _ s use of the rdr , as shown in figure 4 .
this figure presents three different rdrs of the curvature of spacetime for the region between the distant star and earth , but not the paths taken by light emitted by the distant star .
students are asked to match the cases shown in figure 4 with those shown in figure 3 and to rank the different representations of curved spacetime based on the amount of mass they must contain . this reduced - dimensional representation ( rdr ) from the _ lecture - tutorial _
shows the amount of curvature of the region of spacetime between the distant star and earth . ]
used together in this way , the representations provided in figures 3 and 4 help students construct a mental model of the gravitational microlensing process for the case in which the lensing system comprises a single mass and when there is no relative motion between that mass , the distant star , and earth . in reality , gravitational microlensing events are transient phenomena and the lensing system may contain multiple masses .
consequently , we use the representation in figure 5 to introduce and illustrate the dynamic ( as opposed to static ) nature of a microlensing event in which both an extrasolar planet and its parent star lens the light of a distant star .
this representation provides students the opportunity to begin to reason about which object ( the star or planet ) creates a larger warping of spacetime .
it also provides students key information they will need in order to determine which of the light curves shown in figure 6 would result from this particular arrangement of planet and parent star .
the _ lecture - tutorial _ is designed to help students reason as follows : since the planet will move through the region of spacetime between the distant star and earth before the parent star , the increase in brightness of the distant star must occur at an earlier time than the brightening of the distant star resulting from the lensing caused by the parent star .
using these pdrs , the _ lecture - tutorial _ guides students to conclude that graph d in figure 6 corresponds to the situation shown in figure 5 . creating and understanding a complete , correct mapping between these two representations is a critical step in a student s development of a mental model of the microlensing process . this representation from the _
lecture - tutorial _ helps students visualize the movement of an extrasolar planet and its parent star relative to a distant star , as seen from earth . ]
students must determine which of these two light curves corresponds to the situation depicted in fig .
figure 7 shows four different light curves . at this point in the _ lecture - tutorial _
, students have progressed through a carefully sequenced set of questions that guide their thinking to investigate and interpret the pdrs in figures 3 - 6 .
the tasks presented thus far provide students with the foundational discipline knowledge necessary to recognize that each of the light curves in figure 7 must be due to a system of one extrasolar planet and its parent star , which each lens the light of a distant star .
they are asked to rank the relative masses of the extrasolar planets based on the widths of the perturbations in the corresponding light curves of figure 7 .
they are also asked to determine in which cases the extrasolar planet was located to the left or right of its parent star , as seen from earth , during the microlensing event .
four light curves that students reason about in terms of the mass of the detected extrasolar planet and whether or not it is located to the left or right of its parent star , as seen from earth . ]
figure 8 shows two additional light curves .
these curves differ in the temporal separation of their planetary perturbations .
this affords students the opportunity to reason about which light curve must be due to the stellar system in which the extrasolar planet was at the greater physical distance from its parent star , as seen from earth .
two light curves students reason about to determine which of the detected extrasolar planets has the greatest separation from its parent star . ]
the final question of the _ lecture - tutorial _ asks students to reason about the stellar system that must have been responsible for producing the light curve shown in figure 9 .
this question probes whether students have a complete and coherent mental model for the full breadth of physical characteristics of the extrasolar planet system that this light curve represents . after working through all preceding components of the _ lecture - tutorial _
, students should be able to discern how the features of this representation affords information about the number of extrasolar planets in the stellar system , the relative separation of those planets from their parent star , and the relative masses of those planets .
figure 10 shows an analogous question that we use as part of a post-_lecture - tutorial _ homework assignment ; in this question , students must sketch a light curve based on the stellar system that is shown .
our post - instruction assessments of student understanding show that most of the astro 101 students participating in this work
the majority of whom are non - stem majors@xcite are able to successfully complete the reasoning tasks shown in figures 9 and 10 . in the final question of the _ lecture - tutorial _ , students must reason about the number of extrasolar planets in the stellar system , the relative separation of those planets from their parent star , and the relative masses of those planets . ] a question from the _ lecture - tutorial _
s associated homework that asks students to sketch the light curve that the stellar system would produce when it lenses the light of a distant star . ]
while our assessment of students post - instruction understandings of this topic are ongoing , we have found that between 70 - 85% of students can correctly answer multiple - choice midterm and final exam questions that are modeled after the _ lecture - tutorial _ and homework questions in figures 9 and 10 .
these values are consistent with the results from multiple studies conducted over the past fifteen years on how student understanding of difficult - to - understand astrophysics topics is significantly elevated above what is achieved in lecture - based classrooms when they engage with research - based active learning curricula.@xcite we attribute the greater discipline fluency that these students attain to their cognitive engagement with the _ lecture - tutorial _ s carefully targeted sequence of socratic - style questions , which guide students to unpack the content afforded to them by the pdrs used in figures 3 - 9 .
gravitational microlensing is an important application of einstein s general theory of relativity to extrasolar planet detection , which is at the forefront of modern astrophysics .
we have developed a new _ lecture - tutorial _
, detecting exoplanets with gravitational microlensing , " that instructors can use to actively engage their astro 101 students .
this paper details the various representations the _ lecture - tutorial _ employs to help students develop their abilities to reason about this topic . while some of these representations are familiar to experts in the field of extrasolar planet detection ( e.g. , light curves ) ,
others are novel representations that we created specifically for the purposes of teaching and learning , and which we refer to as pedagogical discipline representations " ( pdrs ) .
we used the theoretical framework of linder ( 2013)@xcite and fredlund et al .
( 2014)@xcite to analyze the affordances of the _ lecture - tutorial _ s representations and to justify why we selected this particular set and sequence of representations .
we also discussed the limitations of some representations , which underscores the importance of instructors understanding which conclusions can and can not be drawn from a given representation .
one component of an expert instructor s pedagogical content knowledge@xcite is a knowledge of the affordances and limitations of different pdrs and other representations ; such knowledge enables an instructor to select the most pedagogically appropriate representations to help students efficiently develop greater discipline fluency with a given topic .
we hope that by illustrating our use of this theoretical framework for choosing and developing representations , other instructors and curriculum developers will be able to take complex , advanced topics , such as gravitational microlensing , and translate them into curricula that meaningfully engage and elevate the reasoning abilities of introductory students .
the development of the detecting exoplanets with gravitational microlensing " lecture - tutorial and accompanying curricula was funded by the generous contributions of nasa s exoplanet exploration program . i. a. bond , a. udalski , m. jaroszynski , n. j. rattenbury , b. paczynski , i. soszynski , l. wyrzykowski , m. k. szymanski , m. kubiak , o. szewczyk , k. zebrun , g. pietrzynski , f. abe , d. p. bennett , s. eguchi , y. furuta , j. b. hearnshaw , k. kamiya , p. m. kilmartin , y. kurata , k. masuda , y. matsubara , y. muraki , s. noda , k. okajima , t. sako , t. sekiguchi , d. j. sullivan , t. sumi , p. j. tristram , t. yanagisawa , p. c. m. yock , and the moa and ogle collaborations , ogle 2003-blg-235/moa 2003-blg-53 : a planetary microlensing event , " _ ap .
j. lett .
_ , * 606 * , l155-l158 ( 2004 ) .
committe on undergraduate physics education research and implementation , board on physics and astronomy , and division on engineering and physical sciences , _ adapting to a changing world challenges and opportunities in undergraduate physics education _
( the national academies press , washington , dc , 2013 ) .
s. freeman , s. l. eddy , m. mcdonough , m. k. smith , n. okoroafor , h. jordt , and m. p. wenderoth , active learning increases student performance in science , engineering , and mathematics , " _ proc .
_ , * 111 * , 8410 - 8415 ( 2014 ) . s. r. singer , n. r. nielsen , and h. a. schweingruber ( eds . ) , _ discipline - based education research : understanding and improving learning in undergraduate science and engineering _ ( the national academies press , washington , dc , 2012 ) . e. e. prather , a. l. rudolph , g. brissenden , and w. m. schlingman , a national study assessing the teaching and learning of introductory astronomy . part i. the effect of interactive instruction , " _ am
. j. phys .
_ , * 77 * , 320 - 330 ( 2009 )
. e. e. prather , t. f. slater , j. p. adams , j. m. bailey , l. v. jones , and j. a. dostal , research on a lecture - tutorial approach to teaching introductory astronomy for non - science majors , " _ astron .
_ , * 3 * , 122 - 136 ( 2004 ) . c. s. wallace , e. e. prather , and d. k. duncan ,
a study of general education astronomy students understandings of cosmology . part v. the effects of a new suite of cosmology _ lecture - tutorials _ on students conceptual knowledge , " _ int .
j. sci .
_ , * 34 * , 1297 - 1314 ( 2012 ) .
a. l. rudolph , e. e. prather , g. brissenden , d. consiglio , and v. gonzaga , a national study assessing the teaching and learning of introductory astronomy .
part ii . the connection between student demographics and learning , " _ astron .
rev . _ * 9 * , 010107 ( 2010 ) . k. e. williamson , development and calibration of a concept inventory to measure introductory college astronomy and physics students understanding of newtonian gravity , " ph
. d. thesis , department of physics , montana state university , 2013 . | the detection and study of extrasolar planets is an exciting and thriving field in modern astrophysics , and an increasingly popular topic in introductory astronomy courses .
one detection method relies on searching for stars whose light has been gravitationally microlensed by an extrasolar planet . in order to facilitate instructors abilities to bring this interesting mix of general relativity and extrasolar planet detection into the introductory astronomy classroom
, we have developed a new _ lecture - tutorial _
, detecting exoplanets with gravitational microlensing . " in this paper
, we describe how this new _ lecture - tutorial _
s representations of astrophysical phenomena , which we selected and created based on theoretically motivated considerations of their pedagogical affordances , are used to help introductory astronomy students develop more expert - like reasoning abilities . |
more than 30 years ago , refsdal ( 1964 , 1966 ) pointed out that the differential light propagation time delay between two or more gravitationally lensed images of a background object establishes an absolute physical distance scale ( @xmath8 ) in the system .
thus , the distance to a high - redshift object is directly measured , yielding a value of hubble s constant , @xmath4 .
the theory of this technique has been elaborately developed and its realization has become a major focus of gravitational lens studies [ narayan ( 1991 ) gives an especially elegant treatment ; see blandford & narayan ( 1992 ) and narayan & bartelmann ( 1996 ) for reviews ] .
it may be useful to briefly review the main strengths of the lensing method in determination of the extragalactic distance scale : 1 .
it is a geometrical method based on the well understood and experimentally verified physics of general relativity in the weak - field limit .
by contrast , most conventional astronomical techniques for measuring extragalactic distances rely either on empirical relationships or on our understanding of complex astrophysical processes , or both . 2 .
it provides a direct , single step measurement of @xmath4 for each system and thus avoids the propagation of errors along the `` distance ladder '' which is no more secure than its weakest rung .
it measures distances to cosmologically distant objects , thus precluding the possibility of confusing a local with a global expansion rate .
note that both observed cmb fluctuations and cobe normalized numerical simulations of large - scale structure formation suggest the possibility of 10 - 20% _ rms _ expansion rate fluctuations even on scales of order 10,000 km / s for some cosmological models ( d. spergel 1996 , private communication ; turner , cen & ostriker 1992 ) .
independent measurement of @xmath4 in two or more lens systems with different source and lens redshifts allows a powerful internal consistency check on the answer .
although an inaccurate model of the lens mass distribution or other systematic problems could yield an incorrect distance in any particular system , no known or imagined problem will consistently give _ the same _ wrong answer when applied to different lenses .
thus , if a small number of time delay measurements all give the same @xmath4 , this value can be regarded as correct with considerable confidence . despite these potent virtues , the practical realization ( _ i.e. _ , at a useful and competitive accuracy ) of refsdal s method for measuring @xmath4
has proven quite challenging and has been long delayed . for the lens system 0957 + 561a , b
( walsh , carswell & weymann 1979 ) , by far the best studied case , there are two basic reasons .
first , there has been sufficient ambiguity in detailed models of the mass distribution in the lensing galaxy and associated cluster to allow values of @xmath4 different by a factor of two or more to be consistent with the same measured time delay ( young et al . 1981 ; narasimha , subramanian & chitre 1984 ; falco , gorenstein & shapiro 1991 ; kochanek 1991 ) ; fortunately , this problem has been much alleviated by recent theoretical and observational work .
see [ model.sec ] for details .
second , despite extensive optical ( lloyd 1981 ; keel 1982 ; florentin - nielsen 1984 ; schild & cholfin 1986 ; vanderriest et al .
1989 ; schild & thomson 1995 ) and radio ( lehr et al . 1992 ; haarsma et al .
1996 ) monitoring programs extending over a period of more than 15 years , values of the differential time delay discrepant by more than 30% have continued to be debated in the literature . in particular , most studies have given a delay either in the range 400420 days or one of about 530540 days .
these two rough values , the `` short delay '' and the `` long delay '' have been obtained _ both by applying the same statistical techniques to different data sets and by applying different statistical techniques to the same data _ [ vanderriest et al .
1989 ; lehr et al .
1992 ; press , rybicki & hewitt 1992a , 1992b ( hereafter collectively referred to as prh ) ; pelt et al .
1994 , 1996 ( hereafter collectively phkrs ) ; beskin & oknyanskij 1995 ] !
moreover , even application of a single technique to a single ( radio ) data set has produced best estimate delays that move far outside the nominal formal high confidence interval as additional points in the light curve accumulate ( press et al .
1992b , haarsma et al .
1996 ) . the history of the 0957 + 561a , b time delay , which can certainly be described as confusing and controversial , is reviewed by haarsma et al .
( 1996 ) . in this paper we report a robust determination of the time delay which we believe effectively resolves the controversy in favor of the short delay .
in addition , we use this delay and the results of recent theoretical ( grogin & narayan 1996 , hereafter gn ) and observational ( garrett et al .
1994 ) studies of the lens mass distribution to calculate a global measure of hubble s constant of accuracy comparable to that of the best conventional techniques , both in terms of statistical and systematic errors .
our time delay determination differs from all previous ones in that the appearance of a sharp , large amplitude feature in the @xmath0 band light curve of the trailing image ( 0957 + 561b ) during 1996 was _ predicted in advance _ based on observations of the leading image ( 0957 + 561a ) .
the light curve data showing this sharp @xmath0 band event , plus other weaker features in the @xmath0 and @xmath1 band light curves , is given in kundi et al .
( 1995 , hereafter paper i ) along with predictions of when it would appear in the b image during 1996 for either the short or long delay .
this paper reports 1996 data showing that the short time delay prediction was quantitatively correct while the long time delay prediction is not even qualitatively consistent with the 1996 image b light curve .
oscoz et al .
( 1996 ) have very recently employed the paper i predictions to exclude a delay longer than 500 days , a result entirely consistent with those presented here . [ data.sec
] presents the 1996 light curve data for both images in which the short delay predictions are clearly confirmed . [ stat.sec ] presents a set of statistical analyses of the light curves from which we derive a best fit value of the delay and estimate its uncertainty . in
[ model.sec ] , we derive a global value of hubble s constant @xmath4 from the measured delay and discuss its statistical and systematic uncertainties . finally , in [ summary.sec ] , we comment on some other implications of the data and on the general situation in attempts to apply refsdal s method for determining @xmath4 .
the 0957 + 561 photometric monitoring program at the apache point observatory ( apo ) 3.5 meter telescope which is described in paper i for the late 1994 to mid-1995 season ( hereafter referred to as the 1995 season ) was continued in the late 1995 to mid-1996 season ( hereafter designated as the 1996 season ) . the instrumentation , filter system , observing protocols , and data reduction techniques were as described in paper i except that the large majority of the 1996 observations were made from the apo site rather than remotely from princeton .
the 1996 light curves consist of 51 @xmath0 band and 54 @xmath1 band brightness measurements made from 25 november 1995 to 6 july 1996 , as compared to the 46 @xmath0 and 46 @xmath1 band data points obtained from 2 december 1994 to 27 may 1995 in the 1995 season . in 1996 ,
the mean photometric error of a nightly measurement was 10 millimag in @xmath0 band and 9 millimag in @xmath1 band , compared to respective values of 12 and 12 in the previous season .
the light curves for the two seasons are thus of quite similar quality .
figure [ alldata.fig ] displays our @xmath0 and @xmath1 data for images a and b in the 1995 and 1996 seasons , in other words our complete photometric data set .
this data is available in the elt/:0957 subdirectory of the anonymous ftp area at astro.princeton.edu .
the predictions of the short time delay set forth in paper i can be confirmed by inspection of fig .
[ alldata.fig ] without need for the detailed statistical analysis presented in the next section .
the sharp event which paper i noted in the @xmath0 band light curve of image a in early 1995 is marked in fig .
[ alldata.fig ] , as are the times of its predicted appearance in the 1996 image b light curve .
an event of just the expected amplitude and duration is seen in image b at the time corresponding to the short delay , and no such feature is seen at a time corresponding to the long delay .
although we will proceed to a quantitative analysis below , the qualitative success of the short delay prediction is compelling .
note that fig .
[ alldata.fig ] also displays the rather quiescent behavior of image a in 1996 , which corresponds to the predicted 1997 behavior of image b , modulo the now determined delay and magnitude offset .
it also emphasizes the good fortune of the occurrence of the sharp @xmath0 band event in image a shortly after the beginning of our monitoring program ; no comparable feature has been seen since ! also note that photometric events in the @xmath1 band are consistently smaller in amplitude than those in the @xmath0 band . since most previous monitoring programs have concentrated their attention in red bands ( probably to minimize problems with moonlight ) , this may explain some of the difficulty in obtaining a definitive time delay measurement .
in this section , we present a series of statistical comparisons of our 1995 image a light curve with the 1996 image b light curve .
the goals of these analyses are 1 ) to determine the best estimate differential time delay for 0957 + 561a , b ; 2 ) to quantitatively assess its error ( 95% confidence interval ) ; and 3 ) to demonstrate that the data provide a _ robust _ delay measurement .
this last point deserves brief elaboration : previous attempts to measure the 0957 + 561a , b time delay from radio and optical data have often depended as much on the statistical techniques employed as upon the data , _
i.e. _ , different methods gave different answers when applied to the same data ( see references in [ intro.sec ] ) .
moreover , even the best delay as determined by a single method was unstable against the inclusion or exclusion of a few data points in some cases ( phkrs ) .
we demonstrate below that the data shown in fig .
[ alldata.fig ] produce no such difficulties or ambiguities . table 1 and fig . [ chi2.fig ]
summarize the results of our statistical comparisons of the 1995 image a light curve with the 1996 image b data .
we employ four different statistical methods on four subsets of the complete data set and four parameterized models of how the image a and b light curves are related , although we do not present all 64 of the possible combinations here . of the at least 11 different statistical analyses previously applied to 0957 + 561a ,
b optical and radio light curves by various authors ( tabulated by haarsma et al .
1996 ) , we have selected four representative ones intended to span the range of reasonable approaches for analysis of our data . 1 .
linear method : first and simplest , each light curve and its errors are linearly interpolated , and the data points from the other light curve are shifted so as to minimize @xmath9 ( @xmath10 per degree of freedom ) .
the number of degrees of freedom is equal to the number of overlapping data points in the shifted light curves minus the number of fitted parameters .
time delays with a small overlap of data are thus penalized with respect to delays where the overlap is significant . the 95% confidence interval quoted in table 1 for each fitted parameter
is derived by bootstrap resampling of residuals from the combined light curve smoothed by a 7-point triangular filter with a maximum length of 28 days .
the confidence interval is robust with respect to the choice of the smoothing filter . 2 .
prh method : as emphasized by press et al .
( 1992a , b ) , interpolation of the light curves is a crucial ingredient in any method , and we adopt their rather elegant , though somewhat assumption laden , interpolation scheme plus @xmath10 minimization as our second method .
the prh method relies heavily on a model of the structure function of the source s intrinsic variability .
we have taken a ( measured ) structure function of @xmath11 = -4.3 + 0.86\tau$ ] ; @xmath12 = -4.5 + 0.83\tau$ ] , where @xmath13 is in days and @xmath14 is in magnitudes squared .
the quoted error on the prh scaled @xmath10 method is the @xmath15 interval of @xmath16 .
while the association of a 95% ( or any other specific ) confidence interval with this @xmath15 interval is problematic , it provides a rough estimate of the measurement error ; we refer readers to prh for further discussion .
3 . phkrs method : as an alternative to @xmath10 minimization and as a more assumption - free approach , we adopt a non - parametric technique suggested by phkrs as a third method . in particular , we use their dispersion measure @xmath17 from the second paper with linearly decreasing weights @xmath18 and decorrelation length of @xmath19 days . here
again 95% confidence intervals quoted in table 1 are derived from the bootstrap resampling of residuals .
cross - correlation method : finally , we have employed a conventional cross - correlation analysis , similar to that used by schild & cholfin ( 1986 ) , vanderriest et al . ( 1989 ) , schild ( 1990 ) , and lehr et al . ( 1992 ) , as our fourth method ; this technique also requires interpolation but seems to be rather insensitive to its details . here
we report cross - correlation results using only the simplest linear interpolation scheme .
the errors in table 1 for this method are estimated 1@xmath7 bounds on the position of the correlation peak based on a bootstrapping analysis .
the statistical methods described above have been applied to four subsets of our light curve data .
first , we have considered only the @xmath0 band data ; we believe that this gives the best and most reliable delay due to the high signal - to - noise provided by the sharp event .
second , we have treated the @xmath1 band data separately ; it provides a partially independent , though lower signal - to - noise , delay determination .
third , we have combined the @xmath0 and @xmath1 data to find overall best fit parameter values .
fourth , we have used only the jd 24496892449731 interval of the 1995 image a light curve , _
i.e. _ , the sharp @xmath0 band event alone , to search the 1996 image b light curve for a best fit .
this last subsampling of the data corresponds most closely to testing the paper i prediction and is not an _ a posteriori _
( an hence invalid ) editing of the data only because of that prediction .
the simplest model of the mapping of the a image light curve into that of the b image is just a fixed offset in time , @xmath20 , and a fixed magnitude offset , @xmath21 .
these two parameters correspond to the macrolensing differential delays and image magnifications .
we also consider models in which the magnitude offset is allowed to be a linear function of time at a rate @xmath22 , motivated by the possibility that one or both images are being affected by microlensing events with characteristic time scales long compared to the delay and the extent of the observed light curves ( chang & refsdal 1979 ; young 1981 ; gott 1981 ; falco , wambsganss & schneider 1991 ) .
finally , we know that the image b photometry is contaminated by the light of the lensing galaxy g1 , especially in the @xmath1 band due to the galaxy s red color relative to the quasar source . thus , we consider a parameter @xmath23 which represents a constant flux added to that of the source in image b. the general form of our model of the relation of the image a light curve to that of image b is , therefore @xmath24 } + 10^{-0.4 m_{\rm gal } } \quad , \ ] ] where @xmath25 , @xmath26 are the measured image a and b magnitudes ; @xmath27 is the time delay ; @xmath21 , @xmath22 are the magnitude offset and its rate of change ; @xmath28 = jd 2450000 ( arbitrarily chosen ) ; and @xmath23 is the magnitude of the lensing galaxy g1 which contaminates image b photometry . as shown in table 1 , we have experimented with models in which up to three parameters are allowed to vary as well as ones in which the less interesting ones , @xmath22 and @xmath23 are neglected , that is fixed at zero and infinity , respectively .
table 1 shows the results of a selected set of combinations of our statistical methods , light curve mapping models and versions of the data .
all 64 possible combinations are not shown for the sake of clarity and because some of the statistical methods and some versions of the data do not lend themselves well to the determination of some parameters .
however , we emphasize that we are aware of no combination of method , data subset and model which gives a @xmath20 value significantly different from those shown in table 1 .
[ chi2.fig ] displays the variation of the minimized statistical parameter with respect to @xmath20 for the @xmath0 band light curves with each of our four methods . in this figure microlensing and galaxy terms
are not included in the fits . the most important conclusion to be drawn from table 1 and fig .
[ chi2.fig ] is that the best fit value of @xmath20 is _ extremely robust _ , changing by less than 1%
whatever method , data version or parameterized model is used .
the delay error estimates produced by the various statistical analyses are somewhat more varied , with those methods that assume greater knowledge of the statistical character of the light curves naturally producing smaller error estimates , but in no case is the estimated error greater than 3% .
as we shall see in the next section , the error in the measured delay now contributes negligibly to the error in the deduced global value of @xmath4 .
because it is simple and seems to work as well as any other approach for this data set , we adopt the results of the linear method applied to the @xmath0 band light curves with just a @xmath29 model as our final numbers .
this gives a delay of @xmath2 days and a @xmath0 band magnitude offset of @xmath30 millimag ( b is brighter than a ) .
[ linear.fig ] displays this fit and the corresponding independent fit ( 420 days , @xmath31 millimag ) for the @xmath1 band data ; it also shows the cleanly rejected fit which corresponds to the long delay of about 540 days in the figure insets .
in addition to providing best fits and error estimates , the prh method produces an optimal ( under the assumed statistical properties of the source light curve ) reconstruction of the underlying single light curve sampled in the two images , and a `` snake '' corresponding to its uncertainty at each point .
this reconstruction and its comparison to the data are shown in fig .
[ prh.fig ] .
for a complex gravitational lens system such as 0957 + 561 , reliable measurement of the differential time delay may still leave one far from the goal of a good measurement of @xmath4 ( _ e.g. _ , schechter et al .
1996 ) , simply because of uncertainties in the lens mass distribution and its implied conversion factor between @xmath8 and the source angular diameter distance ( falco , gorenstein & shapiro 1991 ; kochanek 1991 ) .
this was once a major obstacle , but fortunately an extensive recent theoretical modeling study of 0957 + 561 by gn appears to have resolved the major ambiguities with the exception of one effect noted by falco , gorenstein & shapiro ( 1985 ) ; and gorenstein , falco & shapiro ( 1988 ) ; see below .
in particular , gn use 15 known observables to constrain a set of lens mass models having 9 free parameters . the relation between time delay and distance
is controlled primarily by the total projected mass of the lens within the circle defined by the diameter connecting the two source images @xmath32 and by its derivative @xmath33 ( chang & refsdal 1976 ; refsdal 1992 ; wambsganss & paczyski 1994 ) .
gn find that the observed image separation @xmath34 and the observed mapping of the complex 5-component structure in the vlbi images of the image a and b ( garrett et al .
1994 ) nearly fix @xmath32 and @xmath33 at the image positions , respectively , independent of other features of the lens model . for example , compared to their best fit model ( which models g1 as a softened power - law sphere ) , gn find that the best - fit king model ( after falco et al . 1991 ) with a compact nucleus gives only a 3% higher value of @xmath4 despite introducing a central black hole in g1 with a mass of @xmath35 . including ellipticity in g1 at the observed value of @xmath36 and major axis position angle of @xmath37 ( bernstein , tyson & kochanek 1993 ) increases @xmath4 by 2% . adding perturbations from the two closest cluster galaxies
reduces it by 4% .
moreover , changing @xmath38 from unity , the value otherwise assumed throughout this paper , to 0.1 increases @xmath4 by only 7% . introducing the cosmological constant while keeping the universe flat results in an increase of just 4% in the @xmath39 , @xmath40 model .
unfortunately , the distance conversion factor supplied by the gn models remains quite sensitive to one remaining degeneracy which can not be removed by the observed properties of the lensing event itself ( falco et al .
1985 , gorenstein et al .
1988 ) , namely how much of the lensing is contributed by mass associated with the galaxy g1 versus how much is supplied by the associated and superimposed galaxy cluster .
since these two must sum to a known value ( in order to produce the observed splitting ) , the degeneracy may be parameterized by either the one or the other .
gn adopt @xmath41 , the central line - of - sight velocity dispersion of g1 , to measure its contribution and @xmath42 , the dimensionless lensing convergence contributed by the cluster , to parameterize its effect . in order to derive an @xmath4 value
, we need to independently measure one or the other parameter . measuring both
provides an internal consistency check .
fischer et al .
( 1996 ) have recently reported a mapping of the cluster mass distribution based on the observed distortion of faint background galaxies , a technique which is now well developed and has been successfully applied to several other galaxy clusters ( _ e.g. _ tyson , valdes & wenk 1990 ; tyson & fischer 1995 ; squires et al.1996a , b ) .
unfortunately , the mass profile reported by fischer et al.is centered on g1 , rather than the center of mass in the field located some @xmath43 northeast . since the distortion of background galaxies at the two quasar image positions is dominated by the mass of g1 , it is difficult to independently estimate the contribution of the cluster .
this would have been possible if the cluster profile had been extracted around its observed center and the region around g1 excluded from the weak lensing analysis .
assuming circular symmetry , one could have then estimated @xmath42 at the radius corresponding to g1 from the uncontaminated area in the cluster .
one can still make an indirect estimate of @xmath42 from the total mass of the cluster , its core radius and the location of g1 with respect to the cluster center .
following gn and kochanek ( 1991 ) , we assume that the cluster potential corresponds to a softened isothermal sphere @xmath44 , where @xmath45 is the core radius and @xmath46 is the critical radius of the cluster expressed in terms of its velocity dispersion @xmath47 .
this potential results in a density profile slightly different from that of fischer et al .
( 1996 ) , but well within their observational uncertainty . the local convergence at a position @xmath1 in the cluster is then given by @xmath48 .
using direct spectroscopy of cluster galaxies , angonin - willaime , soucail & vanderriest ( 1994 ) find @xmath49 km / s . the cluster mass estimate from weak lensing , @xmath50 ( fischer et al .
1996 ) , implies a slightly higher velocity dispersion of @xmath51 km / s .
we thus adopt @xmath52 km / s as our 95% confidence region . using @xmath53 for the distance of g1 from the center of the cluster and @xmath54 for the cluster core radius ( as measured by fischer et al . ) , our final estimate for the cluster surface density at the position of the lensing event is @xmath55 ( 2@xmath7 ) .
this result is insensitive to the precise value of @xmath45 , but it could increase considerably if g1 were much closer to the center of the cluster . at that point , however , the quadratic approximation of the cluster potential ( which only includes the convergence and shear ) would break down as well . thus ,
including first only the quoted errors in the gn model and those of our time delay and then adding the dominant contribution from the @xmath42 uncertainty , we find @xmath56 where the errors reflect the total 95% confidence interval .
the primary remaining systematic uncertainty in this calculation is associated with the mean redshift of the background galaxy population used to map the cluster mass distribution , somewhat arbitrarily assumed to be 1.2 by fischer et al .
if it were allowed to vary between 0.75 and 3 , the derived value of @xmath4 would range between 58 and 68 km / s / mpc , respectively .
considering very extreme possibilities , @xmath4 would vary from 37 to 70 km / s / mpc for mean background galaxy redshifts between 0.5 and infinity .
falco et al .
( 1997 , as quoted by m. davis 1996 , private communication ) recently obtained a new measurement of the velocity dispersion in g1 from a high signal - to - noise keck lris spectrum .
although there are some puzzling features of their data , a surprisingly rapid variation in @xmath41 along the slit in particular , they obtain a value roughly in the range @xmath57 km / s ( 2@xmath7 ) .
this is much more accurate than but consistent with the only previously reported value of @xmath58 km / s ( rhee 1991 ) , which was based on a significantly lower signal - to - noise spectrum . inserting this new value of @xmath41 and again propagating all the relevant errors , we find @xmath59 where we again quote 2@xmath7 error intervals .
the gn model error , included in our @xmath4 uncertainty , allows for finite aperture effects and possible anisotropy of stellar orbits .
it is important and reassuring that these two entirely independent methods of resolving the galaxy - cluster degeneracy in 0957 + 561 give very similar and entirely consistent results .
for example , if one assumes an extremely low value of 0.5 for the mean redshift of background galaxies in the @xmath42 method , this would yield a very small value of @xmath4 ( see above ) , but would also contradict the falco et al .
( 1996 ) result by implying a g1 velocity dispersion of only 210 km / s .
we can not properly average ( 2 ) and ( 3 ) since they are based on the same lens mass model ( _ i.e. _ , are not entirely independent ) , but they can be approximately summarized by @xmath5 km / s / mpc at 95% confidence , or @xmath60 10% at 1@xmath7 .
our main results may be summarized as follows : the sharp photometric event predicted in the 1996 0957 + 561b light curve in paper i based on the observed 1995 image a light curve has been observed at the time corresponding to the so called short ( about 420 day ) delay , thus confirming its validity .
no such event was observed at a time corresponding to the long ( 540 day ) delay , thus rejecting it . the time delay determined by the data presented here and in paper
i is quite robust and has a best fit value of @xmath2 days . combining this value with the latest theoretical models of the lens mass distribution plus two independent measurements of the relative contributions of the galaxy g1 and
the associated galaxy cluster indicates that the global value of @xmath4 lies in the interval 51 to 77 km / s / mpc with a most probable value in the range 5870 km / s / mpc .
although the possibility of significant systematic error remains , especially in the lens mass distribution model , _ this result is likely as accurate and reliable as the best conventional measures of _ @xmath4 .
though still a significant potential source of systematic error , there is reason to believe that further refinements of the lens mass distribution model will have little effect on the derived value of @xmath4 , as argued in detail at the beginning of [ model.sec ] , following gn .
the point is that the constraints provided by the detailed matching of the a and b vlbi images ( garrett et al .
1994 ) appear to sufficiently constrain the critical physical properties of lens models , particularly the gradient of the projected lens mass distribution at the image positions , that even models which differ substantially in their other properties will yield nearly the same value of @xmath4 ( chang & refsdal 1976 ; refsdal 1992 ; wambsganss & paczyski 1994 ) .
these critical vlbi constraints were not available to earlier modelers , _ e.g. , _
falco et al .
( 1991 ) , who derived values of @xmath4 varying from the gn results by up to 50% for a fixed delay and g1 velocity dispersion .
of course , further exploration of the space of all possible lensing models of 0957 + 561a , b will be useful to determine just how much the derived @xmath4 value can be changed within the constraints of the vlbi maps ; the extensive modeling by gn was not exhaustive of all possibilities .
first , note in figures 3 and 4 that the time shifted and magnitude offset light curves of images a and b do not appear to match perfectly , even though they are strikingly similar . in particular , the image b points fall a few hundredths of a magnitude below ( fainter than ) those of the shifted and offset image a data during the interval jd 2450160 to 2450175 and then even more slightly above them during 2450175 to 2450220 .
we have made no test of the statistical significance of this discrepancy , but no acceptable delay and offset removes it , nor have we been able to identify any plausible shortcoming or anomaly in the data for either image during the intervals in question which might be blamed for a spurious effect .
it is tempting to speculatively attribute the mismatch to microlensing perturbations of the macrolensing event since one expects a substantial microlensing optical depth for at least image b and since claims of such effects have been made before based on earlier light curve data ( schild & smith 1991 , schild & thomson 1995 , schild 1996 ) .
the timescale of the mismatch ( tens of days ) is far shorter than that expected for strong microlensing perturbations by roughly stellar mass objects , or even massive planets , in the low optical depth limit ( years to decades , young 1981 , gott 1981 ) , but these are not large amplitude events and the low optical depth limit probably does not apply .
the small magnitude of these short time scale mismatches and the null detection of a slow offset drift @xmath22 , see table 1 , may be more conservatively interpreted as _
upper limits _ on microlensing effects in the present data .
we may return to this issue in a separate paper . in any case , these minor and marginally detected features do not appear to compromise or complicate the time delay measurement .
second , the shifted and offset light curves of images a and b combine to give an intriguing record of the quasar source s intrinsic variability , possible small microlensing perturbations aside .
see figures 3 and 4 .
such data may contain important clues to the quasar s detailed internal physics , the functioning of its accretion disk in the conventional agn model .
although we have not yet carried out any such analyses , it is apparent that one might ask many interesting statistical question of this and similar quasar light curves .
third , peebles et al .
( 1991 ) and dar ( 1991 ) point out that gravitational lensing events confirm that high - redshift galaxies and quasars are at cosmological distances . the precise repetition of the 1995 image a light curve by image b in 1996 ( to roughly 1% accuracy ) based on an _ a priori _ prediction provides the best such available case .
it both unambiguously proves that 0957 + 561a , b is a _ bona fide _ case of gravitational lensing and gives an order - of - magnitude , model - independent measure of the source distance @xmath61 gpc .
turning finally to a more general perspective , 0957 + 561a , b appears to have provided a first and so far best success for refsdal s method of measuring @xmath4 , but it is clear that there are good prospects of both improving its accuracy and extending its applications to other gravitational lens systems . towards the former goal ,
improved measurements of g1 s position and velocity dispersion @xmath62 and of the galaxy cluster s mass distribution are clearly possible ; some are already underway .
direct measurement of redshifts of the distorted background galaxy images will be challenging but would remove a major systematic uncertainty .
further theoretical exploration of the lens mass model space is also desirable to verify the uniqueness of the gn model .
rapid progress can also be expected using other lens systems .
recently , schechter et al .
( 1996 ) have reported a convincing , though only moderately accurate , pair of time delays in the 1115 + 080a , b , c system . the accuracy can almost certainly be improved .
it remains to be seen whether or not sufficiently unique and well constrained models of this system can be obtained to allow useful @xmath4 measurements .
systems such as b0218 + 357 , for which a very rough estimate of the delay has been obtained from radio polarimetry ( corbett et al .
1996 ) should be powerful for distance determinations because their resolved structure provides strong model constraints .
the rate of discovery of new lens systems is likely to increase , so we may expect yet more suitable rosetta stone " systems to be found in the future . as noted in [ intro.sec ] ,
no known or imagined effect or systematic error would cause different lens systems to give _ the same _ wrong answer consistently .
thus , if good time delay measurements and well constrained lens models yield the same value of @xmath4 for a few different lens systems ( with different angular sizes , lens and source redshift , _ et cetera _ ) , the problem of measuring the global value of hubble s constant will have been effectively solved .
we gratefully acknowledge assistance from and informative discussions with r. blandford , m. davis , d. haarsma , j. hewitt , r. narayan , b. paczyski , u. pen , w. press , r. schild and k. stanek .
the long term , frequent sampling photometric monitoring program at apo would not be possible without the hard work and dedication of its staff , for which we are very thankful .
this work was supported by nsf grants ast94 - 19400 and ast95 - 29120 , and nasa grant nag5 - 2759 .
wnc would like to thank the fannie and john hertz foundation for its continued support .
linear & @xmath0 & @xmath20 , @xmath21 & @xmath63 & @xmath64 & & + & & @xmath20 , @xmath21 , @xmath22 & @xmath65 & @xmath66 & @xmath67 & + & & @xmath20 , @xmath21 , @xmath23 & @xmath68 & @xmath69 & & @xmath70 + linear & @xmath1 & @xmath20 , @xmath21 & @xmath73 & @xmath74 & & + & & @xmath20 , @xmath21 , @xmath22 & @xmath75 & @xmath76 & @xmath77 & + & & @xmath20 , @xmath21 , @xmath23 & @xmath78 & @xmath79 & & @xmath80 + prh & @xmath0 & @xmath20 , @xmath21 & @xmath83 & @xmath84 & & + & & @xmath20 , @xmath21 , @xmath22 & @xmath85 & @xmath86 & @xmath87 & + & & @xmath20 , @xmath21 , @xmath23 & @xmath85 & @xmath88 & & @xmath89 + prh & @xmath1 & @xmath20 , @xmath21 & @xmath90 & @xmath91 & & + & & @xmath20 , @xmath21 , @xmath22 & @xmath92 & @xmath93 & @xmath94 & + & & @xmath20 , @xmath21 , @xmath23 & @xmath92 & @xmath95 & & @xmath96 + angonin - willaime , m .- c . ,
soucail , g. , & vanderriest , c. 1994 , , 291 , 411 bernstein , g. m. , tyson , j. a. , & kochanek , c. s. 1993 , , 105 , 816 beskin , g. m. , & oknyanskij v. l. 1995 , , 304 , 341 blandford , r.d . , & narayan , r. 1992 , , 30 , 311 chang , k. , & refsdal , s. 1976 , international cnrs coll . , 263 , 369 chang , k. , & refsdal , s. 1979 , , 282 , 561 corbett , e. a. , browne , i. w. a. , wilkinson , p. n. , & patnaik , a. r. 1996 , in astrophysical applications of gravitational lensing , ed . c. s. kochanek , & j. n. hewitt ( dodrecht : kluwer ) , 37 dar , a. 1991 , , 382 , l1 falco , e. e. , gorenstein , m. v. , & shapiro , i. i. 1985 , , 289 , l1 falco , e. e. , gorenstein , m. v. , & shapiro , i. i. 1991 , , 372 , 364 falco , e. e. , wambsganss , j. , & schneider , p. 1991
, , 251 , 698 fischer , p. , bernstein , g. , rhee , g. , & tyson , j. a. 1996 , aj , in press ( astro - ph/9608117 ) florentin - nielsen , r. 1984 , , 138 , l19 garrett , m. a. , calder , r. j. , porcas , r. w. , king , l. j. , walsh , d. , & wilkinson , p. n. 1994 , , 270 , 457 gorenstein , m. v. , falco , e. e. , & shapiro , i. i. 1988 , , 327 , 693 gott , j. r. 1981 , , 243 , 140 grogin , n. a. , & narayan , r. 1996 , , 464 , 92 & 473 , 570 ( gn ) haarsma , d. b. , hewitt , j. n. , lehr , j. , & burke , b. f. 1996 , , submitted ( astro - ph/9607080 ) keel , w. c. 1982 , , 255 , 20 kochanek , c. s. 1991 , , 382 , 58 kundi , t. , colley , w. n. , gott , j. r. iii , malhotra , s. , pen , u. , rhoads , j. e. , stanek , k. z. , & turner , e. l. 1995 , , 455 , l5 ( paper i ) lehr , j. , hewitt , j. n. , roberts , d. h. , & burke , b. f. 1992 , , 384 , 453 lloyd , c. 1981 , , 294 , 727 narasimha , d. , subramanian , k. , & chitre , s. m. 1984 , , 210 , 79 narayan , r. 1991 , , 378 , l5 narayan , r. & bartelmann m. 1996 , astro - ph/9606001 oscoz , a. , serra - ricart , m. , goicoechea , l. j. , buitrago , j. , & mediavilla , e. 1996 , , 470 , l19 peebles , p. j. e. , schramm , d. n. , turner , e. l. , & kron , r. g. 1991 , , 352 , 769 pelt , j. , hoff , w. , kayser , r. , refsdal , s. , & schramm , t. 1994 , , 286 , 775 ( phkrs ) pelt , j. , kayser , r. , refsdal , s. , & schramm , t. 1996 , , 305 , 97 ( phkrs ) press , w. h. , rybicki , g. b. , & hewitt , j. n. 1992a , , 385 , 404 ( prh ) press , w. h. , rybicki , g. b. , & hewitt , j. n. 1992b , , 385 , 416 ( prh ) refsdal , s. 1964 , , 128 , 307 rhee , g. 1991 , , 350 , 211 schechter , p. l. et al .
1996 , , submitted schild , r. e. 1990 , , 100 , 1771 schild , r. e. 1996 , , 464 , 125 schild , r. e. , & cholfin , b. 1986 , , 300 , 209 schild , r. e. , & smith , r. c. 1991 , , 101 , 813 schild , r. , & thomson , d. j. 1995 , , 109 , 1970 squires , g. , kaiser , n. , babul , a. , fahlman , g. , woods , d. , neumann , d. m. , bhringer , h. 1996a , , 461 , 572 squires , g. , kaiser , n. , fahlman , g. , babul , a. , woods , d. , 1996b , , 469 , 73 turner , e. l. , cen , r. , & ostriker , j. p. 1992
, , 103 , 1427 tyson , j. a. , valdes , f. , & wenk , r. a. 1990 , , 349 , l1 tyson , j. a. , & fischer , p. 1995
, , 446 , l55 vanderriest , c. , schneider , j. , herpe , g. , chevreton , m. , moles , m. , & wlrick , g. 1989 , , 215 , 1 walsh , d. , carswell , r. f. & weymann , r. j. 1979 , , 279 , 381 wambsganss , j. , & paczyski , b. 1994 , , 108 , 1156 young , p. 1981
, , 244 , 756 young , p. , gunn , j. e. , kristian , j. , oke , j. b. , & westphal , j. a. 1981 , , 244 , 736 | continued photometric monitoring of the gravitational lens system 0957 + 561a , b in the @xmath0 and @xmath1 bands with the apache point observatory ( apo ) 3.5 m telescope during 1996 shows a sharp @xmath0 band event in the trailing ( b ) image light curve at the precise time predicted in an earlier paper .
the prediction was based on the observation of the event during 1995 in the leading ( a ) image and on a differential time delay of 415 days .
this success confirms the so called `` short delay '' , and the absence of any such feature at a delay near 540 days rejects the `` long delay '' for this system , thus resolving a long standing controversy .
a series of statistical analyses of our light curve data yield a best fit delay of @xmath2 days ( 95% confidence interval ) and demonstrate that this result is quite robust against variations in the analysis technique , data subsamples and assumed parametric relationship of the two light curves .
recent improvements in the modeling of the lens system ( consisting of a galaxy plus a galaxy cluster ) allow us to derive a value of the global ( at @xmath3 ) value of hubble s constant @xmath4 using refsdal s method , a simple and direct ( single step ) distance determination based on experimentally verified and securely understood physics and geometry .
the result is @xmath5 km / s / mpc ( for @xmath6 ) where this 95% confidence interval is dominantly due to remaining lens model uncertainties .
however , it is reassuring that available observations of the lensing mass distribution over constrain the model and thus provide an internal consistency check on its validity . _
we argue that this determination of the extragalactic distance scale ( 10% accurate at 1@xmath7 ) is now of comparable quality , in terms of both statistical and systematic uncertainties , to those based on more conventional techniques .
_ finally , we briefly discuss the prospects for improved @xmath4 determinations using gravitational lenses and some other possible implications and uses of the 0957 + 561a , b light curves . |
the requirement that preferences are _ coherent _ aims to make rigorous the idea that elementary restrictions on rational preferences entail that personal probabilities satisfy the axioms of mathematical probability .
this use of coherence as a justification of personal probability is very well illustrated by de finetti s ( @xcite ) approach to the foundations of probability .
de finetti distinguished two senses of coherence : coherence@xmath0 and coherence@xmath1 .
coherence@xmath0 requires that probabilistic forecasts for random variables ( he calls them previsions ) do not lead to a finite set of fair contracts that , together , are uniformly dominated by abstaining .
coherence@xmath1 requires that a finite set of probabilistic forecasts can not be uniformly dominated under brier ( squared error ) score by a rival set of forecasts .
he showed that these two senses of coherence are equivalent in the following sense .
each version of coherence results in using the expectation of a random variable as its forecast .
moreover , these expectations are based on a finitely additive probability without requiring that personal probability is countably additive . [ in appendix [ app : lf ] , we explain what we mean by expectations with respect to finitely additive probabilities .
these are similar in many ways , but not identical to integrals in the sense of @xcite , chapter iii . ]
schervish , seidenfeld and kadane ( @xcite ) extended this equivalence to include a large class of strictly proper scoring rules ( not just brier score ) but for events only .
the corresponding extension to general random variables is included in the supplemental article [ schervish , seidenfeld and kadane ( @xcite ) ] . here
, we refer to the extended sense of coherence@xmath1 as coherence@xmath2 .
we investigate asymmetries between coherence@xmath0 and coherence@xmath2 reflecting differences between cases where personal probabilities are countably additive and where personal probabilities are finitely ( but not countably ) additive .
we give conditions where coherence@xmath2 may be applied to assessing countably many forecasts at once , but where coherence@xmath0 can not be applied to combining infinitely many fair contracts .
also , we study conditional forecasts given elements of a partition @xmath3 , where the conditional forecasts are based on the conditional probabilities given elements of @xmath3 . each coherence criterion is violated by combining infinitely many conditional forecasts when those conditional forecasts are not conglomerable ( see definition [ def : nonc ] ) in the partition @xmath3 .
neither criterion is violated by combining infinitely many conditional forecasts when conditional expectations satisfy the law of total previsions ( see definition [ def : ltp ] ) in @xmath3 .
coherence of preference , as de finetti [ ( @xcite ) , chapter 3 ] formulates it , is the criterion that a rational decision maker respects _ uniform ( strict ) dominance_. in section [ sub : dominance ] , we explain the version of the dominance principle that de finetti uses . in section [ sub :
coherence ] , we review de finetti s two versions of coherence , with a focus on how preferences based on a finitely additive probability are coherent .
let @xmath4 be a set .
the elements of @xmath4 will be called _ states _ and denoted @xmath5 .
random variables are real - valued functions with domain @xmath4 , which we denote with capital letters .
let @xmath6 index a set of options .
consider a hypothetical decision problem @xmath7 specified by a set of exclusive options @xmath8 .
each option @xmath9 is a random variable with the following interpretation : if @xmath5 is the state which occurs , then @xmath10 denotes the decision maker s loss ( negative of cardinal utility ) for choosing option @xmath9 .
the values of @xmath11 ( for all @xmath12 ) are defined up to a common positive affine transformation .
let @xmath9 and @xmath13 be two options from @xmath7 . if there exists an @xmath14 such that for each @xmath15 , @xmath16 , then option @xmath9 _ uniformly strictly dominates @xmath13_. if , for each @xmath5 , @xmath17 , we say that @xmath9 _ simply dominates _ @xmath13 .
uniform strict dominance is clearly stricter than simple dominance . as we explain , next , in order to permit preferences based on maximizing finitely ( and not necessarily countably ) additive expectations , de finetti used the following dominance principle , rather than some other more familiar concepts of admissibility , for example , simple dominance .
there are additional ways to define dominance , which we discuss further in section [ sec : summary ] .
dominance principle : let @xmath9 and @xmath13 be options in @xmath18 . if @xmath9 uniformly ( strictly ) dominates @xmath13 , then @xmath13 is an _
inadmissible _ choice from @xmath7 .
de finetti [ ( @xcite ) , chapter 3 ] formulated two criteria of _ coherence _ that are based on the dominance principle . throughout this paper
, we follow the convention of identifying events with their indicator functions .
[ def : coh][def : condcoh ] a _ conditional prevision _ ( or _ conditional forecast _ ) @xmath19 for a random variable @xmath20 given a nonempty event @xmath21 is a fair price for buying and selling @xmath20 in the sense that , for all real @xmath22 , the option that costs the agent @xmath23 $ ] is considered fair .
[ we call @xmath24 an _ unconditional prevision _ and denote it @xmath25 . ] a collection @xmath26 of such conditional forecasts is
_ coherent@xmath0 _ if , for every finite subset @xmath27 and all real @xmath28 , there exists no @xmath29 such that @xmath30 \geq\varepsilon\ ] ] for all @xmath15 . a collection of conditional forecasts is _ coherent@xmath1 _ if no sum of finitely many ( brier score ) penalties can be uniformly strictly dominated in the partition of states by the sum of penalties from a rival set of forecasts for the same random variables .
that is , for every finite subset @xmath27 , all alternative forecasts @xmath31 , and all positive @xmath28 , there is no @xmath29 such that=-1 @xmath32 ^ 2 \geq\sum_{j=1}^n \alpha_jh_{i_j}(\omega)\bigl[x_{i_j}(\omega ) -q_{i_j}\bigr]^2+\varepsilon\]]=0 for all @xmath5 .
de finetti [ ( @xcite ) , pages 8889 ] proved that a decision maker who wishes to be both coherent@xmath0 and coherent@xmath1 must choose the same forecasts for both purposes .
he also proved that the decision maker s coherent@xmath0 forecasts are represented by a finitely additive personal probability , @xmath33 , in the sense of definition [ def : fap ] below . if @xmath34 , then coherence@xmath0 and coherence@xmath1 place no restrictions on @xmath19 for bounded @xmath20 .
nevertheless , it is possible and useful to make certain intuitive assumptions about conditional forecasts given events with 0 probability . in particular , theorems [ thm:3 ] and [ thm:2008b ] of this paper
assume that @xmath35 is a finitely additive expectation ( in the sense of definition [ def : daniell ] in appendix [ app : lf ] ) satisfying @xmath36 for all @xmath21 and @xmath20 .
this assumption holds whenever @xmath37 , and it captures the idea that @xmath35 is concentrated on @xmath21 .
de finetti [ ( @xcite ) , appendix 16 ] introduces an axiom that places a similar requirement on conditional previsions .
see levi ( @xcite ) , section 5.6 , and @xcite for other ways to augment the coherence criteria of definition [ def : condcoh ] in order to satisfy these added requirements on conditional previsions given a null event . rather than adding such requirements to the definition of coherence , we prefer that individual agents who wish to adopt them do so as explicit additional assumptions .
example 2 in the supplemental article [ schervish , seidenfeld and kadane ( @xcite ) ] illustrates our reason for such a preference . in this way ,
our definition of coherence is slightly weaker than that of de finetti . as an
aside , the meaning of conditional expected value in the finitely - additive theory differs from its meaning in the countably - additive theory in this one major regard : in the finitely - additive theory a conditional expectation can be specified given an arbitrary nonempty event , regardless of whether that event has positive probability . a conditional expectation of a bounded random variable given an event with zero probability is not defined uniquely in terms of unconditional expectations , but @xcite shows that , in the finitely additive theory , conditional expectations can be defined on the set of bounded random variables so that they are finitely additive expectations . in the countably - additive theory , conditional expectation is defined twice : given events with positive probability and given @xmath38-fields .
the two definitions match in a well - defined way , and both provide uniquely defined conditional expectations in terms of unconditional expectations .
[ def : fap ] a probability @xmath33 is _ finitely additive _ provided that , when events @xmath39 and @xmath40 are disjoint , that is , when @xmath41 , then @xmath42 .
a probability is _ countably additive _ provided that when @xmath43 is a denumerable sequence of pairwise disjoint events , that is , when @xmath44 if @xmath45 , then @xmath46 . we call a probability @xmath47 _ merely finitely additive _ when @xmath47 is finitely but not countably additive .
likewise , then its @xmath47-expectations are merely finitely additive . for each pair
@xmath20 and @xmath48 of random variables with finite previsions ( expectations ) , @xmath49 . for countably additive expectations and countably many random variables @xmath50 , conditions under which @xmath51 can be derived from various theorems such as the monotone convergence theorem , the dominated convergence theorem , fubini s theorem and tonelli s theorem .
de finetti ( @xcite ) recognized that coherence@xmath1 ( but not coherence@xmath0 ) provided an incentive compatible solution to the problem of mechanism design for eliciting a coherent set of personal probabilities .
specifically , brier score is a _ strictly proper scoring rule _ , as defined here .
[ def:4 ] a _ scoring rule _ for coherent forecasts of a random variable @xmath20 is a real - valued loss function @xmath52 with two real arguments : a value of the random variable and a forecast @xmath53 .
let @xmath54 be the collection of probability distributions such that @xmath25 is finite and @xmath55 $ ] is finite for at least one @xmath53 .
we say that @xmath52 is _ proper _ if , for every probability @xmath56 , @xmath55 $ ] is minimized ( as a function of @xmath53 ) by @xmath57 .
if , in addition , only the quantity @xmath57 minimizes expected score , then the scoring rule is _ strictly proper_.
the following trivial result connects proper scoring rules with conditional distributions . [
pro : src ] if @xmath21 is a nonempty event and @xmath35 is a probability distribution then @xmath58 $ ] is ( uniquely ) minimized by @xmath59 if @xmath52 is ( strictly ) proper .
some authors reserve the qualification strictly proper for scoring rules that are designed to elicit an entire distribution , rather than just the mean of a distribution .
[ see gneiting ( @xcite ) , who calls the latter kind
_ strictly consistent_. ] for the remainder of this paper ,
we follow the language of definition [ def:4 ] , which matches the usage in gneiting ( @xcite ) .
we present some background on strictly proper scoring rules in section [ sec : scoring ] .
section [ sec : extensions ] gives our main results .
we discuss propriety of scoring rules for infinitely many forecasts in section [ sec : proper ] .
in this section , we introduce a large class of strictly proper scoring rules that we use as generalizations of brier score . associated with this class , we introduce a third coherence concept that generalizes coherence@xmath1 .
let @xmath60 be a class of strictly proper scoring rules .
let @xmath61 be a collection of random variable / nonempty event pairs with corresponding conditional forecasts @xmath62 .
the forecasts are__coherent@xmath2 relative to @xmath60 _ _ if , for every finite subset @xmath63 , every set of scoring rules @xmath64 , and every set @xmath65 of alternative forecasts , there is no @xmath29 such that @xmath66 for all @xmath5 .
that is , no sum of finitely many scores can be uniformly strictly dominated by the sum of scores from rival forecasts .
coherence@xmath1 is the special case of coherence@xmath2 in which @xmath60 consists solely of brier score .
the supplemental article [ schervish , seidenfeld and kadane ( @xcite ) ] includes a proof that , if @xmath60 consists of strictly proper scoring rules of the form ( [ eq : gdef ] ) below , then coherence@xmath2 relative to @xmath60 is equivalent to coherence@xmath0 .
the general form of scoring rule that we will consider is @xmath67 where @xmath68 is a measure that is mutually absolutely continuous with lebesgue measure and is finite on every bounded interval .
it is helpful to rewrite ( [ eq : gdef ] ) as @xmath69 using the convention that an integral whose limits are in the wrong order equals the negative of the integral with the limits in the correct order . another interesting way to rewrite ( [ eq : gdef ] ) , using the same convention , is @xmath70,\ ] ] where , for all @xmath71 and @xmath72 , @xmath73 an immediate consequence of ( [ eq : glin ] ) is that , if @xmath74 and @xmath53 are real numbers , then @xmath75.\ ] ] the form ( [ eq : gdef ] ) is suggested by equation ( 4.3 ) of @xcite .
each such scoring rule is finite , nonnegative and continuous as a function of @xmath76 .
if we wanted to consider only countably additive distributions , we could use a larger class of scoring rules by allowing @xmath68 to be an infinite measure supported on a bounded interval @xmath77 .
but this relaxation would allow functions @xmath52 that are not strictly proper for natural classes of finitely additive distributions .
example [ exa : log ] below illustrates this point .
lemma [ lem : sp ] justifies the use of ( [ eq : gdef ] ) as the form of our scoring rules .
the proofs of all results in the body of the paper are given in appendix [ app : aux ] .
[ lem : sp ] let @xmath52 be a scoring rule of the form ( [ eq : gdef ] ) .
then @xmath52 is strictly proper .
it follows from ( [ eq : glin ] ) that , if @xmath68 is a probability measure with finite mean , then @xmath54 from definition [ def:4 ] is the class of all finitely additive distributions with finite mean because @xmath78 and @xmath79 are both bounded functions of @xmath53 and @xmath80 . even if @xmath68 is not a finite measure , ( [ eq : diff ] ) implies that , if @xmath81 $ ] is finite , then @xmath82 is linear in @xmath80 so that @xmath83 is also strictly proper with @xmath84 equal to the class of all probabilities with finite mean .
for example , if @xmath85 , namely brier score , then @xmath54 is the set of distributions with finite second moment .
however , @xmath86 has @xmath84 equal to the class of all probabilities with finite mean .
let @xmath87 denote the radon nikodym derivative of @xmath68 with respect to lebesgue measure .
some familiar examples of strictly proper scoring rules are recovered by setting @xmath88 equal to specific functions .
brier score corresponds to @xmath89 .
logarithmic score on the interval @xmath77 corresponds to @xmath90 $ ] , but the corresponding measure is infinite on @xmath77 .
hence , logarithmic score is not of the form ( [ eq : gdef ] ) .
in addition , if @xmath52 is this logarithmic score , then @xmath54 does not include all finitely additive distributions that take values in the bounded interval @xmath77 , as the following example illustrates .
[ exa : log ] let @xmath20 be a random variable whose entire distribution is agglutinated at @xmath91 from above .
that is , let @xmath92 and @xmath93 for all @xmath29 .
let @xmath52 be the logarithmic scoring rule that uses @xmath94 from above .
then @xmath95 , but @xmath96 for all @xmath5 , which could not have finite mean even if we tried to extend the definition of random variables to allow them to assume infinite values . on the other hand , for @xmath97 ,
the mean of @xmath98 is @xmath99>0 $ ] , which decreases to 0 as @xmath53 decreases to @xmath91 , and is always finite .
so , @xmath54 is nonempty but does not contain @xmath47 .
some of our results rely on one or another condition that prevents the @xmath68 measures that determine the scoring rules from either being too heavily concentrated on small sets or from being too different from each other .
let @xmath100 be a collection of strictly proper scoring rules of the form ( [ eq : gdef ] ) with corresponding measures @xmath101 .
suppose that , for every @xmath102 , there exists @xmath103 such that for all @xmath12 and all real @xmath104 , @xmath105 implies @xmath106 .
then we say that the collection @xmath60 satisfies the _ uniform spread condition_. suppose that , for every @xmath29 and every @xmath12 , there exists @xmath107 such that for all @xmath108 and all real @xmath104 , @xmath109 implies @xmath110 .
then we say that the collection @xmath60 satisfies the _ uniform similarity condition_. the @xmath111 in ( [ eq : ra2 ] ) can be thought of as the mean of the probability measure on the interval @xmath112 obtained by normalizing @xmath68 on the interval .
the uniform spread condition insures that , the @xmath68 measures are spread out enough to keep the means of the normalized measures on intervals far enough away from both endpoints .
the next result gives sufficient conditions for both the uniform similarity and uniform spread conditions .
it is easy to see that the conditions are logically independent of each other .
[ lem : ubf ] let @xmath100 be a collection of strictly proper scoring rules of the form ( [ eq : gdef ] ) with corresponding measures @xmath101 and corresponding radon
nikodym derivatives @xmath113 with respect to lebesgue measure .
assume that there exists @xmath114 such that @xmath115 , for all @xmath116 and all @xmath12 .
then @xmath60 satisfies the uniform spread condition .
assume that for every @xmath12 , there exists @xmath117 such that @xmath118 , for all @xmath116 and all @xmath108 .
then @xmath60 satisfies the uniform similarity condition . as an example , suppose that each @xmath119 is @xmath120 times lebesgue measure .
if the @xmath121 are bounded above , then @xmath60 satisfies the uniform spread condition .
if the @xmath121 are bounded away from 0 , then @xmath60 satisfies the uniform similarity condition .
these sets of measures correspond to multiples of brier score .
there are collections that satisfy the uniform spread condition without satisfying the conditions of part ( i ) of lemma [ lem : ubf ] .
for example , let @xmath122 which is not bounded above . for this @xmath88 , we have @xmath123 if @xmath124 , and @xmath125 if @xmath126 .
so @xmath127 is no larger than two times the distance between @xmath71 and @xmath72 . also , @xmath111 is always at least @xmath128 of the way from both @xmath71 and @xmath72 .
we can add the corresponding scoring rule to any class that already satisfies the uniform spread condition by ( if necessary ) lowering @xmath129 to @xmath130 .
in section [ sub : dominance3 ] , we investigate when each sense of coherence can be extended to allow combining countably many forecasts into a single act by summing together their individual outcomes . in section [ sub :
dominance4 ] , we introduce the concept of _ conditional _ forecasts and present results about the combination of countably many coherent conditional forecasts . let @xmath131 be a countable set of random variables with corresponding coherent@xmath0 unconditional previsions @xmath132 . let @xmath133 be a sequence of real numbers .
the decision maker s net loss in state @xmath5 , from adding the individual losses from the fair options @xmath134 $ ] is @xmath135.\ ] ] similarly , if the agent s prevision @xmath136 for @xmath137 is scored by the strictly proper scoring rule @xmath138 for each @xmath139 , the total score in each state @xmath5 equals @xmath140 we assume that each of the two series above are convergent for all @xmath15 . [ examp:1 ] de finetti [ ( @xcite ) , page 91 ] noted that when the decision maker s personal probability is merely finitely additive , she / he can not always accept as fair the countable sum ( [ eq : infsumprev ] ) determined by coherent@xmath141 forecasts . that sum may be uniformly dominated by abstaining .
let @xmath142 be a countable state space .
let @xmath143 be the indicator function for state @xmath144 if @xmath145 and @xmath146 if @xmath147 .
consider a collection of merely finitely additive coherent@xmath0 forecasts @xmath148 where @xmath149 .
so @xmath33 is not countably additive . with @xmath150 , for all @xmath139 ,
the loss from combining these infinitely many forecasts into a single option is uniformly positive , @xmath151 = ( 1-c)>0.\ ] ] hence , the decision maker s alternative to abstain , with constant loss @xmath152 , uniformly strictly dominates this infinite combination of fair options .
if , on the other hand , the decision maker s personal probability @xmath47 is countably additive , then @xmath153 . for arbitrary @xmath133
such that @xmath154 is defined and finite , the sum of losses is @xmath155 = \alpha_{i(\omega)}-d,\ ] ] where @xmath156 is the unique @xmath139 such that @xmath157 . because @xmath153 , there is at least one @xmath158 and at least one @xmath159 , hence ( [ eq : gca ] ) must be nonpositive for at least one @xmath139 , and abstaining does not uniformly strictly dominate .
next , we focus on the parallel question whether a coherent@xmath2 set of forecasts remains undominated when strictly proper scores for countably many forecasts are summed together . some conditions will be needed in order to avoid @xmath160 arising in the calculations , and these are stated precisely in the theorems .
the principal difference between dominance for infinite sums of forecasts and dominance for infinite sums of strictly proper scores is expressed by the following result .
[ thm:1 ] let @xmath60 be a collection of strictly proper scoring rules of the form ( [ eq : gdef ] ) that satisfies the uniform spread condition .
let @xmath47 be a coherent@xmath2 prevision defined over a collection @xmath161 of random variables that contains all of the random variables mentioned in the statement of this theorem .
let @xmath131 be random variables in @xmath161 with coherent@xmath2 forecasts @xmath162 for @xmath163 assume that the forecast for @xmath137 will be scored by a scoring rule @xmath164 for each @xmath139 .
finally , assume that @xmath165 & = & v < \infty\quad\mbox{and } \\
\label{eq2 } p \biggl[\sum_{i=1}^\infty g_i(x_i , p_i ) \biggr ] & = & w < \infty.\end{aligned}\ ] ] there does not exist a rival set of forecasts @xmath166 such that , for all @xmath15 , @xmath167 theorem [ thm:1 ] asserts conditions under which infinite sums of strictly proper scores , with coherent@xmath2 forecasts @xmath168 for @xmath50 , have no rival forecasts that simply dominate , let alone uniformly strictly dominate @xmath169 .
that is , even countably many unconditional coherent@xmath2 forecasts can not be simply dominated under the conditions of theorem [ thm:1 ] .
[ examp:3 ] recall that @xmath170 is a countable space . consider the special case in which @xmath47 is a purely finitely additive probability satisfying @xmath171 , for all @xmath139 .
so , @xmath172 , and @xmath173 in the notation of example [ examp:1 ] .
as before , let @xmath174 ( @xmath175 ) be the indicator functions for the states in @xmath4 .
so @xmath176 and combining the losses @xmath177 , for @xmath175 results in a uniform sure - loss of @xmath178 .
but this example , with each @xmath138 equal to brier score times @xmath120 , satisfies the conditions of theorem [ thm:1 ] , if the @xmath121 are bounded above .
that is , there are no rival forecasts @xmath166 for the @xmath179 that simply dominate the forecasts @xmath169 by weighted sum of brier scores , let alone uniformly strictly dominating these forecasts .
we can illustrate the conclusion of theorem [ thm:1 ] directly in this example .
the weighted sum of brier scores for the @xmath136 forecasts is=-1 @xmath180=0 let @xmath166 be a rival set of forecasts with @xmath181 for at least one @xmath139 .
the corresponding weighted sum of brier scores is @xmath182 ^ 2=s(\omega ) -2\sum_{i=1}^\infty \alpha_iq_iw_i(\omega)+\sum _ { i=1}^\infty\alpha_iq_i^2.\ ] ]
let @xmath183 , which must be strictly greater than 0 .
define @xmath156 to be the unique value of @xmath139 such that @xmath157 .
the right - hand side of ( [ eq : exampsum ] ) can then be written as @xmath184 .
if @xmath185 , then the rival forecasts clearly fail to dominate the original forecasts .
if @xmath186 , then @xmath187 . because the @xmath121 themselves are bounded , it follows that all but finitely many @xmath188 are less than @xmath189 . for each @xmath5
such that @xmath190 , we have the weighted sum of brier scores displayed in ( [ eq : exampsum ] ) strictly greater than @xmath191 , hence the rival forecasts do not dominate the original forecasts .
theorem [ thm:1 ] , as illustrated by example [ examp:3 ] , shows that the modified decision problem in de finetti s prevision game modified to include infinite sums of betting outcomes is not isomorphic to the modified forecasting problem under strictly proper scoring rules modified to include infinite sums of scores .
in particular , abstaining from betting , which is the alternative that uniformly dominates the losses for coherence@xmath0 , is not an available alternative under forecasting with strictly proper scores . in summary , the two criteria , coherence@xmath0 and coherence@xmath2 behave differently when probability is merely finitely additive and we try to combine countably many forecasts.=1 we conclude this section with an example to show why we assume that the class of scoring rules satisfies the uniform spread condition in theorem [ thm:1 ] .
this example satisfies all of the conditions of theorem [ thm:1 ] except that the class of scoring rules fails the uniform spread condition .
we show that the conclusion to theorem [ thm:1 ] also fails . for each integer @xmath192 ,
let @xmath193 , and define @xmath194 let @xmath119 be the measure whose radon nikodym derivative with respect to lebesgue measure is @xmath195 , and define @xmath138 by ( [ eq : gdef ] ) using @xmath196 .
the form of @xmath138 is as follows : @xmath197 these scoring rules fail the uniform spread condition because arbitrarily short intervals with both endpoints positive have arbitrarily large @xmath119 measure as @xmath139 increases .
let @xmath198 be a partition of the real line , and let @xmath33 be a finitely additive probability such that @xmath199 for all @xmath139 .
for each integer @xmath192 , let @xmath200 and @xmath201 , and define @xmath202 it follows that @xmath203 for all @xmath139 , and @xmath204=p \biggl [ \sum_{i=1}^\infty a_i|q_i-1-p_i| \biggr]=1,\ ] ] so that ( [ eq9 ] ) holds .
next , compute the various scores : @xmath205 , \\ & & g_i\bigl(x_i(\omega),q_i \bigr)\\ & & \qquad = a_i^c(\omega)\frac{1}{\alpha_i}(p_i - q_i)^2 + a_i(\omega ) ( q_i-1-q_i)^2 \\ & & \qquad = a_i^c(\omega)2^{-i-1}+a_i ( \omega).\end{aligned}\ ] ] define @xmath206 for that unique @xmath139 such that @xmath207 .
when we sum up the scores for the forecasts @xmath169 , we get @xmath208=3 + 2^{-i(\omega)-1}.\ ] ] it follows that @xmath209 , so ( [ eq2 ] ) holds .
the sum of the @xmath166 scores is @xmath210=1.5 - 2^{-i(\omega)-1}.\ ] ] finally , compute the difference in total scores : @xmath211 hence the scores of the @xmath169 forecasts are uniformly strictly dominated by the scores of a set of rival forecasts .
definition [ def : condcoh ] allows mixing conditional forecasts with unconditional forecasts by setting @xmath212 whenever @xmath213 .
de finetti showed that , if @xmath25 , @xmath19 and @xmath214 are all specified , a necessary condition for coherence@xmath0 is that @xmath215 so that @xmath19 is the usual conditional expected value of @xmath20 given @xmath21 whenever @xmath37 .
for this reason , conditional forecasts are often called _
conditional expectations_. the concept of _ conglomerability _ plays a central role in our results about coherence for combining countably many conditional forecasts .
conglomerability in a partition @xmath216 of conditional expectations @xmath217 over a class @xmath161 of random variables @xmath20 is the requirement that the unconditional expectation of each @xmath218 lies within the range of its conditional expectations given elements of @xmath3 .
[ def : nonc ] let @xmath47 be a finitely additive prevision on a set @xmath161 of random variables , and let @xmath219 be a partition of @xmath4 such that conditional prevision @xmath220 has been defined for all @xmath221 .
if , for each @xmath218 , @xmath222 then @xmath47 is _ conglomerable in the partition @xmath3 with respect to @xmath161_. otherwise , @xmath47 is _ nonconglomerable _ in @xmath3 with respect to @xmath161 .
if a decision maker s coherent@xmath0 or coherent@xmath2 forecasts fail conglomerability in a partition @xmath3 , theorem [ thm:2 ] below shows there exist countably many conditional forecasts that are uniformly strictly dominated .
on the other hand , if the decision maker s previsions for random variables satisfy a condition ( see definition [ def : ltp ] ) similar to being conglomerable in @xmath3 , theorem [ thm:3 ] below establishes that no countable set of forecasts , conditional on elements of @xmath3 , can be uniformly strictly dominated .
what we mean by `` similar '' is explained in section [ sec : cdltp ] below .
[ thm:2 ] let @xmath47 be a finitely additive prevision , and let @xmath161 be a set of random variables .
let @xmath223 be a denumerable partition and let @xmath224 be the corresponding conditional previsions associated with @xmath47 .
let @xmath60 be a collection of strictly proper scoring rules of the form ( [ eq : gdef ] ) that satisfies the uniform similarity condition .
assume that the conditional previsions @xmath224 are nonconglomerable in @xmath3 with respect to @xmath161 .
then there exists a random variable @xmath218 with @xmath225 and @xmath226 for all @xmath221 such that ( 2.1 ) the countable sum @xmath227 of individually fair options is uniformly strictly dominated by abstaining , and ( 2.2 ) if the forecast for @xmath20 is scored by @xmath228 and the conditional forecast for @xmath20 given @xmath229 is scored by @xmath230 for @xmath231 then the sum of the scores , @xmath232 is uniformly strictly dominated by the sum of scores from a rival set of forecasts .
we illustrate theorem [ thm:2 ] with an example of nonconglomerability due to .
this example is illuminating as the conditional probabilities do not involve conditioning on null events .
[ examp:2 ] let @xmath233 .
let @xmath234 and let @xmath235 .
define a merely finitely additive probability @xmath47 so that @xmath236 for @xmath237 and let @xmath238 .
note that @xmath239 , so @xmath240 is well defined by the multiplication rule for conditional probability .
evidently , the conditional probabilities @xmath241 are nonconglomerable in @xmath3 since @xmath242 whereas @xmath243 for all @xmath221 . for ( 2.1 ) ,
consider the fair options @xmath244 for @xmath245 and @xmath246 .
choose @xmath247 and @xmath248
. then @xmath249 \biggr ] \\ & & \qquad= \cases { 0.5 - 1.0 = -0.5,&\quad
$ \mbox{if $ \omega\notin f$,}$ \vspace*{2pt}\cr -0.5 + 0.0 = -0.5,&\quad $ \mbox{if $ \omega\in f$.}$}\end{aligned}\ ] ] hence , these infinitely many individually fair options are not collectively fair when taken together .
their sum is uniformly strictly dominated by @xmath152 in @xmath4 , corresponding to the option to abstain from betting . regarding ( 2.2 ) , unlike the situation with theorem [ thm:1 ] involving countably many unconditional forecasts ,
the sum of brier scores from these conditional forecasts are uniformly strictly dominated . in particular , the sum of brier scores for these forecasts
is @xmath250 ^ 2\\ & & \qquad= \cases { 0.25 + 1.00 = 1.25,&\quad $ \mbox{if $ \omega\notin f$,}$ \vspace*{2pt}\cr 0.25 + 0.00 = 0.25,&\quad $ \mbox{if $ \omega\in f$.}$}\end{aligned}\ ] ] consider the rival forecasts @xmath251 and @xmath252 .
these correspond to the countably additive probability @xmath253 and @xmath254 for @xmath255 then the combined brier score from these countably many rival forecasts is @xmath256 ^ 2 \\ & & \qquad=\cases { 9/16 + 9/16 = 1.125,&\quad $ \mbox{if $ \omega\notin f$,}$ \vspace*{2pt}\cr 1/16 + 1/16 = 0.125,&\quad
$ \mbox{if $ \omega\in f$,}$}\end{aligned}\ ] ] which is 0.125 less than the sum of the brier scores of the original forecasts .
we offer one more example to show why we assume that the class of scoring rules satisfies the uniform similarity condition in theorem [ thm:2 ] .
recall that we have a partition @xmath257 and an event @xmath39 with @xmath258 and @xmath259 for all @xmath221 .
let the unconditional forecast for @xmath39 be scored by brier score , and let the conditional forecast for @xmath39 given @xmath229 be scored by @xmath260 times brier score .
these scoring rules fail the uniform similarity condition .
we show that the conclusion to theorem [ thm:2 ] fails . specifically , we show that there is no rival set of forecasts @xmath261 for @xmath39 and @xmath262 for @xmath229 ( @xmath263 ) whose sum of scores uniformly strictly dominates the original forecasts .
the total of the scores for the original forecasts is @xmath264 ^ 2.\ ] ] consider an arbitrary rival set of forecasts with @xmath261 for @xmath39 and @xmath262 for @xmath39 conditional on @xmath229 .
the sum of the scores for the rival forecasts is @xmath265 ^ 2+\sum _ { j=1}^\infty h_j(\omega)2^{-j-1 } \bigl[q_j - f(\omega)\bigr]^2.\ ] ] let @xmath266 when @xmath267 .
then the difference ( [ eq : pscore ] ) minus ( [ eq : qscore ] ) is @xmath268 ^ 2 + 2^{-i(\omega)-1 } \bigl(\bigl[1-f(\omega ) \bigr]^2-\bigl[q_j - f(\omega)\bigr]^2 \bigr).\ ] ] if @xmath269 , then ( [ eq : scorediff ] ) becomes @xmath270.\ ] ] if there exists @xmath5 such that @xmath271 , ( [ eq : scorediff2 ] ) is nonpositive , and the rival forecasts do not strictly dominate . if all @xmath272 , ( [ eq : scorediff2 ] ) is negative for all @xmath273 , and there is no dominance .
if @xmath274 , then ( [ eq : scorediff ] ) is at most @xmath275 ^ 2 + 2^{-i(\omega)}.\ ] ] no matter what @xmath274 we pick , either @xmath276 or @xmath277 is greater than @xmath278 .
let @xmath279 ^ 2,q_f^2\}$ ]
. for @xmath280 , ( [ eq : scorediff3 ] ) is negative either for all @xmath281 or all @xmath282 .
so , there is no dominance .
last , we establish conditions under which combining strictly proper scores from countably many conditional forecasts given elements of a partition , or combining the losses from countably many fair options based on those forecasts , does not result in a uniform sure loss .
a definition is useful first .
[ def : ltp ] let @xmath47 be a finitely additive prevision on a set @xmath161 of random variables , and let @xmath219 be a partition of @xmath4 such that conditional prevision @xmath220 has been defined for all @xmath221 . for each random variable @xmath218 , we let @xmath283 denote the random variable @xmath48 defined by @xmath284 for all @xmath267 and all @xmath221 .
we say that @xmath47 _ satisfies the law of total previsions in @xmath3 with respect to @xmath161 _ provided that for each random variable @xmath218 , @xmath285 $ ] .
[ thm:3 ] let @xmath47 be a finitely additive prevision , and let @xmath161 be a set of random variables such that @xmath47 satisfies the law of total previsions in @xmath223 with respect to @xmath161 .
let @xmath218 be a random variable with finite prevision @xmath225 and finite conditional prevision @xmath226 given each @xmath229 .
assume that @xmath220 is a finitely additive expectation ( in the sense of definition [ def : daniell ] ) that satisfies @xmath286 for every @xmath221 .
let @xmath60 be a collection of strictly proper scoring rules .
( 3.1 ) let @xmath287 be real numbers .
the sum of losses @xmath288,\ ] ] is not uniformly strictly dominated by abstaining .
( 3.2 ) let @xmath289 be elements of @xmath60 .
there is no rival set of forecasts that uniformly strictly dominates the sum of scores @xmath290 we claimed earlier that the law of total previsions in a partition @xmath3 is similar to conglomerability in @xmath3 .
the claim begins with a result of @xcite .
dubins defines _ conglomerability in partition _
@xmath3 of a finitely additive prevision @xmath47 by the requirement that , for all bounded random variables @xmath20 , @xmath291 dubins definition of conglomerability in @xmath3 is equivalent to definition [ def : nonc ] with respect to the set of all bounded random variables .
however , for a set @xmath161 that includes unbounded random variables and/or does not include all bounded random variables , the two definitions are not equivalent without further assumptions .
definition [ def : nonc ] is based on the definition given by de finetti [ ( @xcite ) , section 4.7 ] , which generalizes to unbounded random variables more easily .
@xcite also defines _
disintegrability of @xmath47 in partition _ @xmath3 by the requirement that , for every bounded random variable @xmath20 , @xmath292 where the finitely additive integral is as developed by dunford and schwartz [ ( @xcite ) , chapter iii ] .
moreover , he establishes that conglomerability and disintegrability in @xmath3 are equivalent for the class of bounded random variables.=1 the law of total previsions in definition [ def : ltp ] , with respect to the set of all bounded random variables , is equivalent to disintegrability in dubins sense , but not necessarily for sets that either include some unbounded random variables or fail to include some bounded random variables .
in addition , not all real - valued coherent@xmath0 previsions admit an integral representation in the sense of dunford and schwartz for sets that include unbounded random variables . for discussion of the problem and related issues , see @xcite ; berti and rigo ( @xcite ) ; schervish , seidenfeld , and kadane ( @xcite ) and @xcite .
as described in appendix [ app : lf ] , we use a definition of finitely additive integral that is a natural extension of coherent@xmath0 prevision . in this way
, the law of total previsions extends dubins definition of disintegrability from bounded to unbounded random variables without introducing the technical details of dunford and schwartz . finally , theorem 1 of @xcite gives conditions under which conglomerability ( definition [ def : nonc ] ) is equivalent to the law of total previsions
the following is a translation of that result into the notation and terminology of the present paper .
[ thm:2008b ] let @xmath47 be a finitely additive prevision on a set @xmath161 of random variables .
let @xmath223 be a denumerable partition and let @xmath224 be the corresponding conditional previsions associated with @xmath47 .
assume that , for all @xmath221 , @xmath220 is a finitely additive expectation on @xmath161 .
also assume that , for all @xmath218 : * @xmath25 is finite , * @xmath293 is finite for all @xmath221 , * @xmath294 for all @xmath221 , * @xmath295 for all @xmath221 , and * @xmath296 , where @xmath48 is defined ( in terms of @xmath20 ) in definition [ def : ltp ] .
then @xmath47 is conglomerable in @xmath3 with respect to @xmath161 if and only if @xmath47 satisfies the law of total previsions in @xmath3 with respect to @xmath161 . under the conditions of theorem [ thm:2008b ] , theorems
[ thm:2 ] and [ thm:3 ] show that , when the conditioning events form a countable partition @xmath3 , coherence@xmath0 and coherence@xmath2 behave the same when extended to include , respectively , the countable sum of individually fair options , and the total of strictly proper scores from the forecasts . if and
only if these coherent quantities are based on conditional expectations that are conglomerable in @xmath3 , then no failures of the dominance principle result by combining infinitely many of them .
@xcite show that each merely finitely additive probability fails to be conglomerable in some countable partition .
but each countably additive probability has expectations that are conglomerable in each countable partition .
thus , the conjunction of theorems [ thm:1 ] , [ thm:2 ] and [ thm:3 ] identifies where the debate whether personal probability may be merely finitely additive runs up against the debate whether to extend either coherence criterion in order to apply it with countable combinations of quantities .
we arrive at the following conclusions : * unless unconditional coherent@xmath141 forecasts arise from a countably additive probability , combining countably many unconditional coherent@xmath141 forecasts into a single option may be dominated by abstaining .
* however , under the conditions of theorem [ thm:1 ] , strictly proper scoring rules are not similarly affected .
the scores from countably many coherent@xmath2 unconditional forecasts may be summed together without leading to a violation of the dominance principle . *
unless conditional forecasts arise from a set of conglomerable conditional probabilities , the dominance principle does not allow combining countably many of these quantities into a single option .
hence , only countably additive conditional probabilities satisfy the dominance principle when an arbitrary countable set of conditional quantities are summed together .
scoring an agent based on the values of the fair gambles constructed from coherent@xmath0 forecasts , is not proper . because of the presence of the opponent in the game , who gets to choose whether to buy or to sell the random variable @xmath20 at the decision maker s announced price
, the decision maker faces a strategic choice of pricing .
for example , if the decision maker suspects that the opponent s fair price , @xmath297 , is greater than his own , @xmath25 , then it pays to inflate the announced price and to offer the opponent , for example , @xmath298/2 $ ] , rather than offering @xmath25 .
thus , the forecast - game as de finetti defined it for coherence@xmath0 is not incentive compatible for eliciting the decision maker s fair prices . with a finite set of forecasts and a strictly proper scoring rule for each one , using the finite sum of the scores as the score for the finite set preserves strict propriety . that is , with the sum of strictly proper scores as the score for the finite set
, a coherent forecaster minimizes the expected sum of scores by minimizing each one , and this solution is unique . here , we report what happens to the propriety of strictly proper scores in each of the three settings of the three theorems presented in section [ sec : extensions ] .
that is , we answer the question whether or not , in each of these three settings , the coherent forecaster minimizes expected score for the infinite sum of strictly proper scores by announcing her / his coherent forecast for each of the infinitely many variables .
these findings are corollaries to the respective theorems . under the assumptions of theorem [ thm:1 ] ,
the infinite sum of scores applied to the infinite set of forecasts @xmath169 is a strictly proper scoring rule . under the assumptions used for ( 2.2 ) of theorem [ thm:2 ] ,
namely when the conditional probabilities @xmath299 are nonconglomerable in @xmath3 , then the infinite sum of strictly proper scores applied to the infinite set of conditional forecasts @xmath300 is not proper . under the assumptions used to establish ( 3.2 ) of theorem [ thm:3 ] , namely that @xmath47 satisfies the law of total previsions in @xmath3 , the infinite sum of strictly proper scores applied to the infinite set of conditional forecasts
@xmath300 is a proper scoring rule .
thus , these results about the propriety of infinite sums of strictly proper scores parallel the respective results about extending coherence@xmath2 to allow infinite sums of scores .
we study how two different coherence criteria behave with respect to a dominance principle when countable collections of random variables are included .
theorem [ thm:1 ] shows that , in contrast with fair prices for coherence@xmath0 , when strictly proper scores from infinitely many unconditional forecasts are summed together there are no new failures of the dominance principle for coherence@xmath2 .
that is , if an infinite set of probabilistic forecasts @xmath301 are even simply dominated by some rival forecast scheme @xmath302 in total score , then the @xmath168 are not coherent@xmath2 , that is , some finite subset of them is uniformly strictly dominated in total score . however , because each merely finitely additive probability fails to be conglomerable in some denumerable partition , in the light of theorem [ thm:2 ] , neither of the two coherence criteria discussed here may be relaxed in order to apply the dominance principle with infinite combinations of conditional options .
merely finitely additive probabilities then would become _
incoherent_. specifically , the conjunction of theorems [ thm:1][thm:2008b ] shows that it matters only in cases that involve nonconglomerability whether incoherence@xmath2 is established using scores from a finite rather than from an infinite combination of forecasts . in that one respect , we think coherence@xmath2 constitutes an improved version of the concept of coherence .
coherence@xmath0 applied to a merely finitely additive probability leads to failures of the dominance principle both with infinite combinations of unconditional and infinite combinations of nonconglomerable conditional probabilities .
coherence@xmath2 leads to failures of the dominance principle only with infinite combinations of nonconglomerable conditional probabilities .
a referee suggested that de finetti might have been working with a different dominance principle , here denoted dominance*. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ dominance * _ : let @xmath9 and @xmath13 be two options in @xmath7 . if @xmath9 uniformly ( strictly ) dominances @xmath13 and there exists an option @xmath303 in @xmath7 that is not itself dominated by some @xmath304 in @xmath7 , then @xmath13 is an inadmissible choice from @xmath7 .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ dominance * requires that some option from @xmath7 is undominated if dominance signals inadmissibility . with respect to the decision problems considered in this paper , each of our results formulated with respect to the dominance principle obtains also with dominance*. because dominance * implies dominance as we have defined it , the only result that needs to be checked is theorem [ thm:2 ] . in that case , so long as @xmath7 contains options that correspond to a probability that satisfies the law of total previsions in @xmath3 ( as will all countably additive probabilities ) then theorem [ thm:3 ] says that such options will be undominated .
so , we could replace dominance by dominance * in the results of this paper .
this appendix gives the definitions of infinite prevision and finitely additive expectation along with brief motivation for these definitions .
details are given in the supplemental article [ schervish , seidenfeld and kadane ( @xcite ) ] .
our theorems assume that various random variables have finite previsions . in the proof of theorem [ thm:1 ]
, the possibility arises that some other random variable has infinite prevision .
definition [ def : coh ] makes no sense if infinite previsions are possible .
fortunately , we can extend the concept of coherent@xmath0 ( conditional ) prevision to handle infinite values , which correspond to expressing a willingness either to buy or to sell a gamble , but not both .
[ def : cohinf ] let @xmath305 be a collection of conditional previsions .
the previsions are
_ coherent@xmath0 _
if , for every finite @xmath306 , every @xmath27 , all real @xmath28 such that @xmath307 for all @xmath221 with @xmath308 and @xmath309 for all @xmath221 with @xmath310 , and all real @xmath311 such that @xmath312 for each @xmath221 such that @xmath313 is finite , we have @xmath314\leq0.\ ] ] that is , no linear combination of gambles may be uniformly strictly dominated by the alternative option of abstaining .
notice the restrictions on the signs of coefficients in definition [ def : cohinf ] , namely that for each infinite prevision , @xmath315 has the opposite sign as the prevision .
these restrictions express the meaning of infinite previsions as being _
one - sided _ in the sense that they merely specify that all real numbers are either acceptable buy prices ( for @xmath316 previsions ) _ or _ acceptable sell prices ( for @xmath317 previsions ) but not fair prices for both transactions .
@xcite and @xcite give alternate definitions of coherence for infinite previsions and conditional previsions .
but their definition does not make clear the connection to gambling .
however , the definition of crisma , gigante and millossovich ( @xcite ) and definition [ def : cohinf ] are equivalent for unconditional previsions , as shown in the supplemental article [ schervish , seidenfeld and kadane ( @xcite ) ] . throughout this paper , an expectation with respect to a finitely additive probability will be defined as a special type of linear functional on a space of random variables .
[ see @xcite for the case of bounded random variables . ]
infinite previsions are allowed in the sense of section [ sec : infprev ] .
[ def : daniell ] let @xmath318 be a linear space of real - valued functions defined on @xmath4 that contains all constant functions , and let @xmath319 be an extended - real - valued functional defined on @xmath318 . if ( @xmath320 and @xmath321 ) implies @xmath322 , we say that @xmath319 is _
nonnegative_. we call @xmath319 an _ extended - linear functional _ on @xmath318 , if , for all real @xmath323 and all @xmath320 , @xmath324 whenever the arithmetic on the right - hand side of ( [ eq : el ] ) is well defined ( i.e. , not @xmath160 ) and where @xmath325 in ( [ eq : el ] ) .
a nonnegative extended - linear functional is called a _ finitely additive daniell integral_. [ see schervish , seidenfeld and kadane ( @xcite ) . ] if @xmath326 , we say that @xmath319 is _
normalized_. a normalized finitely additive daniell integral is called a _ finitely additive expectation_. note that , if @xmath160 appears on the right - hand side of ( [ eq : el ] ) , @xmath327 still has a value , but the value can not be determined from ( [ eq : el ] ) .
finitely additive expectations are essentially equivalent to coherent@xmath0 previsions , as we prove in the supplemental article .
finitely additive expectations also behave like integrals in many ways , as we explain in more detail in the supplemental article . in particular , when the finitely additive expectation defined here is restricted to bounded functions , it is the same as the definition of integral developed by @xcite , and it is the same as the integral used by @xcite in his results about disintegrability .
hence , definition [ def : daniell ] is an extension of the definition of integral from sets of bounded functions to arbitrary linear spaces of functions .
let @xmath52 be of the form ( [ eq : gdef ] ) .
let @xmath47 be such that @xmath328 is finite , and let @xmath329 be such that @xmath81 $ ] is finite . if @xmath330 , then @xmath331=\lambda\bigl((q , p)\bigr)\bigl[p - r(q , p , \lambda)\bigr],\ ] ] according to ( [ eq : diff ] ) .
because lebesgue measure is absolutely continuous with respect to @xmath68 , neither @xmath332 nor @xmath333 equals 0 and they have the same sign .
it follows that ( [ eq : pdiff ] ) is strictly positive . since @xmath74 is finite , ( [ eq : pdiff ] ) is finite with @xmath334 , so that @xmath335 $ ] is also finite and so @xmath336 provides the unique minimum value of @xmath55 $ ] .
since @xmath337 , it suffices to assume that @xmath104 .
let @xmath29 .
\(i ) if @xmath109 and @xmath338 is such that @xmath339 , then the probability obtained by normalizing @xmath119 on the interval @xmath112 stochastically dominates the probability obtained by normalizing @xmath119 on the interval @xmath340 . hence
, @xmath341 .
so , it suffices to find a @xmath342 that implies @xmath343 for all @xmath12 and all @xmath104 such that @xmath344 .
for the remainder of the proof , let @xmath104 with @xmath344 , and let @xmath345 be the probability obtained by normalizing @xmath119 on @xmath112 .
let @xmath346 be @xmath347 times lebesgue measure .
then @xmath348 , and @xmath349 .
because @xmath350 , it follows that @xmath345 stochastically dominates the probability obtained by normalizing @xmath346 on @xmath351 , hence @xmath352 , and @xmath353 .
the proof @xmath354 is similar , so @xmath129 can be taken equal to @xmath355 .
\(ii ) let @xmath12 , and assume that @xmath109 .
since @xmath356 for all @xmath116 , we have @xmath357 , so @xmath358 can be taken to be @xmath359 .
because a larger random variable has a larger prevision than a smaller random variable , a necessary condition for ( [ eq : dom ] ) is that @xmath360\leq p \biggl[\sum _ { i=1}^\infty g_i(x_i , p_i ) \biggr]<\infty.\ ] ] hence , we will assume that @xmath361 from now on . also , it is necessary for ( [ eq : dom ] ) that @xmath362 for at least one @xmath139 , so we will assume this also . in light of ( [ eq : diff ] ) , we can write , for each finite @xmath363 , @xmath364 \\ & = & \sum_{i=1}^k\lambda_i \bigl((q_i , p_i)\bigr)[p_i - r_i ] + p \biggl[\sum_{i = k+1}^\infty g_i(x_i , q_i)-\sum _
{ i = k+1}^\infty g_i(x_i , p_i ) \biggr ] \\ & \geq&\sum_{i=1}^k\lambda_i \bigl((q_i , p_i)\bigr)[p_i - r_i]-w,\end{aligned}\ ] ] where the inequality follows because @xmath138 is nonnegative for each @xmath139 and where @xmath365 from ( [ eq : ra2 ] ) .
since @xmath366 does not depend on @xmath367 , it follows that @xmath368 $ ] is finite . because of ( [ eq2 ] ) and ( [ eq : qsum ] ) , the two series @xmath369 and@xmath370 are simultaneously finite with probability 1 .
let @xmath371 be the event that at least one of the two series is finite .
on @xmath372 , both series sum to @xmath316 , hence ( [ eq : dom ] ) fails unless @xmath373 .
hence , we can assume that @xmath374 for the rest of the proof .
it now follows that , for all @xmath5 , @xmath375 \\[-8pt ] \nonumber & & \qquad=\sum _
{ i=1}^\infty\bigl[g_i\bigl(x_i ( \omega),q_i\bigr)-g_i\bigl(x_i ( \omega),p_i\bigr)\bigr].\end{aligned}\ ] ] we complete the proof by showing that @xmath376>0.\ ] ] because a nonpositive random variable has nonpositive forecast , ( [ eq5 ] ) implies that ( [ eq : dom ] ) can not hold for all @xmath5 . in light of ( [ eq : pq ] ) , it suffices to show that @xmath377 \biggr)>0.\ ] ] for each @xmath367 , @xmath378 \biggr ) = \sum_{i=1}^{k } \lambda_i\bigl((q_i , p_i)\bigr ) ( p_i - r_i ) \geq0.\ ] ] next , in light of ( [ eq : diff ] ) and ( [ eq7 ] ) , write @xmath379 \biggr ) \nonumber \\[-8pt ] \\[-8pt ] \nonumber & & \qquad= \sum_{i=1}^{k } \lambda_i \bigl((q_i , p_i)\bigr ) ( p_i - r_i ) + \nonumber p \biggl[\sum_{i = k+1}^{\infty } \lambda_i\bigl((q_i , p_i)\bigr ) ( x_i - r_i ) \biggr].\end{aligned}\ ] ] since the left - hand side of ( [ eq : split ] ) does not depend on @xmath367 and the first sum on the right side is nondecreasing in @xmath367 , it follows that the second sum on the right - hand side is nonincreasing in @xmath367 , and hence , has a limit .
let @xmath380 , which is finite and strictly positive ( because @xmath181 for at least one @xmath139 ) .
then , the right - hand side of ( [ eq : split ] ) becomes @xmath381.\ ] ] the proof will be complete if we can show that the limit in ( [ eq : sumt ] ) is 0 .
first , we show that @xmath382 .
if @xmath383 , then there must exist a subsequence @xmath384 with @xmath385 for all @xmath221 .
for such a subsequence , the uniform spread condition implies that there is @xmath386 such that @xmath387 .
this would make @xmath388 , a contradiction .
it now follows that @xmath389\biggr{\vert}&\leq & \max _ { i\geq k+1}\bigl|\lambda_i\bigl((q_i , p_i ) \bigr)\bigr|p \biggl(\sum_{i=1}^\infty \\ & = & v\max_{i\geq k+1}\bigl|\lambda_i\bigl((q_i , p_i ) \bigr)\bigr|,\end{aligned}\ ] ] which can be made arbitrarily small by increasing @xmath367 , and ( [ eq : xyb ] ) follows .
corollary [ thm:1 ] is equivalent to equation ( [ eq5 ] ) , which is established in the proof of theorem [ thm:1 ] .
let @xmath390 be a denumerable partition .
nononglomerability means that there exists a random variable @xmath20 such that either @xmath391 clearly , if @xmath20 satisfies one of the above inequalities , @xmath392 satisfies the other , hence we will assume that the first inequality holds . specifically , let @xmath225 and @xmath226 for all @xmath221 , and assume that @xmath393 also , for each @xmath15 , let @xmath156 be the unique integer such that @xmath394 .
hence @xmath395 if and only if @xmath396 . (
2.1 ) consider the following sum of individually fair options : @xmath397 and the countably many options @xmath398 $ ] for @xmath399 then , for each @xmath5 , @xmath400 \\ & & \qquad = x(\omega ) - p_x - x(\omega ) + p_{i(\omega ) } = -p_{x}+ p_{i(\omega ) } \geq\varepsilon.\end{aligned}\ ] ] thus , the countable sum of the conditional forecasts for @xmath20 given @xmath229 , combined with the forecast for @xmath20 results in a loss that is uniformly strictly dominated by 0 .
( 2.2 ) for an arbitrary set of forecasts @xmath401 for @xmath20 and @xmath402 for @xmath20 given @xmath229 ( for @xmath403 ) , the sum of the scores in state @xmath5 equals @xmath404\,d \lambda_0(v ) + \int_{s_{i(\omega)}}^{x(\omega)}\bigl[x ( \omega)-v\bigr]\,d\lambda_{i(\omega)}(v ) .
\nonumber\end{aligned}\ ] ] we can substitute the original forecasts @xmath405 and @xmath406 , @xmath403 into ( [ eq16 ] ) to obtain the total score for each @xmath15 . we can also identify dominating rival forecasts @xmath407 and @xmath262 , @xmath408 so that ( [ eq16 ] ) is uniformly larger , for each state @xmath15 with @xmath405 and @xmath406 than with @xmath409 and @xmath410 .
let @xmath411 , and let @xmath412 , where @xmath413 is from part ( ii ) of lemma [ lem : ubf ] .
let @xmath414 be such that @xmath415 .
this makes @xmath416 for all @xmath221 . for each @xmath221 , @xmath417 , so that @xmath418 .
let @xmath419 , and let @xmath262 be such that @xmath420 for all @xmath221 .
this makes @xmath421 for all @xmath221 .
let @xmath407 be such that @xmath422 .
this makes @xmath423 .
we now form the difference between the scores for the original forecasts and the rival forecasts .
subtracting ( [ eq16 ] ) with @xmath424 and @xmath410 ( for all @xmath221 ) from ( [ eq16 ] ) with @xmath425 and @xmath406 ( for all @xmath221 ) yields @xmath426\,d\lambda_0(v)-\int_{q_{i(\omega ) } } ^{p_{i(\omega ) } } \bigl[x(\omega)-v\bigr]\,d\lambda_{i(\omega)}(v).\ ] ] we need to find a positive number @xmath342 such that ( [ eq17 ] ) is strictly greater than @xmath342 for all @xmath5 .
the difference in ( [ eq17 ] ) is greater than @xmath427\lambda_0 \bigl((p_x , q_x)\bigr)-\bigl[x(\omega)-q_{i(\omega ) } \bigr ] \lambda_{i(\omega)}\bigl((q_{i(\omega)},p_{i(\omega)})\bigr ) \\ & & \qquad = w_2(q_{i(\omega)}-q_x)>w_2 \bigl(q'-q_x\bigr)>0.\end{aligned}\ ] ] so , we set @xmath428 , which completes the proof .
corollary [ thm:2 ] is immediate from ( 2.2 ) of theorem [ thm:2 ] , as the existence of the rival set of dominating forecasts , @xmath429 , establishes that the forecaster does not minimize the infinite sum of expected scores by giving the forecast @xmath430 and the conditional forecasts @xmath300 .
( 3.1 ) in order to show that ( [ eq18 ] ) can not be uniformly strictly positive , it is sufficient to show @xmath431=0.\ ] ] of course , @xmath432= p\bigl[\alpha_0(x - p_x)\bigr]+p \biggl[\sum _ { j=1}^\infty\alpha _ jh_j(x - p_j ) \biggr].\ ] ] trivially , @xmath433=0.\ ] ] since @xmath47 satisfies the law of total previsions in @xmath3 , @xmath434=p \biggl [ p \biggl [ \sum_{j=1}^\infty \alpha_jh_j(x - p_i ) \big{\vert}\pi \biggr ] \biggr].\ ] ] for each @xmath139 , @xmath435=0 $ ] for all @xmath436 .
it follows that , for every @xmath139 , @xmath437 = p\bigl[\alpha_ih_i(x - p_i)\mid h_i\bigr],\ ] ] and trivially , @xmath438=0 $ ] .
thus , @xmath439=0\ ] ] for all @xmath5 , and it follows by the law of total previsions that @xmath440 = 0.\ ] ] equations ( [ eq20 ] ) and ( [ eq23 ] ) establish ( [ eq19 ] ) . ( 3.2 ) we must establish that there is no rival set of forecasts @xmath407 , and @xmath429 whose total score uniformly dominates ( [ eq24 ] ) .
that is , there is no rival set of forecasts such that for some @xmath14 and every @xmath5 , @xmath441 it is sufficient to show that @xmath442 \biggr\ } \nonumber \\[-8pt ] \\[-8pt ] \nonumber & & \qquad\geq0.\end{aligned}\ ] ] write the left - hand side of ( [ eq26 ] ) as @xmath443+p \biggl[\sum_{j=1}^\infty h_jg_j(x , q_j)-\sum _
{ j=1}^\infty h_jg_j(x , p_j ) \biggr].\ ] ] that the first expectation in ( [ eq27 ] ) is nonnegative follows from the fact that @xmath52 is strictly proper . from the assumption that @xmath47 satisfies the law of total previsions in @xmath3 , @xmath444\\ & & \qquad = p \biggl[p \biggl[\sum_{j=1}^\infty h_jg_j(x , q_j)-\sum _ { j=1}^\infty h_jg_j(x , p_i ) \big{\vert}\pi \biggr ] \biggr].\end{aligned}\ ] ] using equation ( [ eq : diff ] ) and the same logic as in part ( 3.1 ) , we obtain , for each @xmath139 , @xmath445\\ & & \qquad = p\bigl[h_i \bigl\ { g_i(x , q_i)-g_i(x , p_i ) \bigr \}|h_i\bigr ] \\ & & \qquad = p\bigl[g_i(x , q_i)|h_i\bigr]-p \bigl[g_i(x , p_i)|h_i\bigr ] \\ & & \qquad\geq 0,\end{aligned}\ ] ] where the final inequality follows because @xmath138 is a proper scoring rule and @xmath446 is a finitely additive expectation for all @xmath139 . therefore , since @xmath47 satisfies the law of total previsions in @xmath3 , @xmath447\geq0,\ ] ] which completes the proof of ( [ eq27 ] ) .
corollary [ thm:3 ] is equivalent to the claim that for each set of rival forecasts , @xmath407 and @xmath429 , the second prevision in ( [ eq27 ] ) is nonnegative , which was established in the proof of ( 3.2 ) .
we thank raphael stern , department of statistics , carnegie mellon university , and two anonymous referees for helpful advice about this paper . | we investigate differences between a simple dominance principle applied to sums of _ fair prices _ for variables and dominance applied to sums of _ forecasts _ for variables scored by proper scoring rules .
in particular , we consider differences when fair prices and forecasts correspond to finitely additive expectations and dominance is applied with infinitely many prices and/or forecasts . , |
this paper presents _ hst / wfpc2 _ observations of 77 early - type galaxies .
we investigate the central morphology of the galaxies by characterizing the properties of their nuclei , color gradients , ellipticities , isophote twists , instances of central surface brightness minima , offset centers , dust content , and dust morphologies .
the result is a more complete portrait of the sample galaxies than has been available heretofore and a reference where many different properties are measured and compared in one place .
we also derive high resolution surface photometry profiles of the sample galaxies , which we characterize with `` nuker law '' fits @xcite .
analysis of the fit parameters is presented in paper vi ( lauer et al .
, in preparation ) , in which we combine the present results with previously published nuker law fits to _ hst _ surface photometry profiles to define an extended sample of 264 e / s0 galaxies .
a motivation of both this paper and paper vi is to understand the origin of `` cores '' in early - type galaxies .
_ hst _ images show that nearly all galaxies have singular starlight distributions in the sense that surface brightness diverges as @xmath4 with @xmath5 @xcite . in typically
lower - luminosity early - type galaxies , @xmath6 decreases only slowly as the center is approached and a steep @xmath7 cusp continues into the _ hst _
resolution limit ; @xcite classified these systems as `` power - law '' galaxies . in more luminous galaxies , however , the projected profile transitions or breaks from a steep power law in the envelope to a shallow inner cusp with @xmath8 ; these `` core galaxies '' thus show central deficits of starlight or a core compared to the centrally - steeper `` power - law '' galaxies .
many of the core galaxies are the same systems in which cores were already evident from ground - based observations @xcite ; however , rather than representing regions in which the central stellar density becomes constant , the residual shallow cusps in projected brightness still imply steep and singular cusps in density @xcite .
@xcite and @xcite showed that the distribution of cusp slopes at the _ hst _ resolution limit in both stellar luminosity density and projected surface brightness is bimodal ; power - laws and core galaxies are separated into two distinct groups by their inner cusp slopes . @xcite and @xcite later identified a small number of `` intermediate '' galaxies that have limiting cusp slopes with @xmath9 but showed that the ensemble of cusp slopes in all early - type galaxies is still bimodal .
this topic will be further explored in paper vi , which strongly ratifies the bimodality of central structure .
@xcite also examined how the central structure correlates with other galaxy properties , showing that luminous early - type galaxies preferentially have cores , whereas most fainter spheroids have power - law profiles .
moreover , cores are slowly rotating and have boxy isophotes , while power laws rotate rapidly and are disky .
these ideas resonate well with a revision of the hubble sequence proposed by @xcite , which divided ellipticals into boxy , non - rotating types and disky , rotating types .
the latter were seen to be the close relatives of spirals , whereas the boxy , non - rotating ellipticals were somehow different . "
cores are associated with the boxy , non - rotating sequence , and thus serve as a fundamental morphological marker .
@xcite further argued that the prevalence of cores is directly tied to the presence of nuclear black holes in nearly all galaxies @xcite and the assembly of galaxies by hierarchical mergers .
@xcite argued that the merging of two galaxies , each harboring a central massive black hole , would create a black hole binary in the center of the merger remnant .
gradual hardening of the binary via stellar encounters would scatter stars out of the center , creating a central deficit of stars with respect to inward extrapolation of the envelope in a merged galaxy whose center would otherwise be very dense .
n - body simulations of merging galaxies with central black holes @xcite show that cores can indeed form in such merger remnants . looking for ways to test
the hypothesis that cores reflect the effects of binary black holes on the central structure of galaxies is a subtext for much of the analysis presented in this paper .
we will return to this topic in the summary .
a final section of the paper is devoted to the morphology and prevalence of optical dust absorption in early - type galaxies . as in previous works ( e.g. , @xcite , @xcite , @xcite ) , dust
is found in roughly half of our galaxies .
dust is extremely well correlated with both the presence and strength of nuclear optical emission , which we assume indicates agn activity .
thus , it is hard to avoid concluding that dust is outlining interstellar material that is about to fall onto the black hole , and that it therefore becomes a valuable clue as to how that happens .
we discuss the possibility that the diverse patterns of dust absorption seen in early - type galaxies may be viewed as various stages of a `` settling sequence ; '' under this picture dust may come and go on timescales of a few @xmath10 years .
the implication is that galaxies might be emptying and refilling themselves with dust constantly , raising the question of where the dust comes from and why the process is cyclical .
understanding this could be an important clue to the growth of black holes .
we present images and surface photometry for 77 early - type galaxies observed with _
hst_+ wfpc2 ; 55 of these were obtained under programs go 5512 , 6099 , 6587 , and 9107 which were carried out by our collaboration .
the observations are listed in table [ tab : obs ] .
we also include an independent reduction of the galaxies observed in go 5454 ( pi : franx ) , which were selected to have kinematically decoupled cores ( kdc ) .
analysis of these by @xcite showed no morphological differences between ordinary ellipticals and those with kdc , thus including these galaxies in the present sample causes no obvious bias .
some of the remaining galaxies represent original or archival images obtained in support of go 9107 , in which we proposed stis spectroscopy to explore the mass range of nuclear black holes at a particular galaxy luminosity .
lastly , there are additional galaxies obtained from the archive in support of our other stis programs .
a portion of this photometry has been published previously in papers on the detection of nuclear black holes @xcite or on the morphology of galaxies with double centers or local minima in surface brightness at their centers @xcite ; such galaxies are presented again here to provide complete surface photometry in the context of a comprehensive re - investigation of the central structure of early - type galaxies .
table [ tab : obs ] lists any earlier publication of surface photometry derived from the given dataset .
there is no single criterion that characterizes our sample . in general
, the sample comprises relatively luminous nearby elliptical and early - s0 galaxies that could be investigated spectroscopically with _
hst _ to detect and weigh " nuclear black holes .
the particular selection of galaxies was heavily informed by the central - structure parameter relationships presented by @xcite .
several galaxies were selected , for example , within the luminosity range over which power - law and core galaxies co - exist , to explore the transition between the two types .
others were selected to extend the luminosity range of the core parameter relationships .
the properties of the galaxies , such as luminosity , velocity dispersion , distance , and morphology are listed in table [ tab : glob ] .
a set of galaxies already observed with wfpc1 and presented by @xcite were also re - observed with wfpc2 when it was felt that the improved resolution and dynamic range might lead to a significant improvement in the galaxies central structure . in the end ,
the sample is rich in core galaxies , with relatively few power - law galaxies ; it thus is complementary to the @xcite sample , which is rich in power - law galaxies .
all galaxies were imaged with the high - resolution pc1 ccd in wfpc2 ; the image scale is @xmath11pixel .
nearly all galaxies were imaged in the f555w filter ( equivalent to broad band @xmath12 ) ; images were also obtained in f814w filter ( broad band @xmath13 ) to provide color information for many galaxies .
the exposure time for most galaxies was roughly a single orbit in each filter , typically split into several sub - exposures to allow for cosmic - ray event repair .
the exposure levels in the galaxy centers varied widely over the sample , but were typically @xmath14 which is optimal for point - spread function ( psf ) deconvolution .
beginning in program g0 6587 , many of the image sets were dithered in a @xmath15 square pattern of 0.5 pc pixel substeps .
this procedure allows for the removal of the aliasing present in any single pc image and consequent recovery of the optimal spatial resolution provided by the _
hst_+wfpc2 combination .
the reduction procedure for a given image set depended on whether the images were dithered .
non - dithered image sets were always obtained with the same pointing and could simply be compared to detect and repair cosmic - ray events , before being stacked .
the dithered images for a given galaxy and filter combination were combined into a single nyquist - sampled `` super - image '' using the fourier algorithm presented by @xcite .
the algorithm combines the images in the fourier domain and is designed to recover the nyquist image even if the dither positions are somewhat non - optimal .
the output of the algorithm is a double - sampled image with 0.5 pc pixel steps .
the fourier algorithm is theoretically justified and introduces no blurring of the data ; this is in contrast to _ drizzle _
@xcite , which can not remove aliasing in a small dataset , and introduces additional blurring on the scale of the original pixel sampling .
a drawback of our reduction path is that the pointing of each image in the image - set is slightly different , complicating the repair of cosmic - ray events .
the procedure used for cosmic ray repair was to first intercompare the images ignoring their small offsets and then construct a nyquist image from the repaired images .
this first estimate of the nyquist image was used to improve the initial detection of the cosmic - ray events , thus producing a revised nyquist image .
this procedure was iterated a second time to produce the final image .
the final step was to deconvolve the psf .
hst+_wfpc2 psf introduces significant blurring as far as @xmath16 from the galaxy centers , but deconvolution can recover the intrinsic surface brightness profiles to a few percent accuracy for @xmath17 @xcite .
deconvolution was achieved by 40 iterations of the lucy - richardson @xcite algorithm .
the psfs were calculated from the tiny - tim package @xcite .
for the double - sampled images , the tiny - tim psfs were convolved with the pixel - response function of @xcite .
central images of the sample galaxies are shown in figures [ fig : images ] and [ fig : obs_im ] .
while the classic view of early - type galaxies is that they are purely stellar systems with elliptical isophotes , the majority of the galaxies depart from this picture one way or another , having central dust patches or rings , stellar disks , nuclear clusters , central minima , and so on .
a summary of the morphological features seen is presented in table [ tab : morph ] .
figure [ fig : images ] also shows the galaxy images divided by models reconstructed from the surface photometry ( described below ) to help isolate these features .
those galaxies that had their centers so obscured by dust that surface photometry could not be derived are shown separately in figure [ fig : obs_im ] . the overall importance of dust for any galaxy
was roughly quantified by comparing the integrated light of the observations to model galaxy images reconstructed from the surface photometry within @xmath18 of the galaxy centers .
the dust absorption thus estimated is presented in table [ tab : morph ] in units of hundredths of a magnitude .
these absorption values are not intended for quantitative astrophysical interpretation , but do seem to correspond well to subjective impressions of the strength of central dust seen in figure [ fig : images ] .
surface photometry was derived under the assumption that the isophotes are concentric ellipses , but with position angle and ellipticity than can vary arbitrarily with radius .
the photometry was measured in several stages .
the first step was to mask obvious stars , dust patches , image defects , and so on , from the images .
initial photometry was derived using the least - squares algorithm of @xcite . a model reconstructed from photometry
was then subtracted from the images to isolate and mask more subtle features that were missed during the initial pass . for galaxies with extensive dust
this procedure was often iterated several times , as the isophote parameters would often change significantly as the mask was refined .
the final stage was to measure the inner isophotes ( @xmath19 ) with the high - resolution algorithm of @xcite .
this algorithm uses sinc - interpolation to sample the image , thus preserving the spatial resolution of the images . for this stage ,
image defects were initially filled in with a model constructed from the lower - resolution photometry , and then with a model constructed from the high - resolution photometry .
the final surface photometry profiles are presented in figure [ fig : surf ] .
the zeropoint calibration was obtained from @xcite .
a `` sky '' level for each galaxy was measured from the corner of the wfpc2 wf3 ccd farthest away from the galaxy center ( typically @xmath20 in angle ) , and was subtracted from the photometry . for the more extended galaxies ,
the sky will have some galaxy - light contamination , but given that the present study focuses on the brighter central isophotes , this will cause only a small error .
a number of galaxies in the sample were originally observed with wfpc1 .
the new wfpc2 photometry allows a test of the wfpc1 observations , which required large deconvolution corrections .
figure [ fig : decon ] compares wfpc2 and wfpc1 profiles for several galaxies .
@xcite argued based on simulations that deconvolution could recover profiles from wfpc1 that were accurate to a few percent for @xmath21 with significant residual blurring at smaller radii .
the comparison of wfpc2 and wfpc1 photometry for m32 and m33 presented by @xcite showed this conclusion to be essentially correct , while underscoring the need to deconvolve wfpc2 photometry as well .
figure [ fig : decon ] ratifies this conclusion for a larger set of galaxies . for the most part , the wfpc2 and wfpc1 profiles agree extremely well for @xmath22 the exception to this general rule appears to be galaxies with strong nuclei , such as ngc 3115 , where light from the central point - source spilled beyond @xmath23 and was poorly recovered by wfpc1 .
weak nuclei in core galaxies , such as in ngc 1399 , were also poorly captured by wfpc1 .
the conclusion , overall , is that the wfpc1 profiles can still be used to their stated @xmath24 resolution limit .
we fitted the surface brightness profiles with the `` nuker law '' @xcite parametric form @xmath25^{(\gamma-\beta)/\alpha } , \label{eqn : nuker}\ ] ] which describes the profiles as a `` broken power law ; '' it is an extension of a broken power - law form first used to describe the brightness profile of m32 @xcite . here
@xmath26 is the semi - major axis represented the geometric mean of the semi - major and semi - minor axes . ] .
the outer portion of the profile is a steep power - law with logarithmic slope , @xmath27 this outer power - law `` breaks '' at radius @xmath28 transitioning to a shallower inner power - law with slope @xmath29 the smoothness or speed " of transition is determined by @xmath30 the intensity scale is set by @xmath31 the surface brightness at @xmath32 the break radius is both the point of maximum curvature of the profile in logarithmic coordinates and the location where the local slope is the average , @xmath33 of the inner and outer slopes . in passing ,
we note that we are aware of the limitations of parametric fits to data .
however , a nonparametric analysis of the wfpc1 profiles by @xcite yielded no additional insight into the central structure of elliptical galaxies than was provided by the parametric analysis of @xcite and @xcite using the nuker law . we fitted the major - axis profiles using the minuit package @xcite , which was also used by @xcite to fit the wfpc1 profiles . unlike byun et al .
, however , we allowed the algorithm to find solutions with @xmath34 the discovery of galaxies with central _ minima _ in surface brightness @xcite shows that central brightness profiles with positive slopes must be allowed for .
while in formally @xmath35 would imply that the surface brightness profile goes to zero as @xmath36 which is impossible in projection , the integrated brightness in any aperture remains positive , and more to the point , we mean to imply no behavior of the profile within the resolution limit .
the nuker law parameters for all galaxies in figure [ fig : images ] are listed in table [ tab : cen ] and plotted in figure [ fig : surf ] .
the radial domain fitted was adjusted on an _ ad hoc _ basis based on the structure of a given galaxy and the residuals of an initial nuker law fit over the entire extent of the profile .
typically an inner limit was specified to avoid nuclei ( see below ) , which the nuker law is not intended to represent .
an outer limit was specified in some cases to avoid deviations from the nuker law that might have influenced the quality of the fit to the inner region .
the fits are shown as solid lines in figure [ fig : surf ] , with the extent of the line showing the radial range of the fit .
the fit residuals are plotted at the bottom of the brightness profile panels .
@xcite showed that the brightness profiles could be grouped into two classes . ``
core galaxies '' show an obvious break in logarithmic slope that marks a transition to a shallow inner cusp with @xmath37 where @xmath38 is the slope at the _ hst _ resolution limit ( @xmath39 for the f555w wfpc2 images ) .
`` power - law galaxies , '' in contrast , retain a steep slope , @xmath40 into the resolution limit . as was shown by @xcite and @xcite ,
the distribution of cusp slopes is bimodal , with no examples of galaxies in those papers with @xmath41 this classification scheme will be re - examined in detail in paper vi .
the adopted profile classifications are listed in table [ tab : cen ] . in some cases ,
as noted in table [ tab : cen ] , we have reclassified a number of galaxies previously observed by ourselves or others because we have adopted finer resolution limits or used higher resolution imagery than the previous investigations .
@xcite and @xcite identified a set of galaxies that have @xmath42 which rest et al .
call `` intermediate '' galaxies .
three such galaxies ( ngc 821 , 3585 , and 7723 ) are in the present sample . as @xcite showed , intermediate types are rare , so that the distribution of cusp slopes remains bimodal , even if the region between core and power - law galaxies is no longer completely empty .
we also validate the concern of @xcite that the intermediate classification is highly sensitive to resolution . as it happens , after adopting finer resolution limits than did rest et al .
, we consider _ all _ of their intermediate galaxies to be core galaxies , while we now consider some of their power - law galaxies to be intermediates .
we will explore this issue in detail in paper vi .
figure [ fig : ellip ] plots the average isophote ellipticity interior to @xmath28 weighted by the integrated isophote luminosity ( analogous to equation ( [ eqn : lw ] ) in @xmath43 below ) , as a function of the average exterior ellipticity ( averaged the same way but for @xmath44 ) for both core and power - law galaxies ( ellipticity values are given in table [ tab : ellip ] ) .
there is no clear trend for either type of galaxy to become systematically flatter or rounder as the center is approached .
what is evident in figure [ fig : ellip ] , however , is that the power - law galaxies on average are more elliptical than core galaxies .
this trend is also seen in figure [ fig : ellip_mag ] , which plots inner ellipticity as a function of galaxy luminosity . while @xcite show that elliptical galaxies overall become rounder with increasing luminosity , and cores become more prevalent with luminosity , core galaxies are rounder than power - law galaxies at the same luminosity .
@xcite looked at _ hst _ photometry of a large sample of elliptical galaxies and reached similar conclusions .
in particular , they looked at isophote shape as a function of the fractional distance to the effective radius , finding that the distinction between core and power - law galaxies was strongest at the smallest radii .
@xcite and @xcite noted that the power - law galaxies ( their type ii ) in their sample were highly elliptical , while the core galaxies ( their type i ) were nearly round .
they argued that the difference between the two was _ merely one of projection . _ in their view , all ellipticals harbored disks , which when seen edge - on would make sharply rising cusps ; the stellar density profiles of all ellipticals would have essentially the same form . @xcite
disagreed with this picture , noting the existence of round power - law galaxies , which should not exist at all in the projection model . moreover , the new data in figures [ fig : ellip ] and [ fig : ellip_mag ] show that core galaxies co - exist with power - law galaxies at all ellipticities .
although core galaxies are on average rounder than power - law galaxies , the ellipticity distributions of the two types overlap significantly , whereas the projection model implies a much cleaner separation .
@xcite also claimed that the majority of highly elliptical power - law galaxies showed no evidence of a central stellar disk ; however , we now think that such cases are rare .
first @xcite subsequently noted that power - law galaxies have disky _ outer _ isophotes , which firmly associates power - laws with disks .
now , new wfpc2 imagery here shows that disks can be seen in _
nearly all _ power - law galaxies with inner ellipticity @xmath45 whereas _ no _ disks are seen in power - law galaxies rounder than that ( see figure [ fig : ellip_mag ] ) .
this strong dichotomy suggests that _ all _ power - law galaxies harbor disks ; their visibility is simply due to the viewing angle .
it is true that @xcite found disks in only three out of nine highly elliptical power - law galaxies ; however , some of the disks in the present sample are faint and may have been poorly visible in wfpc1 imagery .
while , in common with @xcite , we can not detect a disk in the power - law galaxy ngc 1426 ( which has @xmath46 ) , we do find a disk in ngc 4458 ( and now classify it as a core rather than power - law galaxy ) .
core galaxies generally have boxy isophotes @xcite , but we find disks in four core galaxies , all of which have high inner @xmath47 three of these are rather dim , having @xmath48 and only one ( ngc 3706 ) of the five highly elliptical core galaxies brighter than this has a disk .
two of the three intermediate galaxies also have disks , and they also lie in the same magnitude transition zone near @xmath49 the overall impression is that disks are frequent if not ubiquitous in power - law galaxies , populate a fair fraction of faint cores and intermediate galaxies in the transition zone , and then peter out amongst bright core galaxies . the disappearance of disks with brightness seems to be related to the core / power - law transition itself and may give a clue as to how it occurs , as discussed in the summary .
an examination of the isophote position angle profiles presented in figure [ fig : surf ] shows that many core galaxies have significant isophote twists close to or interior to their break radii .
this behavior , for example , can be seen in ngc 507 , the first galaxy presented in figure [ fig : surf ] .
the general phenomenon is that most core galaxies have isophote orientations that remain essentially constant over their outer radii ( within the regions imaged by wfpc2 ) , but at some small radius the isophotes smoothly twist away from the outer orientations . to quantify this phenomenon , we fitted the position angle profiles of the core galaxies with a `` hinged '' line .
in detail , a line was first fitted over the outer portion of the profile , starting at an inner limiting `` hinge '' radius .
once the parameters of the outer line were determined , a second line was fitted interior to the hinge radius ( but still with @xmath50 to avoid nuclei ) with the constraint that the inner line had to intersect with the outer line at the hinge radius .
the effect is to fit the entire position angle profile with two lines of differing slope that intersect at a `` hinge . ''
the position of the hinge was selected by finding the location that minimized the residuals of the entire position angle profile fitted by both line segments .
the final fits are shown as dotted lines in the position angle plots in figure [ fig : surf ] .
table [ tab : bend ] presents the values of the inner and outer position angle gradients , the hinge radius , and the local logarithmic slope of the brightness profile , @xmath51 at that location .
figure [ fig : bend ] shows the inner position angle gradients as a function of @xmath38 at the transition radius .
the symbols plotted also show the inner and outer gradients schematically as angled wedges , with the slopes of the segments to the left and right of the vertex proportional to the inner and outer position - angle gradients .
both the table and figure show that the isophote position angle varies only slowly outside the hinge radius , but strong twists occur at small radii .
the transition in behavior occurs near @xmath52 a point typically within the shallow surface brightness cusp interior to the core break - radius .
figure [ fig : bend_pow ] plots the same data for the power - law galaxies .
the power - law galaxies appear to have at most very small twists in their inner isophotes .
only two galaxies not affected by dust have gradients that exceed @xmath53 consideration of the true `` power '' in the twists in power law galaxies , however , suggests that they are as significant as the large twists seen in the inner cusps of the core galaxies .
the significance of any given isophote twist is tied to how much the light distribution in a galaxy must have changed to exhibit the twist .
very little light needs to be redistributed in a portion of a galaxy with a shallow brightness gradient and nearly circular isophotes to cause a strong twist in angle , while the perturbation must be large when the gradient is steep or the isophotes are highly elliptical .
we have attempted to quantify these effects by following the general approach of @xcite , who analyzed the shapes of dwarf ellipticals by introducing a twist - measure that accounts for the true perturbation in the projected light distribution of a galaxy to produce the observed twists .
the basic concept is to measure the relative amount of light required to `` iron - out '' or untwist a twisted galaxy .
the first step is to use the brightness profiles to calculate the luminosity - weighted average position angle of a given galaxy , @xmath54 over a major - axis range @xmath55 in which the twist is to be measured . here
@xmath56 gives the isophote position angle profile , @xmath57 is the brightness profile , @xmath58 is the integrated galaxy luminosity over the radial range , and the area element is @xmath59 where @xmath60 the ratio of isophote minor to major axes at @xmath61 the `` surface - brightness twist '' is estimated as @xmath62 where @xmath63 is an image reconstructed from the surface photometry with the isophotes all forced to have position angle @xmath64 while @xmath65 is an image reconstructed from the surface photometry with the twists as measured , and the sum is over the @xmath66 pixels within @xmath67 the surface - brightness twist thus is a typical relative root - mean - square ( rms ) measure of the difference between the untwisted and observed galaxy .
we emphasize that our @xmath68 measure differs from that of @xcite , which we believe suffers from a conceptual error in its derivation .
we have calculated surface - brightness twists separately for the radius ranges inside and outside the hinge radius ( limited to an outer radius three times larger than the hinge - radius defining the boundary of the inner and outer gradients ) ; they are presented in table [ tab : bend ] and figure [ fig : twist ] .
as can be seen , both power - law and core galaxies exhibit somewhat stronger twists in their inner regions .
the inner - twists are typically @xmath69 stronger for both types of galaxies than the outer twists .
displayed this way , however , we see that the significance of the inner twists in both core and power - law galaxies is essentially the same . the @xmath70 color gradients of the inner regions of the sample galaxies are generically similar to those seen farther out in the envelopes of early - type galaxies .
the great majority of sample galaxies become slightly redder toward the center .
the exceptions are ngc 2300 , 4382 , and 7457 , which have shallow gradients that become _ bluer _ with decreasing radius . as seen in figure [ fig : surf ] , the color gradients are roughly linear with shallow slopes in logarithmic coordinates .
table [ tab : vmi ] lists the color gradients measured over @xmath71 the inner cutoff is set to avoid nuclear sources .
figure 4 of @xcite suggests that the color gradients show more variability over @xmath72 than they do at larger radii , but we do not see this effect .
the strengths of the color gradients superficially appear to depend on the profile type of the galaxy , as already suggested by figure 11 of @xcite , and by high - resolution ground - based images of galaxy centers @xcite ; however , this effect appears largely to be due to the greater prevalence of dust in power - law galaxies .
figure [ fig : grad_gam ] plots the color gradients as a function of the limiting logarithmic profile slope at the resolution limit , @xmath73 the symbol type encodes the central dust indices given in table [ tab : morph ] , with open symbols corresponding to galaxies with dust absorption greater that 0.02 mag .
even though we attempted to exclude visible dust patches from the profile estimation algorithms , the strongest color gradients correlate with the largest scatter about the mean gradient line , supporting the conclusion that the color profiles in these cases are still affected by residual dust absorption and are not likely to reflect true changes in the underlying stellar population .
if dusty galaxies are ignored , there still a suggestion of a weak trend for power - law galaxies to have stronger gradients than core galaxies ; the mean @xmath70 gradients for the two types are @xmath74 mag ( excluding the one power - law galaxy with a positive gradient ) and @xmath75 mag , respectively . clearly , however , the scatter in color gradients at any @xmath38 is substantially larger than the small offset between the two types , and to first order both types of galaxies have essentially the same inner color gradients .
lastly , figure [ fig : grad_mag ] shows that there is no relationship between the strength of the gradient and galaxy luminosity , once dusty galaxies are excluded .
we also examined the color gradients of core galaxies to see if the gradients were smaller interior to the break radius .
because the envelope color gradients in core galaxies are small to begin with , such a change in absolute terms will be a small effect , and indeed for the majority of core galaxies it is difficult to see any change in the gradient associated with the core . however , ngc 3379 , 3608 , 4291 , 4365 , 4406 , and 7619 do appear to have flatter color gradients for @xmath76 ngc 4406 provides the best example , but the significance of this effect is low .
a substantial fraction of the sample galaxies have nuclei , a compact light source that is seen to rise above the inward extrapolated surface brightness cusp at small radii . in the present sample
we find 13/45 ( 29% ) of the core galaxies and 12/20 ( 60% ) of the power - law galaxies to have nuclei .
nuclei were recognized by visual inspection of the residuals of the nuker law fits to the surface brightness profiles . for the fainter nuclei in core galaxies ,
the inner radial limits of the fits were often increased to better isolate nuclei included in the initial fits .
table [ tab : nuclei ] lists the luminosities and colors of the nuclei identified .
the luminosities were derived by integrating the central residuals from the nuker law fits .
this procedure presumes that the nuker law can be extrapolated inwards from the radial domain over which the fit was performed .
the total fluxes of the nuclei may thus have unknown systematic errors . at the same time
, our nuclear luminosities for ngc 4278 , 4365 , and 4552 agree extremely well with those of @xcite .
the color of the nucleus is that of the central light source isolated from the hosting galaxy ; it is estimated from the nuker law residuals in the @xmath12 band and the @xmath70 profiles on the assumption that the @xmath70 color just outside the nuclei gives the color of the background starlight .
the colors have been corrected for galactic reddening .
the @xmath77 colors listed in table [ tab : nuclei ] give the change in color of the nuclei relative to these fiducial values .
note that a faint nucleus associated with a significant decrease in the central total @xmath70 will thus be inferred to be extremely blue .
in general the nuclei are bluer than the background galaxy starlight immediately outside the nuclei , although they are not particularly blue in absolute terms .
most nuclei are also spatially unresolved ( more compact than @xmath78 pc ) ; only the nuclei in ngc 741 and 4365 appear to be resolved with scales of @xmath79 pc and @xmath80 pc , respectively .
if the nuclei are star clusters , then they may comprise stars younger or more metal - poor than those in the host galaxies .
we note that most of the nuclei are too luminous to be classic globular clusters
. a better analog in color , luminosity , and physical compactness would be the m33 nuclear star cluster @xcite . in passing ,
we note that ngc 5419 appears to have a double point source at its center .
the brighter component is at the photocenter of the galaxy ; the second component appears to be weakly resolved .
the detection of nuclei in many power - law galaxies is not surprising in light of previous work .
@xcite identified nuclei in a large fraction of their power - law galaxies ( some of which are in the present sample ) ; @xcite also found nuclei in a number of power - law galaxies .
the frequency of nuclei in core galaxies is less well understood .
@xcite found nuclei in only two of their core galaxies , ngc 6166 , and the brightest cluster galaxy ( bcg ) in a2052 , both of which were known to be agn .
however , the wfpc1 images were poorly suited to detecting faint nuclei . @xcite
were the first to use wfpc2 for this problem ; they only observed a few core galaxies , but found nuclei in most of them .
@xcite also used wfpc2 , but found nuclei in only one of their nine core galaxies .
@xcite found nuclei in 5/13 of their core galaxies , consistent with the present detection rate .
rest et al .
adopted very conservative criteria for identifying nuclei , while ravindranath et al
. always included a nuclear point source in their nuker law fits .
lastly , @xcite detected nuclei in @xmath81 of their bcg sample , which mainly had core galaxies .
the bcg are in general at much larger distances than the galaxies in the present sample and observed in the @xmath13 band , both making the detection of nuclei difficult .
the present sample is also rich in core galaxies , and the 29% detection rate is probably a more representative number than in previous samples .
there is too little information to determine whether nuclei are star clusters or are agns ; both phenomena are probably present .
@xcite argued that at least 11/14 of their nuclei were associated with agn . in the present sample agn
are likely to explain ngc 1316 @xcite , 4552 @xcite , and 7457 @xcite .
table [ tab : morph ] lists what is known about nuclear activity of the sample galaxies based on the optical emission lines .
there is weak evidence that power - law galaxies with nuclei have absorption - line spectra , while core galaxies with nuclei have strong emission lines .
however , exceptions in both directions exist ; in particular , a number of core galaxies with emission lines showed no nuclei in the present surface photometry .
we thus can not conclude that low - level agn , even if present , explain the all nuclear sources seen in the broad - band _ hst _ images .
a basic structural paradigm that underlies the present structural analysis of early type galaxies is that their central starlight distributions can be described by concentric elliptical isophotes with surface brightness increasing monotonically towards the center .
there is however , an increasingly large set of objects that depart from this picture .
m31 @xcite and ngc4486b @xcite have double nuclei , and @xcite identify a number of galaxies that have local _ minima _ in their brightness profiles at or near their centers . in this section
we discuss ngc 4073 and 4382 , which also have central surface brightness minima . in the next section
we discuss an additional set of galaxies in which the location of maximum brightness within the core is spatially offset from the photocenter of the core itself .
we speculate that a substantial majority if not all of these systems may be cases where black holes were critical to either creating or preserving the unusual structures over long times .
@xcite , for example , explains the double nuclei in m31 ( and by implication , ngc 4486b ) as the projected appearance of an eccentric disk of starlight stabilized by a central black hole . in the case of the galaxies with central minima , @xcite advanced two competing explanations for the creation of such systems , both involving black holes .
the first is that a diffuse torus of starlight has been added to a pre - existing core to create the appearance of a central minimum interior to the torus .
this would occur during the final stages of merging in which the tidal field of the black hole in the more massive galaxy disrupts the dense center of a galaxy being cannibalized @xcite . a second
, more exotic idea is that stars have actually been ejected from the galaxy center by the decay of a central black hole binary created in a merger .
@xcite further noted that at least some of galaxies with central minima may be related to the double nucleus systems . strictly speaking , m31 and ngc 4486b have locations near their centers at which a local minimum in surface brightness occurs , and
could thus be included in the class of galaxies with central minima .
conversely , the local minima in some of the @xcite galaxies are bracketed by structures that create the appearance of highly diffuse double nuclei .
separately , the central minimum in ngc 3706 is clearly due to a well defined central ring of stars that might create a double nucleus if it were slightly more eccentric or were viewed from a different angle ; the central structure in ngc 3706 resembles that of m31 in other ways . in the present sample , the center of ngc 4073 resembles that of @xcite galaxies ngc 4406 and 6876 , while ngc 4382 is yet another galaxy with a central minimum that resembles m31 .
lastly , we note that the galaxies with offset centers , which we discuss in @xmath82 may be poorly resolved examples of double nuclei or galaxies with central minima .
the core galaxy ngc 4073 is the brightest galaxy in the poor cluster mkw4 ; it is likely to have cannibalized other cluster members .
@xcite obtained a rotation curve and stellar velocity dispersion profile .
the central few arcseconds of the kinematic profiles reveal a @xmath83 depression in velocity dispersion and counter - rotation .
@xcite thus argue that ngc 4073 has a kinematically - decoupled core .
the envelope of ngc 4073 has a fairly high ellipticity of 0.4 ; however , for @xmath84 the isophotes rapidly become nearly circular and show a strong twist of over @xmath85 as is evident in the profiles shown in figure [ fig : surf ] , and in a contour plot of the galaxy center ( figure [ fig : n4073_con ] ) .
the maximum surface brightness is reached in a `` ring '' at @xmath86 with the brightness falling by 0.14 mag at the center of the galaxy . at larger radii the surface brightness profile ( figure [ fig : surf ] ) first _ increases _ in slope with decreasing radius for @xmath87 before abruptly flattening off just outside the ring .
superficially , a high contrast image of the galaxy center ( figure [ fig : n4073_im ] , left ) suggests that the center is obscured by a dust patch ; however , an archival nicmos-2 image ( figure [ fig : n4073_im ] , right ) ( gto 7820 ) in the @xmath88 band also shows a central minimum , despite its lower angular resolution .
we conclude that we are seeing the intrinsic distribution of starlight rather than obscuration .
a close examination of the image and contour plot shows that the central ring is not exactly symmetric , but is slightly brighter to one side .
a cut through the center at position angle @xmath89 shows this clearly ( figure [ fig : n4073_cut ] ) by going through the brightest part of the ring .
the maximum brightness along the cut occurs at @xmath90 from the central minimum , slightly further away from the center than the diametrically - opposed maximum , which occurs at @xmath91 from the central minimum .
the implied eccentricity of the ring is @xmath92 the slightly asymmetric brightness distribution could be explained by an eccentric torus of stars @xcite .
however , the ring - radius corresponds to @xmath93 pc , a considerably larger scale than that of the central minima discussed in @xcite , and one that is outside the nominal @xmath94 pc sphere - of - influence of a central black hole based on the velocity dispersion of ngc 4073 ( table [ tab : glob ] ) and the @xcite @xmath95 relationship .
if so , the stability of such a torus might therefore be in question . the central structure of ngc 4382
, a bright s0 core galaxy in the virgo cluster , somewhat resembles the double nucleus of m31 ; here the central minimum is a `` valley '' between two brightness peaks .
this might result from the projected appearance of a much more eccentric torus or stellar disk than that seen in ngc 4073 . as can be seen in figures [ fig : n4382_im ] and [ fig : n4382_con ] the isophotes become asymmetric for @xmath96 rounding out and becoming somewhat diffuse at position angle @xmath97 this side of the nucleus suggests a reduced - amplitude version of the off - center p1 component of m31 ; indeed , a local surface brightness peak occurs at @xmath98 separated by @xmath99 from the brightest portion of the nucleus , which itself is within @xmath100 of the galaxy envelope photocenter . based on the velocity dispersion of ngc 4382 ( table [ tab : glob ] ) and the @xcite @xmath95 relationship , the estimated sphere - of - influence of a central black hole is @xmath101 again like m31 ,
the isophotes become more eccentric and twist @xmath102 within @xmath103 as the nucleus is approached ( figure [ fig : surf ] ) . the position angle of the line connecting the two brightness peaks
is itself twisted away from isophotes of semimajor axes of only @xmath104 the isophotes on the side of the photocenter opposite the secondary peak become highly eccentric , appearing as an almost linear extension of the central nucleus ; this in turn resembles the p2 component of m31 .
similar structure is seen in both the @xmath12 and @xmath13 bands ; a color - ratio image ( figure [ fig : n4382_im ] ) is flat ( except for a slight radial gradient ) , showing no evidence for dust or change in stellar population associated with the unusual nuclear structure . from other data ,
ngc 4382 appears to be an excellent candidate for a fairly recent merger ; the central structure noted here may be related to the unusual spectral properties of the central regions detected from ground - based observations .
ngc 4382 has the second - highest fine - structure measure in the @xcite sample , who argue that the fine - structure measure is indicative of the time and strength of the last merger .
@xcite obtained long - slit line - strength measures over the central region of ngc 4382 , finding it to have an exceptionally large @xmath105 absorption line - width of 2.6 ; they argued that the central population of ngc 4382 is as young as 3 gyr .
they also confirm that the galaxy becomes steadily _ bluer _ with decreasing radius for @xmath106 as was first noted by @xcite .
the present @xmath70 color profile of ngc 4382 also shows a centrally blue gradient for @xmath107 as discussed in @xmath108 nearly all galaxies have @xmath70 color gradients that become redder at small radii .
we also note that ngc 4382 has an average @xmath70 color that is bluer than all but one galaxy in the present sample ( table [ tab : vmi ] ) .
we do , however , disagree with some of the finer details of the ngc 4382 color distribution presented by @xcite we find that the nucleus is essentially the same color as the surrounding galaxy for @xmath109 and we do not find any evidence for a central ring of particularly blue emission .
@xcite notes that the stellar velocity dispersion profile of ngc 4382 decreases for @xmath84 which is confirmed in sauron observations @xcite .
high spatial - resolution spectroscopy with the oasis instrument shows that the core within @xmath110 counter - rotates with respect to the outer envelope of the galaxy @xcite . in our data ,
the location of maximum surface brightness is typically concentric with the surrounding galaxy , as measured by the center of the isophotes on @xmath111 scales , with a median precision of @xmath112 75% of the sample has the peak brightness concentric to within @xmath113 in physical units , this means that the galaxy centers are aligned at the sub - parsec level with the isophotes on the scale of the break radii farther out .
any lopsidedness or strong @xmath114 asymmetries on the scale of the galaxy cores is therefore extremely rare ( or of extremely low amplitude ) .
however , we identify five galaxies ( table [ tab : morph ] ) in which the center of the surface brightness cusp appears to be significantly displaced from the surrounding galaxy . in ngc 507
, the brightness peak is displaced @xmath115 from the surrounding core ; this is readily apparent in a contour plot ( figure [ fig : n0507_con ] ) and in the residuals of the surface photometry model ( figure [ fig : images ] ) .
examination of the image of ngc 1374 shows that its brightness peak is clearly offset from its core ( figure [ fig : n1374_con ] ) , but the measured offset is only @xmath116 the image of ngc 7619 looks normal , but the surface photometry model shows a clear pattern associated with an offset peak , which is found to be displaced from the core by @xmath117 ngc 4291 and 5576 show similar central offsets when the model is compared to the images .
in contrast , @xcite identified ngc 1700 as a possible candidate for an offset center , but the improved imagery presented here shows no such effect .
the five galaxies identified here with offset centers are all core galaxies , but the implied @xmath118 fraction of core galaxies with offset centers is too small to rule out similar behavior in the power - law galaxies , which are more poorly represented in the present sample . it is now recognized that many early - type galaxies are triaxial , rather than axisymmetric . the term triaxial " is usually taken to mean that the galaxy is symmetric with respect to three orthogonal principal planes ( the @xmath119 point symmetry group ) ; if this symmetry is present then all isophotes should be concentric .
however , there is no known dynamical argument that prohibits self - consistent stellar systems without this symmetry ( e.g. , lopsided systems ) .
the observation that the centers of most early - type galaxies have concentric isophotes therefore shows that some aspect of the galaxy - formation process strongly favors systems with @xmath119 symmetry . in the rare exceptions such as m31 @xcite
, the lopsidedness appears to be localized within the sphere of influence of the central black hole , which is natural since orbits do not precess in a kepler potential , so a lopsided orbit distribution is not axisymmetrized by phase mixing ( e.g. , * ? ? ?
the coincidence of the point of maximum surface brightness with the centroid of the isophotes on larger scales also suggests that any massive black hole is located at the bottom of the potential well due to the stellar mass distribution to within a fraction of a parsec .
massive black holes are normally expected to be close to the potential minimum because they rapidly lose orbital energy due to dynamical friction ; however , the black hole will execute brownian motion relative to the potential minimum due to gravitational encounters with nearby stars .
the expected rms amplitude @xmath120 and velocity @xmath121 of this motion have been discussed by many authors .
@xcite argued that the black hole should be in energy equipartition with the unbound stars surrounding it ; for a hole of mass @xmath122 embedded in a homogeneous core with core radius @xmath123 we then expect that @xmath124 and @xmath125 where @xmath126 is the typical stellar mass and @xmath127 is the velocity dispersion in the core .
@xcite obtained a similar result , which they confirmed using n - body simulations .
however , these results are for galaxies with homogeneous cores , not the singular density distributions that are found in both core and power - law galaxies
. there has been very little theoretical work on the expected value of @xmath120 in galaxies with cusps . in the case where the galaxy density profile is a power law , but do not discuss @xmath128 .
] however , it is likely that the formula for a homogeneous core provides an approximate upper limit to @xmath128 in a singular core , when the core radius of the homogeneous core is replaced by the break radius of the singular core . for characteristic values @xmath129pc and @xmath130 we find @xmath131pc , too small to be resolved by _ hst . _
black holes in a binary may also be displaced from the center .
the expected binary semi - major axes are in the range 0.01 10 pc @xcite ; binaries at the larger end of this range may be within the resolution limit of _ hst .
_ the conclusion that brownian wandering of a central black hole is not responsible for the offset centers leaves us with the hypothesis that these galaxies may be unresolved analogs to m31 ; this is suggested by the highly local nature of the offset centers plus the increasingly rich sample of early - type galaxies that have complex central structure .
@xcite noted that the m31 nucleus was displaced from the m31 bulge based on ground - based observations , which had @xmath132 pc limiting resolution , prior to the _ hst _ discovery of the complex nuclear morphology @xcite .
if m31 were at the distance of virgo , its double nucleus would remain unresolved by _ hst , _ and would look like an off - center galaxy .
the estimated sizes of the black hole sphere - of - influence range from @xmath133 for ngc 5576 to @xmath134 for ngc 4291 , significantly larger than the observed central offsets for all five galaxies .
the structure of these galaxies is thus not inconsistent with the @xcite model for m31 .
it is already well - established that dust features are common in _ hst _ images of the centers of early type galaxies @xcite . in our own sample , dust is visible in 38 out of 77 galaxies .
this frequency of 49% agrees well with that of previous studies ( @xcite : 40% ; @xcite : 23% ; @xcite : 41% ; @xcite : 48% ; @xcite : 75% ; @xcite : 56% ; @xcite : 43% ( their unbiased sample ) ) . if we combine the samples of @xcite and @xcite with the present sample ( this combination largely encompasses all _ hst _ samples ) , we find the overall frequency of dust to be 47% in a total of 177 galaxies ( see table [ tab : dustfreq ] for details ) .
the sample of @xcite avoided known dust , whereas one quarter of the @xcite sample was strongly biased toward it ; the present sample has no obvious biases for dust .
the combined sample is large enough that the observed dust fraction should approach a representative value for bright ellipticals .
dust is about equally prevalent in both power - law galaxies and cores ; this agrees with frequencies seen in @xcite and @xcite .
dust morphologies were determined by eye using the @xmath135 postage stamps shown in figure [ fig : images ] .
we should clarify that by `` dust '' in this paper we always mean clumpy dust that is visible as dark patches in _ hst _ images . a diffuse ( and possibly more massive ) warm dust component
may also be present , detected by iras ( e.g. @xcite , @xcite , @xcite ) , but that is not the focus here .
following @xcite , we distinguish nuclear rings and disks ( which we lump together as `` rings '' ) from all other structures ( termed `` non - rings '' ) .
ring structures are presumably highly flattened , circular , and axi - symmetric @xcite , implying ages that are at least a few orbital times ( an orbital time at 100 pc is a few million years ) .
the non - ring class is further sub - divided into `` spirals '' ( with sheared dust patterns ) , and `` chaotics '' ( random blobs ) .
a handful of transitional cases lie between the rings and non - rings .
this schema is motivated by the notion that we are seeing various stages of a settling sequence " ( e.g. , @xcite ; @xcite ) .
this idea is discussed further below , but it rests heavily on the fact that non - ring dust _ must _ be in a dynamically short - lived configuration .
the settling hypothesis says that it is falling into the center to form nuclear rings .
the most striking result of the dust morphology census is that the ratio of rings to non - rings is only 0.6 , consistent across all three sub - samples .
this small ratio surprised us ; if rings are the descendants of short - lived non - rings , we expected to find a much larger ratio of rings to non - rings .
in fact , the opposite is true .
we return to this point below .
we studied larger model - subtracted images covering most of the pc1 ccd to learn about the radial distribution of dust beyond @xmath136 which may be relevant to the settling model .
dust concentration was graded qualitatively , with ` 0 ' meaning that the dust is highly _
un_concentrated dust is more prevalent beyond the @xmath135 postage stamp than in it , ` 1 ' means that dust is visible both inside and outside of the postage stamp , and ` 2 ' means that dust is seen only in the postage stamp itself . the three dust properties , strength ( as already quantified in @xmath137 ) , morphology , and radial concentration , are presented in table [ tab : morph ] .
only one of our 38 dusty galaxies ( ngc 6876 ) has unconcentrated dust ( concentration class 0 ) .
thus , when dust appears in the envelope of the galaxy , it essentially always in seen in the center , too .
the converse is not true .
indeed , it is striking that galaxies with nuclear rings are completely devoid of dust outside the ring : not one dust cloud is to be found in these objects beyond the ring itself ! overall , the radial pattern of dust suggests an _ episodic _ settling model in which dust appears at times throughout the galaxy and then falls to the center .
galaxies with lots of dust would be near the beginning of the cycle , when dust is still widely distributed .
nuclear rings would be the endpoint of an episode , by which time the rest of the galaxy would be dust free .
the fact that galaxies with weaker dust also tend to be highly concentrated ( table [ tab : morph ] ) suggests that dust is being destroyed as it sinks to the center ; it falls in , leaving just a few weak shreds by the time it nears the nucleus .
a possible mechanism for dust destruction is sputtering by hot x - ray gas .
survival of some but not all dust would then require that the timescale for dust destruction by sputtering be comparable to the timescale for infall .
this is consistent with the dust - radiative cooling model of @xcite ( see also @xcite ) .
many papers have found positive correlations between dust and nuclear optical emission , extended optical emission , non - thermal radio activity , and far - ir dust emission ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
we decided to explore the relationship between dust and _ central _ optical emission since that relation seemed particularly close @xcite .
optical emission - line strengths were taken from the literature ( sources given in table [ tab : morph ] ) .
the emission is ranked qualitatively into four grades in table [ tab : morph ] : 0 ( no emission ) , 1 ( weak emission ) , 2 ( strong emission ) , and 3 ( very strong emission ) .
some papers measure extended emission , but most are based on spectroscopic data and thus measure only nuclear emission , which in our galaxies has line ratios characteristic of active nuclei ( seyferts and liners ) . since extended and nuclear emission correlate well @xcite , we assume that any optical emission , however weak , signifies _ nuclear activity .
_ the presence of dust in the inner @xmath135 of our galaxies correlates remarkably well with the presence of emission : 96% of galaxies without emission lack dust , whereas 90% of galaxies with emission have at least some dust .
furthermore , the _ amounts _ of dust and emission also go together : mean dust strength in weak emission galaxies is only 1.09 , versus 1.85 in strong - emission and very - strong - emission galaxies ( see table [ tab : dustvsemiss ] ) .
@xcite also found a close connection between dust and emission , though not quite so tight as in our sample .
notably , emission strength does not appear to depend on dust morphology ; as also found by @xcite ; rings and non - rings with the same dust strength have qualitatively similar emission strengths .
this connection between dust and emission is important if it is assumed , as here , that optical emission always signifies gas falling onto a black hole .
it then follows that visible dust actually _ shows _ us gas that is about to fall onto the hole .
the distribution and motions of dust become essential clues to the feeding of local agns , a point that was also stressed by @xcite .
figure [ fig : dust_mag ] plots the histograms of dusty and clean galaxies in virgo as a function of absolute magnitude .
dust is found in roughly half the galaxies brighter than @xmath138 ( which is the normal frequency ) , whereas _ all _ 16 dimmer galaxies in virgo are dust free .
the probability of this happening by chance in 16 galaxies is only @xmath139 if the dividing line were placed at @xmath140 it would still be only @xmath141 .
either dust is intrinsically rare in faint ellipticals , or the cluster environment discourages the accumulation of dust in small galaxies . deciding between these hypotheses is problematic since the only good selection of faint galaxies is in the virgo cluster .
the episodic settling model implies that we should to be able to identify galaxies that are in different stages of settling .
a possible sequence is offered in figure [ fig : sequence ] .
the transitional objects that appear to be in the process of settling from a non - ring configuration into a ring are particularly important .
four ( perhaps five ) of these are in our data , two of which are shown in figure [ fig : sequence ] .
@xcite note similar transitional structures in radio galaxies , which they call lanes . "
lanes are geometrically distinct from rings and are more skewed with respect to stellar principal planes , indicating transient configurations that they suggest are in the process of settling .
if all these galaxies are really parts of a sequence non - rings , rings , to empty then setting the lifetime of any one phase sets the lifetime of the other two phases .
a good place to start is the non - ring phase . since non - ring dust can not be in dynamical equilibrium ,
a reasonable estimate of the lifetime is the local dynamical time .
the orbital time at 1 kpc in a bright elliptical galaxy is @xmath142 yr , and at the edge of a 100 pc nuclear ring is ten times shorter .
dust sputtering times ( even with self - shielding ) are also of order @xmath143 years @xcite , which is consistent with destruction of some fraction of the dust as it falls in .
finally , once gas falls close to a rotating disk , the timescale to settle into a principal plane of the gravitational potential is again a few orbital times @xcite .
the ratio of rings to non - rings can now be used to estimate the lifetime of the ring phase . since that ratio is less than unity ( 0.6 ) , the nuclear rings must be even shorter - lived ! @xcite model the kinematics of the nuclear ring in ngc 4261 , calling attention to the high turbulent velocity near the center .
this central line - broadening is a common feature of nuclear rings ( e.g. , ngc 7052 , @xcite ) and may be ubiquitous .
it implies high turbulent viscosity , say @xcite , which causes the inner gas to flow inward and feed the black hole .
their estimated timescale for emptying out the inner ring is @xmath3 yr , not far from the timescale needed by the settling model .
finally , rings should empty from the inside out .
one might therefore expect a population of ghost " rings that are the remnants of previously filled disks ; four such candidates are in our sample , two of which are exceedingly faint ( ngc 741 , 2778 , 3379 , and 3608 ) .
a further interesting observation is the finding by @xcite that dust grains are smaller in rings than in non - rings .
this is consistent with the hypothesis that the grains in rings were subjected to more sputtering by the hot gas , and are therefore older .
the final step is to deduce the length of time that galaxies remain dust free , which sets the length of the whole cycle .
since half of all bright ellipticals are dust free , the length of the dust - free phase must be of order that of the dusty phase , that is , a few times @xmath144 yr , and the length of the whole cycle is only @xmath145 yr .
the assumption of a settling sequence thus implies that _ clumpy dust is appearing and disappearing continually in ellipticals .
_ this would help to understand why there seems to be little correlation between the presence of dust and merger remnants @xcite ; merger phenomena are comparatively long - lived and would show little correlation with more rapid dust cycling .
the same reason would explain why there is little correlation between mean light - weighted stellar population ages and the presence of dust ( d. forbes , private communication ) .
the `` dust settling '' hypothesis leaves at least three major questions unanswered .
the first is the origin of the dust , whether internal ( shed by stellar winds ) or external ( accreted from outside ) .
we find it striking that there is not the slightest evidence , even in very dusty galaxies , for any extra starlight , either distributed in clumps throughout the galaxy or in the neighborhood of the dust clouds themselves .
our subtracted models should be very sensitive to clumpy starlight , yet no trace of stars is seen .
any dust accreted from the outside must therefore be essentially _ star free _ , which excludes small dwarf galaxies as sources , or even tidal tails .
we can not think of any plausible external sources of dust that would be completely star free .
in contrast , a known source of dust _ within _ the galaxies is m - star winds , which are thought to have been detected in both the mid - ir and far - ir ( e.g. * ? ? ? * ; * ? ? ?
* ; * ? ? ?
however , this dust is injected uniformly throughout the galaxy and is not expected to clump .
perhaps there is a thermal instability that causes the hot gas to cool into clumps , bringing dust with it .
a thermal instability driven by dust cooling @xcite might play such a role .
finding such a mechanism would bolster the internal origin theory .
a second problem stems from the fact that the overall length for the settling cycle seems uncomfortably short , only at most a few times @xmath146 yr .
if true , this points strongly to an internal origin for the dust , as the dynamical infall time from the edge of a galaxy is too long , more like 1 gyr . yet
numerous studies have revealed galaxies in which the distribution and motions of ionized gas and dust seem _ unrelated _ to the motions of stars , and even counter - rotate ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
these data have been widely interpreted as incontrovertible evidence for the _ external _ origin of dust , at least in certain key objects .
we have inspected our sample to identify our best candidates for external dust origin , using the simple criteria that the dust be extended and that its distribution over the galaxy differ strongly from that of the underlying starlight .
six objects meet these criteria ( ngc 1316 , 2768 , 4026 , 4589 , 5018 , and 7727 ) .
looking at them , it is indeed hard to imagine how the dust could have come from any quasi - uniform process _ inside _ these systems ( see the large - scale model - subtracted images noted in the footnote above ) .
five of these objects have strong dust , and four also have strong optical emission , which would imply that strong nuclear activity is triggered by external accretion in at least these galaxies .
this would be consistent with the finding that two of the strongest known liners ( ngc 4278 and ngc 1052 ) have external h i clouds in their vicinity @xcite as does ngc 5018 @xcite , another agn in our sample . on the other hand , @xcite searched for relationships between dust strength and external environments , finding no connection between dust and cluster versus group membership , or the distance to the nearest neighboring galaxy .
new data on the interstellar medium in ellipticals may help to solve this problem by _ overturning _ the old prediction that gas and star kinematics must agree .
a consensus seems to be building that the hot interstellar medium ( ism ) in ellipticals is highly turbulent and is being stirred periodically by nuclear activity @xcite .
stirring of the ism would strongly perturb warm ionized gas and dust , and would explain the many examples of highly irregular , disorganized velocity fields that are detected in warm gas ( e.g. , @xcite ) .
any dust and gas thrown out to the periphery of galaxies by stirring could be further perturbed by the passage of neighboring galaxies or the pressure of hot intracluster winds . in such a way , the motions of dust , gas , and stars might become strongly decoupled , giving the appearance of an external origin for clumpy gas and dust when none is required .
the most important problem with the cycling picture is a clear idea of how all the stages fit together and what drives the whole cycle .
why does dust begin to condense into clouds in the first place ?
why does the black hole accrete only dust - related gas and not other gas ?
what stops dust from condensing ?
if it is feedback from the black hole , what exactly is its nature and how does it interface with dust and dust - related gas ?
these questions are basic to the cycling model but currently lack answers .
we conclude by highlighting the beautiful but puzzling object ngc 3607 ( figure [ fig : n3607 ] ) .
this galaxy contains a large , dusty outer disk that must be dynamically old because of its extraordinary symmetry and tightly wrapped arms .
however , this outer disk appears to transition rapidly but smoothly at the center to a second gas disk that is _ perpendicular _ to the first and is seen nearly edge on !
this inclined disk seems to be settling onto an inclined nuclear ring .
how can this happen ?
a facile explanation would invoke external infall of gas directly into the galaxy center , yet no disturbance in the outer dust or other sign of an interaction is seen .
thanks to its many regularities but schizophrenic nature , ngc 3607 may be a rosetta stone that will reward further study .
we have examined the central structure of early - type galaxies ; a component of this analysis has included searching for morphological differences between core and power - law galaxies .
we conclude by summarizing what this analysis might say about the formation of central structure in the context of hierarchical merging of galaxies with central black holes ; in @xcite we argued that cores and power - laws are different outcomes of this process .
this discussion is largely speculative while modern merger simulations are able to study the gross effects of mergers on central structure , there are few firm theoretical predictions of the phenomena considered in the analysis sections .
the picture presented by @xcite is that power - law galaxies were formed by a process in which central _ gaseous _ dissipation and attendant star formation took place , while core galaxies were generated in the mergers of galaxies largely _ free of gas . _ the first generation of core galaxies
would be made by merging power - laws , which themselves would harbor central black holes formed in the initial collapse of the primordial galaxy .
the cores were formed in the final stage of the merger when the central black holes of the merging galaxies formed a binary that ejected stars from the center as it hardened .
we can distill this scenario to be simply that power - law centers are sites of dissipation and concentration , while cores are sites of mixing and scattering . in the case of color gradients ,
this picture leads us to expect power - law galaxies to have stronger gradients than core galaxies . as star formation progresses following a merger or initial collapse of gas - rich systems ,
the inventory of gas is simultaneously slowly depleted and centrally concentrated , while becoming increasingly metal - enriched .
the stellar density profile would reflect both the concentration of gas over time and the metal - enrichment history . merging
a then gas - free power - law galaxy with a similar system would not completely erase any pre - existing color gradients , but some dilution of the gradient might be expected .
strong mixing of stars interior to the break radius associated with the formation of a core , however , might cause significant flattening of the gradient at small radii .
the present data do show that power - law galaxies on average have steeper color gradients than do core galaxies , although the difference is small and the scatter about the mean gradient for both types is large . whether or not the color gradients in core galaxies flatten further interior to the break radius
is unclear .
there are some core galaxies that do appear to show this effect , but the majority do not .
unfortunately , in absolute terms the envelope gradients in core galaxies are small to begin with , and it is thus especially difficult to detect even smaller gradients over a limited radial range . we conclude that color gradients are broadly consistent with the formation of cores by binary black holes with the caveat that the effects seen are modest in absolute terms
. the discovery that core galaxies twist interior to their break radii at first glance also appears to support the formation of cores by binary black holes . if the shape of the core is related to the angular momentum vector of the binary black hole , which itself may have little to do with the principal axes of the merged galaxy , then it is easy to imagine that twists occur as one transitions from the outer regions of the core into the inner region dominated by the black hole .
however , our analysis shows that the absolute physical significance of the twists must also be considered .
since the inner twists of both cores and power - laws are about the same in terms of the strength of the photometric perturbation required to twist the galaxy , it is not possible to argue that the _ angularly _ larger twists in core galaxies are a consequence of their particular formation mechanism .
the interpretation of nuclei in early - type galaxies depends on whether the nuclei are the non - stellar emission of low - level agn , or are star clusters of the sort seen in m33 .
the existence of nuclear star clusters in cores raises a few questions . in @xcite
we argued that the survival of cores , once created , against infalling material in subsequent mergers requires the central black hole to serve as a `` guardian '' that tidally disrupts any dense stellar aggregates that reach the center , a notion that has been validated by n - body simulations @xcite .
it thus appears that nuclear star clusters can not be pre - existing systems that have arrived at the center through dynamical friction .
the likely alternative is that such nuclei were created _ in situ _ by the central accretion of small amounts of gas with subsequent star formation .
this scenario may be supported by the bluer colors of the nuclei if they indicate that the nuclei are younger than the rest of the central stellar population .
the remaining issue is to understand how converting any infalling gas into stars competes with accretion into the black hole , itself .
the same problem exists in the center of our own galaxy , where a very young star cluster surrounds a central black hole ( see the review by @xcite ) .
we conclude that while formation of nuclear clusters is an interesting problem , their existence does not offer any obvious insight into the formation of core versus power - law galaxies .
this conclusion also applies to agn , which in this sample offer little more than a ratification of the already standard notion that black holes are likely to exist at the centers of both types of galaxies .
the discovery of two more galaxies that have central minima in their brightness distributions in addition to the galaxies discussed in @xcite implies that @xmath147 of core galaxies have such structures . as noted in @xcite , there are two competing ideas for the formation of these objects . in the first scenario , central minima might be explained by diffuse tori of stars added to a pre - existing core .
the tori result from the tidal disruption of dense stellar aggregates transported to the center of the galaxy by dynamical friction following a merger ; again , the black hole prevents the core once established from being filled in by latter mergers . in a second and more speculative scenario ,
the central minima are a direct consequence of stars ejected from the galaxy center by a binary black hole
. it may be noteworthy that ngc 4073 and a number of the galaxies discussed in @xcite have relatively high luminosities . in the discussion in that paper
, we advanced the hypothesis that the mergers of two galaxies each with pre - existing cores , as would occur at the high - mass ends of a merging hierarchy , might be the necessary precondition for generating a merged galaxy in which the core was actually evacuated by the binary black hole . under this picture
, galaxies with central minima would be vivid examples of the action of binary black holes on central structure ; however , present theoretical and observational treatment of this problem leaves this as an unproven hypothesis .
likewise , we presently do not know what bearing galaxies with offset centers have on understanding the formation of central structure .
if these systems are indeed unresolved analogs of the m31 nucleus , then their role may be more to show how nuclear black holes can stabilize non - axisymmetric structures in galaxy centers .
in @xcite we noted that power - law galaxies have disky outer isophotes ; the conclusion here that all inclined power - law galaxies have visible stellar disks , with the implication that disks exist in _ all _ power - law galaxies , supports the dissipative formation scenario for power - law galaxies .
conversely , the corresponding lack of disks in all but the lowest - luminosity core galaxies supports their formation in dissipationless gas - free processes .
a transition zone in galaxy luminosity where the dissipation and dissipationless stellar processes compete in mergers may explain the disks in the lower - luminosity galaxies with cores .
lastly , the rounder shapes of the inner isophotes of core galaxies as compared to power - law galaxies may be a direct consequence of core formation by a binary black hole on the presumption that scattering of stars by the binary is most likely to be nearly isotropic . in the final analysis section , we have reasoned that the high prevalence of clumpy dust in non - equilibrium conditions means that it is generated in processes interior to the galaxies , themselves .
we have further argued that dust clouds once created settle into the center , and are eventual destroyed or processed by the galaxy on time scales of @xmath3 yr .
the continued existence of dust clouds in a substantial fraction of early - type galaxies suggests that dust clouds are formed and destroyed in a cyclical process .
the strong link between the presence of dust and nuclear emission argues that one outcome of this process would be episodic delivery of matter to the centers of the galaxies ; understanding the secular evolution of cores may have to take this into account . at the same time
we see no properties of present - day dust that bear one way or another on the basic formation mechanism of cores versus power - laws . in conclusion
, it appears that the diverse morphological measures derived for galaxies here give at most weak evidence for differences between core and power - law galaxies that might be taken as support for the binary black hole core formation mechanism .
dust is formed equally in both types , and the nuclei offer no apparent insight into this problem . understanding the galaxies with central minima may offer a useful path for understanding core formation , but presently the mechanisms for forming central minima are too uncertain for forming conclusions .
position - angle twists , when properly normalized , seem equally strong in both types .
cores do have shallower color gradients , as expected , but the effect is weak .
the strongest evidence comes from the rounder inner isophotes of cores and , more important , the high frequency of disks in the centers of power - law galaxies and ( to a lesser degree ) in the fainter core galaxies .
overall the evidence is consistent with the binary black hole core formation mechanism , but it is not a strong endorsement .
the real value of our data , we feel , is not in current tests of core formation , but in providing a rich set of observations that can be used to confront future models .
any successful model will have to match the large suite of quantitative data produced in this paper .
finally , we note that this paper does not consider the fundamental radial distribution of starlight in galaxies , which distinguishes cores from power - laws in the first place .
this is the topic that taken up in paper vi , where we demonstrate a strong dichotomy between the two types , validating the assumption of two classes in this paper .
this research was supported in part by several grants provided through stsci associated with go programs 5512 , 6099 , 6587 , 7388 , 8591 , 9106 , and 9107 .
our team meetings were generously hosted by the national optical astronomy observatory , the observatories of the carnegie institution of washington , the aspen center for physics , and the leiden observatory .
we thank dr .
barbara ryden for useful conversations on isophote twists .
bahcall , j. n. , & wolf , r. a. 1976 , , 209 , 214 begelman , m. c. , blandford , r. d. , & rees , m. j. 1980 , , 287 , 307 bender , r. , & m " ollenhoff , c. 1987 , , 177 , 71 bertola , f. , bettoni , d. , rusconi , l. , & sedmak , g. 1984 , , 89 , 356 bertola , f. , buson , l. m. , & zeilinger , w. w. 1988 , , 335 , 705 bertola , f. , cappellari , m. , funes , j. g. , corsini , e. m. , pizzella , a. , & vega beltr ' an , j. c. 1998 , , 509 , l93 bertola , f. , cinzano , p. , corsini , e. m. , rix , h. , & zeilinger , w. w. 1995 , , 448 , l13 beuing , j. , bender , r. , mendes de oliveira , c. , thomas , d. , & maraston , c. 2002 , , 395 , 431 biller , b. a. , jones , c. , forman , w. r. , kraft , r. , & ensslin , t. 2004 , , 613 , 238 bower , g. , et al .
2001 , , 550 , 75 buson , l. m. , et al .
1993 , , 280 , 409 byun , y. i. , et al .
1996 , , 111 , 1889 caldwell , n. 1984 , , 96 , 287 caon , n. , macchetto , d. , & pastoriza , m. 2000 , , 127 , 39 cappellari , m. , renzini , a. , greggio , l. , di serego alighieri , s. , buson , l. m. , burstein , d. , & bertola , f. 1999 , , 519 , 117 carollo , c. m. , franx , m. , illingworth , g. d. , & forbes , d. 1997 , , 481 , 710 carter , d. l. , prieur , j. l. , wilkinson , a. , sparks , w. b. , & malin , d. f. 1988 , , 235 , 813 chatterjee , p. , hernquist , l. , & loeb , a. 2002 , , 572 , 371 corsini , e. m. , & bertola , f. 1998 , journal of korean physical society , 33 , 574 crane , p. , et al .
1993 , , 106 , 1371 de koff , s. et al .
2000 , , 129 , 33 ebisuzaki , t. , makino , j. , & okumura , s. k. 1991 , , 354 , 212 emsellem , e. , et al .
2004 , , 352 , 721 enlin , t. a. , & heinz , s. 2002 , , 384 , l27 fabbiano , g. , fassnacht , c. , & trinchieri , g. 1994 , , 434 , 67 faber , s. m. , tremaine , s. , ajhar , e. a. , byun , y. , dressler , a. , gebhardt , k. , grillmair , c. , kormendy , j. , lauer , t. r. , & richstone , d. 1997 , , 114 , 1771 faber , s. m. , wegner , g. , burstein , d. , davies , r. l. , dressler , a. , lynden - bell , d. , & terlevich , r. j. , 1989 , , 69 , 763 ferrarese , l. , van den bosch , f. c. , ford , h. c. , jaffe , w. , & oconnell , r. w. 1994 , , 108 , 1598 ferrari , f. , pastoriza , m. g. , macchetto , f. , & caon , n. 1999 , , 136 , 269 fisher , d. 1997 , , 113 , 950 fisher , d. , franx , m. , & illingworth , g. 1996 , , 459 , 110 fisher , d. , illingworth , g. , & franx , m. 1995 , , 438 , 539 forbes , d. a. 1991 , , 249 , 779 fruchter , a. s. , & hook , r. n. 2002 , , 114 , 144 gebhardt , k. , richstone , d. , ajhar , e. a. , lauer , t. r. , byun , y. , kormendy , j. , dressler , a. , faber , s. m. , grillmair , c. , & tremaine , s. 1996 , , 112 , 105 gebhardt , k. , richstone , d. , kormendy , j. , lauer , t. r. , ajhar , e. a. , bender , r. , dressler , a. , faber , s. m. , grillmair , c. , magorrian , j. , & tremaine , s. 2000 , , 119 , 1157 gebhardt , k. , et al .
2003 , , 583 , 92 gonzalez , j. g. 1993 , ph.d .
thesis , university of california , santa cruz goudfrooij , p. , & de jong , t. 1995 , , 298 , 784 goudfrooij , p. , de jong , t. , hansen , l. , & norgaard - nielson , h. u. 1994a , , 271 , 833 goudfrooij , p. , hansen , l. , jorgensen , h. e. , & norgaard - nielson , h. u. 1994a , , 105 , 341 ho , l. c. , filippenko , a. v. , & sargent , w. l. w. 1995 , , 98 , 477 . ho ,
l. c. , filippenko , a. v. , & sargent , w. l. w. 1997 , , 112 , 315 holley - bockelmann , k. , & richstone , d. o. 2000 , , 531 , 232 holtzman , j. a. , burrows , c. j. , casertano , s. , hester , j. j. , trauger , j. t. , watson , a. m. , & worthey , g. 1995 , , 107 , 1065 humason , m. l. , mayall , n. , & sandage , a. r. 1956 , , 61 , 97 jaffe , w. , ford , h. , ferrarese , l. , van den bosch , f. , & oconnell , r. w. , 1996 , , 460 , 214 jaffe , w. , ford , h. c. , oconnell , r. w. , van den bosch , f. c. , & ferrarese , l. 1994 , , 108 , 1567 james , f. , & roos , m. 1977 , cern / d internal report 77/506 kannappan , s. j. , & fabricant , d. g. 2001 , , 121 , 140 kim , d .- w .
1989 , , 346 , 653 knapp , g. r. , guhathakurta , p. , kim , d. , & jura , m. a. 1989 , , 70 , 329 kormendy , j. 1985 , , 295 , 73 kormendy , j. , & bender , r. 1996 , , 464 , l119 kormendy , j. , bender , r. , richstone , d. , ajhar , e. a. , dressler , a. , faber , s. m. , gebhardt , k. , grillmair , c. , lauer , t. r. , & tremaine , s. 1996 , , 459 , l57 kormendy , j. , dressler , a. , byun , y .-
i . , faber , s. m. , grillmair , c. , lauer , t. r. , richstone , d. , & tremaine , s. 1994 , in eso / ohp workshop on dwarf galaxies , ed .
g. meylan & p. prugniel ( garching : eso ) , p. 147 .
krist , j. 1995 , asp conf .
ser . 77 : astronomical data analysis software and systems iv , 4 , 349 laine , s. , van der marel , r. p. , lauer , t. r. , postman , m. , odea , c. p. , & owen , f. n. 2003 , , 125 , 478 lauer , t. r. 1985 , , 57 , 473 lauer , t. r. 1986 , , 311 , 34 lauer , t. r. 1999a , , 111 , 227 lauer , t. r. 1999b , , 111 , 1434 lauer , t. r. , ajhar , e. a. , byun , y .-
i . , dressler , a. , faber , s. m. , grillmair , c. , kormendy , j. , richstone , d. , & tremaine , s. 1995 , , 110 , 2622 lauer , t. r. , faber , s. m. , ajhar , e. a. , grillmair , c. j. , & scowen , p. a. 1998 , , 116 , 2263 lauer , t. r. , tremaine , s. , ajhar , e. a. , bender , r. , dressler , a. , faber , s. m. , gebhardt , k. , grillmair , c. , kormendy , j. , & richstone , d. 1996 , , 471 , l79 lauer , t. r. , et al .
1991 , , 369 , l41 lauer , t. r. , et al . 1992a , , 103 , 703 lauer , t. r. , et al . 1992b , , 104 , 552 lauer , t. r. , et al .
1993 , , 106 , 1436 lauer , t. r. , et al .
2002 , , 124 , 1975 laun , f. , & merritt , d. 2004 , astro - ph/0408029 lucy , l. b. 1974 , , 79 , 745 macchetto , f. , pastoriza , m. , caon , n. , sparks , w. b. , giavalisco , m. , bender , r. , & capaccioli , m. 1996 , , 120 , 463 magorrian , j. , et al . 1998 , , 115 , 2285 makino , j. 1997 , , 478 , 58 martel , a. r. , baum , s. a. , sparks , w. b. , wyckoff , e. , biretta , j. a. , golombek , d. , macchetto , f. d. , de koff , s. , mccarthy , p. j. , & miley , g. k. 1999 , , 122 , 81 mathews , w. g. , & brighenti , f. , 2003 , , 590 , l5 mcdermid , r. , et al .
2004 , astron .
nachr . , 325 , 100 michard , r. 1998 , , 335 , 49 milosavljevi ' c , m. , & merritt , d. 2001 , , 563 , 34 m " ollenhoff , c. , & bender , r. 1987 , , 174 , 63 morris , m. , & serabyn , e. 1996 , , 34 , 645 nieto , j .- l . , macchetto , f. d. , perryman , m. a. c. , di serego alighieri , s. , & lelievre , g. 1986 , , 165 , 189 peiris , h. , & tremaine , s. 2003 , , 599 , 237 phillips , m. m. , jenkins , c. r. , dopita , m. a. , sadler , e. m. , & binette , l. , 1986 , , 91 , 1062 prugniel , p. , & simien , f. 1996 , , 309 , 749 quinlan , g. d. , & hernquist , l. 1997 , new astronomy , 2 , 533 raimond , e. , faber , s. m. , gallagher , j. s. , & knapp , g. r. 1981 , , 246 , 708 ravindranath , s. , ho , l. c. , peng , c. y. , filippenko , a. v. , & sargent , w. l. w. 2001 , , 122 , 653 rest , a. , van den bosch , f. c. , jaffe , w. , tran , h. , tsvetanov , z. , ford , h. c. , davies , j. , & schafer , j. 2001 , , 121 , 2431 richardson , w. h. 1972 , j. opt .
a. , 62 , 52 ryden , b. s. , forbes , d. a. , & terlevich , a. i. 2001 , , 326 , 1141 ryden , b. s. , terndrup , d. m. , pogge , r. w. , & lauer , t. r. 1999 , , 517 , 650 sadler , e. m. , & gerhard , o. e. 1985 , , 214 , 177 schweizer , f. , & seitzer , p. 1992
, , 104 , 1039 shields , j. c. , 1991 , , 102 , 1314 sparks , w. b. , macchetto , f. , & golombek , d. 1989 , , 345 , 153 steiman - cameron , t. y. , & durisen , r. h. 1988 , , 325 , 26 tohline , j. e. , simonson , g. f. , & caldwell , n. 1982 , , 252 , 92 tomita , a. , aoki , k. , watanabe , m. , takata , t. , & ichikawa , s. 2000 , , 120 , 123 tonry , j. l. , dressler , a. , blakeslee , j. p. , ajhar , e. a. , fletcher , a. b. , luppino , g. a. , metzger , m. r. , & moore , c. b. 2001 , , 546 , 681 tran , h. d. , tsvetanov , z. , ford , h. c. , davies , j. , jaffe , w. , van den bosch , f. c. , & rest , a. , 2001 , , 121 , 2928 tremaine , s. 1995 , , 110 , 628 tremaine , s. , et al .
2002 , , 574 , 740 van den bosch , f. c. , ferrarese , l. , jaffe , w. , ford , h. c. , & oconnell , r.w .
1994 , , 108 , 1579 van den bosch , f. c. , & van der marel , r. p. 1995
, , 274 , 884 van dokkum , p. g. , & franx , m. 1995 , , 110 , 2027 van gorkom , j. h. , knapp , g. r. , raimond , e. , faber , s. m. , & gallagher , j. s. 1986 , , 91 , 79 verdoes kleijn , g. a. , baum , s. a. , de zeeuw , p. t. , & odea , c. p. , 1999
, , 118 , 2592 verdoes kleijn , g. , & de zeeuw , t. 2005 , in preparation vron - cetty , m .-
p . , & vron , p. 1988
, , 204 , 28 vincent , r. a. , & ryden , b. s. 2005 , in preparation xilouris , e. m. , madden , s. c. , galliano , f. , vigroux , l. , & sauvage , m. 2004 , , 416 , 41 yu , q. 2002 , , 331 , 935 lcclcl ngc 0507 & 6587&f555w&@xmath148 & 22.44 & + ngc 0584 & 6099&f814w&@xmath149 & 21.07 & + & & f555w&@xmath150 & 21.97 & + ngc 0596 & 6587&f555w&@xmath151 & 22.36 & double - sampled + ngc 0741 & 6587&f555w&@xmath152 & 21.84 & + ngc 0821 & 6099&f814w&@xmath149 & 21.28 & @xcite + & & f555w&@xmath150 & 22.18 & + ngc 1016 & 6587&f555w&@xmath148 & 22.37 & + ngc 1023 & 6099&f814w&@xmath153 & 21.48 & @xcite + & & f555w&@xmath153 & 22.38 & + & 6587&f555w&@xmath154 & 21.40 & double - sampled + ngc 1316 & 5990&f814w&@xmath155 & 20.69 & + ngc 1374 & 5446&f606w&@xmath156 & 22.57 & + ngc 1399 & 8214&f606w&@xmath157 & 21.75 & double - sampled + ngc 1426 & 6587&f555w&@xmath151 & 22.22 & double - sampled + ngc 1427 & 5454&f814w&@xmath158 & 21.40 & @xcite + & & f555w&@xmath159 & 22.50 & + ngc 1439 & 5454&f814w&@xmath158 & 21.67 & @xcite + & & f555w&@xmath159 & 22.57 & + ngc 1700 & 5454&f814w&@xmath158 & 21.05 & @xcite + & & f555w&@xmath159 & 21.95 & + & 6587&f555w&@xmath151 & 22.23 & double - sampled + ngc 2300 & 6099&f814w&@xmath160 & 21.56 & + & & f555w&@xmath150 & 22.46 & + ngc 2434 & 5943&f450w&@xmath161 & 23.21 & + & & f814w&@xmath162 & 21.54 & + & & f555w&@xmath163 & 22.26 & + ngc 2768 & 6587&f555w&@xmath164 & 21.99 & double - sampled ; obscured + ngc 2778 & 6099&f814w&@xmath149 & 21.73 & @xcite + & & f555w&@xmath150 & 22.60 & + ngc 2974 & 6587&f555w&@xmath165 & 22.94 & double - sampled + ngc 3115 & 5512&f814w&@xmath166 & 20.83 & @xcite + & & f555w&@xmath166 & 21.65 & + ngc 3377 & 5512&f814w&@xmath166 & 21.38 & @xcite + & & f555w&@xmath166 & 22.29 & + ngc 3379 & 5512&f814w&@xmath167 & 21.40 & @xcite + & & f555w&@xmath168 & 22.30 & + ngc 3384 & 5512&f814w&@xmath166 & 22.35 & @xcite + & & f555w&@xmath166 & 22.35 & + ngc 3557 & 6587&f555w&@xmath169 & & obscured + ngc 3585 & 6587&f555w&@xmath165 & 21.75 & double - sampled + ngc 3607 & 5999&f814w&@xmath170 & 21.24 & + & & f555w&160 & 22.21 & + ngc 3608 & 5454&f814w&@xmath158 & 21.64 & @xcite + & & f555w&@xmath159 & 22.54 & + ngc 3610 & 6587&f555w&@xmath171 & 22.59 & double - sampled + ngc 3640 & 6587&f555w&@xmath165 & 21.82 & double - sampled + ngc 3706 & 6587&f555w&@xmath164 & 21.82 & double - sampled ; @xcite + ngc 3842 & 6587&f555w&@xmath148 & 22.54 & + ngc 3945 & 6633&f450w&@xmath172 & 22.36 & + & & f814w&@xmath173 & 21.77 & + & & f555w&@xmath174 & 22.68 & + ngc 4026 & 9107&f814w&@xmath175 & 21.57 & double - sampled + & & f555w&@xmath175 & 22.83 & double - sampled + ngc 4073 & 6587&f555w&@xmath148 & 22.30 & + ngc 4125 & 6587&f555w&@xmath164 & & double - sampled ; obscured + ngc 4278 & 5454&f814w&@xmath158 & 21.86 & @xcite + & & f555w&@xmath159 & 22.76 & + ngc 4291 & 6099&f814w&@xmath149 & 21.98 & @xcite + & & f555w&@xmath150 & 22.88 & + ngc 4365 & 5454&f814w&@xmath158 & 21.43 & @xcite + & & f555w&@xmath159 & 22.33 & + ngc 4382 & 7468&f814w&@xmath176 & 21.07 & + & & f555w&@xmath177 & 22.07 & + ngc 4406 & 5512&f814w&@xmath168 & 21.40 & @xcite + & & f555w&@xmath168 & 22.30 & + ngc 4458 & 6587&f814w&@xmath148 & 21.55 & + & & f555w&@xmath178 & 22.49 & + ngc 4472 & 5236&f814w&@xmath169 & 21.40 & + & & f555w&@xmath169 & 22.30 & + ngc 4473 & 6099&f814w&@xmath179 & 21.38 & @xcite + & & f555w&@xmath180 & 22.32 & + ngc 4478 & 6587&f555w&@xmath151 & 22.12 & double - sampled + ngc 4486b & 6099&f814w&@xmath179 & & @xcite + & & f555w&@xmath180 & & + ngc 4494 & 5454&f814w&@xmath158 & 21.70 & @xcite + & & f555w&@xmath159 & 22.60 & + ngc 4552 & 6099&f814w&@xmath168 & 21.17 & @xcite + & & f555w&@xmath181 & 22.15 & + ngc 4589 & 5454&f814w&@xmath158 & 22.06 & @xcite + & & f555w&@xmath159 & 22.96 & + ngc 4621 & 5512&f814w&@xmath166 & 21.19 & + & & f555w&@xmath166 & 22.09 & + ngc 4649 & 6286&f814w&@xmath182 & 21.40 & @xcite + & & f555w&@xmath183 & 22.30 & + ngc 4660 & 5512&f814w&@xmath184 & 20.89 & + & & f555w&@xmath185 & 21.85 & + ngc 4709 & 6587&f555w&@xmath169 & 21.98 & + ngc 4786 & 6587&f555w&@xmath171 & & double - sampled ; obscured + ngc 4936 & 6587&f555w&@xmath148 & & obscured + ngc 5018 & 6587&f555w&@xmath171 & 21.94 & double - sampled ; obscured + ngc 5061 & 6587&f555w&@xmath165 & 22.48 & double - sampled + ngc 5322 & 5454&f814w&@xmath158 & & obscured ; @xcite + & & f555w&@xmath159 & & + ngc 5419 & 6587&f555w&@xmath148 & 22.23 & + ngc 5557 & 6587&f555w&@xmath164 & 22.78 & double - sampled + ngc 5576 & 9107&f814w&@xmath151 & 21.60 & double - sampled + & & f555w&@xmath151 & 22.52 & double - sampled + ngc 5813 & 5454&f814w&@xmath158 & 21.51 & @xcite + & & f555w&@xmath159 & 22.41 & + ngc 5845 & 6099&f814w&@xmath186 & & obscured + & & f555w&@xmath167 & & + ngc 5982 & 5454&f814w&@xmath158 & 22.15 & @xcite + & & f555w&@xmath159 & 23.15 & + ngc 6776 & 6587&f555w&@xmath169 & & obscured + ngc 6849 & 6587&f555w&@xmath169 & 22.56 & + ngc 6876 & 6587&f814w&@xmath187 & & @xcite + & & f555w&@xmath187 & & + ngc 7213 & 9107&f814w&@xmath188 & 21.83 & double - sampled + & & f555w&@xmath188 & 22.84 & double - sampled + ngc 7457 & 5512&f814w&@xmath166 & 21.93 & @xcite + & & f555w&@xmath166 & 22.87 & + ngc 7619 & 6554&f814w&@xmath189 & 21.59 & + & & f555w&@xmath189 & 22.49 & + ngc 7626 & 5454&f814w&@xmath158 & & obscured ; @xcite + & & f555w&@xmath159 & & + ngc 7727 & 7468&f814w&@xmath176 & 20.61 & + & & f555w&@xmath190 & 21.50 & + ngc 7785 & 6587&f555w&@xmath165 & 22.43 & double - sampled + ic 1459 & 5454&f814w&@xmath158 & 21.59 & @xcite + & & f555w&@xmath159 & 22.49 & + ic 3370 & 6587&f555w&@xmath164 & & double - sampled ; obscured + ic 4296 & 6587&f555w&@xmath148 & & obscured + ic 4329 & 6587&f555w&@xmath148 & 22.56 & + eso 0462 - 015 & 6587&f555w&@xmath148 & 22.45 & + [ tab : obs ] llrccr ngc 0507 & s0 & 63.8&4&@xmath19123.38&329 + ngc 0584 & e & 22.1&1&@xmath19121.39&225 + ngc 0596 & e & 22.1&1&@xmath19120.99&164 + ngc 0741 & e & 70.9&4&@xmath19123.20&293 + ngc 0821 & e & 25.5&1&@xmath19121.39&209 + ngc 1016 & e & 75.9&3&@xmath19123.03&279 + ngc 1023 & s0 & 12.1&1&@xmath19120.51&212 + ngc 1316 & e & 22.7&1&@xmath19123.34&250 + ngc 1374 & e & 20.9&1&@xmath19120.75&207 + ngc 1399 & e & 21.1&1&@xmath19122.07&359 + ngc 1426 & e & 26.5&1&@xmath19120.80&153 + ngc 1427 & e & 21.0&1&@xmath19120.78&170 + ngc
1439 & e & 26.5&1&@xmath19120.98&159 + ngc 1700 & e & 40.6&2&@xmath19121.94&234 + ngc 2300 & s0 & 30.4&4&@xmath19121.72&252 + ngc 2434 & e & 22.8&1&@xmath19122.01&229 + ngc 2768 & e & 23.3&1&@xmath19121.90&194 + ngc 2778 & e & 24.2&1&@xmath19119.52&166 + ngc 2974 & e & 22.7&1&@xmath19121.33&143 + ngc 3115 & s0- & 10.2&1&@xmath19121.17&271 + ngc 3377 & e & 11.7&1&@xmath19120.04&143 + ngc 3379 & e & 11.7&1&@xmath19120.90&221 + ngc 3384 & s0- & 11.7&1&@xmath19119.88&158 + ngc 3557 & e & 48.3&1&@xmath19123.23&265 + ngc 3585 & e & 21.2&1&@xmath19122.01&216 + ngc 3607 & e & 10.9&1&@xmath19120.64&220 + ngc 3608 & e & 23.0&1&@xmath19121.11&195 + ngc 3610 & e & 22.6&1&@xmath19120.88&167 + ngc 3640 & e & 28.3&1&@xmath19121.97&184 + ngc 3706 & s0- & 46.9&4&@xmath19122.49&302 + ngc 3842 & e ( bgc ) & 97.0&1&@xmath19123.22&316 + ngc 3945 & s0 & 19.9&3&@xmath19121.14&176 + ngc 4026 & s0 & 15.6&1&@xmath19120.32&193 + ngc 4073 & e & 92.2&4&@xmath19123.62&266 + ngc 4125 & e & 25.2&1&@xmath19122.34&233 + ngc 4278 & e & 16.7&1&@xmath19121.05&251 + ngc 4291 & e & 25.0&1&@xmath19120.56&284 + ngc 4365 & e & 21.6&1&@xmath19122.02&269 + ngc 4382 & s0 & 17.9&1&@xmath19122.28&184 + ngc 4406 & e & 17.9&1&@xmath19122.28&246 + ngc 4458 & e & 17.9&1&@xmath19119.32&101 + ngc 4472 & e & 17.9&1&@xmath19122.92&303 + ngc 4473 & e & 17.9&1&@xmath19120.99&179 + ngc 4478 & e & 17.9&1&@xmath19119.98&144 + ngc 4486b & ce & 17.9&1&@xmath19117.91&178 + ngc 4494 & e & 17.9&1&@xmath19121.48&155 + ngc 4552 & e & 17.9&1&@xmath19121.39&263 + ngc 4589 & e & 25.0&1&@xmath19121.33&225 + ngc 4621 & e & 17.9&1&@xmath19121.61&237 + ngc 4649 & e & 17.9&1&@xmath19122.49&343 + ngc 4660 & e & 17.9&1&@xmath19120.09&191 + ngc 4709 & e & 37.0&1&@xmath19122.31&239 + ngc 4786 & e+p & 70.7&4&@xmath19122.58&295 + ngc 4936 & e & 50.1&4&@xmath19123.26&278 + ngc 5018 & e & 42.6&3&@xmath19122.61&212 + ngc 5061 & e & 26.9&4&@xmath19121.94&194 + ngc 5322 & e & 33.0&1&@xmath19122.40&234 + ngc 5419 & e & 62.6&4&@xmath19123.43&315 + ngc 5557 & e & 52.7&4&@xmath19122.61&253 + ngc 5576 & e & 27.1&1&@xmath19121.67&188 + ngc 5813 & e & 28.7&1&@xmath19121.84&230 + ngc 5845 & e & 28.7&1&@xmath19119.84&241 + ngc 5982 & e & 40.5&4&@xmath19122.01&250 + ngc 6776 & e+p & 77.6&4&@xmath19122.80&209 + ngc 6849 & s0- & 84.0&4&@xmath19123.32&207 + ngc 6876 & e & 57.6&4&@xmath19123.74&230 + ngc 7213 & sa & 22.7&3&@xmath19121.51&185 + ngc
7457 & s0- & 14.0&1&@xmath19118.63 & 78 + ngc 7619 & e & 56.1&1&@xmath19122.93&319 + ngc 7626 & e & 56.0&1&@xmath19122.49&269 + ngc 7727 & sa & 21.6&3&@xmath19121.33&200 + ngc 7785 & e & 49.9&4&@xmath19122.14&253 + ic 1459 & e & 30.9&1&@xmath19122.56&311 + ic 3370 & e & 28.3&1&@xmath19121.55&202 + ic 4296 & e ( bcg ) & 45.4&2&@xmath19123.32&340 + ic 4329 & e ( bcg ) & 56.7&2&@xmath19123.70&270 + eso 462 - 15 & e & 80.3&4&@xmath19122.88&289 + [ tab : glob ] lcccccclll ngc 0507 & 0.0 & & & & & offset & a & 0 & 1,3,8 + ngc 0584 & 0.5 & c & 2 & & & & e : & 1 : & 3,5,6,8,12 + ngc 0596 & 0.0 & & & & @xmath192 & & a & 0 & 4,5,8,9,12 + ngc 0741 & 0.0 & nr & 2 & & @xmath192 & & e & 1 & 5,13 + ngc 0821 & 0.0 & & & @xmath192 & & & a & 0 & 1,3,4,5,8 + ngc 1016 & 0.1 & & & & & & & & + ngc 1023 & 0.0 & & & @xmath192 & @xmath192 & & a & 0 & 1 + ngc 1316 & 4.1 & c & 1 & & @xmath192 & & e & 2 & 2,12 + ngc 1374 & 0.3 & & & & & offset & a & 0 & 2 + ngc 1399 & 0.0 & & & & @xmath192 & & a & 0.5 & 2,4,8,12 + ngc 1426 & 0.0 & & & & @xmath192 & & e & 1 & 9 + ngc 1427 & 0.0 & & & @xmath192 & @xmath192 & & a & 0 & 2 + ngc 1439 & 2.4 & nr & 2 & & @xmath192 & & a & 0 & 14 + ngc 1700 & 0.2 & c & 2 & & & & e & 1 & 3,4,9 + ngc 2300 & 0.0 & & & & & & a & 0 & 1,3,4 + ngc 2434 & 3.0 & nr & 2 & & @xmath192 & & a : & 0 : & 2,12 + ngc 2768 & & s & 1 & & & & l2 & 2 & 1,7,9 + ngc 2778 & 0.3 & nr : & 2 & & @xmath192 & & e : & 2 : & 3 + ngc 2974 & 2.5 & s & 1 & & & & e & 2 & 4,6,7,8,11,12 + ngc 3115 & 0.1 & & & @xmath192 & @xmath192 & & a & 0 & 1,5 + ngc 3377 & 0.0 & c & 1 & @xmath192 & & & a & 0.5 & 1,3,4,9 + ngc 3379 & 1.9 & nr & 2 & & & & l2/t2 : : & 1 & 1,3,4,5 + ngc 3384 & 0.0 & & & @xmath192 & @xmath192 & & a & 0 & 1 + ngc 3557 & & nr & 2 & & & & e & 2 & 2,4,8,9 + ngc 3585 & 1.6 & nr & 2 & @xmath192 & & & a & 0 & 6,8 + ngc 3607 & 8.7 & t & 1 & & & & l 2 & 2 & 1,5,6 + ngc 3608 & 0.2 & nr : & 2 & & & & l2/s2 : & 1 & 1,3,4 + ngc 3610 & 0.0 & & & @xmath192 & & & a & 0 & 1,4,9 + ngc 3640 & 0.2 & & & & & & a & 0 & 1,4 + ngc 3706 & 0.0 & & & @xmath192 & & bar & a & 0 & 4,10 + ngc 3842 & 0.0 & & & & & & & & + ngc 3945 & 1.2 & c & 1 & & @xmath192 & & l2 & 2 & 1 + ngc 4026 & 0.0 & c & 1 & @xmath192 & @xmath192 & & a & 0 & 1 + ngc 4073 & 0.0 & & & & & min & a & 0 & 14 + ngc 4125 & & c & 1 & & & & t2 & 2 & 1,4,7 + ngc 4278 & 3.8 & s & 1 & & @xmath192 & & l2 & 3 & 1,3,4 + ngc 4291 & 0.0 & & & & & offset : & a & 0 & 1 + ngc 4365 & 0.0 & & & & @xmath192 & & a & 0 & 1,4,6,8 + ngc 4382 & 0.0 & & & & & min & a & 0 & 1,6,8 + ngc 4406 & 0.0 & & & & @xmath192 & min & a & 0 & 1,6,8 + ngc 4458 & 0.2 & & & @xmath192 & & & & & + ngc 4472 & 0.7 & c & 2 & & @xmath192 & & s2 : : & 1 & 1,3,5,6 + ngc 4473 & 0.0 & & & @xmath192 & & & a & 0 & 1,4,5,6 + ngc 4478 & 0.0 & & & @xmath192 & @xmath192 & & a & 0 & 1,3 + ngc 4486b & 0.0 & & & & & min & & & + ngc 4494 & 5.3 & nr & 2 & & & & l2 : : & 1 & 1 + ngc 4552 & 0.5 & c & 1 & & @xmath192 & & t2 : & 1.5 & 1,3,5,6,9 + ngc 4589 & 6.7 & c & 1 & & & & l2 & 2 & 1,4,7 + ngc 4621 & 0.0 & & & @xmath192 & & & a & 0 & 1,4,6,8 + ngc 4649 & 0.0 & & & & & & a & 0 & 1,3,6,7,8 + ngc 4660 & 0.0 & & & @xmath192 & & & a & 0 & 1,4 + ngc 4709 & 0.0 & & & & @xmath192 & & & & + ngc 4786 & & nr & 2 & & & & e & 1 & 9 + ngc 4936 & & c & 1 & & & & e & 2 & 5,10 + ngc 5018 & & t & 1 & & & & e : & 1 : & 4,7,8,9,10 + ngc 5061 & 0.0 & & & & & & a & 0 & 4,8 + ngc 5322 & & nr & 2 & & & & l2 : : & 1 & 1,4,7,9 + ngc 5419 & 0.1 & & & & @xmath192 double ? & & e : & 1 : & 2,5,12 + ngc 5557 & 0.1 & & & & & & a & 0 & 1,9 + ngc 5576 & 0.0 & & & & & offset : & a & 0 & 1,4,9 + ngc 5813 & 0.5 & t & 1 & & & & l2 : & 1 & 1,3,4,5,8 + ngc 5845 & & nr & 2 & & & & & & + ngc 5982 & 0.1 & & & & & & l2 : : & 1 & 1,9 + ngc 6776 &
& t & & & & & e & 2 & 2,5,9,10 + ngc 6849 & 0.0 & & & & & & a & 0 & 2,10 + ngc 6876 & 0.0 & c & 0 & & & min & e & 1 & 2,5,6 + ngc 7213 & 4.0 & s & 1 & & @xmath192 & & s1 & 2 & 2,8 + ngc 7457 & 0.0 & & & & @xmath192 & & c+a & 0 & 1 + ngc 7619 & 0.0 & & & & & offset & a & 0 & 1,3,5,6 + ngc 7626 & & nr & 2 & & & & l2 : : & 1 & 1,3,5,6,13 + ngc 7727 & 6.2 & s & 1 & & & & e & 2 & 8 + ngc 7785 & 2.7 & c & 2 & & & & e & 1 & 3,9 + ic 1459 & 3.7 & t & 1 & & @xmath192 & & e & 3 & 4,5,6,9,11 + ic 3370 & & s & 1 & & & & e & 2 & 4,8 + ic 4296 & & nr & 2 & & & & e & 1 & 2,5,6,9 + ic 4329 & 0.2 & & & & & & a & 0 & 10 + eso 0462 - 015 & 0.1 & & & & @xmath192 & & & & + [ tab : morph ] lcrrcrcrrl ngc 0507 & @xmath193&343.3&1.11&17.16 & 1.26&1.75&@xmath194&@xmath195 & + ngc 0584 & @xmath193 & 47.1&0.44&15.22 & 0.47&1.61&@xmath196&@xmath197 & + ngc 0596 & @xmath198 & 4.3&0.54&15.90 & 0.45&1.59&@xmath199&@xmath200 & + ngc 0741 & @xmath193&391.9&1.14&17.67 & 2.27&1.29&@xmath201&@xmath202 & + ngc 0821 & @xmath203 & 59.3&0.48&15.69 & 0.43&1.71&@xmath201&@xmath204&@xmath198 in rav . + ngc 1016 & @xmath193&618.2&1.68&18.01 & 0.98&1.90&@xmath205&@xmath202 & + ngc 1023 & @xmath198 & 2.3&1.23&15.71 & 7.23&1.15&@xmath206&@xmath206 & + ngc 1316 & @xmath193 & 41.8&0.38&14.41 & 1.01&1.23&@xmath207&@xmath208&@xmath209 + ngc 1374 & @xmath193 & 9.1&0.09&14.57 & 1.87&1.08&@xmath210&@xmath211&@xmath212 + ngc 1399 & @xmath193&324.3&3.17&17.23 & 1.58&1.63&@xmath205&@xmath205&@xmath212 + ngc 1426 & @xmath198 & 5.1&1.25&16.98 & 0.51&1.80&@xmath213&@xmath214 & + ngc 1427 & @xmath198 & 4.1&0.86&16.25 & 0.79&1.67&@xmath197&@xmath215 & + ngc 1439 & @xmath198 & 5.1&0.65&16.21 & 4.82&1.48&@xmath206&@xmath206 & + ngc 1700 & @xmath193 & 11.8&0.06&13.58 & 1.03&1.19&@xmath207&@xmath216&@xmath198 in l95 + ngc 2300 & @xmath193&330.1&2.24&17.53 & 1.20&1.80&@xmath216&@xmath217 & + ngc 2434 & @xmath198 & 4.4&5.41&19.10 & 1.95&2.05&@xmath218&@xmath218 & + ngc 2778 & @xmath198 & 4.7&0.35&16.06 & 0.41&1.75&@xmath219&@xmath220 & + ngc 2974 & @xmath198 & 4.4&0.44&15.41&15.72&1.10&@xmath221&@xmath221 & + ngc 3115 & @xmath198 & 2.0&0.86&14.48 & 4.30&1.09&@xmath222&@xmath222 & + ngc 3377 & @xmath198 & 2.3&0.31&14.12 & 0.30&1.96&@xmath223&@xmath221 & + ngc 3379 & @xmath193&112.3&1.98&16.15 & 1.54&1.54&@xmath224&@xmath224 & + ngc 3384 & @xmath198 & 2.3&3.15&16.44&15.32&1.81&@xmath225&@xmath225 & + ngc 3585 & @xmath203 & 37.0&0.36&14.72 & 1.62&1.06&@xmath226&@xmath226 & + ngc 3607 & @xmath193&128.4&2.43&16.87 & 2.06&1.70&@xmath227&@xmath227 & + ngc 3608 & @xmath193 & 53.5&0.48&15.73 & 0.92&1.50&@xmath205&@xmath228 & + ngc 3610 & @xmath198 & 4.4&2.84&16.39&48.48&1.86&@xmath229&@xmath229 & + ngc 3640 & @xmath193 & 46.6&0.34&15.66 & 0.64&1.30&@xmath207&@xmath223&@xmath203 in rest + ngc 3706 & @xmath193 & 40.9&0.18&14.18&14.66&1.24&@xmath196&@xmath196 & + ngc 3842 & @xmath193&507.9&1.08&17.76 & 1.84&1.42&@xmath230&@xmath230 & + ngc 3945 & @xmath198 & 3.9&7.38&18.62 & 0.30&2.56&@xmath231&@xmath232 & + ngc 4026 & @xmath198 & 3.0&0.63&15.23 & 0.39&1.78&@xmath230&@xmath233 & + ngc 4073 & @xmath193&129.6&0.29&16.53 & 4.43&0.99&@xmath234&@xmath234 & + ngc 4278 & @xmath193&102.0&1.26&16.20 & 1.40&1.46&@xmath235&@xmath201 & + ngc 4291 & @xmath193 & 72.7&0.60&15.66 & 1.52&1.60&@xmath195&@xmath236 & + ngc 4365 & @xmath193&231.4&2.21&16.88 & 1.74&1.52&@xmath216&@xmath205 & + ngc 4382 & @xmath193 & 80.7&0.93&15.67 & 1.13&1.39&@xmath194&@xmath195 & + ngc 4406 & @xmath193 & 79.8&0.92&16.03 & 3.87&1.04&@xmath237&@xmath237 & + ngc 4458 & @xmath193 & 7.8&0.09&13.72 & 4.55&1.40&@xmath199&@xmath228&@xmath198 in l95 + ngc 4472 & @xmath193&211.7&2.44&16.63 & 1.89&1.17&@xmath195&@xmath195 & + ngc 4473 & @xmath193&386.2&4.45&17.24 & 0.66&2.60&@xmath238&@xmath195 & + ngc 4478 & @xmath193 & 19.1&0.22&15.46 & 1.88&1.01&@xmath207&@xmath201&@xmath198 in l95 + ngc 4486b & @xmath193 & 13.9&0.16&14.53 & 2.82&1.39&@xmath207&@xmath207 & + ngc 4494 & @xmath198 & 3.5&2.82&17.19 & 0.68&1.88&@xmath222&@xmath239 & + ngc 4552 & @xmath193 & 42.5&0.49&15.20 & 1.97&1.17&@xmath207&@xmath240 & + ngc 4589 & @xmath193 & 67.9&0.56&16.08 & 1.05&1.32&@xmath241&@xmath213 & + ngc 4621 & @xmath198 & 3.5&1.00&15.76 & 0.41&1.34&@xmath218&@xmath242 & + ngc 4649 & @xmath193&437.4&5.04&17.28 & 1.67&1.58&@xmath199&@xmath199 & + ngc 4660 & @xmath198 & 3.5&1.76&16.34 & 5.61&1.50&@xmath243&@xmath243 & + ngc 4709 & @xmath193&222.4&1.24&17.46 & 1.89&1.47&@xmath244&@xmath245 & + ngc 5061 & @xmath193 & 37.8&0.29&14.35 & 1.64&1.43&@xmath246&@xmath247 & + ngc 5419 & @xmath193&722.3&2.38&17.87 & 1.43&1.69&@xmath207&@xmath223 & + ngc 5557 & @xmath193&204.4&0.80&16.40 & 0.80&1.68&@xmath236&@xmath216&@xmath203 in rest + ngc 5576 & @xmath193&549.2&4.18&17.81 & 0.43&2.73&@xmath195&@xmath227&@xmath198 in rest + ngc 5813 & @xmath193&162.0&1.17&16.86 & 1.06&1.81&@xmath207&@xmath234 & + ngc 5982 & @xmath193 & 68.7&0.35&15.66 & 2.07&1.03&@xmath247&@xmath247 & + ngc 6849 & @xmath193&138.5&0.34&17.03 & 1.01&1.21&@xmath195&@xmath217 & + ngc 6876 & @xmath193&139.6&0.50&16.99&12.07&0.75&@xmath194&@xmath194 & + ngc 7213 & @xmath193&383.0&3.48&17.71 & 1.03&3.10&@xmath235&@xmath241 & + ngc 7457 & @xmath198 & 2.7&0.22&16.33 & 1.01&1.05&@xmath207&@xmath248&@xmath203 in rav .
+ ngc 7619 & @xmath193&220.3&0.81&16.40 & 1.08&1.62&@xmath240&@xmath195 & + ngc 7727 & @xmath203&601.1&5.74&18.42 & 0.57&1.93&@xmath249&@xmath250 & + ngc 7785 & @xmath193 & 14.5&0.06&15.09 & 0.95&0.92&@xmath207&@xmath247 & + ic 1459 & @xmath193&293.6&1.96&16.36 & 0.80&2.08&@xmath207&@xmath230 & + ic 4329 & @xmath193&261.2&0.79&16.37 & 2.87&1.31&@xmath195&@xmath195 & + eso 462 - 015 & @xmath198 & 15.6&0.41&16.59 & 0.20&1.77&@xmath207&@xmath214 & + [ tab : cen ] lccc ngc 507&@xmath193&0.229&0.190 + ngc 584&@xmath193&0.346&0.255 + ngc 596&@xmath198&0.123&0.072 + ngc 741&@xmath193&0.043&0.147 + ngc 821&@xmath203&0.370&0.382 + ngc 1016&@xmath193&0.099&0.071 + ngc 1023&@xmath198&0.351&0.201 + ngc 1316&@xmath193&0.242&0.367 + ngc 1374&@xmath193&0.036&0.110 + ngc 1399&@xmath193&0.078&0.125 + ngc 1426&@xmath198&0.320&0.315 + ngc 1427&@xmath198&0.287&0.293 + ngc 1439&@xmath198&0.274&0.148 + ngc 1700&@xmath193&0.318&0.294 + ngc 2300&@xmath193&0.196&0.232 + ngc 2434&@xmath198&0.083&0.067 + ngc 2778&@xmath198&0.111&0.204 + ngc 2974&@xmath198&0.296&0.301 + ngc 3115&@xmath198&0.537&0.548 + ngc 3377&@xmath198&0.360&0.509 + ngc 3379&@xmath193&0.102&0.097 + ngc 3384&@xmath198&0.415&0.180 + ngc 3585&@xmath203&0.283&0.456 + ngc 3607&@xmath193&0.129&0.159 + ngc 3608&@xmath193&0.108&0.184 + ngc 3610&@xmath198&0.437&0.516 + ngc 3640&@xmath193&0.110&0.212 + ngc 3842&@xmath193&0.044&0.174 + ngc 3945&@xmath198&0.191&0.299 + ngc 4026&@xmath198&0.469&0.385 + ngc
4278&@xmath193&0.155&0.167 + ngc 4291&@xmath193&0.260&0.247 + ngc 4365&@xmath193&0.286&0.248 + ngc 4382&@xmath193&0.262&0.240 + ngc 4458&@xmath193&0.573&0.187 + ngc 4472&@xmath193&0.088&0.110 + ngc 4473&@xmath193&0.469&0.412 + ngc 4478&@xmath193&0.311&0.220 + ngc 4486&@xmath193&0.431&0.124 + ngc 4494&@xmath198&0.225&0.153 + ngc 4552&@xmath193&0.138&0.050 + ngc 4589&@xmath193&0.399&0.197 + ngc 4621&@xmath198&0.299&0.351 + ngc 4649&@xmath193&0.072&0.140 + ngc 4660&@xmath198&0.396&0.404 + ngc 4709&@xmath193&0.233&0.184 + ngc 5061&@xmath193&0.060&0.038 + ngc 5419&@xmath193&0.180&0.208 + ngc 5557&@xmath193&0.089&0.209 + ngc 5576&@xmath193&0.242&0.302 + ngc 5813&@xmath193&0.124&0.107 + ngc 5982&@xmath193&0.031&0.273 + ngc 6849&@xmath193&0.056&0.353 + ngc 7213&@xmath193&0.108&0.110 + ngc 7457&@xmath198&0.064&0.371 + ngc 7619&@xmath193&0.392&0.266 + ngc 7727&@xmath203&0.262&0.243 + ngc 7785&@xmath193&0.266&0.367 + ic 1459&@xmath193&0.260&0.261 + ic 4329&@xmath193&0.168&0.152 + eso 462 - 15&@xmath198&0.170&0.271 + [ tab : ellip ] lcccrrl ngc 0596&@xmath198 & 21.8 & & & @xmath19110.0&@xmath251 + ngc 0741&@xmath193 & 21.9 & & & @xmath19112.5&@xmath252 + ngc 1023&@xmath198 & 19.5 & 0.9 & @xmath1910.4 & @xmath19111.1&@xmath253 + ngc 1316&@xmath193 & 19.9 & & & @xmath19112.0&@xmath254 + ngc 1399&@xmath193 & 24.4 & & & @xmath1917.2&@xmath255 + ngc 1426&@xmath198 & 19.6 & & & @xmath19112.5&@xmath252 + ngc 1427&@xmath198 & 19.6 & 1.3 & 0.0 & @xmath19112.0&@xmath256 + ngc 1439&@xmath198 & 20.2 & 1.1 & @xmath1910.2 & @xmath19111.9&@xmath254 + ngc 2434&@xmath198 & 20.3 & 0.7 & @xmath1910.5 & @xmath19112.3&@xmath257 + ngc 2778&@xmath198 & 20.0 & 1.3 & @xmath1910.1 & @xmath19111.9&@xmath254 + ngc 3115&@xmath198 & 17.8 & 1.2 & @xmath1910.1 & @xmath19112.3&@xmath258 + ngc 3384&@xmath198 & 18.9 & 1.3 & 0.0 & @xmath19111.5&@xmath259 + ngc 3945&@xmath198 & 20.9 & 1.5 & 0.2 & @xmath19110.7&@xmath260 + ngc 4026&@xmath198 & 19.5 & 1.3 & 0.0 & @xmath19111.6&@xmath259 + ngc 4278&@xmath193 & 19.5 & 0.4 & @xmath1910.8 & @xmath19111.7&@xmath259 + ngc 4365&@xmath193 & 21.0 & 0.6 & @xmath1910.7 & @xmath19110.6&@xmath260 + ngc 4406&@xmath193 & 22.9 & 1.2 & 0.0 & @xmath1918.4&@xmath261 + ngc 4472&@xmath193 & 22.7 & 1.0 & @xmath1910.4 & @xmath1918.6&@xmath261 + ngc 4478&@xmath193 & 19.7 & & & @xmath19111.6&@xmath259 + ngc 4552&@xmath193 & 20.5 & 1.3 & 0.0 & @xmath19110.9&@xmath260 + ngc 4709&@xmath193 & 23.2 & & & @xmath19110.0&@xmath262 + ngc 5419&@xmath193 & 19.8 & & & @xmath19114.4&@xmath263 + ngc 7213&@xmath193 & 16.6 & 0.6 & @xmath1910.8 & @xmath19115.2&@xmath264 + ngc 7457&@xmath198 & 18.1 & 1.2 & 0.1 & @xmath19112.7&@xmath265 + ic 1459&@xmath193 & 18.4 & 1.0 & @xmath1910.4 & @xmath19114.0&@xmath266 + [ tab : nuclei ] lcc ngc 0584&@xmath267&@xmath268 + ngc 0821&@xmath269&@xmath270 + ngc 1023&@xmath271&@xmath272 + ngc 1427&@xmath273&@xmath274 + ngc 1439&@xmath275&@xmath276 + ngc 1700&@xmath277&@xmath278 + ngc 2300&@xmath279&@xmath280 + ngc 2434&@xmath281&@xmath282 + ngc 2778&@xmath283&@xmath284 + ngc 3115&@xmath285&@xmath286 + ngc 3377&@xmath287&@xmath288 + ngc 3379&@xmath289&@xmath290 + ngc 3384&@xmath291&@xmath292 + ngc 3607&@xmath293&@xmath294 + ngc 3608&@xmath277&@xmath295 + ngc 3945&@xmath285&@xmath296 + ngc 4026&@xmath297&@xmath298 + ngc 4278&@xmath299&@xmath288 + ngc 4291&@xmath300&@xmath301 + ngc 4365&@xmath302&@xmath303 + ngc 4382&@xmath304&@xmath305 + ngc 4406&@xmath306&@xmath288 + ngc 4458&@xmath307&@xmath308 + ngc 4472&@xmath309&@xmath310 + ngc 4473&@xmath311&@xmath312 + ngc 4486b&@xmath313&@xmath314 + ngc 4494&@xmath315&@xmath316 + ngc 4552&@xmath317&@xmath318 + ngc 4589&@xmath319&@xmath320 + ngc 4621&@xmath321&@xmath322 + ngc 4649&@xmath323&@xmath324 + ngc 4660&@xmath325&@xmath278 + ngc 5576&@xmath326&@xmath327 + ngc 5813&@xmath328&@xmath329 + ngc 5982&@xmath299&@xmath330 + ngc 7213&@xmath331&@xmath332 + ngc 7457&@xmath333&@xmath334 + ngc 7619&@xmath335&@xmath336 + ngc 7727&@xmath337&@xmath338 + ic 1459&@xmath339&@xmath340 + [ tab : vmi ] lccrrcc ngc 0507&0.50&0.47&@xmath341&@xmath342&@xmath343&@xmath344 + ngc 0584&0.67&0.87&@xmath345&@xmath346&@xmath347&@xmath348 + ngc 0596&3.82&1.17&@xmath349&@xmath350&@xmath344&@xmath351 + ngc 0741&1.08&0.67&@xmath352&@xmath353&@xmath354&@xmath355 + ngc 0821&0.78&0.99&@xmath356&@xmath357&@xmath358&@xmath359 + ngc 1016&0.46&0.48&@xmath360&@xmath361&@xmath362&@xmath363 + ngc 1023&1.44&1.05&@xmath364&@xmath365&@xmath366&@xmath367 + ngc 1316&1.08&0.89&@xmath368&@xmath369&@xmath370&@xmath371 + ngc 1374&0.32&0.98&@xmath372&@xmath357&@xmath373&@xmath374 + ngc 1399&0.48&0.16&@xmath375&@xmath376&@xmath377&@xmath367 + ngc 1426&0.30&0.75&@xmath378&@xmath379&@xmath371&@xmath380 + ngc 1427&0.55&0.86&@xmath381&@xmath382&@xmath383&@xmath384 + ngc 1439&0.78&1.27&@xmath385&@xmath386&@xmath387&@xmath388 + ngc 1700&0.30&0.98&@xmath389&@xmath390&@xmath391&@xmath344 + ngc 2300&0.57&0.34&@xmath392&@xmath393&@xmath394&@xmath395 + ngc 2434&0.50&0.76&@xmath396&@xmath397&@xmath398&@xmath399 + ngc 2778&4.68&1.39&@xmath389&@xmath400 & & + ngc 2974&0.39&0.68&@xmath386&@xmath401&@xmath370&@xmath402 + ngc 3115&0.78&0.75&@xmath364&@xmath403&@xmath404&@xmath348 + ngc 3377&1.76&1.24&@xmath405&@xmath403&@xmath406&@xmath407 + ngc 3379&0.67&0.39&@xmath408&@xmath375&@xmath409&@xmath410 + ngc 3384&4.68&1.81&@xmath411&@xmath412 & & + ngc 3585&0.43&0.74&@xmath413&@xmath349&@xmath414&@xmath415 + ngc 3607&1.08&0.49&@xmath416&@xmath417&@xmath398&@xmath418 + ngc 3608&0.57&0.85&@xmath419&@xmath364&@xmath420&@xmath421 + ngc 3610&1.04&0.76&@xmath422&@xmath423&@xmath371&@xmath344 + ngc 3640&0.34&0.60&@xmath424&@xmath425&@xmath426&@xmath427 + ngc 3842&0.78&0.60&@xmath428&@xmath429&@xmath430&@xmath384 + ngc 3945&2.87&1.07&@xmath431&@xmath432&@xmath370&@xmath433 + ngc 4026&0.30&0.84&@xmath401&@xmath434&@xmath435&@xmath436 + ngc 4278&0.46&0.33&@xmath437&@xmath438&@xmath344&@xmath347 + ngc 4291&3.38&1.50&@xmath439&@xmath440&@xmath441&@xmath367 + ngc 4365&1.76&0.65&@xmath442&@xmath411&@xmath443&@xmath359 + ngc 4382&1.50&0.88&@xmath444&@xmath403&@xmath445&@xmath399 + ngc 4458&1.50&1.40&@xmath403&@xmath446&@xmath410&@xmath447 + ngc 4472&0.67&0.10&@xmath448&@xmath449&@xmath450&@xmath451 + ngc 4473&0.46&0.42&@xmath452&@xmath453&@xmath384&@xmath374 + ngc 4478&0.30&0.60&@xmath454&@xmath455&@xmath456&@xmath457 + ngc 4486b&4.68&1.39&@xmath458&@xmath459 & & + ngc 4494&0.41&0.81&@xmath460&@xmath461&@xmath462&@xmath463 + ngc 4552&0.67&0.72&@xmath464&@xmath465&@xmath466&@xmath363 + ngc 4589&2.08&1.10&@xmath467&@xmath468&@xmath436&@xmath394 + ngc 4621&0.32&0.98&@xmath469&@xmath357&@xmath414&@xmath470 + ngc 4649&2.44&0.49&@xmath471&@xmath472&@xmath473&@xmath474 + ngc 4660&4.68&1.50&@xmath475&@xmath476 & & + ngc 4709&0.36&0.38&@xmath477&@xmath478&@xmath479&@xmath367 + ngc 5061&3.24&1.41&@xmath480&@xmath481&@xmath482&@xmath483 + ngc 5419&0.92&0.27&@xmath484&@xmath485&@xmath457&@xmath466 + ngc 5557&0.48&0.68&@xmath486&@xmath485&@xmath394&@xmath487 + ngc 5576&0.32&0.68&@xmath488&@xmath489&@xmath377&@xmath490 + ngc 5813&0.57&0.51&@xmath491&@xmath492&@xmath493&@xmath466 + ngc 5982&0.27&0.41&@xmath494&@xmath495&@xmath496&@xmath496 + ngc 6849&0.41&0.67&@xmath497&@xmath498&@xmath499&@xmath415 + ngc 7213&0.54&0.45&@xmath500&@xmath501&@xmath502&@xmath503 + ngc 7457&0.27&0.53&@xmath504&@xmath505 & & + ngc 7619&0.67&0.71&@xmath506&@xmath405&@xmath363&@xmath363 + ngc 7727&4.68&1.14&@xmath507&@xmath508 & & + ngc 7785&1.99&0.88&@xmath379&@xmath509&@xmath510&@xmath511 + ic 1459&0.50&0.45&@xmath512&@xmath357&@xmath457&@xmath406 + ic 4329&0.57&0.37&@xmath513&@xmath478&@xmath395&@xmath373 + eso 0462 - 015&0.36&0.83&@xmath514&@xmath515&@xmath370&@xmath516 + [ tab : bend ] | we present observations of 77 early - type galaxies imaged with the pc1 ccd of _ hst_+wfpc2 .
`` nuker law '' parametric fits to the surface brightness profiles are used to classify the central structure into `` core '' or `` power - law '' forms .
core galaxies are typically rounder than power - law galaxies .
nearly all power - laws with central ellipticity @xmath0 have stellar disks , implying that disks are present in most power - laws with @xmath1 but are not visible due to unfavorable geometry .
a few low - luminosity flattened core galaxies also have disks ; these may be transition forms from power - laws to more luminous core galaxies , which lack disks .
several core galaxies have strong isophote twists interior to their break radii , although power - laws have interior twists of similar physical significance when the photometric perturbations implied by the twists are evaluated .
central color gradients are typically consistent with the envelope gradients ; core galaxies have somewhat weaker color gradients than power - laws .
nuclei are found in 29% of the cores and 60% of the power - laws .
nuclei are typically bluer than the surrounding galaxy . while some nuclei are associated with agn , just as many are not ; conversely , not all galaxies known to have low - level agn exhibit detectable nuclei in the broad - band filters .
ngc 4073 and 4382 , are found to have central minima in their intrinsic starlight distributions ; ngc 4382 resembles the double nucleus of m31 . in general , the peak brightness location is coincident with the photocenter of the core to a typical physical scale @xmath2 pc .
five galaxies , however , have centers significantly displaced from their surrounding cores ; these may be unresolved asymmetric double nuclei .
lastly , as noted by previous authors , central dust is visible in about half of the galaxies .
the presence and strength of dust correlates with nuclear emission , thus dust may outline gas that is falling into the central black hole .
the prevalence of dust and its morphology suggest that dust clouds form , settle to the center , and disappear repeatedly on @xmath3 yr timescales .
we discuss the hypothesis that cores are created by the decay of a massive black hole binary formed in a merger .
apart from their brightness profiles , there are no strong differences between cores and power - laws that demand this scenario ; however , the rounder shapes of cores , their lack of disks , and their reduced color gradients may be consistent with it . |
this paper constitutes a survey of a collection of results from @xcite as well as the announcement of new results on the @xmath5-fractal billiard table @xmath6 ( see @xcite ) and a self - similar sierpinski carpet billiard table @xmath7 ( see @xcite ) . in [ sec :
rationalbilliards ] and [ sec : fractalgeometry ] , we survey the necessary background material for understanding the remainder of the article . more specifically , in [ sec : rationalbilliards ] , we introduce the notion of a rational polygonal billiard , a translation surface determined from a rational polygonal billiard and discuss the consequence of a dynamical equivalence between the billiard flow and the geodesic flow .
this dynamical equivalence allows us to express an orbit of a rational billiard table as a geodesic on an associated translation surface , and vice - versa , with the added benefit of being able to determine the reflection in certain types of vertices of a rational billiard table .
furthermore , in [ sec : fractalgeometry ] , we provide additional background material from the subject of fractal geometry necessary for understanding the construction of the koch snowflake @xmath12 , @xmath5-fractal @xmath13,-fractal @xmath13 was previously studied in a different context in @xcite . ] and a sierpinski carpet @xmath15 , as well as particular orbits and _ nontrivial paths_. we then combine the background material presented in [ sec : rationalbilliards ] and [ sec : fractalgeometry ] to analyze the prefractal billiard tables @xmath8 , @xmath9 and @xmath10 , for @xmath11 .
we begin by providing a general language for prefractal billiards and subsequently focus on determining sufficient conditions for what we are calling a _ sequence of compatible periodic orbits_. while [ subsec : theprefractalkochsnowflakebilliard][subsec : aprefractalselfsimilarsierpinskicarpetbilliard ] contain specific results and specialized definitions , there is an over - arching theme that is more fully developed in
[ sec : fractalbilliards ] .
in addition to providing a general language within which to analyze a fractal billiard , we discuss in [ subsec : thekochsnowflakefractalbilliard][subsec : aself - similarsierpinskicarpetbilliard ] how one can determine well - defined orbits of @xmath4 , @xmath6 and @xmath7 , as well as nontrivial paths of @xmath4 and @xmath6 that connect two elusive points of each respective billiard . relying on the main result of @xcite , the second author and joe p. chen have shown that it is possible to determine a periodic orbit of a self - similar sierpinski carpet billiard @xmath7 ; additional results and proofs are forthcoming in @xcite , but a synopsis is provided in [ subsec : aprefractalselfsimilarsierpinskicarpetbilliard ] and [ subsec : aself - similarsierpinskicarpetbilliard ] .
many of the results in
[ sec : prefractalrationalbilliards ] and [ sec : fractalbilliards ] are being announced for the first time .
specifically , [ subsec : thetfractalprefractalbilliard ] and [ subsec : thetfractalbilliard ] contain new results on the prefractal @xmath5-fractal billiard @xmath9 and the @xmath5-fractal billiard @xmath6 ( see @xcite ) ; [ subsec : aprefractalselfsimilarsierpinskicarpetbilliard ] and [ subsec : aself - similarsierpinskicarpetbilliard ] contain new results for a prefractal sierpinski carpet billiard @xmath10 and self - similar sierpinski carpet billiard @xmath7 , where @xmath16 is the single underlying scaling ratio ( see @xcite ) .
as these sections constitute announcements of new results on the respective prefractal and fractal billiards , we will provide in future papers @xcite detailed statements and proofs of the results given therein .
given the nature of the subject of _ fractal billiards _ , we will close with a discussion of open problems and possible directions for future work , some of which are to appear in @xcite and @xcite .
in this section , we will survey the dynamical properties of a billiard ball as it traverses a region in the plane bounded by a closed and connected polygon . in the latter part of this article
, we will remove the stipulation that the boundary be a polygon and focus on billiard tables having boundaries that are fractal or containing subsets that are fractal ( while still being simple , closed and connected curves in the plane ) . under ideal conditions , we know that a point mass making a perfectly elastic collision with a @xmath17 surface ( or curve ) will reflect at an angle which is equal to the angle of incidence , this being referred to as the _ law of reflection_. consider a compact region @xmath1 in the plane with simple , closed and connected boundary @xmath0 .
then , @xmath1 is called a _ planar billiard _ when @xmath0 is smooth enough to allow the law of reflection to hold , off of a set of measure zero ( where the measure is taken to be the arc length measure on @xmath0 ) .
though the law of reflection implicitly states that the angles of incidence and reflection be determined with respect to the normal to the line tangent at the basepoint , we adhere to the equivalent convention in the field of mathematical billiards that the vector describing the position and velocity of the billiard ball ( which amounts to the position and angle , since we are assuming unit speed ) be reflected in the tangent to the point of incidence .. ] that is , employing such a law in order to determine the path on which the billiard ball departs after impact essentially amounts to identifying certain vectors .
such an equivalence relation is denoted by @xmath18 and , in the context of a polygonal billiard , is discussed below in more detail . for the remainder of the article , unless otherwise indicated , when @xmath0 is a simple , closed , connected and piecewise smooth curve so as to allow the law of reflection to hold ( off finitely many points ) , we assume @xmath0 is a closed and connected polygon .
in such a case , we will refer to @xmath1 as a _
polygonal billiard_. one may express the law of reflection in terms of equivalence classes of vectors by identifying two particular vectors that form an equivalence class of vectors in the unit tangent bundle corresponding to the billiard table @xmath1 ; see figure [ fig : billiardmap ] .
( see @xcite for a detailed discussion of this equivalence relation on the unit tangent bundle @xmath19 . )
denote by @xmath20 the unit circle , which we let represent all the possible directions ( or angles ) in which a billiard ball may initially move . to clearly understand how one forms equivalence classes from elements of @xmath19 , we let @xmath21 and say that @xmath22 if and only if @xmath23 and one of the following is true : 1 .
@xmath23 is not a vertex of the boundary @xmath0 and @xmath24 ; 2 . @xmath23 is not a vertex of the boundary @xmath0 , but @xmath23 is a point on a segment @xmath25 of the polygon @xmath0 and @xmath26 , where @xmath27 denotes reflection in the segment @xmath25 ; 3 .
if @xmath23 is a vertex of @xmath0 , then we identify @xmath28 with @xmath29 for every @xmath30 in the group generated by reflections in the two adjacent sides having @xmath31 ( or @xmath32 ) as a common vertex . for now , we shall denote by @xmath33 $ ] the equivalence class of @xmath28 , relative to the equivalence relation @xmath18 .
the collection of vertices of @xmath1 forms a set of zero measure ( when we take our measure to be the arc - length measure on @xmath0 ) , since there are finitely many vertices .
the phase space for the billiard dynamics is given by the quotient space @xmath34 .
in practice , one restricts his or her attention to the space @xmath35 .
the billiard flow on @xmath35 is determined from the continuous flow on @xmath36 as follows .
let @xmath37 be an initial basepoint , @xmath38 be an initial direction and @xmath39 be a flow line corresponding to these initial conditions in the phase space @xmath34 .
the values @xmath40 for which @xmath41 constitute the _ return times _
( i.e. , times at which @xmath39 returns to the section , or intersects it in a non - tangential way ) .
then , the discrete map @xmath42 constitutes the _ section map_. in terms of the configuration space , @xmath42 constitutes the point and angle of incidence in the boundary @xmath0 .
since @xmath1 is the billiard table and we are interested in determining the collision points , it is only fitting that such a map be called the _
billiard map_. more succinctly , we denote @xmath43 by @xmath44 and , in general , such a map is called the _ poincar map _ and the section is called the _
poincar section_. furthermore , the obvious benefit of having a visual representation of @xmath45 in the configuration space is exactly why one restricts his or her attention to the section @xmath35 .
specifically , all one really cares about in the end , from the perspective of studying a planar billiard , are the collision points , which are clearly determined by the billiard map . in order to understand how one determines the next collision point and direction of travel
, we must further discuss the _ billiard map _ @xmath46 . as previously discussed , @xmath47 , where the equivalence relation @xmath18 is the one introduced above .
more precisely , if @xmath38 is an inward pointing vector at a basepoint @xmath37 , then @xmath48 is the representative element of the equivalence class @xmath49 $ ] .
the billiard map then acts on @xmath35 by mapping @xmath50 $ ] to @xmath51 $ ] , where @xmath52 and @xmath53 are collinear in the direction determined by @xmath54 and where @xmath55 is the reflection of angle @xmath54 through the tangent at @xmath53 . in general , we have @xmath56 = [ ( x^k,\theta^k)]$ ] , for every @xmath57 . in the sequel , we will simply refer to an element @xmath58\in ( \omega(d)\times s^1)/\sim$ ] by @xmath59 , since the vector corresponding to @xmath54 is inward pointing at the basepoint @xmath52 . so as not to introduce unnecessary notation ,
when we discuss the billiard map @xmath60 corresponding to the @xmath61th prefractal billiard @xmath62 approximating a fractal billiard @xmath3 , we will simply write @xmath60 as @xmath63 . when discussing the discrete billiard flow on @xmath64 , the @xmath65th point in an orbit @xmath66
will instead be denoted by @xmath67 , in order to keep track of the space such a point belongs to ( namely , with our present convention , @xmath64 ) .
specifically , @xmath68 refers to the number of iterates of the billiard map @xmath63 necessary to produce the pair @xmath67 .
an initial condition of an orbit of @xmath62 will always be referred to as @xmath69 . in
what follows , we are presupposing an orbit can be formed by iterating the billiard map forward in time and backwards in time , whenever @xmath70 is defined .
an orbit making finitely many collisions in the boundary is called a _ closed orbit_. if , in addition , there exists @xmath71 such that @xmath72 , then the resulting orbit is called _ periodic _ ; the smallest positive integer @xmath73 such that @xmath74 is called the _ period _ of the periodic orbit . in the event that a basepoint @xmath75 of @xmath76 is a corner of @xmath1 ( that is , a vertex of the polygonal boundary @xmath0 ) and reflection can not be determined in a well - defined manner ,
then the resulting orbit is said to be _
singular_. in addition , if there exists a positive integer @xmath65 such that the basepoint @xmath77 of @xmath78 is a corner of @xmath1 ( here , @xmath79 denotes the @xmath65th inverse iterate of @xmath46 ) , then the resulting orbit is closed and the path traced out by the billiard ball connecting @xmath75 and @xmath77 is called a _
saddle connection_. finally , we note that a periodic orbit with period @xmath73 is a closed orbit for which reflection is well defined at each basepoint @xmath80 of @xmath81 , @xmath82 and @xmath72 .
we say that an orbit @xmath83 is _ dense _ in a rational billiard table @xmath1 if the path traversed ( forward and backward in time ) by the billiard ball in @xmath1 is dense in @xmath1 .
that is , the closure of the set of points comprising the path traversed by the billiard ball is exactly @xmath1 .
likewise , the points of incidence ( i.e. , the footprint ) of a dense orbit will be dense in the boundary @xmath0 , as explained in remark [ rmk : denseorbits ] .
[ rmk : denseorbits ] consider a rational polygonal billiard @xmath1 .
the associated translation surface @xmath84 can be constructed as described in
[ subsec : translationstructuresandtranslationsurfaces ] . as we will show in [ subsec : unfoldingabilliardorbit ] , the geodesic flow on a translation surface is dynamically equivalent to the billiard flow .
a dense orbit will have an initial direction preventing the path from being parallel to any side of @xmath1 ( except , possibly , for finitely many initial directions , and hence , for a measure - zero set ) .
the corresponding path on the associated translation surface for an explanation of what constitutes a translation surface .
] must also be dense in the surface .
since the path on the surface is arbitrarily close to every side appropriately identified with another side of a copy of @xmath1 and not parallel to any side , the path will be transversal with respect to each side . thus , the collection of basepoints of a dense orbit must be dense in @xmath0 .
[ def : footprintofanorbit ] let @xmath85 be an orbit of a billiard @xmath1 with an initial condition @xmath86 .
then the trace of an orbit on the boundary @xmath0 , @xmath87 is called the _ footprint _ of the orbit @xmath85 and is denoted by @xmath88 .
when we are only interested in a prefractal billiard @xmath89 , we denote the footprint of an orbit by @xmath90 . for the remainder of the article , when discussing polygonal billiards
, we will focus our attention on what are called _ rational polygonal billiards _ , or , more succinctly , _
rational billiards_. if @xmath0 is a nontrivial connected polygon such that for each interior angle @xmath91 of @xmath0 there are relatively prime integers @xmath92 and @xmath93 such that @xmath94 , then we call @xmath0 a _ rational polygon _ and @xmath1 a _ rational billiard_. [ def : ratbilliard ] in this subsection , we will discuss what constitutes a translation surface and how to construct a translation surface from a rational billiard .
then , in [ subsec : unfoldingabilliardorbit ]
, we will see how to relate the continuous billiard flow on @xmath34 with the geodesic flow on the associated translation surface .
[ def : translationstructure ] let @xmath95 be a compact , connected , orientable surface .
a _ translation structure _ on @xmath95 is an atlas @xmath96 , consisting of charts of the form @xmath97 , where @xmath98 is a domain ( i.e. , a connected open set ) in @xmath95 and @xmath99 is a homeomorphism from @xmath98 to a domain in @xmath100 , such that the following conditions hold : 1 .
the collection @xmath101 covers the whole surface @xmath95 except for finitely many points @xmath102 , called _ singular points _ ; 2 .
all coordinate changing functions are translations in @xmath100 ; 3 .
the atlas @xmath96 is maximal with respect to properties @xmath103 and @xmath104 ; 4 . for each singular point @xmath105
, there is a positive integer @xmath106 , a punctured neighborhood @xmath107 of @xmath105 not containing other singular points , and a map @xmath108 from this neighborhood to a punctured neighborhood @xmath109 of a point in @xmath100 that is a shift in the local coordinates from @xmath96 , and is such that each point in @xmath109 has exactly @xmath106 preimages under @xmath108 .
we say that a connected , compact surface equipped with a translation structure is a _ translation surface_. note that in the literature on billiards and dynamical systems , the terminology and definitions pertaining to this topic are not completely uniform ; see , for example , @xcite .
( we note that in @xcite and @xcite , ` translation surfaces ' are referred to as ` flat surfaces ' . )
we have adopted the above definition for clarity and the reader s convenience .
we now discuss how to construct a translation surface from a rational billiard .
consider a rational polygonal billiard @xmath1 with @xmath65 sides and interior angles @xmath110 at each vertex @xmath105 , for @xmath111 , where the positive integers @xmath112 and @xmath113 are relatively prime .
the linear parts of the planar symmetries generated by reflection in the sides of the polygonal billiard @xmath1 generate a dihedral group @xmath114 , where @xmath115 ( the least common multiple of the @xmath113 s ) .
next , we consider @xmath116 ( equipped with the product topology ) .
we want to glue ` sides ' of @xmath117 together and construct a natural atlas on the resulting surface @xmath95 so that @xmath95 becomes a translation surface .
as a result of the identification , the points of @xmath95 that correspond to the vertices of @xmath1 constitute ( removable or nonremovable ) conic singularities of the surface .
heuristically , @xmath116 can be represented as @xmath118 , in which case it is easy to see what sides are made equivalent under the action of @xmath18 .
that is , @xmath18 identifies opposite and parallel sides in a manner which preserves the orientation .
see example [ exa : equilateraltriangleflatsurface ] and figure [ fig : sixequisurface ] for an example of a translation surface constructed from the equilateral triangle billiard @xmath119 .
can be acted on by a particular group of symmetries to produce a translation surface that is topologically equivalent to the flat torus . in this figure , we see that opposite and parallel sides are identified in such a way that the orientation is preserved .
this allows us to examine the geodesic flow on the surface .
we will see in [ subsec : unfoldingabilliardorbit ] that the geodesic flow on the translation surface is dynamically equivalent to the continuous billiard flow . ]
[ exa : equilateraltriangleflatsurface ] consider the equilateral triangle @xmath120 .
the corresponding billiard is denoted by @xmath119 .
the interior angles are @xmath121 .
hence , the group acting on @xmath119 to produce the translation surface is the dihedral group @xmath122 .
the resulting translation surface is topologically equivalent to the flat torus .
we will make use of this fact in the sequel .
consider a rational polygonal billiard @xmath1 and an orbit @xmath83 . reflecting the billiard @xmath1 and the orbit in the side of the billiard table containing the basepoint @xmath123 of the orbit ( or an element of the footprint of the orbit )
partially unfolds the orbit @xmath83 ; see figure [ fig : partiallyunfoldinganorbitofthesquarebilliard ] for the case of the square billiard @xmath124 .
continuing this process until the orbit is a straight line produces as many copies of the billiard table as there are elements of the footprint ; see figure [ fig : unfoldinganorbitofthesquarebilliard ] . that is , if the period of an orbit @xmath83 is some positive integer @xmath125 , then the number of copies of the billiard table in the unfolding is also @xmath125 .
we refer to such a straight line as the _ unfolding of the billiard orbit_. .
the ` r ' is shown so as to provide the reader with a frame of reference . ] . ]
given that a rational billiard @xmath1 can be acted on by a dihedral group @xmath114 to produce a translation surface in a way that is similar to unfolding the billiard table , we can quickly see how the billiard flow is dynamically equivalent to the geodesic flow ; see figure [ fig : equivalenceofgeodesicflow ] and the corresponding caption . and correctly identifying sides so as to recover the flat torus
, we see that the unfolded orbit corresponds to a closed geodesic of the translation surface . ]
one may modify the notion of `` reflecting '' so as to determine orbits of billiard tables tiled by a rational polygon @xmath0 . as an example , we consider the unit - square billiard table . an appropriately scaled copy of the unit - square billiard table can be tiled by the unit - square billiard table by making successive reflections in the sides of the unit square .
one may then unfold an orbit of the unit - square billiard table into a larger square billiard table .
when the unfolded orbit of the original unit - square billiard intersects the boundary of the appropriately scaled ( and larger ) square , then one continues unfolding the billiard orbit in the direction determined by the law of reflection ( that is , assuming the unfolded orbit is long enough to reach a side of the larger square ) .
we will refer to such an unfolding as a _ reflected - unfolding_. we may continue this process in order to form an orbit of a larger scaled square billiard table .
suppose that an orbit @xmath83 has period @xmath125 .
the footprint of the orbit is then @xmath126 . if @xmath127 is a positive integer ( i.e. , @xmath128 ) , then the footprint @xmath129 of an orbit constitutes the footprint of an orbit that traverses the same path @xmath127-many times . for sufficiently large @xmath128 , an orbit that traverses the same path as an orbit @xmath83 @xmath127-many times
can be reflected - unfolded in an appropriately scaled square billiard table to form an orbit of the larger billiard table ; see figure [ fig : areflectedunfoldinginthesquare ] .
such a tool is useful in understanding the relationship between the billiard flow on a rational polygonal billiard @xmath1 and a billiard table tiled by @xmath0 , and will be particularly useful in understanding the nature of particular orbits of a self - similar sierpinski carpet billiard in [ subsec : aself - similarsierpinskicarpetbilliard ] .
as one may expect , if @xmath0 is a rational polygon that tiles a billiard table @xmath130 , then an orbit of @xmath130 may be folded up to form an orbit of @xmath1 .
this is done by making successive reflections in @xmath0 , the result being an orbit of @xmath1 ; see figure [ fig : foldeduporbit ] for the case of a square billiard table . .
in the first image , we see an orbit of unit - square billiard table . partitioning the unit square into nine equally sized squares
, we see that we can fold up the orbit by making successive reflections in the sides of the squares comprising the partition . using sufficiently many reflections results in an orbit of one of the squares of the partition . ]
we are primarily interested in fractals with boundaries either partially or completely comprised of self - similar sets and fractals that are self - similar .
so as to make the material discussed in [ sec : prefractalrationalbilliards][sec : fractalbilliards ] more accessible , we provide a few basic definitions from the subject of fractal geometry .
[ def : contraction ] let @xmath131 be a metric space and @xmath132 . * ( contraction).if there exists @xmath133 such that @xmath134 for every @xmath135 , then @xmath136 is called a _ contraction _ ( or _ contraction mapping _ ) . *
( similarity contraction).if there exists @xmath137 such that @xmath138 for every @xmath135 , then @xmath136 is called a _
similarity contraction_. this unique value @xmath139 is called the _ scaling ratio _ of @xmath136 .
[ def : iteratedfunctionsystemandselfsimilarset ] let @xmath131 be a complete metric space . *
( iterated function system and attractor).let @xmath140 be a family of contractions defined on @xmath141 .
then @xmath140 is called an _ iterated function system _ ( ifs ) .
+ an iterated function system is so named because the map @xmath142 , given by @xmath143 and defined on the space @xmath144 of nonempty compact subsets of @xmath141 , can be composed with itself . indeed , for each @xmath145
, we have @xmath146 furthermore , there exists a unique nonempty compact set @xmath147 ( i.e. , @xmath148 ) , called the _ attractor _ of the ifs , such that @xmath149 * ( self - similar system and self - similar set).in the special case where each @xmath150 is a contraction similarity , for @xmath151 , then the ifs @xmath140 is said to be a _ self - similar system _ and its attractor @xmath2 is called a _ self - similar set _ ( or a _ self - similar subset _ of @xmath141 ) . if @xmath141 is complete , then so is @xmath144 ( equipped with the hausdorff metric ) and hence , since it can be shown that @xmath152 is a contraction , it follows from the contraction mapping theorem that @xmath153 has a unique fixed point ( thereby justifying the definition of the attractor @xmath2 above ) and that for any @xmath154 , @xmath155 , as @xmath156 ( where , as in equation ( [ eqn : phimiterate ] ) , @xmath157 is the @xmath73th iterate of @xmath153 ) .
( see @xcite . )
we state the next property in the special case which will be of interest to us , namely , that of an ifs in a euclidean space .
[ thm : falconerstheorem ] consider an iterated function system given by contractions @xmath140 , each defined on a compact set @xmath158 , such that @xmath159 for each @xmath160 , and with attractor @xmath2 .
then @xmath161 and in fact , @xmath162 for every set @xmath154 such that @xmath163 for all @xmath160 . here ,
the transformation @xmath142 is given as in part _ _ ( _ _ i _ ) _ of definition [ def : iteratedfunctionsystemandselfsimilarset ] .
suppose @xmath2 is a fractal set .
then , the @xmath61th prefractal approximation of @xmath2 is denoted by @xmath164 . in the case of a self - similar fractal @xmath2 , the @xmath61th _ prefractal approximation _ of @xmath2
is usually defined by @xmath165 , where @xmath166 .
not every fractal is self - similar or embedded in euclidean space .
however , such sets represent an important collection of examples of fractal sets . in the next subsection
, we will discuss the fractal subsets ( self - similar or not ) of @xmath167 or of @xmath100 of direct interest to us in this paper .
a cantor set is a set with very rich and counter - intuitive properties ; topologically , it is a compact and totally disconnected ( i.e. , perfect ) space . in order to illustrate some of the properties that make a cantor set so interesting
, we refer to the canonical example of a cantor set : the ternary cantor set .
we focus on three methods for constructing the ternary cantor set : 1 ) by tremas , 2 ) as the unique fixed point attractor of an iterated function system , and 3 ) in terms of an alphabet . before we discuss the ternary cantor set
, we mention that this set was first discovered by henry j. s. smith in 1875 .
later , in 1881 , vito volterra independently rediscovered the ternary cantor set .
smith s and volterra s records being obscured over the years for one reason or another , it was the german mathematician georg cantor whom , in 1883 , history credits with the discovery of a bounded , totally disconnected , perfect and uncountable set with measure zero , that is now commonly referred to as `` the cantor set '' . we now proceed to construct the ternary cantor set , hereafter denoted by @xmath168 , by the method known as _ construction by tremas _ , which is latin for ` cuts ' .
begin with the unit interval @xmath169 and remove the middle open third @xmath170 from @xmath169 , leaving the two closed intervals @xmath171 $ ] and @xmath172 $ ] .
next , remove the middle open ninth from each closed subinterval .
what remains are the closed intervals @xmath173 $ ] , @xmath174 $ ] , @xmath175 $ ] , @xmath176 $ ] . continuing this process ad infinitum
, we construct the ternary cantor set ; see figure [ fig : ternarycantorset ] .
one may also construct @xmath177 by utilizing an appropriately defined iterated function system .
consider the following contraction maps defined on the real line @xmath178 : @xmath179 then , @xmath180 , where @xmath153 is given as in part ( i ) of definition [ def : iteratedfunctionsystemandselfsimilarset ] .
moreover , since @xmath181 is a family of similarity contractions and @xmath182 , we have that @xmath168 is a self - similar set . a third and equivalent
construction of the ternary cantor set can be given in terms of the symbols @xmath183 , @xmath184 , and @xmath185 .
recall that the elements of @xmath167 can be expressed in terms of a base-@xmath186 number system .
we focus our attention on elements of the unit interval @xmath169 .
so - called ternary numbers is a _ ternary number _ if @xmath187 , @xmath188 , @xmath189 . ] in @xmath169 have two equivalent expansions : one that is finite and one that is infinite .
for example , @xmath190 can be written in base-@xmath186 as @xmath191 or , equivalently , as @xmath192 ( where the overbar indicates that the digit @xmath193 is repeated infinitely often ) .
we next discuss a similar _ addressing system _ that has the benefit of preventing ternary numbers from having a finite representation .
the characters @xmath183 , @xmath184 and @xmath185 are to remind the reader of choosing _ left _ , _
center _ and _ right_. we identify an element of the unit interval @xmath169 by an infinite address that indicates _ where _ in @xmath169 the element is located .
motivated by the construction of @xmath168 by tremas , one can identify any point of @xmath169 by an infinite address consisting of the characters @xmath183 , @xmath184 and @xmath185 . while elements of @xmath168 can be represented by infinite addresses consisting of @xmath184 s
, we make the stipulation that no element of @xmath168 will be represented by an infinite address containing @xmath184 s . to be approximated by a sequence @xmath194 of elements of @xmath195 , where @xmath195 is the complement of the ternary cantor set in @xmath169 .
] moreover , this method of representing elements of @xmath169 ( or @xmath177 ) provides every element with an infinite representation and never a finite representation .
the values @xmath196 , @xmath190 and @xmath197 have the ternary representations @xmath198 , @xmath199 and @xmath200 , respectively .
has a representation given by @xmath201 .
although , we will not consider this as a representation for @xmath190 on account of @xmath202 .
] while @xmath190 has a finite ternary expansion given by @xmath203 , it does not have a finite ternary representation .
it should be noted that elements like @xmath196 and @xmath197 will play an important role in our analysis of the koch snowflake fractal billiard .
the occurrence of infinitely many @xmath184 s or infinitely many @xmath183 s and @xmath185 s is critical to developing some of the theory regarding the koch snowflake fractal billiard .
so that some of the results concerning the koch snowflake fractal billiard can be more succinctly expressed , we introduce a notation used for describing a value s _ type of ternary representation_. [ not : typeofternaryrepresentation ] the _ type of ternary representation _ can be defined as follows . if @xmath204 , then the first coordinate of @xmath205 $ ] describes the characters that occur infinitely often and the second coordinate of @xmath206 $ ] describes the characters that occur finitely often .
if we want to discuss many different types of ternary representations , then we use ` or ' .
that is , the notation @xmath206\vee [ \cdot,\cdot]\vee ... \vee[\cdot,\cdot]$ ] is to be read as _
@xmath206 $ ] or @xmath206 $ ] or ... or @xmath206$]_. if the collection of characters occurring finitely often is empty , then we denote the corresponding type of ternary representation by @xmath207}$ ] .
the value @xmath197 has a ternary representation of @xmath200 .
hence , @xmath197 has a type of ternary representation given by @xmath208}$ ] .
moreover , the value @xmath209 has a ternary representation given by @xmath210 , which means that @xmath209 has a type of ternary representation given by @xmath211}$ ] .
we note that `` the '' type of representation of a point @xmath204 is not unique , in general .
for instance , the value @xmath190 has a ternary representation of type @xmath212}$ ] or @xmath213}$ ] .
a thorough understanding of the ternary cantor set is not only important for understanding many of the results on the koch snowflake prefractal and fractal billiard . in general
, cantor sets will be ever - present and instrumental in our analysis of other fractal billiard tables .
in each example of a fractal billiard , we will clearly indicate where and how a particular cantor set is important in analyzing a particular fractal billiard table .
the koch curve @xmath214 is constructed as shown in figure [ fig : constructionofkochcurvefrom2dcompactset ] and is the unique fixed point attractor of the following iterated function system on the euclidean plane ( here , @xmath215 ) : @xmath216 since each contraction map in the iterated function system is a similarity transformation ( i.e. , @xmath217 is a self - similar system ) and @xmath218 , we have that @xmath214 is a self - similar set ; see part ( ii ) of definition [ def : iteratedfunctionsystemandselfsimilarset ] .
there are additional properties of the koch curve that are reminiscent of the cantor set ; this is more than just a coincidence and is discussed in more detail below .
if we allow the iterated function system to act on the triangle @xmath219 , as shown in figure [ fig : constructionofkochcurvefrom2dcompactset ] , sequential iterates of the iterated function system very quickly produce a prefractal that is visually indiscernible from the true limiting set .
but there is a more common construction that allows us to visualize the curve @xmath214 more readily , this being depicted in figure [ fig : kochcurveconstruction ] .
the technical caveat which we are brushing under the carpet is that each polygonal approximation shown in figure [ fig : kochcurveconstruction ] does not contain the koch curve @xmath214 , while each approximation in the sequence shown in figure [ fig : constructionofkochcurvefrom2dcompactset ] does contain @xmath214 . and theorem [ thm : falconerstheorem ] that for a set @xmath2 to be the unique fixed point attractor of an ifs
, each @xmath164 must be such that @xmath220 , so that @xmath221 . ] . here
, the self - similar set @xmath214 is viewed as a limit of the prefractal approximations @xmath222 , where @xmath223 is the initial triangle and the map @xmath153 is defined in terms of the ifs given by equation ( [ eqn : ifsforthekochcurve ] ) , as in definition [ def : iteratedfunctionsystemandselfsimilarset ] .
( see theorem [ thm : falconerstheorem ] and the text preceding it . ) ] when learning about fractal sets .
beginning with the unit interval @xmath169 , one removes the middle third and replaces it with the two other sides of an equilateral triangle , as shown .
one then repeats this process infinitely often for every remaining interval ; the resulting limiting set is @xmath214 .
such a sequence @xmath224 of approximations converges to @xmath214 , because it is a subsequence of the convergent sequence of prefractal approximations @xmath222 shown in figure [ fig : constructionofkochcurvefrom2dcompactset ] .
( here , we are using the notion of convergence in the sense of the hausdorff metric . ) ] for each integer @xmath225 , we denote by @xmath226 the @xmath61th ( inner ) polygonal approximation of the koch curve @xmath214 .
intuitively , one expects the koch curve to have finite length , since it is the limit of a sequence of polygonal approximations . on the contrary
, the koch curve @xmath214 has infinite length , which can be seen by the following calculation given in terms of the @xmath61th prefractal @xmath226 , where @xmath226 is one of the polygonal approximations indicated in figure [ fig : kochcurveconstruction ] : @xmath227 then , @xmath228 .
the koch snowflake @xmath12 is a fractal comprised of three abutting copies of the self - similar koch curve ; see figure [ fig:3kochcurveslabeled ] . for each integer @xmath225 , we denote by @xmath229 the @xmath61th ( inner ) polygonal approximation of the koch snowflake @xmath12 .
is the union of three abutting copies of @xmath214 . ] as a closed ( simple ) curve , the koch snowflake @xmath12 bounds a region of the plane ; furthermore , the area of this region can be calculated as follows : @xmath230 then , as @xmath61 increases , the right - hand side of ( [ eqn : areaofkochsnowflake ] ) tends to a finite value .
the area bounded by the koch snowflake is thus given by @xmath231 , assuming the sides of @xmath232 have length one .
as we noted at the end of [ subsec : cantorsets ] , cantor sets are ever present in the context of self - similarity . in the case of the koch snowflake
, @xmath233 is the union of @xmath234 self - similar ternary cantor sets , each spanning a distance of @xmath235 .
such a fact will be important in determining certain sequences of what we will call _ compatible orbits _
( see definitions [ def : compatibleinitialconditions][def : sequenceofcompatibleorbits ] ) and certain families of well - defined orbits of @xmath4 .
the @xmath5-fractal @xmath13 , discussed in @xcite in a different context , is not a self - similar set .
however , @xmath13 contains , as a proper subset , a set that is constructed in a way that is reminiscent of an iterated function system acting on a compact set so as to produce a self - similar set . and theorem [ thm : falconerstheorem ] that each prefractal approximation @xmath164 must contain the unique fixed point attractor @xmath2 .
] as shown in figure [ fig : t - fractal ] , one constructs the @xmath5-fractal by appending scaled copies of the initial @xmath5 shape @xmath236 to each successive approximation .
specifically , @xmath237 is constructed from @xmath238 by appropriately appending @xmath239 copies of @xmath240 to @xmath241 .. ] -fractal @xmath13 . ]
the overall height of @xmath13 can be calculated and the total area bounded by @xmath242 can be shown to be finite , as shown in the following calculations ( we assume here that the base of @xmath236 is two units in length ) : @xmath243 then , @xmath244 , which is the height of @xmath13 . furthermore , the area bounded by @xmath238 is calculated as follows .
there are eight squares , each with side - length one , comprising @xmath236 ; see figure [ fig : t0tiledbyunitsquare ] . hence , the area of @xmath236 is eight square - units .
therefore , @xmath245 then , @xmath246 , which is the total area bounded by @xmath13
. can be tiled by the unit square @xmath247 . ]
there is a natural fractal subset of @xmath13 , but , for each @xmath225 , no point of @xmath238 is in this fractal subset , which is unlike what we have seen in the case of the koch snowflake fractal @xmath12 .
in fact , the fractal subset in question is given by @xmath248 .
we note that this fractal subset is not self - similar and each point a priori fails to yield a well - defined tangent necessary for calculating the angle of reflection of a billiard ball traversing the billiard table .
a sierpinski carpet can be constructed by systematically removing particular open subsquares from the unit square @xmath249 . depending on how one chooses the sizes of the open subsquares to be removed
, one can either construct a self - similar sierpinski carpet or a non - self - similar sierpinski carpet , these being defined below .
this method of construction is called _ construction by tremas _ and is described in the caption of figure [ fig : polygonalappxof1 - 3sierpinskicarpet ] , using the standard `` @xmath250-sierpinski carpet '' as an example
. such a construction process should be very familiar , since `` removing middle thirds '' is exactly what we did to construct the ternary cantor set in
[ subsec : cantorsets ] .
-sierpinski carpet is a self - similar carpet constructed in one of two ways : 1 ) by tremas and 2 ) an iterated function system ( in fact , a self - similar system ) .
we describe here the construction of the @xmath250-sierpinski carpet by tremas , the latter being further discussed in the main text . beginning with the unit square ,
one then removes the middle open square with side - length @xmath190 . from each remaining subsquare of side - length @xmath190 ,
one then removes the middle open square of side - length @xmath251 .
one continues this procedure of removing subsquares of remaining squares until there is no area left .
as one would expect , each step of the construction process can be emulated by applying the correct iterated function system , which is given in equation ( [ eqn : theifsfor1 - 3sierpinskicarpet ] ) . ]
as referred to in the caption of figure [ fig : polygonalappxof1 - 3sierpinskicarpet ] , one may also construct the @xmath250-sierpinski carpet by applying an appropriately defined iterated function system to the unit square @xmath247 .
consider the following iterated function system , which is a self - similar system .
@xmath252 then , denoting the @xmath250-sierpinski carpet by @xmath253 , we have that @xmath254 .
since each contraction in the iterated function system is a similarity contraction and @xmath255 , it follows that @xmath253 is a self - similar set .
we discuss here the relevant results and material from @xcite . for our purposes
, the first level approximation of a sierpinski carpet @xmath256 will always be the unit square @xmath247 and denoted by @xmath257 .
since every ( self - similar and non - self - similar ) sierpinski carpet has the same zeroth level approximation and zero is never a scaling ratio , such notation will never cause any confusion .
what follows is a general description on how to construct a sierpinski carpet by removing appropriately sized middle open squares .
consider the unit square @xmath258 .
let @xmath259 for some @xmath260 .
partition @xmath257 into @xmath261 squares of side - length @xmath262 .
next , remove the middle open subsquare .
let @xmath263 for some @xmath264 .
each subsquare may then be partitioned into @xmath265 many squares with side - length @xmath266 .
we then remove each middle open subsquare of side - length @xmath266 ; see figure [ fig : polygonalappxof1 - 3sierpinskicarpet ] . continuing this process , let @xmath267 where @xmath268 and let @xmath269 for some @xmath270 .
then we partition a subsquare of side - length @xmath271 into @xmath272 many squares .
we then remove the middle open square from each subsquare in the partition .
continuing in this manner ad infinitum , one constructs a sierpinski carpet denoted by @xmath273 , where @xmath274 .
[ def : aselfsimilarsierpinskicarpet ] if @xmath275 , with @xmath276 and @xmath277 , is a periodic sequence of rational values , then the sierpinski carpet @xmath256 is called a _ self - similar sierpinski carpet_. we have described the construction of a self - similar sierpinski carpet @xmath15 in terms of the removal of particular open squares . as the name would suggest , there exists a suitably defined iterated function system @xmath140 such that @xmath278 . viewing @xmath279 as the unique fixed point attractor of an appropriately defined iterated function system will be useful in stating some of the results in the subsequent sections . more precisely , @xmath15 is viewed as the self - similar set associated with a self - similar system , as in part ( ii ) of definition [ def : iteratedfunctionsystemandselfsimilarset ] . while we do not discuss any results concerning non - self - similar sierpinski carpet billiards in this paper , we provide the definition for completeness .
if @xmath275 , with @xmath276 and @xmath277 , is an aperiodic sequence of rational values , then the sierpinski carpet @xmath256 is called a _
non - self - similar sierpinski carpet_. [ def : acellofsai ] let @xmath280 , @xmath281 .
consider a partition of the unit square @xmath258 into @xmath282 many squares of side - length @xmath262 .
a subsquare of the partition is called a _ cell of @xmath257 _ and is denoted by @xmath283 .
furthermore , let @xmath279 be a sierpinski carpet .
consider a partition of the prefractal approximation @xmath284 into subsquares with side - length @xmath285 .
a subsquare of the partition of @xmath284 is called a _ cell of @xmath284 _ and is denoted by @xmath286 and has side - length @xmath285 . in accordance with the convention adopted in @xcite , the boundary of an open square removed in the construction of @xmath256
is called a _ peripheral square _ of @xmath256 . furthermore , by convention , the unit square @xmath258 is not a peripheral square .
[ def : nontriviallinesegmentofsa ] a _ nontrivial line segment of @xmath256 _ is a ( straight - line ) segment of the plane contained in @xmath256 and which has nonzero length .
unless otherwise indicated , in what follows , we assume that @xmath256 is a self - similar sierpinski carpet with a single scaling ratio @xmath16 ; that is , @xmath287 , where @xmath288 for some fixed @xmath289 .
in addition , when @xmath290 , @xmath15 is denoted by @xmath291 .
we next state the following theorem , due to durand - cartagena and tyson in @xcite and which will be very useful to us in this context ( see [ subsec : aprefractalselfsimilarsierpinskicarpetbilliard ] and [ subsec : aself - similarsierpinskicarpetbilliard ] ) .
[ thm : asetbset ] let @xmath291 be a self - similar sierpinski carpet
. then the set of slopes @xmath292 of nontrivial line segments of @xmath291 is the union of the following two sets _ _ : _ _ @xmath293 moreover , if @xmath294 , then each nontrivial line segment in @xmath291 with slope @xmath295 touches vertices of peripheral squares , while if @xmath296 , then each nontrivial line segment in @xmath291 with slope @xmath295 is disjoint from all peripheral squares .
[ not : aabaabbb ] let @xmath297 be odd positive integers such that @xmath298 and let @xmath292 and @xmath299 be the set of slopes of nontrivial line segments of @xmath291 and @xmath300 , respectively .
we denote by @xmath301 ( resp . ,
@xmath302 ) the subset @xmath303 ( resp . , @xmath304 ) given in equation ( [ eqn : aset ] ) of theorem [ thm : asetbset ] .
similarly , we denote by @xmath305 ( resp . ,
@xmath306 ) the subset @xmath307 ( resp . , @xmath299 ) given in equation ( [ eqn : bset ] ) of theorem [ thm : asetbset ] .
( resp . , @xmath306 )
, @xmath16 should of course be replaced by @xmath308 in equation ( [ eqn : aset ] ) ( resp .
, equation ( [ eqn : bset ] ) ) . ]
if @xmath291 and @xmath300 are self - similar sierpinski carpets with @xmath309 , then it is clear that @xmath310 .
moreover , in this case , we also have that @xmath311 and @xmath312 .
we note that if @xmath295 is the slope of a nontrivial line segment in @xmath291 , then so is @xmath313 , @xmath314 and @xmath315 by symmetry of the carpet .
however , we restrict our attention in this paper to the slopes described in the above result of @xcite .
in the previous sections , we surveyed basic facts and results from mathematical billiards and fractal geometry , with most of our attention being focused on the subject of rational billiards and sets exhibiting self - similarity .
we also discussed the importance of examining the dynamically equivalent geodesic flow on an associated translation surface . in this section
, we will examine examples from particular classes of prefractal ( rational ) billiards .
we are interested in tables that can be tiled by a single polygon which can also tile the ( euclidean ) plane .
the main examples we will discuss are the koch snowflake prefractal billiard table , the @xmath5-fractal prefractal billiard table and a self - similar sierpinski carpet prefractal billiard table .
each example of a prefractal billiard table constitutes a rational billiard table , but is an element of a sequence of rational billiard tables approximating a fractal billiard table with radically different qualities when compared to the others .
that is , the koch snowflake has an everywhere nondifferentiable boundary ; the @xmath5-fractal billiard table is certainly a fractal billiard table , since its boundary @xmath13 contains a fractal set , but the portion of the boundary that is nondifferentiable has lebesgue measure zero ; a sierpinski carpet billiard table can possibly have no area , yet yield billiard orbits of finite length .
we restrict our attention to billiard tables with fractal boundary @xmath2 , where @xmath2 can be approximated by a suitably chosen sequence of rational polygons @xmath316 .
more specifically , we are interested in a fractal billiard table @xmath317 with the property that , for every @xmath225 , @xmath89 can be tiled by a single polygon @xmath318 , where @xmath319 . here , @xmath320 is a suitably chosen scaling ratio and @xmath321 is a polygon that tiles both the ( euclidean ) plane as well as the rational billiard @xmath322 .
is exactly @xmath321 . in general , however , @xmath321 does not always equal @xmath322 , but certainly tiles @xmath322 .
an example of this situation is @xmath323 ; see [ subsec : thetfractalprefractalbilliard ] .
such a billiard is tiled by the unit square , which is the associated polygon @xmath321 . ]
the focus in this subsection is on developing a general framework for discussing billiards on prefractal approximations .
if @xmath62 and @xmath324 are two prefractal billiard tables approximating a given fractal billiard table @xmath3 , then we want to have a systematic way of determining how and if two orbits @xmath325 and @xmath326 of @xmath62 and @xmath324 are related .
we will primarily measure angles relative to a fixed coordinate system , with the origin being fixed at a corner of a prefractal approximation @xmath327 .
however , we will sometimes measure an angle relative to a side of @xmath164 on which a billiard ball lies . in such situations , we will write the angle as @xmath328 in order to indicate that the inward pointing direction is @xmath329 , measured relative to the side on which the vector is based . to motivate our general discussion ,
consider the orbit @xmath330 of @xmath331 , where @xmath332 ; see the first image in figure [ fig : notapwfagnanoorbit ] ( and recall our earlier discussion in
[ subsec : cantorsets ] ) .
the same orbit , viewed as a continuous curve embedded in @xmath333 , does not constitute an orbit of @xmath333 ; see the second image in figure [ fig : notapwfagnanoorbit ] .
consider the orbit @xmath334 shown in the third image in figure [ fig : notapwfagnanoorbit ] .
such an orbit does intersect the boundary of @xmath333 and appears to be related to @xmath335 , but in what way we have not yet explicitly said .
initially , we notice that , as a continuous curve embedded in @xmath333 , the orbit @xmath335 is a subset of @xmath334 .
being eager to establish a proper notion of `` related '' , we may be inclined to declare that two orbits are related if one is a subset of the other , when viewed as continuous curves in the plane .
unfortunately , we quickly see that such a definition is highly restrictive .
a more general observation is that @xmath336 and @xmath337 are collinear in the direction of @xmath338 , without any portion of @xmath339 intersecting the segment @xmath340 .
we then say that @xmath341 and @xmath342 are _ compatible initial conditions_. we state the formal definition as follows . of @xmath331 . in the second image
, we see that the orbit @xmath335 , when embedded in @xmath333 , is not an orbit of @xmath333 . in the third image ,
the given orbit of @xmath333 intersects sides of @xmath333 and appears to be related to @xmath335 in some way . ]
[ def : compatibleinitialconditions ] without loss of generality , suppose that @xmath61 and @xmath73 are nonnegative integers such that @xmath343 .
let @xmath344 and @xmath345 be two initial conditions of the orbits @xmath325 and @xmath346 , respectively , where we are assuming that @xmath347 and @xmath348 are both inward pointing . if @xmath349 and if @xmath350 and @xmath351 lie on a segment determined from @xmath347 ( or @xmath348 ) that intersects @xmath89 only at @xmath350 , then we say that @xmath69 and @xmath352 are _ compatible initial conditions_. when two initial conditions @xmath69 and @xmath352 are compatible , then we simply write each as @xmath353 and @xmath354 .
if two orbits @xmath346 and @xmath325 have compatible initial conditions , then we say such orbits are _
compatible_. depending on the nature of @xmath317 , not
every orbit must pass through the region of @xmath62 corresponding to the interior of @xmath322 , let alone pass through the interior of @xmath355 , for any @xmath356 .
because of this , it may be the case that an initial condition @xmath353 is not compatible with @xmath354 , for any @xmath356 .
as such , in definitions [ def : sequenceofcompatibleinitialconditions ] and [ def : sequenceofcompatibleorbits ] , we consider sequences beginning at @xmath357 , for some @xmath358 .
[ def : sequenceofcompatibleinitialconditions ] let @xmath359 be a sequence of initial conditions , for some integer @xmath358 .
we say that this sequence is a _ sequence of compatible initial conditions _ if for every @xmath360 and for every @xmath361 , we have that @xmath69 and @xmath352 are compatible initial conditions .
in such a case , we then write the sequence as @xmath362 .
[ def : sequenceofcompatibleorbits ] consider a sequence of compatible initial conditions @xmath363
. then the corresponding sequence of orbits @xmath364 is called _ a sequence of compatible orbits_. if @xmath365 is an orbit of @xmath355 , then @xmath365 is a member of a sequence of compatible orbits @xmath364 for some @xmath358 .
it is clear from the definition of a sequence of compatible orbits that such a sequence is uniquely determined by the first orbit @xmath366 .
since the initial condition of an orbit determines the orbit , we can say without any ambiguity that a sequence of compatible orbits is determined by an initial condition @xmath367 .
let @xmath368 be a property ( resp .
, @xmath369 a list of properties ) .
if every orbit in a sequence of compatible orbits has the property @xmath368 ( resp .
, a list of properties @xmath369 ) , then we call such a sequence _ a sequence of compatible @xmath368 _ _ ( _ _ resp .
, @xmath369 _ _ ) _ _ orbits_. the following theorem can be deduced from theorem 3 of gutkin s paper @xcite ; see @xcite .
[ thm : topologicaldichotomyforfn ] consider a prefractal rational billiard @xmath62 .
if @xmath370 is tiled by a rational polygon @xmath318 such that @xmath318 tiles the euclidean plane , then , for a fixed direction @xmath347 , every orbit @xmath371 of @xmath62 is closed or every orbit @xmath371 is dense in @xmath62 , . ] regardless of the initial basepoint @xmath350 . when @xmath62 is tiled by @xmath318 , where @xmath318 is a rational polygon tiling the plane , then @xmath62 is more generally referred to as an _ almost integrable billiard _ ,
this being the language used in @xcite .
the following is a generalization to this broader setting of corollary 16 from @xcite .
it is established in the same manner .
[ thm : generaltopologicaldichotomyforsequencesofcompatibleorbits ] let @xmath317 be a fractal billiard table approximated by a suitable sequence of rational polygonal billiard tables @xmath372 .
if there exists a polygon @xmath321 that tiles the plane and such that for every @xmath225 there exists @xmath320 with @xmath373 tiling @xmath62 , then any sequence of compatible orbits is either entirely comprised of closed orbits or entirely comprised of orbits that are dense in their respective billiard tables .
the billiard @xmath8 can be tiled by equilateral triangles .
specifically , if @xmath120 is the equilateral triangle with sides having unit length , then @xmath8 is tiled by @xmath374 , for every @xmath225 . moreover , as is well known , @xmath375 tiles the plane .
therefore , theorems [ thm : topologicaldichotomyforfn ] and [ thm : generaltopologicaldichotomyforsequencesofcompatibleorbits ] hold for the prefractal billiard @xmath8 .
our goal for this subsection and [ subsubsec : thecorrespondingprefractaltranslationsurface ] is to survey some of the main results of @xcite .
we will focus on pertinent examples that will motivate a richer discussion in [ subsec : thekochsnowflakefractalbilliard ] .
initially , we focus on properties of orbits with an initial direction of @xmath338 and @xmath376 . and
@xmath377 , since @xmath377 is the rotation of @xmath376 through the angle @xmath338 , the angle @xmath338 being an angle that determines an axis of symmetry of @xmath229 , for @xmath225 . ]
if @xmath330 is an orbit of @xmath331 , so long as @xmath378 is not a corner of @xmath232 , the orbit will be periodic , as expected .
however , depending on the nature of the ternary representation of @xmath378 , the compatible orbit @xmath334 may be singular in @xmath333 . of [ subsec : cantorsets ] . ] let @xmath379 .
if @xmath337 has a ternary representation of type @xmath380}\vee { [ r , lc]}$ ] , then there exists @xmath358 such that the compatible orbit @xmath381 will be singular in @xmath382 .
moreover , for every @xmath383 , @xmath384 will also be singular in @xmath8 .
if @xmath378 has a ternary representation of the form @xmath385}\vee{[lc , r]}\vee{[cr , l]}\vee{[lcr,\emptyset]}\vee{[lr , c]}$ ] , then the sequence of compatible orbits given by @xmath386 is a sequence of compatible periodic orbits .
the length and period of an orbit @xmath387 is dictated by the ternary representation of @xmath378 .
_ _ ( _ _ see @xcite for the corresponding specific formulas .
_ ) _ see 4.4 of @xcite for a precise statement and proof of this result , as well as for additional properties of orbits with an initial direction of @xmath338 .
[ exa : asequenceofcompatiblehookorbits ] let @xmath388 have a ternary representation given by @xmath389 .
such a representation indicates that , in each prefractal approximation @xmath229 , @xmath337 is an element of an open , connected neighborhood contained in @xmath229 .
the point @xmath337 corresponds to the value @xmath390 .
if we consider an orbit of @xmath331 with an initial direction of @xmath376 , the ternary representation of the basepoints at which the billiard ball path forms right angles with the sides of @xmath331 is of the type @xmath385}$ ] .
this is a degenerate periodic hybrid orbit , meaning that it doubles back on itself , and the next orbit in the sequence of compatible periodic hybrid orbits has the initial condition @xmath391 .
since the ternary representation of the basepoint of @xmath392 is @xmath393 and @xmath394 , it follows that the basepoint of @xmath395 is a point which , for every prefractal approximation @xmath229 , is an element of an open , connected neighborhood contained in @xmath229 .
then the basepoint of @xmath396 ( where @xmath397 denotes the second iterate of the billiard map @xmath398 ) has a ternary representation of type @xmath385}$ ] .
this same pattern is repeated for every subsequent orbit in the sequence of compatible orbits .
it follows that the resulting sequence of compatible orbits forms a sequence of orbits that is converging to a set which is well defined .
that is , such a set will be some path in the fractal billiard table @xmath4 with finite length which is effectively determined by the law of reflection in each prefractal approximation of @xmath4 .
such orbits are introduced in @xcite and referred to as _ hook orbits _ , because they appear to be `` hooking '' into the koch snowflake ; see figure [ fig : hybridpi6lv5 ] .
the hook orbits of example [ exa : asequenceofcompatiblehookorbits ] are special cases of a general class of orbits called hybrid orbits , which were introduced , as well as studied , in @xcite .
[ def : hybridorbit ] let @xmath325 be an orbit of @xmath8 . if all but at most two basepoints @xmath399 have ternary representations ( determined with respect to the side @xmath400 on which each point resides ) of type @xmath385}\vee{[cl , r]}\vee{[cr , l]}\vee{[lcr,\emptyset]}\vee{[lr,\emptyset]}$ ] , then we call @xmath371 a _ hybrid orbit _ of @xmath8 . a hybrid orbit is so named for the fact that it _ may _ have qualities reminiscent of an orbit @xmath326 that is identical to the compatible orbit @xmath325 and an orbit @xmath401 that is visually different from the compatible orbit @xmath402 ; see figure [ fig : hybrid012 ] and its caption .
if @xmath325 is a hybrid orbit with property @xmath403 , then we say that it is a @xmath403 hybrid orbit .
[ prop : denseorbitisadensehybridorbit ] if @xmath325 is a dense orbit of @xmath8 , then @xmath325 is a dense hybrid orbit . applying the results in theorem [ thm : generaltopologicaldichotomyforsequencesofcompatibleorbits ] and proposition [ prop : denseorbitisadensehybridorbit ] , we state the following result .
[ thm : atopologicaldichotomy ] let @xmath364 be a sequence of compatible orbits .
then we have that @xmath364 is either entirely comprised of closed orbits or is entirely comprised of dense hybrid orbits .
, just before definition [ def : footprintofanorbit ] . ]
[ thm : hybridorbitofks0implieshybridorbitofksn ] if @xmath404 is a periodic hybrid orbit of @xmath331 with no basepoints corresponding to ternary points _ _ ( _ _ i.e . , points having ternary representations of the types @xmath380}\vee{[r , lc]}$ ] _ _ ) _ _ , then for every @xmath225 , the compatible orbit @xmath325 is a periodic hybrid orbit of @xmath8 . in order to fully understand the following result , we define what it means for a vector to be _ rational with respect to a basis _ @xmath405 of @xmath100 .
if @xmath406 , for some @xmath407 , then we say that @xmath408 is _ rational with respect to the basis _
otherwise , we say that @xmath408 is _ irrational with respect to _ @xmath405 .
let @xmath388 and consider a vector @xmath409 that is rational with respect to the basis @xmath410 .
then , we have the following _ _ : _ _ 1 .
if @xmath16 and @xmath308 are both positive integers with @xmath308 being odd , @xmath411 , for some @xmath412 with @xmath413 , @xmath414 being odd and @xmath415 , then the sequence of compatible closed orbits @xmath416 is a sequence of compatible periodic hybrid orbits .
if @xmath417 , @xmath308 is a positive odd integer , @xmath418 , for some @xmath412 with @xmath413 , @xmath419 being odd and @xmath415 , then the sequence of compatible closed orbits @xmath416 is a sequence of compatible periodic hybrid orbits .
[ thm : bodd ] we want to emphasize that the angle @xmath38 in part ( 1 ) and part ( 2 ) of theorem [ thm : bodd ] is not necessarily @xmath338 , @xmath377 or @xmath376 , but can assume countably infinitely many values .
[ exa : asequenceofcompatibleperiodichybridorbits ] in figure [ fig : hybrid012 ] , three periodic hybrid orbits are displayed .
these three orbits constitute the first three terms in a sequence of compatible periodic hybrid orbits.(2 ) , the angle @xmath420 determined by the initial segment of the orbit and the initial basepoint @xmath421 both guarantee that the sequence of compatible orbits @xmath416 is a sequence of compatible periodic hybrid orbits . ] if we choose @xmath422 and @xmath423 to be an angle such that @xmath337 connects with the midpoint of the lower one - third interval on the side of @xmath331 , we can see that @xmath404 is a periodic hybrid orbit .
more importantly , there are elements of the footprint @xmath424 with ternary representations of type @xmath211}$ ] .
this observation is key for constructing what we call nontrivial paths of @xmath4 , a topic which is discussed in more detail in [ subsec : thekochsnowflakefractalbilliard ] . .
in order to understand exactly what is discussed in the paragraph immediately following definition [ def : hybridorbit ] , compare and contrast the hybrid orbits shown here with the hybrid orbits shown in figures [ fig : notapwfagnanoorbit ] , [ fig : hybridpi6lv5 ] and [ fig : compatiblecantororbit7 - 12 ] .
certain segments of the hybrid orbits shown here remain intact and become subsets of subsequent compatible periodic hybrid orbits , yet the orbits are visually different from one another . ]
given a nonnegative integer @xmath425 , we say that a sequence of compatible orbits @xmath364 is a _ constant sequence of compatible orbits _ if the path traversed by @xmath326 is identical to the path traversed by @xmath325 , for every @xmath383 .
furthermore , we say that a sequence of compatible orbits @xmath416 is _ eventually constant _ if there exists a nonnegative integer @xmath425 such that @xmath364 is constant , in the above sense .
[ thm : sufficientconditionforcantororbit ]
let @xmath404 be an orbit of @xmath331 such that every @xmath426 has a ternary representation of type @xmath211}$ ] .
then @xmath416 is a sequence of compatible periodic hybrid orbits .
moreover , there exists @xmath358 such that @xmath364 is a constant sequence of compatible periodic hybrid orbits .
[ exa : aconstantsequenceofcompatibleperiodichybridorbits ] consider @xmath427 in the base of the equilateral triangle .
such a value has a ternary representation of type @xmath211}$ ] .
consider the initial condition @xmath428 .
then the sequence of compatible orbits @xmath429 is a constant sequence .
this follows from the fact that the ternary representation of @xmath336 is @xmath389 .
moreover , the representation of every basepoint of @xmath384 is @xmath198 . in figure
[ fig : compatiblecantororbit7 - 12 ] , we show the first three orbits in this ( eventually ) constant sequence of compatible periodic hybrid orbits . as of now , the only examples of constant sequences of compatible periodic hybrid orbits are those for which the initial direction is @xmath338 and @xmath376 ( and , equivalently , @xmath377 ) . when the initial angle of an orbit of a constant sequence of compatible periodic hybrid orbits is @xmath376 ( or , equivalently , @xmath377 ) , then the orbit will be degenerate .
for example , the orbit @xmath430 traverses a path that is a vertical line .
this orbit has period @xmath431 .
while @xmath432 is an important example of a constant sequence of compatible periodic hybrid orbits , it is arguably less interesting than the constant sequence of compatible periodic hybrid orbits @xmath433 .
lies on the middle third of the unit interval .
the basepoint @xmath336 of the compatible initial condition @xmath342 has a ternary representation of type @xmath434}$ ] . ]
in [ subsec : translationstructuresandtranslationsurfaces ] we saw how to construct a translation surface from a rational billiard table . in the case of the equilateral triangle billiard table @xmath436
, there are @xmath437 copies of @xmath119 used in the construction of the associated translation surface @xmath438 ; see example [ exa : equilateraltriangleflatsurface ] and the associated figure [ fig : sixequisurface ] . in the case of the prefractal billiard table @xmath8 ,
only six copies of @xmath8 are needed in the construction of the associated translation surface @xmath439 , for every @xmath225 ; see figure [ fig : thethreecorrespondingflatsurfaces ] .
( we refer to @xcite for further discussion of the topics in the present subsection . ) , @xmath440 and @xmath441 associated with the koch snowflake prefractal approximations @xmath339 , @xmath442 and @xmath443 , respectively.,title="fig : " ] , @xmath440 and @xmath441 associated with the koch snowflake prefractal approximations @xmath339 , @xmath442 and @xmath443 , respectively.,title="fig : " ] , @xmath440 and @xmath441 associated with the koch snowflake prefractal approximations @xmath339 , @xmath442 and @xmath443 , respectively.,title="fig : " ] the vertices of @xmath8 correspond to conic singularities of the translation surface
. however , only certain singularities are removable .
the vertices with angles measuring @xmath338 ( measured from the interior ) , constitute removable singularities of the translation surface .
that is , the geodesic flow can be appropriately defined at these points .
the vertices with angles measuring @xmath444 constitute nonremovable singularities .
hence , it is possible to define reflection at certain vertices of the prefractal billiard @xmath8 , but impossible to define at others .
moreover , defining reflection at acute corners of @xmath8 in this way is independent of @xmath61 .
that is , for a given vertex @xmath445 of @xmath8 with an acute angle @xmath338 , the general rule for reflection in @xmath445 states that an incoming trajectory reflect through the angle bisector of @xmath445 .
a billiard ball entering @xmath445 along the same path in @xmath446 as in @xmath8 will then reflect in @xmath445 in @xmath446 in exactly the same way as it did when considering @xmath445 as a vertex of @xmath8 .
such insight is clearly helpful in further understanding the behavior of a billiard ball on the koch snowflake fractal billiard @xmath4 , but we must be careful not to extrapolate more than is possible from this observation . knowing that we can determine an orbit of a prefractal billiard @xmath8 by unfolding the orbit of @xmath331 in @xmath8 , we are inclined to allow orbits of @xmath331 that make collisions with corners .
however , a priori , we can not conclude that such orbits do not unfold to form saddle connections in @xmath8 connecting two nonremovable singularities . in the event an orbit @xmath346 of @xmath447
intersects the boundary @xmath448 solely in acute corners , then such an orbit is an element of a sequence of compatible orbits @xmath364 with @xmath449 , for every @xmath450 .
we refer to [ subsec : thetfractal ] for a discussion of the @xmath5-fractal @xmath13 and of its prefractal approximations @xmath451 , for @xmath452 ; see , in particular , figure [ fig : t - fractal ] .
recall that the base of @xmath236 has a length of two units .
the prefractal billiard @xmath323 can be tiled by the unit square @xmath247 ; see figure [ fig : t0tiledbyunitsquare ] . in general
, for every @xmath225 , @xmath9 can be tiled by the square @xmath453 . as such , and
since @xmath247 obviously tiles the plane , we can apply theorems [ thm : topologicaldichotomyforfn ] and [ thm : generaltopologicaldichotomyforsequencesofcompatibleorbits ] . much like the case of the prefractal koch snowflake billiard @xmath8 , we are interested in forming sequences of compatible orbits of prefractal billiards exhibiting particular properties .
the results in this subsection appear here for the first time and will be further discussed in @xcite .
it is true that if a periodic orbit has an initial condition @xmath454 , then there may exist a compatible orbit @xmath455 that forms a saddle connection if @xmath378 has a finite binary expansion .
this is not to suggest that @xmath455 _ must _ form a saddle connection .
however , if every basepoint @xmath456 of a periodic orbit @xmath457 of the unit square has an infinite binary expansion ( with no equivalent finite binary expansion ) , then viewing @xmath458 in @xmath323 as the reflected - unfolding of @xmath457 , the corresponding sequence of compatible orbits @xmath416 will be a sequence of compatible periodic orbits .
we state this formally in the following theorem .
let @xmath454 be an initial condition of an orbit @xmath459 of @xmath124 .
suppose every element of the footprint @xmath460 has an infinite binary expansion _ _ ( _ _ and no equivalent finite binary expansion _ _ ) _ _ and @xmath454 is then the initial condition of an orbit of @xmath323 that constitutes the reflected - unfolding of @xmath457 in @xmath323 .
then the sequence of compatible orbits @xmath416 _ _ ( _ _ where @xmath461 _ _ ) _ _ of the prefractal billiards @xmath9 is a sequence of compatible periodic orbits .
[ exa : t - fractalsequenceofcompatibleorbits ] let @xmath462 and @xmath463 .
then , @xmath464 is a nonconstant sequence of compatible periodic orbits ; see figure [ fig : t - fractalsequenceofcompatibleorbits ] . , @xmath465 and @xmath466 , respectively . ]
the following two theorems are ultimately concerning the prefractal billiard @xmath9 . determining which intercepts and slopes yield line segments in the plane that avoid lattice points of the form @xmath467 is equivalent to specifying an initial condition of an orbit of a square billiard table that avoids corners of the billiard table .
then , using the fact that an appropriately scaled square billiard table tiles @xmath9 , we can reflect - unfold such an orbit in @xmath9 in order to determine an orbit of @xmath9 .
[ thm : whattheslopecannotbeintfrac ] let @xmath468 with @xmath469 , @xmath470 and @xmath186 relatively prime , @xmath471 and @xmath472 .
further , let @xmath473 . if for every @xmath474 , @xmath475 , we have that @xmath476 then the line @xmath477 does not contain any point of the form @xmath467 , @xmath478 , with @xmath479 . note that the condition ( [ eqn : whatmcannotbe ] ) above is automatically satisfied if the slope @xmath73 is irrational .
[ thm : whattheslopecanbeintfrac ] let @xmath468 , with @xmath480 , @xmath470 and @xmath186 relatively prime , @xmath471 and @xmath472 .
if @xmath481 with @xmath482 , @xmath483 , then , for every @xmath474 with @xmath475 , the point @xmath484 does not lie on the line @xmath477
. finally , theorems [ thm : whattheslopecannotbeintfrac ] and [ thm : whattheslopecanbeintfrac ] combined with the fact that an initial condition of an orbit of @xmath485 , @xmath358 , determines a sequence of compatible orbits @xmath364 , allows us to determine a countably infinite family of sequences of compatible periodic orbits . for every @xmath225 ,
the interior angles of @xmath451 are @xmath377 and @xmath487 .
to form the associated translation surface @xmath488 , we appropriately identify four copies of @xmath9 ; see figure [ fig : t - fractalflatsurfaces ] for a depiction of the first three translation surfaces . , @xmath489 and @xmath490 associated with the @xmath5-fractal prefractal approximations @xmath491 , @xmath492 and @xmath493 , ] then , every point of @xmath488 associated with a vertex of @xmath9 measuring @xmath377 constitutes a removable singularity of @xmath488 .
similarly , every point of @xmath488 associated with a vertex of @xmath9 of interior angle measuring @xmath487 constitutes a nonremovable singularity of @xmath488 .
therefore , not every vertex of @xmath9 will present a problem for the billiard flow . consider an orbit of @xmath124 , where the orbit has basepoints corresponding to vertices of @xmath247 , the unit square . since such vertices correspond to removable singularities in the corresponding translation surface
( this being the flat torus , see [ subsec : translationstructuresandtranslationsurfaces ] ) , we see that the same orbit reflected - unfolded in the billiard @xmath9 ( if one first scales the billiard @xmath124 and the orbit contained therein by @xmath494 , see [ subsec : unfoldingabilliardorbit ] ) can potentially intersect vertices of @xmath451 that are associated with nonremovable singularities in the corresponding translation surface
. let @xmath14 be a self - similar sierpinski carpet , as defined in definition [ def : aselfsimilarsierpinskicarpet ] , and let us denote its natural prefractal approximations by @xmath495 for @xmath496 ( as in [ subsec : asierpinskicarpet ] ) .
the corresponding billiard is then denoted by @xmath7 . in this subsection , we examine the behavior of the billiard flow on the rational polygonal billiard given by the prefractal approximations @xmath497 . in the event
a billiard ball collides with a corner of a peripheral square , we must terminate the flow and such a trajectory is then called _
singular_. in addition to being singular , such a trajectory will form a saddle connection ( see the beginning of [ sec : rationalbilliards ] for a discussion of closed billiard orbits that form saddle connections ) .
as we have discussed , an examination of the corresponding translation surface may prove useful in determining whether or not a billiard ball can reflect in a vertex .
[ def : obstacletobilliardflow ] let @xmath1 be a polygonal billiard
. then @xmath1 can be modified by placing in its interior a piecewise smooth segment that inhibits the billiard flow and causes a billiard ball to reflect .
such a segment is called an _ obstacle _ of @xmath1 .
clearly , each prefractal billiard @xmath497 can be interpreted as a _ square billiard with obstacles_. [ nota : alphafortheta ] due to the fact that theorem [ thm : asetbset ] refers to the slope of a nontrivial line segment and we make heavy use of this theorem , we will denote the initial condition @xmath69 of an orbit of @xmath10 by @xmath498 , where @xmath499 .
consider the boundary of a cell @xmath500 of @xmath501 as a barrier .
is a cell of the @xmath65th prefractal approximation @xmath502 , as given in definition [ def : acellofsai ] with all numbers @xmath503 equal to @xmath16 . ]
then an orbit with an initial condition contained in the cell is called an _ orbit of the cell @xmath500 of @xmath501_. so as to be clear , the boundary of the cell does not form an obstacle to the billiard flow , as defined in definition [ def : obstacletobilliardflow ] .
rather , we are treating the cell @xmath500 as a billiard table in its own right , embedded in the larger prefractal approximation @xmath501 .
recall from
[ subsec : asierpinskicarpet ] that a self - similar sierpinski carpet @xmath291 is the unique fixed point attractor of a suitably chosen iterated function system @xmath504 consisting of similarity contractions . in light of this ,
an orbit of a cell @xmath500 of @xmath501 is the image of an orbit @xmath505 of the unit - square billiard @xmath506 under the action of a composition of contraction mappings @xmath507 , with @xmath508 and @xmath509 , determined from the iterated function system @xmath510 of which @xmath291 is the unique fixed point attractor .
[ lem : segmentbeginningatmidpointisnontrivialifalphafromb ] consider a self - similar sierpinski carpet @xmath291 .
let @xmath57 and @xmath511 be a prefractal approximation of @xmath291 . if @xmath512 , that @xmath305 is the set of slopes given by equation ( [ eqn : bset ] ) .
] then the line segment beginning at a midpoint of a cell @xmath500 of @xmath511 is a nontrivial line segment _ _ ( _ _ in the sense of definition [ def : nontriviallinesegmentofsa]_)_. moreover , such a segment avoids the boundary of the peripheral squares of @xmath291 with side - length @xmath513 , @xmath514 .
the statement in lemma [ lem : segmentbeginningatmidpointisnontrivialifalphafromb ] asserts that a segment beginning at a midpoint of a cell with slope @xmath512 will be a nontrivial line segment in @xmath291 .
in addition to this , any line segment contained in @xmath100 that contains a nontrivial line segment of @xmath291 must necessarily avoid the peripheral squares in a tiling of @xmath100 by @xmath291 .
otherwise , there exists @xmath515 such that scaling the line segment in @xmath100 and the tiling of @xmath100 by @xmath516 results in a segment contained in the nontrivial line segment which intersects peripheral squares of @xmath291 .
this is a contradiction of the fact that the segment beginning at @xmath517 with slope @xmath512 is a nontrivial line segment of @xmath291 .
we then deduce the following result .
[ thm : constantsequenceofcompatibleorbitsinsa ] consider a self - similar sierpinski carpet @xmath291 .
let @xmath57 and @xmath511 be a prefractal approximation of @xmath291 .
furthermore , let @xmath512 and @xmath518 with @xmath519 a positive , odd integer .
if @xmath520 is an orbit of @xmath501 , then the initial condition @xmath521 determines a sequence of compatible periodic orbits @xmath522 of the prefractal approximations @xmath10 .
as one may suspect , there exists @xmath523 such that a sequence of compatible orbits @xmath524 is a constant sequence of compatible orbits .
moreover , @xmath525 , for every @xmath383 .
this is not any different from the case of a constant sequence of compatible orbits of prefractal billiards @xmath8 , as discussed in theorem [ thm : sufficientconditionforcantororbit ] and example [ exa : aconstantsequenceofcompatibleperiodichybridorbits ] .
however , in the context of a self - similar sierpinski carpet billiard table , every sequence of compatible orbits we will examine will be a sequence for which there exists @xmath358 such that @xmath524 is a constant sequence of compatible orbits . in much the same way the billiard @xmath497 can be interpreted as a square billiard with obstacles
, the corresponding translation surface can be interpreted as a `` torus with obstacles '' ; see figure [ fig : squaretoruswithobstacles ] . as a flat torus with obstacles . ] in light of the fact that @xmath527 can be interpreted as a torus with obstacles and the presence of a dynamical equivalence between the billiard flow and the geodesic flow on the corresponding translation surface ( see [ subsec : unfoldingabilliardorbit ] ) , we see that reflection in the vertices with angles measuring @xmath377 ( relative to the interior ) can be defined .
more specifically , the geodesic flow can be defined at points corresponding to vertices with angles measuring @xmath377 , because these points constitute removable singularities of the geodesic flow .
this fact is crucial in determining orbits of @xmath7 for which the slope @xmath295 is an element of @xmath301 and not @xmath305 ( see notation [ not : aabaabbb ] ) , and the orbit avoids all peripheral squares of @xmath7 . while one may say that this contradicts part of theorem [ thm : asetbset ] ( and he / she would be right ) , in @xcite a more precise formulation of theorem [ thm : asetbset ] is given that clarifies which slopes are permissible and which ones are not . that is , if @xmath528 , it may be possible for an orbit @xmath529 to begin at the origin and avoid the peripheral squares of each billiard @xmath530 , for every @xmath531 .
we do not give here an explicit reformulation of theorem [ thm : asetbset ] , but example [ exa : orbitexamplecontradictingtysonstheoremslope2 - 3origins7 ] in [ subsec : aself - similarsierpinskicarpetbilliard ] exhibits a situation showing that the latter half of theorem [ thm : asetbset ] is not stated precisely enough .
the theme that will tie together all of the examples in [ sec : prefractalrationalbilliards ] is that suitable limits of sequences of compatible orbits may constitute billiard orbits of each respective fractal billiard table .
we have shown that in the case of @xmath4 , @xmath6 and @xmath7 , we can determine a sequence of compatible periodic orbits .
we will see that in each case of a fractal billiard , under certain conditions , a sequence of compatible periodic orbits ( or a proper subset of points from each footprint @xmath532 ) will converge to a set which can be thought of as a true orbit of a fractal billiard table ( or such a sequence will yield a subsequence of basepoints converging to what we are calling an _ elusive point _ in @xcite ) .
we restrict our attention to the family of fractal billiard tables @xmath3 where @xmath2 is a fractal approximated by a suitable sequence of rational polygons @xmath316 , with each @xmath164 tiled by @xmath534 for suitably chosen @xmath535 $ ] and @xmath321 a polygon that tiles the plane . specifically , we are interested in developing a general framework for dealing with a fractal billiard table @xmath317 which is similar to that of @xmath4 , @xmath6 and @xmath533 . before we begin our discussion of the fractal billiard tables @xmath4 , @xmath6 and @xmath7
, we define certain terms .
the following definitions were initially motivated by the work in @xcite , but later generalized for this paper in order to account for a larger class of fractal billiard tables .
( from now on , we assume that @xmath317 is a fractal billiard table with prefractal billiard approximations @xmath536 as described just above . ) [ def : acorner ] let @xmath537 .
if there exists @xmath225 such that @xmath538 and @xmath408 is a vertex of @xmath164 , then @xmath408 is called a _ corner _ of @xmath2 .
[ def : acantorpoint ] let @xmath537 be such that @xmath408 is not a corner of @xmath2 .
if there exists @xmath358 such that for every @xmath383 , @xmath538 and every connected neighborhood of @xmath408 contained in @xmath164 becomes totally disconnected when intersected with @xmath2 , then @xmath408 is called a _ cantor point _ of @xmath2 . in the koch snowflake @xmath12 ,
every cantor point is a smooth point of infinitely many prefractals @xmath229 approximating @xmath12 .
that is , if @xmath408 is a cantor point in @xmath12 , then there exists @xmath539 such that for every @xmath383 , there exists a well - defined tangent at @xmath540 .
below , @xmath408 is viewed as a point of the smooth subarc of @xmath164 to which it belongs .
] we deduce from this that the law of reflection holds at @xmath540 , for every @xmath383 .
moreover , since the billiard ball reflects at @xmath540 at the same angle for every @xmath383 , we deduce that the tangent at @xmath408 is the same for each @xmath229 , @xmath383 .
this observation then prompts us to generalize the definition of a cantor point in order to account ( for example ) for points of the @xmath5-fractal which are not cantor points , but are points for which a well - defined tangent can be found in infinitely many prefractal approximations .
[ def : smoothfractalpoint ] let @xmath537 and @xmath358 be such that @xmath538 for every @xmath383 . if there exists a well - defined tangent at @xmath538 for every @xmath383 , then @xmath408 is called a _ smooth fractal point_. to be clear , a cantor point of @xmath2 is an example of a smooth fractal point of @xmath2 . the special nature of a cantor point warrants a formal definition . in the @xmath5-fractal billiard , there are certainly corners and elusive points .
there are also smooth connected segments contained in the boundary of @xmath6 .
points contained in such segments that do not correspond to corners are then called smooth fractal points .
[ def : anelusivepoint ] let @xmath537 .
if @xmath541 , then @xmath408 is called an _ elusive point _ of @xmath2 . consider a piecewise linear path in @xmath3 , such that every linear segment of the path is joined at the endpoint of another segment with the coincidental endpoints intersecting the boundary @xmath2 at a smooth fractal point of @xmath2 ( in the sense of definition [ def : smoothfractalpoint ] ) . in the following definition ,
we define a particular type of piecewise linear curve in a fractal billiard @xmath3 .
suppose that there exists a piecewise linear curve in @xmath3 as described immediately above .
if at each point @xmath408 for which the piecewise linear path intersects the boundary @xmath2 , the angle formed by the first segment is equal to the angle formed by the second segment , relative to the side of @xmath164 on which @xmath408 lies , that a smooth fractal point @xmath408 of @xmath2 is necessarily a point of infinitely many prefractal approximations @xmath542 . hence , there is a least nonnegative integer @xmath61 such that @xmath538 . ]
then the piecewise linear path is called a _ nontrivial path _ of @xmath3 . in @xcite , a nontrivial path
was called a _ nontrivial polygonal path_. the change in name is purely based on aesthetics .
[ def : acantororbit ] suppose @xmath455 is an orbit of @xmath543 , for some @xmath358 , such that every point of the footprint @xmath544 corresponds to a smooth fractal point of @xmath2 .
this then readily implies that @xmath325 is the same as @xmath455 for every @xmath383 .
is a constant sequence of compatible orbits , where a _ sequence of compatible orbits _ was defined in definition [ def : sequenceofcompatibleorbits ] . ]
then @xmath325 is called a _ cantor orbit _ of @xmath3 and is denoted by @xmath83 . if @xmath3 is a fractal billiard table , then it may or may not be possible to construct cantor orbits or nontrivial paths of @xmath317 .
we will next discuss three examples of fractal billiard tables with different dynamical properties that lend themselves well ( or not ) to determining well - defined billiard orbits .
we note that applying definitions [ def : acorner ] , [ def : acantorpoint ] and [ def : anelusivepoint ] to @xmath4 and the sequence of rational polygon prefractal approximations @xmath8 which we have discussed in [ subsec : thekochcurveandkochsnowflake ] yields exactly the sets of points we are considering as corners , cantor points and elusive points of @xmath4 , respectively . moreover , applying definitions [ def : acorner ] , [ def : smoothfractalpoint ] and [ def : anelusivepoint ] to @xmath6 and the prefractal approximations @xmath9 which we discussed in [ subsec : thetfractal ] yields exactly the sets of points that we are considering as corners , smooth fractal points and elusive points of @xmath6 . finally , applying definitions [ def : acorner ] and [ def : smoothfractalpoint ] to @xmath533 and the prefractal approximations @xmath10 which we discussed in [ subsec : asierpinskicarpet ] yields exactly the set of points we are considering as corners and smooth fractal points of @xmath533 .
as we have noted before at the end of [ subsec : thekochcurveandkochsnowflake ] , for each @xmath225 , @xmath545 can be realized as the union of @xmath234 self - similar ternary cantor sets , each spanning a distance of @xmath235 . within each cantor set
, we find cantor points and corners of the koch snowflake .
we begin our discussion of orbits of @xmath4 by examining the limiting behavior of a particular sequence of compatible orbits with the initial condition @xmath546 , where @xmath547 is a cantor point of @xmath12 ( i.e. , @xmath547 is a point of @xmath548 with a well - defined tangent in @xmath229 for every @xmath383 ) . for the sake of simplicity ,
we let @xmath549 and @xmath550 be on the base of the equilateral triangle @xmath232 ( recall that we are assuming that the left corner of @xmath232 is at the origin and the length of each side is one unit ) .
then , @xmath330 is an orbit that remains fixed as one constructs @xmath333 from @xmath331 .
more correctly , @xmath551 is a sequence of compatible orbits with @xmath552 for every @xmath225 ( that is , with the same footprint in each prefractal approximation ) .
in general , if @xmath546 is an initial condition of an orbit of @xmath382 and @xmath547 is a cantor point , then the sequence of compatible orbits is such that for every @xmath383 , the footprints @xmath553 and @xmath554 are the same .
if @xmath555 is a cantor point , then there exists a well - defined orbit of @xmath4 with an initial condition @xmath556 , where the angle @xmath557 is determined with respect to the side on which @xmath31 lies in a prefractal approximation @xmath8 .
there are many more properties of @xmath558 which we could discuss here .
these properties largely rely on the nature of the ternary representation of @xmath547 , and are elaborated upon in @xcite .
we now proceed to illustrate how we can connect two elusive points of @xmath4 .
such a result has already been presented in greater detail in @xcite , so we will be brief . in
[ subsec : thetfractalbilliard ] , we will show that an identical construction holds for the billiard table @xmath6 .
recall from example [ exa : asequenceofcompatibleperiodichybridorbits ] that we were able to construct a sequence of compatible periodic hybrid orbits .
from such a sequence we can derive a sequence of basepoints that is converging to an elusive point of @xmath4 .
the latter sequence of basepoints constitutes the vertices of a nontrivial path ; see figure [ fig : nontrivialpath ] . one may consider a direction @xmath559 that is the reflection of @xmath423 through the normal at @xmath378 .
then , the resulting sequence of compatible periodic hybrid orbits @xmath560 yields a sequence of basepoints converging to another elusive point . again , such a sequence of basepoints constitutes the vertices of a nontrivial path of @xmath4 ; see figure [ fig : twonontrivialpaths ] .
together , these two nontrivial paths constitute a single nontrivial path connecting two elusive points of @xmath4 .
beginning at @xmath561 . ] .
( as is explained in the text , these two paths can be concatenated to obtain a single nontrivial path connecting the two elusive points . ) in the first figure , we only show the relevant portions of the koch snowflake . in the second figure
, we magnify the regions containing the nontrivial paths so as to highlight the fact that such paths are converging to elusive points .
actually , there is an obvious geometric similarity one can take advantage of in order to produce more segments of the nontrivial path . ] in conjunction with theorem [ thm : bodd ] , we can determine countably infinitely many initial conditions @xmath69 , each of which determines a sequence of compatible periodic hybrid orbits yielding a sequence of basepoints converging to an elusive point of @xmath4 .
the results in this subsection appear here for the first time and will be further discussed in @xcite .
we begin our discussion of the billiard @xmath6 by recalling ( and referring the reader back to ) example [ exa : t - fractalsequenceofcompatibleorbits ] from [ subsec : thetfractalprefractalbilliard ] .
the sequence of compatible periodic orbits provided by example [ exa : t - fractalsequenceofcompatibleorbits ] gives rise to a nontrivial path that connects @xmath562 with an elusive point of @xmath6 .
furthermore , considering the sequence of compatible periodic orbits @xmath563 , we determine another nontrivial path that connects @xmath562 with another elusive point of @xmath6 ; see figure [ fig : t - fractalnontrivialpaths ] .
this behavior is analogous to the one which we observed for the koch snowflake billiard in
[ subsec : thekochsnowflakefractalbilliard ] . .
] as was the case with @xmath4 , we can analogously build upon theorems [ thm : whattheslopecannotbeintfrac ] and [ thm : whattheslopecanbeintfrac ] in order to determine a sequence of basepoints converging to an elusive point .
that is , theorems [ thm : whattheslopecannotbeintfrac ] and [ thm : whattheslopecanbeintfrac ] guide our search for a sequence of compatible periodic orbits which yields a sequence of basepoints converging to an elusive point of @xmath6 .
let @xmath364 be a sequence of compatible orbits .
then , there are countably infinitely many directions and countably infinitely many points from which to choose so that @xmath364 is a sequence of compatible periodic orbits yielding a sequence of basepoints @xmath564 that converges to an elusive point of @xmath6 .
the collection of basepoints @xmath564 constitute the vertices of a nontrivial path of @xmath6 . moreover , once such a nontrivial path is constructed , letting @xmath565 , an additional nontrivial path can be determined from a sequence of compatible periodic orbits @xmath566 in exactly the same fashion . in @xcite ,
nontrivial line segments of sierpinski carpets are constructed .
building on the main results of @xcite , the second author and joe p. chen have been able to construct a family of cantor periodic orbits of a self - similar sierpinski carpet , in the sense of @xcite recalled in definition [ def : acantororbit ] .
such orbits constitute cantor orbits of the self - similar sierpinski carpet . as of
yet , we have not attempted to construct a nontrivial path of a sierpinski carpet . in light of theorem
[ thm : constantsequenceofcompatibleorbitsinsa ] , we say that the trivial limit of a constant sequence of compatible periodic orbits constitutes a periodic orbit of a self - similar sierpinski carpet billiard @xmath7 . in the event
an orbit has an initial direction @xmath567 , we may still be able to determine a constant sequence of compatible periodic orbits .
the trivial limit of such a sequence then constitutes a periodic orbit of @xmath7 . using the fact that reflection can be defined in the vertices with interior angles measuring @xmath377
, we can state the following result .
( recall from
[ subsec : aprefractalselfsimilarsierpinskicarpetbilliard ] that @xmath568 is the @xmath61th prefractal approximation of @xmath291 . )
[ thm : alphainaformsorbitofsa ] recall from notation [ nota : alphafortheta ] that if @xmath329 is the initial direction of a billiard orbit , then @xmath569 .
let @xmath570 , @xmath571 and let @xmath572 be an orbit of @xmath506 . if @xmath573 , as an orbit of @xmath574 , avoids the middle peripheral square , then the initial condition @xmath575 will determine an orbit of @xmath576 . specifically
, the path traversed by the orbit @xmath573 of @xmath574 is exactly the path traversed by the orbit of @xmath7 determined by @xmath575 .
[ exa : orbitexamplecontradictingtysonstheoremslope2 - 3origins7 ] let @xmath570 , @xmath577 .
consider an orbit of @xmath578 with an initial condition @xmath575 ; see figure [ fig:2 - 3orbitstartingfrom1 - 2 ] .
we see that the orbit avoids the peripheral square of @xmath579 . by theorem [ thm : alphainaformsorbitofsa ]
, the initial condition @xmath575 determines an orbit of @xmath580 .
the path traversed by the orbit of @xmath580 is exactly the path traversed by the orbit @xmath573 . and with an initial direction constituting a slope of @xmath581 , where @xmath582 is defined as in notation [ not : aabaabbb ] .
while it would appear that this orbit intersects corners of peripheral squares , it in fact remains away from all peripheral squares .
the same is true for finer approximations . ]
it is clear from the preceding sections that much work remains to be developed in order to determine a well - defined phase space @xmath583 for the yet to be defined fractal billiard flow .
we have discussed several examples of what clearly constitute periodic orbits of @xmath4 and @xmath7 .
furthermore , for both @xmath4 and @xmath6 , we were able to connect two elusive points of each billiard table via suitably chosen nontrivial paths .
these nontrivial paths were determined from suitably chosen sequences of compatible periodic orbits .
[ ques : determineadditionalnontrivialpaths ] let @xmath2 be either @xmath12 or @xmath13 .
suppose that two suitably chosen nontrivial paths converge to two distinct elusive points of @xmath3 .
for each of the two elusive points , is it possible to determine another nontrivial path converging to a different elusive point ? if we can answer question [ ques : determineadditionalnontrivialpaths ] in the affirmative ( or answer it in the affirmative under specific conditions ) , will this help us gain insight into how to determine a well - defined phase space for the billiard flow on @xmath3 ?
an alternate approach , discussed in the concluding remarks of @xcite , entails determining a well - defined fractal translation surface .
following this line of thought to its logical end , for certain fractal billiard tables ( e.g. , @xmath6 ) , is it possible to determine which directions produce recurrent orbits ? more generally , can one prove that , in almost every direction , the billiard flow is ergodic in @xmath317 ? regarding a self - similar sierpinski carpet billiard @xmath533 , we have determined a countable set of points from which a periodic billiard orbit can begin .
can we show that the set of points from which a periodic orbit can begin is in fact uncountable and , furthermore , a set of full _ _ ( _ _ lebesgue _ ) _ measure in the base of the unit square @xmath257 ?
it is possible to construct a nontrivial line segment of @xmath291 beginning from @xmath584 with slope @xmath585 , that , when translated to @xmath586 , no longer lies entirely in @xmath14 .
however , if we consider the sequence of compatible periodic orbits given by @xmath587 , is it possible to determine a well - defined limit ?
the work of @xcite may prove useful in further exploring the behavior of a sequence of compatible periodic orbits .
building on the work of @xcite , the author of @xcite has examined the behavior of nonperiodic orbits in what is an example of what is called a _ wind - tree billiard _ , and what is also strongly suggestive of a sierpinski carpet .
such work may provide insight into examining the behavior of a sequence of compatible dense orbits .
[ ques : determiningawelldefinedbilliardflow ] in analogy with the prefractal billiard and associated translation surface , can a thorough understanding of the geodesic flow on the limiting _ _ ( _ _ and still to be mathematically defined _
_ ) _ _ _ _ _ _ fractal translation surface__ _ _ @xmath588 aid us in determining a well - defined billiard flow on @xmath3 ?
approaching the problem of determining a well - defined billiard flow on a fractal billiard table from many different points of view may prove useful .
the theories of translation surfaces and rational billiards are intimately tied together and more deeply understood by knowing the structure of what is called the _ veech group _ ( this being the group studied in , for example , @xcite ) . in short ,
the veech group of a translation surface @xmath84 determined from a rational polygon @xmath0 is the stabilizer of @xmath84 .
let @xmath317 be a fractal billiard table , with @xmath2 being approximated by a suitably chosen sequence of rational polygons @xmath316 .
is it then possible to construct a veech group for @xmath317 _ _ ( _ _ or rather , of @xmath588 _ _ ) _ _ , presumably in terms of the veech groups for the prefractal approximations @xmath62 _ _ ( _ _ or rather , of the associated translation surfaces @xmath589 _ _ ) _ _ ?
will the knowledge of such a group aid us in determining a well - defined billiard flow on @xmath317 ?
barnsley , m. f. : _ superfractals _ : _ patterns of nature _ , cambridge univ . press , new york , 2006 .
chen , j. p. , niemeyer , r. g. : periodic billiard orbits of self - similar sierpinski carpets , 29 pages , e - print , arxiv:1303.4032v1 , 2013 .
( to appear in the j. of math .
anal . and appl .
) lapidus , m. l. , niemeyer , r. g. : towards the koch snowflake fractal billiard computer experiments and mathematical conjectures , in : _ gems in experimental mathematics _ ( t. amdeberhan , l. a. medina and v. h. moll , eds . ) , contemporary mathematics , amer . math . soc . ,
providence , ri , * 517 * ( 2010 ) , pp . 231263 .
[ e - print : arxiv : math.ds.0912.3948v1 , 2009 . ]
lapidus , m. l. , niemeyer , r. g. : sequences of compatible periodic hybrid orbits of prefractal koch snowflake billiards , _ discrete and continuous dynamical systems ser .
a _ , in press , 2012 .
[ e - print : ihes / m/12/16 , 2012 ; arxiv:1204.3133v1 [ math.ds ] , 2012 . ] smillie , j. : dynamics of billiard flow in rational polygons , in : _ dynamical systems _ , encyclopedia of math .
sciences , vol .
100 , math .
physics 1 ( ya .
g. sinai , ed . ) , springer - verlag , new york , 2000 , pp . | if @xmath0 is a rational polygon , then the associated rational billiard table is given by @xmath1 .
such a billiard table is well understood .
if @xmath2 is a closed fractal curve approximated by a sequence of rational polygons , then the corresponding fractal billiard table is denoted by @xmath3 . in this paper
, we survey many of the results from [ * lapnie1 - 3 * ] for the koch snowflake fractal billiard @xmath4 and announce new results on two other fractal billiard tables , namely , the @xmath5-fractal billiard table @xmath6 ( see @xcite ) and a self - similar sierpinski carpet billiard table @xmath7 ( see @xcite ) . we build a general framework within which to analyze what we call a sequence of compatible orbits .
properties of particular sequences of compatible orbits are discussed for each prefractal billiard @xmath8 , @xmath9 and @xmath10 , for @xmath11 . in each case
, we are able to determine a particular limiting behavior for an appropriately formulated sequence of compatible orbits .
such a limit either constitutes what we call a nontrivial path of a fractal billiard table @xmath3 or else a periodic orbit of @xmath3 with finite period . in our examples ,
@xmath2 will be either @xmath12 , @xmath13 or @xmath14 .
several of the results and examples discussed in this paper are presented for the first time .
we then close with a brief discussion of open problems and directions for further research in the emerging field of fractal billiards . |
our motivation for gaining theoretical understanding of polymer thin film growth stems from their technological applications in microelectronic interconnects @xcite , organic electronics @xcite , and biotechnology .
various experimental methods like vapor deposition polymerization ( vdp ) @xcite , ionization assisted polymer deposition @xcite , sputtering growth @xcite , pulsed laser deposition @xcite , and organic molecular beam deposition @xcite have been developed to produce a variety of polymer thin films .
polymer film growth is complex compared to the conventional inorganic thin film growth process due to polymer s complicated structure and interactions that include internal degrees of freedom , limited bonding sites , chain - chain interactions , etc .
many experimental efforts have focussed on the formation of polymer thin films using vdp @xcite . in a typical vdp experiment ,
a wafer ( 2-d substrate ) is exposed to one or more volatile gas phase precursors that produce free - monomers .
the free - monomers impinge on the substrate at random locations and react on the substrate surface to produce the desired deposit .
polymer thin films grown by vdp are made up of long polymer chains formed through the polymerization reaction occurring during the growth process .
the polymerization process involves the interaction of two free - monomer molecules in a chemical reaction to initiate a dimer ( polymer chain of length = 2 ) .
the free - monomers moving towards the substrate are consumed by either of the two processes : first being chain initiation in which new polymer molecules are generated ; and secondly , chain propagation in which the existing polymer molecules are extended to higher molecular weight . besides these two mechanisms , the free - monomer adsorption , diffusion , and polymer merger can be considered as other mechanisms that determine the overall film morphology . the chemical nature of the linear polymer chain restricts the number of bonding sites .
a free - monomer can only bond to either of the two active ends of a polymer , or to another free - monomer .
this bonding constraint leads to the formation of an entangled or an overhang configuration , which blocks the region it covers from the access of other incoming free - monomers . in conventional physical vapor deposition ( pvd )
processes @xcite , atoms can nucleate at the nearest neighbors of the nucleated sites and atomistic processes such as surface diffusion , edge diffusion , step barrier effect , etc .
effect the growth , resulting in the films being compact and dense@xcite .
recent investigations by zhao et al .
@xcite have shown that the submonolayer growth behavior of vdp is very different from that of pvd due to long chain confinement and limited bonding sites , indicating that the detailed molecular configuration can drastically change the growth behavior @xcite . in experiments ,
the growth behavior of polymer thin films have been investigated through their morphological evolution study .
the vdp processes for producing parylene - n ( pa - n ) films typically are far from equilibrium .
the precursor material di - p - xylylene ( dimer ) is sublimed at @xmath21c and then pyrolized into free - monomers at @xmath22c .
the free - monomers impinge at random angles onto the si - wafer at room temperature and eventually condense and polymerize to form the polymer film . by varying the growth rates in the pa - n growth experiments , zhao et al .
@xcite reported an average roughness exponent @xmath23 and an average growth exponent @xmath24 .
however , by considering the tip effect of the atomic force microscope @xcite , the range of @xmath25 was estimated to be between @xmath26 to @xmath27 and the authors found the absence of dynamic scaling hypothesis in the pa - n film growth . in the recent experiments done by lee et al .
@xcite , the authors observed unusual changes in the roughening behavior during the poly(chloro - p - xylylene ) growth . in the early rapid growth regime , they observed @xmath28 ( larger than the random deposition @xmath29 ) and upon complete coverage of the substrate ( around @xmath30 ) , they found @xmath31 and the interface width did not evolve with the film thickness . finally , during the continuous growth regime , the surface roughness again was found to increase steadily with a new power law of @xmath32 @xcite , which is close to the results of zhao et al .
@xcite . of the known theoretical results for dynamic roughening ,
the mbe nonlinear surface diffusion dynamics proposed by lai and das sarma @xcite predicted similar exponents as obtained in the experimental study of zhao .
however , their nonlinear surface diffusion theory could not explain the findings of varying local slope in the experiments .
zhao et al .
proposed a stochastic growth model based on bulk diffusion @xcite , which correctly predicted the kinetic roughening phenomena observed in their experiment @xcite but lacks the details on how polymers evolve .
insufficient theoretical studies coupled with inconsistencies in the experiments motivate us to model the polymer film growth and seek a better understanding of the growth processes that determine the roughening mechanism . in this paper , we study the 1 + 1d lattice model for the polymer films grown by vdp and examine the effects of random angle deposition , free - monomer diffusion , free - monomer adsorption in determining the evolution of the film s morphology .
schematic of the 1 + 1d growth model .
1(a ) : free - monomer deposition at random angles ; 2(a ) : adsorbed free - monomer diffuses along along the substrate ; 2(b ) : adsorbed free - monomer diffuses on a polymer chain ; 3(a , b ) : polymer chain initiation resulting from random angle deposition ; 3(c , d ) : polymer chain initiation resulting from free - monomer diffusion on substrate and polymer chain respectively ; 4(a , b ) : free - monomer deposits onto the active end of a polymer chain resulting in chain propagation ; 4(c ) chain propagation due to free - monomer diffusion . ]
in our simulation , the free - monomers were deposited at random angles on a 1-d substrate of lattice size @xmath33 at a deposition rate @xmath0 ( in units of monomers per site per unit time ) . the kiss random number generator @xcite was employed and one deposition time unit corresponded to the deposition of @xmath33 free - monomers . the incident free - monomers were released from a height three lattice units above the highest point on the surface with an initial abscissa randomly chosen from @xmath34 .
the depositing free - monomers have a uniform `` launch angle '' distribution which corresponds to a nonuniform flux distribution @xmath35 of particles above the surface , where @xmath36 is defined as the angle between the direction of impinging monomer and the substrate normal @xcite .
our vdp growth model is similar to the square - lattice disk model studied by ref .
@xcite along with additional constraints of free - monomer diffusion , limited bonding , polymer - initiation , propagation and merger .
the particles followed a ballistic trajectory until contacting the surface .
the impinging free - monomer was then moved to the lattice position nearest to the point of contact .
the deposited free - monomers were allowed to diffuse via nearest - neighbor hops with a diffusion rate @xmath1 ( nearest neighbor hops per monomer per unit time ) . a free - monomer that is deposited on top of an existing polymer chain gets adsorbed on the chain and is also allowed to diffuse .
the excluded volume constraint was implemented by rejecting the diffusion or deposition moves to an already occupied site . as the monomer coverage increases on the substrate , the polymer film grows along a direction perpendicular to the substrate .
this `` two dimensional '' growth is often referred to as the 1 + 1d growth in literature .
figure [ fig_1_schematic ] shows the schematic of various processes that occur during the non - equilibrium film growth on a 1-d substrate of length @xmath33 with periodic boundary conditions .
processes 1(a ) , 3(a ) , 3(b ) , 4(a ) , and 4(b ) show the gas phase free - monomers depositing onto the substrate at random locations with uniform launch angle distribution .
these free - monomers get adsorbed either on the substrate ( shown by process 2(a ) ) or on the polymer chains ( process 2(b ) ) .
adsorbed free - monomers are allowed to diffuse along the adsorbent to any of the nearest - neighboring unoccupied sites with equal probability , the rate of diffusion @xmath1 is assumed to be the same on both the substrate and the polymer chain .
we define the chain length @xmath37 of a polymer as the number of monomers forming a polymer chain .
when an impinging free - monomer encounters another free - monomer on the substrate as its nearest neighbor ( process 3(a ) ) , both are frozen and undergo a chemical reaction to form a dimer ( polymer chain of length @xmath38 ) ; polymers with chain length @xmath39 can also be formed after deposition ( process 3(b ) ) .
free - monomer diffusion along the substrate ( process 3(c ) ) and along the polymer chain ( process 3(d ) ) can also result in the polymer chain initiation @xcite .
when an impinging free - monomer encounters the active end of a polymer chain , it attaches itself to the chain and increases the chain length by one unit ( 4(a ) , 4(b ) ) .
a diffusing free - monomer can meet an active end of a polymer chain in its neighborhood and get bonded to that polymer chain ( process 4(c ) ) . in linear polymer systems ,
the free - monomers are allowed to form a maximum of two chemical bonds and at any given time only the two ends of the polymer chain are chemically active , resulting in the chain propagation at these two end locations . the chain portion ( of the polymer ) excluding the two chemically active ends , is not allowed to form chemical bonds with neighboring free - monomers .
free - monomers can however be physically adsorbed on the chain and can diffuse along the chain ( process 2(b ) ) .
another interesting process that occurs during the film growth is the _ polymer - merger_. two different polymers can merge when their respective active - ends meet as nearest neighbors . during the polymer - merger ,
the nearest neighboring ends of merging polymers react chemically to join the two polymers into one longer polymer chain with higher molecular weight .
the resulting polymer chain is left with two active - ends , one from each of the parent polymers . in the case
when both the active - ends belonging to the same chain appear as nearest neighbors , the chemical bond between the ends is prohibited . in this study
we do not attempt to study the effects of re - emission which allows the free - monomers to `` bounce around '' before they settle at appropriate sites on the surface .
instead , we assume that the impinging free - monomers will always stick to the particle that comes on its deposition path ( processes 4(a ) , 4(b ) ) . at each stage of the simulation , either a deposition or a diffusion
is performed . in order to keep track of the competing rates of diffusion and deposition we adapted the method suggested by amar et al.@xcite and carried out the deposition with a probability @xmath40 , @xmath41 } , \ ] ] where @xmath42 is the free - monomer density ( per site ) and @xmath2 .
the diffusion was carried out with the probability @xmath43 , @xmath44 } .\ ] ] in our simulations the incoming free - monomer flux @xmath0 was fixed for different @xmath1 , thus an increase in @xmath1 was parametrized as an increase in @xmath4 . throughout the growth process the list of all free - monomers and polymer chains were continually updated .
if a free - monomer encountered another free - monomer or an active end of a polymer as its nearest neighbor , it was added to the polymer chain and removed from the free - monomer list . in cases where a free - monomer was the nearest neighbor to the active ends of more than two polymers ,
we selected a random pair of polymers and performed polymer - merger .
in fig . [ fig_2_six_eps_figs_grey ] we show typical snapshots of the polymer films generated using @xmath33 = @xmath45 substrate for two extreme cases : @xmath4 = @xmath5 ( figs .
[ fig_2_six_eps_figs_grey]a , b , c ) and @xmath4 = @xmath46 ( figs .
[ fig_2_six_eps_figs_grey]d , e , f ) after a deposition time of @xmath20 = 15 ( figs .
[ fig_2_six_eps_figs_grey]a , d ) , @xmath20 = 30 ( figs .
[ fig_2_six_eps_figs_grey]b , e ) , and @xmath20 = 60 ( figs .
[ fig_2_six_eps_figs_grey]c , f ) respectively . for both the values of @xmath4 ,
the films show the presence of columnar structures , overhangs , and unoccupied regions .
these structures were observed to persist throughout the growth process .
presence of these morphological structures can be explained by the shadowing effect inherent in the growth process and is attributed to the cos@xmath47 distribution of the impinging free - monomer flux @xcite .
shadowing effects arise when the columnar structures of the surface `` stick out '' and shadow their neighboring sites , thus inhibiting the growth in their neighboring sites . due to the angular flux distribution of the impinging free - monomers the taller surface features
prevent the incoming flux from entering the lower lying areas of the surface . for comparable deposition times
@xmath20 , the films grown at @xmath48 ( figs .
[ fig_2_six_eps_figs_grey]a , b , c ) are characterized by small unoccupied - regions and short polymer chains , resulting in shorter , denser , and compact films .
whereas for @xmath4 = @xmath46 ( figs .
[ fig_2_six_eps_figs_grey]d , e , f ) , the films are characterized by large unoccupied - regions and longer polymer chains resulting in taller , more porous , and less dense films . for a relatively low diffusion rate at @xmath4 = 10 ,
the free - monomers deposited on the film have a higher probability of encountering another impinging free - monomer as nearest neighbor and thereby initiating new polymers .
many such polymer initiations inhibit the occurrence of unoccupied - regions and make the film dense and compact .
in contrast , at a higher diffusion rate of @xmath4 = @xmath46 , the free - monomers have a higher probability to diffuse upward towards the growth front and the upward diffusion of free - monomers is favored due to the non - symmetric nature of the lattice potential associated with diffusion over a step @xcite .
the diffusing free - monomers arrive towards the growth front and bind to the active ends of the polymers and increase their chain length .
this explains the occurrences of longer polymer chains at higher @xmath4 observed in fig .
[ fig_2_six_eps_figs_grey ] ( the longest polymer chains are highlighted in black ) . throughout the growth process the longest chains for @xmath4 = @xmath46 are much longer than those obtained at @xmath49 . in fig .
[ fig_2_six_eps_figs_grey ] for a fixed @xmath20 even though the films have the same number of particles , the film morphology looks different for @xmath48 and @xmath50 .
the difference arises from the variation in the position of growth front @xmath51 , defined as the set of occupied sites in the film that are highest in each column , and @xmath52 represents the horizontal lattice site on the substrate .
the growth front @xmath51 studied here is a crude approximation of a more complex aggregate that is growing .
however , our method of quantifying the growth front is justified because , in the afm experiments , the measured 1-d height - height correlation function is based only on height profiles along the fast scanning axis .
moreover the finite size of the afm tip is known to distort the growth front and the measured growth front is a convolution in which the interaction with the tip dilates the surface details @xcite . in fig .
[ fig_2_six_eps_figs_grey ] the growth fronts are shown vertically displaced by @xmath53 pixels in the growth direction for clarity .
the height fluctuation frequencies in @xmath51 are observed to occur at different length scales depending on the ratio @xmath4 . for a fixed @xmath20 ,
it is intuitive to think that the height profiles corresponding to @xmath50 ( figs .
[ fig_2_six_eps_figs_grey]d , e , f ) are more rougher than those obtained using @xmath48 ( figs .
[ fig_2_six_eps_figs_grey]a , b , c ) .
a detailed analysis of the dynamics of interface roughness evolution is presented in the later section .
generalized correlation function @xmath54 calculated for @xmath55 and @xmath56 .
the data are averaged over @xmath57 independent runs and the error - bars are smaller than the symbol size . ]
the _ _ q__th - order generalized height - height correlation function @xmath54 ( defined below ) is commonly used in identifying multi - affine surfaces @xcite .
@xmath58 the scaling properties of _ multi - affine _ surfaces can be described in terms of an infinite set of _ hurst exponents _
@xcite @xmath59 which are obtained using , @xmath60 for multi - affine interfaces , the exponents @xmath59 are known to vary with @xmath61 @xcite . in fig .
[ fig_3_generalized_correlation_fns_g1_only ] , we plot the generalized correlation function @xmath54 using @xmath61 = @xmath62 , @xmath63 , @xmath64 , and @xmath65 for @xmath4 = @xmath5 at deposition time @xmath20 = @xmath7 . for smaller @xmath66 , we observe a power - law dependence of @xmath54 on @xmath66 in accordance to eq .
( [ hurst_exponent ] ) .
the slopes of the log - log plots shown in fig .
[ fig_3_generalized_correlation_fns_g1_only ] depend on @xmath61 and indicate the presence of multi - scaling in the films generated by our vdp growth model .
the average height of the growth front @xmath67 , is defined as , @xmath68 and quantifies the overall thickness of the film .
figure [ fig_4_avg_h_vs_g ] shows the plot of @xmath69 versus @xmath20 for different @xmath4 and @xmath70 . for the studied values of @xmath4 , the @xmath69 versus
@xmath20 plots show a linear relationship .
this linear behavior is a consequence of restricting the growth to 1-d and the excluded volume constraint implemented in the growth process . in fig .
[ fig_4_avg_h_vs_g ] with an increase in @xmath4 , we observe a systematic increase in the slope of the @xmath69 versus @xmath20 plot .
we define the _ growth rate _ @xmath71 of the polymer film as , @xmath72 variation of @xmath69 as a function of deposition time @xmath20 for different @xmath4 and @xmath33 = 512 .
the data are averaged over 500 independent runs .
the statistical errors are smaller than the symbol size . ] in general one expects the polymer film s growth rate to be proportional to the incoming free - monomer flux only , i.e. @xmath71 @xmath73 @xmath0 and it is natural to expect @xmath71 to be independent of @xmath74 since @xmath75 is a constant for a fixed @xmath33 . however , in fig .
[ fig_4_avg_h_vs_g ] our simulations show a strong dependence of @xmath71 on @xmath4 as well , and the growth rate @xmath71 is observed to increase monotonically with @xmath4 .
this @xmath4 dependent growth rate @xmath71 indicates that the growth rate is effected by monomer diffusion directly , i.e. , there is a net uphill monomer diffusion current that is responsible for this @xmath76 relationship .
we thus incorporate an additional term @xmath77 that determines @xmath71 in addition to its dependence on @xmath0 .
variation of @xmath77 as a function of @xmath4 .
the data were obtained for @xmath33 = @xmath45 and averaged over @xmath78 independent runs .
the statistical errors are smaller than the symbol size . ]
@xmath79 where @xmath80 is the growth rate due to the random deposition flux only ( @xmath1 = 0 ) and @xmath77 is the growth rate induced only by the uphill diffusion of free - monomers . in fig .
[ fig_5_r_vs_g ] we show the variation of @xmath77 as a function of @xmath4 .
we find a monotonic increase in @xmath77 with an increase in @xmath4 .
this shows the strong influence of @xmath4 in determining the polymer film s growth rate .
specifically , at higher @xmath4 we find @xmath77 @xmath73 @xmath81 .
this indicates that the growth rate due to uphill diffusion @xmath77 asymptotically follows a power - law dependence with @xmath4 .
lateral film density profiles @xmath82 of polymer films grown on a @xmath70 substrate with @xmath48 at deposition times of @xmath83 , @xmath84 , @xmath85 , and @xmath86 respectively .
the statistical errors are smaller than the symbol size . ]
the morphology of the growth front can be characterized by studying its spatial and temporal evolution . in morphological scaling studies , typically two kinds of scaling behaviors are associated with the roughening kinetics : the global scaling and the local scaling @xcite .
the growth models with anomalous kinetic roughening are known to have different scaling exponents in the local and the global scales .
the local roughness exponents have been employed in studying the irregularly growing mound morphology and are often used in the experimental analysis @xcite . in general , the local roughness exponent @xmath87 and the global roughness exponent @xmath10 take different values depending on the type of scaling exhibited by the growth process . in the case of super - rough surfaces generated by nonequilibrium mbe growth models @xcite and growth models with horizontal diffusion @xcite , the assumption of the equivalence between the global and local descriptions of the surface is not valid and such behavior has been termed as anomalous scaling @xcite .
the differences in the global and local scaling exponents have been attributed to the _ super - roughening _ and _ intrinsically anomalous _ spectrum observed in the anomalous scaling of surfaces @xcite .
for studying the morphological evolution of vdp generated polymer films , it is essential to identify their steady growth regime .
to do so we employed the _ lateral film density _
@xmath88 ( at a height @xmath89 lattice units ) of polymer film defined as , @xmath90where @xmath91 represents the number of occupied lattice sites at a height _ y _ above the substrate .
calculations were performed on films with @xmath70 , @xmath48 at deposition times of @xmath83 , @xmath84 , @xmath85 , and @xmath86 , respectively .
figure [ fig_6_g1_l512_lateral_f_density_vs_t ] shows the variation of @xmath92 with @xmath20 .
similar plots were obtained for other @xmath4 values also .
three distinct regions of the polymer film growth referred to as the _ interface region _
, the steady - state _ film region _ and the _ growth front _ were identified based on the ref .
@xcite . to evaluate the scaling exponents for kinetic roughening
, we only concentrate on films with the deposition time @xmath93 .
variation of @xmath94 with @xmath33 for @xmath4 varying from @xmath5 to @xmath8 .
the data are averaged over @xmath95 independent runs and the statistical errors are smaller than the symbol size.the error in the exponents were obtained from the curve - fit . ] the global scaling of the growth front can be determined by studying the lattice size dependent _ interface width _
@xmath96 defined as , @xmath97 ^2}.\ ] ] in many 1-d morphological growth processes such as ballistic deposition @xcite , eden model @xcite , and solid - on - solid models @xcite , @xmath96 usually follows the scaling law , @xmath98 where exponent @xmath99 is known as the _ growth exponent _ that characterizes the time - dependence of surface roughening . for
any given @xmath33 , the power - law increase in @xmath96 does not continue indefinitely with @xmath20 , and is followed by a saturation regime during which the @xmath96 reaches a saturation value @xmath94 .
the power - law growth regime and the saturation regime are separated by a _ crossover time _ @xmath100 .
as @xmath33 increases , @xmath94 also follows a power - law @xcite , @xmath101 and @xmath102 is referred to as the _ global roughness exponent _ @xcite .
the time @xmath100 at which the behavior of @xmath96 crosses over from eq .
( [ short_time ] ) to that of eq .
( [ long_time ] ) depends on @xmath33 and scales as , @xmath103 where @xmath13 is called the _
global dynamic exponent_. typically the scaling exponents are independent of specific interactions involved in the growth process and depend on the dimensionality and symmetries of the system @xcite . for some growth processes ,
the exponents @xmath10 , @xmath99 , and @xmath13 are unified using the _ dynamic scaling hypothesis _
@xcite ( also known as the _ family - vicsek scaling _ ) given as , @xmath104where @xmath105 is referred to as the _ scaling function _ and satisfies , @xmath106 and @xmath107 in fig .
[ fig_7_combined_log_width_log_time_g1_g3 ] we show two representative plots of @xmath96 versus @xmath20 on a log - log scale for ( a ) @xmath48 , ( b ) @xmath108 and varying @xmath33 . from both figs .
[ fig_7_combined_log_width_log_time_g1_g3](a ) and [ fig_7_combined_log_width_log_time_g1_g3](b ) , for @xmath109 , the @xmath110 versus @xmath111 shows a linear dependence implying a power - law behavior .
for the same @xmath4 and @xmath112 , the @xmath96-@xmath20 plots overlap with one another for different @xmath33 .
we estimated @xmath99 by fitting eq .
( [ short_time ] ) to the plots of @xmath96 in fig .
[ fig_7_combined_log_width_log_time_g1_g3 ] for the film region @xmath112 .
we obtained average @xmath113 and @xmath114 for @xmath4 = @xmath5 and @xmath7 respectively .
for other @xmath4 values , the @xmath99 obtained was close to @xmath115 and we observed an invariance of @xmath99 ( within the error - bars ) with @xmath4 .
the statistical average and error bars in @xmath99 were obtained from @xmath116 independent simulations .
the vdp model studied here is similar to the ballistic deposition model with additional degree of freedom including diffusion , polymer initiation , extension and merger .
however , unlike the atomic diffusion in mbe growth , the free - monomer diffusion is confined by the linear geometry of the polymer chain . from fig .
[ fig_2_six_eps_figs_grey ] one can observe that in most cases the polymer chains are more or less perpendicular to the substrate . since the free - monomers can only move along the polymer chain , most of the diffusion happens in vertical direction rather than in lateral direction ( which is the case for mbe growth ) .
yet , the vertical diffusion does not contribute significantly to extra roughness increasing or decreasing ( the total particle number should be conserved ) in the growth front due to the porous nature of the film as compared to the random deposition model .
thus , it is expected that the growth exponent will be close to that of random deposition @xmath117 .
we calculated @xmath94 by averaging @xmath96 for _ @xmath118 _ from the data shown in figs .
[ fig_7_combined_log_width_log_time_g1_g3](a ) and [ fig_7_combined_log_width_log_time_g1_g3](b ) .
figure [ fig_8_combined_alpha_g3_regression_using_l150_upto_l600 ] shows the plot of @xmath94 versus @xmath33 on the log - log scale for @xmath4 varying from @xmath5 to @xmath8 .
the exponent @xmath10 was estimated for each @xmath4 using eq .
( [ long_time ] ) . with an increase in @xmath4 from @xmath5 to @xmath8 ,
@xmath10 was found to decrease from @xmath11 to @xmath12 , the error - bars in @xmath10 were obtained from the curve fitting .
the exponent @xmath10 is known to be closely related to the surface fractal dimension @xcite .
the smaller the @xmath10 , the larger is the fractal dimension .
our observation of a decrease in @xmath10 with an increase in @xmath4 shows that the fractal dimension of the growth front increases with @xmath4 , i.e. there are more spatial frequency fluctuations in the film morphology with an increase in @xmath4 .
this finding is consistent with the growth front profiles shown in fig .
[ fig_2_six_eps_figs_grey ] .
this result also demonstrates that @xmath4 induces a large effective vertical growth rate @xmath119 ( shown in fig .
[ fig_5_r_vs_g ] ) and the large @xmath119 in turn produces a much rougher film surface .
+ to determine whether the vdp process obeys the dynamic scaling behavior , we start with an assumption that vdp growth follows the dynamic scaling hypothesis and obtain @xmath13 through eq .
( [ defn_z ] ) .
as @xmath96 scales with both @xmath20 ( eq . ( [ short_time ] ) ) and @xmath33 ( eq . ( [ long_time ] ) ) we can rescale the @xmath96 curve shown in fig .
[ fig_7_combined_log_width_log_time_g1_g3 ] by plotting @xmath120 versus @xmath121 to see whether those curves `` collapse '' . for @xmath48
we obtain @xmath10 = @xmath11 , @xmath99 = @xmath122 and according to eq .
( [ defn_z ] ) @xmath13 = @xmath14 and for @xmath108 we get @xmath10 = @xmath123 , @xmath99 = @xmath114 , and @xmath13 = @xmath124 .
we use the data from figs .
[ fig_7_combined_log_width_log_time_g1_g3](a ) and [ fig_7_combined_log_width_log_time_g1_g3](b ) and divide @xmath96 by @xmath125 .
this shifts the curves of varying @xmath33 vertically on the log - log scale . according to eq .
( [ long_time ] ) , these curves now saturate at the same value of the ordinate @xmath120 , however their saturation times do not overlap .
we then rescale the time axis and plot @xmath121 for both cases of @xmath4 .
this rescaling of time axis according to eq .
( [ time_crossover ] ) leads to a horizontal shift of the curves and the curves now saturate at the same abscissa @xmath121 . in figs .
[ fig_9_combined_rescaled_plots_log_width_time](a ) and [ fig_9_combined_rescaled_plots_log_width_time](b ) we show the rescaled plots of @xmath96 for @xmath48 , @xmath7 and consequently observe the `` collapse '' of individual curves for varying @xmath33 onto a single curve .
this characteristic `` collapsed '' curve shown in figs .
[ fig_9_combined_rescaled_plots_log_width_time](a ) and [ fig_9_combined_rescaled_plots_log_width_time](b ) is the _ scaling function _
@xmath105 mentioned in the eq .
( [ scaling_function_property ] ) .
the scaling functions obtained in figs . [ fig_9_combined_rescaled_plots_log_width_time](a ) and [ fig_9_combined_rescaled_plots_log_width_time](b ) are observed to follow eq .
( [ scaling_function_property ] ) for both cases of @xmath49 and @xmath7 . for other @xmath4 values , we observed similar `` collapse '' behavior indicating the global dynamic scaling of growth fronts of the polymer films grown using vdp .
auto - correlation function @xmath126 calculated for @xmath55 and @xmath56 ( averaged over @xmath57 independent runs ) . ]
estimates for column separation @xmath127 and lateral correlation length @xmath128 obtained for @xmath129 and @xmath130 .
open symbols correspond to @xmath131 and filled symbols correspond to @xmath127 .
the data are averaged over @xmath57 independent simulations with @xmath55 and @xmath56 .
the error in the exponents were obtained from the curve - fit . ] the local scaling behavior of the growth front can be understood from studying the spatial correlation functions : the auto - correlation function @xmath126 and height - height correlation function @xmath3 that describe the local properties of growing interfaces , @xmath132 @xmath133 ^ 2\rangle\ } = 2[w^2-c(r , t)],\]]where @xmath66 is the translation distance also referred to as the lag or slip @xcite and @xmath134 denotes a spatial average over the entire system .
the functions @xmath126 and @xmath3 are directly related as shown in eq .
( [ h_r_equation ] ) and differ only by a constant pre - factor of @xmath135 .
the @xmath3 scales in the same way as the interface width and is often used in studying the kinetic roughening @xcite . in order to obtain accurate parameters from the correlation functions ,
it is important to account for the accuracy of statistical averages . for the random gaussian model surface studied by ref .
@xcite , converging @xmath3 were obtained within an order of @xmath136 where @xmath137 is the system dimension and @xmath128 is the lateral correlation length .
the accuracy depends not only on the number of data points , but also on the sample size .
it is the ratio @xmath138 , not the number of data points , that determines the accuracy .
once the ratio @xmath138 is known , one may not be able to increase the accuracy no matter how many data points are collected @xcite .
this rule is different from the law of large numbers for independent random variables .
this important difference needs to be recognized while studying spatially correlated systems .
ideally , one would like to have @xmath139 , i.e , @xmath140 .
however , due to the computational constraints we performed calculations using @xmath55 and @xmath56 . for a random self - affine surface
, @xmath126 usually decays to zero with an increase in @xmath66 .
the shape of the decay depends on the type of the surface and the decay rate depends on the distance over which two points @xmath52 , @xmath141 become uncorrelated . in fig .
[ fig_2_six_eps_figs_grey ] the polymer growth front does not appear to be a random rough surface , instead it has regular fluctuations of the columnar structures . figure [ fig_10_auto_correlation_fns_g1_only ] shows the @xmath126 for @xmath55 after a deposition time @xmath56 for @xmath48 , @xmath6 , and @xmath7 used to characterize the surface morphology as a function of @xmath20 .
we can define two different lateral length scales : the lateral correlation length @xmath128 and the average column - separation @xmath127 .
the lateral correlation length @xmath128 defines a representative lateral dimension of a rough surface and is estimated through @xmath126 using @xmath142 . within a distance of @xmath128
the surface heights of any two points are correlated .
the parameter @xmath127 characterizes a wavelength - selection of a surface and is determined by measuring the value of @xmath66 corresponding to the first zero - crossing of @xmath126 @xcite .
the variation of @xmath127 with @xmath20 represents how the columnar structures coarsen with deposition time . in general
, the evolution of the columnar feature size follows a power law with @xmath20 given by @xcite @xmath143 and @xmath144 can be referred to as the _ coarsening exponent_. in fig .
[ fig_11_feature_size_l2k_g1_3_t_1k ] we plot @xmath127 versus @xmath20 for @xmath55 and @xmath48 and @xmath7 along with the estimates for @xmath144 . with an increase in @xmath4 from @xmath5 to @xmath7 , @xmath144 is found to remain close to @xmath145 .
the invariance of @xmath144 indicates that the coarsening of the columnar structures follow a power law that is unaffected by the ratio @xmath4 . and
we believe that this is dominated by the shadowing effect due to monomer vapor coming from all different angles uniformly .
the _ local roughness exponent _ @xmath146 of the growth front can be obtained from @xmath3 using @xcite @xmath147 in figs .
[ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k](a ) and [ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k](b ) we plot @xmath148 for @xmath149 , @xmath150 , and @xmath151 along with the estimates for @xmath152 for @xmath55 and @xmath48 , @xmath7 respectively . for a given @xmath4 and varying @xmath20 , the estimates of @xmath17
were observed to remain invariant within statistical - errors . from figs .
[ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k](a ) and [ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k](b ) , we obtained an average @xmath17 of @xmath153 and @xmath154 for @xmath48 and @xmath7 respectively . a comparison between @xmath10 and @xmath87 ( shown in fig .
[ fig_8_combined_alpha_g3_regression_using_l150_upto_l600 ] and fig .
[ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k ] ) shows @xmath155 for studied @xmath4 and the scaling behavior is observed to be different at short and large length scales .
for small @xmath66 , the plots of @xmath148 do not overlap at varying coverage and show the presence of non - stationary anomalous scaling @xcite .
the vertical temporal shift in the @xmath3 observed in fig [ fig_12_h_h_sq_l2k_g1_3_t_250_500_750_1k ] is due to the difference between @xmath10 and @xmath87 and indicates the presence of anomalous scaling . the mechanisms that lead to anomalous scaling can be separated into two classes : _ super - roughening _ @xmath156 and _ intrinsic _ anomalous scaling @xcite .
the _ anomalous _ growth exponent @xmath157 @xcite measures the difference between @xmath10 , @xmath87 and can be obtained from the scaling of @xmath3 @xcite @xmath158and the _ anomalous _ scaling function @xmath159 @xcite satisfies @xmath160 @xmath161 @xmath162 in figs .
[ fig_13_rescaled_h_h_sq_l2k_g1_3_t_250_500_750_1k](a ) and [ fig_13_rescaled_h_h_sq_l2k_g1_3_t_250_500_750_1k](b ) we show the plots of @xmath163 versus @xmath164 and obtain the `` data - collapse '' for @xmath48 and @xmath7 respectively .
the `` collapsed '' curves shown in figs .
[ fig_13_rescaled_h_h_sq_l2k_g1_3_t_250_500_750_1k](a ) and [ fig_13_rescaled_h_h_sq_l2k_g1_3_t_250_500_750_1k](b ) are the scaling functions @xmath159 ( for @xmath48 and @xmath7 ) mentioned in eq .
( [ rescaled_h_function ] ) . for both @xmath4 , the scaling functions obtained in fig .
[ fig_13_rescaled_h_h_sq_l2k_g1_3_t_250_500_750_1k ] satisfy eq . ( [ g_a_scaling_function ] ) in accordance with the theory for anomalous scaling @xcite .
the exponents @xmath165 and @xmath166 were obtained from @xmath159 plots using eq .
( [ g_a_scaling_function ] ) .
for @xmath48 we have @xmath167 , @xmath168 and for @xmath108 we obtained @xmath169 , @xmath170 . for both @xmath4 , we find that the exponents @xmath165 and @xmath166 obtained from the curve - fit satisfy eqs .
( [ equation_kappa_1 ] ) and ( [ equation_kappa_2 ] ) for the numerical estimates of @xmath171 and @xmath172 obtained from eq .
( [ long_time ] ) and eq .
( [ local_alpha_equation ] ) .
variation of exponents @xmath10 , @xmath99 , @xmath13 , @xmath17 , and @xmath19 with @xmath4 .
the global exponents were calculated after @xmath173 and the data were averaged over @xmath95 independent runs .
the local exponents were calculated using @xmath55 and @xmath56 .
the data were averaged over @xmath57 runs and the error in the exponents were obtained from the curve - fit . ] the dynamic scaling exponents of the kinetic roughening are summarized in figs .
[ fig_14_exponents_alpha_beta_z_vs_g](a ) , [ fig_14_exponents_alpha_beta_z_vs_g](b ) , and [ fig_14_exponents_alpha_beta_z_vs_g](c ) which shows the variation of global scaling exponents @xmath10 , @xmath99 , @xmath13 and local exponents @xmath87 , @xmath19 with @xmath4 . with an increase in @xmath4 from @xmath5 to @xmath8 , we found @xmath9 , @xmath10 decreased from @xmath11 to @xmath12 , and @xmath13 decreased from @xmath14 to @xmath15 . on the local length scale , an increase in @xmath4 did not effect a noticeable change in @xmath87 @xmath174 and @xmath19 was observed to decrease from @xmath175 to @xmath176 .
we have performed 1 + 1d monte carlo simulation of vdp growth process by considering free - monomer deposition , free - monomer diffusion , polymer initiation , extension and polymer merger .
the ratio @xmath4 of the free - monomer diffusion coefficient @xmath1 to the deposition rate @xmath0 was found to have a strong influence on the film s growth morphology .
the growth rate @xmath119 of the polymer film was found to increase monotonically with an increase in @xmath4 .
this is due to the consequence of an upper diffusion flux of free - monomers .
the detailed analysis of the surface morphology indicated the presence of very different scaling behavior at global and local length scales .
the kinetic roughening study of film interface indicates anomalous scaling and multiscaling . with an increase in @xmath4 from @xmath5 to @xmath8 ,
the global growth exponent @xmath9 was found to be invariant , whereas the global roughness exponent @xmath10 decreased from @xmath11 to @xmath12 along with a corresponding decrease in the global dynamic exponent @xmath13 from @xmath14 to @xmath15 .
the global scaling exponents were found to follow the dynamic scaling hypothesis with @xmath16 for various @xmath4 . with an increase in @xmath4 from @xmath5 to @xmath8 ,
the average local roughness exponent @xmath87 remained close to @xmath18 with @xmath177 , this observation is unlike the ones obtained in self - affine surfaces @xcite .
the anomalous growth exponent @xmath19 was also found to decrease from @xmath175 to @xmath176 with an increase in @xmath4 . even though our model is in 1 + 1d as compared to the 2 + 1d experiments , our estimates of @xmath178 and @xmath19 are close to the experimental findings of @xmath179 to @xmath27 and @xmath180 obtained from the afm studies of linear pa - n films grown by vdp @xcite .
the similarity between the experimental and simulational estimates appears to be a coincidence since the dimension of the two systems are totally different .
we also did not observe the changes in the dynamic roughening behavior reported by ref .
@xcite , perhaps due to the limitations of our current simulation model in considering the effect of free - monomer diffusion only .
this makes us believe that the kinetic roughening of the polymer films is sensitive to the specific molecular - level interactions , relaxations of polymer chains through inter - polymer interactions , and the intrinsic nature of polymerization process that need to be accounted for in the future simulations .
st and dpl were partially supported by the nsf grant dmr-0810223 .
st and ypz were also partially supported by nsf grant cmmi-0824728 .
st would like to thank s. j. mitchell for help with the visualization tools . | the morphological scaling properties of linear polymer films grown by vapor deposition polymerization ( vdp ) are studied by 1 + 1d monte carlo simulations .
the model implements the basic processes of random angle ballistic deposition ( @xmath0 ) , free - monomer diffusion ( @xmath1 ) and monomer adsorption along with the dynamical processes of polymer chain initiation , extension , and merger .
the ratio @xmath2 is found to have a strong influence on the polymer film morphology .
spatial and temporal behavior of kinetic roughening has been extensively studied using finite - length scaling and height - height correlations @xmath3 .
the scaling analysis has been performed within the no - overhang approximation and the scaling behaviors at local and global length scales were found to be very different .
the global and local scaling exponents for morphological evolution have been evaluated for varying free - monomer diffusion by growing the films at @xmath4 = @xmath5 , @xmath6 , @xmath7 , and @xmath8 and fixing the deposition flux @xmath0 . with an increase in @xmath4 from @xmath5 to @xmath8 ,
the average growth exponent @xmath9 was found to be invariant , whereas the global roughness exponent @xmath10 decreased from @xmath11 to @xmath12 along with a corresponding decrease in the global dynamic exponent @xmath13 from @xmath14 to @xmath15 .
the global scaling exponents were observed to follow the dynamic scaling hypothesis , @xmath16 . with a similar increase in @xmath4
however , the average local roughness exponent @xmath17 remained close to @xmath18 and the anomalous growth exponent @xmath19 decreased from 0.23(4 ) to 0.18(8 ) .
the interfaces display anomalous scaling and multiscaling in the relevant height - height correlations .
the variation of @xmath3 with deposition time @xmath20 indicates non - stationary growth .
a comparison has been made between the simulational findings and the experiments wherever applicable . |
the f - theory vacuum @xcite constructed from an elliptically fibered threefold @xmath2 with section determines an effective theory with @xmath3 supersymmetry in six dimensions .
such supersymmetric theories will have fields in hypermultiplets , vector supermultiplets and tensor supermultiplets .
( see , for example , @xcite for a discussion of such theories . ) for any particular f - theory vacuum , the taxonomy of the supermultiplets may be derived from the geometry of @xmath2 as an elliptic fibration via seemingly straightforward methods in the case of the vector and tensor multiplets @xcite .
the classification of the hypermultiplet content has always been a little harder to carry out .
many methods have been proposed which allow the hypermultiplets to be determined from the geometry in certain cases @xcite .
the purpose of this paper is to outline a systematic approach to the problem of determining the gauge symmetry and hypermultiplet content of a given six - dimensional theory obtained from f - theory .
( note that as far as the moduli space of hypermultiplets in concerned , our methods utilize the associated type iia compactification and thus also apply directly to the compactification of m - theory on @xmath2 giving an @xmath4 theory in five dimensions and to the compactification of the type iia string on @xmath2 to yield an @xmath5 theory in four dimensions , provided that the expectation values of certain ramond
ramond fields have been tuned appropriately . )
the methods we employ will not be particularly new but we will see that the process of analyzing the gauge group and matter content can be quite a bit more subtle than had previously been appreciated . in particular , the case of monodromy of the fibration leading to non - simply - laced lie algebras requires some care . a particularly awkward case which has caused some confusion
is when a @xmath6 monodromy acts on a curve of @xmath0 singularities , i.e. , a curve of @xmath7 fibers in f - theory language . in this paper
we resolve this problem in agreement with an observation by intriligator and rajesh in @xcite concerning anomaly cancellation . in section [ s : lcy ] we will show how many features of a lie algebra structure arise naturally from an elliptically fibered threefold .
this will allow us to elucidate the method for determining the gauge algebra . in section [ s :
hyper ] we discuss exactly how to analyze the hypermultiplet content in the cases where the associated curves and surfaces within the threefold are smooth .
we discuss the cases where these curves and surfaces are singular in section [ s : odd ] .
this section includes some unexpected rules we are forced to adopt for @xmath1-brane wrapping
. although the results of this section are less rigorous than the preceding section , we are able to give precise results in many instances which can be extended to the general case under the fairly conservative assumption that the relevant physics is determined locally from the geometry of the singularities .
finally in section [ s : num ] we emphasize the peculiar numerical predictions which arise from anomaly cancellation in the f - theory compactification on @xmath2 .
we begin with a threefold @xmath2 which admits an elliptic fibration @xmath8 , where @xmath9 is a complex surface , and also assume that this elliptic fibration has a section .- field has been turned on in the base @xcite .
] the type iia string compactified on @xmath2 yields an effective four - dimensional theory with @xmath5 supersymmetry ; its strong - coupling limit , known as `` m - theory compactified on @xmath2 , '' yields an effective five - dimensional theory .
one more effective spatial dimension is obtained in a limit in which the areas of all components of the elliptic fibers shrink to zero this is the `` f - theory limit . ''
see , for example , @xcite for an explanation of this .
we point out that most of the following analysis does not really depend upon this elliptic fibration structure and applies to m - theory and type iia compactifications of @xmath2 .
we use the f - theory language as an organizational tool to give examples later on .
one also has the advantage in f - theory of being able to use anomaly cancellation as a powerful tool in checking the consistency of results concerning spectra of massless particles . in the f - theory context
we can freely exchange the notion of , say , an @xmath10 fiber and an @xmath11 singularity .
the former is the elliptic fibration description for the latter .
recall @xcite that this is because although @xmath10 is really the _ extended _ dynkin diagram of @xmath11 and that one always ignores the components of the fiber which hit the chosen section of the elliptic fibration .
thus , in the zero - area fiber limit of f - theory , a shrunken @xmath10 fiber gives the same physics as an @xmath11 singularity one dimension lower .
whenever rational curves in @xmath2 are shrunk down to zero size we expect @xmath1-branes of the type iia string wrapped around these curves to contribute massless particles to the spectrum .
it is precisely these massless states which are the focus of our interest in this paper .
actually we need to be careful with the statement that massless states appear automatically when a brane wraps a vanishing cycle .
there is always the subtlety of @xmath12-fields and r - r fields which should be tuned to the right value ( usually denoted `` zero '' by convention ) to really obtain a massless state .
as emphasized in @xcite the relevant parameters to worry about in this context are the r - r fields .
we may see this as follows .
if one considers the type iia string compactified on @xmath2 then deformations of the khler form ( and @xmath12-field ) on @xmath2 are given by vector moduli .
suppose we use these khler moduli to shrink down a holomorphic @xmath1-cycle to obtain an enhanced gauge symmetry .
once we reach this point of enhanced symmetry we may have a phase transition releasing new hypermultiplet degrees of freedom .
thus at the point of phase transition , these new parameters , which include r - r fields , are fixed at some value . reversing this point of view
, we may tune parameters in the hypermultiplet moduli space to achieve an enhanced gauge symmetry but these parameters include r - r fields . thus we need to assume always that the r - r parameters have been tuned to the appropriate values required to obtain the enhanced gauge symmetries we discuss below .
witten @xcite analyzed how to determine the massless particle content for a given configuration of rational curves .
let us assume that a given rational curve lives in a family parametrized by a moduli space @xmath13
. in the simplest case one has an embedding @xmath14 .
an isolated rational curve is a trivial example of this where @xmath13 is simply a point . according to witten s calculation
, one half - hypermultiplet may be associated to the fact that a @xmath1-brane breaks half of the supersymmetry .
this half - hypermultiplet is then tensored with the total cohomology of @xmath13 in an appropriate sense .
the result is that if @xmath13 is a point , then we simply obtain a single half - hypermultiplet .
if @xmath13 is an algebraic curve of genus @xmath15 then we obtain a single vector multiplet and @xmath15 hypermultiplets .
this was also argued by a different method in @xcite .
note that for any wrapping we may also wrap with the opposite orientation to double this spectrum . of central interest to us
is the fact that compactifying a type iia string theory ( and thus m - theory and f - theory ) on a threefold @xmath2 produces a theory with a yang mills sector .
the gauge fields may be viewed as arising from integrating the r - r @xmath16-form of the type iia string over @xmath1-cycles in @xmath2 to produce @xmath17-forms .- form field over @xmath1-cycles . ]
these @xmath17-forms play the role of the yang - mills connection .
in addition , the @xmath18-cycles in @xmath2 which are dual ( via intersection theory ) to these @xmath1-cycles will play an important role .
let @xmath19 denote the @xmath18-form field strength of the r - r @xmath16-form in the type iia string .
note that the @xmath1-branes of the type iia theory are _ electrically _ charged under this field that is @xmath20 for a @xmath21-dimensional shape @xmath22 ( such as a six - sphere ) enclosing the seven directions transverse to a fundamental @xmath1-brane . upon compactification we will be wrapping @xmath1-branes around a @xmath1-cycle in @xmath2 to produce a point particle in four - dimensional space - time .
to find the charge of this resulting particle we may take @xmath23 , where @xmath24 is a sphere in four dimensional space - time enclosing the particle and @xmath25 is a @xmath18-cycle within @xmath2 .
it follows that in the type iia compactification 1 .
we have @xmath26 gauge symmetries of the type @xmath27 , each labelled by an element of @xmath28 , in addition to the @xmath29 gauge symmetry coming from the r - r @xmath17-form ( whose charge is measured using @xmath30 , the generator of @xmath31 ) .
2 . if a @xmath1-brane wraps a @xmath1-cycle @xmath32 to produce a particle then the `` electric '' charge of this particle under the @xmath27 symmetry associated to a @xmath18-cycle @xmath25 will be the intersection number @xmath33 .
we thus obtain a perturbative @xmath34 gauge symmetry in type iia . in the m - theory compactification
, there is no r - r @xmath17-form , and the eight transverse directions to the m - theory @xmath1-brane are enclosed by @xmath35 , so the total `` perturbative '' gauge symmetry is given by @xmath36 . in the f - theory limit ,
the only @xmath18-cycles which contribute gauge fields are those with intersection number zero with the elliptic fiber ; moreover , @xmath18-cycles which are the inverse images of @xmath1-cycles in the base @xmath9 are associated to tensor multiplets rather than gauge fields .
thus , in the f - theory limit , we get a `` perturbative '' gauge symmetry group of @xmath37 .
as is now well - known , and as we will discuss , the wrapped @xmath1-branes will elevate this @xmath38 gauge symmetry to a non - abelian lie group ( since certain wrapped branes include vector multiplets in their spectra ) , where @xmath39 , @xmath40 , or @xmath41 for iia , m - theory or f - theory , respectively .
from now on we will concern ourselves only with the lie _ algebra _ of the gauge symmetry .
it was noted in @xcite that , at least in f - theory , the global structure of the gauge group may be recovered from the mordell - weil group of @xmath2 as an elliptic fibration . of the gauge group
is equal to the mordell - weil group ( including both the free and torsion parts ) . ]
if this @xmath42 appears as the cartan subalgebra of our gauge algebra then the discussion above implies that we may make the following identifications .
let @xmath43 be the ( real ) cartan subalgebra , let @xmath44 be the dual space , and let @xmath45 be the coroot lattice and @xmath46 be the weight lattice so that the cartan subgroup @xmath38 is naturally identified with @xmath47 . for the iia compactification , we take @xmath48 and @xmath49 , and in m - theory , we take @xmath50 and @xmath51 . in f - theory , we begin with the orthogonal complement within @xmath28 of the elliptic fiber @xmath52 , and then we mod out by @xmath53 ( that is , @xmath54^\perp/\pi^{-1}h_2(\sigma)\subset h_4(x)/\pi^{-1}h_2(\sigma)$ ] ) ; we then take @xmath55 to be the dual lattice of @xmath56 . in each case , a @xmath1-brane wrapped around a particular @xmath1-cycle is then naturally associated with an element of the weight lattice and its charges under the cartan subalgebra are given in the standard way .
we work this out in detail in several particular cases .
consider first the case that @xmath2 contains a `` ruled '' complex surface @xmath57 admitting a fibration @xmath58 , for some @xmath13 , where all fibers are isomorphic to @xmath59 . the fibers will shrink down to zero size in the f - theory limit .
the simplest example of this is @xmath60 where @xmath13 is a riemann surface of genus @xmath15 and that @xmath61 is in the fiber direction .
that is to say , in our elliptic fibration @xmath8 we have a curve @xmath62 over which the fiber is i@xmath63 .
clearly we have massless states appearing for the @xmath1-branes wrapped around @xmath64 .
we also have a @xmath65 symmetry associated to @xmath66 .
let us consider the normal bundle of a single @xmath64 curve .
this normal bundle may be written as @xmath67 where @xmath68 by the adjunction formula and the fact that @xmath2 is a space . since this curve
may be translated along the @xmath13 direction one of these line bundles must be trivial .
thus the normal bundle is @xmath69 where the @xmath70 describes the normal bundle direction which is also normal to @xmath71 .
therefore @xmath72 .
this tells us that we have a vector supermultiplet and @xmath15 hypermultiplets from wrapping @xmath1-branes around @xmath73 all with charge @xmath74 with respect to the @xmath65 gauge symmetry associated to this divisor .
similarly by wrapping with the opposite orientation we obtain a copy of this except with charge @xmath75 .
these vector supermultiplets enhance the @xmath65 symmetry to @xmath76 in the usual way and we have an additional @xmath15 hypermultiplets in the adjoint representation .
the key point is to notice in this construction that the condition @xmath77 has played the role of the _ cartan matrix _ of @xmath76 .
the next simplest case is where we have a set of curves @xmath78 which may intersect each other and are each isomorphic to @xmath59 and lying in the fiber direction .
we assume that @xmath79 embeds algebraically into @xmath2 .
distinct @xmath59 fibrations over @xmath13 along appropriate disjoint sections , forming a chain . in the remainder of this paper , we will continue to explain by example and will not explicitly state the most general form of the algebraic surfaces which contract to @xmath13 in the f - theory limit .
] we now have a cartan matrix given purely by the configuration of @xmath78 . applying the above method we obtain the standard f - theory result of a simply - laced enhanced gauge symmetry as listed , for example , in table 4 of @xcite .
( 8865,6298)(511,-6347 ) ( 1201,-1861 ) ( 5 , 2)4500 ( 1201,-1861 ) ( 0,-1)3900 ( 1201,-5761 ) ( 5 , 2)4500 ( 5701,-61 ) ( 0,-1)3900 ( 3631,-511 ) ( 2,-1)4350 ( 1984,-1180 ) ( 2,-1)4350 ( 1201,-2774 ) ( 5 , 2)4500 ( 1201,-3689 ) ( 5 , 2)4500 ( 1201,-4591 ) ( 5 , 2)4500 ( 1201,-5253 ) ( 5 , 2)4500 ( 3619,-1428 ) ( 2,-1)4350 ( 3579,-2326 ) ( 2,-1)4350 ( 3469,-3190 ) ( 2,-1)4350 ( 3290,-4023 ) ( 2,-1)4350 ( 2692,-3462 ) ( 2,-1)4350 ( 2272,-2172 ) ( 2,-1)4350 ( 8401,-1861)(-1,-1)1275 ( 9301,-3961)(-1 , 0)1275 ( 3301,-5761)(-3 , 4)1125 ( 6901,-1261)(-1 , 0)1125 ( 3081,-4165 ) ( 2,-1)4350 ( 601,-5161 ) ( 0 , 1)2955 ( 2401,-1381 ) ( 1973,-1178 ) ( 6332,-3356 ) ( 8476,-1936)(0,0)[lb ] ( 9376,-4036)(0,0)[lb ] ( 6976,-1336)(0,0)[lb ] ( 3376,-5911)(0,0)[lb ] ( 411,-5561)(0,0)[lb ] as noted first in @xcite the real power of this cartan matrix approach is that it gives a clear way of describing _ non_-simply - laced gauge algebras .
consider a less trivial example of ruled surfaces as shown in figure [ fig : one ] . in this example
the moduli space @xmath80 of the curve @xmath64 is different from the moduli space @xmath81 the curve @xmath82 .
think of @xmath80 as the vertical direction in the figure .
we obtain ruled surfaces @xmath83 and @xmath84 .
we have a two - fold cover @xmath85 branched at one point in figure [ fig : one ] .
any other branch points are not shown in the figure .
the intersection matrix of this configuration may be written @xmath86 this is the cartan matrix for @xmath87 ( or @xmath88 ) and so our enhanced gauge algebra should be @xmath87 .
this phenomenon of obtaining a non - simply - laced symmetry algebra was first noted in @xcite inspired by the construction of @xcite .
there it was explained by monodromy acting on the fibers as follows .
let @xmath80 be embedded in @xmath9 and let @xmath81 be a two - fold cover of @xmath80 ( branched at various points ) . over a generic point in @xmath80
we see that , ignoring the component meeting the chosen section , the kodaira fiber consists of one line from @xmath71 and two lines from @xmath89 forming the ( dual ) dynkin diagram of @xmath90 . moving along a closed path in the complement of the set of branch points of @xmath85 we will exchange the two lines in @xmath89 .
this action on the dynkin diagram is induced by an _
outer automorphism _ of @xmath90 and the invariant subgroup under this outer automorphism can be taken to be @xmath87 .
one might therefore suspect that the effective gauge algebra is the monodromy - invariant subalgebra of the simply - laced gauge symmetry generated locally by the vanishing cycles .
this was the assertion in @xcite .
unfortunately it is an ambiguous statement .
let us analyze carefully all possible outer automorphisms of @xmath91 .
an element @xmath15 of @xmath91 satisfies @xmath92 .
complex conjugation @xmath93 is an example of an outer automorphism .
indeed this acts on the dynkin diagram of @xmath91 by reflection about the middle node .
clearly the invariant subgroup under this outer automorphism is given by @xmath15 real .
but this yields the group @xmath94not what we were expecting !
a general outer automorphism of @xmath91 can be obtained by combining complex conjugation with an arbitrary inner automorphism , yielding @xmath95 , where @xmath96 ( there are no other possibilities since that would imply further symmetries of the dynkin diagram ) . since this outer automorphism acts on the dynkin diagram as the reflection , it is also a viable candidate for the monodromy action on the gauge group . in this general situation ,
the invariant subalgebra satisfies @xmath97 the case @xmath98 yields @xmath94 as stated before . now
if we put @xmath99 where @xmath100 is the @xmath101 identity matrix , we obtain the group @xmath102as desired . in this case
, the outer automorphism is an involution , but this is not a requirement in general .
we see then that the method of directly working out the cartan matrix from intersection theory is a better way to determine the effective gauge algebra in f - theory than trying to find subalgebras invariant under outer automorphisms .
the latter method is ambiguous .
one might try to assert that f - theory picks out the `` maximal '' invariant subalgebra under all possible outer automorphisms .
indeed , @xmath103 is `` bigger '' than @xmath104 in as much as it has a larger dimension ( although in general @xmath105 ) .
however , even this approach is inadequate as will be shown by examples in section [ s : odd ] .
note that in the m - theory or type iia compactifications , an ambiguity of the sort we have discovered is actually to be expected . as we have already pointed out , if the ramond
ramond fields have non - zero expectation values then some of the non - abelian gauge fields will become massive ; when these are integrated out , the gauge group becomes smaller .
this is precisely what happens when the outer automorphism of the covering lie group is varied in the construction above .
the gauge algebra which we wish to determine is the one in which these effects have been turned off so that the f - theory limit can be taken .
( a similar phenomenon of variable gauge group depending on the precise value of an outer automorphism has been observed in a closely related context by witten @xcite , and applied in @xcite ) .
in the last section we described how to determine the gauge algebra in f - theory ( or m - theory or iia string theory ) by determining the cartan matrix from intersection theory .
similar methods will in principle determine the hypermultiplet spectrum completely as we now discuss .
first there can be the case of a family of rational curves acquiring extra rational curves at certain points in the family . in the context of elliptic fibrations
this can be seen as collisions of curves in @xmath9 over which there are singular fibers .
the simplest example is a transverse collision of i@xmath106 and i@xmath107 .
the resolution of singularities associated to this collision was explained in @xcite , and applied to the case of an i@xmath106-i@xmath108 collision in the context of string theory in section 8.2 of @xcite .
the key point is that there exist rational curves within the collision with normal bundle @xmath109 .
one of these curves @xmath110 is the intersection of two ruled @xmath18-cycles , one lying over the curve of @xmath111 fibers , and the other lying over the curve of @xmath112 fibers
. the normal bundle of @xmath110 in @xmath2 is naturally the direct sum of the normal bundles of @xmath110 in each of these @xmath18-cycles , and each of these is @xmath113 .
thus this curve appears as ( minus ) a fundamental weight .
the above rules imply that we have found a curve representing the ( lowest ) weight of the @xmath114 representation of @xmath115 . by adding other ( possibly reducible ) curves in the collision fiber we may indeed build up the full @xmath114 representation .
thus _ the transverse collision of a curve of i@xmath106 and i@xmath107 fibers yields a hypermultiplet in the @xmath114 representation of @xmath115 . _
( this same result had earlier been determined using quite different methods in @xcite .
another approach which is closer to ours appeared in @xcite . )
many other `` collisions '' can be explained in similar ways .
however , if the extra rational curves at the collision point have normal bundles other than @xmath109 , then witten s calculation does not directly apply .
general methods for evaluating the corresponding contribution to the hypermultiplet spectrum are not known .
the case of non - simply - laced symmetry algebras raises even more complicated possibilities .
some of the hypermultiplet matter can appear in a somewhat `` non - local '' manner as we now explain .
suppose we are in a situation analogous to figure [ fig : one ] .
let us consider the example of a type @xmath116 fiber ( where we again ignore the component passing through the chosen section ) .
let the middle component in the chain have a moduli space given by @xmath80 and the other components have moduli space @xmath81 where @xmath85 is a double cover .
that is , we have a @xmath6-monodromy acting on the @xmath116 fiber ( in the only possible way ) .
figure [ fig : one ] is the case @xmath117 .
according to @xcite we should obtain @xmath118 hypermultiplets for @xmath1-branes wrapping the middle component and @xmath119 hypermultiplets for @xmath1-branes wrapping each of the other components .
note that @xmath120 from the double cover .
there are additional hypermultiplets arising from wrapping connected unions of these components .
in fact , each of the positive roots of the covering algebra @xmath121 is represented by such a connected union , some of which are fixed by the monodromy , and others of which are exchanged in pairs under the monodromy .
the ones which are fixed under monodromy have @xmath80 as moduli space , while those which are exchanged in pairs have @xmath81 as their moduli space .
when we organize these weights in terms of representations of @xmath103 , we find that the invariant subspace describes the adjoint of @xmath103 while the anti - invariant subspace describes the remaining weights in the adjoint of @xmath121 . on the other hand
, each invariant positive root contributes to the invariant subspace , while the roots exchanged in pairs contribute to both the invariant and anti - invariant subspaces .
we conclude that the adjoint of @xmath103 occurs @xmath118 times while the weights in the anti - invariant subspace each occur @xmath122 times .
( 6024,6924)(2389,-6673 ) ( 3001,-661 ) ( 5401,-661 ) ( 7801,-661 ) ( 3001,-3061 ) ( 7801,-3061 ) ( 3001,-5461 ) ( 5401,-5461 ) ( 7801,-5461 ) ( 5401,-3061 ) ( 5401,-661 ) ( 3001,-3061 ) ( 5401,-3061 ) ( 7801,-3061 ) ( 5401,-5461 ) ( 5401,239 ) ( 0,-1)6900 ( 7801,239 ) ( 0,-1)6900 ( 3001,239 ) ( 0,-1)6900 ( 2401,-3061 ) ( 1 , 0)6000 ( 2401,-5461 ) ( 1 , 0)6000 ( 2401,-661 ) ( 1 , 0)6000 ( 5401,-3061 ) ( 1 , 0)2400 ( 5401,-3061)(-1 , 1)2400 ( 6776,-2986)(0,0)[lb ] ( 4501,-2086)(0,0)[lb ] ( 5551,-511)(0,0)[lb ] ( 7951,-511)(0,0)[lb ] we demonstrate which weights appear in the example of @xmath117 in figure [ fig : asym ] .
we show the weights of the adjoint representation .
the dots represent weights associated to @xmath80 , i.e. , from @xmath64 .
the circles represent weights associated to @xmath81 , i.e. , from @xmath82 .
it is important to note that _ reducible _ curves may also be wrapped by @xmath1-branes .
that is , two rational curves intersecting transversely at a point may be viewed together as a nodal rational curve .
these wrappings of reducible curves are required to obtain all the adjoint weights of the vector multiplets of the previous section .
the reducible curve @xmath123 has moduli space given by @xmath81since @xmath82 has a moduli space given by @xmath81 .
looking at figure [ fig : one ] we see that there is also a chain of rational curves in the class @xmath124 but note this this combination is invariant under the @xmath6 monodromy and so has moduli space given by @xmath80 .
these circles form the weights of @xmath125 of @xmath87 . ignoring the zero weights for now we see that the adjoint appears @xmath118 times and the @xmath125 of @xmath87 appears @xmath122 times . indeed
the zero weights also work out correctly .
a zero weight must represent an uncharged hypermultiplet and therefore a modulus .
we may use the work of wilson @xcite to demonstrate this .
wilson showed that a threefold containing a ruled surface @xmath73 has a moduli space which preserves this ruled surface only in codimension @xmath126 .
that is , there are @xmath126 deformations of the threefold which destroy this ruled surface . applying this to both ruled surfaces , we get @xmath127 deformations . on the other hand
, each @xmath87 adjoint contains a two - dimensional weight zero eigenspace while each @xmath125 contains a one - dimensional weight zero eigenspace .
thus the dimension of the weight zero eigenspace is @xmath128 , which simplifies to @xmath127 , as claimed .
the above construction may be easily generalized to @xmath103 : [ ieven ] let @xmath6 monodromy act on an i@xmath129 fiber in f - theory so that the central component of the fiber has moduli space @xmath80 and the outer components have moduli space @xmath81 . thus @xmath85 is a double cover .
then the resulting gauge algebra is @xmath103 and we have @xmath118 hypermultiplets in the adjoint representation and @xmath122 hypermultiplets in the @xmath130 representation ( which has dimension @xmath131 ) .
similarly @xmath132 with @xmath6 monodromy will yield an @xmath133 gauge algebra with @xmath122 hypermultiplets in the @xmath134 representation ( in addition to the usual @xmath118 adjoints ) .
also @xmath104 with @xmath6 monodromy will yield an @xmath135 gauge algebra with @xmath122 hypermultiplets in the vector @xmath136 representation . in the case of @xmath137 with @xmath138 or @xmath139 monodromy , a similar analysis yields @xmath122 hypermultiplets in the @xmath140 representation of @xmath141 .
this agrees with the various computations in @xcite where @xmath142 .
let @xmath81 be the double cover of @xmath80 branched at @xmath143 points .
thus @xmath144 .
then , for example , in the @xmath133 case we should have @xmath145 @xmath134 s .
this agrees with section 4.3 of @xcite by identifying the branch points with the @xmath146 zeroes of @xmath147 .
one might note that the above cases with @xmath6 monodromy may be combined into a simple rule as follows .
let @xmath148 be the simply - laced lie algebra which contains the actual gauge symmetry algebra @xmath149 as a subalgebra invariant under an outer automorphism given by the monodromy action .
( in fact , in each of the above cases , the outer automorphism which we use actually has order @xmath1 as an automorphism of @xmath148 , not merely order @xmath1 as an automorphism of the dynkin diagram . )
we may then decompose the adjoint representation of @xmath148 as follows @xmath150 where @xmath151 is a ( possibly reducible ) representation of @xmath149 on which the generator of the @xmath6 outer automorphism acts as @xmath152 .
the above rules may be combined to say that we obtain @xmath122 hypermultiplets in the @xmath151 representation . as we have already noted above in the case @xmath153
, the rule will be different if the monodromy group is not @xmath6 .
in fact , we will see more generally in section [ ss : small ] that if the outer automorphism representing the monodromy has higher order , the simple rule expressed in equation must be modified .
in addition to these `` non - local '' hypermultiplets coming from rational curves moving in families one may also obtain further hypermultiplets from collisions of curves of reducible fibers as in the i@xmath106-i@xmath107 collision discussed above .
note that some simple collisions may just induce monodromy without further contributions ( that is , their contributions are completely accounted for by the representation @xmath154 obtained in eq .
( [ eq : rul1 ] ) ) . as an example
we show in figure [ fig : so9 ] the generic case of a spin(9 ) gauge symmetry in f - theory .
( 7512,3045)(2389,-2803 ) ( 2401,-1561 ) ( 1 , 0)6000 ( 6901,-61 ) ( 0,-1)2700 ( 3301,-2761 ) ( 8551,-1636)(0,0)[lb ] ( 9901,-1561)(0,0)[lb ] ( 3451,-2761)(0,0)[lb ] ( 7051,-2761)(0,0)[lb ] ( 4501 , 14)(0,0)[lb ] ( 6751 , 14)(0,0)[lb ] this figure shows an @xmath155 fiber along a section of the hirzebruch surface @xmath156 .
this section has self - intersection @xmath157 and is denoted @xmath158 in the notation of @xcite . in the most generic situation ,
the rest of the discriminant locus of the elliptic fibration will consist of @xmath159 fibers along curves which intersect @xmath158 as shown in figure [ fig : so9 ] .
generically there are two types of collisions occurring with the frequencies shown .
a lengthy computation shows that the @xmath160 cubic collisions where @xmath161 and @xmath162 are affine coordinates in @xmath163 . ]
produce extra rational curves in the fiber but no monodromy while the @xmath164 transverse collisions produce monodromy but no extra rational curves .
thus the @xmath164 collisions produce @xmath165 of the vector @xmath166 s of @xmath167 .
an analysis of the rational curves in the cubic collision shows that we have @xmath160 spinor @xmath168 s . assuming @xmath169 , the existence of these spinors shows that the gauge group must be spin(9 ) .
this agrees perfectly with section 4.6 of @xcite .
similarly all the other results of @xcite may be confirmed .
finally in this section let us return to the case of @xmath6 monodromy acting on a curve of @xmath121 singularities to give an effective @xmath103 gauge symmetry .
we will consider the higgs branch in which we give expectation values to the hypermultiplets so as to break completely this @xmath103 gauge symmetry . recall that the geometry of moduli spaces of supersymmetric field theories in question imply that the dimension of this higgs branch should equal the total dimension of the representations of charged hypermultiplets minus the dimension of the gauge group which is broken .
we will observe that the geometry is in accord with theorem [ ieven ] .
we do this by describing the deformations after shrinking all of the curves in the fibers to zero volume . in section [ ss : below ] we will use the ideas introduced here towards the justification of our main assertion stated in the next section , which states that the gauge algebra in the case of @xmath170 with @xmath6 monodromy is @xmath103 .
we let @xmath171 be an unramified ( for simplicity ) double cover of @xmath80 .
in addition , we denote by @xmath172 the involution which exchanges sheets of the double cover . we now describe a local threefold @xmath2 containing the geometry of @xmath121 with @xmath6 monodromy over @xmath80 .
first we construct a threefold @xmath173 with an @xmath121 fibration over @xmath81 without monodromy .
then @xmath2 will be constructed as a @xmath6 quotient of @xmath173 .
we construct a singular threefold inside the bundle @xmath174 as the variety defined by the equation @xmath175 where @xmath176 and @xmath177 are in @xmath178 and @xmath179 is in @xmath180 .
this threefold has an @xmath181 singularity along @xmath81 , which is identified with the zero section of @xmath182 .
it has trivial canonical bundle by the adjunction formula .
in a moment we will construct a nowhere vanishing holomorphic @xmath16-form on it , giving independent verification of this fact .
the desired threefold @xmath173 is obtained by blowing up the singular locus @xmath183 times in the usual way to obtain a chain of @xmath184 ruled surfaces over @xmath81 . to obtain the desired geometry ,
we take the quotient of @xmath173 by the fixed point free involution obtained by using @xmath185 on @xmath81 while sending @xmath186 to @xmath187 .
note that the fibers of @xmath180 over @xmath188 and @xmath189 are canonically identified with the fiber of @xmath190 over @xmath191 , so this map makes sense . using the explicit description of the blowup and the fact that @xmath176 and @xmath177 are interchanged
, it follows that there is @xmath6 monodromy . to show that the quotient @xmath2 by this involution has trivial canonical bundle , it suffices to show that the involution preserves the holomorphic @xmath16-form on @xmath173 .
it suffices for our purposes to compute on the singular model .
let @xmath192 be any holomorphic @xmath17-form on @xmath81 .
then @xmath193 is a holomorphic @xmath18-form on @xmath182 with values in @xmath194 .
now , thinking of @xmath192 as a section of @xmath180 , we divide by @xmath192 to obtain the nowhere vanishing @xmath18-form @xmath195 on @xmath182 with values in @xmath196 which is independent of @xmath192 . finally , the residue @xmath197 is the holomorphic @xmath16-form on the singular model of @xmath173 .
it is clearly invariant under the involution .
the deformations of @xmath2 may be described as the deformations of @xmath173 in @xmath182 which preserve the involution .
the general deformation of @xmath173 ( up to change of coordinates ) is given by @xmath198 where the @xmath199 are sections of @xmath200 .
note that we are implicitly assuming @xmath201 to construct these deformations .
the invariance condition is that @xmath199 lies in the @xmath202-eigenspace of @xmath185 .
we now count parameters .
the @xmath203-eigenspace of @xmath204 has dimension @xmath205 , while the @xmath152-eigenspace has dimension @xmath206 .
thus the dimension of the higgs branch is @xmath207 this is exactly the number of parameters freed up from the @xmath118 adjoints and @xmath122 copies of the @xmath130 representation by higgsing an @xmath103 , as expected .
we have implicitly assumed that there are no global obstructions to the local deformations that we have constructed above , and our parameter count is consistent with this assumption .
while we appear to have given fairly general rules in the previous section for computing the massless particle spectrum of an f - theory compactification , there are actually many cases where the rules we have given so far become difficult to apply
. in particular , witten s analysis of the moduli space of rational curves in @xcite assumes that everything is smooth ( and reduced ) .
this need not be the case .
we will discuss some awkward cases which appear quite commonly in f - theory .
we begin with a discussion of a case which has caused some confusion in the literature that of @xmath6 monodromy acting on a gauge algebra of @xmath209 .
the approach of asking for the largest subalgebra invariant under an outer automorphism is not that helpful in this case .
putting @xmath98 in eq .
( [ eq : inv ] ) shows that @xmath210 is one possibility .
decomposing @xmath211 as @xmath212 and putting @xmath213 ( where @xmath100 is the @xmath101 identity matrix ) shows that @xmath103 is another possibility .
( this form of @xmath214 is nicely adapted to the action on the dynkin diagram . )
now it so happens that @xmath215 ( and that this is the largest dimension which can occur ) .
so which is the gauge algebra that f - theory actually wants ? using the approach of section [ s : lcy ] we immediately run into a problem .
one of the ruled surfaces , which we will denote @xmath71 , swept out by the reducible components of the fibers will look inevitably locally like the surface @xmath216 in @xmath217 .
we show a sketch of ( the real version of ) this surface in figure [ fig : nd ] .
each line @xmath64 in this surface crosses another line @xmath218 in the same class . in the case of @xmath170 for @xmath219
there will be other smooth surfaces .
this case is a little hard to visualize . in figure
[ fig : nd2 ] we show the case of monodromy acting on @xmath220 .
( in this case @xmath89 is the surface @xmath221 . ) the thick lines at the bottom of this sketch show the fiber over a branch point of @xmath85 .
the problem is that it is not clear what value we should give to @xmath222 since @xmath64 meets the singular line in @xmath71 .
the most nave interpretation of figure [ fig : nd2 ] is to completely ignore the fact that @xmath71 is singular and from the figure read off the intersection matrix @xmath223 this would imply that the gauge algebra is @xmath224 . in general , according to this argument , @xmath6 monodromy acting on @xmath170 would produce @xmath225neither of the possibilities suggested above ! it would be the most obvious algebra suggested by `` folding the dynkin diagram up '' by the outer automorphism .
we could get an @xmath103 cartan matrix from the case in question if we could somehow tie @xmath64 and @xmath218 in figure [ fig : nd ] together .
that is , somehow the rules of @xmath1-brane wrapping would have to assert that @xmath64 may not be wrapped alone
one must also wrap the intersecting curve @xmath218 simultaneously .
since @xmath71 is singular , there is no known reason for ruling such a possibility out . by considering the reducible curve @xmath226 as a single curve , we effective
replace @xmath71 by a simple ruled surface .
thus we would reduce figure [ fig : nd2 ] to figure [ fig : one ] .
that is , the case @xmath170 is reduced to @xmath121 and so we get @xmath103 under monodromy . at this point
therefore we do not really seem to know what the gauge algebra is .
the geometry seems to suggest @xmath225 or @xmath103 while the outer automorphism argument suggests @xmath103 or @xmath227 .
we will now give various arguments in support of the following assertion .
[ conj : x ] for an f - theory compactification on an elliptic threefold with a curve of @xmath228 fibers ( which locally suggests a symmetry of @xmath170 ) with @xmath6 monodromy , the resulting gauge symmetry is @xmath103 ( provided that the r - r fields are set to `` zero '' ) .
this assertion corrects some statements which appeared in earlier literature where it had been assumed the resulting gauge symmetry was @xmath227 for reasons we discussed above . as mentioned in the introduction
one can show that the spectrum of various f - theory models , such as point - like instantons on a @xmath229 singularity , is anomaly free for @xmath103 but inevitably would have anomalies in some cases had the gauge algebra contained @xmath227 .
note that the outer automorphism of @xmath230 which yields @xmath103 actually has order @xmath18 , since its square is conjugation by the matrix @xmath231 where @xmath100 is the @xmath101 identity matrix ; @xmath232 is not a central element of @xmath230 .
this outer automorphism of course still induces the required reflection of the dynkin diagram , as we explained near the end of section [ s : lcy ] .
this modifies the analysis which led to eq . as follows
. we let the outer automorphism of order @xmath18 act on @xmath148 and decompose into eigenspaces : @xmath233 each of which will be a representation of @xmath149 ( possibly reducible ) .
as before , the eigenspaces for eigenvalues @xmath234 can be accounted for by certain positive roots which are left invariant under the involution and by certain pairs of positive roots which are exchanged under that involution .
the moduli space for the former is @xmath80 and for the latter is @xmath81 ; when we consider the quantization of the d2-branes wrapped on the corresponding curves , we find @xmath235 half - hypermultiplets for each of the invariant roots , and @xmath236 half - hypermultiplets for each of the pairs .
since each pair contributes to both the @xmath203 and @xmath152 eigenspace , this adds up to a total of @xmath235 half - hypermultiplets in the adjoint representation of @xmath149 , and @xmath237 half - hypermultiplets in the representation @xmath154 .
we have yet to account for the representations @xmath238 and @xmath239 .
in fact , these are the roots which contain either @xmath110 or @xmath240 , which as we have argued above cannot occur as wrapped d2-branes at the generic point of the parameter curve @xmath80 if we are to reproduce the cartan matrix compatible with our main assertion ( or at least , such wrapped branes can not produce vector multiplets ) .
however , as we will see in section [ ss : small ] , the representations @xmath238 and @xmath239 _ do _ occur in the hypermultiplet spectrum perhaps because at the branch points of the map @xmath85 , @xmath110 and @xmath240 are identified and there is no apparent obstruction to wrapping the d2-brane there . to be more concrete concerning the case at hand , with @xmath241 , @xmath242 , and the outer automorphism determined by the @xmath214 in eq .
, we have @xmath243 the second exterior power of the fundamental of @xmath103 ( which has a trivial one - dimensional summand ) , and @xmath244 the fundamental representation of @xmath103 .
thus , the predicted spectrum is : * @xmath118 hypermultiplets in the adjoint representation * @xmath122 hypermultiplets in the second exterior power representation ( including its trivial summand ) , and * additional hypermultiplets in the fundamental representation .
in fact , an anomaly calculation @xcite predicts that there will be precisely @xmath245 such hypermultiplets .
one possible interpretation of this formula is that there are two fundamentals ( @xmath238 and @xmath239 ) associated to the parameter curve @xmath81 and an additional half - fundamental at each branch point . of @xmath80 which corresponds to the order four element of @xmath246 , and express things in terms of the genus of @xmath247 . ] a kodaira type iv fiber
would intrinsically produce an @xmath224 gauge symmetry but monodromy may act on this fiber producing the case of interest . at first sight
this might not look like such a good candidate for examination since @xmath249 !
however , the hypermultiplet spectrum will allow us to distinguish the cases .
consider the case of amassing point - like @xmath250-instantons on an orbifold point of a k3 surface along the lines analyzed in @xcite .
we will be interested in the case of four instantons and six instantons on a @xmath251 quotient singularity . from result 3 and figure 7 of @xcite we may deduce the spectrum without encountering any difficulties . in the case of four instantons , the @xmath138 singularity may actually be partially resolved to a @xmath6 singularity without affecting the particle spectrum .
this @xmath6 singularity may then be effected by a `` vertical '' line of @xmath252 fibers ( in the notation of @xcite ) . the six instanton case is effected by a vertical line of @xmath253 fibers .
the results are * for four point - like @xmath250-instantons on a @xmath138 singularity we have a nonperturbative enhanced gauge algebra of @xmath76 with hypermultiplets in four @xmath254 representations . * for six point - like @xmath250-instantons on a @xmath138 singularity we have a nonperturbative enhanced gauge algebra of @xmath255 with hypermultiplets as @xmath256 .
we may also produce exactly the same physics by using a vertical line of type iv fibers .
the configurations of curves of kodaira fibers in the base of the elliptic fibration is shown in figure [ fig : iv ] for the cases of four and six instantons respectively .
these diagrams are again similar to those presented in @xcite and represent the situation after the base has been blown up the requisite number of times .
the short curved lines represent fragments of the the curve of @xmath159 fibers .
( 10148,2967)(365,-2773 ) ( 377,-61 ) ( 1 , 0)3824 ( 1577,-661 ) ( 1 , 0)1500 ( 2777,-361 ) ( 0,-1)2100 ( 1577,-2161 ) ( 1 , 0)1500 ( 377,-2761 ) ( 1 , 0)3824 ( 5701,-61 ) ( 1 , 0)4800 ( 5701,-2761 ) ( 1 , 0)4800 ( 6301,-511 ) ( 1 , 0)1800 ( 6301,-2311 ) ( 1 , 0)1800 ( 7651,-2536 ) ( 4 , 3)900 ( 8326,-586 ) ( 0,-1)1650 ( 7651,-286 ) ( 4,-3)900 ( 2552,-811 ) ( 2552,-1111 ) ( 2552,-1711 ) ( 2552,-2011 ) ( 7976,-698 ) ( 8126,-751 ) ( 8149,-1988 ) ( 8044,-2086 ) ( 1352,-2236)(0,0)[lb ] ( 1352,-736)(0,0)[lb ] ( 2852,-2611)(0,0)[lb ] ( 3901 , 14)(0,0)[lb ] ( 3901,-2686)(0,0)[lb ] ( 10201,-2686)(0,0)[lb ] ( 10276 , 14)(0,0)[lb ] ( 6601,-436)(0,0)[lb ] ( 6601,-2236)(0,0)[lb ] ( 8401,-1486)(0,0)[lb ] ( 8626,-1936)(0,0)[lb ] ( 8626,-1036)(0,0)[lb ] let us begin with the case of six instantons on the right of figure [ fig : iv ] . the lines of type ii fibers produce no gauge symmetry enhancement .
the upper and lower diagonal lines of type iv fibers each collide once transversely with a line of type ii fibers and once non - transversely with the curve of @xmath159 fibers .
actually these collisions are very similar .
fibers can be turned into a line of type ii fibers intersecting the line of type iv fibers transversely . ]
each of these collisions produces @xmath6 monodromy in the type iv fiber producing the geometry of figure [ fig : nd ] .
thus each of these diagonal lines of type iv fibers produce an @xmath76 ( or @xmath257 ) gauge symmetry .
the remaining vertical line of type iv fibers collides with the other two lines of type iv fibers . resolving
this collision shows that no monodromy is induced .
thus this vertical line represents an @xmath224 gauge symmetry .
an analysis of the collisions shows that there would be an induced hypermultiplet in the @xmath258 of @xmath259 for each collision _ if there were no monodromy_. clearly from the desired spectrum above , this @xmath258 must break up as @xmath260 of @xmath261 .
this tells us immediately that the inclusion @xmath262 produced by the action of the monodromy produces a decomposition of the fundamental of @xmath224 via @xmath263 .
_ this rules out the natural embedding @xmath264 _ for which @xmath265 .
we are left with having to account for a hypermultiplet @xmath254 in each of the @xmath76 s .
this must come from the monodromy - inducing collisions of the diagonal lines of type iv fibers with the lines of type ii and type @xmath159 fibers .
as these collisions are all the isomorphic locally , each collision must produce a half - hypermultiplet @xmath254 .
this is in agreement with our comments concerning the @xmath238 and @xmath239 representations at the end of the previous subsection .
the collision point is the point around which the monodromy acts and so it associated with the location of the curve denoted @xmath266 in figure [ fig : nd ] .
the choice of associating this @xmath76 as the @xmath267 case of @xmath103 or @xmath225 differs as explained above by whether we view the positive root of @xmath76 as being associated to @xmath64 or to @xmath226 .
clearly in the latter case we have @xmath268 as divisor classes and so @xmath266 naturally generates the @xmath254 as required .
if only @xmath64 were identified as the positive root then @xmath266 would produce nothing new .
therefore we can only correctly identify the spectrum f - theory in the case of the geometry on the right - hand side of figure [ fig : iv ] if we take one of roots of the gauge algebra to be @xmath226 .
that is , there really does appear to be a rule in string theory which allows @xmath1-branes to wrap @xmath226 together but not @xmath64 or @xmath218 individually .
we can further verify our picture by considering the spectrum for four instantons on the left of figure [ fig : iv ] .
there are four collisions with the vertical line of type iv fibers , each producing monodromy .
thus , @xmath269 and there are @xmath270 branch points . following the arguments at the end of the previous subsection , we thus predict a spectrum consisting of one hypermultiplet in the @xmath271 representation ( from @xmath154 ) , and @xmath272 hypermultiplets in the fundamental representation ( i.e. @xmath273 ) .
this precisely agrees with the spectrum found above : there are four @xmath254 s of @xmath76 .
even the @xmath154 representation `` @xmath274 '' occurs correctly : it is the deformation _ la _
wilson @xcite of @xmath80 , or in physical terms of the heterotic string , it is the deformation of the @xmath138 singularity to a @xmath6 singularity which does not affect the spectrum as noted earlier .
the rules of @xmath1-brane wrapping are therefore rather unusual for the curves @xmath64 and @xmath218 .
as observed previously , away from the branch points , a @xmath1-brane can _ never _ wrap @xmath64 or @xmath218 individually .
however , as we have just seen , at the branch points where @xmath64 and @xmath218 coincide , the @xmath1-brane is allowed to wrap the curve .
in fact , when this wrapping is taken with both orientations , a hypermultiplet in the @xmath254 of @xmath76 is produced for each branch point . since @xmath64 lies in the singular surface @xmath71
it is perhaps not surprising that the usual rules of @xmath1-brane wrapping appear to break down . anyway , since this same @xmath71 appears as the `` end '' component for the higher rank gauge groups of this type , assuming string theory wraps branes around curves in @xmath71 in a similar way in that context , we arrive at our main assertion
. we can also give a different argument in favor of the @xmath103 gauge group in case @xmath80 has positive genus .
let us start with the case of @xmath276 fibers with @xmath6 monodromy . by theorem [ ieven
] , this leads to a @xmath277 gauge group and at least one adjoint hypermultiplet .
we will show in a moment that the corresponding @xmath278 singularity can be smoothed to an @xmath279 singularity with @xmath6 monodromy .
this corresponds to giving a nonzero vev to a semisimple element of the adjoint hypermultiplet , and the @xmath277 gauge group gets higgsed to some rank @xmath183 subgroup @xmath280 .
we still have to determine what @xmath149 without knowledge of the @xmath281 that acquires a vev .
clearly @xmath282 is possible , so we could have @xmath242 .
we now argue that @xmath283 is impossible .
there is no embedding of @xmath227 in @xmath277 for @xmath219 . * proof * this argument is due to r. zierau .
suppose that there were an embedding of @xmath227 in @xmath277 .
then the fundamental @xmath284 dimensional representation @xmath285 of @xmath277 would restrict to a representation of @xmath227 , which necessarily decomposes as a fundamental representation @xmath286 of @xmath227 plus a trivial representation .
the alternating form on @xmath285 restricts to a alternating form on @xmath286 . since @xmath286 is odd dimensional , this form is degenerate .
nullspace @xmath287 is invariant under @xmath227 , and is a proper subspace since the alternating form on @xmath286 is not identically zero .
this is a contradiction . the singular surface in question given by eq .
( [ eq : prob ] ) may be written as @xmath288 and thought of as a double cover of the @xmath289-plane branched along @xmath290 and doubly along @xmath291 .
it is the double branching that makes the surface singular .
we may smooth the surface by deforming to @xmath292 now the double branching has been split to @xmath291 and @xmath293 . for a fixed value of @xmath179
this process replaces a nodal rational curve by a smooth rational curve , where the nodal rational curve can be viewed as two rational curves intersecting transversely at a point .
that is , each pair of intersecting lines in figure [ fig : nd ] is replaced by a single line and the surface is smoothed .
this smoothing process is remarkably benign at the level of global geometry .
it is often possible to perform it even when the geometry of the ambient threefold , @xmath2 , is completely smooth at all times .
we can then derive the @xmath103 gauge symmetry indirectly as follows .
the existence of the deformation shows that @xmath1-branes are not allowed to wrap the individual lines of figure [ fig : nd ] .
the deformation converts each pair of lines into a single new line .
thus if physics is not discontinuously affected by the deformation , the @xmath1-branes contributing to vector particles must only be allowed to wrap the pairs of line in figure [ fig : nd ] together . as we have observed in the discussion immediately preceding the main assertion
, we can now conclude that we do indeed obtain @xmath103.-branes around the individual lines is not compromised by this argument .
when we deform the curve of @xmath279 singularities to a curve of @xmath181 singularities we may affect the geometry of some points on this curve .
thus hypermultiplets which were `` spread '' over the whole curve of singularities may be localized to isolated rational curves by this deformation process .
massless vectors can not come from such isolated curves . ]
there is of course a problem with the proof of our main assertion by this argument there may be global obstructions to such a deformation .
wilson s criterion suggests that such deformations only occur when @xmath294 , where @xmath80 is the base of the fibration of figure [ fig : nd ] as a ruled surface .
we address this in part by giving an example of a deformation when @xmath294 .
we return to the setup introduced at the end of section [ s : hyper ] . using the notation leading to eq .
( [ eq : su2kdef ] ) , we write the equation @xmath295 this gives @xmath170 with @xmath6 monodromy at the generic point of @xmath80 .
the deformation is simply @xmath296 to make sense of this , we have to say a little more about the blowup .
the @xmath279 blowups are determined by a procedure given in @xcite after choosing an ordering of the @xmath211 factors of @xmath297 .
choosing the @xmath298 factor to be in the middle , we obtain the desired geometry .
the last blowup creates a single ruled surface , which smooths out the singular component .
it is immediate to see from the description in @xcite that for generic @xmath199 this is a smooth deformation of the desired type . in this model ,
we have placed a restriction on the genus and have introduced localized matter at the zeros of @xmath299 . however
, if we are willing to accept that the process of gauge symmetry enhancement is dictated by local geometry then this example is enough to justify our main assertion .
finally we close with a note on the peculiar numerical predictions dictated by anomaly cancellation in the six - dimensional physics produced by f - theory compactified on @xmath2 .
this has been discussed in many places before ( for example @xcite ) and is often used as a method of enumerating the spectrum of hypermultiplets .
here we have outlined a systematic way of constructing the hypermultiplet spectrum and so the anomaly constraint becomes a peculiar numerical property of the geometry of an elliptic threefold . for completeness we will repeat the anomaly condition here .
we consider an elliptic fibration @xmath300 with a section .
let @xmath301 be the gauge group ( or algebra ) in six dimensions and @xmath302 the picard number of @xmath9 . then anomaly cancellation along the lines of @xcite yields the following @xmath303 where the hypermultiplets fall into representations @xmath304 of @xmath301 and @xmath305 is equal to @xmath17 if the representation is real or @xmath306 if the representation is complex or quaternionic ( pseudoreal ) .
note that the trivial representations also contribute to the sum .
these can be determined from the fact that the number of neutral hypermultiplet moduli are equal to @xmath307 . as an example consider the extreme case of @xmath308 corresponding to @xmath309 point - like @xmath250-instantons on a binary icosahedral quotient singularity in the heterotic string @xcite .
the threefold for the f - theory description of this has @xmath310 . applying the methods of sections [ s : hyper ] and [ ss : small ] to this threefold we also arrive at a spectrum of hypermultiplets of a @xmath311 for each of the @xmath312 copies of @xmath313 .
these representations are quaternionic .
equation ( [ eq : anom ] ) then reads @xmath314 the anomaly condition in eq .
( [ eq : anom ] ) has been verified in situations illustrating theorem [ ieven ] and our main assertion ( see the footnote in section 4 of @xcite ) .
note that one may obtain further conditions from the anomaly cancellation condition .
for example one may require the vanishing of coefficient of each `` @xmath315 '' term in the anomaly .
see for example @xcite .
it would be very satisfying to give a purely geometric proof of eq .
( [ eq : anom ] ) and the other anomaly conditions . ( a geometric proof which covers a wide variety of cases has recently been constructed @xcite . ) sadly at present the origin of this formula without using string theory is something of a mystery .
note that the existence of a section in the fibration @xmath300 is necessary for this to work .
if this requirement is not satisfied then there is no six - dimensional physics and the condition need not be satisfied @xcite ( for an example , see @xcite ) .
it is a pleasure to thank a. grassi , k. intriligator , j. morgan , r. plesser , g. rajesh , and r. zierau for useful conversations and insights .
is supported in part by a research fellowship from the alfred p. sloan foundation .
the work of d.r.m . is supported in part by by nsf grant dms-9401447 .
the work of s.k . is supported by nsa grants mda904 - 96 - 1 - 0021 and mda904 - 98 - 1 - 0009 , and nsf grant dms-9311386 .
s.k . also thanks the mittag - leffler institute for support during the early stages of this project . | the f - theory vacuum constructed from an elliptic calabi yau threefold with section yields an effective six - dimensional theory .
the lie algebra of the gauge sector of this theory and its representation on the space of massless hypermultiplets are shown to be determined by the intersection theory of the homology of the calabi
yau threefold .
( similar statements hold for m - theory and the type iia string compactified on the threefold , where there is also a dependence on the expectation values of the ramond
ramond fields . )
we describe general rules for computing the hypermultiplet spectrum of any f - theory vacuum , including vacua with non - simply - laced gauge groups .
the case of monodromy acting on a curve of @xmath0 singularities is shown to be particularly interesting and leads to some unexpected rules for how @xmath1-branes are allowed to wrap certain @xmath1-cycles .
we also review the peculiar numerical predictions for the geometry of elliptic calabi
yau threefolds with section which arise from anomaly cancellation in six dimensions . 0 |
the quest for understanding ultra - energetic events such as gamma - ray bursts , active - galactic nuclei , quasars etc . passes through understanding of the compact massive objects which are thought to be their central engine the rotating black holes , magnetars , binary neutron stars , etc .. for this , one needs to make a clear difference between theoretical , numerical and observational black hole ( bh ) and also , to keep account of the strengths and the weaknesses of the methods numerical and observational when working with the different types of objects . in the case of black holes candidates in the role of central engines ,
the theoretical energy extraction processes of ultra - relativistic jets ( such as the penrose process or the blandford - znajek process ) should depend on the rotational bh parameter ( @xmath0 ) , with a predicted threshold for effective jet launching to be around @xmath1 , with more conservative estimate at @xmath2 @xcite .
observationally , the measured rotational parameters seem to follow this requirement a list of measured @xmath0 can be found in table 1 in @xcite . from it , one can see that in general , the measurements correspond to the expectations .
there is , however , a catch .
the measurement of the spins of the black hole candidates is not independent it depends critically on the measurement of their masses , which comes with serious incertitude @xcite .
one such example is the black hole candidate gro j0422 + 32 .
its mass was estimated to @xmath3 @xcite but further analysis showed that its mass can be as low as @xmath4 @xcite thus putting it in the mass - interval of neutron stars ( @xmath5 ) ) . in the same article , a number of other black hole candidates have their masses re - examined and corrected , which poses a serious question about all the results on black hole candidates .
additionally , in @xcite it was demonstrated that numerous black hole candidates can be fitted by super - spinars ( objects with @xmath6 ) , thus showing that the x - ray spectrum is not enough to fix the spin of the black hole candidate .
considering the importance of the mass - spin measurement for the modeling of astrophysical jets , it is clear that there is an acute need for a better understanding of the theory of the rotating black hole .
the vacuum solution of the einstein equations which may be used for description of a rotating massive compact object ( black hole or super - spinar ) is the kerr metric .
the kerr metric depends only on 2 parameters the rotation @xmath7 to denote the theoretical rotational parameter , while the observational will be @xmath0 .
usually the mass of the kerr metric is taken for a constant so @xmath8 . ] and the mass @xmath9 . while there is a known theoretical problem in using the kerr metric for objects different from rotating black holes ( the problem of finding interior matter solution matching the exterior vacuum kerr solution ) , it is thought that it approximates well enough any distant rotating compact massive object , the precise type of the object being specified by the boundary conditions imposed additionally on the given by the metric spacetime to fix its physical content , like black holes , super - spinars , etc . .
confirming that the observational spectra indeed coincide with the kerr _ black hole _
spectra , is an important task facing astronomy and astrophysics . for a long time , the theory of perturbations was the only viable method of dealing with different type of matter in the kerr metric , or to study evolution of several interacting black holes . in the past decade , the perturbation theory has been replaced by the much more sophisticated , but also computationally expensive and complicated numerical relativity simulations method . the theory of perturbations , however , still has place in the study of this type of objects , allowing a clearer view of the purely gravitational properties of the system .
moreover , the important role of perturbative approach has been confirmed by full - gr numerical calculations . in the theory of perturbations , the characteristic ring - down of the black hole , as a result of perturbation ,
is dominating in later epochs through the complex frequencies called the quasi - normal modes ( qnms ) .
the qnms were used for the first wave - templates for gravitational detectors .
they were also observed in the late - time evolution of grmhd simulations .
finally , qnms were used in full gr simulations as a marker of the end of the merger phase @xcite . in @xcite
we presented the quasinormal and quasibound spectra of the kerr metric obtained by using a novel approach of solving the governing equations in terms of the confluent heun functions and imposing the required boundary conditions directly on them .
while this method allowed us to obtain both well - known old and also some new results , the novelty of the method makes its numerical stability uncharted . in the current paper
, we study in detail the numerical stability of those spectra which is crucial for further use of the method and the results .
in particular , we demonstrate the numerical stability with respect to deviations in the actual radial infinity and the position in the complex plane of the radial variable , proving that the so obtained quasinormal and quasibound spectra are indeed stable , as expected from the theory , and we show a counter - example of a spurious spectrum which despite being a root of the system of equations under consideration , does not satisfy the criteria for numerical stability and is thus non - physical .
the linearized perturbations of the kerr metric are described by the teukolsky master equation , from which under the substitution @xmath10 one obtains the teukolsky radial equation ( tre ) and the teukolsky angular equation ( tae ) @xcite ( where @xmath11 for integer spins and @xmath12 is the complex frequency ) .
the tae and tre describe on the same footing perturbations with different spin ( scalar waves , fermions , electromagnetic and gravitational waves ) without taking into account of the origin of the perturbation .
their numerical solutions have provoked the creation of numerical methods aimed at solving the non - trivial transcendental system of differential equations through different approximations , the most well - known being the method of continued fractions and that of direct integration performed by leaver and andresson ( @xcite , + @xcite . for a review on the methods for studying the qnm spectra ,
see + @xcite .
the tre and the tae are coupled differential equations for whose solutions one needs to impose the appropriate boundary conditions on the boundaries of the physical interval of interest .
depending on the boundary conditions , one obtains as solutions of system the quasinormal spectra ( black hole boundary conditions , see below ) , the quasibound spectra ( resonant waves ) or total - transmission modes ( waves moving in one direction , i.e. entirely to the left or to the right ) . all of them are discrete spectra of complex , possibly infinite number of frequencies ( numbered by @xmath13 ) , which dominate the late - time evolution of the object s response to perturbations . unlike normal modes , qnms do not form a complete set , because the mathematical problem is not self - adjoint .
therefore , one can not analyze the behavior of the system entirely in terms of qnms .
the qnms , however , can be used to discern a black hole from other compact massive object
such observational test was proposed in @xcite , and for a theoretical estimate of the em output for non - rotating bh @xcite . explicitly , the tae and the tre are : @xmath14 and @xmath15 where @xmath16 , @xmath17 , @xmath18 and @xmath19 .
we fix @xmath20 for the spin of the perturbation and @xmath21 are the two horizons of the bh .
the problem has two known parameter : the mass @xmath9 and the rotation @xmath7 and two unknown parameters : the frequency @xmath12 and the constant of separation of the angular equation @xmath22 .
depending on the sign of the imaginary part of the frequency one can consider both perturbations damping with time , @xmath23 , ( the qnm case ) and also perturbations growing exponentially with time , @xmath24 ( the qbm case )
. for details on the different classes of solutions of the tae and the tre em perturbations ( @xmath20 ) , see @xcite . here
we will present only the equations we use to solve the system . *
the equation governing the spectrum of the tae is * : @xmath25 where @xmath26 , @xmath27 and the parameters are : for the case @xmath28 : @xmath29 and for the case @xmath30
: @xmath31 * the equation governing the spectrum of the tre is * : @xmath32 where @xmath33 and the parameters are : @xmath34 , @xmath35 , @xmath36 , + @xmath37 , + @xmath38.$ ] here the @xmath39 is the confluent heun function the unique local frobenius solutions of the second - order linear ordinary differential equation of the fuchsian type @xcite , + @xcite with 2 regular singularities ( @xmath40 ) and one irregular ( @xmath41 ) , normalized to 1 at @xmath42 ( @xcite , + @xcite ) .
the singularities of the tae are the poles of the sphere ( regular ) and infinity ( irregular ) , those of the tre are the two horizons : @xmath43 ( regular ) and @xmath44 ( irregular ) .
in order to obtain the spectrum , one needs to impose specific boundary conditions . for the angular equation ,
one requires angular regularity .
the wronskian of two solutions regular around one of the poles , @xmath45 and @xmath46 , should become equal to zero , i.e. @xmath47=0 $ ] , i.e. @xmath48=\frac{\text{heunc}'(\alpha_1,\beta_1,\gamma_1,\delta_1,\eta_1,\left ( \cos \left ( \pi/6 \right ) \right ) ^{2})}{\text{heunc}(\alpha_1,\beta_1,\gamma_1,\delta_1,\eta_1,\left ( \cos \left ( \pi/6 \right ) \right ) ^{2})}+\notag\\ \frac{\text{heunc}'(\alpha_2,\beta_2,\gamma_2,\delta_2,\eta_2,\left ( \sin \left ( \pi/6 \right ) \right ) ^{2})}{\text{heunc}(\alpha_2,\beta_2,\gamma_2,\delta_2,\eta_2,\left ( \sin \left ( \pi/6 \right ) \right ) ^{2})}+ p=0 \label{wr1}\end{aligned}\ ] ] where the derivatives are with respect to @xmath49 , and we use @xmath50 ( the qnm should not depend on the value of the parameter @xmath51 in the spectral conditions ) . the parameter
p can be found in @xcite . for the radial equation
, one can impose the black hole boundary conditions ( bhbc ) : waves going simultaneously into the event horizon ( @xmath52 ) and into infinity ; or the quasibound boundary conditions ( qbbc ) : waves going simultaneously out of the event horizon ( @xmath52 ) and out of infinity @xcite .
those two conditions lead to : bhbc : for @xmath28 , the only valid solution is @xmath53 , while for @xmath54 , the solution @xmath53 is valid for frequencies for which @xmath55 . also , it requires that : @xmath56 ( the direction of steepest descent ) .
qbbc : for @xmath28 , the only valid solution is @xmath57 , while for @xmath54 , the solution @xmath57 is valid for frequencies for which @xmath55 . also , it requires that : @xmath58 .
the total - transmission modes ( ttm ) do not appear in the em case , in contrast to the gravitational case . here , we will use the @xmath59-method @xcite in its most direct form @xmath60 to vary @xmath61 , and afterwards we will check whether the bhbc or the qbbc are satisfied .
in order to obtain the spectrum @xmath62 , we will set the mass to @xmath63 and the radial infinity to @xmath64 ( where not stated otherwise ) and we will solve eqs . and .
the system of two equations featuring the confluent heun functions will be solved by the two - dimensional mller algorithm , developed by the team @xcite .
the algorithm is realized in maple code and the numbers presented below are obtained using maple 18 and 2015 .
the software floating point number ( the parameter digits ) is set to 64 ( unless stated otherwise ) , the precision of the algorithm to 25 digits .
note that the numerical error due to the integration is hidden in the evaluation of the confluent heun function which depends on the maple parameter digits `` . with current algorithms ,
one is not able to set the final number of stable digits in the heunc evaluation , but one can request the integration to be performed with certain precision , controlled by ' ' digits `` . in general , with our choice of the parameters of the root - finding algorithms , we can expect at least 15 digits of precision of the final result for the qnm / qbm , a number which can be increased by increasing further ' ' digits , but increasing also the computational cost of the calculations .
the numerical results of the qnm and qbm spectra are presented in + @xcite . here
we will focus only on the tests of the numerical stability of those modes .
we will consider two types of deviations : deviation in the absolute value of the radial variable and deviations in the argument of radial variable .
the qnm / qbm spectra should not depend on the radial variable at all , because @xmath12 is a separation constant of the teukolsky master equation .
if the spectra depend on @xmath65 , then either the actual numerical infinity is too low and we need to go further away from the horizon ( i.e. increasing @xmath66 ) , or the spectrum is not qnm / qbm , but a numerical artifact .
one should keep in mind that the qnm / qbm are expected to be stable on @xmath61 only in a limited region around the direction of steepest descent @xcite , which in principle can be a curve and not a straight line , as in the first approximation . as discussed in @xcite
, we obtained 3 types of spectra : the qnm , the qbm and a spurious spectrum .
the last one is not independent on the value of the radial variable , as we demonstrate below .
thus , it can not be considered a physical solution of the spectral problem , even though its @xmath67 are formal roots of the spectral system . to study numerically the stability of the qnm / qbm spectra
, we will work only with the @xmath68 case ( no rotation ) because it is computationally simpler .
the results are readily extended to the case with rotation ( @xmath69 ) . for the non - rotating case ,
the spectra are found directly from the radial equation
@xmath70 ( @xmath71 ) for the qnm ( qbm ) since in this case the angular solutions are known : @xmath72 , @xmath73 .
this means that we can work with the one - dimensional mller algorithm , used very successfully in the case of the jet - spectra of the black hole and naked singularities @xcite .
the main difficulty encountered in our previous works was the limited precision of the algorithms evaluating the confluent heun functions . in maple version 18 and
higher , those algorithms have been redesigned to allow arbitrary precision for heunc .
this facilitated significantly the current stability study and because of it , now it is possible to obtain the qnm and the qbm with arbitrary digits of significance .
the tre , just as the tae , has 16 classes of solutions , valid in different sectors of the complex plane .
while for the qnm the solution @xmath74 covers the needed sectors of the complex plane , for the qbm the above solution @xmath75 requires a significant increase in the parameter digits compared to @xmath74 to obtain the same precision . for this reason , we used one of the other , better convergent , solutions of the tre which are linearly independent of @xmath74 and have the same roots .
the new solution differs from @xmath75 only because in it @xmath76 .
similarly , one may find another solution which produces the qnm and it differs from @xmath74 by change @xmath77 . over all , there are 4 solutions from which one can obtain the qnm spectra and 4 for the qbm spectra , the only difference being the convergence of the root - finding algorithm .
as mentioned above the qnm / qbm are a discrete set of frequencies numbered @xmath78 in the direction of increasing absolute value of their imaginary part , thus denoted here as @xmath79 .
one can see the qnm / qbm frequencies with @xmath80 on fig .
[ fig1 ] a ) . on fig .
[ fig1 ] b ) we have plotted the boundary conditions corresponding to the 2 types of spectra . the study of the stability in this case can be seen on fig .
[ fig2]-[fig4 ] . on them , we focused on the following cases : 1 .
stability of the modes with respect to the parameter digits in maple + the results are on fig .
[ fig2 ] a ) where on the x - axis we have plotted the digits in use and on the y - axis the real part of the difference between two consecutive frequencies evaluated for those digits : @xmath81 for 4 different modes @xmath13 .
clearly the difference quickly falls to a very small number and for digits>60 for @xmath82 , one has @xmath83 .
the general tendency is that one needs higher digits for higher modes . 2 .
stability with respect to the absolute value of the radial variable + on fig .
[ fig2 ] b ) we have plotted @xmath84)|(|r|)$ ] in log - scale on the y - axis , where @xmath85)=\omega[n](r)-\omega[n](r')$ ] . from the plot it is clear that the frequencies depend on @xmath66 only near the horizon , with higher modes depending on @xmath65 stronger . for the modes on the plot ( @xmath86 ) , @xmath87 corresponds to the actual radial infinity after which the modes stabilize and remain the same with further increase of @xmath66 ( with the difference between two consecutive evaluations @xmath88 ) .
3 . stability with respect to the argument of the radial variable + the qbm and the qnm spectra are stable to the precision required by the root - finding algorithm in a very large interval of epsilon . if we consider @xmath89 , we find that the qnm frequencies with positive real parts appear for @xmath90 ( i.e. @xmath91 ) and remain stable in an interval on @xmath59 depending on @xmath13 for @xmath92 , @xmath93 $ ] and for @xmath94 $ ] , for @xmath95 , @xmath96 $ ] and for @xmath97 $ ] , for @xmath98 , @xmath99 $ ] and for @xmath100 $ ] .
the symmetry with respect to the imaginary axis in the dependence on @xmath59 is clear we see that for @xmath90 the frequencies with the positive real parts disappear and on their place , the root - finding algorithm finds the qnm frequencies with the negative real parts . for the qbm frequencies we observe the inverse behavior
those with the negative real part appear for @xmath101 .
this symmetry is due to the fact that the @xmath102 frequencies of each type are obtained with the same differential equation in a different region of the complex plane .
+ additionally , one can see that the interval on @xmath59 moves to the left with the increase of @xmath13 , which , combined with the need of increasing of the digits makes the detailed study of the @xmath59-dependence for the modes with @xmath103 computationally expensive . on fig .
[ fig2 ] c ) the dependence @xmath104 for @xmath92 is plotted , but it is representative for the behavior of all modes .
generally , what we observe is a plateau - like region of the @xmath59-interval where the frequencies remain stable with more than 20 digits , followed by a region where the frequencies start changing smoothly , but in a very steep manner .
all the numerical data point to the stability of the so - obtained frequencies in certain intervals on @xmath59 and @xmath66 , just as expected by the theory .
* near - extremal regime ( @xmath105 ) * let us define the angular velocity of the kbh : @xmath106 , with @xmath107 being the inner and the outer bh horizon ; the area of the kbh horizon : @xmath108 and the hawking temperature : @xmath109 .
hod in @xcite derived two analytical formulas for the frequency @xmath12 in the near - extremal regime in terms of @xmath110 and @xmath111 , i.e. @xmath112 and @xmath113 , where @xmath114 is a complex number .
as part of our test of the validity of our numerical results , we tested how well this theoretical formula approximates the so - obtained frequencies .
the results can be seen on fig .
[ fig3 ] . clearly , @xmath115 fits the numerical results better . on those plots we have used @xmath116 ,
although for the case @xmath117 the best fit is for @xmath118 .
the deviation from the theoretical formula is less than @xmath119 for @xmath120 , which is surprising considering the fact that the confluent heun function in this regime is expected to become numerically unstable .
the fact that @xmath121 does not approximate the frequencies better , even though @xmath122 , may be due to the low @xmath13 we are using , since the hod s original approximation was for infinitely damped frequencies ( i.e. @xmath13 very high ) .
* qnms in physical units * finally , we would like to calculate the qnm frequencies in physical units for two observationally well - studied objects .
from @xcite we know that : lmc x-1 has @xmath123 , @xmath124 and cygnus x-1 @xmath125 , @xmath126 . using the formula provided in the appendix
, we find the physical frequencies of the first two qnms to be : @xmath127 , + @xmath128 .
clearly , frequencies of less than a khz damping in a manner of milliseconds , are inaccessible to even the best earth radio telescopes because of the ionospheric absorption and scattering .
a possible lunar or space - based radio telescope could listen below @xmath129mhz , but even then , hearing the ringing of stellar black holes would be extremely difficult because of their matter - rich surrounding which will absorb the em spectrum .
such observations can be made only if the qnms are up - scattered by some unknown mechanism to frequencies observable on the earth .
this is an open question facing theoretical and observational astrophysics . on fig .
[ fig4 ] , one can see an example of a spurious spectrum corresponding to @xmath130 .
once again we have a discrete infinite spectrum , although the points on the figure are very closely spaced , and thus the central part looks like a line ( see fig . [
for @xmath131 those frequencies change , like in the qnm / qbm case , however the change is less pronounced . from fig .
[ fig4]b , it is clear that even though the spurious spectrum is very similar to the qnm / qbm , it does nt follow the same boundary conditions , with some points corresponding to infinity bhbc ( qnm ) ( @xmath132 ) and the rest to infinity qbbc ( qbm ) . because they are roots to both @xmath57 and @xmath53 , the boundary condition on @xmath133 is ambiguous in those points the functions @xmath75 and @xmath74 stop being linearly independent ( up to certain numerical precision ) . those roots have been confirmed with the 3 methods of heunc evaluation in maple the direct integration , the series expansion and the taylor expansion .
the study on the stability in this case is on fig .
[ fig4]-[fig6 ] : 1 .
stability of the modes with respect to the parameter digits + this case repeats our study of the digits dependence on qnm / qbm , so we did not plot this dependence again . as expected , the modes with a bigger imaginary part require higher value for digits .
2 . stability with respect to the absolute value of the radial variable + the results are on fig .
[ fig5 ] a ) and b ) , where we plotted 5 consecutive spurious modes . on fig .
[ fig5 ] a ) one can see how the absolute value of the frequency corresponding to each mode changes with the increase of @xmath66 ( the actual infinity ) .
one can see that it decreases , and upon a closer inspection , it seems like the imaginary part of the modes tends to a constant while the real part tends to zero .
we could not reach that probable limit up to @xmath134 , where the calculations become computationally too expensive .
one can see hints of such a limit also on fig .
[ fig5 ] b ) , where a complex plot of 5 consecutive modes in the interval for @xmath135 is plotted . on it @xmath64
is to the right and there is a hint of convergence for the different modes . for the moment ,
however , such a limit can not be definitely confirmed numerically .
3 . stability with respect to the argument of the radial variable + the study of the dependence of the modes on @xmath136 using the @xmath59-method can be seen on fig .
[ fig6 ] . on it , we have varied @xmath137 for one mode with @xmath138 ( 0.044511+.297566i ) and one with @xmath131 ( 0.023201 + 0.110030i ) . the interval for @xmath59 means that the argument is moved in a complete circle in the unit - sphere . as can be seen from the figure , the complex plot of those modes also draws a circle .
one can see also the boundary conditions in this case .
as mentioned in @xcite , the spurious modes are obtained from both equations ( @xmath75 and @xmath74 ) for the same value of argument the left branch for @xmath139 , the right for @xmath140 . on the plots we have used only points calculated from @xmath74 but the values obtained from @xmath75
coincide with them to a precision of @xmath141 .
4 . stability with respect to rotation + on fig .
[ fig5 ] c ) we plotted the behavior of one spurious mode with respect to an increase of the rotation from @xmath142 .
one can see the characteristic for the confluent heun functions loops , which occur also for the qnm modes . since those modes evolve with rotation all the way to the naked singularity regime , it is important to be able to distinguish them from the qnm / qbm modes by checking the @xmath66 dependence of the resulting frequencies and the boundary conditions .
this problem is the most serious for the qnm mode with @xmath143 which lies near a spurious mode , ( see fig .
[ fig1 ] ) , but considering how densely spaced is the spurious modes spectrum , it could occur also near other modes ( for @xmath144 $ ] ) .
it is important to note that those modes are roots not only to the radial equation , but to the spectral system .
they can be seen to propagate for @xmath69 and even in the naked singularity case for @xmath145 while the qnm / qbm spectra do disappear for @xmath145 or at least for the moment can not be located by the root - finding algorithm .
because of their persistence in the numerical results , the only way to recognize the spurious spectra is through the radial stability checks we performed .
it is important when discussing the numerical stability of a spectrum to make the difference between the numerical stability of the frequencies , the stability of the obtained spectrum and the stability of the spacetime with the given metric .
above , we emphasized that we discussed the numerical stability of the results as an essential test of how trustable they are .
when dealing with numerical solutions , we always work with certain precision and in a certain numerical domain , where the numerical algorithm is applicable .
the latter depends on both the theory of the function in question and the numerical routines evaluating it . in our case ,
the function we are working with is the confluent heun function . because of the irregularity of the singular point at infinity , there are no currently known series expansions working at infinity , and the available asymptotics of the solutions of the confluent heun equation is not properly developed , too .
thus , one has limited options when dealing with infinity . the method of the continued fraction is one possible way to deal with the problem . with it ,
one moves the singularities , so that : @xmath146 . in this way , the series is convergent from @xmath42 to the next nearest singularity @xmath147 which corresponds to infinity . with this method ,
one obtains a continued fractions for the coefficients of the recurrence , from which the @xmath148variable is missing .
the limits of applicability of this method are only partially known for example , it does not work for purely imaginary modes .
what is more important , this method does not address the fact that there are 16 classes of solutions of the confluent heun type valid in different parts of the complex plane .
there is also the problem with the branch cuts spectra needed for the full description of the perturbation ( @xcite ) .
finally , observations of the effect of the stokes and anti - stokes lines were noted in works utilizing the phase - amplitude method by andersson @xcite .
all those details point to the need of proper development of the theory of the confluent heun equation and its solutions , in order to have clarity about the area of applicability of the different methods and their physical interpretation . in our case , to test the numerical stability of our results , we compare the output of the 3 methods of evaluation of the confluent heun function currently implemented in maple direct integration , taylor expansion and continuation of the series expansion around @xmath147 , and then we check if small deviations in the parameters lead to significant changes in the frequencies .
we also plot the stokes and anti - stokes contours , as well as a 3rd complex plot of the function around a root for @xmath68 .
all those tests point to numerical stability of the qnm and qbm spectra and to instability of the third one .
it is important to note that the qnm problem in black holes is an analogue to the one of finding the poles of the s - matrix of the schrodinger equation . in quantum mechanics ,
those poles depending on the boundary conditions may correspond to bound states , virtual states or resonant states .
all of them correspond to different boundary conditions in which one may deal with solutions with exponentially decreasing or increasing asymptotics ( bound states or resonances ) . in astrophysics , because of the existence of the horizon on one of the boundaries of the s - matrix and , furthermore , the fact that all the frequencies are complex , one can not apply directly the same terminology and methods to the problem . for this reason , we work with quasinormal modes and quasibound modes , even though the bound states in quantum mechanics are purely imaginary modes . in quantum mechanics , the branch cut of
the s - matrix also plays an important and well - known role . in the theory of qnm
we have currently a very limited number of works on this approach ( @xcite , + @xcite ) .
apart from the numerical stability , one also may discuss the physical stability of the spectrum . here with
spectrum , we refer to the solutions of the spectral problem defined by the exact solutions of the differential equations and the boundary conditions we impose on them . as known from the theory of the differential equations , different boundary conditions describe different physical objects , even though they are governed by the same differential equation . in this context
, we can divide the different spectra into stable and unstable .
for example , the spectrum obtained from black hole boundary conditions in our work has been found to be physically stable , i.e. corresponding to perturbations which damp with time . on the other hand
, the spectrum corresponding to the quasibound modes is clearly physically unstable , because the perturbations grow with time .
there could be an other explanation , however . as mentioned in @xcite , the ranges of the boyer - lindquist coordinates for the kerr metric
are : @xmath149 . in the case of the standard substitution @xmath10 , the perturbation will damp with time for @xmath23 and @xmath150 . for @xmath151 the perturbation will grow exponentially for the same condition . for the qbm spectra , the situation reverses and for @xmath24 the perturbations will damp with time for @xmath151 and grow exponentially for @xmath150 .
considering the full range of the variables is not unusual .
for example in @xcite , the standard extension @xmath152 , equivalent to @xmath153 was already considered . also , it is known that the metric is invariant under the changes @xmath154 , meaning that the inversion of time can be considered as inversion of the direction of rotation . therefore ,
both qnm and qbm are valid frequencies in the whole range for the variables of the problem .
finally , the question of the stability of the physical object described by the metric is more complicated and has a long history ( for example @xcite ) . in many of those studies ,
linear perturbation theory was used with certain ambiguity in the definitions .
when one speaks of the stability of the schwarzschild black hole , one implies that the boundary conditions will be those of a black hole ( bhbc ) .
the schwarzschild metric , however , just like the kerr metric , in linear perturbation theory , allows one to impose other types of boundary conditions both the inverse bh conditions which we call qbbc or the total - transmission modes boundary conditions . in those cases , however , one deviates from the classical description of a black hole since the waves are not entering the horizon , but moving away from it .
this problem becomes even more complicated when we work with the kerr metric .
the kerr metric according to the theory has 2 horizons ( @xmath155 ) and one physical singularity ( @xmath156 ) , which is singularity of the scalar invariants of the metric .
this definition usually ignores the differential invariants where the horizons are also visible .
what is more important , the singularities visible for the differential equation when @xmath157 are only the two horizons and infinity .
because of this , following the theory of ordinary differential equations , one needs to impose the boundary conditions on two of those singularities in order to specify the singular spectral problem .
usually , one regards the inner horizon as physically not interesting and imposes the boundary conditions on the outer event horizon and infinity .
but there have been studies in which one imposes the boundary conditions on two horizons , which have shown that the region between the two horizons is _ unstable _ with respect to linear perturbations .
this poses the question of the physical interpretation of the different regions of the kerr metric and their analytical extensions . in any case
, there are 3 different types of boundary conditions bhbc , qbbc and ttm which correspond to different physical situations occurring on the background of the kerr metric and their study has its theoretical value .
the quasinormal modes are a well - known and seriously studied problem in gravitational physics .
they have been extensively used in different studies , but one significant problem has remained and it is related to the theory of the confluent heun functions and their roots . in series of articles ,
we have studied different physical configurations using the confluent heun functions in maple and we demonstrated that this method can lead to interesting scientific results , however , this novel approach also requires the development of new numerical tools and methods that are able to deal with the complexity of the heun differential equations and their solutions . in this article , by using the exact solutions of the tre and tae , we were able to impose different boundary conditions on them , describing qnm and qbm . by checking the numerical stability of the so obtained spectra in both @xmath66 and @xmath61
, we were able to sift out the physical solutions from the spurious ones , with clear theoretical justification .
this proves that the new approach indeed leads to a better understanding of the problem in question and also points to the huge potential which any improvement in the theory of the heun functions offers to physics .
the authors would like to thank dr .
edgardo cheb - terrab for useful discussions of the algorithms evaluating the heun functions in maple , and for continuing the improvement of those algorithms in the latest versions of maple .
the authors would like to thank d. sc .
svetlana pacheva for critical reading of the manuscript and the constructive comments .
this article was supported by the foundation `` theoretical and computational physics and astrophysics '' , by the bulgarian national scientific fund under contracts do-1 - 872 , do-1 - 895 , do-02 - 136 , and sofia university scientific fund , contract 185/26.04.2010 , grants of the bulgarian nuclear regulatory agency for 2013 , 2014 and 2015 .
p.f . posed the problem of the evaluation of the em qnms of rotating bhs as a continuation of previous studies of the applications of the confluent heun functions in astrophysics .
he proposed the epsilon method and supervised the project .
d.s . is responsible for the numerical results , their analysis and the plots and tables presented here .
both the authors discussed the results at all stages .
the manuscript was prepared by d.s . and edited by p.f ..
table [ table_n0_om ] presents some of the values obtained for the em qnm , converted to physical units using the relations : @xmath158 note that in those formulas a factor of @xmath159 is missing because the em qnms were obtained for @xmath160 and not for @xmath161 . then if @xmath9 is the mass of the object in physical units ( we use @xmath162 ) , @xmath163 the mass of the sun ( @xmath164 $ ] ) and @xmath165 , c=2.99792458\,10 ^ 8[m / s]$ ] , one obtains @xmath166,\tau^{phys}\approx \frac{0.4925\,10^{-5}\,m / m_{\odot}}{\im(\omega)}[s].\ ] ] & & & + & @xmath167[hz ] & @xmath168[ms ] & @xmath169[hz ] & @xmath170[ms ] & @xmath171[hz ] & @xmath172[ms ] + 0 & 802.1512449166 & 0.5325890917 & 802.1512449166 & 0.5325890917 & 802.1512449167 & 0.5325890917 + 0.2 & 804.9393652797 & 0.5343356142 & 849.8315682698 & 0.5388677452 & 763.6902591869 & 0.5299212818 + 0.6 & 829.2637578502 & 0.5526525743 & 996.9258848852 & 0.5772488810 & 704.6451920585 & 0.5313970112 + 0.98 & 884.6086875757 & 0.6427140687 & 1445.8841670353 & 1.2178343064 & 661.8628389523 & 0.5373275077 + + & & & + & @xmath167[hz ] & @xmath168[ms ] & @xmath169[hz ] & @xmath170[ms ] & @xmath171[hz ] & @xmath172[ms ] + 0 & 472.3043572607 & 0.0638131626 & 472.3043572599 & 0.0638131626 & 472.3043572609 & 0.0638131626 + 0.2 & 479.7880705182 & 0.0641624606 & 549.5382706431 & 0.0658306413 & 415.5213067645 & 0.0624052214 + 0.6 & 530.3546588895 & 0.0677548111 & 792.4033820584 & 0.0741706889 & & + 0.98 & 408.1590806664 & 0.0756272782 & & & &
+ & & & + & @xmath173 & @xmath174 & @xmath175 + 0 & 2.0000000000 + @xmath176 & 2.0000000000 + @xmath177 & 2.0000000000 + @xmath178 + 0.2 & @xmath179 & @xmath180 & @xmath181 + 0.6 & @xmath182 & @xmath183 & @xmath184 + 0.98&@xmath185 & @xmath186 & @xmath187 + + & & & + & @xmath173 & @xmath174 & @xmath175 + 0 & 2.0000000000+@xmath188i & 2.0000000000 + 4.82 @xmath189i & 2.0000000000 + 4.83 @xmath190 + 0.2&@xmath191 & @xmath192 & @xmath193 + 0.6&@xmath194 & @xmath195 & + 0.98&@xmath196&@xmath197& + , _ a numerically accurate investigation of black - hole normal modes _ , proc .
london a**439 * * no.1905 : 47 - 58 ( 1992 ) , _ outflows from accreting super - spinars _ , phys.rev.d . *
81*:104004,2010 , arxiv:1003.4821 [ gr - qc ] , _ rotating black holes : locally nonrotating frames , energy extraction , and scalar synchrotron radiation _ ,
apj * 178 * : 347 - 370 ( 1972 ) , _ black hole quasinormal modes : hints of quantum gravity ? _ , arxiv : gr - qc/0411025v1 ( 2004 ) , _ quasinormal modes of black holes and black branes _ , class .
quantum grav .
* 26 * 163001 ( 108pp ) ( 2009 ) , _ on the stability of the kerr metric _ , commun.math.phys .
* 221 * ( 2001 ) 659 - 676 , ( 2001 ) . _
the branch cut and quasi - normal modes at large imaginary frequency in schwarzschild space - time _
d. * 86 * , 024021 ( 2012 ) .
_ analytic investigation of the branch cut of the green function in schwarzschild space - time_,phys .
d * 87 * , 064010 ( 2013 ) , _ the quasinormal modes of the schwarzschild black hole _
london a * 344 * : 441 - 452 ( 1975 ) , _ how to tell gravastar from black hole _ , class .
* 24 * : , 4191 - 4206 , ( 2007 ) , arxiv:0706.1513v2 [ gr - qc ] , _ the electromagnetic signature of black hole ring - down _ , apj * 613 * : 492 - 505 , ( 2004 ) , arxiv : astro - ph/0310323 , ann .
buxelles * 92 * 53 , ( 1978 ) , 1978 ann .
buxelles 92 151 .
1995 _ heun s differential equations ed roneaux a _ , oxford : oxford univ . press , ( 1995 ) , _ on the equations governing the electromagnetic perturbations of the kerr black hole _ , proc .
london a * 349 * , 217 - 230 ( 1976 ) , _ black holes and gravitational waves .
iii - the resonant frequencies of rotating holes _ ,
apj:*239 * , 292 - 295 , ( 1980 ) , _ gravitational instabilities in kerr space - times _ , class.quant.grav .
* 25*:245012 , ( 2008 ) , _ black hole spectroscopy : testing general relativity through gravitational wave observation _ , class.quant.grav.*21*:787-804,2004 , arxiv : gr - qc/0309007 , _ quasi - normal modes and gravitational wave astronomy _ ,
gen.rel.grav.*40 * : 945 - 970 ( 2008 ) , arxiv:0709.0657v2 [ gr - qc ] ( 2007 ) , _ exact solutions of regge - wheeler equation and quasi - normal modes of compact objects _ , class
. quant . grav .
* 23 * 2447 - 2468 ( 2006 ) , arxiv:0509123v5 [ gr - qc ] , _ classes of exact solutions to the teukolsky master equation _ class . quantum grav .
* 27 * 135001 ( 2010 ) , arxiv:0908.4234v4 [ gr - qc ] , _ application of the confluent heun functions for finding the qnms of non - rotating black hole _ , phys .
d. * 84 * , 127502 ( 2011 ) , arxiv:1109.1532 [ gr - qc ] , _ solving systems of transcendental equations involving the heun functions .
_ , american journal of computational mathematics vol . *
02 * : 02 , pp.95 ( 2012 ) , _ are known maximal extensions of the kerr and kerr - newman spacetimes physically meaningful and analytic ? _ ,
arxiv:1205.5848 [ gr - qc ] , _ cgro j0422 + 32 : the lowest mass black hole ?
_ , astrophys.j.599:2 , 1254,2003 , arxiv : astro - ph/0308490 hod s. , _ slow relaxation of rapidly rotating black holes _ , phys.rev .
d * 78*:084035,(2008 ) , arxiv:0811.3806 [ gr - qc ] , _ gravitational radiation from gamma - ray burst progenitors _ , apj * 589 * : 861 - 870 , 2003 , _ constraints on the spacetime geometry around 10 stellar - mass black hole candidates from the disk s thermal spectrum _ , astrophys.j . *
797*:78,2014 , arxiv:1405.1508 [ gr - qc ] , _ quasinormal modes of black holes : from astrophysics to string theory _ ,
reviews of modern physics , * 83 * : 793 - 836 , issue 3 , ( 2011 ) , arxiv:1102.4014v1 [ gr - qc ] ( 2011 ) , _ super - spinning compact objects and models of high - frequency quasi - periodic oscillations observed in galactic microquasars _ , a&a * 572 * , a79 ( 2014 ) , arxiv:1410.6129 [ astro-ph.he ] , _ constraints on the spacetime geometry around 10 stellar - mass black hole candidates from the disk s thermal spectrum _ ,
astrophys.j.:*757 * , 36 , ( 2012 ) arxiv:1205.1805 [ astro-ph.he ] , _ extracting black - hole rotational energy : the generalized penrose process _
, physical review d * 89 * , 024041 ( 2014 ) , arxiv:1310.7499 [ gr - qc ] , , _ an analytic representation for the quasinormal modes of kerr black holes _ , proc .
london a*402 * : 285 - 298 ( 1985 ) , _ solutions to a generalized spheroidal wave equation : teukolsky s equations in general relativity , and the two - center problem in molecular quantum mechanics _ , j.math .
phys.*27 * ( 5):1238 ( 1986 ) .
_ unconventional gravitational excitation of a schwarzschild black hole _ , class.quant.grav .
* 20 * l217 ( 2003 ) , _ on the role of black hole spin and accretion in powering relativistic jets in agn _ , 2012 j. phys . :
. ser . * 355 * 012016 , _ observational evidences for spinning black holes : a proof of general relativity for spacetime around rotating black holes _ , ijmpd * 21 * , no .
11 ( 2012 ) 1250086 .
arxiv:1210.2441 [ astro-ph.he ] , _ a detailed study of quasinormal frequencies of the kerr black hole _ , phys.rev .
d * 55 * : 3593 - 3602 ( 1997 ) , arxiv : gr - qc/9610048v1 , _ special functions , a unified theory based on singularities _ ( oxford : oxford mathematical monographs ) ( 2000 ) , _ the spectrum of electromagnetic jets from kerr black holes and naked singularities in the teukolsky perturbation theory_,astrophysics and space science , * 332 * , pp.385 - 401 , arxiv:1002.0480v2 [ astro-ph.he ] , ( 2010 ) ] arxiv - kerr staicova d. , fiziev p. , _ new results for electromagnetic quasinormal modes of black holes _ , arxiv:1112.0310 [ astro-ph.he ] , _ new results for electromagnetic quasinormal and quasibound modes of kerr black holes _ , astrophysics and space science , june 2015 , * 358*:10 ( 2015 ) , arxiv:1412.4111 [ astro-ph.he ] , _ observational testability of a kerr bound in the x - ray spectrum of black hole candidates _ , 2010 class . quantum grav . * 27 * 075003 ( 2010 ) , arxiv:1002.0421 [ astro-ph.he ] , _ rotating black holes : separable wave equations for gravitational and electromagnetic perturbations _ , phys.rev.lett .
* 29 * : 1114 - 1118 ( 1972 ) , _ review article : the kerr metric _ , to appear in cqg for its `` milestones of general relativity '' focus issue to be published during the centenary year of gr , arxiv:1410.2130 .
_ analytic treatment of black - hole gravitational waves at the algebraically special frequency _ , phys .
d * 62 * 064009 ( 2000 ) , _ stability of the schwarzschild metric _ , phys .
d * 1 * , 2870 , ( 1970 ) , _ note on the stability of the schwarzschild metric _ , j. math .
* 20 * , 1056 ( 1979 ) | the proper understanding of the electromagnetic counterpart of gravity - waves emitters is of serious interest to the multimessenger astronomy . in this article , we study the numerical stability of the quasinormal modes ( qnm ) and quasibound modes ( qbm ) obtained as solutions of the teukolsky angular equation and the teukolsky radial equation with appropriate boundary conditions .
we use the epsilon - method for the system featuring the confluent heun functions to study the stability of the spectra with respect to changes in the radial variable .
we find that the qnm and qbm are stable in certain regions of the complex plane , just as expected , while the third `` spurious '' spectrum was found to be numerically unstable and thus unphysical .
this analysis shows the importance of understanding the numerical results in the framework of the theory of the confluent heun functions , in order to be able to distinguish the physical spectra from the numerical artifacts . |
gamma - ray bursts ( grbs ) that have a large fluence in the x - ray energy band ( 2 - 30 kev ) relative to the gamma - ray energy band ( 30 - 400 kev ) are receiving increased attention .
the burst and transient source experiment ( batse ) on board the _ compton gamma ray observatory _ detected 2704 grbs @xcite .
the spectra of 156 bright bursts exhibit a distribution of low - energy power - law indices @xmath13 whose centroid is @xmath14 , and a distribution of observed break energies @xmath15 whose centroid is @xmath16 230 kev @xcite , where @xmath17 .
here @xmath13 , @xmath18 , and @xmath5 are the slope of the low - energy power - law index , the high - energy power - law index , and the energy of the peak of the @xmath12 spectrum of the band function @xcite , an expression that satisfactorily represents the spectra of almost all grbs . in contrast , 36% of the bright bursts observed by _
ginga _ have peak energies @xmath5 in their photon number spectrum at a few kev and large x - ray to @xmath19-ray fluence ratios @xcite .
the _ beppo_sax wide field camera ( wfc ) detected events that are very similar to the soft _ ginga _ grbs ; these events have been termed `` x - ray flashes '' ( xrfs ) @xcite . >
-0.5 $ ] and 0.0 , respectively . ]
the energy flux of these xrfs lies in the range 10@xmath20 erg @xmath21 s@xmath22 and the low - energy photon index @xmath13 of their spectra ranges from @xmath233 to @xmath231.2 .
the sky distribution of xrfs is consistent with isotropy , and there is no evidence that the sources are galactic .
the xrfs have t@xmath24 durations between 10 and 200 sec .
the event rate of xrfs detected by the wfc is 3.4 events per year .
clarifying the connection between xrfs and grbs could provide a breakthrough in our understanding of the prompt emission of grbs .
@xcite made a detailed spectral comparison of grbs and xrfs , using a sample of eighteen grbs that were observed by batse and a sample of nine xrfs that were observed by both the wfc and batse . according to their joint analysis of wfc / batse spectral data ,
the low - energy and high - energy photon indices of xrfs are @xmath25 and @xmath26 , respectively , which are no different from those of grbs . on the other hand , xrfs have much lower values of @xmath5 than do grbs .
thus the only temporal or spectral difference between grbs and xrfs appears to be that xrfs have lower @xmath27 values .
kippen et al .
therefore suggest that xrfs might represent an extension of the grb population to events with low peak energies . analyzing 35 hete-2 grbs seen by fregate , @xcite
demonstrate that the spectral properties of `` x - ray rich '' grbs form a continuum with those of ordinary grbs and suggest that xrfs may represent a further extension of this continuum .
batse s low - energy threshold of @xmath4 20 kev made it difficult for batse to detect xrfs .
@xmath28 and @xmath29sax had the capability of detecting xrfs ; however , @xmath28 could not determine the direction of the burst and the @xmath29sax grbm had difficulty in triggering on xrfs .
consequently , these missions could not carry out in depth investigations of xrfs .
in contrast , hete-2 @xcite has the ability to trigger on and localize xrfs , and to study their spectral properties , using the wide - field x - ray monitor [ wxm ; 2 - 25 kev energy band ; @xcite ] and the french gamma telescope [ fregate ; 6 - 400 kev energy band ; @xcite ] , which have energy thresholds of a few kev . in this letter ,
we report the detection and localization of xrf 020903 by hete-2 @xcite and present the results of a detailed study of its properties .
since this event was extremely soft and there was very little signal ( a @xmath30 excess in the best selected energy range ) in fregate , we focus our analysis on the wxm temporal and spectral data for the event .
xrf 020903 was detected with the hete-2 wxm and the soft x - ray camera [ sxc ; 0.5 - 10 kev energy band ; @xcite ] instruments at 10:05:37.96 ut on 2002 september 3 @xcite .
the wxm flight localization was correct , but was not sent out because hete-2 was pointing at the galactic bulge region at the time and wxm triggers were therefore not being sent to the gcn in order not to overwhelm the astronomical community with x - ray burst localizations .
a gcn notice reporting the localization of the burst , based on ground analysis @xcite of the wxm data , was sent out 231 minutes after the burst .
the wxm localization can be expressed as a 90% confidence circle that is 16.6@xmath31 in radius and is centered at r.a .
= 22@xmath3249@xmath3325@xmath34 , dec .
= @xmath2320@xmath3553@xmath3159@xmath36 ( j2000 ) .
a localization of the burst based on ground analysis @xcite of the sxc data was distributed as a gcn notice about 7 hours after the burst .
only a one - dimensional localization was possible using the sxc data , but this significantly reduced the area of the localization region for xrf 020903 .
the improved localization produced by combining the sxc and wxm localizations can be described as a 90% confidence quadrilateral that is 4@xmath31 in width and @xmath431@xmath31 in length ( see figure 1 ) .
it is centered at r.a .
= 22@xmath3249@xmath3301@xmath34 , dec .
= @xmath2320@xmath3555@xmath3147@xmath36 ( j2000 ) , and its four corners lie at ( r.a . , dec . )
= ( 22@[email protected]@xmath34 , @xmath2320@[email protected]@xmath36 ) , ( 22@[email protected]@xmath34 , @xmath2320@[email protected]@xmath36 ) , ( 22@[email protected]@xmath34 , @xmath2321@[email protected]@xmath36 ) , and ( 22@[email protected]@xmath34 , @xmath2321@[email protected]@xmath36 ) ( j2000 ) .
detections of candidate optical and radio afterglows of xrf 020903 , and the host galaxy of the candidate optical and radio afterglows , have been reported .
@xcite discovered an optical transient within the hete-2 sxc + wxm localization region at r.a .
= @xmath37 , dec = @xmath38 , using the palomar 200-inch telescope .
these authors mention that the optical transient brightened by @xmath39 magnitudes between about 7 and 24 days after the xrf , and suggest that the re - brightening might be due to an associated supernova .
however , the optical transient apparently faded by over a magnitude only three days later @xcite .
spectroscopic observations of the optical transient , using the magellan 6.5 m baade and clay telescopes , detected narrow emission lines from an underlying galaxy at a redshift @xmath40 , suggesting that the host galaxy of the optical transient is a star - forming galaxy [ @xcite ; see also @xcite ] .
a fading bright radio source at the position of optical transient was detected using the very large array @xcite .
hubble space telescope observations of the xrf 020903 field reveal the optical transient and show that its host galaxy is an irregular galaxy , possibly with four interacting components @xcite .
these detections likely represent the first discoveries of the optical and radio afterglows , together with the host galaxy , of an xrf . in our analysis of the prompt emission of xrf 020903
, we apply a `` cut '' to the wxm photon time- and energy - tagged data ( tag data ) , using only the photons from the pixels on the three wires in the x - detectors ( xa0 , xa1 , and xa2 ) and the four wires in the y - detectors ( ya1 , ya2 ; yb0 , yb1 ) that were illumnated by the burst _ and _ that maximize the s / n of the burst light curve , in the same manner as we did for grb 020531 @xcite .
we use this optimized tag data when performing our temporal and spectral analyses of this event .
figure 2 shows the light curve of xrf 020903 in four wxm energy bands .
the time history of the burst in the 2 - 5 and 5 - 10 kev energy bands has two peaks .
clearly , there is no significant flux above 10 kev .
table 1 gives the t@xmath41 and t@xmath24 durations of the burst in the 2 - 5 kev , 5 - 10 kev , and 2 - 10 kev energy bands .
the duration of the burst is longer in the lower energy band ; this trend is similar to that seen in long bright grbs @xcite .
as we have seen , the light curve of xrf 020903 shows two peaks : the first occurring in the time interval 0 - 8 s , and the second occurring in the time interval 8 - 13 s. the s / n of the first peak is much higher than that of the second .
in addition , inspection of the burst light curve in the 2 - 5 and 5 - 10 kev energy bands suggests that the second peak is much softer than the first . for these reasons ,
we analyze the spectrum of the burst in three time intervals : 0 - 8 s , 8 - 13 s , and the total duration of the burst , 0 - 13 s. the background region we use is 40 seconds in duration and starts 45 seconds before the burst .
the wxm detector response matrix has been well - calibrated using observations of the crab nebula @xcite .
in the spectral fits , we include only the photons that registered on the four wires in the x - detectors and the five wires in the y - detectors that were illuminated by the burst , as mentioned above .
since the variation in the gain is not uniform at the ends of the wires in the wxm detectors @xcite , we use only the photon counts that registered in the central @xmath4250 mm region of the wires to construct the spectra of the burst .
we include all of the photons that register in the central regions of these wires ( i.e , we use the full 2 - 25 kev energy range of the wxm ) .
the relation between pulse height and energy in the wxm is non - linear and is different for each wire . in order to extract the strongest possible constraints on the parameters of the spectral models we consider , we treat each individual wxm wire separately but take the normalizations on all wires to be the same . for the same reason , we do not re - bin any of the pulse height channels in the wxm and in the fregate , and we carry out a set of fits for the total duration of the burst ( 0 - 13 s ) that include the spectral data from both the wxm and the fregate .
we use the xspec v11.2.0 software package to do the spectral fits .
table 2 presents the results of our time - resolved and time - integrated spectral analysis of the burst . in this analysis
, we consider the following models : ( 1 ) blackbody , ( 2 ) power - law model , ( 3 ) power - law times exponential ( ple ) [ the comp model in @xcite ] , and ( 4 ) band function @xcite .
the galactic value of @xmath43 in the direction of the burst is @xmath44 @xmath21 @xcite , which is negligible ( i.e. , it is undetectable in the wxm energy range ) .
furthermore , the wxm and fregate data do not request @xmath43 as a free parameter ( i.e. , introducing @xmath43 as an additional free parameter produces only a small change in @xmath45 ) .
there is therefore no need to include @xmath43 as a parameter in the spectral fits , and we do not do so . in the band model fits , we have fixed @xmath46 , which is a typical value for grbs , in order to better constrain the remaining parameters .
all of the models provide acceptable fits to the data ; i.e. , the data do not request models more complicated than a blackbody or a power - law .
however , essentially all grb spectra are well - described by the band function @xcite , and the analysis of @xcite shows that at least some xrf spectra are also well - described by the band function
. fits to the wxm data for all three time intervals using the power - law model give spectral slopes @xmath47 with high significance .
for example , comparing the minimum value @xmath48 corresponding to the best - fit value of the spectral slope @xmath49 and the value @xmath50 at @xmath51 for the power - law fit to the average spectrum of the burst ( i.e. , the time interval 0 - 13 s ) , we find that @xmath52 for one additional parameter .
thus , using the maximum likelihood ratio test , @xmath47 at the 99.3% confidence level . from this evidence ,
we conclude that the peak @xmath11 in @xmath53 lies near or below 2 kev , the lower limit of the energy range of the wxm detectors .
there is evidence of spectral softening between the first and second time intervals . in particular
, a power - law fit to the first time interval gives @xmath54 and @xmath55 , while a power - law fit to the second time interval gives @xmath56 and @xmath57 .
in contrast , a power - law fit to the first and second time intervals , but with @xmath58 gives @xmath59 and @xmath60 .
the first ( more complicated ) model includes the second model as a special case ( i.e. , the models are nested ) .
comparing @xmath61 for the two models , we find that @xmath62 for one additional parameter .
thus , using the maximum likelihood ratio test , there is evidence of spectral softening at the 98.6% confidence level .
figure 3 shows a comparison of the observed count spectrum and the count spectrum predicted by the best - fit power - law model , for the time intervals 0 - 8 s and 8 - 13s .
figure 4 shows the same comparison , except for the total duration of the burst ( 0 - 13 s ) .
table 3 gives the peak photon number and energy fluxes ( in 1 s ) and the fluence of xrf 020903 , assuming the power - law model .
using the power - law model that best fits the burst - averaged wxm plus fregate spectral data , we find fluences @xmath63 erg @xmath21 and @xmath64 erg @xmath21 where the quoted errors give the 90% confidence intervals .
thus the ratio of fluences @xmath65 = 0.7 $ ] , with a 90% lower limit of 0.3 , making this burst not only an xrf but the most extreme example of an xrf observed so far by hete-2 .
a comparison of the power - law and band function fits to the first peak , which has a much higher s / n than the second peak , provides modest evidence for an @xmath66 near 2 kev , the lower limit of the energy range of the wxm detectors .
specifically , we find that@xmath67 for one additional parameter , which means that the data requests the ( more complicated ) band function model at the 89% confidence level .
however , the evidence is clearly not of high statistical significance , and in this fit we fixed @xmath13 at @xmath231 , its typical value for grbs .
we therefore choose to place an upper limit on @xmath68 .
the appropriate model to use is the band function , since ( as we have already mentioned ) the spectra of almost all grbs and at least some xrfs are well - described by this function .
however , this presents a problem : the band function has two distinct ways of representing a power - law spectrum in the detector energy range .
first , it can do so by having @xmath69 , so that only the high - energy , pure power - law part of the band function is visible in the energy range of the detector .
second , it can do so by having @xmath70 and @xmath71 , where @xmath72 is the `` cutoff energy '' of the cutoff power - law that constitutes the low - energy part of the band function . in this limit
, the limiting power - law is actually the cutoff power - law , but the cutoff energy is so large that the curvature of the model is imperceptible in the detector energy range .
we solve this problem by developing a new statistical method .
this method uses a _ constrained _ band function which is parameterized by two quantities , @xmath66 and @xmath18 .
the _ constrained _ band function is perfectly able to make both pure power - law spectra and power - law times exponential spectra of the required curvature in the detector energy range , but only the high - energy part of the band function is allowed to produce a pure power - law spectrum .
we describe this new method in detail in appendix a. this method has general applicability to all instruments when the spectra of the bursts considered have @xmath66 near or below the low - energy threshold of the detector .
it is necessary to demonstrate that the photon index @xmath73 before applying this method .
in applying the constrained band function method to xrf 020903 , we jointly fit the wxm and the fregate data .
figure 5 shows the posterior probability density distribution for @xmath66 that we find using this approach . from this posterior probability density distribution
, we find a best - fit value @xmath74 kev , that @xmath75 with 68% probabilty , and that @xmath76 4.1 and 5.0 kev with 95% and 99.7% probabilities , respectively .
we conclude that the properties of xrf 020903 are very similar to those of long grbs , with the exception that the observed peak energy @xmath77 kev is @xmath78 times smaller .
the extremely low value of @xmath5 seen in xrf 020903 is similar to the smallest value found among the 9 xrfs whose spectra were determined by jointly fitting _
beppo_sax wfc and batse data @xcite .
we exclude the possibility that xrf 020903 is a type i x - ray burst ( xrb ) on the following grounds .
first , its galactic latitude is @xmath79 ( using the center of the combined wxm plus sxc error box ) , and there is no known persistent x - ray source or globular cluster in this error box . since type i x - ray burst sources lie in the galactic plane or in globular clusters , and have persistent x - ray emission , xrf 020903 is unlikely to be an x - ray burst on locational grounds alone .
second , the time history of xrf 020903 is not fred - like ( i.e. , it does not exhibit a fast rise and an exponential decay ) , while those of xrbs typically are .
third , although the blackbody model gives an acceptable fit to the spectra of the first and second peaks in the time history of xrf 020903 , the derived blackbody temperatures are @xmath4 1.0 kev .
these temperatures are lower than those of almost all type i x - ray bursts [ which typically have temperatures @xmath80 kev ; see , e.g. @xcite ] . for these reasons , we conclude that xrf 020903 is an xrf and not an xrb .
the extremely low value of @xmath66 observed for xrf 020903 is remarkable .
if the observed spectrum of xrf 020903 were the redshifted spectrum of a typical grb , the implied redshift would be @xmath81 , using the best - fit value of @xmath82 kev observed for xrf 020903 and the mean value of @xmath83 for the sample of 5500 spectra formed from the brightest 156 batse grbs @xcite
. a redshift of this magnitude would be hard to understand , and is certainly not expected if long grbs are associated with the collapse of massive stars @xcite .
it is also wildly inconsistent with the measured redshift @xmath8 of the host galaxy @xcite of the candidate optical @xcite and radio @xcite afterglows of xrf 020903 .
it is therefore difficult to attribute the low observed value of @xmath66 for xrf 020903 to cosmological redshift . in figure 6 ,
we plot xrf 020903 in the ( @xmath84,@xmath85)-plane , where @xmath86 and @xmath84 are the energy fluences of the bursts in the 7 - 30 and 30 - 400 kev energy bands .
for the value of @xmath86 and @xmath84 , we use the best - fit power - law model for the average spectrum of wxm and fregate .
also plotted in this figure are the 35 grbs whose spectra have been determined using hete-2 fregate data @xcite .
figure 6 shows that the properties of xrf 020903 are consistent with the relation between @xmath86 and @xmath87 found by @xcite . in figure 7 , we plot xrf 020903 in the ( @xmath88,@xmath89)-plane , where @xmath5 is the peak of the observed @xmath12 spectrum . for @xmath5
, we plot the 99.7% upper limit ( 5.0 kev ) .
the properties of xrf 020903 are consistent with an extension by two decades of the hardness - intensity correlation @xcite between @xmath84 and @xmath5 found by @xcite .
@xcite demonstrated that there is a relation between @xmath10 and the burst - averaged value of @xmath11 ( i.e. , @xmath11 for the time - averaged spectrum of the burst ) . assuming that the candidate optical and radio afterglows of xrf 020903 are indeed the afterglows of xrf 020903 , and
therefore that the redshift of the underlying host galaxy is the redshift of the xrf , we can calculate the isotropic - equivalent radiated energy @xmath10 and the upper limit on the burst - averaged peak energy @xmath11 of the @xmath12 spectrum in the source frame , in the same way as did @xcite .
figure 8 shows that the properties of xrf 020903 are consistent with an extension by a factor of @xmath9 in @xmath10 of the relation found by @xcite .
figures 6 - 8 provide evidence that xrfs , `` x - ray - rich grbs , '' and grbs form a continuum , and are therefore the same phenomenon . a variety of theoretical models have been proposed to explain xrfs [ see , e.g. , @xcite for a comparative discussion of several of these models ] . in the off - axis grb jet model ( yamazaki , ioka , & nakamura 2002 , 2003 ) ,
xrfs are the result of viewing the jet of an ordinary grb off - axis , so that relativistic beaming shifts the @xmath19-rays into the x - ray range . in the clean fireball model ,
xrfs are due to the relativistic pair plasma in the grb jet becoming optically thin much later than usual , at which time the relativistic bulk lorentz factor @xmath90 has already decreased to a relatively low value @xcite . in the dirty fireball model ,
xrfs occur when there is significant baryon loading of the grb jet , so that @xmath90 never reaches large values @xcite . in the universal jet model ,
xrfs are the result of viewing the grb jet off - axis , where @xmath90 is lower because of the structure of the jet @xcite . in the uniform jet model ,
the different properties of xrfs , `` x - ray - rich '' grbs , and grbs are due primarily to different jet opening angles , with larger jet opening angles associated with lower values of @xmath90 @xcite .
any such model of xrfs must reproduce the correlation found by @xcite between @xmath86 and @xmath84 , and the evidence we report in this paper for correlations between @xmath91 and @xmath5 , and especially , @xmath10 and @xmath11 the latter spanning nearly five decades in @xmath10 .
in this paper , we have reported hete-2 wxm / fregate observations of the x - ray flash , xrf 020903 .
this event was extremely soft : the spectrum had a best - fit peak energy @xmath92 kev and @xmath93 kev ( 99.7% probability upper limit ) and no photons were detected above @xmath4 10 kev .
the burst is shorter at higher energies , which is typical of long grbs .
we considered the possibility that the burst lies at very high redshift and that the low value of @xmath5 is therefore due to the cosmological redshift , and showed that this is very unlikely .
we find that the properties of xrf 020903 are consistent with the relation between @xmath94 and @xmath95 found by barraud et al .
( 2003 ) for grbs and x - ray - rich grbs , and are consistent with an extension by two decades of the hardness - intensity correlation @xcite between @xmath95 and @xmath5 demonstrated by the same authors . assuming that xrf 020903 lies at a redshift @xmath8 as implied by the host galaxy of the candidate optical afterglow of this burst , we find that the the properties of xrf 020903 are consistent with an extension by a factor @xmath9 of the relation between @xmath10 and @xmath11 in the source frame of the burst found by @xcite for grbs .
when combined with earlier results , the results reported in this paper provide strong evidence that xrfs , x - ray - rich grbs , and grbs form a continuum and are a single phenomenon . the correlation found by @xcite between @xmath85 and @xmath84 , and the evidence we find in this paper for correlations between @xmath95 and @xmath5 , and especially , @xmath10 and @xmath11 , provide strong constraints on any model of xrfs and x - ray - rich grbs .
we would like to thank the anonymous referee for comments and suggestions that materially improved the paper .
the hete-2 mission is supported in the u.s . by nasa contract nasw-4690 ; in japan , in part by the ministry of education , culture , sports , science , and technology grant - in - aid 12440063 ; and in france , by cnes contract 793 - 01 - 8479 .
k. hurley is grateful for _ ulysses _ support under contract jpl 958059 and for hete-2 support under contract mit - sc - r-293291 .
g. pizzichini acknowledges support by the italian space agency .
one of the authors ( ts ) is partially supported by the junior research associate ( jra ) program at riken .
in the spectral analysis of grbs , one occasionally encounters events ( such as xrf 020903 ) that are so soft that they present themselves as pure power - laws with power - law index @xmath73 in the energy range of the detector .
the natural interpretation of such spectra is that the break energy @xmath96 separating the two functional parts of the band function is near or below the lower boundary of the detector energy range .
1 . @xmath97 , so that only the high - energy , pure power - law part of the band function is visible in the energy range of the detector .
2 . @xmath98 , @xmath99 , where @xmath72 is the `` cutoff energy '' of the cutoff power - law that constitutes the low - energy part of the band function . in this limit
, the limiting power - law is actually the cutoff power - law , but the cutoff energy is so large that the curvature of the model is imperceptible in the detector energy range .
therefore , despite the fact that the numerical value of the power - law index is such that we are certain that we should be dealing with the high - energy part of the band function ( i.e. , the index is @xmath100 ) , the low - energy part of the function can `` horn in '' on the fit , altering the physical inferences drawn from the spectrum .
this situation is particularly a problem for the estimation of @xmath101 .
since we know that we are in case 1 , we also know that we ought to have at least a firm upper limit on @xmath11 , since @xmath11 is always necessarily less than @xmath96 , which is at the low end of the detector energy range . on the other hand , the case 2 limit implies @xmath102 .
unfortunately , the data do nt care which side of the band function makes the power - law , so no discrimination is possible between the two cases .
consequently , we ca nt constrain @xmath11 at all using a normal band function fit .
the approach we have chosen to deal with this situation in the case of xrf 020903 is to fit a _ constrained _ band function to the data .
that is , we consider a three - dimensional subspace of the full four - dimensional band function parameter space , choosing the subspace with a view to satisfying the following criteria : 1 .
it is perfectly possible to make both pure power - laws and cutoff power - laws of the any desired curvature in the detector energy range .
2 . only the high - energy part of the band function is allowed to produce a pure power - law .
we define the three - dimensional subspace in the following way : consider a band function parametrized by low- and high - energy indices @xmath13 and @xmath18 , and by a cutoff energy @xmath72 .
the well - known relation between @xmath72 and @xmath96 is @xmath103 .
we impose the constraint condition on our family of fitting functions @xmath104 where @xmath105 is some suitably chosen energy , in the general neighborhood where the grb has appreciable emission . @xmath96 and @xmath72 are then inversely related , and are equal to each other when both are equal to @xmath105 . on the other hand , when @xmath108 , then @xmath109 , and as @xmath71 , @xmath110 . in other words ,
when the low - energy part of the band function is trying to imitate a power law , the break energy becomes small enough to force the low - energy part of the function below the energy range of interest , where it can not be seen and therefore can do no harm .
any pure power - law work must thus be done by the high - energy part of the band function .
the resulting spectral function has three parameters ( including the scale ) , rather than four .
the two input shape parameters can be chosen arbitrarily from the set \{@xmath13 , @xmath18 , @xmath72 , @xmath96 , @xmath11}. the remaining parameters may then be determined by algebraic relationships .
we have found it most convenient to adopt @xmath11 and @xmath18 as our parameters .
the choice of @xmath11 is dictated by the necessity of estimating its value , or at least an upper bound on its value .
the choice of @xmath18 is convenient because one may then impose the parameter bound @xmath73 , which guarantees that the formal expression for @xmath11 may be meaningfully interpreted as the energy of the peak of the @xmath12 distribution .
this bound on @xmath18 is an important part of the specification of the fitting family of models .
were it not imposed , it would be possible for the formal expression for @xmath11 to exceed @xmath96 , so that at large values of @xmath11 the fit could always produce a @xmath111 power - law in the detector energy range .
the result would be an extended tail of constant @xmath45 for arbitrarily large values of @xmath11 .
figure 9 shows the constrained band function , with @xmath112 and @xmath113 kev , for different values of @xmath11 .
this figure shows that @xmath11 increases , @xmath96 also necessarily increases , so that @xmath72 is forced to smaller and smaller values by the constraint , which increases the curvature ( and the value of @xmath13 ) .
figure 10 shows the constrained band function , with @xmath114 kev and @xmath113 kev , for different values of @xmath18 .
the progression from some curvature at low energy ( @xmath115 ) to almost none ( @xmath116 ) is evident , as is the fact that as the curvature disappears , the resulting power - law is produced by the high - energy part of the band function .
figures 9 and 10 show that the constrained band function is perfectly able to make both pure power - laws , and cutoff power - laws with any desired curvature in the detector energy range .
figure 10 demonstrates that , in the constrained band function , a power - law spectrum is always produced by the high - energy part of the band function . 1
. @xmath105 must be low enough to prevent the low - energy part of the band function from making a power - law in the energy range of interest .
if @xmath105 were 1 gev ( say ) , then the band function would have no difficulty making @xmath72 large and @xmath117 , which is what we are trying to prevent by introducing the constraint .
so @xmath105 should be `` as low as possible .
@xmath105 must not be so low that we can not adequately fit any curvature that may exist in the spectrum .
if @xmath105 were 1 ev ( say ) , then whenever @xmath96 was in or above the energy range where the spectrum is appreciable , @xmath72 would be so tiny that the curvature of the model would be huge , much too large to fit the data well .
one way of choosing @xmath105 is to calculate its value using the best - fit parameters from a fit of a _
free _ band function , using @xmath118 .
this choice , which effectively chooses the unique constrained subspace of the full parameter space that contains the best - fit free band function , allows the constrained family of functions to optimally fit whatever curvature the data may seem to hint is required .
we must require that the inferences that we draw from the spectral fit should be robust , in the sense that they should not depend strongly on the specific choice of @xmath105 .
so the proper use of this constrained band model involves not only choosing a representative value of @xmath105 , but also varying @xmath105 in some reasonable range , to make sure that the conclusions about parameter estimates and bounds are unaffected by the choice of @xmath105
. figure 11 shows the constrained band function , with @xmath114 kev and @xmath119 , for different values of @xmath105 .
once again , as the low - energy curvature disappears , the resulting power - law is produced by the high - energy part of the band function .
figure 11 also shows that the shape of the spectrum in the detector energy range is insensitive to the specific choice of @xmath105 , within a reasonable range .
thus the conclusions about parameter estimates and bounds are unaffected by the choice of @xmath105 .
figure 12 shows the constrained band functions with parameters that best fit the 13 s spectrum of xrf 020903 , for different fixed values of @xmath105 .
this figure illustrates the fact that the shape of the best - fit model is essentially unchanged in the energy range of the wxm for choices of @xmath105 within a reasonable range .
finally , we give the algebraic relationships necessary to recover the remaining band function parameters assuming that @xmath11 and @xmath18 are given .
let @xmath120 .
then , @xmath121 in equation ( [ alpha ] ) , we have resolved the ambiguity in the choice of root of a quadratic equation by requiring that when @xmath122 , then @xmath123 , so that @xmath11 is in fact the peak energy of the @xmath12 distribution .
amati , l. , et al .
2002 , , 390 , 81 atteia , j - l , et al .
2003 , in gamma - ray bursts and afterglow astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 17 band , d. l. , et al .
1993 , , 413 , 281 barraud , c. , et al .
2003 , a&a , 400 , 1021 berger , e. , kulkarni , s. r. , frail , d. a. , soderberg , price , p. a. , fox , d. w. , harrison , f. a. , & yost , s. 2002 , gcn circ .
1555 chornock , r. , and filippenko , a. v. 2002 , gcn circular 1609 covino , s. , et al , 2002 , gcn circ .
1563 dermer , c. d. , chiang , j. , & bttcher , m. 1999 , , 513 , 656 dickey & lockman 1990 araa , 28 , 215 fenimore , e. e. , int zand , j. j. m. , norris , j. p. , bonnell , j. t. , & nemiroff , r. j. 1995 , , 448 , l101 graziani , c. , et al . 2003 , in gamma - ray bursts and astronomy , eds . g. r. ricker and r. vanderspek ( new york : aip ) , 114 heise , j. , int zand , j. , kippen , r. m. , & woods , p. m. , in proc .
2nd rome workshop : gamma - ray bursts in the afterglow era , eds .
e. costa , f. frontera , j. hjorth ( berlin : springer - verlag ) , 16 huang , y. f. , dai , z. g. & lu , t. 2002 , mnras , 332 , 735 kawai , n. , et al .
2003 , in gamma - ray bursts and afterglow astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 25 kippen , r. m. , woods , p. m. , heise , j. , int zand , j. , briggs , m. s. , & preece , r. d. 2002 , in gamma - ray bursts and afterglow astronomy , eds . g. r. ricker and r. vanderspek ( new york : aip ) , 244 lamb , d. q. et al .
2003 , submitted to ( astro - ph/0206151 ) lamb , d. q. & reichart , d. e. 2000 , apj , 536 , 1 lamb , d. q. , donaghy , t. q. , and graziani , c. 2003 , to be submitted to levan , a. , fruchter , a. , strolger , l. , burud , i. , & rhoads , j. 2002 , gcn circ .
1761 lewin , w. h. g. , van paradijs , j. & taam , r. e. 1993 , space sci .
, 62 , 223 lloyd - ronning , n. , petrosian , v. , & mallozzi , r. s. 2000 , , 534 , 227 mallozzi , r. s. , paciesas , w. s. , pendleton , g. n. , briggs , m. s. , preece , r. d. , meegan , c. a. & fishman , g. j. 1995 , , 454 , 597 mszros , p. , ramirez - ruiz , e. , rees , m. j. , & zhang , b. 2002 , , 578 , 812 monnelly , g. , et al .
2003 , in gamma - ray bursts and afterglow astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 25 mochkovitch , r. , daigne , f. , barraud , c. , & atteia , j. l. 2003 , asp conference series ( san francisco : asp ) , in press ( astro - ph/0303289 ) paciesas , w. s. , et al . 1999 , , 122 , 465 preece , r. d. , briggs , m. s. , mallozzi , r. s. , pendleton , g. n. , & paciesas , w. s. 2000 , , 126 , 19 ricker , g. r. , et al . 2002 , gcn circ .
1530 ricker , g. r. , et al .
2003 , in gamma - ray bursts and astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 3 rossi , e. , lazzati , d. , & rees , m. j. 2002 , mnras , 332 , 945 shirasaki , y. , et al .
2000 , spie , 4012 , 166 shirasaki , y. , et al .
2002 , spie in press .
shirasaki , y. , et al .
2003 , in gamma - ray bursts and astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 117 soderberg , a. m. , et al .
2002 , gcn circ .
1554 strohmayer , t. e. , fenimore , e. e. , murakami , t. , & yoshida , a. 1998 , , 500 , 873 villasenor , j. n. , et al .
2003 , in gamma - ray bursts and astronomy , eds .
g. r. ricker and r. vanderspek ( new york : aip ) , 33 woosley , s. e. , zhang , w. & heger , a. 2003 , , submitted ( astro - ph/0206004 ) yamazaki , r. , ioka k. & nakamura t. 2002 , , 571 , l31 yamazaki , r. , ioka k. & nakamura t. 2003 , , submitted ( astro - ph/0212557 ) zhang , b. & mszros , p. 2002
, , 571 , 876 zhang & mszros , p. 2003
, , in press ( astro - ph/0206158 ) | we report hete-2 wxm / fregate observations of the x - ray flash , xrf 020903 .
this event was extremely soft : the ratio @xmath0 , where @xmath1 and @xmath2 are the fluences in the 2 - 30 and 30 - 400 kev energy bands , is the most extreme value observed so far by hete-2 .
in addition , the spectrum has an observed peak energy @xmath3 kev ( 99.7% probability upper limit ) and no photons were detected above @xmath4 10 kev .
the burst is shorter at higher energies , which is similar to the behavior of long grbs .
we consider the possibility that the burst lies at very high redshift and that the low value of @xmath5 is due to the cosmological redshift , and show that this is very unlikely .
we find that the properties of xrf 020903 are consistent with the relation between the fluences @xmath6 and @xmath7 found by barraud et al . for grbs and x - ray - rich grbs , and
are consistent with the extension by a decade of the hardness - intensity correlation ( mallozzi et al .
1995 ) found by the same authors . assuming that xrf 020903 lies at a redshift @xmath8 as implied by the host galaxy of the candidate optical and radio afterglows of this burst
, we find that the properties of xrf 020903 are consistent with an extension by a factor @xmath9 of the relation between the isotropic - equivalent energy @xmath10 and the peak @xmath11 of the @xmath12 spectrum ( in the source frame of the burst ) found by amati et al . for grbs .
the results presented in this paper therefore provide evidence that xrfs , x - ray - rich grbs , and grbs form a continuum and are a single phenomenon .
the results also impose strong constraints on models of xrfs and x - ray - rich grbs . |
the 1809 kev line of @xmath0al was the first astronomical gamma - ray line from radioactive decay discovered in the galaxy @xcite , confirming a prediction made by @xcite and @xcite .
it is the brightest galactic line after the line from positron annihilation at 511 kev .
this first result was from _
heao 3 , which used high - resolution germanium detectors .
@xcite found that the line width was consistent with their instrumental resolution of about 3 kev , but the statistical significance of the result was low ( 4.8 @xmath1 ) . _
this isotope is thought to be produced by proton capture on magnesium , and therefore can occur in any site of nucleosynthesis where these two components are abundant @xcite .
environments suggested for its creation and release into the interstellar medium include type - ii supernovae , novae , and the winds of wolf - rayet and asymptotic giant branch stars . because its half - life is around 10@xmath3 yr , much shorter than scales of galactic evolution , its distribution on the sky reflects the current galactic distribution of the relevant parent events .
maps of the galactic 1809 kev emission ( * ? ? ?
* and references therein ) were made with the comptel instrument on the _ compton gamma - ray observatory and have given us the first detailed look at where the emission is concentrated .
the maps correlate better with early than late stellar populations , suggesting that supernovae and/or wolf - rayet stars are likely to be the primary contributors .
_ many other observations of this line , with high- and low - resolution balloon instruments and with low - resolution satellite instruments , have occurred since _ heao 3 ( see @xcite for a comprehensive review ) .
only one high - resolution measurement has had a statistical significance comparable to the _ heao 3 data and thus been able to further advance our understanding of the shape of the line .
this was performed by the gamma - ray imaging spectrometer ( gris ) balloon @xcite .
the authors found the line to be significantly broadened , with an intrinsic width of @xmath4 kev derived from a measured width of @xmath5 kev by subtracting their instrumental width of @xmath6 kev in quadrature .
the significance of their overall detection was @xmath7 , slightly higher than that of _ heao 3 .
_ _ _ this doppler broadening corresponds to isotropic velocities of 540 km s@xmath8 or a temperature of @xmath9k @xcite , and there is no model for how either could be maintained by a gas in the interstellar medium for anything approaching the lifetime of the isotope .
this result has stimulated interesting theoretical work centered on concentrating the @xmath0al in grains , which can maintain their birth velocities much longer than gaseous material and even be re - accelerated in supernova shocks @xcite .
the _ reuven ramaty high energy solar spectroscopic imager ( rhessi ) is a nasa small explorer satellite in a nearly circular @xmath10600 km orbit ( 96minute period ) with inclination @xmath1038@xmath11 .
its primary mission is to make high - resolution images and spectra of solar flares in the range from 3 kev to 17 mev @xcite .
its detector array is a set of nine high - purity coaxial germanium detectors cooled to liquid nitrogen temperature by a stirling - cycle refrigerator @xcite .
each of _ rhessi s detectors is segmented into a thin front segment facing the sun ( to stop hard x - rays ) and a much thicker rear segment meant to record solar gamma - rays that penetrate the front segment .
because the front segments have little effective area at 1809 kev for any incident angle , i did not use them for this analysis .
one of the nine detectors operates in an unsegmented mode with poor energy resolution , and is excluded as well .
the spacecraft rotates at close to 15 rpm about the axis pointing at the sun .
_ _ because the array is unshielded and the spacecraft is very light , the effective area of the array to highly penetrating 1809 kev photons is nearly independent of the direction of incidence , either in azimuth or zenith angle with respect to the spacecraft axis .
this has been verified with monte carlo simulations using the geant3 package with a highly detailed model of the spacecraft , which show a maximum deviation of @xmath1210% at any angle from the mean effective area averaged over all angles .
thus the annual rotation of the spacecraft with respect to the stars can not be used to modulate the galactic signal as was done with data from the _ solar maximum mission gamma - ray spectrometer @xcite . as an additional complication
, there is a background line in the instrument at 1809 kev due to cosmic - ray interactions with aluminum in the spacecraft .
this must be understood and subtracted before the galactic line can be studied .
_ the lack of directionality and strong background line can be overcome by using the earth as an occulter . to begin the analysis
, i divided nine months of rhessi data ( 1 march 2002 to 2 december 2002 ) into one - minute ( and therefore spin - averaged ) intervals .
i defined the `` inner galaxy '' as a box running from @xmath13 in galactic longitude and @xmath14 in galactic latitude .
`` source '' intervals were times when this entire box was unocculted by the earth , and `` background '' intervals were defined as times when it was entirely occulted . the rest of the data were discarded .
because the earth subtends less than half the sky , the set of source pointings is larger .
the total amount of time accumulated in the source spectra is 75.0 dy , and the total accumulation of background data is 33.7 dy .
data contaminated by the precipitation of magnetospheric electrons were identified with the onboard particle detector and were removed from consideration , as were data taken during the x4.8 solar flare of 23 july , the only flare observed so far with emission at energies approaching 1809 kev .
the background subtraction algorithm is adapted from one used to derive the most precise value to date for the flux of positron - annihilation radiation from the galactic center region : @xmath15@xmath16 ph @xmath17 s@xmath8 @xcite . in that case
the instrument was the burst and transient source experiment ( batse ) on the _ compton gamma - ray observatory ( cgro ) .
the principle is to sort the background spectra into a two - dimensional library based on the two most important parameters that control the background . for each source spectrum ,
a background spectrum is generated by selecting among the library spectra using the values of the controlling parameters for that source spectrum .
the computer code tracks the repeated contributions of each original raw background spectrum in order to propagate statistical errors correctly . _
all background lines with half - lives less than a few hours are fairly well controlled by choosing two particular parameters on which to sort the library : the longitude of the ascending node ( lan ) of the current orbit , and the orbital phase since the ascending node ( pan ) .
these parameters define a particular spot on the earth , so that the cosmic - ray flux ( which is controlled primarily by geomagnetic latitude ) is reproduced .
the lines with half - lives of seconds or less are proportional to the instantaneous cosmic - ray flux : this includes the 1809 kev background line due to cosmic - ray production of @xmath18 mg in aluminum , which decays with a 0.49 ps half - life @xcite .
unlike simple geographic or geomagnetic coordinates , however , lan and pan also contain a memory of the recent history of the spacecraft trajectory
most usefully , how recently and how deeply the spacecraft passed through the saa . thus matching lan and
pan also produces a good match for the saa - induced background lines , which have half - lives from minutes to hours .
we see such a line around 1811 kev that blends with the 1809 kev line .
the only candidate in the gamma - ray tables of @xcite is the 1810.77 kev line of @xmath19mn ( half - life 2.6 hr ) , presumably created by interactions of saa protons with iron near the detectors .
this line was suggested as a contributor to the background of the germanium spectrometer on the _ wind spacecraft @xcite , and the published 1809 kev background line from _ heao 3 @xcite shows an excess in the blue wing that is consistent with the @xmath19mn line appearing at a low level .
there is no very long - lived background line ( such as the 1275 kev line from @xmath20na ) near 1809 kev .
such a line would require a background - selection system which is cognizant of the total time since the start of the mission , rather than relying entirely on lan and pan . _ _
figure 1 ( left ) shows the rhessi background - subtracted spectrum of the galactic @xmath0al line . the smooth curve shown is 3% of the average background for comparison .
fitting the spectrum from 17901825 kev to a gaussian plus a linear background , i find a center energy of @xmath21 kev , 1.2@xmath1 from the expected rest energy of the line at 1808.65 kev @xcite .
the area of the line is @xmath22@xmath23 counts s@xmath8(11@xmath1 ) , which is 12.9% of the background line ( averaged over the entire set of background spectra interpolated from the background library ) .
its fwhm , which includes instrumental broadening , is @xmath24 kev .
the fit is good ( @xmath25 = 26.7 with 30 degrees of freedom ) . by fitting seven narrow background lines ranging from 186 kev to 2243 kev ,
i interpolate the intrinsic instrumental resolution to its value at 1809 kev and find @xmath26 kev .
i avoid the instrumental background line at 1809 kev itself because of the blend with the line at 1811 kev , which makes it much broader than the interpolation ( @xmath27 kev fwhm ) and raises its center energy to @xmath28 kev . subtracting the instrumental resolution in quadrature from the galactic line
gives an intrinsic width of @xmath2 kev , corresponding to velocities of @xmath10200 km / s or a temperature of @xmath29k . to demonstrate the quality of the background subtraction , i show in figure 1 ( right ) the region around a background line at 1014 kev .
this line , like the background line at 1809 kev , is due to the prompt decay of a short - lived isomer created by cosmic - ray interactions in aluminum .
thus , if it subtracts well , one should be able to assume that the 1809 kevbackground line does also .
the figure demonstrates subtraction better than the 3% shown for comparison . there is a hint of a residual 1014 kev line which
, when fit with a gaussian , gives a flux of @xmath30% of the background line .
the averaged cosmic - ray count rate in the detectors over all the periods used for source pointings is 0.46% higher than the same average over the background periods .
this is consistent with the slight undersubtraction hinted at in figure 1 ( right ) .
i find that subtracting out an extra 0.46% or 1.2% of the 1809 kevbackground line has no significant effect on the galactic 1809 kev line width , whether i use the natural form of the background spectrum ( which includes the 1811 kev line ) or whether i substitute an artificial single gaussian line at exactly 1808.65 kev with the instrumental resolution .
since the galactic line flux is 12.9% of the average background , these extra subtractions would reduce its value by 4% and 9% , respectively . to further demonstrate that the cosmic - ray - induced background line is well - subtracted , i divided the data set into two roughly equal parts by the value of the cosmic - ray count rate during each one - minute spectrum .
the rate averaged 158 counts s@xmath8 for the lower data set and 237 counts s@xmath8 for the higher data set , for a ratio of 1.50 .
the count rate in the total , unsubtracted 1809 kev line had a similar ratio of 1.42 , with 8.00@xmath23 counts s@xmath8 and 11.36@xmath23 counts s@xmath8 in the two sets , respectively .
once the backgrounds were subtracted , however , the values for the residual ( galactic ) line were @xmath31@xmath23 counts s@xmath8and @xmath32@xmath23 counts s@xmath8 , respectively , a ratio of only 1.09 and well within statistical agreement .
the derived galactic signal thus appears to be independent of the intensity of the background line .
the conversion of the count rate in the galactic line to a total galactic flux is strongly dependent on the assumed galactic distribution .
the effective area of the eight rear segments used , calculated at 1809 kev using the distribution of angles of incidence over the observation period assuming the source is concentrated at the galactic center , is 20.5 @xmath33 .
the incident flux is then @xmath34@xmath35 ph @xmath17 s@xmath8 for the artificial case of a point source .
since the effective area of the instrument is nearly independent of angle , the model distribution is important in determining the flux only because of occultation of parts of the distribution by the earth .
as noted above , i required the inner 60@xmath11 of galactic longitude to be visible in each pointing , but on average a considerable additional amount of the plane is also included .
a certain amount of this high - longitude plane emission also appears in the background intervals , however , and is subtracted off .
adding the gris best - fit value of 5.4 kev for the galactic line width in quadrature with _ rhessi s instrumental resolution would give a 6.78 kev width .
fixing the width at that value for the fit from 17901825 kev causes @xmath36 to increase by 17.3 to 44.0 ( now with 31 degrees of freedom ) .
the probability of _ rhessi s result being consistent with the gris best fit is then 4@xmath37 for one parameter of interest . _ _ the _ rhessi value for the galactic line s width is only marginally consistent with zero .
it is expected , however , that the interstellar @xmath0al will share in galactic rotation and display appropriate doppler shifts @xcite . @xcite
combined a three - dimensional model @xmath0al distribution derived from comptel data with a galactic rotation model to produce a map of the radial velocity versus galactic longitude ( integrated over galactic latitude and line of sight ) .
this map has a galactic bulge component with velocities up to @xmath1075 km / s and spiral arm components with velocities up to @xmath10150 km / s .
although an integration of this map over the inner galaxy has not been published , from the slices shown at different longitudes it appears that a integrated fwhm of not more than 150 km / s or about 0.9 kev would be obtained .
this is within 1@xmath1 of our result .
additional broadening due to more local motions of the @xmath0al - bearing gas or dust is possible , and of course the most recently - created portion of the @xmath0al can be traveling near its birth velocity in supernovae , etc .
, even if this portion is small .
_ _ rhessi s flux value of @xmath38@xmath35 ph @xmath17 s@xmath8 is comparable to previous measurements . a value of @xmath39@xmath35 ph @xmath17 s@xmath8 rad@xmath8 was considered consistent with the existing ensemble of data sets at the time by @xcite . for a uniform distribution in the plane
, this would imply about 1.4 rad for _
rhessi s effective field of view ( including the effect of subtracted flux in the background pointings ) .
realistic distributions fall off with longitude , however , and would require an even larger effective field of view .
@xcite , however , using the gris data , found higher fluxes per radian , with the value depending on the model used and ranging up to @xmath40@xmath35 ph @xmath17 s@xmath8 rad@xmath8 assuming the comptel distribution available at that time .
they suggested that comptel , which subtracts background by imaging with the compton telescope technique , would be insensitive to a spatially diffuse component that the large , integrating field of view of gris would accept . at first glance
the _ rhessi result seems to support this higher value , since the effective field of view required to match the gris data , 1.04 radians , is identical to the region i required to be visible for the source intervals . in future work
, these _ rhessi source and background fields of view , with the addition of data from those times in which the central radian is partially occulted , will be convolved with different model distributions to provide a normalization for each distribution and to help choose among them .
comparison with upcoming results from integral / spi , which has a smaller field of view , will allow us to test for the existence of very large - scale diffuse emission .
_ _ _ _ i would like to thank the rest of the _ rhessi team for making this work possible , particularly those who worked on the spectrometer and the data - analysis software .
g. h. share made valuable suggestions related to this analysis , and r. p. lin and h.
hudson thoughtfully reviewed the manuscript .
this work was supported by nasa contract nas5 - 98033 .
_ smith , d. m. , cheng , l. x. , leventhal , m. , tueller , j. , gehrels , n. , & fishman , j. 1997 , in aip conf .
410 , fourth compton symposium , ed . c. d. dermer , m. s. strickman , & j. d. kurfess , ( new york : aip ) , 1012 wheaton , w. a. , jacobson , a. s. , ling , j. c. , mahoney , w. a. , & varnell , l. s. 1989 , in aip conf .
186 , high - energy radiation background in space , ed .
a. rester , jr . &
j. i. trombka , ( new york : aip ) , 304 | observations of the central radian of the galaxy by the _ reuven ramaty high energy solar spectroscopy imager ( rhessi ) have yielded a high - resolution measurement of the 1809 kev line from @xmath0al , detected at 11@xmath1 significance in nine months of data .
the _ rhessi result for the width of the cosmic line is @xmath2 kev fwhm .
the best fit line width of 5.4 kev fwhm reported by @xcite using the gamma - ray imaging spectrometer ( gris ) balloon instrument is rejected with high confidence . _ _ |
a ubiquitous feature of the optical / uv emission from type 1 seyfert galaxies ( sy1s ) is the highly correlated variability of the continuum flux across a wide range of wavelengths ( e.g. , peterson et al .
1998 ) . when this behavior was first observed during the 19891990 broad line region reverberation mapping campaign for ngc 5548 ( peterson et al.1991 ; clavel et al . 1991 ) , it was immediately recognized that the upper limits for the relative lags between the optical and uv wavebands were too small for the variability to have been mediated by hydrodynamic processes in the putative accretion disk the differences in the characteristic radii for the emission in the respective wavebands were too large .
in fact , the upper limits for the interband lags of @xmath0days required signal speeds of @xmath1 which could not have occurred within the disk itself .
this fact and the observation that the optical / uv continuum became bluer as the flux increased led to the conclusion that the continuum variations were due to thermal reprocessing in the disk of higher energy radiation produced by a single varying source , such as the x - ray source posited to exist at the disk s inner regions , near the central black hole ( krolik et al .
1991 ; courvoisier & clavel 1991 ; collin - souffrin 1991 ) .
subsequent observations of other sy1s showed that highly correlated broad band optical / uv variability is generic to these objects ( e.g. , courvoisier & clavel 1991 ) .
recently , the thermal reprocessing picture has been challenged by simultaneous optical , uv , and x - ray observations of sy1s .
the first object so observed was ngc 7469 by the iue and rxte satellites over @xmath2days in 1997 ( nandra et al .
although the uv and x - ray light curves showed some similar variability characteristics both in amplitude and time scales , these light curves were not nearly as well correlated as would be expected if the uv emission was due to thermal reprocessing .
naively , the variations in the uv should follow those in the x - rays with some characteristic time delay , but the maxima in the uv light curves preceded similarly shaped maxima in the x - ray light curve by @xmath3days , while minima in both light curves occurred nearly simultaneously .
the 1998 observations of ngc 3516 by hst , asca , and rxte seem to be equally troubling for the thermal reprocessing model ( edelson et al.2000 ) .
these observations consisted of intensive monitoring in optical , uv , and x - ray bands over 3 days and so were able to probe correlations on far shorter time scales than any previous set of observations .
the x - rays showed very strong variations ( @xmath4% peak - to - peak ) , while the changes in the optical continuum were much smaller ( @xmath5% ) .
because of the accurate relative photometry afforded by the hst / stis ccds , the measurement uncertainties were sufficiently small to reveal significant variability that was strongly correlated across the optical bands .
hence , the original motivation for the thermal reprocessing model was still present , but as for ngc 7469 , the x - ray and optical flux variations were not strongly correlated on any relevant time scale . more recently , nandra et al . ( 2000 )
have re - examined the 1997 observations of ngc 7469 and found that while the relatively narrow band 210kev x - ray flux was nt well - correlated with the uv flux the spectral index of the x - ray emission was .
this led them to suggest that the uv emission is actually correlated with the _ bolometric _ x - ray flux . in this letter , we expand upon that idea and describe what is required for the multiwaveband phenomenology exhibited by ngc 3516 to be due to disk thermal reprocessing in the context of a thermal comptonization model for the x - ray emission .
most of the important details of the model are described in zdziarski , lubiski , & smith ( 1999 ; hereafter zls ) which the interested reader is encouraged to consult .
we will proceed by summarizing the significant features and results from that work .
this model was proposed by zls to explain the apparent correlation between the x - ray spectral index , @xmath6 , and the relative magnitude , @xmath7 , of the so - called compton reflection hump seen in the hard x - ray spectra of sy1s and x - ray binaries .
zls argued that this correlation results from the reprocessing and comptonizing interactions between the disk and the hot plasma believed to produce the x - ray continuum . expanding upon the work of previous authors ( e.g. , poutanen , krolik , & ryde 1997 ; zdziarski et al .
1998 ) , zls noted that the @xmath7-@xmath6 correlation could arise in a geometry in which the comptonizing plasma occupies a spherical region of radius @xmath8 centered on the black hole .
the accretion disk has an inner radius @xmath9 which penetrates this sphere at its equator by varying amounts .
this varying penetration could be due to changes in @xmath8 , @xmath9 , or both .
smaller values of @xmath10 result in a larger fraction of the x - ray photons from the sphere being intercepted by the disk .
the covering factor , @xmath7 , for compton reflection and thermal reprocessing is larger , and a larger fraction of the disk soft photons re - enter the plasma precipitating additional cooling via compton losses and thereby reducing the temperature of the plasma .
this results in a softer x - ray spectrum and a larger value for the photon index @xmath6 . enforcing energy balance under the assumption that the disk emits only reprocessed radiation ,
zls derived a @xmath7-@xmath6 relation which looks very similar to the observed correlations .
we follow zls by assuming that the region of comptonizing plasma is homogeneous and optically thin and that we can neglect internal viscous dissipation in the disk , at least initially .
reprocessing of incident x - rays by a razor - thin disk then gives a radial effective temperature distribution @xmath11^{1/4}. \label{t(r)}\ ] ] here @xmath12 is the total luminosity of the comptonizing plasma , @xmath13 is the stefan - boltzmann constant , and @xmath14 is the disk albedo . the dimensionless function @xmath15 is given by eq . 1 of zls . assuming local blackbody emission and that the optical portion of the observed spectrum originates at large radii
, we may use the asymptotic form @xmath16 in eq .
[ t(r ) ] to obtain the disk spectrum @xmath17 where @xmath18 , @xmath19^{1/4}$ ] , and @xmath20 is the photon energy . neglecting variations in the disk albedo and assuming that @xmath9 is fixed on the time scales of interest , a constant optical flux implies a constant value of @xmath21 . in the presence of variations of the x - ray luminosity
, this requires @xmath22 . for @xmath23 ,
i.e. , at higher frequencies , it turns out that the large radius limit of @xmath15 is a surprisingly good approximation for evaluating the disk spectrum since it enters as the 1/4-power in @xmath24 , and at @xmath25 , the function takes the value @xmath26 ( zls ) , and we have @xmath27 .
hence , for @xmath28 , adopting the asymptotic form for @xmath15 throughout and setting @xmath22 , all @xmath8-dependence drops out of eq .
[ fopt ] , and in this approximation , the resulting disk spectrum has a constant shape and magnitude for different values of @xmath8 .
the exact expression for @xmath15 flattens for @xmath29 , so when @xmath30 , the disk spectrum predicted by the full calculation will be somewhat softer , but not appreciably at the wavelengths which have actually been observed .
given the x - ray luminosity @xmath12 , we estimate the x - ray spectrum produced by thermal comptonization in the sphere . for this
we use a second result from zls which relates the amplification factor for comptonization , @xmath31 , to the ratio @xmath10 ( zls , eq . 2 ) . following zls , we use the empirical result of beloborodov ( 1999a ) which gives the x - ray spectral index as a function of @xmath31 : @xmath32 . to determine the overall magnitude of the x - ray spectrum , we use another empirical result of beloborodov ( 1999a ) to estimate the temperature of the comptonizing plasma , @xmath33 , where the compton parameter is given by @xmath34 and @xmath35 .
we assume a typical value for the thomson depth , @xmath36 , and solve the preceding equations to find the electron temperature @xmath37 .
we model the x - ray continuum as a power - law with an exponential roll - over , @xmath38 , where the cut - off energy is taken to be @xmath39 appropriate for a thomson depth @xmath40 .
the normalization , @xmath41 , of the x - ray spectrum is given by @xmath42 for @xmath43 , the precise value of the minimum photon energy is not very important so we have set it at a nominal value of @xmath44kev .
we have computed thermal comptonization and disk reprocessing spectra in this model for ngc 3516 .
the x - ray luminosity @xmath12 is found by fitting the disk spectrum to the mean optical fluxes measured by hst ( edelson et al .
we fix the disk inner radius at @xmath45 cm and plot the resulting spectral energy distributions ( seds ) in figure [ seds ] for values of the sphere radius in the range @xmath46@xmath47 cm ( solid curves ) . as claimed , the calculated disk spectrum is strikingly constant , with only relatively small variations in the bluer parts of the spectrum . for comparison ,
we plot the mean fluxes measured by hst . given the crudeness of this model , the quality of the fit is satisfactory even for the uv flux . since the uv flux was only monitored over a significantly shorter period than the optical data due to saa passages .
in addition , the uv data point , at @xmath48 , is in a part of the spectrum which may be contaminated by the broad wings of the siiv emission line ( see edelson et al .
2000 , fig . 3 ) .
unfortunately , similar contamination problems may exist for the optical data as well ( see edelson et al .
2000 , fig . 4 ) . ]
since we have thus far neglected local viscous dissipation in the disk , the black hole mass does not enter into the calculation of the disk emission .
however , the value of the disk inner radius can , in principle , be constrained by the optical and uv data .
the dashed curve in figure [ seds ] is the sed obtained for @xmath49 cm .
this shows that while the disk spectrum is relatively insensitive to factor @xmath5 changes in @xmath8 for a given optical flux , comparably sized changes in @xmath9 produce strong color variations in the uv .
therefore , in order to account for the uv flux , the disk inner radius can not be too much larger than @xmath50 cm .
for a central black hole mass of @xmath51m@xmath52(e.g . , edelson & nandra 1999 ) ,
this value for the disk inner radius is roughly consistent with innermost radius implied by asca observations of the broad fek@xmath53 emission line ( nandra et al.1999 ) .
despite the fact that the optical / uv continuum barely changes as a function of @xmath8 , the x - ray spectrum varies drastically . in the 210kev band in particular , the flux changes by factors of @xmath54 for the spectra shown . to relate this result more concretely to the x - ray observations of ngc 3516 , we plot in figure [ gamma_vs_flux ] the x - ray spectral index versus 210kev flux for a family of curves corresponding to different optical fluxes with the position on any given curve being parameterized by the value of @xmath8 .
we propose that the hst / asca / rxte monitoring observations of ngc 3516 can be explained in this model by changes in @xmath8 and accompanying changes in @xmath12 so that @xmath22 is approximately maintained , and the source takes a path in the @xmath6-@xmath55 plane which lies nearly along a curve of constant optical flux .
any significant variations in the optical / uv continuum will correspond to deviations from the constant flux curve and therefore from the @xmath22 relation .
furthermore , the changes associated with such departures will be broad band and will show the same color variations that have been previously associated with thermal reprocessing . for comparison , we have plotted in figure [ gamma_vs_flux ] values of @xmath6 and @xmath55 which we have obtained from fitting the publicly archived rxte ngc 3516 data for the observations in question .
although this simple model overpredicts the x - ray flux by a factor @xmath3 for the observed range of spectral indices , the correlation of spectral index and flux given by the model is strikingly similar to that of the data .
the flux discrepancy could be reduced somewhat by selecting a smaller value for the thomson depth @xmath56 since for a given spectral index ( and @xmath57-parameter ) that will imply a larger value of @xmath37 and more of the x - ray luminosity can be `` hidden '' at higher energies . alternatively
, a large reduction can be obtained by including contributions to the disk emission from local viscous dissipation .
this corresponds to adding a term @xmath58 inside the square brackets in the expression for the disk temperature ( eq . [ t(r ) ] ) where @xmath59 for a maximal kerr metric and @xmath60 for a schwarzschild metric .
since the viscous dissipation term is independent of @xmath8 , it acts essentially as a constant offset for the required value of the x - ray luminosity , and @xmath22 is still needed to maintain a constant optical flux . in order to match the observed x - ray fluxes , roughly 1/4 of the total disk luminosity
must be due to thermal reprocessing with the remaining 3/4 being due to viscous dissipation .
assuming @xmath61m@xmath52 and @xmath59 , we have solved for the energy balance condition and find that an accretion rate of @xmath62m@xmath52 yr@xmath63 is required to match the mean x - ray flux .
this solution is plotted as the thick solid curve in figure [ gamma_vs_flux ] .
although the energy balance condition we have used only restricts the properties of the comptonizing plasma , _ global _ energy balance is roughly satisfied by this solution . the bolometric x - ray luminosity which gives the observed mean x - ray flux and spectral index is @xmath64erg s@xmath63 , while the above accretion rate implies a disk luminosity of @xmath65erg s@xmath63 due solely to viscous dissipation . for a disk extending all the way to @xmath59 ( rather than being truncated at @xmath8 ) , the total implied luminosity is @xmath66erg s@xmath63 , and we have @xmath67 this would be expected if the accretion luminosity that would otherwise emerge as local black body emission from the inner regions of the untruncated disk were instead partitioned to the heating of the comptonizing plasma .
we note that we have arbitrarily set the thomson depth to @xmath36 in computing this solution . by fine tuning the values of @xmath56 and @xmath68
, we find that we can satisfy the global energy balance condition arbitrarily well while simultaneously matching the observed mean x - ray flux , spectral index and optical continuum levels .
the fact that an approximately constant optical / uv continuum can in principle result from thermal reprocessing of a highly variable x - ray flux is connected with the fact that the radial distribution of the flux incident upon the disk is well approximated by a power - law , even very close to the innermost disk radius . as we showed above , the constant flux constraint , in conjunction with the assumed constancy of the disk inner radius and disk albedo , causes the radius of the comptonizing sphere to drop out of the integral describing the disk spectrum .
however , this constraint does not demand that the shape of the incident flux distribution be a power - law .
the only requirement is that the relevant physical parameters affecting the incident flux can be considered independently from the expression which describes the _ shape _ of the distribution .
for example , consider the model proposed by beloborodov ( 1999b ) in which the comptonizing medium consists of a blob of plasma which is at some height @xmath69 above the disk .
the blob moves with mildly relativistic velocities either vertically away from or towards the underlying accretion disk , and the effective reflection fraction is controlled by relativistic aberration of the flux from the blob . using the formulae in beloborodov ( 1999b ) ,
the radial distribution of the flux incident upon the disk is given by @xmath70,\ ] ] where @xmath71 is the blob velocity in units of @xmath72 , @xmath73 , and @xmath74 is the disk radius in units of the blob height .
for the range of blob velocities required by this model to reproduce the observed @xmath7-@xmath6 correlation , @xmath750.7 , the factor in the square brackets which gives the shape of the flux distribution is largely insensitive to the exact value of @xmath71 . therefore , since the lower limit of the radial flux integral can be taken to be zero ( cf .
eq . [ fopt ] ) , a nearly constant optical flux in this model can be obtained if @xmath76 .
although neither of the aforementioned models offer any reason for the luminosity scalings which seem to be required by the ngc 3516 observations , any plausible explanation would still have to contend with why ngc 7469 and probably other sy1s violate it . therefore , we will not attempt to posit a physically motivated rationale for the specific relationship between x - ray luminosity and geometry required for any particular object , but a cursory examination of the important time scales may provide some insights . for ngc 3516 ,
the typical variability time scale exhibited by the 210 kev light curve is @xmath7730ks ( edelson et al .
assuming a radial thomson depth @xmath36 and @xmath28 , the cooling time is very short : @xmath78s , where @xmath79 is the total number of electrons in the sphere with mean temperature @xmath80k .
unfortunately , the thermal expansion time for a sphere of this size is fairly long , approximately @xmath81ks .
this assumes that the protons and the electrons are at the same temperature .
if we have a two - temperature plasma in which the protons possess a virial temperature of @xmath82k , then the expansion time scale is reduced substantially .
alternatively , this could also be achieved if we have a pair plasma instead .
we have assumed for simplicity that the thermal emission from the disk is locally blackbody .
detailed modeling of bare disk atmospheres appropriate for sy1s indicates that this is a moderately reasonable approximation in the optical , but significant deviations can occur at higher frequencies ( e.g. , hubeny et al .
2000 , 2001 ) .
this appears to remain true for disks powered exclusively by external x - ray illumination ( sincell & krolik 1997 ) .
our treatment of the reprocessed flux in the optical is therefore probably fairly good , but the full seed photon spectrum illuminating the comptonizing plasma will not be very accurate .
more detailed modeling of the disk spectrum will likely affect the specific values we deduce for the disk inner radius , x - ray luminosity , and comptonizing sphere radius , but it will not change our general conclusions .
the existing simultaneous optical , uv , and x - ray data plus the global energy balance condition ( eq . [ global_energy_balance ] ) provide just enough information to infer unique values for the five model parameters , @xmath9 , @xmath8 , @xmath37 , @xmath56 , and @xmath68 , required to describe the entire spectral energy distribution .
therefore , a crucial test of this general scheme for linking thermal comptonization and thermal reprocessing will be provided by simultaneous optical , uv , and _ broad band _ x - ray monitoring observations which include reliable measurements of the thermal roll - over at energies @xmath83kev .
these latter measurements are necessary for making estimates of the x - ray luminosity @xmath12 which can then be compared with the luminosities required by the model . forthcoming _
integral _ observations of sy1s will provide some of the necessary spectral coverage to address this issue , but what is really required is a hard x - ray / soft gamma - ray telescope which can measure the x - ray spectrum and thermal roll - over on time scales comparable to the shortest variability time scales seen from these objects .
we thank andzrej zdziarski and the anonymous referee for helpful comments .
jc was partially supported by nasa atp grant nag 5 - 7723 during the course of this work .
ob acknowledges support from nasa grant nag 5 - 7075 and nsf grant phy 99 - 07949 .
this research has made use of data obtained from the high energy astrophysics science archive research center ( heasarc ) , provided by nasa s goddard space flight center .
beloborodov , a. m. 1999a , in high energy processes in accreting black holes , eds .
j. poutanen and r. svensson , ( san francisco : astronomical society of the pacific ) , 295 beloborodov , a. m. 1999b , apj , 510 , l123 clavel , j. , et al .
1991 , apj , 366 , 64 collin - souffrin , s. 1991 , a&a , 249 , 344 courvoisier , t. j .-
, & clavel , j. 1991 , a&a , 248 , 389 edelson , r. e. , & nandra , k. 1999 , 514 , 682 edelson , r. e. , et al .
2000 , apj , 534 , 180 hubeny , i. , agol , e. , blaes , o. , & krolik , j. h. 2000 , apj , 533 , 710 hubeny , i. , blaes , o. , krolik , j. h. , & agol , e. 2001 , apj , in press ( astro - ph/0105507 ) krolik , j. h. et al .
1991 , apj , 371 , 541 nandra , k. , et al .
1998 , apj , 505 , 594 nandra , k. , george , i. m. , mushotzky , r. f. , turner , t. j. , & yaqoob , t. 1999 , apj , 523 , l17 nandra , k. , et al .
2000 , apj , 544 , 734 peterson , b. m. , et al .
1991 , apj , 368 , 119 peterson , b. m. , et al .
1998 , pasp , 110 , 660 poutanen , j. , krolik , j. h. , & ryde , f. 1997 , mnras , 292 , l21 sincell , m. w. , & krolik , j. h. 1997 , apj , 476 , 605 zdziarski , a. a. , et al .
1998 , mnras , 301 , 435 zdziarski , a. a. , lubiski , p. , & smith , d. a. , 1999 , mnras , 303 , l11 ( zls ) | we present an application of the thermal comptonization / disk reprocessing model recently proposed by zdziarski , lubiski , & smith .
we show that the absence of strong optical variations in the presence of strong concurrent x - ray variations , similar to those found by hst / rxte monitoring observations of ngc 3516 , can be explained by changing the geometry of the comptonizing plasma rather than the accretion disk itself .
the total x - ray luminosity of the comptonizing plasma must decrease as its spatial extent increases .
in contrast , the disk inner radius must be roughly fixed in order not to produce optical / ultraviolet color variations stronger than observed . by including emission due to internal viscous dissipation in the disk
, we can roughly match the optical and x - ray flux levels and variability amplitudes seen from ngc 3516 during the hst / rxte campaign . |
van maanen s star , also known as van maanen 2 was discovered in 1917 by adrian van maanen @xcite .
its discovery was quite accidental , as van maanen was searching for common proper motion companions to hd4628 ( lalande 1299 ) and noticed the even larger proper motion of a nearby star . at 4.4 pc ,
van maanen 2 is the nearest single white dwarf @xcite .
@xcite reported the astrometric detection of a substellar companion to van maanen 2 through analysis of _ hipparcos _ data .
the data suggested an orbital solution with a lower bound companion mass of @xmath8 , a period of 1.57 yr , and a maximum separation on the sky of @xmath9 .
based on the published parameters , a binary orbital calculator ( a. ghez 2004 , private communication ) indicated a separation of @xmath10 at position angle @xmath11 near the middle of january 2004 .
this paper presents the results of a direct imaging search carried out during that time frame .
van maanen 2 was observed using the facility adaptive optics system @xcite and the nirc2 camera at keck observatory on 12 jaunuary 2004 .
five dithered images were obtained through an @xmath12 ( @xmath13 m ) filter . to reduce thermal backgrounds and provide a more symmetric psf ,
a circular pupil stop was used which describes an inscribed circle on the hexagonal keck primary , with a diameter of @xmath14 meters .
each image consisted of 100 coadds of 0.4 seconds each .
van maanen 2 itself was used as the guide star , with the adaptive optics system running at a rate of 80 hz .
the correction was quite good , with a strehl ratio ( measured relative to the psf of the 9 meter pupil ) of 0.75 and a full width at half maximum of @xmath15 , essentially diffraction limited .
the plate scale was @xmath16 per pixel .
the adaptive optics images were reduced using standard programs in the iraf environment . for a given raw image ,
the sky was extracted by taking the median of the four remaining dithered images with rejection of hot and cold pixels .
a flat frame was created by averaging all five sky frames and subsequently normalizing .
each dithered frame was then sky subtracted , flat fielded , and the resulting five frames were registered and averaged . due to the uncertainties in the wings of the adaptive optics psf and weather that was not photometric , it was difficult to perform photometry on the keck data .
therefore photometric observations of van maanen 2 were acquired at the irtf . @xmath1 m data on van maanen 2 and two standard stars were acquired on 1 february 2004 by alan stockton at the 3 meter nasa infrared telescope facility with nsfcam @xcite .
for the standards , a 9 point dither pattern was used with each frame consisting of 0.15 second exposure times 100 coadds yielding a total integration time of 135 seconds .
the ukirt faint standard sa92 - 342 ( @xmath17 mag ) was observed immediately before van maanen 2 and the elias standard star hd22686 ( @xmath18 mag ) was observed immediately afterward @xcite .
two sets of images , acquired in the same manner as the standards , were obtained for van maanen 2 for a total of 18 frames and 270 seconds integration .
conditions during the observations were reported as photometric .
the data on the two standards were reduced using standard programs in the iraf environment . for a given raw image ,
the sky was extracted by taking the median of the eight remaining dithered images with rejection of hot and cold pixels .
a flat frame was created by averaging all nine sky frames and subsequently normalizing .
each dithered frame was then sky subtracted , flat fielded , and the resulting nine frames were registered and averaged . reducing the two sets of images on van maanen 2 was more difficult due to the faintness of the target in each raw frame .
it was found that the target was more easily seen in pairwise subtracted images rather than in sky subtracted images .
all 18 pairwise subtracted frames were flat fielded , registered and averaged , creating one final image for photometric measurements .
the flux of both standards was measured with aperture radii of @xmath19 and @xmath20 and corrected for extinction .
the error in determining the zero point was 0.03 mag .
van maanen 2 was measured in the smaller aperture to minimize noise , corrected to the larger aperture ( with an error of 0.04 mag ) and extinction corrected .
this yielded a signal to noise of 48 and a measurement error of 0.03 mag .
the photometric measurement including all errors was @xmath21 mag for van maanen 2 .
isocam observations of van maanen 2 were carried out in 1997 in an effort to provide observational constraints on the origin of metals in the photospheres of cool white dwarfs @xcite .
data were taken at @xmath5 m and @xmath22 m on van maanen 2 and several other white dwarfs as well as an a0v calibrator star .
the data are listed in table [ tbl-1 ] .
in order to estimate the brightness of the reported companion , a model for a @xmath3 brown dwarf at 5 gyr was chosen .
this mass lies conservatively in the lower range of possible values and required no interpolation within available models .
the model age is likely to be greater than the 3.67 gyr cooling age @xcite of van maanen 2 plus the @xmath23 gyr main sequence lifetime @xcite for a @xmath24 progenitor of the @xmath25 white dwarf @xcite .
a substellar companion of this mass and age would be a late t dwarf ( @xmath26 k ) and have @xmath27 mag which is @xmath28 mag at 4.4 pc @xcite .
photometric @xmath12 band data on known brown dwarfs do exist and the measurements agree with the models used here to within 0.3 mag for spectral type t6 @xcite .
van maanen 2 is predicted to have @xmath29 mag based on the model predicted @xmath30 color of a 6750 k helium white dwarf @xcite and the @xmath31 color of a 6750 k blackbody .
@xmath32 is the measured color the extrapolation was done from @xmath33 in case of any contamination by a companion at @xmath34 .
the reduced @xmath12 image ( figure [ fig1 ] ) shows no indication whatsoever of a companion with the brightness expected from a brown dwarf of the type reported by @xcite . from the published orbital parameters , the companion should have been at a separation of @xmath10 and position angle of @xmath11 on the date of the observation
the full width at half maximum of van maanen 2 in the reduced image is @xmath15 and the distance to the first airy ring is @xmath35 .
there are two extremely faint features within the airy ring at position angles of 284 and @xmath36 .
small aperture flux measurements , relative to the primary , at eight different evenly spaced locations around the airy ring indicate that these features are unlikely to be real . the flux at 284 and @xmath36 , @xmath37 and 0.073 respectively , are both within @xmath38 ( 0.018 ) of the average flux ( @xmath39 ) in the ring and are almost certainly artifacts due to imperfect optical guide star corrections . if the feature at @xmath36 were real , its brightness after subtracting the flux of the primary in the airy ring implies @xmath40 mag and a mass of @xmath41 , assuming an age of 5 gyr @xcite .
this is simply not massive enough to cause the reported astrometric wobble .
an artificial star was planted at @xmath10 from the primary in order to test the ability to detect faint companions at this separation .
the adaptive optics psf , extracted from the reduced image of van maanen 2 , was used for the stellar profile of the planted star . to be conservative , this artificial star is a full 2.0 magnitudes fainter than the white dwarf this was confirmed by placing the star at many positions on the image and measuring its flux without contamination by the primary .
the simulated star is readily seen in figure [ fig2 ] .
large errors in the reported orbital parameters allow for the possibility that the companion remained unresolved due to chance alignment during the observation .
the values and computed errors in separation and position angle , derived from the binary orbital calculator , are @xmath42 and @xmath43 respectively .
hence the putative companion could have been located at almost any position in its orbit by the epoch of the observation
. however , there is further evidence against this possibility .
in the case of a chance alignment , any substellar companion would still cause excess emission at mid infrared wavelengths .
this is primarily due to the size difference between brown dwarfs and white dwarfs ; a ratio of 10:1 in radius .
it is not possible for a significant occultation to occur . at most ,
a white dwarf could block only @xmath44% of the light from an orbiting brown dwarf .
as mentioned above , a value of @xmath21 mag was measured for van maanen 2 .
this is to be compared with the value predicted for a 6750 k white dwarf , @xmath29 mag .
the combined flux of the white dwarf plus a @xmath3 brown dwarf at 5 gyr would have a magnitude of @xmath45 mag .
hence it is concluded there is no excess emission at this wavelength .
furthermore , the published _ iso _
observations place even stronger constraints on the absence of flux from a brown dwarf around van maanen 2 .
a recent model @xcite of the flux from a 5 gyr , @xmath46 brown dwarf was integrated over the _ iso _ @xmath22 m filter lw3 .
this results in a flux of about 1.0mjy in this filter .
the reported measurement taken at van maanen 2 in 1997 was @xmath47 mjy .
this measurement is consistent with photospheric flux from the white dwarf and an _ excess _ of @xmath48 most likely would have been detected .
if the _ iso _ results are correct and the models are right , any companion with @xmath49 k is ruled out
this includes 10 gyr old brown dwarfs with @xmath50 @xcite .
table [ tbl-1 ] summarizes all existing photometric data on van maanen 2 and the corresponding fluxes are plotted in figure [ fig3 ] .
it should be clear from the figure that the measured fluxes are all consistent with a single white dwarf with a temperature around 6800 k.
three compelling infrared observations of van maanen 2 are presented which rule out the presence of a substellar companion warmer than @xmath51 k. first , @xmath1 m adaptive optics images reveal no significant flux at the predicted magnitude and position of the putative companion reported by @xcite .
second , @xmath1 m photometric observations are consistent with no excess emission at this wavelength as would be expected from a @xmath52 brown dwarf at 5 gyr . and third , @xmath22 m data also are consistent with no excess emission , ruling out any possible companion massive enough to induce the reported astrometric wobble .
part of the data presented herein were obtained at keck observatory , which is operated as a scientific partnership among the california institute of technology ( cit ) , the university of california and the national aeronautics and space administration ( nasa ) .
this publication makes use of data acquired at the infrared telescope facility , which is operated by the university of hawaii under cooperative agreement no .
ncc 5 - 538 with nasa , office of space science , planetary astronomy program .
some data used in this paper are part of the two micron all sky survey ( 2mass ) , a joint project of the university of massachusetts and the infrared processing and analysis center ( ipac)/cit , funded by nasa and the national science foundation ( nsf ) .
2mass data were retrieved from the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , cit , under contract with nasa .
the authors wish to express their gratitude to a. stockton , j. rayner and a. tokunaga for their assistance in acquiring the irtf observations of van maanen 2 .
sincere thanks go to j. larkin and m. barczys for donating some of their keck adaptive optics time for this project .
acknowledgement also goes to b. hansen for bringing this system to our attention , to b. zuckerman for a careful reading of this manuscript and many discussions , to a. ghez for the binary orbit calculator , and to b. vacca for help in the data reduction .
part of this work was performed under the auspices of the u.s .
department of energy , national nuclear security administration by the university of california , lawrence livermore national laboratory under contract no . w-7405-eng-48 .
this research has been supported in part by grants from nasa to ucla .
@xmath54 & 0.44 & 12.91 & @xmath55 + @xmath33 & 0.55 & 12.39 & @xmath56 + @xmath57 & 0.64 & 12.13 & @xmath58 + @xmath59 & 0.80 & 11.90 & @xmath60 + @xmath61 & 1.22 & 11.69 & @xmath62 + @xmath63 & 1.65 & 11.57 & @xmath64 + @xmath65 & 2.16 & 11.50 & @xmath66 + @xmath12 & 3.76 & 11.40 & @xmath67 + lw2 & 6.75 & & @xmath68 + lw3 & 15.0 & & @xmath47 + | the results of a comprehensive infrared imaging search for the putative @xmath0 astrometric companion to the 4.4 pc white dwarf van mannen 2 are reported .
adaptive optics images acquired at @xmath1 m reveal a diffraction limited core of @xmath2 and no direct evidence of a secondary .
models predict that at 5 gyr , a @xmath3 brown dwarf would be only 1 magnitude fainter than van maanen 2 at this wavelength and the astrometric analysis suggested a separation of @xmath4 . in the case of a chance alignment along the line of sight , a 0.4 mag excess should be measured .
an independent photometric observation at the same wavelength reveals no excess .
in addition , there exist published _ iso _
observations of van maanen 2 at @xmath5 m and @xmath6 m which are consistent with photospheric flux of a 6750 k white dwarf .
if recent brown dwarf models are correct , there is no substellar companion with @xmath7 k. |
the authors would like to acknowledge fapesp , capes and cnpq for financial support .
the calculations were carried out at cce - usp , center for high performance computing at ufabc and cenapad / sp . | nitrogen - doped carbon nanotubes can provide reactive sites on the porphyrin - like defects .
it s well known that many porphyrins have transition metal atoms , and we have explored transition metal atoms bonded to those porphyrin - like defects in n - doped carbon nanotubes .
the electronic structure and transport are analized by means of a combination of density functional theory and recursive green s functions methods .
the results determined the heme b - like defect ( an iron atom bonded to four nitrogens ) as the most stable and with a higher polarization current for a single defect .
with randomly positioned heme b - defects in a few hundred nanometers long nanotubes the polarization reaches near 100% meaning an effective spin filter .
a disorder induced magnetoresistance effect is also observed in those long nanotubes , values as high as 20000% are calculated with non - magnectic eletrodes . since their discovery by iijima in 1991 ,
carbon nanotubes@xcite ( cnt ) have become the subject of intense research due to their potential for applications,@xcite such as in novel electronic devices.@xcite furthermore , in a seminal paper by tsukagoshi _
et al _ cnts entered the realm of spintronics , whereby one envision the possibility of using the electron spin , instead of its charge , as information carrier.@xcite in that work the authors demonstrated that the spin coherence length of polarized electrons injected onto cnts is larger than 300 nm .
thus , carbon nanotube devices could be used to manipulate spins in a coherent manner .
the prototypical spintronics device uses spin - polarized electrons , which are injected from a source into an unpolarized region and analyzed by a polarized drain . within this arrangement the so - called giant magnetoresistance effect ( gmr)@xcite manifests itself by altering - via an external magnetic field - the relative orientations of the magnetic moments of the electrodes . from a practical point of view
this setup usually involves sandwiching different materials .
an alternative to this has been given by kirwan et al .
whereby initially unpolarized electrons are scattered by magnetic impurities adsorbed on the surface of a segment of a carbon nanotube .
this way , both the electrodes as well as the device itself are made of the same material .
an alternative to this has been given by kirwan _
et al . _
@xcite whereby initially unpolarized electrons are scattered by magnetic impurities adsorbed on the surface of a segment of a carbon nanotube.@xcite this way , both the electrodes as well as the device itself are made of the same material , thus avoiding issues related to surface matching at the interface consequently hindering spurious scattering .
however , one of the issues concerning the use of cnts as spintronics devices is the need to incorporate dopants or defects in order to change their electronic transport properties . closed shell species
do not interact very strongly with the pristine wall of a carbon nanotube.@xcite furthermore , transition metal atoms are more likely to form clusters when interacting with the pristine wall of the nanotube .
@xcite even linear chains of fe atoms are more energetically favorable than free standing iron atoms .
@xcite one possible path to circumvent this problem is to incorporate doping agents during the growth process . in that context , carbon nanotubes sythetized in a nitrogen - rich atmosphere - the so - called @xmath0 nanotubes - are potential candidates.@xcite it has already been demonstrated that these nanotubes could be , for example , used as gas sensors for a variety of chemical species.@xcite the most stable defect in these structures is a pyridine - like defect consisting of a 4 nitrogen divacancy ( 4nd).@xcite this defect ( shown in fig .
[ figure1]a ) is formed by two vacancies surrounded by four substitutional nitrogen atoms .
we have previously exploited the reactivity of this defect to attach ammonia molecules and study the behavior of the system as a sensor.@xcite interestingly , this defect is similar to molecules in the porphyrin class , in particular , to a molecule known as heme - b ( shown in fig .
[ figure1]b ) , which is found , for instance , in hemoglobin and myoglobin .
this heme b molecule has an iron atom bonded to the site with four nitrogens .
thus it is intuitive to assume that an iron atom - and other transition metal ( tm ) atoms - gets bonded to the 4nd defect of the carbon nanotubes .
the heme b - like defect has been recently synthesized by lee et , al .
@xcite their stability was studied by means of repeated cyclic voltammograms , and they have not observed significant differences after @xmath1 cycles , which is attributed to the stability of the covalent incorporation of the atoms . the use of carbon nanotubes - or any other long one dimensional system - with defects , however , poses an additional problem ; the position of the defects can not be controlled during growth .
the result is a device with a large number of randomly positioned defects .
thus , in order to obtain quantitatively meaningful theoretical predictions , one must take into consideration the effects of disorder .
it is generally assumed that disorder has a detrimental effect .
recently , however , we have shown that one - dimensional boron - doped graphene nanoribbons can present near - perfect conductance polarization due to disorder.@xcite such an effect should not depend on the system under consideration provided there was a difference in transmission probabilities for majority and minority spin cases for a single scatterer .
furthermore , the introduction of a large number of defects with non - zero magnetic moment leads not only to structural disorder , but also to a magnetic one .
in fact , the magnetic moments of the impurities might be pointing in random directions .
we can then consider that , in the absence of a magnetic field , approximately half of the moments are pointing up and half of them are pointing down .
a magnetic field would tend to align the magnetic moments leading to an analogous to the magnetoresistance effect without the need to rely on a multilayered material .
in this work we show that porphyrin - like cnts can exhibit a nearly - perfect polarization and extremely large disorder induced magnetoresistance ( as high as 20000% ) driven by disorder .
the disorder was simulated using cnts containing a large number of magnetic impurities randomly positioned along the tube .
for those calculations we used a combination of density functional theory@xcite coupled to recursive green s functions methods@xcite . we have initially performed _
ab initio _ calculations within density functional theory ( dft)@xcite for a segment of a ( 5,5 ) @xmath0 nanotube containing a 4nd defect ( figure [ figure1]a ) . as we have said before
, the porphyrin molecules can have different tm in the 4nd site , as shown in figure [ figure1]b .
we , thus , performed calculations using iron , cobalt , manganese and nickel atoms in the middle of the 4nd defect .
the final arrangement for the case of iron is presented in figure [ figure1]c .
as one can see , nine irreducible cells of the pristine system were used to describe the region containing the defect .
the computational code used was siesta @xcite which uses a linear combination of atomic orbitals ( lcao ) as basis set . in the particular case of this work
we have employed a double zeta basis set with polarization orbitals .
we used the generalized gradient approximation ( gga ) as parametrized by perdew - burke - ernzerhof @xcite ( pbe ) for the exchange correlation functional .
finally , the atomic coordinates have been optimized using a conjugated gradient scheme until the forces on atoms were lower than 0.03 ev / . furthermore , in order to assess whether the transition metal atoms in adjacent cells are magnetically coupled@xcite we simulated a supercell with 18 irreducible cells ( twice the initial size ) with two heme - b - like defects ; one case with ferromagnetic ordering ( @xmath2 ) , and another with a antiferromagnetic one ( @xmath3 ) .
the total energy difference ( e@xmath4-e@xmath5 ) is negligible for all transition metal atoms considered in this work , so we infer that there s no magnetic coupling in our system . for the electronic transport calculations , the system - following the procedure proposed by carolli _ _
et al__@xcite - is initially divided in three regions namely , the right and the left electrodes and a central scattering region .
the electrodes for our system are taken as semi - infinite repetitions of the pristine carbon nanotube . in the absence of spin - orbit interactions
one assumes the two spin fluid approximation , whereby one can calculate the electronic transport properties of the majority and minority spins independently of each other .
we then use the landauer - bttiker @xcite formula to calculate the transmission coefficients of the system . in order to access the electronic transport properties of a more realistic system with hundreds of nanometers we use a combination of dft and recursive green s function methods .
@xcite to do so , we split up a long nanotube into small pieces .
each piece is simulated using a separate dft calculation as already described previously , and the hamiltonian and overlap opperators respectively @xmath6 and @xmath7 are stored . for our ( 5,5 )
@xmath0 nt we also have to consider five different rotations for the position of the defect . with those smaller blocks we build up a long nanotube , ranging from 20 to 600 nanometers , by randomly placing the segments with defects together with pristine pieces , as shown in fig .
[ long ] .
one then recursively reduce the system to two renormalized electrodes coupled via an effective scattering potential that contains all the information about the central region . in the low bias limit
the differential spin - dependent conductance can be calculated by the landauer - buttiker formula for the current @xcite @xmath8 where @xmath9 is the fermi distribution function for a given temperature .
the total conductance is then given by the sum of the majority and minority conductances , @xmath10 .
we are interested in two quantities .
firstly , in order to quantify the spin filtering effect of this device , we calculate the degree of polarization@xcite @xmath11 secondly , as discussed earlier one also needs to take into consideration the relative orientations of the magnetic moments .
one can analyze the changes in conductance due to an external magnetic field that tends to align the local magnetization of each impurity .
the magnetic field in our calculations is taken into consideration only in the alignment of the magnetic moments , it has no other effect on the electronic structure of our system .
consequently , the value of this disorder induced magnetoresistance ( mr ) is given by @xmath12 where @xmath13 corresponds to the total conductance for the case where all of the magnetic moments are pointing along the same direction , and @xmath14 is the total conductance for the case where there is an equal distribution of positive and negative magnetic moments .
finally , in these long and disordered @xmath0 nanotubes , different defect distributions along the @xmath0 give different values of conductance . in order to get statistically meaningful values of conductance
we have calculated about 200 random arrangements for each concentration and length of the nanotubes .
for cn@xmath15 nanotube containing an iron atom ( heme - b - like defect ) .
the net magnetic moment of the system is localized in the iron atom.,scaledwidth=50.0% ] upon placing the different tm atoms in the 4nd , we observe that they strongly bind with a similar final structure in all situations .
table [ table1 ] shows the binding energies and the final magnetization of the system for each one of the tm atoms .
as can be seen , the binding energies are relatively high ( the reference is the isolated atom infinitelly separated from the nanotube ) .
one also sees a local magnetic moment in all cases , except for the nickel atom . a similar behavior was also observed by shang , _ et al .
_ .@xcite in fig .
[ rhoupdn ] we present the local magnetic moment , @xmath16 of the heme b - like defect . from this figure
we can notice a highly localized magnetic moment in the iron atom . .binding energy , magnetization and polarization for different transition metal atoms bonded to the 4nd defect . [ cols="^,^,^,^,^",options="header " , ] [ lengths100 ] in order to address the effect of a magnetic field applied to the system and the possibility that not all the magnetic moments are completely aligned we also considered that only 80% of the defects are magnetically aligned .
the results are shown in figure [ condpol2 ] , and the linear fits for the localization lengths also are presented in table [ lengths100 ] .
we can note a decrease ( increase ) in localization length for majority ( minority ) spin compared to the 100 % case .
this is to be expected since one is moving towards higher magnetic disorder , _ i.e. _ the 50 % case . from fig .
[ condpol2]c we can see a polarization near 100% for 700 nm long @xmath0 nanotubes .
most importantly , the magnetoresistance ( fig . [ condpol2]d ) presents values which are one order of magnitude lower than the fully aligned arrangement , but it is still in the 1000 % range .
thus , even in the case where not all spins are aligned , there is still an extremely large disorder induced magnetoresistance .
thus , this disorder - driven gmr effect is extremely robust toward fluctuations in the alignment of the magnetic moments .
we have used an ideal paramagnet model to estimate the needed magnetic field to obtain a 80% magnetization at 300k and 3k .
unfortunately the needed magnetic field at ambient temperature is about 200 t , making it impracticable for ambient temperature devices .
for a temperature of 3k the needed field will be about 2 t for an 80% magnetization and about 5 t for a 95% magnetization , so it s possible to observe the predicted effects in this paper in low temperature experiments .
in summary we have observed that transition metal atoms bind strongly to nitrogen defects in cn@xmath15 carbon nanotubes in a fashion similar to heme b molecules .
the end result is a scattering site with a localized magnetic moment on the transition metal atom that leads to a small conductance polarization in the case of a single impurity . for a large number of such impurities randomly distributed along carbon nanotubes a few hundred nanometers long
, it leads to near perfect polarization and a large magnetoresistance ( up to 20000% ) .
an interesting feature of the system proposed here is the fact that they do not need polarized electrodes as it is usually the case .
we estimate that , at low temperature , this effect could be measured experimentally . |
in this paper we continue our investigation of @xmath0 scattering at high energies ( see ref.@xcite for our previous attempts to study this process in the dglap dynamics ) .
we concentrate our efforts here on the case of two photons with large but almost equal virtualities .
it has been argued @xcite that this process is the perfect tool to recover the bfkl dynamics @xcite which is the key problem in our understanding of the low @xmath7 ( high energy ) asymptotic behavior in qcd .
it is well known that the correct degrees of freedom at high energy are not quarks or gluon but colour dipoles @xcite which have transverse sizes @xmath8 and the fraction of energy @xmath9 .
therefore , two photon interactions occur in two successive steps .
first , each virtual photon decays into a colour dipole ( quark - antiquark pair ) with size @xmath8 . at large value of photon virtualities the probability of such a decay
can be calculated in pqcd .
the second stage is the interaction of colour dipoles with each other .
the simple formula ( see for example ref .
@xcite ) that describes the process of interaction of two photons with virtualities @xmath10 and @xmath11 ( @xmath12 ) is ( see ) [ pps ] ( q_1 , q_2,w)= d^2 b_t ^n_f_a , b@xmath13 where the indexes @xmath14 and @xmath15 specify the flavors of interacting quarks , @xmath16 and @xmath17 indicate the polarization of the interacting photons where @xmath18 denote the transverse separation between quark and antiquark in the dipole ( dipole size ) and @xmath19 are the energy fractions of the quark in the fluctuation of photon @xmath20 into quark - antiquark pair .
@xmath21 is the imaginary part of the dipole - dipole amplitude at @xmath7 given by [ x ] x = for massless quarks ( @xmath22 is the energy of colliding photons in c.m.f . ) .
@xmath1 is the impact parameter for dipole - dipole interaction and it is equal the transverse distance between the dipole centers of mass .
the wave functions for virtual photon are known @xcite and they are given by ( for massless quarks ) @xmath23 with @xmath24 where @xmath25 denote the faction of quark charge of flavor @xmath14 . since the main contribution in
is concentrated at @xmath26 and @xmath27 where @xmath28 is the soft mass scale , we can safely use pqcd for calculation of the dipole - dipole amplitude @xmath29 in . in this paper
we study this process in the region of high energy and large but more - less equal photon virtualities ( @xmath30 ) in the framework of the bfkl dynamics . in the region of very small @xmath7 ( high energies )
the saturation of the gluon density is expected @xcite .
we will deal with this phenomenon using glauber - mueller formula @xcite which is the simplest one that reflects all qualitative features of a more general approach based on non - linear evolution @xcite .
for @xmath0 scattering with large but equal photon virtualities , the glauber - mueller approach is the only one on the market since the non - linear equation is justified only for the case when one of the photon has larger virtuality than the other . in the next section
we discuss the dipole - dipole interaction in the bfkl approach of pqcd .
the solution to the bfkl equation , that describes the dipole - dipole interaction in our kinematic region , has been found @xcite and our main concern in this section is to find the large impact parameter ( @xmath1 ) behavior of the solution . as was discussed in ref .
@xcite , we have to introduce non - perturbative corrections in the region of @xmath1 larger than @xmath31 where @xmath32 is the pion mass .
we argue in this section that it is sufficient to introduce the non - perturbative behavior into the born approximation to obtain a reasonable solution at large @xmath1 .
section 3 is devoted to glauber - mueller formula in the case of the bfkl emission @xcite . here
, we use the advantage of photon - photon scattering with large photon virtualities , since we can calculate the gluon density without uncertainties related to non - perturbative initial distributions in hadronic target .
we consider the low @xmath7 behavior of the dipole - dipole cross section and show that the large impact parameter behavior , introduced in the born cross section , fulfills the unitarity restrictions ( unitarity bound @xcite ) .
therefore , we confirm that the large @xmath1 behavior can be concentrated in the initial condition ( see refs . @xcite without changing the kernel of the non - linear equation that governs evolution in the saturation region as it is advocated in ref.@xcite . in the last section
we summarize our results .
in this section we discuss the one parton shower interaction in the bfkl dynamics ( see ) .
we start with the born approximation which is the exchange of two gluons ( see ) or the diagrams of without emission of a gluon .
these diagrams have been calculated in ref .
@xcite using the approach of ref .
@xcite and they lead to the following expression for the dipole - dipole amplitude : @xmath33 where @xmath19 is the fraction of the energy of the dipole carried by quarks ; @xmath34 and @xmath35 .
@xmath36 is the coordinate of quark @xmath20 ( see ) .
all vectors are two dimensional in .
each diagrams in is easy to calculate @xcite and the first diagram is equal to [ ba2 ] _
s^2 ^2_1,1^2_2,2. summing all diagrams we obtain .
we are interested mostly in the limit of large @xmath37 where the dipole - dipole amplitude can be reduced to a simple form .
[ balb ] n^ba ( r_1 , t , r_2 , t ; b_t)_s^2 , after integration over azimuthal angles .
therefore , we have a power - like decrease of the dipole - dipole amplitude at large @xmath1 , namely @xmath38 .
such behavior can not be correct since it contradicts the general postulates of analyticity and crossing symmetry of the scattering amplitude @xcite .
since the spectrum of hadrons has no particles with mass zero , the scattering amplitude should decrease as @xmath39 @xcite . in ref .
@xcite we suggested a procedure of how to cure this problem which is based on the results of qcd sum rules @xcite.following this procedure we rewrite the dipole - dipole amplitude as the integral over the mass of two gluons in @xmath40-channel ; and we assume , as in qcd sum rules , that this integral describes all hadronic states on average . restricting the integral over mass by the minimal mass of hadronic states ( 2 @xmath32 )
we obtain the model which provides the exponential fall at large @xmath41 and does not change the power like behavior for small @xmath42 .
we choose for the born amplitude the following formula [ bacorr ] n^ba(r_1 , t , r_2 , t ; b_t)=_s^2 k_4(2m_b_t ) .
one can easily see that reproduces and leads to [ bacorlb ] n^ba ( r_1 , t , r_2 , t ; b_t)_s^2 e^ - 2m_b_t . at large @xmath43 .
the emission of a gluon is described by the bfkl equation @xcite which was solved in ref.@xcite for fixed @xmath1 ( see ref.@xcite for many useful discussion of the different aspects of the solution ) .
the solution can be presented in factorized form ( see ) .
[ bfkl1 ] n(x , r_1 , t , r_2 , t ; b_t)= @xmath44 with [ omega ] ( ) = ( 2 ( 1 ) -(-
i)- ( + i ) ) ; where @xmath45,@xmath46 is euler gamma function and where [ v ] v(r_i , t , r_i;)= ( ) ^ -i using the following notations : @xmath47;@xmath48 is the size of the colour dipole @xmath49 and @xmath50 is the position of the center of mass of this dipole . in function
@xmath51 should be found from the initial condition which determines the dipole amplitude at fixed @xmath52 , namely , @xmath53 .
it should be stressed that the bfkl equation is a linear equation in which the kernel does not depend on @xmath1 ( see ref.@xcite ) .
therefore , @xmath54 could be an arbitrary function on @xmath1 . in we can take the integral over @xmath55 which leads to [ v2 ] v(r_2,t , r_2;-)= ( ) ^ + i. we are interested in the large @xmath1 behavior , namely , @xmath56 .
it is instructive to consider two cases : * @xmath57 .
+ this is so called double log approximation of pqcd ( dla ) in which we consider @xmath58 and @xmath59 while @xmath60 as well as @xmath61 and @xmath62 . we have considered this case in ref.@xcite and found that the emission of gluons does not induce any additional dependence on @xmath1 which is concentrated only in the born amplitude .
indeed , we can see this property directly from the solution of .
+ integrating over @xmath63 we find that the integrand of this integral falls down rapidly for @xmath64 due to @xmath63 dependence of the vertex @xmath65 ( see ) providing a good convergence for the integral .
for @xmath66 we can neglect @xmath63 dependence of the vertex @xmath65 and consider it as @xmath67 .
the integral over @xmath63 of @xmath68 for @xmath66 gives @xmath69 .
+ therefore , @xmath70 .
finally , taking @xmath71 , the dipole amplitude has a form [ dla1 ] n^dla(x , r_1 , t , r_2 , t ; b_t ) = n^ba(x , r_1 , t , r_2 , t ; b_t)e^ ( ) y + ( -i)(r^2_1,t / r^2_2,t ) considering @xmath72 and taking into account that @xmath73 at @xmath74 one can take the integral in explicitly .
the answer is well known ( see ref.@xcite for example ) , namely , at low @xmath7 [ dla2 ] n^dla(x , r_1 , t , r_2 , t ; b_t)=n^ba(x , r_1 , t , r_2 , t ; b_t)i_0 ( 2 ) for fixed coupling constant . ] .
* @xmath75 . + for such small values of @xmath76 the integral over @xmath63 is convergent for @xmath77 ( see ref.@xcite ) and , therefore , we neglect the @xmath63 dependence in @xmath78 . introducing a new variable @xmath79
we see that [ df1 ] d^2 r_1 v(r_1,t , r_1 ; ) = ( r^2_1,t)^+ i d^2 ( + ) ^2 ( - ) ^2)^+ i where @xmath80 is a unit vector in the direction of @xmath81 .
the integral is a function of @xmath76 only and can be absorbed in @xmath82 in . for @xmath83 at @xmath84 we have [ vdf ] v(r_2,t , b_t,-)=()^+ i therefore , the dipole amplitude is [ ndf ] n^df(x , r_1 , t , r_2 , t ; b_t)= _ in ( ; b_t ) e^()y ( ) ^+ i. we choose @xmath82 to be of in the form [ incon ] _ in ( ; b_t)= _ s(m_b_t)^4k_4 ( 2m_b_t ) . at small values of @xmath76
we can expand @xmath85 [ kerndf ] ( ) = _ l - d^2 with [ omsn ] _
l=42 ; d=14(3 ) ; finally , we can evaluate the integral over @xmath76 in using the method of steepest decent and obtain the following expression for dipole amplitude : [ andf ] n^df(x , r_1 , t , r_2 , t ; b_t)= @xmath86 + at @xmath87 [ andfb1 ] n^df(x , r_1 , t , r_2 , t ; b_t)@xmath88 + while at @xmath89 [ andfb2 ] n^df(x , r_1 , t , r_2 , t ; b_t ) @xmath90
the glauber - mueller approach @xcite takes into account the interaction of many parton showers with the target as is shown in . in our case of
more or less equal but large virtualities of both photons this approach gives a unique opportunity to study the high energy asymptotic behavior of the dipole amplitude since other methods based on non - linear evolution equation @xcite do not work in the case of two dipoles with more - less equal sizes .
the main idea of this approach is that the colour dipoles are the correct degrees of freedom for high energy scattering ( this idea was formulated by a.h .
mueller in ref .
@xcite ) . indeed ,
the change of the value of the dipole size @xmath8 ( @xmath91 ) during the passage of the colour dipole through the target is proportional to the number of rescatterings ( or the size of the target @xmath92 ) multiplied by the angle @xmath93 where @xmath94 is the energy of the dipole and @xmath95 is the transverse momentum of the @xmath40-channel gluon which is emitted by the fast dipole .
[ dof1 ] r_t r. since @xmath95 and @xmath8 are conjugate variables and due to the uncertainty principle @xmath96 therefore , [ dof2 ] r_trr_t rr^2_tex .
since the colour dipoles are correct degrees of freedom , they diagonalize the interaction matrix at high energy as well as the unitarity constraints , which have the form [ un ] 2n ( x , r_1,t , r_2,t;b_t)=|a_el ( x , r_1,t , r_2,t;b_t)|^2+g_in(x , r_1,t , r_2,t;b_t ) , where @xmath97 is the elastic amplitude of the dipole - dipole interaction and @xmath29 is the imaginary part of @xmath97 ( @xmath98 ) . assuming that the amplitude is pure imaginary at high energy
, one can find a simple solution to , namely @xmath99 where @xmath100 is the arbitrary real function . in glauber - mueller approach
the opacity @xmath100 is chosen as @xmath101 where @xmath102 is the dipole - dipole amplitude for one parton shower interaction that has been found in the previous section ( see ) .
one can see that if we substitute the explicit solution to the bfkl equation of at any fixed @xmath1 the opacity @xmath103 increases at @xmath104 .
therefore , the dipole - dipole amplitude given by glauber - mueller formula of tends to unity in the region of low @xmath7 .
this statement is called saturation @xcite since the physical interpretation of @xmath29 is the density of colour dipoles at least when @xmath29 is not very large . in this discussion
the saturation appears to be the consequence of unitarity for fixed @xmath1 .
however , we have learned several examples where the dipole density could reach a maximum value without having any effect on the elastic dipole - dipole amplitude at fixed @xmath1 ( see ref .
@xcite and paper of kovchegov and mueller in ref.@xcite ) .
however , for @xmath0 scattering of two small dipoles the initial condition is given by born amplitude of which is small .
therefore , we have no reason to expect that the dipole density will be high due to the final state interaction . to obtain the unitarity bound for the dipole - dipole cross section we have to integrate over @xmath1 , namely [ ddxs ] ( dipole - dipole)=2d^2b_tn^gm(x , r_1,t , r_2,t;b_t ) = 2d^2b_t(1-e^- n^df(x , r_1,t , r_2,t;b_t ) ) following froissart @xcite , we divide the region of integration over @xmath1 in in two parts [ ddun1 ] ( dipole - dipole)=2 ^b^2_0(x)_0 db^2_tn^gm( ... ;b_t)+^_b^2_0(x ) d b^2_t
n^gm( ... ;b_t ) where @xmath105 is defined from the equation [ b0 ] n^df(x , r_1,t , r_2,t;b_0(x))=1 it is easy to see that for @xmath106 @xmath107 since @xmath108 , while for @xmath109 and for @xmath110 @xmath111 therefore , we have the following unitarity bound [ ddun2 ] ( dipole - dipole)2(b^2_0(x)+^_b_0(x)d b^2_t n^gm( ... ;b_t ) ) let us consider two possibilities .
the first one that @xmath112 . in this case
the solution to follows directly from for the amplitude @xmath113 and for @xmath114 [ b0s ] ( ) = -2y or [ b0s1 ] b^2_0(x)r_1,tr_2,te^y substituting into we can obtain [ un3 ] ( dipole - dipole)2b^2_0(x)\{1 + 2 } e^y where the second term is calculated by integrating first over @xmath1 and after that using saddle point approach . since @xmath115 turns out to be small at low @xmath7 and we neglect it .
therefore , in this kinematic region we face a power -like increase of the dipole - dipole cross section as was pointed out in refs .
@xcite .
however , this power - like increase will stop for @xmath116 .
indeed , for such large values of @xmath1 we should use for the dipole amplitude @xmath113 . for such large values of @xmath105
has a solution which at low @xmath7 is [ sol1 ] b_0(x ) = y + o(ln y ) which leads to [ un4 ] ( dipole - dipole)2b^2_0(x)=^2(x_0/x ) which comes from the first term in .
it is easy to understand that the second term in this equation gives a term which does not increase with @xmath117 .
is the unitarity bound which has the same energy dependence as for hadron - hadron collisions @xcite but in our approach we are able to calculate the coefficient in front of @xmath118 .
the bound of is the same as was derived in @xcite .
it should be stressed that the diffusion approximation that we used was derived only at small values of saddle point in @xmath76 integration in which is equal to [ daddle ] | _ saddle| = 1 at @xmath119 from .
does not have a solution at any values of @xmath120 and @xmath121 ( formally , we obtain a negative values of @xmath105 ) .
the same equation at @xmath122 , namely [ satsc ] n(x , r_sat , r_2,t;b_t = 0 ) = 1 , determines the saturation scale . at @xmath123
the opacity @xmath100 in glauber - mueller formula is larger than unity ( @xmath124 ) , has a solution and we are in the saturation region with for the unitarity bound .
if @xmath125 , opacity @xmath126 at any value of @xmath1 .
this is a domain of perturbative qcd in virtual photon scattering .
@xmath127 we can find from integrating over @xmath63 , namely [ bfklb0 ] n(x , r_1 , t , r_2 , t ; b_t=0)= _ in ( ; b_t = 0 ) d^2 r_1 e^ ( ) y v(r_1,t , r_1;)v(r_2,t , r_1;- ) . since @xmath128 from is much smaller than @xmath121 we need to find only for @xmath129 .
this observation simplifies the calculations .
indeed , the main contribution in the integral over @xmath63 stems from @xmath130 .
therefore , we can neglect the @xmath63 - dependence of vertex @xmath131 which has the form [ v2s ] v(r_2,t , r_1;- ) = ( ) ^+ i. to perform the integration over @xmath63 we use the following formula ( see equation 3.198 of ref.@xcite ) [ fp ] b(- i , - i ) ( ) ^ -i= @xmath132 where @xmath133 is the euler beta - function .
integrating over @xmath63 using we obtain that [ bfklb01 ] n(x , r_1 , t , r_2 , t ; b_t=0)= _ in ( ; b_t ) e^ ( ) y ( ) ^+ inu the born approximation at @xmath134 and at @xmath135 can be reduced to @xcite [ bab0 ] n^ba(r_1 , t , r_2 , t ; b_t=0)_s^2 .
it is easy to choose @xmath136 in such a way that the final answer for @xmath137 is : [ finb0 ] n(x , r_1 , t , r_2 , t ; b_t=0)= _ s^2 , e^ ( ) y+ ( + i ) ( r^2_1,t / r^2_2,t )
. we can find the solution to in the saddle point approximation for the integral over @xmath76 in @xcite .
introducing a new variable @xmath138 we have the following equation for the saddle point value of @xmath139 : [ sadga ] |_= _ s y + ln(r^2_1,t / r^2_2,t)=0 substituting in we obtain [ satsc1 ] n(x , r_1 , t , r_2 , t ; b_t=0)e^ ( _ s)y- _ s ( r^2_2,t / r^1_2,t)=e^ y \{(_s)-
_ s|_= _ s } using we can solve in semiclassical approximation ( see ref .
@xcite in which we can not calculate the numerical factor in front of .
indeed , @xmath140 is constant on the line [ crl ] |_= _ 0 y + ln(r^2_sat(x)/r^2_2,t)=0 with @xmath141 is the solution to the equation @xcite was called @xmath142 in ref.@xcite and @xmath143 in ref.@xcite .
the numerical solution of leads to @xmath144 . ] [ crg ] = |_= _ 0 leads to a power - like increase of the saturation momentum ( @xmath145 ) at high energies ( low @xmath7 ) .
namely , [ satsc2 ] q^2_sat(x ) ( ) ^q^2_2 ( ) ^ actually , the pre - exponential factors in the steepest decent method of taking integral over @xmath146 could change the @xmath7-dependence of the saturation scale adding some log(1/x ) dependence in ( see ref .
@xcite for an analysis of such corrections ) . to obtain the unitarity bounds for @xmath0 scattering we need to substitute the unitarity bound for dipole - dipole cross section ( see ) into and to perform integrations over @xmath148 and @xmath19 .
@xmath149 is convergent while @xmath150 has a logarithmic divergence that we need to deal with
. holds only for @xmath151 since if @xmath152 dipole - dipole cross section is small and proportional to @xmath153 .
as has been mentioned we consider @xmath154 . on the other hand @xmath155 at @xmath156 .
finally , one can see [ int ] ^1/|q_1_r_sat d^2r_t ^1_0dz_1 |_t(q , z , r_t)|^2= c_q ( q^2_sat(x)r^2_2,t ) , where @xmath157 . we recall that @xmath158 does not depend on @xmath121 .
the integration over @xmath121 is concentrated at the limits @xmath159 and leads to @xmath160 finally , for @xmath0 cross sections we have @xmath161 since the saturation scale increases as a power of @xmath162 one can see that the energy behavior of the unitarity constraints is @xmath163 note only @xmath164 has the same energy dependence as hadron - hadron collisions @xcite but even this cross section has a different coefficient in front .
@xmath165 as well the as numerical factor @xmath166 come from the photon wave function while @xmath167 reflects the bfkl dynamics making and quite different from the unitarity bound for hadronic reactions .
in this paper we use @xmath168 scattering as the laboratory for studying the large impact parameter behavior of the amplitude in the saturation region . at first sight , this processes occurs at short distances for both photons with large virtualities and could be calculated in perturbative qcd .
we demonstrated that the non - perturbative qcd corrections have to be introduced for large @xmath1 even for this process .
the main result of this paper is the statement that it is enough to include the non - perturbative qcd corrections in the born approximation and neglect them in the kernel of the bfkl equation .
this result confirms the mechanism suggested in refs . @xcite but it contradicts the arguments of ref .
@xcite .
this result does not mean that the bfkl kernel correctly describes the large @xmath1 behavior .
the uncertainties in the large @xmath1 tail of the kernel will not affect the high energy asymptotic behavior of the dipole amplitude .
let us assume that kernel of the bfkl equation can be written as @xmath169 where @xmath170 is normal bfkl kernel in pqcd and @xmath171 includes the non - perturbative contribution .
we know that @xmath172 from general properties of the strong interaction @xcite .
let us treat @xmath171 as a small correction and calculate the first digram of the order of @xmath171 ( see ) .
the sum of all diagrams in leads to a contribution [ dltk ] k ( 1-n^gm(x , r_1,t , r_2,t , b_t ) ) = k e^ - n^ops(x , r_1,t , r_2,t , b_t ) since for @xmath106 @xmath173 is very close to unity , the above corrections are suppressed . only for @xmath174
we can expect a considerable contribution . however , this contribution is proportional to @xmath175 .
therefore they turn out to be very small .
this simple discussion shows why the strategy to include the non - perturbative corrections in the born amplitude , works actually , the main result of this paper , namely , is based on a simple physics ( see ref .
@xcite ) . we have demonstrated here that the multi rescattering processes embraced by the glauber - mueller formula lead to a different resulting parton cascade than is given by the bfkl approach .
the principle difference is the fact that the multi parton shower interaction creates a new scale or mean parton transverse momentum ( saturation scale ) given by .
@xmath176 denotes the parton density , consequently the fact that @xmath177 can be understood as the fact that the partons reach a maximal density at low @xmath7 .
this phenomenon is called saturation @xcite . therefore , at low @xmath7 we have the parton distribution in the transverse plane presented in : the uniform distribution of partons ( dipoles ) with sizes of the order of @xmath178 in the disc of radius @xmath179 .
if one of the dipole inside of the disc will emit one extra parton this emitted parton will interact with others partons and as a result of this interaction its transverse momentum will be of the order of @xmath180 .
it means that this emitted gluon will not change its position in impact parameter space since due to uncertainty principle [ unpr ] b_t p_t1 its @xmath181 .
however , for the parton at the edge of the disc the situation is different since the emitted parton in the direction outside of the disc can move freely without any interaction .
this parton changes the size of the disc by its displacement in @xmath1 , namely @xmath182 . in this estimate
we consider the non - perturbative emission with @xmath183 because , as have been discussed , a non - perturbative emission is needed to provide the unitarization of our process .
since the emission that leads to a growth of the disc occurs in one direction ( the exterior of the disc ) it leads to @xmath184 where @xmath185 is the number of emission at given @xmath7 .
since the emission takes place at the edge of the disc where the parton density is rather small , @xmath186 is determined by the bfkl dynamics only @xcite . in the bfkl approach @xcite@xmath187 since @xmath188 .
therefore , we obtain , namely , @xmath189 .
we have discussed in this paper the structure of dipole - dipole interaction in the glauber - mueller approach which is the only one on the market for the interaction of two dipoles of the same sizes . however , for two dipole with small but different sizes the non - linear evolution equation @xcite should be solved to which the bfkl emission is only an approximation in the region of small partonic densities .
comparison of the result of this paper with the dipole - dipole interaction in , so called , double log approximation @xcite shows that the bfkl dynamics does not change physics at large @xmath1 .
the non - linear evolution equation at fixed @xmath1 was solved @xcite in the case when the bfkl kernel was replaced by the double log one .
the solution leads to the answer in the saturation region with geometrical scale@xcite [ gescale ] n(x , r_1,t , r_2,t;b_t)= f(=r^2_1,tq^2_sat(x)e^-4m_b_t ) .
therefore , we believe that for the bfkl dynamics will hold . this belief is based on the similarity between double log and bfkl approximation for @xmath0 processes .
* acknowledgments : * we wish to thank jochen bartels , errol gotsman and uri maor for very fruitful discussions on the subject .
one of us ( e.l . ) would like to thank the desy theory group for their hospitality and creative atmosphere during several stages of this work .
he is indebted to the alexander - von - humboldt foundation for the award that gave him a possibility to work on low @xmath7 physics during the last year .
this research was supported in part by the gif grant @xmath190 i-620 - 22.14/1999 and by israeli science foundation , founded by the israeli academy of science and humanities .
the integration over @xmath63 in can be taken explicitly @xcite and can be reduced to [ a1 ] d^2r_1v(r_1,t , r_1 ; ) v(r_1,t,|_1 -_t| ; -)= @xmath191 where @xmath192 is the hypergeometrical function ( see ref .
@xcite ) ; @xmath7 is the complex anharmonic ratio : [ a2 ] x= and @xmath193 .
@xmath194 gives [ a5 ] xx^*= one sees that is invariant with respect to rotation in the plane .
however , one can see that does not reproduce the born term of at @xmath198 . to understand why it is so we should consider the vertex @xmath199 in momentum representation ( see ref .
@xcite ) , namely , [ a6 ] v(r_1,t , q ; ) = d^2r_1e^ i v(r_1,t , r_1 ; ) . it turns out @xcite that @xmath200 at small @xmath201 leads to the following behavior of vertex @xmath202 : [ a8 ] v(r-1,t , q ; ) | _
q^2 0()^-i ( 1-(q^2)^2 i()^i 2 ) as have been discussed the matching with the born approximation occurs at @xmath203 . in this limit [ a9 ] v(r_1,t , q ; ) ( 1 - q^2 ) , which has correct analytical behavior
actually this behavior dictates the choice of the coefficients @xmath195 and @xmath196 in and
. however , at @xmath204 the low @xmath201 behavior has a singularity @xmath205
. therefore the symmetry of with respect to sign of @xmath76 is broken .
mueller and tang @xcite pointed out that this problem can be cured by adding to the expression of @xmath206 of , namely , @xmath207 as was found@xcite such terms can be added due to gauge invariance of qcd . in momentum representation ( see ) @xmath208 can be written as a sum of three terms as it is shown in .
the mueller - tang vertex leads to the born approximation amplitude in the form of .
however , as was discussed in refs .
@xcite , it has not been proven that this vertex will satisfy the bfkl equation .
the solution in the form of has a different form of the born amplitude , namely , [ a10 ] n^ba ( ) ( ) .
however , these two expressions for the born amplitude are equivalent due to gauge invariance of qcd @xcite .
using we can calculate the dipole - dipole amplitude at @xmath209 and , therefore , the saturation scale with better accuracy than within . on the other hand the saturation momentum increases for @xmath210 .
such an increase guarantees that approaches the amplitude given by in the region of low @xmath7 .
this is the reason why we prefer to use a simple solution of instead of full expression of .
99 m. kozlov and e. levin , arxiv : hep - ph/0211348 .
j. bartels , a. de roeck and h. lotter , phys .
b389 * ( 1996 ) 742 .
s. j. brodsky , f. hautmann and d.e .
soper , phys . rev .
* d56 * ( 1997 ) 6957 .
kuraev , l.n .
lipatov and v.s .
fadin , sov .
jetp * 45 * ( 1977 ) 199 ; + ia.ia .
balitsky and l.n .
lipatov , sov .
* 28 * ( 1978 ) 822 ; + l.n .
lipatov , sov .
jetp * 63 * ( 1986 ) 904 .
a. h. mueller , nucl . phys .
* b335 * ( 1990 ) 115 .
e. m. levin and m. g. ryskin , sov .
j. nucl . phys.*45 * ( 1987 ) 150 .
a. zamolodchikov , b. kopeliovich and l. lapidus , jetp lett .
* 33 * ( 1981 ) 595 .
a. h. mueller , nucl . phys . *
b415 * ( 1994 ) 373 .
a. donnachie , h. g. dosch and m. rueter , eur .
j. * c13 * ( 2000 ) 141 .
nikolaev and b.g .
zakharov , z. phys .
* c49 * ( 1991 ) 607 ; + e.m .
levin , a.d .
martin , m.g .
ryskin and t. teubner , z. phys . *
c74 * ( 1997 ) 671 .
l. v. gribov , e. m. levin and m. g. ryskin , phys . rep . * 100 * ( 1983 ) 1 .
a. h. mueller and j. qiu , nucl . phys . * b 268 * ( 1986 ) 427 .
l. mclerran and r. venugopalan , phys .
d 49 * ( 1994 ) 2233 , 3352 ; * d 50 * ( 1994 ) 2225 , * d 53 * ( 1996 ) 458 , * d 59 * ( 1999 ) 09400 .
balitsky , nucl . phys .
* b 463 * ( 1996 ) 99 ; + yu . kovchegov , phys . rev . *
d 60 * ( 2000 ) 034008 .
l. n. lipatov , sov .
jetp * 63 * ( 1986 ) 904 .
e. m. levin and m. g. ryskin , phys .
* 189 * ( 1990 ) 267 .
a. kovner and u. a. wiedemann , _ `` taming the bfkl intercept via gluon saturation , '' _ hep - ph/0208265 ; phys . lett .
* b551 * ( 2003 ) 311 [ arxiv : hep - ph/0207335 ] ; phys .
* d66 * ( 2002 ) 034031 [ arxiv : hep - ph/0204277 ] ; phys .
* d66 * ( 2002 ) 051502 [ arxiv : hep - ph/0112140 ] .
e. ferreiro , e. iancu , k. itakura and l. mclerran , nucl .
phys . a * 710 * ( 2002 ) 373
[ arxiv : hep - ph/0206241 ] m. froissart , phys . rev .
* 123 * ( 1961 ) 1053 ; a. martin , scattering theory : unitarity , analitysity and crossing . "
lecture notes in physics , springer - verlag , berlin - heidelberg - new - york , 1969 .
i. f. ginzburg , s.l .
panfil and v.g .
serbo , nucl . phys .
* b284 * ( 1987 ) 685 , * b296 * ( 1988 ) 569;i .
f. ginzburg and d. yu .
ivanov , nucl .
b388 * ( 1992 ) 376;d .
y. ivanov and r. kirschner , phys .
* d58 * ( 1998 ) 114026 , hep - ph/9807324 .
colangelo and a. khodjamirian , _ `` qcd sum rules : a modern perspective , '' _ hep - ph/0010175;v .
e. markushin , acta phys . polon . *
b31 * ( 2000 ) 2665 ; o. i. yakovlev , r. ruckl and s. weinzierl , _
`` qcd sum rules for heavy flavors , '' _ hep - ph/0007344;m .
a. shifman , _
`` qcd sum rules , '' _ in lenz , f. ( ed . ) et al . :
ilectures on qcd : foundations " , pp.170 - 187;s .
narison , _ `` qcd spectral sum rules , '' _ world sci . lect .
notes phys . ,
* 26 * ( 1989 ) 1;m .
a. shifman , a. i. vainshtein and v. i. zakharov , nucl . phys .
* b147 * ( 1979 ) 385 , 448 , 519 .
j. bartels , j. r. forshaw , h. lotter , l. n. lipatov , m. g. ryskin and m. wusthoff , phys .
* b348 * ( 1995 ) 589 [ arxiv : hep - ph/9501204 ] .
h. navelet and r. peschanski , nucl . phys .
* b634 * ( 2002 ) 291 [ arxiv : hep - ph/0201285 ] ; phys .
* 82 * ( 1999 ) 137 , , [ arxiv : hep - ph/9809474 ] ; nucl . phys .
* b507 * ( 1997 ) 353 , [ arxiv : hep - ph/9703238 ] .
j. c. collins and j. kwiecinski , ; j. bartels , j. blumlein , and g. shuler , ; a. l. ayala , m. b. gay ducati , and e. m. levin , , * b 510 * ( 1990 ) 355 ; yu .
kovchegov , , * d 55 * ( 1997 ) 5445 , * d 61 * ( 2000 ) 074018 ; a. h. mueller , _ nucl . phys . _ * b 572 * ( 2000 ) 227 , * b 558 * ( 1999 ) 285 ; yu . v. kovchegov , a. h. mueller , ; e. iancu , a. leonidov , and l. mclerran , _ nucl
_ * a 692 * ( 2001 ) 583 ; m. braun , _ eur . phys . j. _ * c 16 * ( 2000 ) 337 . j. jalilian - marian , a. kovner , l. mclerran , and h. weigert , ; j. jalilian - marian , a. kovner , and h. weigert , ; j. jalilian - marian , a. kovner , a. leonidov , and h. weigert , , erratum - ibid . ; a. kovner , j.guilherme milhano , and h. weigert , _ phys . rev . _
* d 62 * ( 2000 ) 114005 ; h. weigert , _ nucl .
phys . _ * a 703 * ( 2002 ) 823 .
j. bartels and e. levin , nucl .
* b387 * , 617 ( 1992 ) .
a. m. stasto , k. golec - biernat and j. kwiecinski , phys .
lett . * 86 * , 596 ( 2001 ) [ arxiv : hep - ph/0007192 ] ; e. iancu , k. itakura and l. mclerran , nucl . phys .
* a708 * ( 2002 ) 327 [ arxiv : hep - ph/0203137],phys .
* b510 * ( 2001 ) 145 [ arxiv : hep - ph/0103032];_``understanding geometric scaling at small x , '' _ arxiv : hep - ph/0205198 .
a. h. mueller and w. k. tang , phys .
* b284 * ( 1992 ) 123 . | in this paper @xmath0 scattering with large , but more or less equal virtualities of two photons is discussed using bfkl dynamics , emphasizing the large impact parameter behavior ( @xmath1 ) of the dipole - dipole amplitude .
it is shown that the non - perturbative contribution is essential to fulfill the unitarity constraints in the region of @xmath2 , where @xmath3 is pion mass .
the saturation and the unitarization of the dipole - dipole amplitude is considered in the framework of the glauber - mueller approach .
the main result is that we can satisfy the unitarity constraints introducing the non - perturbative corrections only in initial conditions ( born amplitude ) .
# 1eq .
( [ # 1 ] ) # 1fig
. [ # 1 ] = # 1#2#3 _ ann . phys .
( ny ) _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ ann .
rev .
nucl . part .
sci . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ nucl .
phys . _
* b#1 * ( 19#2 ) # 3 # 1#2#3 _ phys .
lett . _
* b#1 * ( 19#2 ) # 3 # 1#2#3 _ phys . rev . _
* d#1 * ( 19#2 ) # 3 # 1#2#3 _ phys . rep . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ phys .
rev .
lett . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ prog .
theor .
phys . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ rev .
mod .
phys . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ z. phys . _
* c#1 * ( 19#2 ) # 3 # 1#2#3 _ mod .
phys .
lett . _ * a#1 * ( 19#2 ) # 3 # 1#2#3 _ nuovo cim . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ yad .
fiz . _
* # 1 * ( 19#2 ) # 3 # 1#2#3 _ sov .
j. nucl .
phys . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ sov .
phys .
_ jetp * # 1 * ( 19#2 ) # 3 # 1#2#3 _ jetp lett . _
* # 1 * ( 19#2 ) # 3 # 1#2#3 _ ( sov .
j. nucl .
phys . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ ( sov .
. jetp _
* # 1 * ( 19#2 ) # 3 # 1#2#3_(jetp lett . _
* # 1 * ( 19#2 ) # 3 # 1#2#3 _ zh . etf _
* # 1*(19#2 ) # 3 # 1#2#3 _ comm .
math .
phys . _ * # 1 * ( 19#2 ) # 3 # 1#2#3 _ comp .
phys .
commun . _ * # 1 * ( 19#2 ) # 3 # 1#2 _ dissertation , _ # 1 19#2 # 1#2#3 _ diplomarbeit , _ # 1 # 2 19#3 # 1#2#3 _ ibid .
_ * # 1 * ( 19#2 ) # 3 # 1#2#3 _
j. phys_. * g#1*#2#3 + * @xmath4-@xmath4 scattering : saturation and * + * unitarization in the bfkl approach .
* _ @xmath5 ) hep department _ + _ school of physics and astronomy _ + _ raymond and beverly sackler faculty of exact science _
+ _ tel aviv university , tel aviv , 69978 , israel _ + 0.3 cm _
@xmath6 desy theory group _ + _ 22603 , hamburg , germany _ |
the development of technologies to trap and cool neutral atoms has led to the emergence of cold atom physics as a new subfield of atomic physics .
the atoms can be cooled to temperatures that are orders of magnitude smaller than the tiny differences between the hyperfine energy levels of the atoms .
many of the most important loss processes for ultracold atoms involve deeply inelastic reactions , which produce atoms with kinetic energies much larger than those of the cold trapped atoms .
one such process is three - body recombination , in which a collision of three low - energy atoms results in the binding of two of the atoms into a diatomic molecule .
if the binding energy of the molecule is large compared to the energy scale of the cold atoms , the momenta of the molecule and the recoiling atom are often large enough for them to escape from the trapping potential for the atoms .
ultracold atoms can be described by a local nonrelativistic effective field theory for which the coupling constant is the scattering length @xcite .
local effective field theories can be applied most rigorously to a system in which there is an energy gap separating the low - energy particles described explicitly by the effective field theory from the higher - momentum particles @xcite . in a system consisting of low - energy atoms ,
conservation of energy ensures that a high - momentum atom can only appear through a virtual state that , by the time - energy uncertainty principle , has a short lifetime . during that short lifetime ,
the high - momentum atom can propagate only over a short distance .
its effects on low - energy atoms are therefore restricted to short distances .
these effects can be reproduced by local hermitian operators in the hamiltonian for the effective field theory .
local effective field theories have also been applied to systems with deeply inelastic reactions that produce particles with momenta too large to be described accurately within the effective field theory .
for example , a deeply inelastic three - body recombination event produces a molecule and an atom whose momenta may be outside the domain of validity of the effective theory .
the large energy release from the inelastic reaction comes from the conversion of rest energy into kinetic energy .
the standard argument for the locality of the effective field theory does not apply .
the particles with large momenta produced by the inelastic reaction can propagate over arbitrarily long distances , so their effects on low - energy particles are not obviously restricted to short distances .
nevertheless , there are general arguments based on the uncertainty principle that indicate that their effects can be taken into account systematically through local anti - hermitian operators in the effective hamiltonian @xcite .
the effective hamiltonian can be expressed as @xmath0 , where @xmath1 and @xmath2 are hermitian . the dynamics of a multi - atom system with deeply inelastic reactions
is conveniently described by a density matrix . a system that starts as a pure quantum state with @xmath3 low - energy atoms evolves into a mixed quantum state that is a superposition of a state with @xmath3 low - energy atoms and states with fewer low - energy atoms , as the inelastic reactions shift probability from the low - energy atoms into the high - momentum atoms .
an effective density matrix @xmath4 can be defined by tracing the density matrix over states containing high - momentum atoms .
naively we might expect the time evolution equation for @xmath4 to be @xmath5 - i \left\{k,\rho\right\}. \label{eq : drhodt - naive}\end{aligned}\ ] ] as we will demonstrate in this paper , the correct evolution equation for the effective density matrix is the _
lindblad equation _
@xcite , which has an additional term .
the lindblad equation arises in the quantum information theory of _ open quantum systems_. an open quantum system consists of all the degrees of freedom of both the subsystem of interest and the environment . under special circumstances ,
the density matrix for the subsystem evolves in time according to the lindblad equation . in the lindblad equation for the density matrix of an effective field theory obtained by integrating out deeply inelastic reactions , the additional lindblad term is local , and it can be deduced from the local anti - hermitian terms in the effective hamiltonian . an open quantum system
in which the subsystem of interest is a field theory is called an _
open effective field theory _ @xcite .
grozdanov and polonyi have proposed that an open effective field theory for the hydrodynamic modes of a quantum field theory can be used as a framework for deriving dissipative hydrodynamics @xcite .
burgess , holman , tasinato , and williams have applied open effective field theory to the super - hubble modes of primordial quantum fluctuations in the early universe @xcite . in the stochastic inflation framework ,
the master equation is the lindblad equation .
since the density matrix for an effective field theory in which deeply inelastic reaction products have been integrated out satisfies the lindblad equation , this system can also be regarded as an open effective field theory @xcite . in this case ,
the environment consists of the high - momentum particles produced by the deeply inelastic reactions .
the paper is organized as follows . in section [ sec : densitymatrix ] , we summarize the basic properties of the density matrix and we introduce the lindblad equation . in section [ sec : effective theory ] , we explain how integrating out deeply inelastic reaction products results in local operators in the effective lagrangian density .
we also explain why the effective density matrix obtained by tracing over states that include deeply inelastic reaction products must satisfy the lindblad equation . in section [ sec : atomloss ] , we apply the lindblad equation to the mean number of low - energy atoms .
we derive the universal relation for the two - atom inelastic loss rate for fermions with two spin states and the universal relation for the three - atom inelastic loss rate for identical bosons .
our results are summarized briefly in section [ sec : summary ] . in an appendix
, we demonstrate how the lindblad equation for the density matrix emerges from the diagrammatic analysis of a simple field theory model with a deeply inelastic two - atom scattering reaction .
the density matrix can be used to describe pure quantum states , mixed quantum states , and statistical ensembles of quantum states .
the dynamics of a multi - atom system with atom losses must be described by the density matrix in order to be able to track the losses in the different @xmath3-atom sectors . to lay the foundation for the rest of the paper , we summarize the key properties of the density matrix and we introduce the lindblad equation
. a pure state in a quantum system can be represented by a vector @xmath6 in a complex vector space .
the time evolution of the quantum state is described by the schrdinger equation : @xmath7 , where the hamiltonian @xmath1 is a hermitian operator .
the time evolution preserves the norm @xmath8 of a state .
an arbitrary statistical ensemble of quantum states can be represented by a density matrix @xmath4 .
the density matrix has the following basic properties : * it is hermitian : @xmath9 . *
it is positive : @xmath10 for all nonzero states @xmath6 .
* it can be normalized to have unit trace : @xmath11 .
the density matrix can also describe a pure quantum state , in which case @xmath12 .
the time evolution of the density matrix is described by the von neumann equation : @xmath13 , \label{evol : schrodinger}\ ] ] which follows from the schrdinger equation for @xmath6 .
this evolution equation has the following properties : * it is linear in @xmath4 .
* it preserves the trace of @xmath4 .
it therefore respects the normalization condition @xmath11 . *
it is _ markovian _ : the future is determined by the present only .
there is no additional dependence on the past history .
the system average of an operator @xmath14 can be expressed as the trace of its product with the density matrix : @xmath15 if the operator has no explicit time dependence , the time evolution of the system average is determined by the evolution equation of @xmath4 in eq . . in the field of quantum information theory ,
the full system is often separated into a _
subsystem _ of interest and its _ environment _ @xcite .
of particular interest is the decoherence of the subsystem due to the effects of the environment .
the basis for the quantum states of the full system can be chosen to be direct products of basis states for the subsystem and those for the environment .
a density matrix @xmath4 for the subsystem can be obtained from the density matrix @xmath16 for the full system by the partial trace over the states of the environment : @xmath17 the density matrix for the full system evolves according to the von neumann equation in eq . .
given the initial density matrix @xmath18 for the subsystem and a specified initial state of the environment , the von neumann equation can in principle be solved to determine @xmath19 at future times .
it is possible in principle to construct a self - contained differential equation for @xmath19 , but its evolution is non - markovian @xcite .
the time derivative of @xmath20 is determined by the present density matrix and by its history from time 0 to @xmath21 .
the previous history is needed to take into account the flow of information between the subsystem and the environment .
there are situations in which the time evolution of the subsystem can be described approximately by a self - contained markovian differential equation .
the time during which the subsystem is observed must be much larger than the time scale for correlations between the subsystem and the environment .
we must also restrict our consideration to the low - frequency behavior of the subsystem , which can be accomplished by smoothing out over times larger than the correlation time scale .
the density matrix @xmath4 for the subsystem necessarily satisfies the three basic properties listed before eq .
: it is hermitian , it is positive , and it can be normalized .
it would be desirable for its time evolution to also be in accord with the three basic properties listed after eq .
: it should be linear in @xmath4 , it should preserve the trace of @xmath4 , and it should be markovian .
in 1976 , lindblad and , independently , gorini , kossakowski , and sudarshan showed that these conditions together with one additional technical condition require the time evolution equation to have the form @xcite @xmath22 - \frac{i}{2 } \sum_n \left ( l_n^\dagger l_n \rho + \rho l_n^\dagger l_n -2 l_n \rho l_n^\dagger \right ) , \label{evol : lindblad}\ ] ] where @xmath1 is a hermitian operator and the @xmath23 s are an additional set of operators called _
lindblad operators_. the evolution equation in eq .
is called the _
lindblad equation_. the linear operator acting on @xmath4 defined by the right side of eq .
is called the _
lindbladian_. thus the lindbladian is the generator of the time evolution of the density matrix .
the additional technical condition required to derive the lindblad equation is that the linear operator that determines the time evolution of @xmath4 should be completely positive @xcite . for any complex vector space
, this linear operator has a natural extension to an operator acting on the tensor product of the quantum - state spaces for the subsystem and the other complex vector space .
complete positivity requires the extension of the operator to be positive in the larger space .
an accessible derivation of the lindblad equation can be found in lecture notes on quantum information and computation by john preskill @xcite . the lindblad equation in eq .
can be expressed in the form @xmath22 - i \left\{k,\rho\right\ } + i \sum_n l_n \rho l_n^\dagger , \label{evol : lindblad-2}\ ] ] where the hermitian operator @xmath2 is @xmath24 comparison with the naive evolution equation in eq .
reveals that @xmath25 can be interpreted as a non - hermitian effective hamiltonian for the subsystem . the additional _ lindblad term _ in eq .
is necessary to preserve the trace of @xmath4 .
in this section , we review how effective field theories for cold atoms are constructed , with particular focus on the treatment of deeply inelastic reactions .
we first discuss the construction of the effective lagrangian , and how integrating deeply inelastic reaction products out of a theory results in new local operators in the effective lagrangian density .
we then discuss how such a formalism can be used to study multi - particle states by means of an effective density matrix obtained by tracing over states that include deeply inelastic reaction products .
has contributions from intermediate states @xmath26 , where @xmath27 is a different hyperfine state .
if @xmath27 is a higher hyperfine state , the virtual atoms can propagate only over short distances , so their effects can be mimicked by local @xmath28 operators .
if @xmath27 is a lower hyperfine state , the virtual atoms can propagate over long distances , but the imaginary part of the amplitude is still local and can again be mimicked by local @xmath28 operators . , width=377 ] an effective field theory is obtained by removing ( `` integrating out '' ) states from a field theory .
the simplest applications involve removing particles with much higher energies .
we consider atoms with mass @xmath29 in a specific hyperfine state labelled @xmath30 and with kinetic energies much smaller than the splitting @xmath31 between hyperfine states . as illustrated in fig .
[ fig:2to2-elastic ] , there is a contribution to the elastic scattering amplitude from scattering into a virtual pair of particles in a higher hyperfine state labelled @xmath27 , which then rescatter into atoms in the original hyperfine state .
the virtual atoms are off their energy shells by large amounts of order @xmath31 . by the time - energy uncertainty principle ,
the virtual states have short lifetimes of order @xmath32 , during which the @xmath27 atoms can propagate only over short distances much smaller than the wavelengths of the @xmath30 atoms .
the effects of the higher hyperfine states can be modeled in the effective field theory by a contact interaction of the form @xmath33 , where @xmath30 is the quantum field that annihilates an atom in the specific hyperfine state @xmath30 .
( we use the same symbol for the label of the hyperfine state and for its quantum field . )
the @xmath28 contact interaction is the leading term in a series of local operators in the effective lagrangian that mimic the effects of the higher hyperfine states to arbitrary precision .
this series is obtained at tree - level by taylor expanding the scattering amplitude in the momentum transfer and then writing down terms in the effective lagrangian density that reproduce this expansion : @xmath34 where the coefficients @xmath35 , @xmath36 , are real valued .
the individual operators are renormalized when loop corrections are included , but the effective theory is still capable of reproducing the original theory to arbitrary precision provided operators with a sufficient number of gradients are retained .
operators with @xmath3 gradients correct the theory at order @xmath37 , where @xmath38 is the momentum transfer and @xmath39 is the momentum scale associated with the energy splitting @xmath31 between the hyperfine states .
we have integrated the higher hyperfine states out of the theory by replacing them by the interactions terms in eq . . , where @xmath27 is a lower hyperfine state , to the rate for the deeply inelastic scattering process @xmath40 .
, width=566 ] a less obvious opportunity to remove states from the theory arises when atoms in the hyperfine state @xmath30 scatter inelastically into atoms in a lower hyperfine state @xmath27 .
such a scattering process is an example of a _ deeply inelastic reaction _ , in which the difference in the rest energies of the initial - state and final - state atoms is converted into large kinetic energies of the final - state atoms .
the rate for the deeply inelastic scattering process @xmath40 is related by the optical theorem to the imaginary part of the amplitude for @xmath41 ( see fig . [ fig:2to2-inelastic ] ) . that amplitude is analytic in the momentum transfer @xmath42 , because the nearest non - analyticity in the amplitude is at the threshold energy for the @xmath26 state , which is lower by an amount of order @xmath43 . as a result
, we can taylor expand the amplitude in powers of @xmath44 and reproduce this expansion by terms in the effective lagrangian as in eq . .
in this case , the coefficients @xmath35 , @xmath36 ,
are complex valued .
we are particularly interested in their imaginary parts , which come from the deeply inelastic scattering reaction .
these terms mimic the effects of the lower hyperfine states in the effective theory .
it may seem nonintuitive that the effects of inelastic scattering to on - shell particles can be mimicked by local operators , because once an atom in the lower hyperfine state @xmath27 is created , it can propagate over long distances .
in fact , as far as low - energy atoms in the hyperfine state @xmath30 are concerned , the inelastic scattering process is quite local .
this is because the reaction region in which the low - energy atoms disappear can be reconstructed by tracking the inelastic scattering products back to their origin .
the inelastic reaction products have high momenta of order @xmath45 and therefore short wavelengths of order @xmath46 , so they can locate the decay with a resolution of order @xmath46
. therefore the reaction is localized over a region of size @xmath46 , which is very small compared to the wavelengths of the incoming low - energy atoms .
some aspects of the effective field theory obtained by integrating out deeply inelastic reactions are most easily understood by considering a deeply inelastic 1-body reaction , namely the decay of an atom .
the atom could be a metastable excited state , such as the @xmath47 state of @xmath48he , which was one of the earliest atoms for which bose - einstein condensation was achieved @xcite .
the radiative decay of the atom into its ground state is a deeply inelastic reaction .
for simplicity , we assume the ground - state atoms interact weakly with the metastable atoms and that , once they are produced , the ground - state atoms quickly escape from the system .
the hamiltonian @xmath49 for the effective theory is the sum of the hermitian hamiltonian for the metastable atoms and an anti - hermitian piece @xmath50 .
the hermitian operator @xmath2 can be expressed as @xmath51 , where @xmath52 is the decay rate of the metastable atom and @xmath53 is the number operator that counts the metastable atoms : @xmath54 the quantum mechanics of such a theory is unconventional .
the effective hamiltonian @xmath49 commutes with the number operator @xmath53 , so the time evolution generated by @xmath49 does _ not _ change the number of atoms in a state . instead the effects of atom decay must be taken into account by transferring the probability carried by the @xmath55atom component to states containing fewer atoms .
an @xmath55atom state evolves into a superposition of states with @xmath3 and fewer atoms .
the norm of a state containing @xmath3 atoms decreases to zero exponentially with the decay rate @xmath56 .
this is the correct result : if the probability for one atom to still be an atom after time @xmath21 is @xmath57 , the probability for @xmath3 atoms to still be @xmath3 atoms is @xmath58 .
we typically want more information about where the probability goes .
an @xmath55atom state evolves into a superposition of states with @xmath3 , @xmath59 , @xmath60 atoms . in the full theory ,
the superposition can be described by the density matrix @xmath16 . in the effective theory
, we describe the superposition of states with different atom numbers using an _ effective density matrix _ @xmath4 , from which we remove ground - state atoms by tracing over states that include them : @xmath61 the subscript `` deep '' indicates that the partial trace is over states that include deeply inelastic reaction products .
this effective density matrix , like @xmath16 , is hermitian , positive , and has unit trace : @xmath62 .
these properties follow from the definition in eq .
( [ eq : rho - eff - defn ] ) .
naively we might expect the time evolution equation for @xmath4 to be eq . , which reduces to @xmath63 - \frac{i}{2 } \gamma \left\ { n , \rho\right\ } .\
] ] however this equation does _ not _ conserve the total probability @xmath64 .
the correct evolution equation is the lindblad equation : @xmath65 - \frac{i}{2 } \gamma \left\ { n , \rho\right\ } + i \gamma \int d^3{\bm{r}}\,\psi({\bm{r}})\,\rho\,\psi^\dagger({\bm{r } } ) .
\label{eq : muon - evol - eq}\ ] ] time evolution preserves the trace of @xmath4 , because the trace of the additional lindblad term cancels the trace of the anticommutator term in eq . .
the role of the lindblad term is easily understood if we use the evolution equation to calculate the rate of change of the probability @xmath66 for finding @xmath3 metastable atoms in the system .
this probability can be expressed as the partial trace of @xmath4 over states @xmath67 that contain @xmath3 atoms : @xmath68 the partial trace of the evolution equation in eq .
gives @xmath69 the partial trace of the commutator term in eq .
is 0 , because the operator @xmath1 does not change the atom number .
this allows a complete set of @xmath55atom states to be inserted between @xmath1 and @xmath4 . the partial trace of the anticommutator term in eq .
gives @xmath70 , which is the rate at which probability leaves the @xmath55atom sector because of the decay of an atom .
the partial trace of the lindblad term in eq .
gives @xmath71 , which is the rate at which probability enters the @xmath55atom sector from the decay of an atom in the @xmath72atom sector .
this expression can be obtained by inserting complete sets of @xmath72atom states on the left and right of @xmath4 in the lindblad term in eq .
( [ eq : muon - evol - eq ] ) and then rearranging the factors to give @xmath73 the number operator @xmath53 , which is given in eq . , was obtained by replacing the sum of @xmath74 between @xmath75 and @xmath76 by the identity operator .
the number operator in eq . can be replaced by its eigenvalue @xmath77 . the lindblad term in eq .
( [ eq : muon - evol - eq ] ) is essential to get the correct physical behavior for the time evolution of the total number of metastable atoms .
the expectation value of the atom number is @xmath78 we can use eq . to determine the time dependence of @xmath79 : @xmath80 .\ ] ] after shifting the index of the second term on the right side , we obtain @xmath81 this implies that @xmath82 , as expected for a collection of atoms with decay rate @xmath52 .
we now consider a more general effective field theory obtained by integrating out the reaction products from deeply inelastic reactions . by the arguments presented in sec .
[ sec : locality ] , the effects of the deeply inelastic reactions can be reproduced by local anti - hermitian terms in the effective hamiltonian .
the effective hamiltonian can be expressed in the form @xmath83 , where @xmath1 and @xmath2 are both hermitian operators .
the operator @xmath2 can be expressed in the form @xmath84 where the local operators @xmath85 annihilate low - energy atoms in configurations corresponding to inelastic reactions .
the real constants @xmath86 can be determined by matching low - energy observables in the full theory and in the effective theory . the hamiltonian @xmath87 commutes with the number operator @xmath53 for low - energy atoms , so the time evolution generated by @xmath49 does _ not _ change the number of low - energy atoms in a state . instead
the effects of the deeply inelastic reactions must be taken into account by transferring the probability carried by the @xmath55atom component to states containing fewer atoms .
the norm of a state containing @xmath3 atoms decreases to zero at a rate that increases with @xmath3 .
an @xmath55atom state evolves into a superposition of states with @xmath3 and fewer atoms .
we describe this superposition of states using an _ effective density matrix _ @xmath4 defined by tracing over the deeply inelastic decay products as in eq . .
more precisely , we trace out any state containing an atom with momentum exceeding the ultraviolet cutoff of the effective field theory .
the effective density matrix , like the underlying density matrix , is hermitian , positive , and it has unit trace : @xmath62 .
these properties follow from the definition in eq .
( [ eq : rho - eff - defn ] ) . as is familiar in the field of quantum information @xcite , the definition of the effective density matrix @xmath4 as the partial trace in eq .
( [ eq : rho - eff - defn ] ) implies that @xmath20 satisfies a self - contained time evolution equation that is completely determined by @xmath4 and by the full density matrix @xmath88 at some initial time .
this evolution equation is however non - markovian : the time derivative @xmath89 is determined not only by @xmath20 but also by @xmath90 at the earlier times @xmath91 .
this dependence on the history takes into account the effects of high - momentum atoms that are produced by inelastic reactions and subsequently interact with the low - energy atoms .
however the density matrix @xmath20 for the effective theory does not need to reproduce the full density matrix on short time scales of order @xmath92 .
it is sufficient to reproduce its effects after time averaging over time intervals much larger than @xmath92 .
this time average removes transients associated with high - momentum atoms that can not be described accurately within the effective field theory .
it is only after removing these transients that it is possible to have a density matrix @xmath20 that is markovian .
thus the proper definition of the effective density matrix requires that the partial trace in eq .
( [ eq : rho - eff - defn ] ) be supplemented by an appropriate time average that makes @xmath20 markovian .
naively we might expect the time evolution equation for @xmath4 to be eq . , where @xmath2 is the hermitian operator in eq . .
however , this equation does _ not _ conserve the total probability @xmath64 .
the correct evolution equation is the lindblad equation : @xmath93- i \left\{k,\rho\right\ } + 2i \sum_i \gamma_i \int \!\!d^3r\ , \phi_i \rho \phi_i^\dagger .
\label{eq : lindbladphi}\ ] ] as in eq .
, the trace of the additional lindblad term cancels the trace of the anticommutator term .
the local operators @xmath85 , which annihilate configurations of low - energy atoms that can undergo a deeply inelastic reaction , are the _
lindblad operators_. in quantum information theory , a lindblad operator is sometimes called a _
quantum jump operator_. in the low - energy effective theory , a lindblad operator produces a jump to a state with a different low - energy atom number .
given the form of @xmath2 in eq . , the lindblad equation is a necessary consequence of our physical requirements on the effective density matrix : @xmath4 is hermitian , positive , has unit trace , and it is markovian .
an explicit diagrammatic illustration of the emergence of the lindblad equation from tracing over states that include deeply inelastic reaction products is given in appendix [ sec : deductive ] . in order to obtain the lindblad equation in eq . , it is essential that @xmath2 have the structure shown in eq .
( [ eq : k - phi ] ) . a more generic form for this operator is @xmath94 where @xmath95 is a positive hermitian matrix .
it is a hermitian matrix , because @xmath2 is hermitian by definition .
it is guaranteed to be a positive matrix by the optical theorem , which implies that the @xmath96 matrix satisfies @xmath97 .
matrix elements of @xmath2 reproduce the anti - hermitian parts of scattering amplitudes @xmath98 , where @xmath99 and @xmath100 are states in the effective theory that are connected by intermediate deeply inelastic reaction channels .
the hermitian parts of these amplitudes are reproduced by @xmath1 .
the optical theorem guarantees the positivity of the anti - hermitian parts . the double sum in eq .
is easily rewritten in the canonical form in eq . by expanding @xmath95 in terms of outer products of its eigenvectors .
in this section , we use the lindblad equation to determine the time evolution of the mean number of low - energy atoms , which can be changed by deeply inelastic reactions .
we also derive universal relations for the inelastic two - atom loss rate for fermions with two spin states and for the inelastic three - atom loss rate for identical bosons .
we consider an open effective field theory with effective hamiltonian @xmath101 , where @xmath1 and @xmath2 are hermitian . the hermitian operator @xmath2 in the anti - hermitian part can be expressed in terms of local lindblad operators @xmath85 as in eq . .
the time evolution of the density matrix @xmath4 for the effective field theory is given by the lindblad equation in eq . .
we reconsider an atom number operator @xmath53 that is conserved by the reactions in the effective field theory but is violated by deeply inelastic reactions in the full theory . since lindbladian time evolution preserves the trace of @xmath4 , the system average of @xmath53 can be expressed as @xmath102 the statement that @xmath53 is conserved by the reactions in the effective theory can be expressed as the commutation relation @xmath103=0 .
\label{[n , h]}\ ] ] the inelastic reaction responsible for the @xmath104 term in @xmath2 changes the atom number @xmath53 by some integer @xmath105 .
this statement can be expressed as a commutation relation : @xmath106=-k_i \phi_i(\bm{r } ) .
\label{[n , phi]}\ ] ] we now consider the time evolution of @xmath107 . by taking the time derivative of @xmath108 , inserting the time evolution equation in eq . , and rearranging the terms , we get @xmath109 ) - \frac{1}{\hbar } \sum_i \gamma_i \int \!\ !
d^3r \ , { \rm tr } \big ( \rho ( n\phi_i^\dagger \phi_i + \phi_i^\dagger \phi_i n
-2 \phi_i^\dagger n \phi_i ) \big ) .
\label{nevol : lindblad}\end{aligned}\ ] ] the first term on the right side vanishes , because @xmath53 commutes with @xmath1 . by using the commutation relation in eq . in the second term on the right side
, the rate of change of atom number reduces to @xmath110 the rate of change in @xmath107 has been expressed as a sum of expectation values of the same terms that appear in the expression for @xmath2 in eq . but multiplied by integers @xmath111 . the atoms that are studied in cold atom experiments can form deeply bound diatomic molecules whose binding energies are orders of magnitude larger than the energy scales of the cold trapped atoms .
three - body recombination , in which three low - energy atoms collide and two of them bind into a molecule , is a deeply inelastic reaction if the molecule is deeply bound .
the momenta of the outgoing molecule and the recoiling atom are often large enough for them to escape from the trapping potential . in a locally homogeneous gas of atoms in thermal equilibrium
, the rate of decrease in the local number density @xmath112 of low - energy atoms can be expressed as a rate equation : @xmath113 where @xmath114 is the 3-body recombination event rate constant , which depends on the temperature @xmath96 .
bosonic atoms that are all in the same hyperfine state can be described by an effective field theory with a quantum field @xmath30 . at the low temperatures in cold atom experiments ,
the only relevant interaction between the atoms is s - wave scattering , whose strength is determined by the scattering length @xmath115 .
the interaction term in the hamiltonian is @xmath116 this interaction term can be treated using perturbation theory , provided the scattering length @xmath115 is not much larger than the range of the interactions between the atoms , which for ultracold atoms is the van der waals length scale .
we now consider a deeply inelastic three - body recombination reaction that produces a diatomic molecule with binding energy much larger than the temperature @xmath96 .
its effect on low - energy atoms can be described in the effective field theory by adding to the effective hamiltonian the anti - hermitian term @xmath117 , where the local hermitian operator @xmath2 is @xmath118 and where @xmath119 is a constant that does not depend on the temperature .
the expression for @xmath2 in eq . has a single term with lindblad operator @xmath120 and coefficient @xmath121 .
the expression for the atom loss rate in eq . thus has a single term with integer @xmath122 . in a locally homogeneous system of atoms , the local version of the loss rate in eq .
is @xmath123 the expectation value of @xmath124 is the local number density : @xmath125 . in a thermal gas ,
the expectation value of @xmath126 is @xmath127 . comparing with eq .
, we see that the constant @xmath119 in eq . is the @xmath128 limit of the 3-body recombination event rate constant @xmath114 .
its temperature dependence is negligible , because @xmath129 is much smaller than the binding energy of the molecule . in a bose - einstein condensate at zero temperature
, the expectation value of @xmath126 can be expressed as @xmath130 , where @xmath131 is the mean field .
the atom loss rate is given by eq .
except that the prefactor on the right side is smaller than for a thermal gas by a factor of 1/6 .
in cold atom experiments , the scattering length @xmath115 can be controlled experimentally by tuning an external magnetic field to near a feshbach resonance .
if @xmath115 is much larger than the range of the interactions between the atoms , which for ultracold atoms is the van der waals length scale , the interactions must be treated nonperturbatively .
fermionic atoms in two hyperfine states 1 and 2 with scattering length @xmath115 can be described by an effective field theory with quantum fields @xmath132 and @xmath133 .
the interaction term in the hamiltonian can be expressed as @xmath134 where @xmath135 is the bare coupling constant .
if the ultraviolet cutoff @xmath136 is imposed on the momenta of virtual atoms , the bare coupling constant is @xmath137 if the pair of atoms in the hyperfine states 1 and 2 has a _ spin - relaxation _ scattering channel into a pair atoms in lower hyperfine states 3 and 4 , the optical theorem implies that the scattering length @xmath115 is complex with a small negative imaginary part .
the energy release from the spin - relaxation reaction is much larger than the energy scales for ultracold atoms , so this is a deeply inelastic reaction .
the high - momentum atoms in the hyperfine states 3 and 4 can be integrated out to get an effective field theory for low - energy atoms in the hyperfine states 1 and 2 only .
the interaction term in the hamiltonian is still that in eq . , except that now @xmath138 has an anti - hermitian part because the complex scattering length @xmath115 makes the bare coupling constant @xmath135 in eq .
complex . determining the loss rate of atoms due to
the deeply inelastic spin - relaxation reaction is not trivial , because the large scattering length makes the problem nonperturbative .
however , an exact result for the inelastic two - atom loss rate in any state has been proposed by shina tan @xcite .
the rates of decrease in the numbers @xmath139 and @xmath140 of atoms in the two hyperfine states are given by the universal relation @xmath141 where @xmath142 is a property of the system called the _
contact_. the coefficient of @xmath142 , which is proportional to the imaginary part of @xmath143 , is determined by 2-body physics .
the contact was first introduced by shina tan in 2005 @xcite .
it is the thermodynamic variable conjugate to @xmath143 .
the contact @xmath142 is an extensive variable : it is the integral over space of the contact density @xmath144 , which measures the number of 1 - 2 pairs per ( volume)@xmath145 .
( the unusual power of volume arises from an anomalous dimension associated with a non - gaussian renormalization group fixed point @xcite . )
shina tan derived other universal relations involving the contact that hold for any state of the system @xcite .
they include the `` adiabatic relation '' that specifies how the free energy of a system is affected by a change in the scattering length @xcite : @xmath146 many other universal relations involving the contact were subsequently derived @xcite . the universal relation for the inelastic two - atom loss rate in eq . ( [ uni2loss ] )
can be derived by expressing the hermitian operator @xmath2 in the effective hamiltonian @xmath147 as @xmath148 where the local lindblad operator is @xmath149 the operator product @xmath150 is singular : its matrix elements diverge linearly with the cutoff @xmath136 .
the multiplication by @xmath135 , which according to eq .
decreases as @xmath151 for large @xmath136 , makes @xmath152 a finite operator whose matrix elements have well - behaved limits as @xmath153 . in a spin - relaxation scattering reaction ,
the decreases in the atom numbers @xmath139 and @xmath140 are both @xmath154 . according to eq . , the rates of decrease in their system averages are therefore @xmath155 the universal loss rate in eq .
( [ uni2loss ] ) then follows from the identification of the contact density as @xcite @xmath156 this identification can be verified by using the adiabatic relation in eq . .
the only dependence of the hamiltonian on the scattering length @xmath115 is in the interaction term in the hamiltonian density @xmath157 in eq . .
if the tiny imaginary part of the scattering length is neglected , the derivative of the hermitian part @xmath1 of the effective hamiltonian with respect to @xmath143 is @xmath158 by the feynman - hellman theorem , the system average of the left side of this equation is the left side of the adiabatic relation in eq . .
with the identification of the contact density in eq . , the system average of the right side of eq .
is the right side of the adiabatic relation in eq .
this completes the derivation of the inelastic two - atom loss rate in eq .
( [ uni2loss ] ) and the adiabatic relation in eq . . in ref .
@xcite , the authors presented an incorrect derivation of the universal relation for the inelastic two - atom loss rate that suggests that there are additional terms on the right side of eq . .
they assumed the density matrix @xmath4 satisfies the naive time evolution equation in eq . .
this equation implies that the number of atoms is conserved , but that the probability @xmath159 decreases with time . in ref .
@xcite , the authors assumed that the mean atom number @xmath107 was given by @xmath160 .
the resulting expression for the time derivative of @xmath107 can be expressed as a double integral over space of an expression involving the product of the number density operator and the hamiltonian density operator . upon applying the operator product expansion to the product of these operators
, they obtained an expansion for @xmath161 in terms of increasingly higher - dimension operators with the leading term given by eq . .
the derivation of the universal relation using the lindblad equation makes it clear that there are no additional terms beyond the contact term in eq .
( [ uni2loss ] ) .
bosonic atoms that are all in the same hyperfine state can be described by a nonrelativistic quantum field theory with a quantum field @xmath30 @xcite .
if the scattering length @xmath115 is much larger than the range of interactions of the atoms , two - atom interactions and three - atom interactions must both be treated nonperturbatively .
the resulting quantum field theory is characterized by a discrete scaling symmetry in which lengths and time are multiplied by integer multiples of @xmath162 and @xmath163 , respectively , where @xmath164 and @xmath165 is a universal number .
three - atom interactions are determined by a parameter @xmath166 with dimensions of momentum , upon which physical observables can only depend log - periodically .
if @xmath115 is positive , two - atom interactions produce a weakly bound diatomic molecule called the _
shallow dimer_. regardless of the sign of @xmath115 , three - atom interactions produce a sequence of weakly bound triatomic molecules called _
efimov trimers _ @xcite . in ref .
@xcite , several universal relations were derived that hold for any system consisting of identical bosons .
the universal relations relate various properties of the system to two extensive thermodynamic variables : the 2-body contact @xmath167 , which is the analog of the contact for fermions with two spin states , and the 3-body contact @xmath168 .
the atoms that are studied in cold atom experiments form deeply bound diatomic molecules ( _ deep dimers _ ) whose binding energies are orders of magnitude larger than the energy scales of the cold trapped atoms .
the deep dimers provide various pathways for deeply inelastic reactions that result in the loss of three atoms .
they include ( a ) three - body recombination , in which three low - energy atoms collide and two of them bind into a deep dimer , ( b ) dimer relaxation , in which a shallow dimer and an atom scatter into a deep dimer and an atom , and ( c ) efimov trimer decay into a deep dimer and an atom . in ref .
@xcite , a universal relation for the three - atom inelastic loss rate was presented but not derived . in this section , we use the lindblad equation to sketch the derivation of this relation . to derive the universal relation for the three - atom inelastic loss rate ,
we follow the formalism laid out in ref . @xcite . in the absence of deep dimers ,
the interaction term in the hamiltonian can be written as @xmath169 since this effective field theory has ultraviolet divergences , it must be regularized . the bare coupling constants @xmath170 and @xmath171 depend on an ultraviolet momentum cutoff @xmath136 : @xmath172 where @xmath115 is the two - body scattering length of the bosons and @xmath166 is a three - body parameter introduced in ref
the three - body parameter @xmath166 can be determined from any three - body datum , such as the binding energy of an efimov trimer or the atom - dimer scattering length .
( see ref .
@xcite for explicit relations ) .
the numerical prefactor @xmath173 in eq . depends on the ultraviolet cutoff prescription and has the value @xmath174 if @xmath136 is a sharp momentum cutoff @xcite .
three - atom losses from deeply inelastic reactions involving deep dimers can be taken into account by adding an imaginary part to the coupling constant @xmath171 in the hamiltonian in eq . .
the resulting anti - hermitian part of the hamiltonian is @xmath117 , where @xmath2 is @xmath175 physical observables can be expressed particularly conveniently in terms of @xmath166 and an additional real three - body loss parameter @xmath176 @xcite defined by analytically continuing the three - body coupling constant @xmath171 in eq . to complex values using the substitution @xmath177 inserting @xmath122 into the expression in eq .
for the three - atom inelastic loss rate , we obtain the universal relation @xmath178 where @xmath168 is the three - body contact , which can be expressed as @xmath179 the prefactor @xmath180 depends on the ultraviolet cutoff @xcite .
since its precise form is not very insightful , it will not be given here .
the three - body contact can also be defined in terms of the expectation value of the logarithmic derivative with respect to @xmath166 of the hermitian part of the effective hamiltonian in eq . at
fixed @xmath115 @xcite : @xmath181 the universal relation in eq . , with the factor @xmath182 approximated by @xmath183 ,
was previously given in ref .
it was also given previously in ref .
@xcite for the special case of an efimov trimer state .
in this paper , we have shown that integrating out the high - momentum atoms produced by deeply inelastic reactions produces an open effective field theory in which the time evolution of the effective density matrix is given by the lindblad equation .
the effective density matrix can be obtained from the density matrix of the full theory by taking a partial trace over states that include high - momentum atoms and then carrying out an appropriate time average to eliminate high - frequency transients .
the lindblad operators are local , and they can be deduced from the anti - hermitian terms in the effective hamiltonian .
the lindblad terms in the evolution equation are essential to get the correct evolution equation for the mean atom number .
we used the lindblad equation to present the first correct derivation of the universal relation for the two - atom inelastic loss rate for fermionic atoms with two spin states that interact through a large scattering length .
we also used the lindblad equation to present the first derivation of the universal relation for the three - atom inelastic loss rate for identical bosonic atoms with a large scattering length .
the lindblad equation has many other applications to atom loss processes in cold atom physics .
an obvious extension is to heteronuclear systems , for which there are two types of efimov trimers . due to the smaller discrete scaling factor associated with one type of efimov trimer , atom loss processes in such systems
have recently been of great theoretical and experimental interest @xcite .
the lindblad equation could be applied to fermionic atoms with two spin states on the upper branch of the two - atom spectrum at scattering lengths for which three - body recombination into the shallow dimer is a deeply inelastic reaction @xcite .
it could also be applied to losses of dipolar atoms from three - body recombination into deep dimers @xcite .
e.b . and
h .- w.h .
acknowledge useful discussions with shina tan .
the research of e.b . was supported in part by the department of energy under grant de - fg02 - 05er15715 , by the national science foundation under grant phy-1310862 , and by the simons foundation .
the research of h .- w.h . was supported in part by the bmbf under contract 05p15rdfn1 , and by the deutsche forschungsgemeinschaft ( sfb 1245 ) .
the research of g.p.l . was supported in part by the national science foundation .
in this section , we demonstrate how the lindblad structure of the evolution equation for the density matrix emerges from a diagrammatic analysis of a simple field theory model with a deeply inelastic 2-body reaction . to reduce visual clutter ,
we introduce compact notation for the integral over space and for the integral over a momentum : @xmath184 we also set @xmath185 .
we consider a quantum field theory with two fields @xmath30 and @xmath27 .
we refer to the particles annihilated by these field operators as @xmath30 atoms and @xmath27 atoms , respectively .
the field operators satisfy canonical commutation relations if the atoms are bosons and canonical anticommutation relations if they are fermions .
the hamiltonian for the full theory has the form @xmath186 the free - field terms in the hamiltonian are where @xmath29 is the mass of the atoms .
we have chosen the rest energies of a @xmath30 atom and a @xmath27 atom to be zero and @xmath188 , respectively .
the on - shell energies of atoms with momentum @xmath189 are the interaction hamiltonian includes a term that allows a pair of @xmath30 atoms to scatter into a pair of @xmath27 atoms : @xmath191 where the ellipses denote further local interaction terms , such as @xmath192 and @xmath193 .
their precise form will not be needed in the following discussion .
the leading contribution to the transition amplitude for the process @xmath194 from inelastic reactions with a @xmath195 intermediate state comes from the imaginary part of the one - loop diagram in fig .
[ fig : scattering ] . the energy release in the reaction
@xmath196 is @xmath197 , and the corresponding momentum scale is @xmath198 .
we are interested in systems consisting of @xmath30 atoms whose momenta are small compared to @xmath199 .
the reaction @xmath196 is therefore a deeply inelastic reaction .
we refer to the momentum scale @xmath199 as _
high momentum_. because of the large energy release , the deeply inelastic scattering process @xmath196 is effectively local and instantaneous .
it takes place over a spatial region of size @xmath202 and during a time interval of length @xmath203 .
we proceed to show how this locality can be exploited to remove high - momentum @xmath27 atoms from the theory and construct a low - energy effective field theory for @xmath30 atoms .
we do this first for two @xmath30 atoms in this subsection and then for a system containing many @xmath30 atoms in subsection [ sec : effrho ] . for simplicity ,
we assume the coupling @xmath35 is small and we work to leading order in @xmath35 .
the leading contribution to the @xmath194 scattering amplitude from a @xmath195 intermediate state is given by the diagram in fig .
[ fig : scattering ] . using time - ordered perturbation theory ( or using feynman perturbation theory and integrating by contours over the loop energies ) , the off - shell scattering amplitude is @xmath204 where @xmath205 and @xmath189 are the total energy and the total momentum of the two incoming @xmath30 atoms , respectively . for convenience ,
we consider @xmath194 scattering in the center - of - mass frame . in this reference frame
, the incoming @xmath30 atoms have momenta @xmath206 and energies @xmath207 .
the on - shell scattering amplitude is therefore @xmath208 .
the total cross section for @xmath196 scattering can be obtained using the optical theorem by evaluating the imaginary part of @xmath96 on the energy shell .
the real part of the scattering amplitude in eq .
is ultraviolet divergent .
the divergence can be canceled by a renormalization of the coupling constant for the @xmath196 interaction .
after renormalization , the integral over @xmath42 in eq . is dominated by high momenta of order @xmath199 .
consequently we can expand the on - shell scattering amplitude @xmath208 in powers of @xmath209 .
we are primarily interested in constructing an effective hamiltonian that takes into account the leading term @xmath210 .
successively higher powers of @xmath209 could be taken into account through successively higher - order gradient terms in the effective hamiltonian .
similarly , successively higher powers of @xmath211 in the expansion of @xmath96 about the on - shell energy could be taken into account through successively higher - order time - derivative terms in the effective lagrangian . in the sector with only two @xmath30 atoms ,
the leading effect of the scattering amplitude @xmath212 can be taken into account by adding a local term to the free hamiltonian @xmath213 .
the resulting effective hamiltonian is @xmath214 this equation defines the hermitian operators @xmath1 and @xmath2 .
the real part of @xmath210 , which is ultraviolet - divergent , can be cancelled by renormalizing the coupling constant for the @xmath215 interaction term contained in the ellipses in eq .
( [ eq : hint ] ) .
the anti - hermitian part @xmath50 of the effective hamiltonian comes from the imaginary part of @xmath210 , which is @xmath216 the locality of the inelastic scattering process , which implies @xmath217 , allows us to simplify the contributions to correlators involving the operator @xmath218 from its annihilation of @xmath27 atoms that come from @xmath219 scattering .
for example , the correlator @xmath221 illustrated in fig .
[ fig : scattering - simp ] can be simplified as follows : @xmath222 where @xmath96 is given in eq .
( [ psi - selfenergy ] ) .
generally , we can make the following replacements in such situations : replacing the free hamiltonian @xmath213 by the effective hamiltonian @xmath49 is all that is needed to analyze the impact of inelastic scattering processes on states with only two @xmath30 atoms .
analyzing multi-@xmath30 states is more complicated , however , because a system that is described initially by a state with @xmath53 @xmath30 atoms evolves into a superposition of states with @xmath53 , @xmath224 , @xmath225 , @xmath30 atoms .
the state with two @xmath30 atoms also evolves into a superposition , but there are only two states , @xmath226 and @xmath227 , and we do not care about the second one . for @xmath228 , we need the density matrix @xmath229 to track the superposition of states containing different numbers of @xmath30 atoms over time . where @xmath232 and @xmath233 are the vacuum states annihilated by @xmath234 and @xmath235 , respectively .
the full density matrix @xmath16 can be expanded in the basis of direct product states .
its time evolution is given by @xmath236 where @xmath237 is the full hamiltonian in eq .
( [ eq : hamiltonian ] ) .
we can define an effective density matrix @xmath20 by tracing over the @xmath27 states : this operator acts only on @xmath30 states . a convenient basis for the effective density matrix consists of outer products of the @xmath30 states defined in eq . of the form @xmath239 .
the time derivative of the effective density matrix can be obtained by differentiating eq . and using the time dependence of the full density matrix in eq . : @xmath240 the contributions from the kinetic terms in the full hamiltonian in eq .
( [ eq : hamiltonian ] ) are simple : the first equation holds because @xmath213 does not act on @xmath27 states .
the second equation holds because @xmath242 depends only on @xmath27 fields .
holds for any operator @xmath243 constructed out of the field @xmath27 and any operator @xmath244 .
this can be verified by expressing the partial trace as a sum over a complete set of @xmath27 states and inserting a complete set of @xmath27 states between @xmath243 and @xmath245 . ]
the evolution equation reduces to @xmath246 + \mathrm{tr}_\phi\big(h_{\rm int}\,\rho_{\mathrm{full}}- \rho_{\mathrm{full}}\,h_{\rm int } \big).\ ] ] since there are two terms in the interaction hamiltonian in eq . , there are four contributions to the partial trace in eq . .
we first consider the contribution to the partial trace of @xmath247 from the first term in @xmath248 in eq .
( [ eq : hint ] ) .
this contribution is dominated by terms in which @xmath249 annihilates @xmath27 atoms generated ( through interactions ) by the @xmath30-sector of @xmath16 , leading to correlators like eq .
( [ eq : phi2psid ] ) .
we can therefore use the substitution in eq .
( [ eq : phi^2approx ] ) to obtain @xmath250 \longrightarrow -\tfrac{1}{4}t(0,0)\int_{\bm{r } } \psi^{\dagger 2}(\bm{r})\psi^2(\bm{r } ) \ ; \rho.\end{aligned}\ ] ] similarly , we can use the substitution in eq .
( [ eq : phid^2approx ] ) to obtain the contribution to the partial trace of @xmath251 from the second term in @xmath248 : @xmath252 \longrightarrow -\tfrac{1}{4}t^*(0,0)\int_{\bm{r } } \rho \ ; \psi^{\dagger 2}(\bm{r})\psi^2(\bm{r}).\end{aligned}\ ] ] the contributions to the sum of the partial traces in eqs . and
from the real part of @xmath210 can be added to the term @xmath253 $ ] in eq . to get @xmath254 $ ] , where @xmath1 is the hermitian part of the effective hamiltonian defined in eq .
( [ eq : heff ] ) .
the contributions to the sum of the partial traces in eqs .
and from the imaginary part of @xmath210 gives @xmath255 , where @xmath50 is the anti - hermitian part of the effective hamiltonian defined in eq .
( [ eq : heff ] ) .
note that we are making a key physical assumption about @xmath16 when we use the substitution in eq .
( [ eq : phi^2approx ] ) to replace @xmath256 by @xmath257 . as we indicated above
, this substitution is valid provided @xmath249 annihilates @xmath27 atoms produced by the @xmath30-sector of @xmath16 . in principle , it is also possible for @xmath249 to annihilate @xmath27 atoms from the @xmath27-sector of @xmath16 .
we assume that such contributions can be ignored because the probability for finding two @xmath27 atoms at the same space - time point is vanishingly small ( and therefore the probability of an inverse reaction , @xmath258 , is negligible ) .
this is the case , for example , if @xmath16 describes a situation in which all @xmath27 atoms are produced by inelastic @xmath219 reactions and , once produced , they either escape from the system or they interact so weakly with the low - energy @xmath30 atoms that they decouple .
the contribution to the partial trace of @xmath247 from the second term in the interaction hamiltonian @xmath248 in eq .
is @xmath259 = \mathrm{tr}_\phi\left[\tfrac{1}{2}g\int_{\bm{r}}\psi^2(\bm{r})\rho_{\mathrm{full}}\phi^{\dagger 2}(\bm{r})\right].\end{aligned}\ ] ] in the second expression , the factor of @xmath260 has been moved to the right side of @xmath16 . to verify this equality , we begin by inserting the definition of the partial trace on the left side of eq . : @xmath261 the @xmath262 factors remove @xmath27 atoms from the bras on the left side of this equation .
taking into account the symmetry under permutations of the integration variables , their effect can be taken into account by the substitution @xmath263 the kets on the right side of eq .
can be expressed as @xmath264 making the substitutions in eqs .
( [ eq : brasub ] ) and and using the delta functions to integrate over @xmath265 and @xmath266 , we obtain the expression on the right side of eq . .
upon making the substitution in eq . , we obtain @xmath267 \longrightarrow -\tfrac{1}{4}t^*(0,0 ) \int_{\bm{r } } \psi^2(\bm{r})\,\rho\ , \psi^{\dagger 2}(\bm{r}).\end{aligned}\ ] ] similarly , the contribution to the partial trace of @xmath251 from the second term in @xmath248 in eq .
( [ eq : hint ] ) is @xmath268 \longrightarrow -\tfrac{1}{4}t(0,0)\int_{\bm{r } } \psi^2(\bm{r})\,\rho\,\psi^{\dagger 2}(\bm{r}).\end{aligned}\ ] ] the real part of @xmath210 cancels in the difference between the contributions in eqs . and .
the contributions from the imaginary part of @xmath210 to the partial trace in eq . is proportional to the integral over space of the operator @xmath269 . adding the four contributions in eqs .
, , , and to the partial trace in eq .
, we see that the evolution equation for the effective density matrix is @xmath270 - \frac{i}{4}\,{\rm im}t(0,0 ) \int_{\bm{r } } \big [ \psi^{\dagger2}\psi^2(\bm{r})\ , \rho + \rho\ , \psi^{\dagger2}\psi^2(\bm{r } ) -2\ , \psi^2(\bm{r})\ , \rho\ , \psi^{\dagger2}(\bm{r } ) \big],\end{aligned}\ ] ] where @xmath271 is given in eq . .
this has the standard form of the lindblad equation .
the last term removes @xmath30 atoms two at a time to account for their disappearance due to inelastic scattering into pairs of high - momentum @xmath27 atoms .
e. braaten and h .- w .
hammer , universality in few - body systems with large scattering length , phys .
rept . * 428 * , 259 ( 2006 ) [ cond - mat/0410417 ] .
h. georgi , effective field theory , ann .
nucl . part .
sci . * 43 * , 209 ( 1993 ) .
kaplan , effective field theories , nucl - th/9506035 .
manohar , effective field theories , lect .
notes phys .
* 479 * , 311 ( 1997 ) [ hep - ph/9606222 ] .
r. shankar , effective field theory in condensed matter physics , in _ conceptual foundations of quantum field theory _ , ed .
cao ( cambridge university press , 1999 ) [ cond - mat/9703210 ] .
burgess , introduction to effective field theory , ann .
nucl . part .
sci . * 57 * , 329 ( 2007 ) [ hep - th/0701053 ] .
e. braaten , h .- w . hammer , and g.p .
lepage , open effective field theories from deeply inelastic reactions , arxiv:1607.02939 .
g. lindblad , on the generators of quantum dynamical semigroups , commun . math . phys . * 48 * , 119 ( 1976 ) .
v. gorini , a. kossakowski , and e.c.g .
sudarshan , completely positive dynamical semigroups of @xmath53-level systems , j. math .
* 17 * , 821 ( 1976 ) .
s. grozdanov and j. polonyi , viscosity and dissipative hydrodynamics from effective field theory , phys .
d * 91 * , 105031 ( 2015 ) [ arxiv:1305.3670 ] .
burgess , r. holman , g. tasinato , and m. williams , eft beyond the horizon : stochastic inflation and how primordial quantum fluctuations go classical , jhep * 1503 * , 090 ( 2015 ) [ arxiv:1408.5002 ] .
burgess , r. holman , and g. tasinato , open efts , ir effects and late - time resummations : systematic corrections in stochastic inflation , jhep * 1601 * , 153 ( 2016 ) [ arxiv:1512.00169 ] .
j. preskill , lecture notes for physics 229 : quantum information and computation ( unpublished ) , chapter 3 .
f. pereira dos santos , j lonard , ju . wang , c.j .
barrelet , f. perales , e. rasel , c.s .
unnikrishnan , m. leduc , and c. cohen - tannoudji , bose - einstein condensation of metastable helium , phys .
lett . * 86 * , 3459 ( 2001 ) [ cond - mat/0103387 ] .
e. braaten , universal relations for fermions with large scattering length , lect .
notes phys .
* 836 * , 193 ( 2012 ) [ arxiv:1008.2922 ] .
e. braaten and l. platter , exact relations for a strongly interacting fermi gas from the operator product expansion , phys .
lett . * 100 * , 205301 ( 2008 ) [ arxiv:0803.1125 ] .
e. braaten and h .- w .
hammer , universal relation for the inelastic two - body loss rate , j. phys .
b * 46 * , 215203 ( 2013 ) [ arxiv:1302.5617 ] .
e. braaten , d. kang , and l. platter , universal relations for identical bosons from 3-body physics , phys .
lett . * 106 * , 153005 ( 2011 ) [ arxiv:1101.2854 ] .
d. h. smith , e. braaten , d. kang , and l. platter , two - body and three - body contacts for identical bosons near unitarity , phys .
lett . * 112 * , 110402 ( 2014 ) [ arxiv:1309.6922 ] .
bedaque , h .- w . hammer , and u. van kolck , renormalization of the three - body system with short range interactions , phys .
lett . * 82 * , 463 ( 1999 ) [ nucl - th/9809025 ] .
e. braaten and h .- w .
hammer , enhanced dimer deactivation in an atomic / molecular bec , phys . rev . a * 70 * , 042706 ( 2004 ) [ cond - mat/0303249 ] .
f. werner and y. castin , general relations for quantum gases in two and three dimensions .
ii . bosons and mixtures , phys .
a * 86 * , 053633 ( 2012 ) [ arxiv:1210.1784 ] .
k. helfrich , h .- w . hammer , and d.s .
petrov , three - body problem in heteronuclear mixtures with resonant interspecies interaction , phys .
a * 81 * , 042715 ( 2010 ) [ arxiv:1001.4371 ] .
tung , k. jimenez - garcia , j. johansen , c.v .
parker , and c. chin , geometric scaling of efimov states in a @xmath272li@xmath273cs mixture , phys .
lett . * 113 * , 240402 ( 2014 ) [ arxiv:1402.5943 ] .
r. pires , j. ulmanis , s. hfner , m. repp , a. arias , e.d .
kuhnle , and m. weidemller , observation of efimov resonances in a mixture with extreme mass imbalance , phys .
lett . * 112 * , 250404 ( 2014 ) [ arxiv:1403.7246 ] . | the loss of ultracold trapped atoms due to deeply inelastic reactions has previously been taken into account in effective field theories for low - energy atoms by adding local anti - hermitian terms to the effective hamiltonian . here
we show that when multi - atom systems are considered , an additional modification is required in the equation governing the density matrix .
we define an effective density matrix by tracing over the states containing high - momentum atoms produced by deeply inelastic reactions .
we show that it satisfies a lindblad equation , with local lindblad operators determined by the local anti - hermitian terms in the effective hamiltonian .
we use the lindblad equation to derive the universal relation for the two - atom inelastic loss rate for fermions with two spin states and the universal relation for the three - atom inelastic loss rate for identical bosons . |
since the pioneering work by anderson and fazekas@xcite , the spin-1/2 heisenberg antiferromagnet on a triangular lattice has been extensively investigated as a promising candidate to realize a spin - liquid ground state induced by geometric frustration and quantum fluctuations . yet , in spite of a large amount of theoretical and experimental works , we do not have any unified picture for this system . on the theoretical side ,
most of the numerical studies carried out over the past decade with a variety of different techniques do not support that the suggested spin - liquid ground state is realized in this model .
instead they provide evidences to indicate the ground state with the three - sublattice order where the average direction of neighboring spins differs by a @xmath1 angle@xcite .
then the linear spin wave theory ( lswt)@xcite well describes numerical results calculated on lattices with finite sizes . on the experimental side ,
several novel materials with triangular structures have been investigated recently .
one of these materials is cs@xmath2cucl@xmath3 @xcite , which is supposed to reduce to the one - dimensional spin-1/2 quantum heisenberg antiferromagnet because of its anisotropy@xcite .
other interesting materials are @xmath4-(bedt - ttf)@xmath2cu@xmath2(cn)@xmath5@xcite and etme@xmath5sb[pb(dmit)@xmath2]@xmath2 @xcite , which are considered to be close to the heisenberg antiferromagnet on an isotropic triangular lattice
. these materials , however , do not show any magnetic long - range order down to the quite low temperature compared with the exchange interactions . through further studies motivated by these experiments
, theorists have found that fundamental properties on a triangular lattice are quite different from those on a square lattice , while antiferromagnets on both lattices have the semi - classical long - range orders .
the dispersion relation is one of the properties that have been investigated to compare systems with different geometries .
recently the series expansion study@xcite and the full 1/s calculation of the spin wave theory@xcite on this relation show that on a triangular lattice one sees a downward renormalization of the higher energy spectra , while on a square lattice one sees an upward renormalization .
the former authors also point out that the roton minimum is present in the relatively flat region of the dispersion relation on the triangular lattice .
these features are quite different from the predictions of the lswt . in these somewhat confusing situations
one needs unbiased numerical studies which do not depend on any special physical assumption .
the stochastic state selection ( sss ) method , a new type of monte carlo method which we have developed in several years@xcite , has such a good property .
one can therefore employ the method to evaluate any quantity in any system one wants to investigate . in the algorithm of the sss method we consider the full hilbert space of the system and stochastically
reduce it to relatively small one so that we can manage it in numerical studies . more concretely
, we use a set of stochastic variables which are as many as basis states of the whole vector space under consideration , but most of these variables are valued to be zero
. then any inner product related to an arbitrary operator is calculable using the survived non - zero stochastic variables .
statistical averaging processes guarantee in a mathematically justified manner that the result becomes the correct value of the inner product .
it is found that several constraints on the set of stochastic variables are helpful to obtain better results with less statistical errors . using this constrained sss method
we started our numerical study on the spin-1/2 quantum heisenberg antiferromagnet on a 48-site triangular lattice .
we have estimated lowest energy eigenvalues of the model for each sectors with @xmath7 , where @xmath8 denotes the @xmath9 component of the total spin @xmath10 of the system@xcite . in this paper
we make a further investigation of the model by means of the constrained sss method with two new applications .
one of them is to accurately calculate expectation values of operators which contain many off - diagonal elements in their representations . by evaluating the sublattice magnetization and the static structure function
we demonstrate that it is possible to obtain accurate knowledge of the ground state in this method .
it should be noted that in the usual quantum monte carlo method these physical quantites are not easy to calculate even for non - frustrated systems .
another is an extension to employ a new set of basis states with which complex coefficients are inevitable in an expansion of an arbitrary state . using this set of basis states in the constrained sss method we successfully calculate low - lying one magnon spectra with non - zero wave vectors .
it should also be noted that even for non - frustrated systems such as the quantum heisenberg antiferromagnet on a square lattice we can not do without complex numbers in calculations with non - zero wave vectors .
our study in this paper performed by means of the constrained sss method gives reliable results calculated from the first principle .
we see that our results are compatible with those in refs.@xcite .
it therefore supports the realization of an ordered ground state in the model . at the same time
, however , it adds an evidence that dynamical properties of the system are not described by the lswt .
the plan of this paper is as follows . in section [ sec2 ]
we make brief descriptions of the model and the method .
subsection [ subsec21 ] is to define the hamiltonian of the model we study .
in addition we comment on the power method .
an operator related to the hamiltonian is introduced here so that we can obtain the lowest eigenvalue of the hamiltonian using the power method . in subsection [ subsec22 ]
the sss method is shortly reviewed . here
an introduction of a two - valued stochastic variable that might be zero is essential .
then the constrained sss method is explained . in section [ sec3 ]
we present our numerical results .
subsection [ subsec31 ] is to show our definition and results on the sublattice magnetization and the static structure function .
these results are compared with those reported on smaller lattices . on the sublattice magnetization
we also make a comparison with results calculated by the lswt . in subsection [ subsec32 ]
we show our results on the one magnon spectra . for nine independent states in the brillouin zone of a 48-site triangular lattice
we present estimated values of energy eigenvalues .
then the one - magnon dispersion relation is studied in comparison with results by the lswt .
section [ sec4 ] is devoted to summary .
here we also count up characteristic features of our method . finally an appendix is given to explain our sets of basis states in detail .
the hamiltonian of the model on the @xmath11-site lattice is @xmath12 where @xmath13 denotes the spin-1/2 operator on the @xmath14-th site and the sum runs over all @xmath15 bonds of the @xmath11-site lattice .
the coupling @xmath16 is set to 1 throughout this paper . in our calculations
we employ the power method .
the sss method is used to calculate expectation values of powers of an operator @xmath17 , @xmath18 where @xmath19 denotes the identity operator and @xmath20 is a positive number . for each quantum number
we choose one value of @xmath20 which ensures that the lowest energy eigenvalue of the system corresponds to the eigenvalue of @xmath21 whose absolute value is the largest .
detailed explanations are given in a previous paper@xcite .
in the power method we can estimate the energy eigenvalue @xmath22 by @xmath23 where @xmath24 denotes a state which approximates the exact eigenstate @xmath25 of @xmath21 , @xmath26 since @xmath27 holds .
we estimate an expectation value of an arbitrary operator @xmath28 in the eigenstate @xmath25 using @xmath29 note that in this calculation a contamination appears at @xmath30 order because @xmath31 the stochastic state selection is realized by a number of stochastic variables .
let us expand a normalized state @xmath32 in an @xmath33-dimensional vector space by a basis @xmath34 , @xmath35 then for each @xmath36 we generate a stochastic variable @xmath37 following to the on - off probability function , @xmath38 with a positive parameter @xmath39 which is common to all @xmath40 @xmath41 .
note that ( [ probf ] ) implies @xmath42 or @xmath43 , statistical averages being @xmath44 and @xmath45 . in numerical studies
we can neglect basis states whose @xmath37=0 .
the parameter @xmath46 therefore controls the size of the partial vector space we should consider in each sampling .
then calculations in the full vector space are to be recovered through statistical averaging processes . a random choice operator @xmath47 is then defined by @xmath48 using this @xmath47 we obtain a state @xmath49 , @xmath50 which has less non - zero elements than @xmath32 has .
it should be kept in mind that with any operator @xmath28 the statistical average @xmath51 is exactly equal to the expectation value @xmath52 because @xmath53 for all @xmath36 . in calculations using the power method
we generate independent random choice operators the stochastic variable @xmath37 in @xmath54 is generated independently to the one in @xmath55 with @xmath56 . ]
@xmath57 , @xmath58 , @xmath59 to obtain an improved state @xmath60 from @xmath24 .
namely , we define @xmath61 and @xmath62 in the constrained sss method , we introduce @xmath63 ( @xmath64 ) constraints which can be fulfilled by making @xmath63 @xmath37s depend on other @xmath65 independently generated stochastic variables for each random choice operator .
a detailed description on this improvement is given in our previous paper@xcite together with numerical studies with a @xmath66 case .
in the same manner as ref.@xcite , we use the state @xmath67 instead of the above mentioned @xmath60 . here
@xmath67 is defined by @xmath68 and @xmath69 with random choice operators @xmath70 , @xmath71 , @xmath59 , each of which is generated under a constraint @xmath72 we can evaluate an expectation value of an operator @xmath28 by a statistical average of the ratio @xmath73 with large enough value of @xmath74 . actually , in order to save the computing time , either of the following two ratios is employed in our numerical study instead of ( [ ratioo ] ) , @xmath75 or @xmath76 where @xmath77 denotes another random choice operator , which is generated independently to @xmath78 with the constraint @xmath79 .
it should be noted that the method is applicable for a state which has complex coefficients as well as a state with real coefficients . for a complex coefficient @xmath80
we obtain , after the stochastic selection , @xmath81 or @xmath82 if @xmath83 . with complex coefficients we newly have two technical difficulties .
one of them is that we need double computer memory to treat complex numbers .
another is that we can not use the symmetries of rotation and reflection which reduce the number of independent basis states .
in this subsection we present our results of the sublattice magnetization and of the static structure function in the ground state , which we evaluate on the 48-site lattice using the ratio defined by ( [ ratiomo ] ) .
we start from an approximate state constructed in the sector with @xmath84 since we want evaluate them in the ground state of the system .
detailed explanations of the way to calculate this approximate state are given in a previous paper@xcite .
since the squared sublattice magnetization , which we denote by @xmath85 for the system with @xmath11 sites , is defined in ref.@xcite as @xmath86 the operator @xmath28 in ( [ ratioo ] ) for @xmath85 should be @xmath87 where a , b and c are sublattices of the @xmath11-site triangular lattice and @xmath88 our estimation of the sublattice magnetization with @xmath89 is @xmath90 .
[ fig : mag ] plots this result together with those reported on @xmath91 , 27 and 36 lattices@xcite .
as is shown by calculations up to the 36-site lattice , the finite - size analysis on the magnetization suggests a finite value of the magnetization in the thermodynamic limit .
we see that our result is consistent with those for smaller lattices , which supports the scaling predicted by the lswt , @xmath92 where @xmath93 , @xmath94 and @xmath95@xcite .
the finiteness of the sublattice magnetization , @xmath96 , is also observed in fig .
[ fig : mag ] .
next we show results of the static structure function @xmath97 in the ground state , for which the operator @xmath28 is defined by @xmath98 with @xmath99 possible values of @xmath100 on the 48-site lattice , where @xmath101=@xmath102 , are obtained from the brillouin zone shown in fig .
[ fig : bz ] .
they are @xmath103 , @xmath104 , @xmath105 , @xmath106 , @xmath107 , @xmath108 , @xmath109 and 1 .
numerical results with a normalization factor @xmath110 are plotted in fig .
[ fig : sf ] as a function of @xmath100 . in order to make a comparison between our data and previously reported ones for a smaller @xmath11
, we also plot the structure function calculated on an @xmath111 lattice@xcite in fig .
[ fig : sf ] .
we observe that in the full range of @xmath100 our @xmath89 results lie slightly below those calculated on the 36-site lattice . in this subsection
we discuss the one magnon spectra with @xmath112 .
detailed descriptions of wave vectors @xmath113 and a set of basis states @xmath114 we use here are given in the appendix .
for each value of @xmath115 on the 48-site lattice , we start from an approximate state @xmath116 expanded by a set of basis states @xmath114 with complex coefficients , @xmath117 which is used as @xmath24 in ( [ phindef ] ) , ( [ constcd ] ) and so on .
the state @xmath116 is prepared following descriptions in a previous paper@xcite .
in @xmath116 we constructed for each @xmath113 , the number of basis states with non - zero @xmath118 , which we denote by @xmath119 , ranges from 270 million to 1.6 billion .
we evaluate an upper bound of the lowest energy eigenvalue @xmath120 , using the ratio @xmath121 in ( [ ratioom ] ) with the operator @xmath122 .
the results are presented in table [ tab : result ] together with values of @xmath20 and @xmath119 .
[ cols="<,^,<,^,^",options="header " , ] in fig .
[ fig : spectra ] we plot a gap energy of one magnon with a wave vector @xmath123 , which is calculated from data in table [ tab : result ] by @xmath124 along the path @xmath125 in fig .
[ fig : bz ] .
we also plot the dispersion relation calculated by the lswt@xcite in fig .
[ fig : spectra ] .
we see that our data is consistent with the gapless state at the ordering wave vector , which is expected by the full breaking of the @xmath126 symmetry .
the gap energy at the central region of the brillouin zone is much smaller than that calculated by the lswt analysis .
this downward renormalization is in contrast with the upward renormalization found in the square lattice . in the same central region
we also find that the spectra is flat or dispersionless . 0.5 cm
in this paper we have reported further developments on numerical calculations by means of the stochastic state selection ( sss ) method , which has the following good properties . * with this method we can study large systems compared to the exact diagonalization@xcite . *
this method presents a solution for the so - called sign problem .
note that the ordinary quantum monte carlo method , which has been quite powerful to study spin systems on bipartite lattices@xcite , does not work in most frustrated systems .
* one does not need any physical assumption .
this is in contrast to the situation that recent variational approaches are based on some assumptions which are specific to the system to be studied .
for example , the coupled - cluster method@xcite is based on an eigenstate with a special form . in the density matrix renormalization group method @xcite
, on the other hand , one assumes that the ground state is represented by the matrix product state .
another example is the stochastic reconfiguration method@xcite whose origin is assigned to the fixed - node monte carlo method@xcite . in these methods
an effective hamiltonian is used instead of the exact one .
* in the sss method it is easy to treat complex coefficients in the expansion of states .
this is impossible in the ordinary monte carlo method . using this method we numerically study the spin-@xmath105 quantum heisenberg antiferromagnet on the 48-site triangular lattice .
we evaluate the sublattice magnetization and the static structure function in the ground state of the system .
we also calculate the one magnon spectra with non - zero wave vectors .
our results on the magnetization and the structure function show that the sss method is quite effective to estimate these quantities .
the accuracy of the data , which depends on how many samples are available to calculate statistical averages , is satisfying within our computing resources .
we observe that the results we obtain are consistent with the reported ones on smaller lattices .
therefore we see that the linear spin wave theory describes the results well so that an analysis based on the spontaneous symmetry breaking on the semi - classical nel order ground state is acceptable . on the magnon spectra , we make sure that the one magnon energy at the ordering wave vector is consistent with the goldstone mode , as is expected in the spin wave analysis . for other values of the wave vectors , however , we observe flat spectra with strong downward renormalization , which are quite different from the spectra predicted by the lswt .
we find our results are qualitatively consistent with results by the series expansion study@xcite and by the order @xmath0 calculation of the spin wave analysis@xcite .
finally let us again emphasize that we can get physical insight for the frustrated quantum spin systems from numerical results by the sss method because the method is well developed to calculate various observables such as energy , the magnetization and the magnon spectra .
especially it is notable that using this method we can evaluate the magnon spectra .
while neither the ordinary monte carlo method nor the variational monte carlo method is applicable to calculate this quantity , we can use the sss method in the same manner as the one used to evaluate the ground state energy . in this appendix
we summarize our definition of wave vectors @xmath113 and two sets of basis states @xmath127 and @xmath114 employed in this paper .
the former set @xmath127 , which is introduced for the 48-site lattice in the previous work with @xmath128 @xcite , is used to calculate the sublattice magnetization and the static structure function . while the latter set @xmath114 is newly introduced here in calculations of the one magnon spectra with non - zero wave vectors .
let us denote two basis vectors of the triangular lattice by @xmath129 and @xmath130 , which are described by unit vectors @xmath131 and @xmath132 in @xmath133- and @xmath134-directions , @xmath135 the wave vectors are defined by @xmath136 where @xmath137 @xmath138,2 @xmath139 are two basis vectors of the reciprocal lattice for which @xmath140 holds . to construct a state with a fixed wave vector @xmath113 , we first define translation operators @xmath141 in the directions @xmath142 @xmath143 . for any state
@xmath144 the translation operators @xmath145 and @xmath146 produce new states if @xmath144 is not an eigenstate of them .
we define , with @xmath147 @xmath143 , @xmath148 using these states we make an eigenstate common to @xmath145 and @xmath146 , @xmath149 here the sum runs over all @xmath150 and @xmath151 for which @xmath152 s are linearly independent .
the normalization factor @xmath153 in ( [ comeigens ] ) is therefore @xmath154 when all @xmath155 are independent to each other .
it is easy to see that @xmath156 with the periodic boundary conditions in both directions . in order to describe our sets of basis states we consider a state @xmath157 having a fixed @xmath9 component @xmath158 or @xmath159 on each site @xmath14 of the lattice .
then @xmath160 @xmath161 s form a set of basis states @xmath162 .
the set of basis states @xmath127 is a complete set constructed from @xmath163 s for which the @xmath9-component of the total spin @xmath164 .
each state @xmath127 is defined by a linear combination of several @xmath163 s so that it has translational symmetries for the zero momentum and even under the @xmath165 rotation , @xmath166 rotation and the reflection .
total number of the basis states , namely @xmath33 , in this set amounts @xmath167 @xmath168c@xmath169 @xmath170 @xmath171 for the @xmath89 lattice @xmath161 s , but some contains @xmath161 s less than @xmath172 . ] .
the set of basis states @xmath114 should be constructed for each wave vector @xmath113 .
a basis state @xmath173 is determined by a state @xmath174 in the @xmath112 sector , using it as @xmath144 in ( [ phivrdef ] ) .
then the set of basis states @xmath175 , which fulfil @xmath176 is completed with as many states @xmath161 s as to span the full hilbert space with @xmath112 .
note that with this set of basis states one should use complex coefficients in expansions of states in the calculations .
the total number of the basis states for the 48-site lattice is @xmath174 s . ]
@xmath177 @xmath178 @xmath179 . | we numerically study the magnetization and the dispersion relation of a frustrated quantum spin system .
our method , which is named the stochastic state selection method , is a kind of monte carlo method to give eigenstates of the system through statistical averaging processes . using the stochastic state selection method with some constraints
, we make a successful study of the spin-1/2 heisenberg antiferromagnet on a 48-site triangular lattice .
we calculate the sublattice magnetization and the static structure function in the ground state .
our result on the sublattice magnetization is consistent with the value given by the linear spin wave theory .
this adds an evidence for the analysis based on the spontaneous symmetry breaking of the semi - classical nel order in the ground state .
we also evaluate the low - lying one magnon spectra of the model with all wave vectors available on a 48-site triangular lattice .
we find that at the ordering wave vector there is a goldstone mode , which is in good agreement with the result from the spin wave analysis .
the magnon spectra with other wave vectors , however , are quite different from results obtained by the linear spin wave theory .
we observe a flat dispersion relation with a strong downward renormalization .
our results are compatible with those recently reported in the series expansion study and in the order @xmath0 calculation of the spin wave analysis . |
the most interesting objects in theoretical physics are arguably black holes . to understand their dynamics we need to put together two widely accepted theories of nature : general relativity ( einstein s classical theory of gravity ) and quantum mechanics .
black hole thermodynamics is the crossroad between the classical and the quantum pictures .
discovery of hawking radiations lead to the identification of black holes as thermodynamic objects with physical temperature and entropy @xcite .
this discovery paved the way for progress in the understanding of spacetime , quantum mechanically @xcite .
variation in the mass , @xmath1 , of a rotating black hole having angular momentum , @xmath2 , and electric charge @xmath3 , obeys the formalism @xmath4known as first law of thermodynamics . here
@xmath5 is the angular velocity of the horizon , @xmath6 is the electric potential on the horizon and @xmath7 is the entropy of the event horizon @xmath8 .
the important results of the black hole thermodynamics are the association of temperature ( @xmath9 ) and entropy @xmath10 with surface gravity @xmath11 and area @xmath12 of the black hole event horizon respectively . the phenomena of phase transition in black hole thermodynamics was first observed long ago @xcite .
schwarzschild black hole has negative specific heat and evaporates via hawking radiation .
the ads schwarzschild black holes have a different behavior towards temperature and heat capacity .
there are two types of black holes in ads spacetime : the smaller one , like the usual schwarzschild black hole , with negative specific heat ( unstable ) , and the big black holes having positive specific heat ( locally stable ) .
axisymmetric , stationary , and electrically charged black holes in einstein - maxwell theory with arbitrary surrounding matter , always have regular inner horizon ( cauchy horizon ) ( @xmath13 ) and an outer horizon ( event horizon ) ( @xmath8 ) , if the angular momentum and charge of the black hole do not vanish at the same time @xcite . in recent years products of thermodynamics parameters , specially area and entropy , at both horizons of black holes has gained attention in general relativity and string theory @xcite . it is observed that the area product of the outer and inner horizons is independent of black hole mass @xmath1 . for a regular axisymmetric and stationary spacetime in einstein maxwell gravity
these products are @xcite : @xmath14 and @xmath15 note that both the above given products are mass independent so these are universal quantities .
this universal sense of area and entropy products holds for all known five dimensional asymptotically flat black rings , and for black strings @xcite .
the microscopic degrees of freedom of the black hole are described in terms of those of a conformal field theory ( cft ) .
the area product of the inner and outer horizons of a black hole in three dimensions is @xmath16 where @xmath17 and @xmath18 are the number of right and left moving excitations of the two - dimensional cft @xcite , i.e. @xmath19 in other words , the products of the areas of the killing horizons is independent of the mass of any asymptotically flat black hole in d - dimensional spacetime , therefore depends on the quantized charges @xcite .
hence mass independence of area products , is necessary condition for holographic cft description . in @xcite
the thermal products for rotating black holes are studied . in @xcite area products for stationary black hole horizons
are calculated .
it has been shown that the area products are independent of adm ( arnowitt - deser - misner ) mass parameter and depends on the quantized charge and quantized angular momentum parameter for all known five dimensional asymptotically flat black rings and black strings .
it may sometimes also fail , e.g. in @xcite authors show that entropy products are not mass independent in general myers - perry black holes when spacetime dimension @xmath20 , and kerr - ads black holes with @xmath21 .
the kerr / cft correspondence from the thermodynamics of both outer and inner horizons was investigated recently @xcite .
authors prove that the first law of thermodynamics of the outer horizon guaranty that of the inner horizon , under some assumption , and mass independence of the entropy product @xmath22 is equivalent to the condition @xmath23 .
furthermore , using the thermodynamics method , information of the dual cft could be obtained easily , because thermodynamics of the outer and inner horizons give the thermodynamics in the left and right moving sectors of the dual cft @xcite .
so the central charges and the temperatures in all possible pictures can be obtained in a simple way .
newman et .
al . , obtained the solution of the einstein - maxwell equation in kerr space - time , as a rotating ring of mass and electric charge @xcite .
applying the ernst s formulation ( for axisymmetric stationary fields ) , tomimatsu and sato discovered the series of solutions for the gravitational field of a rotating mass @xcite .
yamazaki obtained the charged kerr - tomimatsu - sato family of solutions with some distortion parameter @xmath24 ( integer ) in the gravitational fields of rotating masses @xcite .
static spherically symmetric julia - zee dyon solution in curved spacetime were obtained by kasuya et .
later , an exact stationary rotating dyon solution in tomimatsu - sato- yamazaki space - time , was proposed @xcite . furthermore , both the `` schwinger '' and the `` julia - zee '' dyon exact solutions in kerr - newman space - time , i.e. for @xmath25 , were studied .
this solution is known as kerr - newman - kasuya ( knk ) black hole in literature @xcite .
this solution is featured by four physical parameters ( mass @xmath1 , angular momentum @xmath2 , electric charge @xmath3 , and magnetic charge @xmath26 ) . in this work
we consider the knk black hole and discuss its thermodynamic at horizons and generalize some already existing results for cauchy horizon @xmath13 .
we also find the knk / cft correspondence using the thermodynamics method .
plan of the work is as follows : in section 2 , we discuss the knk black hole metric and the products of surface gravity , temperature and komar energy , angular velocities and electromagnetic potentials calculated on the inner and outer horizons .
products of area and entropy at @xmath27 are also calculated , and these products are independent of mass of the black hole . in subsection 2.1 ,
the smarr formula for the black hole is derived and first law of thermodynamics is verified . in subsection 2.2 , we have discussed the irreducible mass and obtained the expression relating , the rest mass and irreducible mass . in subsection 2.3 , heat capacity of black hole is calculated , and phase transition is discussed . in 2.4 , we find the cft relation of knk black hole and re - obtained the expressions of temperature and entropy . in 2.5 section
we concluded the work .
we use units in which @xmath28 .
the spherically asymmetric and rotating solutions for einstein - maxwell field equations , in the presence of electric and magnetic field , was proposed by kasuya @xcite .
metric in the boyer- lindquist coordinates is : @xmath29 where @xmath30the electromagnetic potential is : @xmath31 + \epsilon p d\phi,\ ] ] here @xmath1 is mass of the black hole , @xmath3 is the electric charge , @xmath26 is the magnetic charge , @xmath2 is the angular momentum , and @xmath32 .
the horizons of the black hole are obtained by solving @xmath33 , given as @xmath34 where @xmath35 is the outer horizon named as event horizon @xmath8 and @xmath36 is the inner horizon known as cauchy horizon @xmath37 . both cauchy horizon and event horizon
are null surfaces with infinite blue and red - shifts respectively .
cauchy horizon is a null surface of infinite blue - shift , while the event horizon is an infinite red - shift surface @xcite .
product of the two horizons : @xmath38 does not directly depend on mass of the black hole but on charge ( electric and magnetic ) and spin parameter of the black hole .
area and bakenstein - hawking entropy at the horizons @xmath39 of the black hole are @xcite : @xmath40 and @xmath41 surface gravity is the acceleration due to gravity at the horizon of a black hole .
it is defined as the force required to an observer at infinity , for holding a particle ( of unit mass ) in place at the event horizon , given as @xcite @xmath42 @xmath43 hawking temperature of the horizons is : @xmath44 the komar energy is given by @xcite : @xmath45 electromagnetic potential due to electric and magnetic charges is @xmath46 and @xmath47 respectively .
angular velocity at @xmath39 is @xmath48 product of surface gravities , surface temperatures and komar energies at @xmath39 are @xmath49},\label{k11}\ ] ] @xmath50},\ ] ] and @xmath51 respectively .
products of electromagnetic potentials @xmath52 at @xmath53 , ( due to electric and magnetic charges ) are @xmath54 and products of angular velocities @xmath5 at @xmath39 is @xmath55it is clear that all these products ( except product of horizons @xmath56 ) are depending on mass of the black hole directly , so these quantities are not universal .
we also calculate the products of areas and entropies at @xmath39 , @xmath57,\label{k area}\ ] ] and @xmath58.\ ] ] note that both products are independent of mass , so these are universal quantities .
the physical parameters defined in previous section for knk black hole obey the symmetries @xmath59 solving together @xmath60 , the black hole parameters @xmath61 and @xmath62 can be written in terms of horizons ( @xmath63 ) , @xmath3 and @xmath26 as @xmath64 @xmath65},\nonumber\\ & & { \cal a}_{\pm}= 4\pi [ r_{\pm}(r_++r_-)-q^2-p^2],\nonumber\\ & & \phi_{\pm}^q= \frac{q r_{\pm}}{r_+(r_++r_-)-q^2-p^2},\nonumber\\ & & \phi_{\pm}^p= \frac{p r_{\pm}}{r_+(r_++r_-)-q^2-p^2},\nonumber\\ & & \omega_{\pm}=\frac{\sqrt{r_+r_--q^2-p^2}}{r_{\pm}(r_++r_-)-q^2-p^2}.\end{aligned}\ ] ] using the above relations the first law of thermodynamics @xmath66 can be verified easily .
the first law at the inner horizon can be written by exchanging @xmath67 in all the quantities given in eq .
( [ k2.5 ] ) , because @xmath63 are on equal footing .
this symmetry of @xmath63 gives @xcite @xmath68 further @xmath69 these products are independent of @xmath1 . in @xcite authors prove that mass independence of entropy products together with first laws at both the horizons imply that @xmath70 for several black holes .
the expression for area of the black hole can be rewritten using the idea proposed by smarr @xcite as : @xmath72.\ ] ] the area of both horizons must be constant given by : @xmath73.\ ] ] using eq .
( [ k15 ] ) mass of the black hole or adm mass is expressed in terms of the areas of the horizons @xmath74 since the first law of thermodynamics states that change in mass of a black hole is related to change in its area , angular momentum and electric charge , in other words @xmath75 where @xmath76\nonumber\\ \phi_{\pm } ^q&= & \text{electromagnetic potential at horizons due to electric charge $ q$}\nonumber\\ & = & \frac{q r_{\pm}}{2{\cal m}r_{\pm}-q^2-p^2}\nonumber\\ \phi_{\pm } ^p&= & \text{electromagnetic potential at horizons due to magnetic charge $ p$}\nonumber\\ & = & \frac{p r_{\pm}}{2{\cal m}r_{\pm}-q^2-p^2}\nonumber\\ \omega_{\pm}&= & \text{angular velocity at horizons}\nonumber\\ & = & \frac{a}{r_{\pm}^2+a^2}.\end{aligned}\ ] ] we can rewrite effective surface tension as : @xmath77 \nonumber\\ & & = \frac{1}{32 \pi { \cal m}}\big[1-\frac{16\pi { \cal m}^2_{\pm}-{\cal a}_{\pm}-8 \pi ( p^2+q^2)}{{\cal a}}\big]\nonumber\\ & & = \frac{1}{32 \pi { \cal m}}\big[1-\frac{(8\pi { \cal m}^2 - 4 \pi(p^2+q^2))}{4\pi(r_{\pm}^2+a^2 ) } \big ] \nonumber\\ & & = \frac{1}{8 \pi}\big[\frac{r_{\pm}^2-{\cal m}}{r_{\pm}^2+a^2}\big]\nonumber\\ & & = \frac{\kappa_{\pm}}{8\pi},\end{aligned}\ ] ] hence first law of thermodynamics is derived from smarr formula . the mass of a black hole could be increased or decreased but there is no way to decrease the irreducible mass @xmath78 of a black hole . since second law of black hole thermodynamics states that the surface area @xmath79 of both the horizons @xmath80 never decreases @xcite i.e. @xmath81 so there is a one to one connection between the surface area and irreducible mass of a black hole .
christodoulou showed that the irreducible mass of a black hole is proportional to the square root of its surface area @xcite .
the irreducible mass of knk black hole expressed in terms of its area is:@xmath82 the irreducible mass defined on inner and outer horizons is @xmath83 and @xmath84 respectively .
the product of the irreducible mass at the horizons @xmath85 is : @xmath86 this product is universal because it does not depends directly on mass of the black hole .
the expression for the rest mass of the rotating charged black hole given by christodoulou and ruffini in terms of its irreducible mass , angular momentum and charge is @xcite : @xmath87 for knk black hole , expression of mass in terms of irreducible mass is : @xmath88 or @xmath89 where @xmath90 .
the nature ( positivity or negativity ) of heat capacity reflects the change in the stability properties of the thermal system ( black hole ) .
a black hole with negative heat capacity is in unstable equilibrium state i.e. by emitting hawking radiations it may decay to a hot flat space or by absorbing a radiation it may grow without limit @xcite .
heat capacity of a black hole is given by @xmath92 where mass @xmath1 in terms of @xmath63 is : @xmath93 the partial derivatives of mass @xmath1 and temperature @xmath94 with respect to @xmath35 are : @xmath95 and from eq .
( [ k9 ] ) we have @xmath96 and @xmath97 while positive for @xmath98 , so it undergoes phase transition from instable to stable region .
heat capacity diverges at @xmath99 .
here we chose @xmath100 , @xmath101 , @xmath102 , and divergence occurs at @xmath103 .
, width=340 ] the expression for heat capacity @xmath104 for knk black hole becomes : @xmath105 note that heat capacity will be positive if + @xmath106 , and @xmath107 are positive or negative together , and heat capacity is negative if these factors are of opposite signs .
the region where heat capacity is negative , corresponds to an instable region around black hole , whereas a region in which the heat capacity is positive , represents a stability region .
behavior of heat capacity given in eq .
( [ k21 ] ) is shown in fig .
( [ 1k ] ) .
in this section we calculate the central charge , temperature and entropy of knk .
there are three parameters @xmath2 , @xmath3 and @xmath26 of knk black hole besides @xmath108 .
these three parameters correspond to three @xmath109 symmetries ( rotational and gauge symmetries ) of the back ground .
since for each @xmath109 symmetry there is a cft dual , we call these j - pictures , q - picture and p - picture .
the first law of thermodynamics is satisfied for both , the left and right moving sectors independently @xcite .
temperatures and entropies of the left and right moving sectors can be defined as @xmath110 where @xmath111 the temperature of these two sectors can be calculated by considering some results as mentioned in @xcite , i.e. mass independence of the area products guarantees the existence of cft description , further the central charges of the two sectors of dual cft would be equal , since in knk black hole , product of entropies given in eq .
( [ k entropy ] ) is mass independent , so these results hold for knk black hole . to find the temperatures directly from the thermodynamical relations we use the same procedure as discussed in @xcite . as in @xcite authors started discussion with btz black hole , ( then extended to other black holes , in 4d and 5d ) , that the first law of thermodynamics for left and right sectors can be written as @xmath112where @xmath113 are the left and right moving temperatures in dual cft , also known as j - picture cft temperatures .
note that both sides of eq .
( [ k5.17 ] ) are dimensionally correct .
the cardy formula for entropies of the left and right - moving sectors becomes @xmath114 whereas @xmath115 , as mentioned earlier .
once we have exact values of the central charges , the temperatures can be calculated using cardy formula .
these temperatures are proportional to the microscopic temperatures obtained by hidden conformal symmetries in low frequency scattering @xcite .
the hidden conformal symmetries of knk black hole have been discussed in @xcite , these are exactly same as that of kerr newman black hole , but replacing @xmath116 by @xmath117 . in this work
we find the temperatures and central charges of cft duals , by applying thermodynamics method .
we use eqs .
( [ k5.2 ] ) and ( [ k5.3 ] ) with eq .
( [ k2.5 ] ) to find the temperatures and entropies of the left and right moving sectors for knk black hole , @xmath118}\nonumber\\ t_l&=&\frac{1}{4\pi(r_++r_- ) } \nonumber\\ { \cal s}_r&=&\frac{\pi}{2}(r_+^2-r_-^2)\nonumber\\ { \cal s}_l&=&\frac{\pi}{2}[(r_++r_-)^2 - 2q^2 - 2p^2]\nonumber\\ \omega_r&=&\frac{\sqrt{r_+r_--q^2-p^2}}{(r_++r_-)^2 - 2q^2 - 2p^2}\nonumber\\ \omega_l&=&0\nonumber\\ \phi^q_r&=&\frac{q(r_++r_-)}{2[(r_++r_-)^2 - 2q^2 - 2p^2]}\nonumber\\ \phi^q_l&=&\frac{q}{2(r_++r_-)}\nonumber\\ \phi^p_r&=&\frac{p(r_++r_-)}{2[(r_++r_-)^2 - 2q^2 - 2p^2]}\nonumber\\
\phi^p_l&=&\frac{p}{2(r_++r_-)},\end{aligned}\]]where @xmath119 and @xmath120 are the electromagnetic potentials of right and left moving sectors , due to the electric and magnetic charges , @xmath3 and @xmath26 , defined as @xmath121the first law of thermodynamics for the right and left moving sectors is @xmath122 for @xmath123 , eq .
( [ k5.19 ] ) reduces to eq .
( [ k5.16 ] ) , so we calculate the microscopic temperatures for knk in j - picture , defined in eq .
( [ k5.17 ] ) , as @xmath124 using eq .
( [ k5.4 ] ) with eq .
( [ k5.22 ] ) we get the central charges in so called j - picture as @xmath125 if we set @xmath126 we get @xmath127 to balance eq .
( [ k5.27 ] ) dimensionally we use quantized charges , and define the electric and magnetic charges @xcite@xmath128 where @xmath129 are integers . using these charges
one can define @xmath130 and @xmath131 for both horizons @xcite @xmath132 the corresponding electromagnetic potentials for left and right sectors are @xcite : @xmath133 using eqs .
( [ k9 ] ) , ( [ 4.13rn ] ) and ( [ 4.14rn ] ) , we have @xmath134,\end{aligned}\ ] ] @xmath135}{2(r_+r_-)(r_+^2+r_-^2 + 2a^2)},\end{aligned}\ ] ] and @xmath136,\end{aligned}\ ] ] @xmath137}{2(r_+r_-)(r_+^2-r_-^2)},\end{aligned}\ ] ] the angular velocities at left and right sectors are : @xmath138 and we obtain @xmath139and the first law of thermodynamics in terms of the quantized charges becomes @xmath140 or we can write it as @xmath141 for knk black hole , three cft pictures are there , named as the q , p and j - picture . setting @xmath142 , we get cft dual in q - picture @xmath143 with temperature as @xmath144 where @xmath145 is the scale factor , we have @xmath146 @xmath147 from the cardy formula given in eq .
( [ k5.4 ] ) , we can find central charges for q - picture using @xmath148 given as @xmath149 and @xmath150 setting @xmath151 , we get cft dual in p - picture @xmath152 with temperature as @xmath153 where @xmath154 , and are given by @xmath155},\ ] ] and @xmath156}.\ ] ] the central charges for p - picture are @xmath157for knk black hole it becomes @xmath158},\ ] ] and @xmath159}.\ ] ] the q - picture and p - pictures are electronic and magnetic pictures . for @xmath160 , we are left with cft dual in j - picture and the change in @xmath2 is linked with temperature and entropy as @xmath161 where @xmath162 where @xmath163 and we get @xmath164and @xmath165 using @xmath166 we obtain the central charges for j - picture as @xmath167 if there are two dual pictures , there could be a class of dual pictures related by @xmath168 transformations with each other @xcite , as explained for @xmath169d kerr - newman and @xmath170d myers - perry black holes in @xcite .
similarly , one can generate more dual pictures from @xmath168 for knk black hole .
some important thermodynamical parameters for kerr - newman - kasuya black hole with reference to their event and cauchy horizons are studied .
we find the products of these parameters calculated at both horizons of the black hole , and observe that the product of surface gravities , surface temperature product , product of komar energies , products of electromagnetic potentials and angular velocities at the horizons are not universal quantities , being mass dependant , while products of areas and entropies at both the horizons are independent of mass of the black hole .
we derive the expression for heat capacity at the horizons . using the heat capacity expression , stability regions of the black hole
can be estimated , and it is observed graphically for knk black hole . the first law of thermodynamics is also obtained from the smarr formula approach .
the relations that are independent of the black hole mass are of particular interest because these may turn out to be `` universal '' and hold for more general solutions , with nontrivial surroundings .
the relation @xmath171 holds for knk black hole , so it has a cft dual in the einstein gravity and knk / cft correspondence is achievable .
we redefine the thermodynamics by applying the quantization approach on the first laws .
the temperatures of the q , p , and j - pictures of cft duals and the central charges are calculated , with the use of cardy formula .
it is observed that the central charges of the two sectors are identical , which is in agreement with the work done in @xcite , i.e. the mass independence of the area product of the horizons implies the similarity of the central charges of the two sectors .
furthermore , we obtained the temperature of the black hole horizon by thermodynamics approach and by microscopic cft , both temperatures differ by a scale factor .
we would like to thank prof .
bin chen , and jia - ju zhang for valuable discussions on cft correspondence of knk black hole .
99 d. christodoulou , _ phys .
lett . _ * 25 * , 1596 ( 1970 ) .
d. christodoulou and r. ruffini , _ phys
. rev . _ * d 4 * , 3552 ( 1971 ) .
r. penrose and r. m. floyd , _ nature _ * 229 * , 177 ( 1971 ) ; s. hawking , _ phys .
* 26 * , 1344 ( 1971 ) .
j. d. bekenstein , _ phys .
* d 7 * 2333 ( 1973 ) ; j. d. bekenstein , _ phys .
_ * d 9 * 3292 ( 1974 ) .
s. w. hawking , _ nature _ * 248 * 30 ( 1974 ) .
s. w. hawking , " _ commun .
phys . _ * 43 * , 199 ( 1975 ) .
f. larsen , _ phys .
rev . _ * d56 * , 1005 ( 1997 ) .
[ hep - th/9702153 ] .
a. castro and m. j. rodriguez , _ phys .
* d 86 * , 024008 ( 2012 ) .
j. wang , w. xu and x. meng , _ j. high energy phys .
_ * 01 * , 031 ( 2014 ) .
a. curir , _ nuovo cimento . _
* b 51 * , 262 ( 1979 ) .
d. pavon , _ phys .
* d 43 * , 2495 ( 1990 ) ; s. w. hawking , d. n. page , _ commun . math .
phys . _ * 87 * , 577 ( 1983 ) .
m. cvetic and f. larsen , _ j. high energy phys . _ * 0909 * , 088 ( 2009 ) .
e. newman , k. chinnaparad , a. exton , a. prakash and r. torrence , _ j. math
. phys . _ * 6 * , 918 ( 1965 ) .
a. chamblin , r. emparan , c.v .
johnson and r.c .
myers , _ phys .
rev . _ * d 60 * , ( 1999 ) 104026 .
a. strominger , _ j. high energy phys . _ * 9802 * , 009 ( 1998 ) [ hep - th/9712251 ] . t. hartman , k. murata , t. nishioka and a. strominger , _ j. high energy phys . _ * 0904 * , 019 ( 2009 ) [ arxiv:0811.4393 [ hep - th ] ] . h. lu , j. mei and c. n. pope , _ j. high energy phys . _ * 0904 * , 054 ( 2009 ) . b. chen , s. -x .
liu and j. -j .
zhang , _ j. high energy phys .
_ * 1211 * 017 ( 2012 )
. m. ansorg and j. hennig , _ class .
* 25 * ( 2008 ) 222001 ; m. ansorg and j. hennig , _ phys .
* 102 * ( 2009 ) 221102 .
m. cvetic , g. w. gibbons , and c. n. pope , _ phys .
* 106 * ( 2011 ) 121301 ; a. castro and m. j. rodriguez , _ phys .
* d 86 * ( 2012 ) 024008 [ arxiv:1204.1284 [ hep - th ] ] ; m. cvetic , h. lu and c. n. pope , arxiv:1306.4522 [ hep - th ] .
p. pradhan , _ eur .
j. _ * c 74 * , 2887 ( 2014 ) .
prarx14 p. pradhan , arxiv : 1503.04514 [ gr - qc ] . m. visser , _ phys .
* d 88 * , 044014 ( 2013 ) .
h. liu , h. lu , m. luoa and k. shao , _ j. high energy phys . _
* 12 * , ( 2010 ) 054 .
a. tomimatsu and h. sato , _ phys . rev .
* 29 * , 1344 ( 1972 ) .
f. j. ernst , _ phys .
rev . _ * 167 * , 1175 ( 1968 ) ; * 168 * , 1415 ( 1968 ) ; phys . rev .
d 7 , 2520 ( 1973 ) .
m. yamazaki , _
phys . _ * 18 * , 2502 ( 1977 ) ; * 19 * , 1376 1978 .
m. kasuya and m. kamata , nuovo cimento * a 66 * , ( 1981 ) 1 . m. kamata and m. kasuya , _ phys .
_ * b 103 * , 351 ( 1981 ) .
m. kasuya _ phys .
* d 25 * , ( 1982 ) 4 .
s. chandrashekar , the mathematical theory of black holes , clarendon press , oxford ( 1983 ) .
s. hawking , _ phys .
* 26 * , 1344 ( 1971 ) .
a. komar , _ phys .
rev . _ * 113 * , 934 ( 1959 ) .
l. smarr , _ phys .
_ * 30 * , ( 1973 ) 71 ; l. smarr , _ phys . rev . _ * d 7 * , 289 ( 1973 ) .
j. m. bardeen , b. carter , s. w. hawking , _ commun .
phys . _ * 31 * , 161 ( 1973 ) .
d. j. gross , m. j. perry , and l. g. yaffe , _ phys .
_ * d 25 * , 330 ( 1982 ) .
m. cvetic and f. larsen , _ phys .
* d 56 * , 4994 ( 1997 ) .
m. cvetic and f. larsen , _ nucl .
* b 506 * , 107 ( 1997 ) .
b. chen , s. liua and j. j. zhanga , arxive : [ hep - th/ 1206.2015 ] .
d. chen , p. wang , h. wu , h. yang , _ grg _ , * 43 * , 181 ( 2011 ) .
e. witten , _ phys .
b 86 _ , 283 ( 1979 ) . b. chen and j. -j .
zhang , _ j. high energy phys . _
* 1108 * , 114 ( 2011 ) . | * abstract : * we investigate the thermodynamics of kerr - newman - kasuya black hole on the inner and outer horizons .
products of surface gravities , surface temperatures , komar energies , electromagnetic potentials , angular velocities , areas , entropies , horizon radii and the irreducible masses at the cauchy and the event horizons are calculated .
it is observed that the product of surface gravities , surface temperature product and product of komar energies , electromagnetic potentials and angular velocities at the horizons are not universal quantities for kerr - newman - kasuya black hole .
products of areas and entropies at both the horizons are independent of mass of the black hole . heat capacity is calculated and phase transition is observed , under certain conditions on @xmath0 . using the thermodynamics method with quantized charges ( known as refined thermodynamics ) , the central charges and the holographic pictures ( j - picture , q - picture , and p - picture ) of the dual cft for kerr - newman - kasuya black hole are determined . |
the optical properties of small metal particles have been of interest to physicists since the time of maxwell@xcite . such particles , if subjected to light of wavelength much larger than their linear dimensions , exhibit optical resonances due to localized electronic excitations known as `` particle '' or surface " plasmons
. these plasmons can give rise to characteristic absorption peaks , typically in the near - infrared or the visible , which may play an important role in the optical response of dilute suspensions of metal particles in a dielectric host@xcite .
because of recent advances in sample preparation , it has become possible to study _ ordered _ arrays of metal particles in a dielectric host@xcite . in one - dimensional ordered arrays of such closely - spaced particles , waves of plasmonic excitations can propagate along the chains , provided that the interparticle spacing is small compared to the wavelength of light@xcite . in this limit ,
the electric fields produced by the dipole moment of one nanoparticle induces dipole moments on the neighboring nanoparticles .
the dispersion relations for both transverse ( @xmath1 ) and longitudinal ( @xmath0 ) plasmonic waves can then be calculated in the so - called quasistatic approximation@xcite , in which the curl of the electric field is neglected .
while this approximation neglects some significant coupling between the plasmonic waves and free photons@xcite , it gives reasonable results over most of the brillouin zone .
interest in such plasmonic waves has grown vastly in recent years@xcite . in this paper , we extend the study of propagating plasmonic waves in two ways .
first , we calculate the dispersion relation for such plasmonic waves when the metallic chain is immersed in an anisotropic host , such as a nematic liquid crystal ( nlc ) . using a simple approximation
, we show that both the @xmath0 and @xmath1 waves have modified dispersion relations when the director is parallel to the chain axis .
if the director is perpendicular to that axis , we show that the previously degenerate @xmath1 branches are split into two separate branches . secondly , we consider the effects of a static magnetic field applied either parallel and perpendicular to the chain .
for the parallel case we show that a linearly polarized @xmath1 wave is rotated as it propagates along the chain . for a field of 2 tesla and reasonable parameters for the metal
, this faraday rotation may be as large as 1 - 2@xmath3 over ten interparticle spacings . a perpendicular field mixes together the @xmath0 branch and one of the @xmath1 branches , leading to two elliptically polarized branches .
these results suggest that either an nlc host or an applied magnetic field could be used as an additional `` control knob '' to manipulate the properties of the propagating waves in some desired way .
the remainder of this paper is organized as follows . in the next section
, we present the formalism which allows one to calculate the dispersion relations for @xmath0 and @xmath1 waves in the presence of either an anisotropic host or an applied dc magnetic field . in section iii
, we give simple numerical examples , and we follow this by a brief concluding discussion in section iv .
we consider a chain of identical metal nanoparticles , each a sphere of radius @xmath4 , arranged in a one - dimensional periodic lattice along the @xmath5 axis .
the n@xmath6 particle is assumed centered at @xmath7 ( @xmath8 ) .
the propagation of plasmonic waves along such a chain of nanoparticles has already been considered extensively for the case of isotropic metal particles embedded in a homogeneous , isotropic medium@xcite .
various works have considered the quasistatic case in which the electric field is assumed to be curl - free ; this is roughly applicable when both the radius of the particles and the distance between them are small compared the wavelength of light @xcite .
the extension of such studies to include radiative corrections , i. e. , to the case when the electric fields can not be approximated as curl - free , has also been carried out ; these corrections can be very important even in some long - wavelength regimes @xcite .
here we consider how the plasmon dispersion relations are modified when the particle chain is immersed in an anisotropic dielectric , such as an nlc , or subjected to an applied dc magnetic field .
for the case of metallic particles immersed in an nlc , we assume that the host medium is a uniaxial dielectric , with principal dielectric constants @xmath9 , @xmath9 , and @xmath10 . for metal particles in the presence of an applied magnetic field ,
we take the host medium to be vacuum , with dielectric constant unity . in the absence of a magnetic field ,
the medium inside the particles is assumed to have a scalar dielectric function @xmath11 .
if there is a magnetic field along the @xmath5 axis , the dielectric function of the particles becomes a tensor , whose components may be written @xmath12 with all other components vanishing @xcite . in the calculations below , we will assume that the nanoparticles are adequately described by a drude dielectric function .
in this case , the components of the dielectric tensor take the form @xcite @xmath13 and @xmath14 here @xmath15 is the plasma frequency , @xmath16 is a relaxation time , and @xmath17 is the cyclotron frequency , where @xmath18 is the magnetic field , @xmath19 is the electron mass , and @xmath20 is its charge .
we will use gaussian units throughout . in the limit @xmath21 , we may write @xmath22 we first assume that the host has a dielectric tensor @xmath23 with principal components @xmath9 , @xmath9 , and @xmath10 .
such a form is appropriate , for example , in a nematic liquid crystal below its nematic - to - isotropic transition .
we begin by writing down the electric field at @xmath24 due to a sphere with a polarization @xmath25 . in component form
, this field takes the form ( see , e. g. , ref @xcite ) @xmath26 where repeated indices are summed over , and we use the fact that @xmath27 . in eq .
( [ eq : polfield ] ) , @xmath28 is the polarization of the metallic particle , @xmath29 is the dielectric function of the metal particle , and @xmath30 denotes a 3@xmath313 matrix whose elements are @xmath32 where @xmath33 is a green s function which satisfies the differential equation ( see , e. g. , ref .
@xcite ) @xmath34 if the host dielectric tensor is diagonal and uniaxial with diagonal components @xmath9 , @xmath9 and @xmath10 , which we take for the moment to be parallel to the @xmath35 , @xmath36 , and @xmath5 axes respectively , this green s function is given by @xcite @xmath37^{-1/2}. \label{eq : gfaniso}\ ] ] physically , @xmath38 represents the i@xmath6 component of electric field at @xmath24 due to a unit point dipole oriented in the j@xmath6 direction at @xmath39 , in the presence of the anisotropic host .
the next step is to use this result to obtain a self - consistent equation for plasmonic waves along a chain immersed in an anisotropic host . to do this
, we consider the polarization of the n@xmath6 particle , which we write as @xmath40 , where @xmath41 is the electric field within the n@xmath6 particle and @xmath42 .
this field , in turn , is related to the external field acting on the n@xmath6 particle and arising from the dipole moments of all the other particles .
we approximate this external field as uniform over the volume of the particle , and denote it @xmath43
. this approximation should be reasonable if the particle radius is not too large compared to the separation between particles ( in practice , an adequate condition is probably @xmath44 , where @xmath4 is the particle radius and @xmath45 the nearest neighbor separation )
. then @xmath46 and @xmath43 are related by @xcite @xmath47 where @xmath48 is a " depolarization matrix defined , for example , in ref .
@xmath43 is the field acting on the n@xmath6 particle due to the dipoles produced by all the other particles , as given by eq .
( [ eq : polfield ] ) .
hence , the dipole moment of the n@xmath6 particle may be written @xmath49 where @xmath50 is a `` t - matrix '' describing the scattering properties of the metallic sphere in the surrounding material .
finally , we make the assumption that the portion of @xmath43 which comes from particle n@xmath51 is obtained from eq .
( [ eq : polfield ] ) as if the spherical particle @xmath52 were a point particle located at the center of the sphere ( this approximation should again be reasonable if @xmath53 ) . with this approximation , and combining eqs .
( [ eq : polfield ] ) , ( [ eq : pinn ] ) , and ( [ eq : tmatrix ] ) , we obtain the following self - consistent equation for coupled dipole moments : @xmath54 let us first assume that the principal axis of the anisotropic host coincides with the chain direction , which we take as the @xmath5 axis . in this case , the @xmath0 and @xmath1 waves decouple and can be treated independently , because one of the principal axes of the @xmath48 tensor coincides with the chain axis .
first , we consider the longitudinally polarized waves . to find their dispersion relation , we need to calculate @xmath55 . from the definition of this quantity , and from the fact that @xmath56 , we can readily show that @xmath57 .
hence , we obtain the following equations for the @xmath58 s : @xmath59 for transverse modes , the relevant green s function takes the form @xmath60 .
the resulting equation for the dipole moments takes the form @xmath61 in the isotropic case with a vacuum host , @xmath62 , and @xmath63 , the equations for both the parallel and perpendicular cases reduce to the results obtained in ref .
@xcite for both @xmath0 and @xmath1 modes , as expected . for an anisotropic host and only nearest neighbor interactions , the dispersion relation for the @xmath1 waves
is implicitly given by @xmath64 and for the @xmath0 waves by @xmath65 the forms of @xmath66 and @xmath67 are well known ( see , e. g. , ref .
@xcite , where they are denoted @xmath68 and @xmath69 ) .
we rewrite them here for convenience : @xmath70 \nonumber \\ \gamma_{\perp } & = & -\frac{1}{2 } \left[\gamma_{\| } + \frac{1}{\sqrt{\epsilon_{\perp}\epsilon_{\|}}}\frac{\sin^{-1}\sqrt{\lambda}}{\sqrt{\lambda}}\right ] \label{eq : gammaxz}\end{aligned}\ ] ] where @xmath71 .
if we assume that the metallic particle has a drude dielectric function of the form @xmath72 , then the dispersion relation for @xmath1 waves becomes @xmath73 where @xmath74 and that of the @xmath0 waves is @xmath75 where @xmath76 eqs .
( [ eq : tdisp1 ] ) and ( [ eq : ldisp1 ] ) neglect damping of the waves due to dissipation within the metallic particles . to include this effect
, one can simply solve eq .
( [ eq : twave1 ] ) or ( [ eq : lwave1 ] ) for @xmath77 , using the drude function with a finite @xmath16 .
the resulting @xmath77 will be complex in both cases ; the inverse of the imaginary part of @xmath77 will give the exponential decay length of the @xmath1 or @xmath0 wave along the chain .
now , let us repeat this calculation but with the principal axis of the liquid crystalline host parallel to the @xmath35 axis , while the chain itself again lies along the @xmath5 axis .
the self - consistent equation for the dipole moments again takes the form ( [ eq : selfconsist ] ) , but the diagonal elements of @xmath78 are given by @xmath79 , where now @xmath80^{-1/2}.\ ] ] for the case of interest , @xmath81 , @xmath82 , and one finds that @xmath83 , @xmath84 and @xmath85 . the self - consistency condition determining the relation between @xmath86 and @xmath87
can be written out , for all three polarizations , including only nearest neighbor dipole - dipole interactions , in the form @xmath88 .
substituting in the values of @xmath89 for the three cases , we obtain @xmath90 here @xmath91 , @xmath92 , where @xmath93 and @xmath94 are given by eqs .
( [ eq : gammaxz ] ) .
similarly , @xmath95 , while @xmath96 .
these equations can again be solved for @xmath77 in the three cases , with or without a finite @xmath16 , leading to dispersion relations with or without single - particle damping .
next , we turn to a chain of metallic nanoparticles in an external magnetic field , which we initially take to be parallel to the @xmath5 axis .
for such a system , we assume that the metal dielectric tensor is of the form ( [ eq : dielectric_matrix ] ) , ( [ eq : epsw ] ) and ( [ eq : aw ] )
. the cases of a dilute suspension of metallic particles@xcite , or of a random composite of ferromagnetic and non - ferromagnetic particles@xcite , have been treated previously .
once again , we take the chain of particles to lie along the @xmath5 axis , with the n@xmath6 particle centered at @xmath97 .
the self - consistent equation for the dipole moments is still eq .
( [ eq : selfconsist ] ) , but now the elements of both @xmath78 and @xmath48 are different from the case of an nlc host . for a chain of particles parallel to the @xmath5 axis , @xmath78 is diagonal , with non - zero elements @xmath98 , @xmath99 , where @xmath100 and we have assumed that the host has a dielectric constant equal to unity .
the tensor @xmath48 is also diagonal , with nonzero elements @xmath101 , i = 1 , 2 , 3 .
the quantity @xmath102 , where @xmath11 is now the dielectric tensor of the metallic particle . using eq .
( [ eq : dielectric_matrix ] ) to evaluate this tensor , we obtain the following result for the tensor @xmath103^{-1}$ ] , to first order in the quantity @xmath104 , which is assumed to be small : @xmath105 using these expressions , we can now write out the self - consistent linear equations for the oscillating dipole moments and obtain dispersion relations for the modes . once again , the longitudinal and transverse modes decouple . for the longitudinal modes ,
the self - consistent equation simplifies to @xmath106 this is the same as the equation for the longitudinal modes in the absence of a magnetic field and gives the same dispersion relation . for the transverse modes , the @xmath35 and @xmath36 components of the polarization
are now coupled , and satisfy the equations [ writing @xmath107 @xmath108 \nonumber \\
p_{ny } & = & -\frac{4\pi}{3}a^3 \sum_{n^\prime \neq n } \left[t_{yx}{\cal g}_{xx}(z_n - z_{n^\prime})p_{n^\prime , x } + t_{yy}{\cal g}_{yy}(z_n - z_{n^\prime})p_{n^\prime , y}\right ] \label{eq : pxymag}\end{aligned}\ ] ] we can simplify these equations using the fact that @xmath109 , @xmath110 , and @xmath111 to obtain @xmath112 and @xmath113 where @xmath114 and @xmath115 . thus , the equations for left- and right - circularly polarized waves are decoupled . to obtain explicit dispersion relations for left- and right - circularly polarized waves , we assume , as before , that @xmath116 , and substitute the known forms for the quantities @xmath117 , @xmath118 , and @xmath119 , with the following result : @xmath120 ^ 2}\right]\sum_{n=1}^\infty\frac{cos(nk_\pm d)}{n^3}. \label{eq : dispers1}\ ] ] in the special case where we include dipolar interactions only between nearest neighbors , this relation becomes @xmath121 ^ 2}\right]\cos(k_\pm d ) .
\label{eq : dispers2}\ ] ] since the frequency - dependence of both @xmath11 and @xmath104 is assumed known , these equations represent implicit relations between @xmath86 and @xmath122 for these transverse waves . by solving for @xmath123 in eq .
( [ eq : dispers1 ] ) or ( [ eq : dispers2 ] ) , one finds that left and right circularly polarized transverse waves propagating along the nanoparticle chain have slightly different wave vectors @xmath124 and @xmath125 for the same frequency @xmath86 . since a linearly polarized wave
is composed of an equal fraction of right and left circularly polarized waves , this behavior corresponds to a _
rotation _ of the plane of polarization of a linearly polarized wave , as it propagates down the chain , and is analogous to the usual faraday effect in a _
the angle of rotation per unit chain length may be written @xmath126.\ ] ] in the absence of damping , @xmath127 is real .
if @xmath16 is finite , the electrons in each metal particle will experience damping within each particle , leading to an exponential decay of the plasmons propagating along the chain .
this damping is automatically included in the above formalism , and can be seen most easily if only nearest neighbor coupling is included .
the quantity @xmath128 is then the _ complex _ angle of rotation per unit length of a linearly polarized wave propagating along the chain of metal particles . by analogy with the interpretation of a complex @xmath127 in a homogeneous bulk material , re@xmath129 represents the angle of rotation of a linearly polarized wave ( per unit length of chain ) , and im@xmath129 as the corresponding faraday ellipticity - i. e. , the degree to which the initially linearly polarized wave becomes elliptically polarized as it propagates along the chain . for a magnetic field perpendicular to the chain (
let us say , along the @xmath35 axis ) , the elements of the matrix @xmath130 become @xmath131/[\epsilon(\omega)+2 ] \nonumber \\
t_{yz } = -t_{zy } & = & -3ia(w)[\epsilon(\omega)-1]/[\epsilon(\omega)+2]^2 , \end{aligned}\ ] ] with other elements equal to zero .
the transverse waves polarized parallel to the @xmath35 axis are unaffected by the magnetic field , and are described by the equations @xmath132 the @xmath36 and @xmath5 polarized waves , with components @xmath133 and @xmath58 , are coupled , however , and satisfy @xmath134 \nonumber \\
p_{nz } & = & -\frac{4\pi}{3}a^3\sum_{n^\prime\neq n}\left[t_{zy}{\cal g}_{yy}(z_n - z_{n^\prime})p_{n^\prime , y } + t_{zz}{\cal g}_{zz}(z_n - z_{n^\prime})p_{n^\prime , z}\right]\end{aligned}\ ] ] the tensor @xmath78 is still diagonal , with the same nonzero elements as in the case of @xmath2 parallel to the chain . assuming propagating waves of the form @xmath135 , @xmath136 , we find that the amplitudes @xmath137 and @xmath138 satisfy the equations @xmath139exp(ikn^\prime d ) \nonumber \\ p_{0z } & = & -\frac{4\pi}{3}a^3\sum_{n^\prime \neq 0}\left[-t_{yz}{\cal g}_{yy}(-z_{n^\prime})p_{0y}+t_{zz}{\cal g}_{zz}(-z_{n^\prime})p_{0z}\right]\exp(ikn^\prime d ) .
\label{eq : rotyz}\end{aligned}\ ] ] in the special case where only nearest neighbor interactions are included , these equations simplify to @xmath140\cos(kd ) \nonumber \\ p_{0z } & = & -\frac{8\pi}{3}a^3\left[-t_{yz}{\cal g}_{yy}(d)p_{0y } + t_{zz}{\cal g}_{zz}(d)p_{0z}\right]\cos(kd ) .
\label{eq : rotyz1}\end{aligned}\ ] ] substituting in the explicit forms of @xmath141 and @xmath142 , we find that these equations take the form @xmath143\cos kd \nonumber \\ p_{0z } & = & -\frac{a^3}{d^3}\left[\frac{2}{3}t_{yz}p_{0y } + \frac{4}{3}t_{zz}p_{0z}\right]\cos kd .
\label{eq : rotyz2}\end{aligned}\ ] ] if we solve the pair of equations ( [ eq : rotyz ] ) or ( [ eq : rotyz2 ] ) for @xmath137 and @xmath138 for a given value of @xmath87 , we obtain a nonzero solutions only if the determinant of the matrix of coefficients vanishes . for a given real frequency @xmath86 , there will , in general , be two solutions for @xmath77 which decay in the @xmath144 direction .
these correspond to two branches of propagating plasmon ( or plasmon polariton ) waves , with dispersion relations which we may write @xmath145 and @xmath146 .
the frequency dependence appears because both @xmath147 and @xmath148 depend on @xmath86 [ through @xmath11 and @xmath104 ] .
the corresponding solutions @xmath150 , @xmath151 are no longer linearly polarized but will instead be elliptically polarized . however , unlike the case where the magnetic field is parallel to the @xmath5 axis , the waves are not circularly polarized , and the two solutions are non - degenerate . because @xmath104 is usually small
, the ellipse has a high eccentricity , and the change in propagation characteristics due to the magnetic field will usually be small for this magnetic field direction .
as a first numerical example , we calculate the plasmon dispersion relations for a chain of spherical drude metal particles immersed in an nlc . we consider two cases : liquid crystal director parallel and perpendicular to the chain axis , which we take as the @xmath5 axis . for @xmath10 and @xmath9 ,
we take the values used in ref .
these are taken from experiments described in ref .
@xcite , which were carried out on the nlc known as e7 . for comparison
, we also show the corresponding dispersion relations for an isotropic host of dielectric constant which is arbitrarily taken as @xmath152 .
the results of these calculations are shown in figs . 1 and 2 in the absence of damping ( @xmath153 in the drude expression ) .
as can be seen , both the @xmath0 and @xmath1 dispersion relations are significantly altered when the host is a nematic liquid crystal rather than an isotropic dielectric ; in particular , the widths of the @xmath0 and @xmath1 bands are changed . when the director is perpendicular to the chain axis ,
the two @xmath1 branches are split when the host is an nlc , whereas they are degenerate for an isotropic host , or an nlc host with director parallel to the chain .
next , we turn to the effects of an applied magnetic field on these dispersion relations for the case @xmath154 .
the @xmath0 waves are unaffected by a magnetic field , but the @xmath1 waves are split into left- and right - circularly polarized waves . to illustrate the predictions of our simple expressions , we again take @xmath155 , and we assume a magnetic field such that the ratio @xmath156 . for a typical metallic plasma frequency of @xmath157
sec@xmath158 , this ratio would correspond to a magnetic induction @xmath159 t. we consider both the undamped case ( @xmath153 ) and the damped case ( @xmath160 ) . using these parameters , the dispersion relations for the two circular polarizations
are shown in fig . 3 both with and without single - particle damping .
the splitting between the two circularly polarized @xmath1 waves is not visible on the scale of the figure .
we also plot the corresponding rotation angles @xmath161 for a distance equal to one interparticle spacing in fig .
when there is no damping , @xmath127 diverges near the edge of the @xmath1 bands , but this divergence disappears for a finite @xmath16 ( e. g. @xmath160 , as shown in in fig . 4 ) . in this case , re@xmath162 never exceeds about @xmath163 rad per interparticle separation , corresponding to a rotation of less than 0.1@xmath3 over this distance .
im@xmath164 is also small , showing that a linear incident wave acquires little ellipticity over such distances . over a distance of thirty or so interparticle separations ,
a linearly polarized transverse wave would typically rotate by only 1 - 3@xmath3 .
since theory and experiment suggests that the wave _ intensity _ typically has an exponential decay length of no more than around 20 interparticle spacings @xcite , the likely faraday rotation of such a wave in practice will probably not exceed a degree or two , at most , even for a field as large as 2 t. thus , while the rotation found here is likely to be measurable , it may not be large , at least for this simple chain geometry with one particle per unit cell .
the present expressions also indicate that @xmath127 is very nearly linear in b , so a larger rotation could be attained by increasing b. as can be seen , @xmath127 depends strongly on frequency in the case of zero damping , but less so at finite damping . in the ( unrealistic ) no - damping case , at the very top and the very bottom of the plasmonic band , only one of the two circularly polarized waves can propagate down the chain . because this filtering occurs only over a very narrow frequency range ( of order @xmath165 ) , and
because this calculation assumes no damping , it would be quite difficult to detect a region where only one of the two circularly polarized waves can propagate . finally , we mention the case where @xmath166 for the same parameters as in the parallel case .
the effect of a finite b causes two non - degenerate dispersion relations ( one an @xmath0 and the other a @xmath1 wave ) for the perpendicular case to become mixed .
we have not computed the rotation angles for this perpendicular case , but we expect that they would be similar in magnitude to the values shown in fig . 4 .
the calculations and formalism presented in the previous section leave out several effects which may be at least quantitatively important .
first , in our numerical calculations , but not in the formalism , we have included only nearest neighbor dipolar coupling .
inclusion of further neighbors will quantitatively alter the dispersion relations in all cases considered , as well as the faraday rotation angle when there is an applied magnetic field , but these effects should not be very large , as is already suggested by the early calculations in ref .
@xcite for an isotropic host .
another possible effect will appear when @xmath167 is significantly greater than @xmath168 , namely , the emergence of quadrupolar and higher quasistatic bands@xcite .
these will mix with the dipolar band and alter its shape .
for the separations we consider , this multipolar effect should be small . also , even if @xmath169 , the plasmon dispersion relations will still be altered by an nlc host or by an applied magnetic field in the manner described here .
the present treatment also omits radiative damping . in the absence of a magnetic field ,
such damping is known to be small but non - zero in the long - wavelength limit , but it becomes more significant when the particle radius is a substantial fraction of a wavelength .
even at long wavelengths , radiative damping can be very important at certain characteristic values of the wave vector@xcite .
we have not , as yet , extended the present approach to include such radiative effects .
we expect that , just as for an isotropic host in the absence of a magnetic field , radiative effects will further damp the propagating plasmons in the geometries we consider , but will not qualitatively change the effects we have described . for the case
when the host is an nlc , the present work oversimplifies the treatment of the nlc host by assuming that the director field is _ uniform _ , i. e. , position - independent . in reality
, the director is almost certain to be modified close to the metal nanoparticle surface , i. e. , to become nonuniform , as has been pointed out by many authors@xcite .
the effects of such complications on the optical properties of a single metallic particle immersed in an nlc have been treated , for example , in ref .
@xcite , and similar approaches might be possible for the present problem also . to summarize
, we have shown that the dispersion relations for plasmonic waves propagating along a chain of closely spaced nanoparticles of drude metal are strongly affected by external perturbations .
first , if the host is a uniaxially anisotropic dielectric ( such as a nematic liquid crystal ) , the dispersion relations of both @xmath0 and @xmath1 modes are significantly modified , compared to those of an isotropic host , and if the director axis of the nlc is perpendicular to the chain , the two degenerate transverse branches are split . secondly ,
if the chain is subjected to an applied magnetic field parallel to the axis , the @xmath1 waves undergo a small but measurable faraday rotation , and also acquire a slight ellipticity .
a similar ellipticity develops if the magnetic field is perpendicular to the chain , but its effect will likely be more difficult to observe , because the magnetic field couples two non - degenerate branches of the dispersion relation .
all these effects show that the propagation of such plasmonic waves can be tuned , by either a liquid crystalline host or a magnetic field , so as to change the frequency band where wave propagation can occur , or the polarization of these waves
. this control may be valuable in developing devices using plasmonic waves in future optical circuit design .
this work was supported by the center for emerging materials at the ohio state university , an nsf mrsec ( grant no .
dmr0820414 ) .
see , e. g. , t. c. lubensky , d. pettey , n. currier , and h. stark , phys .
e * 57 * , 610 ( 1998 ) ; p. poulin and d. a. weitz , phys .
e * 57*,626 ( 1998 ) ; h. stark , phys . rep . * 351 * , 387 ( 2001 ) ; r. d. kamien and t. d. powers , liq .
* 23 * , 213 ( 1997 ) ; d. w. allender , g. p. crawford , and j. w. doane , phys.rev . lett . *
67 * , 1442 ( 1991 ) . for plasmon waves along a chain of metallic nanoparticles , in the presence of an nlc host .
we plot @xmath170 , where @xmath15 is the plasma frequency , as a function of @xmath171 , where @xmath45 is the distance between sphere centers .
green and blue ( x s and + s ) : @xmath0 and @xmath1 modes for a chain embedded in an nlc with director parallel to the chain .
the nlc is assumed to have principal dielectric tensor elements @xmath172 and @xmath173 parallel and perpendicular to the director , corresponding to the material known as e7 . in this and
all subsequent plots @xmath174 , where @xmath4 is the metallic sphere radius . also shown are the corresponding @xmath0 and @xmath1 dispersion relations ( black and red solid lines , respectively ) when the host is isotropic with dielectric constant @xmath175 . ]
modes ( asterisks , in green ) and @xmath1 modes ( + s and x s , shown in dark and light blue ) , divided by the plasma frequency @xmath15 , are plotted versus @xmath171 .
the nlc has the same dielectric tensor elements as in fig . 1 .
also shown are the corresponding @xmath0 ( solid black ) and @xmath1 ( solid red ) branches for an isotropic host with @xmath176 .
note that the @xmath1 branches which were degenerate in fig .
1 are split into two branches in this nlc geometry . ] | a chain of metallic particles , of sufficiently small diameter and spacing , allows linearly polarized plasmonic waves to propagate along the chain . in this paper
, we consider how these waves are altered by an anisotropic host ( such as a nematic liquid crystal ) or an applied magnetic field . in a liquid crystalline host , with principal axis ( director ) oriented either parallel or perpendicular to the chain , we find that the dispersion relations of both the longitudinal ( @xmath0 ) and transverse ( @xmath1 ) modes are significantly altered relative to those of an isotropic host . furthermore ,
when the director is perpendicular to the chain , the doubly degenerate @xmath1 branch is split by the anisotropy of the host material . with an applied magnetic field
@xmath2 parallel to the chain , the propagating transverse modes are circularly polarized , and the left and right circularly polarized branches have slightly different dispersion relations . as a result ,
if a linearly polarized transverse wave is launched along the chain , it undergoes faraday rotation . for parameters approximating that of a typical metal and for a field of 2 t ,
the faraday rotation is of order 1@xmath3 per ten interparticle spacings , even taking into account single - particle damping . if @xmath2 is perpendicular to the chain , one of the @xmath1 branches mixes with the @xmath0 branch to form two elliptically polarized branches .
our calculations include single - particle damping and can , in principle , be generalized to include radiation damping .
the present work suggests that the dispersion relations of plasmonic waves on chain of nanoparticles can be controlled by immersing the chain in a nematic liquid crystal and varying the director axis , or by applying a magnetic field . |
in ultracold atoms , the typical diluteness condition and short - range interactions can give rise to interesting universal relations among thermodynamic quantities of a many - body system .
these universal relations are independent of the microscopic details of two - body interactions , and have stimulated much research interest . in a unitary fermi gas close to an @xmath0-wave scattering , for instance , the central quantity of these universal relations , tan s contact , has been extensively studied both theoretically and experimentally .
furthermore , the recent radio - frequency ( r.f . ) spectroscopic measurement near a @xmath1-wave feshbach resonance in @xmath2k atoms opens up new possibilities of studying universal relations and thermodynamics in higher partial - wave scattering channels @xcite .
theoretically , it has also been shown recently that a whole family of contacts and universal relations exist in systems with higher partial - wave scattering @xcite .
motivated by this progress , we investigate universal relations and thermodynamics of a two - component degenerate fermi gas with coexisting @xmath0- and @xmath1-wave interactions .
such a system can be experimentally prepared and probed with @xmath2k atoms near the magnetic field @xmath3 g . as the @xmath1-wave feshbach resonance here is also close to an @xmath0-wave feshbach resonance , the @xmath0- and @xmath1-wave interactions can be comparable in strength .
it is then expected that the co - existence of both @xmath0- and @xmath1-wave scattering channels can lead to interesting many - body properties .
indeed , it has been shown very recently that , at zero temperature , an interesting hybridization of @xmath0- and @xmath1-wave superfluid can be stabilized , which exhibit nontrivial pairing correlations @xcite . in this work , we focus on thermodynamic properties of the normal state above the superfluid transition temperature .
we first derive the general universal relations of a two - component ultracold fermi gas with coexisting @xmath0- and @xmath1-wave interactions . due to the orthogonality of two - body wave functions of different scattering channels ,
the universal thermodynamic relations of the system , such as the adiabatic relations , pressure , and the energy functional , appear to be direct summations of contributions from @xmath0- and @xmath1-wave interactions .
these universal relations are governed by a set of contacts , which can be associated with scatterings within different partial - wave channels .
we then numerically evaluate the interaction energy as well as contacts in the high - temperature normal phase using second - order virial expansions .
interestingly , due to the interplay of @xmath0- and @xmath1-wave interactions on the many - body level , we find that the contacts , and hence the relevant thermodynamic properties behave differently from those with only @xmath0- or @xmath1-wave interaction potentials . the interplay of multiple partial - wave interactions also leaves experimentally detectable signatures in asymptotic behavior of momentum distribution and r.f .
spectrum in the high - momentum , high - frequency regime .
another interesting finding is that the interaction energy of the repulsive branch features abrupt changes across the @xmath1-wave resonances , which may serve as a unique signature for the coexistence of @xmath0- and @xmath1-wave interactions in the system .
our results can be readily checked experimentally for @xmath2k atoms near the @xmath3 g @xmath1-wave feshbach resonance , where the two scattering channels naturally coexist .
the paper is organized as follows : in sec . [ 2 ] , starting from the two - body density matrix , we derive the general universal relations such as interaction energy , adiabatic relations , pressure relations , and the energy functional . in sec .
[ 3 ] , we present the formalism of the quantum virial theorem , and express thermodynamic quantities such as the thermodynamic potential and interaction energy in the normal state using second virial coefficients . in sec .
[ 4 ] , we numerically evaluate interaction energy and contacts using virial expansion calculations , adopting typical experimental parameters of @xmath2k atoms near the @xmath3 g @xmath1-wave feshbach resonance .
we also calculate the asymptotic behavior of the momentum distribution and the r.f .
spectrum in the large - momentum and large - frequency limit , respectively .
finally , we summarize in sec . [ 5 ] .
in this section , we study the universal relations for a two - component , partially polarized fermi gas with both @xmath0- and @xmath1-wave interactions . following the standard treatment ,
we start from the two - body density matrix of the many - body system , decompose it into pair wave functions , and expand the pair wave functions in terms of the two - body wave functions . employing the asymptotic forms of the two - body wave functions in the regime @xmath4 , where @xmath5 is the range of interaction potentials , we derive universal thermodynamic relations based on the two - body density matrices .
importantly , these universal relations are governed by a set of contacts , which are associated , respectively , with @xmath0- and @xmath1-wave scattering channels . for the derivations in this section
, we consider the general case where @xmath0-wave interactions between the two spin components ( @xmath6 ) and @xmath1-wave interactions between the same spins coexist .
following the standard treatment , we consider the two - body density matrix for our many - body system @xmath7 , where @xmath8 are spin indices , and @xmath9 creates a fermion with spin @xmath10 at position @xmath11 .
the two - body density matrix can be decomposed as @xcite @xmath12 where @xmath13 , @xmath14 is the center - of - mass momentum of a pair , @xmath15 are the eigenvalues , and @xmath16 label the quantum numbers of the relative radial direction , the angular momentum , and the @xmath17-axis projection of angular momentum , respectively .
moreover , the associated pair wave functions @xmath18 form an orthonormal set . for the system under consideration
, the only nonvanishing pair wave functions are @xmath19 with @xmath20 .
the spin wave functions are defined as @xmath21 and the spatial wave functions @xmath22 here , @xmath23 , @xmath24 , @xmath25 is the quantization volume , and @xmath26 is the relevant spherical harmonic function .
correspondingly , we denote the relevant eigenvalues as @xmath27 and @xmath28 , for the @xmath0- and @xmath1-wave pair wave functions , respectively .
we may then write down the interaction energy of the many - body system in terms of the two - body density matrix , and expand the pair wave functions therein using two - body radial wave functions .
the resulting interaction energy reads @xmath29.\nonumber\\\end{aligned}\ ] ] here , we have defined a set of contacts @xmath30 and @xmath31 , which are related to the scattering length @xmath32 , the scattering volume @xmath33 , and effective range @xmath34 , respectively @xmath35 note that the integration over @xmath36 here includes the contributions of the bound states .
note that the orthogonality of the two - body wave functions in different scattering channels renders the interaction energy in eq . a direct summation of contributions from the three scattering channels : one with the @xmath0-wave interaction between @xmath37 and @xmath38 , and two with the @xmath1-wave interaction between two @xmath37s or two @xmath38s . as a result
, different scattering channels are decoupled in the definition of contacts as well .
as we will see , this effectively allows us to write down the universal relations based on an intuitive extension of the previous works on single - partial - wave interactions .
the importance of contacts is that they appear in various universal relations and provide a link between few - body parameters and many - body quantities .
this link is most apparent in the so - called adiabatic relations .
following the treatment in refs .
@xcite , we have @xmath39 where @xmath40 is the total energy , @xmath41 is the reduced planck constant , and @xmath42 is the atomic mass . again , in the universal relations , we see that contacts associated with different scattering channels are decoupled .
based on the adiabatic relations , it is straightforward to derive other related universal relations . as two illustrating examples , we show below the pressure and the energy functional . the pressure for a uniform gas is given by @xcite @xmath43.\end{aligned}\ ] ] we note that although eq .
( [ ph ] ) is only exact in the absence of external potential , it is approximately valid for each local part of a trapped fermi gas within the local density approximation . in an external harmonic trap @xmath44 , where @xmath45 is the frequency of the harmonic trapping potential , the average energy density is replaced by the local internal energy density , and
the contacts are replaced by the local contacts .
furthermore , the total energy in the presence of a trapping potential is given by @xcite @xmath46,\end{aligned}\ ] ] where @xmath47 is the energy of the trapping potential .
the relation above is essentially the virial theorem in refs .
to further study the thermodynamic properties of the system , we evaluate the thermodynamic potential using virial expansions . for a spin - polarized fermi gas ,
it is necessary to introduce two fugacities @xmath48 and @xmath49 to distinguish different spin configurations , where @xmath50 , @xmath51 is the chemical potential for spin @xmath10 , @xmath52 is the temperature , and @xmath53 is the boltzmann constant . generally , the thermodynamic potential can be written as @xcite @xmath54 where @xmath55 is the @xmath56th virial coefficient associated with a cluster containing @xmath57 spin-@xmath58 fermions and @xmath59 spin-@xmath60 fermions . here , @xmath61 and the thermal de broglie wavelength is @xmath62 . following the common practice , we define @xmath63 , where the superscript ( 1 ) " denotes an ideal noninteracting system with the same fugacity . for a spin - polarized fermi gas with coexisting @xmath0- and @xmath1-wave interactions
, we may rewrite the thermodynamic potential as ( up to the second order ) @xmath64,\end{aligned}\ ] ] where @xmath65 is the thermodynamic potential for a noninteracting ideal fermi gas with the same fugacities as @xmath66 . @xmath67 and @xmath68 are , respectively , the second virial coefficients for the @xmath0- and @xmath1-wave scattering channels . in an ideal spin - imbalanced fermi gas ,
the thermodynamic potential for each spin species takes the form @xmath69 where the standard fermi - dirac integral is @xmath70 here , @xmath71 is the gamma function .
the total thermodynamical potential of an ideal spin - imbalanced fermi gas is a direct sum of the thermodynamic potentials of each spin component @xmath72.\end{aligned}\ ] ] accordingly , we can rewrite the thermodynamic potential of a strongly interacting polarized fermi gas as ( up to the second order ) @xmath73.\end{aligned}\ ] ] therefore , the particle number densities of spin - up and spin - down atoms are given by @xmath74,\\ \label{nh2}n_{\downarrow}&&=-\frac{1}{v}\frac{\partial\omega}{\partial\mu_{\downarrow}}\nonumber\\ & & = \frac{1}{\lambda^{3}}\left[f_{3/2}(z_{\downarrow})+2 z_{\uparrow}z_{\downarrow}\delta b_{2,s}+4z_{\downarrow}^{2}\delta b^{(\downarrow\downarrow)}_{2,p}\right].\end{aligned}\ ] ] the total energy density can be written as @xmath75 where the logarithm of the grand canonical partition function @xmath76 is given by @xmath77.\end{aligned}\ ] ] we can further separate the total energy density into the kinetic and the interaction part : @xmath78 .
the kinetic energy density is @xcite @xmath79\nonumber\\ & & ~~+\frac{3 n_{\downarrow}k_{b}t}{2}\left[1 + 2^{-5/2}(n_{\downarrow}\lambda^{3})\right].\end{aligned}\ ] ] and the interaction energy density is given by @xmath80,\end{aligned}\ ] ] where @xmath81 and @xmath82 . for a spin - polarized fermi gas with a fixed polarization @xmath83 , and a fixed total number density @xmath56 ,
we can obtain two different particle number densities @xmath84 .
we then substitute the number densities into eq .
( [ eint ] ) to get the corresponding interaction energy density with the second virial coefficients given in sec .
[ 3.4 ] .
the contacts can also be expressed in term of the grand thermodynamic potential @xmath85 @xcite @xmath86_{t , v,\mu_{\uparrow},\mu_{\downarrow}},\\ \label{cpr2 } & & c^{(\sigma\sigma)}_{p , r_{m } } = -\frac{8\pi m}{\hbar^{2 } } \left[\frac{\partial \omega}{\partial ( r_{m}^{(\sigma\sigma)})^{-1}}\right]_{t , v,\mu_{\uparrow},\mu_{\downarrow}}.\end{aligned}\ ] ] with these expressions , it is straightforward to numerically evaluate contacts and the associated universal thermodynamic properties once the second virial coefficients are known . for a uniform system , the second virial coefficients can be expressed in terms of the phase shifts of the corresponding two - body scattering problem . for a spin-1/2 fermi gas
, it takes the form @xcite @xmath87 where the summation is over all the two - body bound states ( with the bounding energy @xmath88 ) and @xmath89 is the phase shift of the @xmath90th partial wave . in the case of @xmath0-wave scattering @xcite @xmath91 where @xmath92 is the effective range of interactions .
close to resonance , @xmath93 , the second virial coefficient for @xmath0-wave scattering can then be written as @xcite , @xmath94 where @xmath95 , @xmath96 is the fermi temperature , @xmath97 is the fermi wave vector , @xmath98 is the complementary error function , the heaviside step function @xmath99 in this paper is defined as @xmath100 for @xmath101 and @xmath102 for @xmath103 , the sign function @xmath104 satisfies @xmath105 with @xmath103 , @xmath106 with @xmath107 , and @xmath108 with @xmath101 , and @xmath109 is the @xmath0-wave two - body bound - state energy . in the case of @xmath1-wave scattering @xcite @xmath110 where @xmath111 and @xmath112 are , respectively , the scattering volume and the effective range of the @xmath1-wave interaction with a magnetic number @xmath113 . the second virial coefficient in this case can be analytically given by @xcite @xmath114,\end{aligned}\ ] ] where @xmath115 is the @xmath1-wave two - body bounding energy .
in this section , we present our numerical results for the thermodynamic quantities based on the second virial coefficient calculations . to make connections with experiments , we consider @xmath2k atoms close to the @xmath3 g @xmath1-wave feshbach resonance between the same hyperfine states @xmath116 . as
this @xmath1-wave resonance is close to a wide @xmath0-wave feshbach resonance at @xmath117 g between the hyperfine states @xmath116 and @xmath118 , both @xmath0- and @xmath1-wave interactions exist in the system @xcite .
here , both the @xmath0- and @xmath1-wave scattering parameters are functions of the magnetic field . for the @xmath0-wave scattering ,
the scattering length @xmath32 is given by @xcite @xmath119 where @xmath120 g , @xmath121 , @xmath122 is the bohr radius , and @xmath123 g . for the @xmath1-wave scatterings
, the scattering volume @xmath111 near the @xmath1-wave feshbach resonance can be conveniently calculated using @xcite @xmath124 here , there are two @xmath1-wave feshbach resonances @xmath125 g and @xmath126 g .
the interacting ranges for the @xmath1-wave interactions are @xcite @xmath127 in the following , we will define @xmath128 and @xmath129 . therefore , the only @xmath1-wave interactions in the system are between two @xmath37 states . for the numerical calculations , we take @xmath130 nm , where the fermi wave vector is @xmath131 with the total particle number density @xmath56 . with @xmath132 ,
the interaction energy density becomes @xmath133.\end{aligned}\ ] ] of @xmath2k atoms as a function of the magnetic field @xmath134 ( in units of g ) near @xmath0- and @xmath1-wave resonances for @xmath135 @xcite with fixed spin polarization @xmath136 .
the blue solid line is for the @xmath0-wave feshbach resonance at @xmath137 g , the black solid line denotes the @xmath1-wave feshbach resonance with @xmath138 at @xmath125 g , and the green solid line denotes the @xmath1-wave feshbach resonance with @xmath139 at @xmath126g.,width=302 ] we show in fig .
[ interactionenergy ] the interaction energy density as a function of the magnetic field @xmath134 ( in units of g ) for @xmath135 @xcite with a fixed spin polarization @xmath136 . for comparison
, we show the numerical results for a fermi gas at the same temperature with only @xmath0-wave interactions ( dashed black curve ) , with only @xmath1-wave interactions ( dashed - dotted blue curve ) , and with both @xmath0- and @xmath1-wave interactions ( solid red curve ) .
we also show the interaction energy densities of the repulsive branch ( thin curves ) , for which the bound - state contributions to the second virial coefficients are excluded in eqs .
( [ b2s ] ) and ( [ b2p ] ) . an interesting feature of fig .
[ interactionenergy ] is the behavior of the interaction energy of the repulsive branch across the @xmath1-wave feshbach resonances .
we consider anisotropic @xmath1-wave interactions , in which the @xmath1-wave resonance for angular momentum @xmath140 occurs @xmath141 g , and the @xmath1-wave resonance for angular momentum @xmath142 occurs at @xmath143 g . in the case when only @xmath1-wave interactions are present , the interaction energy of the repulsive branch features an abrupt jump when the magnetic field is swept across the @xmath144 g resonance .
physically , this is due to the coexistence of @xmath1-wave interactions with different magnetic angular momentum . for the case with coexisting @xmath0- and @xmath1-wave interactions , similar sudden changes of the repulsive - branch interaction energy show up at both @xmath1-wave resonances at @xmath145 g and at @xmath146 g , which originate from the coexistence of @xmath0- and @xmath1-wave interactions .
note that while the interaction energy in the repulsive branch approaches the pure @xmath0-wave value as the magnetic field decreases away from the @xmath1-wave resonance , throughout the @xmath1-wave resonance region , the interaction energy is quite different from that of the pure @xmath0-wave case .
this behavior would provide a strong experimental signal for the existence of both scattering channels in the system . in order to calculate the contacts ,
we first expand the thermodynamic potential eq .
( [ omegah ] ) with respect to a small fugacity @xmath147 : @xmath148,\end{aligned}\ ] ] where we have taken @xmath132 .
the particle number densities eqs .
( [ nh1 ] ) and ( [ nh2 ] ) can also be expanded with respect to a small fugacity @xmath147 : @xmath149,\\ \label{nh22}n_{\downarrow}&&\simeq \frac{1}{\lambda^{3}}\left(z_{\downarrow}-\frac{z_{\downarrow}^{2}}{2^{5/2}}+2 z_{\uparrow}z_{\downarrow}\delta b_{2,s}\right).\end{aligned}\ ] ] with eqs .
( [ cs2 ] ) , ( [ cpv2 ] ) , ( [ cpr2 ] ) , and eq . ( [ omegahcontact ] )
the analytical expressions for the contacts become @xmath150_{t , v,\mu_{\uparrow},\mu_{\downarrow}}\nonumber\\ & & = \frac{16\pi m}{\hbar^{2 } } k_{b}t\frac{v}{\lambda^{3}}\sum_{m}\left[\frac{\partial\delta b^{(\uparrow\uparrow)}_{2,p}}{\partial(\nu_{m}^{(\uparrow\uparrow)})^{-1}}\right]_{t}z_{\uparrow}^{2},\\ \label{cprup } c^{(\uparrow\uparrow)}_{p , r}&&=\sum_{m}c^{(\uparrow\uparrow)}_{p , r_{m } } = -\frac{8\pi m}{\hbar^{2 } } \sum_{m}\left[\frac{\partial \omega}{\partial ( r_{m}^{(\uparrow\uparrow)})^{-1}}\right]_{t , v,\mu_{\uparrow},\mu_{\downarrow}}\nonumber\\ & & = \frac{16\pi m}{\hbar^{2 } } k_{b}t\frac{v}{\lambda^{3}}\sum_{m}\left[\frac{\partial\delta b^{(\uparrow\uparrow)}_{2,p}}{\partial ( r_{m}^{(\uparrow\uparrow)})^{-1}}\right]_{t}z_{\uparrow}^{2}.\end{aligned}\ ] ] numerically , we calculate the two fugacities by solving the coupled number equations eqs .
( [ nh12 ] ) and ( [ nh22 ] ) , and then substitute the fugacities into eqs .
( [ csup ] ) , ( [ cpvup ] ) , and ( [ cprup ] ) to get the corresponding contacts . for convenience , in the following discussions , we only calculate the attractive branch of the contacts .
figure [ contactp ] shows the contacts as functions of spin polarization @xmath151 for a homogeneous fermi gas up to the second - order virial expansion with @xmath152 and @xmath153 g ( solid red line ) . here
, the fermi temperature is given by @xmath154 .
for comparison , we also show the numerical results for system with only @xmath0-wave interactions ( dashed black line ) , and with only @xmath1-wave interactions ( dashed - dotted blue line ) .
note that @xmath155 ( @xmath156 ) for systems with only @xmath1-wave ( @xmath0-wave ) interactions . in fig .
[ contactt ] , we show the temperature dependence of various contacts with a polarization @xmath157 , and a magnetic field @xmath153 g . evidently , at lower temperatures , @xmath158 and @xmath159 for the case with coexisting @xmath0- and @xmath1-wave interactions can be significantly different from those with purely @xmath0- or @xmath1-wave interactions . at high temperatures , as thermal effects dominate over interaction effects , all the contacts approach zero .
finally , in fig . [ contactb ] , we show the dependence of contacts on the magnetic field with a polarization @xmath160 at a given temperature @xmath152 . apparently , the interplay of @xmath0- and @xmath1-wave interactions is most significant near the @xmath1-wave resonances .
we also note an interesting sign change of @xmath161 as the magnetic field is swept between two @xmath1-wave resonances . from figs .
[ contactp],[contactt ] , and [ contactb ] , it appears that , in the presence of mixed @xmath0- and @xmath1-wave interactions , despite the apparent decoupling of different scattering channels in the universal relations , both the @xmath0- and the @xmath1-wave contacts are different from those with a single scattering channel .
this clearly demonstrates the interesting interplay between @xmath0- and @xmath1-wave interactions on the many - body level .
physically , as the spin - up atoms are involved simultaneously in both scattering channels , the inclusion of any one of the two scattering channels on top the other would lead to a spin - dependent shift of the chemical potentials for the many - body system .
this amounts to a change of the many - body environment , although the microscopic two - body physics in each channel remains decoupled and unchanged . in terms of numerical calculation , this interplay is reflected in the solutions of the coupled number equations eqs .
( [ nh12 ] ) and ( [ nh22 ] ) .
experimentally , contacts can be determined by measuring the asymptotic behavior of the momentum distribution at large momenta . in a system with coexisting @xmath0- and @xmath1-wave interactions
, this high - momentum tail would also behave differently . to see this
, we define the single - body density matrix in the asymptotic regime @xmath162 with the interaction potential range @xmath5 and @xmath163 : @xmath164 similarly to the approach in sec .
[ 2 ] , it is straightforward to derive @xmath165\nonumber\\ & & ~~ + \frac{1}{v}\int d^3 \vec{r}_3 \sum_{\vec{p}_{c},j}n^{(\uparrow\downarrow)}_{\vec{p}_{c},j,0,0}e^{i\vec{p}_{c}\cdot(\vec{r}_2-\vec{r}_1)/2}\int dk~a_{j , k}\int dk'~a_{j , k'}\frac{y_{0,0}^{*}(\theta,\varphi)}{|\vec{r}_1-\vec{r}_3|}\frac{y_{0,0}(\theta',\varphi')}{|\vec{r}_3-\vec{r}_2|},\\ \label{rho1down } \rho_1^{(\downarrow)}(\vec{r}_1,\vec{r}_2)&&= \frac{1}{v}\int d^3 \vec{r}_3 \sum_{\vec{p}_{c},j}n^{(\uparrow\downarrow)}_{\vec{p}_{c},j,0,0}e^{i\vec{p}_{c}\cdot(\vec{r}_2-\vec{r}_1)/2}\int dk~a_{j , k}\int dk'~a_{j , k'}\frac{y_{0,0}^{*}(\theta,\varphi)}{|\vec{r}_1-\vec{r}_3|}\frac{y_{0,0}(\theta',\varphi')}{|\vec{r}_3-\vec{r}_2|}.\end{aligned}\ ] ] note that we have already assumed that no @xmath1-wave interactions exist between spin-@xmath38 atoms .
fourier - transforming eqs .
( [ rho1up ] ) and ( [ rho1down ] ) , we obtain the following asymptotic behavior of @xmath166 for @xmath167 : @xmath168\nonumber\\ & & ~~ + \pi\frac{\sum_{\vec{p}_{c},j , m}n^{(\uparrow)}_{\vec{p}_{c},j,1,m}(\int dk'~b_{j , m , k'}^{(\uparrow\uparrow)})^{2}p^{2}_{c}}{vk^{4}}\nonumber\\ & & ~~ + \frac{c_{s}}{vk^{4}},\\ \label{nhdown } n_{\downarrow}(\vec{k } ) & & = \frac{c_{s}}{vk^{4}}.\end{aligned}\ ] ] here , we assume that the distribution of @xmath169 is isotropic , so we have @xmath170 and @xmath171 . equations .
( [ nhup ] ) and ( [ nhdown ] ) show that the momentum distributions are highly asymmetric in spin .
especially , the momentum distribution of the spin-@xmath58 component exhibits a quite nontrivial high - momentum tail , which is due to the coexistence of @xmath0- and @xmath1-wave interactions .
another experimental tool for the detection of contacts is the r.f .
spectroscopy , in which universal scalings in the high frequency tail exist and are governed by contacts @xcite .
coupling hamiltonian @xmath172 transfers fermions into a third spin state @xmath173 , where @xmath174 is the radio - frequency rabi frequency determined by the strength of the radio - frequency signal with frequency @xmath175 .
furthermore , the transfer rate can be evaluated as @xmath176,\end{aligned}\ ] ] where @xmath177 , @xmath178 label the initial and final states , @xmath179 denotes the initial state distribution , and @xmath180 , @xmath181 denote the initial and final energies . in the region @xmath182
( @xmath183 is the fermi energy ) , the asymptotic radio - frequency transition rates for the two spin species become @xmath184\nonumber\\ & & = \frac{m\omega_{rf}^{2}}{2\pi\hbar}\sum_{m=-1}^{1}\left[\frac{c^{(\uparrow\uparrow)}_{p,\nu_{m}}}{(m\omega/\hbar)^{1/2 } } + \frac{3c^{(\uparrow\uparrow)}_{p , r_{m}}}{2(m\omega/\hbar)^{3/2}}\right]\nonumber\\ & & ~~ + \frac{m\omega_{rf}^{2}}{4\pi\hbar}\frac{c_{s}}{(m\omega/\hbar)^{3/2}},\\ \label{rfdown } \gamma_{rf,\downarrow}(\omega)&&=\frac{m\omega_{rf}^{2}}{4\pi\hbar}\frac{c_{s}}{(m\omega/\hbar)^{3/2}}.\end{aligned}\ ] ] with spin-@xmath58 as a function of the reduced frequency @xmath185 at magnetic field @xmath153 g for high frequency at temperature @xmath152 with spin polarization @xmath186 .
the line styles are the same as in fig .
[ contactp].,width=302 ] again , eqs .
( [ rfup ] ) and ( [ rfdown ] ) show that the radio - frequency transition rates are highly asymmetric in spin . in fig .
[ rf ] , we show the reduced radio - frequency transition rate @xmath187 with spin-@xmath58 as a function of the reduced frequency @xmath185 at magnetic field @xmath153 g for the temperature @xmath152 with spin polarization @xmath186 . like the high - momentum tail of the momentum distribution , the high - frequency tail of the r.f .
spectroscopy demonstrates the effects of coexistence of @xmath0- and @xmath1-wave interactions .
in this work , we study the universal relations and normal - phase thermodynamic properties of a two - component fermi gas with coexisting @xmath0- and @xmath1-wave interactions . similarly to the purely @xmath0-wave or @xmath1-wave case , we start from the two - body density matrix and derive the adiabatic relations , through which a set of contacts can be defined . due to the orthogonality of the two - body wave functions in different scattering channels ,
contacts from different scattering channels are decoupled .
we may therefore identify @xmath0-wave contact and @xmath1-wave contacts based on the adiabatic relations .
we then numerically evaluate the interaction energy and contacts using second - order virial expansions . for typical experimental parameters corresponding to @xmath2k atoms
, we find that when the interplay of @xmath0- and @xmath1-wave interactions is properly taken into account , all contacts behave differently from those with a single scattering channel .
we characterize in detail the dependence of contacts on quantities such as spin polarization , temperature , and magnetic field .
an interesting finding is that the interaction energy of the repulsive branch features abrupt changes as the @xmath1-wave resonances are crossed .
our results are closely related to the ongoing efforts of realizing and probing strongly interacting quantum gases with high partial - wave interactions .
as the interplay of different partial - wave interactions can have important effects on the many - body properties , our work has interesting implications for experiments exploring the @xmath3 g @xmath1-wave resonance in @xmath2k atoms , where interactions in different partial - wave scattering channels naturally coexist .
we thank zhenhua yu , shizhong zhang , shao - liang zhang , su wang , and jian - song pan for helpful discussions . this work is supported by the national key r&d program ( grant no .
2016yfa0301700 ) , the national natural science foundation of china ( grants no . 11374177 , no . 11374283 , no .
11404106 , no .
11522545 , no . 11534014 , no .
11622436 ) , and the programs of the chinese academy of sciences .
f.q . acknowledges support from the project funded by the china postdoctoral science foundation ( grant no .
2016m602011 ) .
w.y . acknowledges support from the `` strategic priority research program ( b ) '' of the chinese academy of sciences , grant no . xdb01030200 . | we study the universal relations and normal - phase thermodynamics of a two - component ultracold fermi gas with coexisting @xmath0- and @xmath1-wave interactions . due to the orthogonality of two - body wave functions of different scattering channels ,
the universal thermodynamic relations of the system appear to be direct summations of contributions from each partial - wave scattering channels .
these universal relations are dictated by a set of contacts , which can be associated with either @xmath0- or @xmath1-wave interactions .
interestingly , due to the interplay of @xmath0- and @xmath1-wave interactions on the many - body level , the contacts , and hence all the relevant thermodynamic quantities , behave differently from those with only @xmath0- or @xmath1-wave interactions .
these are manifest in our numerical calculations based on second - order virial expansions for @xmath2k atoms under typical experimental parameters .
a particularly interesting finding is that , due to the coexistence of @xmath0- and @xmath1-wave scatterings , the interaction energy of the repulsive branch features abrupt changes across the @xmath1-wave resonances .
our results can be readily checked experimentally for @xmath2k atoms near the @xmath3 g @xmath1-wave feshbach resonance , where multiple partial - wave scatterings naturally coexist . |
results related to the problem of counting spanning trees for a graph date back to 1847 . in @xcite
, kirchhoff showed that the number of spanning trees of a graph @xmath13 is closely related to the cofactor of the laplacian matrix of @xmath13 .
later , a number of related results followed . in 1889
, cayley @xcite derived the number of spanning trees for the complete graph on @xmath4 vertices which is @xmath14 . later formulas for various families of graphs have been derived .
for example , it was shown by baron et al .
@xcite that the number of spanning trees of the square of a cycle @xmath15 equals to @xmath16 where @xmath17 is the @xmath4th fibonacci number .
speaking about a seemingly unrelated branch of mathematics , euler studied around 1778 a special class of numbers allowing him to find large primes .
he called such numbers _ idoneal numbers _ ( _ numerus idoneus _ ) .
he was able to find 65 such numbers :
, , 0 _ @xmath18 _ @xmath19 see also @xcite
. gauss @xcite conjectured that the set of idoneal numbers _
i _ is complete .
it was later proved by chowla @xcite that the set of idoneal numbers is finite .
we denote by _ @xmath20 _ the set of idoneal numbers not present in _ @xmath18 _ and remark that if the generalized riemann hypothesis is true , then _ @xmath20 _ = @xmath21 @xcite .
it is also known that any idoneal number in _
@xmath20 _ has at least six odd prime factors @xcite . in this paper
we use the definition of idoneal numbers stating that @xmath4 is _ idoneal _ if and only if @xmath4 is not expressible as @xmath22 for integers @xmath23 . for other characterizations of idoneal numbers see @xcite .
we use this number theoretical result to improve the answer related to the question sedlek @xcite posed in 1970 : given a number @xmath24 what is the least number @xmath1 such that there exists a graph on @xmath1 vertices having precisely @xmath4 spanning trees ?
sedlek denoted this function by @xmath0 .
he was able to show that @xmath25 for almost all numbers .
more precisely he proved that @xmath26 whenever @xmath27 and @xmath28 whenever @xmath29 .
nebesk @xcite later showed that the only fixed points of @xmath0 are @xmath30 and @xmath31 , i.e. these are the only numbers @xmath4 such that @xmath32 he also defined the function @xmath3 as the least number of edges @xmath33 for which there exists a graph with @xmath33 edges and with precisely @xmath4 spanning trees .
he showed that @xmath34 except for the fixed points of @xmath35 in which case it holds that @xmath36 .
moreover , as it is observed in @xcite , from the construction used by sedlek @xcite we have @xmath37 whenever @xmath38 in this paper we improve their result by showing that @xmath39 whenever @xmath40 and @xmath41 we also prove that @xmath42 , proof of which in @xcite we found to be incomplete as it only states that there is no graph with cyclomatic number 2 or 3 that has 22 spanning trees and that every graph with a greater cyclomatic number has more than 22 spanning trees . we will refer to the number of spanning trees of a graph @xmath13 by @xmath43 . throughout the paper we will often use the following identity used to compute @xmath43 : @xmath44 for every @xmath45 . here
@xmath46 denotes the graph obtained from @xmath13 by contracting the edge @xmath47 of @xmath13 and removing the loop that could possibly be created .
note that the resulting graph may not be simple .
if by @xmath48 we denote the graph that is obtained after introducing an edge into @xmath13 and by @xmath49 we denote the graph obtained after interconnecting two vertices of @xmath13 with a path @xmath50 of length @xmath1 , then we will occasionally use the fact that : @xmath51 for a connected graph @xmath13 and @xmath52 .
the first inequality follows from the fact that we can form at least two spanning trees in @xmath48 that are not spanning trees in @xmath13 by taking a spanning tree @xmath53 of @xmath13 and obtain new trees @xmath54 , @xmath55 after removing an edge ( not equal to @xmath47 ) from the cycle that is obtained in @xmath56 the second inequality is equally easy to prove .
graphs with @xmath0 vertices with @xmath4 spanning trees possess some structure .
for example , it follows directly from equation ( [ rec ] ) that graphs having @xmath4 spanning trees with @xmath0 vertices are always @xmath57-edge - connected . a simple argument
can then be used to show that such graphs have cycles of length at most @xmath58 provided that @xmath59 for nonnegative integers
@xmath60 , let @xmath61 be the graph comprised of two vertices connected by three internally disjoint paths of length @xmath62 and @xmath63 , respectively .
we refer to these paths as @xmath64 , @xmath65 and @xmath66 .
note that @xmath61 is simple if and only if at most one of @xmath60 equals 1 .
for @xmath67 denote by @xmath68 the graph obtained after identifying a vertex of an @xmath69-cycle with a vertex of a disjoint @xmath70-cycle .
notice that @xmath71 is isomorphic to @xmath72
in this section , we examine the number of spanning trees that arise in @xmath61 when interconnecting two distinct vertices by a disjoint path of length @xmath73 . in order to do
so we define simple graphs @xmath74 , @xmath75 , @xmath76 that are obtained from @xmath61 by introducing a path .
let @xmath77 be the 3-vertices of @xmath61 .
first we construct @xmath78 .
we assume @xmath79 . for integers
@xmath80 and @xmath81 with @xmath82 , let @xmath83 and @xmath84 be the vertices of @xmath64 such that @xmath85 and @xmath86 . then @xmath74 is the graph obtained by interconnecting @xmath83 and @xmath84 with a disjoint path of length @xmath73 , see the first graph of figure [ pic ] . as we are only dealing with simple graphs we require that @xmath87 if @xmath88 .
we now construct @xmath89 .
let @xmath90 and let @xmath84 be a vertex on @xmath64 such that @xmath91 .
then @xmath75 is the graph obtained by interconnecting @xmath83 and @xmath84 by a disjoint path of length @xmath73 .
see the second graph of figure [ pic ] .
notice that the possibility @xmath92 is not excluded ; in that case @xmath75 has two 4-vertices .
since we wish that the resulting graph be simple we require @xmath93 whenever @xmath94 and @xmath93 whenever @xmath95 .
moreover , we assume @xmath96 . finally , define @xmath76 by choosing a vertex @xmath83 on @xmath64 , @xmath97 such that @xmath85 and similarly let @xmath84 be a vertex on @xmath65 , @xmath98 such that @xmath99 .
we always assume @xmath80 and @xmath100 then @xmath76 is the graph obtained by connecting @xmath83 and @xmath84 by a disjoint path of length @xmath73 as shown on the third graph of figure [ pic ] .
observe that this construction requires @xmath101 , @xmath102 and @xmath103 in order to preserve simplicity of @xmath104 we now present some formulas and inequalities for the number of spanning trees for the graphs we have just defined .
the following equalities are easy to derive using equation ( [ rec ] ) . obtained after adding a path of length @xmath73 between two distinct vertices @xmath105 of @xmath61.,title="fig : " ] obtained after adding a path of length @xmath73 between two distinct vertices @xmath105 of @xmath61.,title="fig : " ] obtained after adding a path of length @xmath73 between two distinct vertices @xmath105 of @xmath61.,title="fig : " ] [ lm0 ] the following three equalities hold : 1 .
@xmath106 , 2 .
@xmath107 , 3 .
@xmath108 we use the identities presented in lemma [ lm0 ] in order to derive some lower bounds for the number of spanning trees for the graphs @xmath78 , @xmath89 and @xmath104 [ lm1 ] we have @xmath109 let @xmath110
. by lemma [ lm0].(1 ) we have to show @xmath111 where @xmath112 as @xmath113 is a quadratic concave function it is enough to verify the claim for @xmath114 if @xmath93 and @xmath115 if @xmath116 .
suppose first that @xmath117 , i.e. @xmath118 . if @xmath119 then @xmath120 and if @xmath121 , then @xmath122 suppose now that @xmath93 and @xmath123
. then , we have @xmath124 as @xmath125 finally , assume @xmath116 and hence @xmath126 then @xmath127 as @xmath128 [ lm2 ] @xmath129 by lemma [ lm0].(2 ) it is enough to show @xmath130 for each @xmath131 $ ] , where @xmath132 .
as @xmath113 is a quadratic concave function , it is enough to show this inequality only for @xmath133 and @xmath134 . for @xmath133
, this inequality reduces to @xmath135 which trivially holds for @xmath96 . for @xmath134 , we have to show the inequality @xmath136 .
notice that @xmath96 and at least one of @xmath70,@xmath63 is @xmath137 , say @xmath70 .
then @xmath138 and hence : @xmath139 [ lm3 ] we have @xmath140 by lemma [ lm0].(3 ) , we have to show that @xmath141 it will be enough if we show that : @xmath142 the first inequality follows immediately by the known fact @xmath143 for @xmath144 ( just set @xmath145 and @xmath146 ) . for the second inequality ,
it is enough to observe that @xmath147 and @xmath148 as @xmath149 whenever @xmath150 . using the derived bounds from lemmas [ lm1].[lm3 ] .
we can state the following corollary that turns out to be useful in the next section .
[ cor ] let @xmath13 be the graph obtained after interconnecting two vertices of @xmath61 with a @xmath73-path @xmath151 then @xmath152 let @xmath13 be the graph obtained after interconnecting two vertices @xmath105 of @xmath61 with a @xmath73-path .
if @xmath116 or @xmath153 then @xmath83 and @xmath84 are distinct and we see from lemmas [ lm1 ] , [ lm2 ] and [ lm3 ] that the stated inequality holds . if @xmath154 and @xmath83 , @xmath84 are distinct it follows from lemmas [ lm1].[lm3 ] . that the number of spanning trees of @xmath13 is at least @xmath155 times greater than @xmath156 .
however , if @xmath157 then we obtain a graph that has @xmath158 and @xmath61 as blocks and thus @xmath159 which implies our claim .
this section presents our main result .
we derive some bounds and equalities for @xmath35 and @xmath160 for some small values and afterwards we show that for @xmath40 @xmath161 except for a few cases .
[ fixedpoint ] @xmath162 it is enough to show @xmath163 let us assume @xmath13 is a graph on @xmath164 vertices with precisely @xmath31 spanning trees .
since @xmath31 is not expressible as @xmath165 for @xmath166 , @xmath167 and since it is not expressible as @xmath168 for @xmath67 it follows that there exist integers @xmath166 and @xmath167 such that @xmath169 .
the only @xmath170 s graphs having less than 22 spanning trees and vertices are @xmath171 , @xmath172 , @xmath173 , @xmath174 , @xmath175 , @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 and @xmath181 , so @xmath13 contains at least one of them .
moreover , we see from corollary [ cor ] that introducing a @xmath73-path @xmath182 to any of the graphs @xmath175 , @xmath176 , @xmath177 , @xmath178 , @xmath179 , @xmath180 or @xmath181 yields a graph with at least @xmath183 spanning trees .
thus we can conclude that at least one of @xmath171 , @xmath172 , @xmath173 or @xmath174 is a proper subgraph of @xmath13 .
assume @xmath184 since @xmath171 has four vertices and since any graph on four vertices has at most 16 spanning trees it follows that there exists a subgraph @xmath185 of @xmath13 that is obtained after interconnecting two vertices of @xmath171 with some @xmath1-path @xmath186 if @xmath187 then from corollary [ cor ] we have that @xmath185 is a graph with at least @xmath188 spanning trees , so @xmath189 observe that this implies that we can assume @xmath190 , @xmath191 or @xmath192 .
moreover , using corollary [ cor ] we see that by adding a @xmath73-path ( @xmath93 ) to any of the graphs @xmath172 , @xmath173 , @xmath174 one obtains a graph with at least @xmath193 spanning trees . combining these two facts
we conclude that @xmath194 for an edge @xmath47 and @xmath195 after consulting table [ tb2 ] we see that all such graphs @xmath196 have more than 22 spanning trees with the exception of @xmath197 and @xmath198 which have 20 and 21 spanning trees , respectively .
we now deduce from inequality ( [ inequ ] ) that @xmath199 for an edge @xmath200 since the addition of an edge or a path to @xmath198 produces a graph with at least 23 spanning trees while introducing a @xmath201 2-path to @xmath202 yields a graph having at least 40 spanning trees . using the recurrence ( [ rec ] )
we see that @xmath203 since @xmath204 regardless of the choice of @xmath205 this implies that there is no graph having 22 spanning trees and less than 22 vertices , thus proving the stated lemma .
.graphs constructed in the proof of proposition [ prop1 ] and theorem [ thm2 ] . [ cols="^,<,^,^",options="header " , ] [ prop1 ] the relations @xmath206 , @xmath207 , @xmath208 , @xmath209 , @xmath210 , @xmath211 hold .
we consider each claim individually : * @xmath212 it is easy to verify that there is no graph on less thna 6 edges with 9 spanning trees and that @xmath213 * @xmath214 we see from the first equality in lemma [ lm0 ] that @xmath215 has 30 spanning trees and 8 edges from where the stated inequality follows .
* @xmath216 according to the second identity stated in lemma [ lm0 ] , @xmath217 this implies @xmath209 since @xmath218 has 9 edges .
* @xmath219 from the second identity derived in lemma [ lm0 ] we see that @xmath220 is a graph with 10 edges having precisely 58 spanning trees , which implies @xmath219 * @xmath221 clearly , @xmath222 is a graph of order 9 with @xmath223 spanning trees .
assume now @xmath224 and let @xmath13 be a graph with 25 spanning trees having less than 9 vertices .
the only graph of the form @xmath68 which has 25 spanning trees is @xmath222 but @xmath225 moreover , there exist no integers @xmath166 , @xmath167 such that @xmath226 and @xmath227 .
thus , it follows that @xmath13 contains as a subgraph some @xmath61 with at most 8 vertices , i.e @xmath228 .
it can be verified that adding a @xmath73-path @xmath182 to any of such graphs produces a graph with at least 25 spanning trees with the exception of the graphs @xmath171 , @xmath172 , @xmath174 , @xmath175 , @xmath173 , @xmath177 .
moreover , we see from corollary [ cor ] that adding a @xmath57-path to the graph @xmath61 yields a graph with at least @xmath229 spanning trees .
we now split the proof into two cases : * * _ @xmath172 , @xmath174 , @xmath175,@xmath173 or @xmath177 is a subgraph of @xmath230 _ by the above observation , adding a @xmath57-path to any of the graphs covered in this case produces a graph with at least @xmath231 spanning trees .
thus @xmath232 for @xmath233 and an edge @xmath47 .
table [ tb2 ] lists all the graphs ( up to isomorphism ) and the respective number of spanning trees that can be constructed after adding an edge to @xmath234 the fact that adding an edge or a path to a connected graph increases its number of spanning trees by at least 2 reduces our choice for @xmath235 to the elements of the set @xmath236 moreover , we saw in the proof of proposition [ fixedpoint ] that @xmath237 for any edge @xmath205 as we can not add a @xmath201 2-path to @xmath197 without obtaining a graph with less than 40 spanning trees ( by the second inequality stated in ( [ inequ ] ) ) it follows that @xmath238 for an edge @xmath200 .
using the deletion - contraction recurrence stated in ( [ rec ] ) , we now see that @xmath239 since the graph @xmath240 contains @xmath171 as a subgraph independently of the choice of @xmath200 . *
* _ @xmath171 is a subgraph of @xmath230 _ observe that @xmath241 and @xmath242 .
moreover , we see from lemmas [ lm1].[lm3 ] . that adding a 3-path to @xmath171 produces a graph with at least @xmath243 spanning trees unless the path interconnects the same vertex in which case we obtain a graph @xmath244 with @xmath245 spanning trees .
observe that a similar argument holds when we add a longer path to @xmath171 .
since by virtue of inequality ( [ inequ ] ) we can not introduce an edge or a path to @xmath244 that would produce a graph with less than @xmath246 spanning trees we conclude that @xmath247 for @xmath248 . but
this implies that @xmath190 or @xmath191 and we can use the same reasoning as in case 1 to conclude that there is no graph @xmath13 satisifying the stated properties .
+ the proof of the stated inequality is now complete since the above two cases show that there exist no graph on less than @xmath249 vertices that has @xmath223 spanning trees .
* @xmath250 since @xmath251 , it follows that @xmath252 to prove that @xmath253 and thus @xmath250 one can now use an argument similar to the previous case @xmath221 [ prop2 ] if
0 @xmath254 then @xmath12 let @xmath4 be a number from the set defined in the statement of this lemma .
if @xmath4 is of the form @xmath255 for @xmath256 then @xmath257 since @xmath258 for every @xmath259 therefore , the inequality holds for every @xmath4 divisible by 5 since every such @xmath4 that is in the list satisfies the required condition .
the same reasoning can be applied to the cases when @xmath4 is a divisor of 6 or a divisor of 7 leaving us to verify the inequalities for the numbers @xmath260 it is easy to verify that for @xmath261 the graphs @xmath262 , @xmath263 , @xmath264 have @xmath265 , @xmath266 and @xmath267 spanning trees , respectively , and that the inequalities on the number of edges and vertices are satisifed .
if @xmath268 then we use the fact that @xmath4 has at least three odd prime factors .
let @xmath269 be two of these prime factors of @xmath4 and let @xmath270 .
we construct the graph @xmath271 by taking disjoint cycles @xmath272 , @xmath273 , @xmath274 and identifying a vertex of every cycle .
observe that the graph is well defined as @xmath154 .
the obtained graph clearly has @xmath4 spanning trees and @xmath275 edges . the last inequality following from the fact that @xmath276 is clearly greater than 30 . [ thm2 ]
let @xmath277 be an integer .
then , @xmath7 if and only if , ,
0 @xmath278 moreover , if @xmath279 and @xmath280 then @xmath12 if @xmath4 is not idoneal then @xmath4 is expressible as @xmath281 for some integers @xmath23 .
the reader may verify that the graph @xmath61 has @xmath282 edges and precisely @xmath4 spanning trees .
therefore for every non idoneal @xmath4 , @xmath283 observe also that at most one of @xmath69 , @xmath70 , @xmath63 is 1 , therefore @xmath282 is maximal when @xmath284 , @xmath285 and @xmath286 the latter also implies that @xmath287 otherwise , if @xmath288 then the maximum for @xmath282 is attained whenever @xmath289 , @xmath285 and @xmath290 summing up the resulting equalities we conclucde that @xmath291 if @xmath29 and @xmath292 otherwise .
we now consider the case when @xmath4 is an idoneal number from the set
, , 0 @xmath293 if @xmath4 is from the set,,0 @xmath294 then the inequality immediately follows from proposition [ prop2 ] ; moreover , if @xmath4 is from @xmath295 then it can be verified that the graphs @xmath296 , @xmath173 , @xmath297 , @xmath177 , @xmath298 , @xmath299 , @xmath300 , @xmath301 , @xmath302 , @xmath303 , @xmath304 have 8 , 12 , 15 , 16 , 21 , 24 , 28 , 33 , 57 , 93 and 177 spanning trees respectively while also satisfying the stated inequality for the number of edges .
we are now left with the idoneal numbers 9 , 18 , 25 , 30 , 37 and 58 .
we see from proposition [ prop1 ] that for @xmath305 the inequality holds and that for @xmath306 the inequality does not hold , which now proves our theorem .
generalizing the graph @xmath61 to the graph containing four disjoint paths of lengths @xmath307 that interconnect two vertices , one obtains a graph with @xmath308 spanning trees . using this generalization for a higher number of disjoint paths one could lower the constant factor in our inequality to an arbitrary amount .
the fact that there is no result known to the authors related to the solvability of the respective equations and the fact that there could exist a superior construction yielding a better lower bound , motivates the following question : given a real number @xmath309 is there an integer @xmath310 such that for every @xmath311 we have @xmath312
the authors are thankful for the numerous corrections suggested by the anonymous referees . | let @xmath0 be the least number @xmath1 for which there exists a simple graph with @xmath1 vertices having precisely @xmath2 spanning trees .
similarly , define @xmath3 as the least number @xmath1 for which there exists a simple graph with @xmath1 edges having precisely @xmath2 spanning trees . as an @xmath4-cycle has exactly @xmath4 spanning trees
, it follows that @xmath5 . in this paper , we show that @xmath6 and @xmath7 if and only if @xmath8 , which is a subset of euler s idoneal numbers . moreover , if @xmath9 and @xmath10 we show that @xmath11 and @xmath12 this improves some previously estabilished bounds .
* keywords : * number of spanning trees ; extremal graph + * ams subj . class .
( 2010 ) * : 05c35 , 05c05 |
riemann s xi - function is defined by @xmath2 where @xmath3 is the euler @xmath4-function and @xmath5 is the riemann @xmath6-function .
the @xmath7-function is an entire function of order 1 with functional equation @xmath8 and its only zeros are the complex zeros of @xmath9 .
thus , if the riemann hypothesis ( rh ) is true , all the zeros of @xmath0 have real part one half , and the same would hold for all the zeros of the derivative @xmath1 .
we assume rh throughout this paper and investigate the distribution of the zeros @xmath10 of @xmath11 .
the distribution of zeros of @xmath12 is of interest for number - theoretic reasons connected to the problem of landau - siegel zeros , and also in connection to the general behavior of zeros of entire functions under differentiation .
we discuss those motivations in the next section .
we calculate @xmath13 where the sum is over pairs of ordinates of zeros of @xmath11 and @xmath14 is a weight function .
the normalizing factor @xmath15 in front of the sum is the number of zeros of @xmath12 with ordinates in @xmath16 $ ] , which ( on rh ) differs from the number of zeros of @xmath7 by at most 1 .
[ thm : f1 ] let @xmath17 be an arbitrary large positive integer . assuming rh we have @xmath18 as @xmath19 , for @xmath20 .
this theorem is an analogue of hugh montgomery s result @xcite on the pair correlation of zeros of the @xmath6-function .
he considered the function @xmath21 where the sum is over pairs of ordinates of zeros of @xmath0 and @xmath22 .
( we use @xmath23 for zeros of @xmath7 because @xmath24 refers to zeros of @xmath12 in this paper ) . in the terminology of random matrix theory ( rmt ) , @xmath25
is called the `` 2-point form factor '' , although sometimes it is mistakenly referred to as the pair correlation function .
in fact , @xmath25 is the fourier transform of the pair correlation function .
montgomery proved that @xmath25 has main term @xmath26 for @xmath27 .
that is , @xmath25 behaves like a dirac @xmath28-function at @xmath29 and is asymptotically @xmath30 when @xmath31 .
we see the same @xmath28-like behavior in @xmath32 ; this is not surprising since the spike at @xmath33 is simply a consequence of the density of zeros and the weight function .
the behavior of @xmath34 and @xmath35 for @xmath36 , however , is quite different , as illustrated in figure [ fig : f1 ] . [
0.75 ] for @xmath37 and @xmath38 . for comparison ,
the dotted line in both plots is the form factor @xmath39 for the zeros of the @xmath7-function .
, title="fig : " ] 0.25 in [ 0.75 ] for @xmath37 and @xmath38 . for comparison ,
the dotted line in both plots is the form factor @xmath39 for the zeros of the @xmath7-function .
, title="fig : " ] montgomery used his result on @xmath39 to obtain various estimates on the spacing and multiplicity of zeros of the @xmath7-function . applying the same methods to theorem [ thm : f1 ] we obtain [ cor : xipgaps ] assume rh .
a positive proportion of gaps between zeros of @xmath12 are less than 0.91 times the average spacing , and more than 3.5% of the normalized neighbor gaps between zeros of @xmath12 are smaller than average .
[ cor : xipsimple ] assume rh .
more than 85.84% of the zeros of @xmath12 are simple .
it is not surprising that the first corollary is weaker than the corresponding result for the @xmath7-function and the second is stronger .
the reasons are discussed in the next section .
conrey @xcite has shown unconditionally that at least 79.874% of the zeros of @xmath12 are simple and on the critical line .
thus , corollary [ cor : xipsimple ] gives a conditional improvement of his result .
inserting this estimate into formula ( 6 ) of farmer @xcite improves ( again on rh ) the unconditional estimate there of @xmath40 for the number of distinct zeros of the @xmath7-function .
[ cor : xidistinct ] assume rh .
the number of distinct zeros of the zeta - function in @xmath16 $ ] is larger than @xmath41 . in the next section
we discuss the motivations for our work arising from the distribution of zeros of entire functions and the problem of landau - siegel zeros . in section [ sec : explicit ] we state an explicit formula relating the zeros of @xmath12 to prime numbers , and in section [ sec : startofproof ] we begin the proof of theorem [ thm : f1 ] and identify the main terms . in section [ sec : endofproof ] we complete the proof , except for an arithmetic proposition which is proven in section [ sec : arithmetic ] . in section [ sec : explicitformula ] we prove the explicit formula used in section [ sec : explicit ] .
one important investigation we have not carried out here is the heuristic determination of @xmath32 when @xmath42 .
montgomery s study of the statistical behavior of zeros of the riemann zeta - function was motivated by the problem of landau - siegel zeros .
the idea is that a real zero very close to @xmath43 of @xmath44 would have a profound effect on the zeros of the riemann zeta - function : in a certain range all the zeros would be on the critical line and would have a peculiar spacing .
set @xmath45 and denote consecutive zeros of the zeta - function by @xmath46 , so that @xmath47 is 1 on average .
the existence of a siegel zero implies that in a certain range almost all the zeros of the zeta - function satisfy @xmath48 .
thus , one could disprove the existence of landau - siegel zeros by showing that @xmath49 , say , sufficiently often .
montgomery s result , however , only allows one to conclude that @xmath50 a positive proportion of the time .
montgomery refers to the connection to landau - siegel zeros in his paper @xcite , and similar connections are mentioned in unpublished work of heath - brown . at present
the only published account is due to conrey and iwaniec @xcite .
they show that the existence of landau - siegel zeros implies that , in a certain range , most of the spacings between consecutive zeros of the zeta function are close to multiples of half the average spacing .
that is , @xmath51 is close to @xmath52 or @xmath43 , or @xmath53 , etc .
although conrey and iwaniec give explicit estimates only in the case of small spacings between zeros , their main result can be used to show , for example , that if @xmath51 was often close to @xmath54 , then there are no landau - siegel zeros . the estimates in such cases , however , have not been worked out yet .
if the statistics of the zeros of the zeta - function are governed by random matrix theory ( rmt ) , then one could immediately conclude there are no landau - siegel zeros because the neighbor spacing is supported on all of @xmath55 .
since there are no immediate prospects of proving that the zeros of the zeta - function follow random matrix statistics ( or disproving landau - siegel zeros by another method ) , it is interesting to probe the boundary of what distributions are possible for zeros of the zeta - function .
the following has been proposed as a test case : * the alternative hypothesis ( ah ) * _ there exists a real number @xmath56 such that if @xmath57 , then @xmath58 that is , almost all the normalized neighbor spacings are an integer or half - integer .
_ one can also formulate weaker versions , where the normalized spacings are approximately integers or half - integers .
ah is obviously absurd , but it has not been disproven .
a sufficiently strong disproof would show that there are no landau - siegel zeros .
ah implies that montgomery s function @xmath25 is periodic with period two .
thus , on ah the graph of @xmath25 would look like this : [ 1.25 ] on the alternative hypothesis ( the heavy vertical lines represent the dirac @xmath28-functions at the even integers ) .
, title="fig : " ] similarly , the triple correlation function form factor @xmath59 ( see @xcite ) would be periodic with period two in both @xmath60 and @xmath61 .
unfortunately , this does not seem to completely determine @xmath59 because @xmath59 is only known on the hexagon @xmath27 , @xmath62 , @xmath63 . if we let @xmath64 denote the proportion of @xmath65 such that @xmath66 , a straightforward calculation assuming ah leads to @xmath67 , @xmath68 , and @xmath69 .
thus , one could disprove ah by showing that more than 30% of the normalized neighbor gaps of @xmath7 are less than @xmath70 .
we note that rmt predicts that 53.39% of the neighbor gaps are smaller than average .
montgomery s result implies that more than 12.3% of the neighbor gaps are smaller than average ( set @xmath71 at the bottom of page 192 of @xcite ) .
corollary [ cor : xipgaps ] gives information about consecutive small gaps between zeros of @xmath7 .
this is discussed further in the next subsection .
one wonders whether ah and the existing results on zero correlations determine the distribution of neighbor spacings . to specify
@xmath72 seems to require using the triple correlation @xmath59 to determine how often two consecutive normalized neighbor gaps of size @xmath52 can occur .
so this question may be equivalent to the question of whether ah determines all the correlation functions .
goldston and montgomery @xcite showed that the pair correlation conjecture is equivalent to @xmath73 for @xmath74 in a certain range depending on @xmath75 . here
@xmath76 , where @xmath77 is the von mangoldt function : @xmath78
if @xmath79 with @xmath80 prime , and @xmath29 otherwise . montgomery and soundararajan @xcite interpret as saying @xmath81 has mean @xmath74 and variance @xmath82 , and they note that the cramr model of the primes predicts a larger variance of @xmath83 .
it would be interesting to see what the right - hand side of equals if one assumes the alternative hypothesis .
these connections indicate the value of studying the statistics of the zeros of the zeta - function .
we now explain the connection to the zeros of @xmath12 .
the statistics of the zeros of @xmath12 are interesting because of their connection with the alternative hypothesis and also as an illustration of the general behavior of the zeros of derivatives of an entire function upon repeated differentiation .
one motivation for studying the analogue of montgomery s function for the zeros of @xmath12 is the expectation that our theorem [ thm : f1 ] might contradict ah .
the zeros of @xmath12 are influenced by the zeros of @xmath7 in complicated ways , so it seems unlikely that rmt and ah would predict the same behavior for @xmath84 for @xmath27 .
but there are several caveats .
first , as described in the previous section , it is not known whether or not ah determines all the correlation functions of the zeros .
this may lead to some flexibility in @xmath84 for @xmath27 , which may be consistent with theorem [ thm : f1 ] .
second , it is not known how to transfer a measure on the zeros of @xmath7 to a measure on the zeros of @xmath12 .
thus , even if ah determined all the correlations of the zeros of @xmath7 , it is still an unsolved problem to determine the correlations of the zeros of @xmath12 .
third , merely contradicting ah is not sufficient to disprove the existence of landau - siegel zeros .
ah is an extreme example of a possible consequence of landau - siegel zeros .
presumably an extension of the work of conrey and iwaniec @xcite would show that landau - siegel zeros imply that @xmath25 approximately follows figure [ fig : ah ] for some range of @xmath60 .
another motivation is to understand the general behavior of zeros under differentiation .
the riemann @xmath85-function is defined as @xmath86 .
the @xmath85-function is an entire function of order 1 that is real on the real axis . for such functions ,
repeated differentiation causes the zeros to migrate to the real axis @xcite .
thus , in any bounded region the riemann hypothesis is true for the @xmath87th derivative @xmath88 for sufficiently large @xmath87 .
it is conjectured @xcite that for real entire functions of order 1 , whose zeros lie in a strip around the real axis , not only do the zeros migrate toward the real axis , but they also approach equal spacing
. that is , the derivatives approach a multiple of @xmath89 .
this conjecture has been proven with some restrictions on the distribution of zeros @xcite and for some special cases , such as the @xmath85-function @xcite and the reciprocal of the gamma function @xcite .
the reason differentiation leads to equally spaced zeros is that , locally , the zeros of @xmath90 move away from concentrations of zeros of @xmath91 and towards regions with fewer zeros of @xmath91 .
thus , small gaps become larger and large gaps become smaller .
figure [ fig : xiplot ] illustrates these ideas . [ 1.3 ] , and the dots are the zeros of @xmath90.,title="fig : " ] a consequence is that if @xmath12 has a small gap between consecutive zeros , then @xmath7 must have had an even smaller gap .
thus , one should expect that the existence of landau - siegel zeros could be disproven by showing that @xmath12 has sufficiently many zeros separated by less than half the average spacing .
that is , one should be able to extend the result of conrey - iwaniec @xcite to the zeros of @xmath92 for any @xmath93 .
however , we have not worked out the necessary details to show that the required number of small gaps between zeros of @xmath12 leads to the same number of small gaps between the zeros of @xmath7 .
thus , at present we just mention this as a motivating principle .
see @xcite for a discussion .
based on these ideas , we see that corollary [ cor : xipgaps ] does not contradict the alternative hypothesis . on average
each normalized gap of size @xmath52 is adjacent to a gap of size @xmath52 or @xmath43 , so ah implies that at least 29.7% of the normalized neighbor gaps of @xmath12 are smaller than 1 . if corollary [ cor : xipgaps ] could be improved to show that @xmath12 has normalized zero gaps smaller than @xmath94 , then that would imply , on ah , that @xmath7 has consecutive gaps of size @xmath52 .
our final motivation is to understand the manner in which differentiation causes the zeros to become equally spaced .
let @xmath95 denote the analogue of @xmath35 involving the zeros of @xmath96 rather than the zeros of @xmath97 .
since the zeros of @xmath96 approach equal spacing as @xmath87 increases , @xmath95 approaches a sum of dirac @xmath28-functions supported at the integers .
we would like to understand this transition to a sum of @xmath28-functions .
the only case we know of where this has been worked out explicitly is for random trigonometric polynomials @xcite .
theorem [ thm : f1 ] is a first step in this direction , and figure [ fig : f1 ] does seem to illustrate the expected behavior . in a recent ph.d .
thesis , jim bian has now worked out explicit formulas for @xmath98 for @xmath99 and @xmath27 .
it would be interesting to know the random matrix analogue of theorem [ thm : f1 ] . at present , this is one of the few calculations carried out for the riemann zeta - function for which a random matrix analogue seems out of reach .
one issue is that in the random matrix world there seems to be no direct analogue of the riemann @xmath7-function .
see section 1.2 of @xcite for a discussion .
the closest match is @xmath100 where @xmath101 is the characteristic polynomial of a random matrix from the unitary group @xmath102 , chosen uniformly with respect to haar measure .
however , this is more nearly an analogue of the hardy @xmath103-function defined by @xmath104 .
this was the motivation of conrey and soundararajan @xcite , who did similar calculations to ours for the @xmath103-function .
but it turns out that the zeros of @xmath105 and @xmath106 should have similar correlation functions .
since @xmath107 is approximately @xmath108 , the corresponding zeros of @xmath105 and @xmath106 generally differ by @xmath109 .
so one should expect the correlation functions of their zeros to be equal , to leading order .
we provide a more rigorous explanation in section [ sec : z ] . and those correlation functions should equal , to leading order , the correlation function of the zeros of the derivative of the completed characteristic polynomial .
the general outline of our calculation is similar to the case of zeros of the zeta - function .
we start with an explicit formula relating a sum over zeros of @xmath12 to a sum involving number - theoretic expressions , and then compute the mean - square of both sides .
our main complication is that the number - theoretic side of the explicit formula is not a dirichlet series , but an `` approximate '' dirichlet series , by which we mean a dirichlet series whose coefficients may depend on @xmath110 .
the explicit formula is derived in section [ sec : explicitformula ] .
here we just outline the calculation and state the formula .
we begin by noting that @xmath111 has simple poles with residue 1 at the zeros of @xmath11 and no others ; here zeros are counted as many times as their multiplicity .
our first goal is to write @xmath111 as an `` approximate '' dirichlet series . from the definition of the @xmath7-function
we have @xmath112 where @xmath113 and @xmath114 in the region @xmath115 , @xmath116 , say .
multiplying both sides of by @xmath117 and calculating the logarithmic derivative leads to the following lemma , the proof of which is in section [ sec : explicitformula ] .
[ lem : logderiv ] for @xmath118 , @xmath119 , and @xmath17 a large positive integer we have @xmath120 here we have written @xmath121 where @xmath122 the function @xmath123 for @xmath124 is the @xmath125-fold convolution of the von mangoldt function , defined by @xmath126 for @xmath127 .
simple estimates of @xmath123 and @xmath128 will be used repeatedly .
we have the trivial bound @xmath129 which follows from iterating @xmath130 for some divisor @xmath131 of @xmath87 .
similarly , @xmath132 from which it follows that @xmath133 when @xmath134 , say .
a contour integral of @xmath135 times an appropriate kernel leads to the following explicit formula ; the details are in section [ sec : explicitformula ] .
[ prop : explicitformula ] for @xmath136 , @xmath137 , @xmath138 , and @xmath17 a positive integer @xmath139 where @xmath140 .
set @xmath141 in proposition [ prop : explicitformula ] , write the resulting equation as @xmath142 , and calculate @xmath143 the left - hand side may be treated in exactly the same way as the corresponding expression in montgomery @xcite , to whom we refer the reader ( cf .
. 187188 ) .
we find @xmath144 where @xmath145 .
next we begin the calculation of the right - hand side of . in various ranges of @xmath146 , one or another of the integrals @xmath147 dominates the others , and we record the following useful formula for later . given @xmath146 ,
let @xmath148 be the largest of @xmath149 , @xmath150 the next largest , and so on .
then we have @xmath151 we find @xmath152 @xmath153 and @xmath154 to estimate @xmath155 , we need the following [ integral lemma ] let @xmath156 or @xmath157 .
then for @xmath158 we have @xmath159 where @xmath160 .
the same result holds for @xmath161 we prove only , as the proof of is almost identical .
first observe that from we have @xmath162 the case @xmath163 follows immediately .
now suppose that @xmath164 , but @xmath165 . integrating by parts
, we find that our integral equals @xmath166 by , , and , this is @xmath167 this completes the proof .
we now come to the term @xmath168 say . recalling
that @xmath169 we see by lemma [ integral lemma ] that @xmath170 from the standard inequality @xmath171 and we see that the @xmath172-term is @xmath173 thus , we have @xmath174 similarly , we find that @xmath175 for @xmath176 we have combining these estimates , we have @xmath177 in the next section we complete the proof of theorem [ thm : f1 ] , subject to an arithmetic proposition which we prove in section [ sec : arithmetic ] .
it remains to evaluate and to put the expressions in the form of theorem [ thm : f1 ] .
the main terms from can be obtained from a stieltjes integral involving @xmath178 here @xmath128 is defined in and @xmath179 .
we write @xmath180 where @xmath181 the following proposition , which is proven in section [ sec : arithmetic ] , is sufficient to evaluate the leading order asymptotics of .
[ prop : akell ] we have @xmath182 and @xmath183 if @xmath184 , then @xmath185 and if @xmath186 , then @xmath187 as a consequence , we have @xmath188 for @xmath189 with @xmath190 , where @xmath191 is any fixed positive number .
we first evaluate @xmath155 given by .
we have @xmath192 note that the boundary terms canceled when we integrated by parts .
the typical term in the series for @xmath193 has the form @xmath194 and , assuming that @xmath195 , we have @xmath196 and @xmath197 using the first formula and , we have @xmath198 for @xmath199 , the @xmath200 term contributes no more than @xmath201 to this .
hence , @xmath202 similarly , using the second formula and , we obtain @xmath203 combining terms we obtain @xmath204 provided @xmath205 . recall from and
that we write @xmath206 where , for a given @xmath146 , @xmath207 is the largest of @xmath208 and @xmath209 is the next largest .
now , from the various estimates we see that our @xmath207 term is given by @xmath210 and by @xmath211 in each of these ranges , it happens that the @xmath212 term is @xmath213 .
hence , taking @xmath189 in , we find that for @xmath214 and @xmath215 large , @xmath216 using and noting that @xmath217 is an even function of @xmath60 , we have proved theorem [ thm : f1 ] .
we prove proposition [ prop : akell ] , which is the arithmetic portion of the calculation .
we first reduce @xmath218 to a sum involving the arithmetic functions @xmath123 . in section [ sec : lemmas ] we state some lemmas which are needed in the calculation , and in section [ sec : lambda_j ] we evaluate the sums of the @xmath123
. then we complete the proof of proposition [ prop : akell ] in section [ sec : proof of akell ] recall that @xmath219 where @xmath128 is given by .
[ lem : akell ] we have @xmath220 and @xmath183 for @xmath221 we have @xmath222 the form of @xmath223 is different when @xmath224 and @xmath43 from what it is for larger @xmath225 , and this will be reflected in our estimates for @xmath226 .
we therefore treat these cases separately .
first consider the case of @xmath227 . by the prime number theorem we have @xmath228 and @xmath229 for @xmath230 we have @xmath231
we have used the fact that @xmath232 and that this function vanishes unless @xmath87 is a product of at least @xmath225 ( not necessarily distinct ) primes and @xmath233 .
next we consider @xmath234 for @xmath235 . by the prime number theorem , @xmath236 if @xmath237 then @xmath238 exactly as in .
we now come to the general case of @xmath239 with @xmath240 .
we have @xmath241 say . the estimate on the third line above was done as follows . by symmetry , it is sufficient to estimate @xmath242 by the trivial bound this is @xmath243 as claimed .
one can use lemma [ lambdaj pn ] on the inner summand of to improve this bound by a power of @xmath244 , but this will not affect our final result .
it is clear that @xmath245 equals the second main term in , so it remains to put @xmath246 in the appropriate form .
the point is that if @xmath80 is prime and @xmath247 then @xmath248 see the proof of lemma [ lambdaj pn ] .
thus @xmath249 the second term can be estimated using the trivial bound , exactly as in , showing that it is @xmath250 .
for the first term , removing the conditions @xmath251 and @xmath252 and estimating with the trivial bound gives an even smaller error term .
this completes the proof of lemma [ lem : akell ] .
the following lemmas concerning the @xmath123 function and sums over primes are required in the next subsection .
the first lemma is an improvement on the trivial bound for @xmath253 when @xmath87 has a known prime factor .
[ lambdaj pn ] if @xmath254 , then @xmath255 if @xmath256 then @xmath257 setting @xmath258 , we obtain .
now @xmath259 hence @xmath260 setting @xmath261 , this is @xmath262 [ lem : logplogxp ] if @xmath263 and @xmath264 then @xmath265 let @xmath266 , so @xmath267 , where @xmath268 and @xmath269 .
we have @xmath270 say . in @xmath271
replace @xmath272 by @xmath273 and use euler s beta - integral to obtain @xmath274 which is the main term above . in the error term
we integrate by parts and find that @xmath275 the first term vanishes . in the integral make the change of variable @xmath276 and again use the beta - integral to obtain @xmath277 as claimed .
we evaluate the sums over @xmath123 which appear in lemma [ lem : akell ] let @xmath279 in this subsection we prove the following theorem .
[ thm : sksum ] if @xmath280 , then @xmath281 if @xmath235 , then @xmath282 the theorem is proved by induction , using the following proposition .
[ prop : skl ] for @xmath283 we have @xmath284 we assume @xmath285 and begin by unfolding @xmath286 in the sum : @xmath287 say .
we split @xmath288 into two sums @xmath289 and @xmath290 according to whether @xmath291 in the inner sum is or is not coprime to @xmath80 . by , if @xmath292 then @xmath293 .
hence @xmath294 by lemma [ lambdaj pn ] , removing the coprimality condition here introduces a change of @xmath295 that is , @xmath296 for @xmath290 we find that @xmath297 by lemma [ lambdaj pn ] this is @xmath298 hence , combining @xmath289 and @xmath290 , we obtain @xmath299 which is the main term in the proposition . by the trivial bound and lemma [ lambdaj pn ] we have
@xmath300 combining this with completes the proof of proposition [ prop : skl ] .
we first prove the bound . if @xmath237 then using the basic properties of @xmath77 and @xmath286 we have @xmath301 which is much smaller than the claimed bound .
now suppose holds for some @xmath302 and all @xmath303 . by proposition [ prop : skl ] ,
the induction hypothesis , and lemma [ lem : logplogxp ] : @xmath304 as required .
this proves .
now we prove .
when @xmath305 we have @xmath306 so holds in this case .
suppose holds for some @xmath307 .
then by proposition [ prop : skl ] , the induction hypothesis , and lemma [ lem : logplogxp ] : @xmath308 as required .
we are now ready to estimate @xmath239 .
suppose first that @xmath309 .
using theorem [ thm : sksum ] and lemma [ lem : akell ] we have @xmath310 by lemma [ lem : logplogxp ] the first term is @xmath311 and the second term is also @xmath312 .
when @xmath313 and @xmath314 , we obtain the same bound .
the only difference is that this time the first term in is omitted because @xmath315 when @xmath316 .
this proves .
next suppose that @xmath317 .
by lemma [ lem : akell ] and then theorem [ thm : sksum ] we have @xmath318 applying lemma [ lem : logplogxp ] we obtain @xmath319 as claimed
. it remains to do the case @xmath320 .
the only change in the above analysis is in the first term of . since @xmath321 if @xmath322 and 0 otherwise , using lemma [ lem : akell ] and then theorem [ thm : sksum ] and lemma [ lem : logplogxp ] we have @xmath323 note that this is the same as the general case with @xmath324 .
this completes the proof . recalling equation , @xmath325 and using the fact that @xmath326 if @xmath327 with @xmath328 , gives the final statement in proposition [ prop : akell ] .
we prove lemma [ lem : logderiv ] and proposition [ prop : explicitformula ] . since @xmath3 has simple poles at @xmath329 @xmath330 has simple poles with residue @xmath331 at @xmath332 and a simple pole with residue 1 at @xmath333 .
it is not difficult to show that @xmath334 has only simple real zeros at @xmath335 , with @xmath336 as @xmath337 . multiplying both sides of by @xmath117 and calculating the logarithmic derivative ,
we obtain @xmath338 suppose now that @xmath339 .
then there exists an absolute constant @xmath340 such that for @xmath118 , @xmath341 and @xmath342 hence , by , there exists an absolute constant @xmath343 such that @xmath344 for @xmath118 and @xmath345 .
now let @xmath17 be an arbitrary large integer . then by , , , and , @xmath346 for @xmath347 and @xmath348 . using the definition of the @xmath125-fold von mangoldt function ,
this can be rewritten as @xmath349 this completes the proof of lemma [ lem : logderiv ] .
suppose @xmath351 and let @xmath352 .
we have @xmath353 by choosing @xmath354 appropriately , one can obtain either the riemann @xmath7-function or the hardy @xmath103-function .
in either case , @xmath355 and @xmath356 which is all that was used in the calculation of the form factor @xmath32 .
the lower order terms in @xmath357 are different in those two cases , and this should have an effect on the lower order terms of @xmath84 . presumably the lower order terms also have an arithmetic component , so both should differ from the analogous expression from random matrix theory .
we will integrate both sides of against the following kernel : @xmath358 it is easy to see that @xmath359 and @xmath360 moreover , as a function of @xmath361 , @xmath362 has simple poles at @xmath363 and @xmath364 with residues 1 and @xmath331 , respectively . from the partial fraction decomposition for @xmath365 and the fact that @xmath366 has @xmath367 zeros with ordinates in the interval @xmath368 $ ]
, it follows that one can find an increasing , unbounded sequence @xmath369 such that @xmath370 and @xmath371 uniformly for @xmath372 . using these in and
, we find that @xmath373 uniformly for @xmath372 .
we now write @xmath374 where @xmath375 , and @xmath376 is the positively oriented rectangle with vertices at @xmath377 , where @xmath378 , @xmath379 , and @xmath380 is a large positive number .
the integrand has simple poles at @xmath381 , @xmath382 , and @xmath383 .
now , since @xmath379 , @xmath384 . therefore @xmath385 lies outside @xmath376 .
thus , by the calculus of residues , @xmath386 we now estimate the contributions of the horizontal and left edges of @xmath376 to @xmath387 . besides we need the estimate @xmath388 for @xmath389 . by , , , and this holds in @xmath390 .
hence it holds in @xmath391 by the functional equation . by and ,
the top and bottom edges of @xmath376 contribute @xmath392 to @xmath387 .
the left edge contributes @xmath393 letting @xmath380 and @xmath394 both tend to infinity , we obtain @xmath395 we evaluate the integral here by replacing @xmath396 by . since this representation holds only when @xmath397 , we have @xmath398 where @xmath399 now @xmath400 furthermore , by @xmath401
clearly we also have @xmath402 therefore , @xmath403 we split the integral on the right - hand side into two parts , namely , @xmath404 in @xmath405 we pull the contour left to @xmath406 and in doing so we pass a pole of @xmath362 at @xmath383 , the poles of @xmath407 at @xmath408 , and the poles of @xmath409 at the points @xmath410 , where the @xmath411 s are the zeros of @xmath334 .
we find that @xmath412 where @xmath413 using this and , we find that the sum over @xmath411 is @xmath414 the sum over @xmath291 is @xmath415 hence @xmath416 we treat @xmath417 by moving the contour away to @xmath418 .
this time we pass the pole from @xmath419 at @xmath420 and the poles of @xmath421 at @xmath422 and @xmath423 .
thus we find that @xmath424 we treat the second of these two terms as we did the corresponding term in @xmath405 and find that it is @xmath425 hence , @xmath426 combining , , , , and , we see that @xmath427 by the functional equation and , @xmath428 for @xmath429 . by and this equals @xmath430 .
hence , writing @xmath431 , and using the definition , we finally obtain the explicit formula in proposition [ prop : explicitformula ] . | the complex zeros of the riemannn zeta - function are identical to the zeros of the riemann xi - function , @xmath0 .
thus , if the riemann hypothesis is true for the zeta - function , it is true for @xmath0 . since @xmath0 is entire , the zeros of @xmath1 , its derivative ,
would then also satisfy a riemann hypothesis .
we investigate the pair correlation function of the zeros of @xmath1 under the assumption that the riemann hypothesis is true .
we then deduce consequences about the size of gaps between these zeros and the proportion of these zeros that are simple . |
the object of this paper is to carry on studying the local potential approximation of the exact renormalization group ( rg ) equation for the scalar theory @xcite . in a previous publication @xcite ( to be considered as the part i of the present work )
, we had already considered this approximation with a view to qualitatively discuss the connection between the standard perturbative renormalization of field theory ( as it can be found in most textbooks on field theory , see for example @xcite ) and the modern view @xcite in which the renormalized parameters of a field theory are introduced as the `` relevant '' directions of a fixed point ( fp ) of a rg transform .
actually the local potential approximation , which allows us to consider all the powers of the field @xmath7 on the same footing , is an excellent textbook example of the way infinitely many degrees of freedom are accounted for in ( nonperturbative ) rg theory .
almost all the characteristics of the rg theory are involved in this approximation .
the only lacking features are related to phenomena highly correlated to the non local parts neglected in the approximation and when the critical exponent @xmath8 is small ( especially for @xmath9 and @xmath10 ) , one expects the approximation to be qualitatively correct on all aspects of the rg theory @xcite . in the following we look at the domains of attraction or of repulsion of fixed points in the @xmath0 scalar theory in three and four dimensions ( @xmath10 and @xmath9 ) . at first sight
, one could think that the issue considered is very simple since , with regard to criticality , the @xmath0-symmetric systems in three dimensions are known to belong to the same class of universality ( the ising class ) .
now , because the ising class is associated to the domain of attraction of the unique ( non - trivial ) wilson - fisher fixed point @xcite , then by adjusting one parameter ( in order to reach the critical temperature ) any @xmath0 scalar hamiltonian should be driven to the wilson - fisher fixed point under the action of renormalization .
consequently there would have only two domains for the @xmath0 scalar theory : the critical subspace @xmath1 ( of codimension 1 ) in the wilson space ( @xmath11 ) of infinite dimensions of the hamiltonian parameters ( in which the rg transforms generate flows ) and the complement to @xmath11 of @xmath1 ( corresponding to noncritical hamiltonians ) .
in fact , that is not correct because there is another fixed point in @xmath11 : the gaussian fixed point which , although trivial , controls tricritical behaviors in three dimensions .
now each fp has its own basin of attraction in @xmath11 @xcite .
the attraction domain of the gaussian fp is the tricritical subspace @xmath4 of codimension 2 ( with no intersection with @xmath1 ) .
in addition , we show that there is a second subspace of codimension 1 in @xmath11 , called @xmath2 , which is different from @xmath1 , and thus which is not a domain of attraction to the wilson - fisher fixed point .
there is no fp to which a point of @xmath2 is attracted to .
@xmath2 is characterized by a negative sign of the @xmath12-hamiltonian parameter @xmath13 and is associated with systems undergoing a first - order phase transition @xcite .
we show that an endless attractive rg trajectory is associated to this domain of first - order transitions .
it is a renormalized trajectory ( denoted below by t@xmath14 ) that emanates from the gaussian fixed point .
the frontier between @xmath15 and @xmath1 corresponds to the tricritical subspace @xmath4 which is the domain of attraction of the gaussian fixed point while @xmath16 and @xmath1 are two distinct domains of repulsion for the gaussian fixed point .
actually , the situation is conform to the usual view .
considering the famous @xmath5-model [ landau - ginzburg - wilson ( lgw ) hamiltonian ] in which the associated coupling @xmath13 is positive , then the hamiltonian at criticality is attracted exclusively to the wilson - fisher fixed point , but if @xmath13 is negative , a @xmath17-term is required for stability , but then one may get either a tricritical phase transition or a second- or a first - order transition @xcite . in the present study we do not truncate the hamiltonian which involves all the powers of the field @xmath7 .
we explicitly show that a system which would correspond to an initial point lying very close to the frontier @xmath4 in the critical side ( in @xmath18 ) would display a retarded classical - to - ising crossover @xcite .
this result is interesting with regard to ionic systems ( for example ) in which a classical behavior has been observed while an ising like critical behavior was expected .
the eventuality of a retarded crossover from the classical to the ising behavior has previously been mentioned but without theoretical explanation on how this kind of crossover could develop @xcite . in @xcite
a calculation suggests that the rpm model for ionic systems would specifically correspond to a scalar hamiltonian with a negative sign for the @xmath5-hamiltonian parameter ( but the order parameter chosen is not the bulk density @xcite ) .
that calculation has motivated the present study .
we also indicate that the renormalized trajectory t@xmath19 still exists in four dimensions .
this makes the gaussian fixed point ultraviolet stable and the scalar field theory _ formally _ asymptotically free .
however the associated `` perfect '' action @xcite would have the wrong sign to provide us with an acceptable ( well defined ) field theory .
the organisation of the paper is as follows . in section [ lpa - erge ]
we briefly present the local potential approximation of the exact rg equation to be studied .
we introduce the strategy we have chosen to solve the resulting nonlinear differential equation with a view to show the trajectories of interest in the space @xmath11 of infinite dimension .
because the practical approach to the gaussian fixed point is made difficult due to the logarithmic slowness characteristic of a marginally irrelevant direction ( for @xmath10 ) , we found it useful to first test our numerical method with a close approach to the wilson - fisher fixed point .
we present the characteristic results of this approach and the various kinds of domains corresponding to @xmath20 ( somewhat a summary of @xcite ) in section [ uneg ] we describe the various kinds of attraction or repulsion domains of the gaussian fixed point ( for a negative value of the @xmath12-hamiltonian parameter ) corresponding to tricritical , critical and first - order subspaces .
then we discuss the consequences and especially explicitly show how a retarded crossover from the classical to the ising behavior can be obtained .
we then shortly discuss the case @xmath9 when @xmath21 . in two appendices we report on some technical aspects of the numerical treatment of the rg equation
studied , in particular on the appearance of spurious nontrivial tricritical fixed points ( appendix [ spurious ] ) .
the local potential approximation has been first considered by nicoll et al @xcite from the sharp cutoff version of the exact rg equation of wegner and houghton @xcite , it has been rederived by tokar @xcite by using approximate functional integrations and rediscovered by hasenfratz and hasenfratz @xcite . as in @xcite
we adopt the notation of the latter authors and consider the following nonlinear differential equation for the simple function @xmath22 : @xmath23 in which a prime refers to a derivative with respect to the constant dimensionless field @xmath7 ( at constant @xmath24 ) and @xmath25 is the derivative of the dimensionless potential @xmath26 ; @xmath27 stands for @xmath28 in which @xmath24 is the scale parameter defined by @xmath29e@xmath30 and corresponding to the reduction to @xmath31 of an arbitrary initial momentum scale of reference @xmath32 ( the initial sharp momentum cutoff ) . finally , @xmath33 is the surface of the @xmath34-dimensional unit sphere divided by @xmath35 .
a fixed point is solution of the equation @xmath36 .
the study of the resulting second order differential equation provides the following results : * @xmath37 , no fp is found except the gaussian fixed point . *
@xmath38 , one nontrivial fp ( the wilson - fisher fixed point @xcite ) is found @xcite * a new nontrivial fp emanates from the origin ( the gaussian fixed point ) below each dimensional threshold @xmath39 , @xmath40 @xcite .
if one represents the function @xmath22 as a sum of monomials of the form : @xmath41 then , for @xmath10 , the wilson - fisher fixed point @xmath42 is located in @xmath11 at @xcite : @xmath43 , @xmath44 , @xmath45 , @xmath46 , etc .
once the fp is known , one may study its vicinity which is characterized by orthogonal directions corresponding to the infinite set of eigenvectors , solutions of the differential equation ( [ eq4 ] ) linearized at @xmath42 .
the eigenvectors associated to positive eigenvalues are said relevant ; when the eigenvalues are negative they are said irrelevant and marginal otherwise @xcite .
the relevant eigenvectors correspond to directions along which the rg trajectories go away from the fp and the irrelevant eigenvectors correspond to directions along which the trajectories go into the fp .
a marginal eigenvector may be relevant or irrelevant .
our present fp @xmath42 has only one relevant direction and infinitely many irrelevant directions ( no marginal direction , however see @xcite ) . as already explained and shown in @xcite , in order to approach @xmath42 starting from an initial point in @xmath11
, one must adjust one parameter of the initial function @xmath47 .
this amounts to fixing the temperature of a system to its critical temperature . starting with a known initial function ( at `` time '' @xmath48 )
say : @xmath49 we adjust @xmath50 to the critical value @xmath51 corresponding to @xmath52 so that @xmath22 [ solution at time @xmath24 of the differential equation ( [ eq4 ] ) ] approaches @xmath42 when @xmath53 . the approach to @xmath54
is characterized by the least negative eigenvalue @xmath55 ( @xmath56 was noted @xmath57 in @xcite ) .
this means that , in the vicinity of @xmath42 any parameter @xmath58 evolves as follows ( @xmath53 ) : @xmath59 fig .
[ fig1 ] illustrates this feature for the first four @xmath58 s in the approach to @xmath42 . in @xcite the two associated attractive trajectories ( locally tangent to the least irrelevant eigenvector in the vicinity of @xmath42 )
was noted t@xmath60 and t@xmath61 .
one may also constrain the trajectory to approach @xmath42 along the second irrelevant direction ( with the associated attractive trajectories noted t@xmath62 or t@xmath63 in @xcite and associated with the second least negative eigenvalue @xmath64 ) . in this case a second parameter of the initial @xmath65 must be adjusted , e.g. , @xmath66 must be adjusted to @xmath67 _ and _ _ simultaneously _ @xmath68 to the corresponding @xmath69 , see @xcite .
then , in the vicinity of @xmath42 , any parameter @xmath70 will evolve as follows : @xmath71 looking for this kind of approach to @xmath42 , we have found that @xmath72 and @xmath73 this has allowed us to estimate @xmath74 .
although the shooting method is certainly not well adapted to the determination of the eigenvalues ( see the huge number of digits required in the determination of @xmath75 and @xmath67 ) , our estimate is close to @xmath76 found by comellas and travesset @xcite .
because @xmath67 can not be perfectly determined , the trajectory leaves the trajectory t@xmath77 before reaching @xmath42 to take one of the two directions t@xmath60 or t@xmath78 ( corresponding to @xmath56 ) .
fig [ fig2 ] illustrates this effect with the evolution , for @xmath79 , of the following effective eigenvalue : @xmath80 the definition of which does not refer explicitly to @xmath42 .
the evolution of @xmath81 shows a flat extremum ( or a flat inflexion point ) at an rg eigenvalue of @xmath42 each time the rg flow runs along an eigendirection in the vicinity of @xmath42 .
similarly to @xmath67 , the value @xmath69 can not be perfectly determined , consequently the trajectory ends up going away from the fixed point .
this provides us with the opportunity of determining the only positive ( the relevant ) eigenvalue corresponding to the critical exponent @xmath82 [ @xmath81 shows then a flat extremum at @xmath83 when the flow still runs in the close vicinity of @xmath42 ] .
finally , far away from the fixed point , the rg trajectory approaches the trivial high temperature fixed point characterized by a classical eigenvalue ( equal to @xmath85 ) .
the global picture summarizing the evolution of @xmath81 along a rg trajectory initialized in such a way as to approach @xmath42 first along t@xmath62 , is drawn on fig .
[ fig3 ] .
the values we have determined by this shooting method are ( for eigenvalues other than @xmath86 already mentioned ) : @xmath87 which are close to the values found , for example , in @xcite : @xmath88 and @xmath89 .
in the preceding section , we have obtained a rg trajectory approaching the wilson - fisher fixed point @xmath42 along t@xmath77 by adjusting two parameters of the initial hamiltonian ( @xmath67 and @xmath69 ) .
this is exactly the procedure one must follow to determine a tricritical rg trajectory approaching the gaussian fixed point in three dimensions ( because of its two relevant directions ) .
the only difficulty is to discover initial points in @xmath90 which are attracted to the gaussian fixed point . to this end
, we again use the shooting method . from usual arguments on the lgw hamiltonian and from the work done by aharony on compressible ferromagnets @xcite ,
one expects to find the tricritical surface in the sector @xmath21 ( and with @xmath91 ) .
thus we have tried to approach the gaussian fixed point starting with initial function @xmath92 of the form : @xmath93 with ( not large ) negative values of @xmath94 , for example @xmath95 . because the gaussian fixed point is twice unstable , we must adjust two parameters to approach it starting with ( [ eq - init - tri1 ] ) .
we do that by successive tries ( shooting method ) . for example , if we choose @xmath96 and @xmath97 and determine a value of @xmath98 such as to get a trajectory which does not go immediately towards the trivial high temperature fixed point , the best we obtain is a trajectory which approaches the wilson - fisher fixed point ( thus the corresponding initial point belongs to the attraction domain of @xmath42 although @xmath99 @xcite ) .
but if @xmath100 , the adjustment of @xmath50 with a view to counterbalance the effect of the most relevant direction of the gaussian fixed point ( which would drive the trajectory toward the high temperature fp ) yields a runaway rg flow towards larger and larger negative values of @xmath101 .
from now on , the target is bracketted : the tricritical trajectory corresponding to @xmath102 can be obtained with a value of @xmath103 in the range @xmath104 2,3\right [ $ ] ( we actually find a rather close approach to the gaussian fixed point for @xmath105 and @xmath106 ) . in order to understand the origin of the direction of runaway in the sector of negative values of @xmath13 , it is worth to study the properties of the gaussian fixed point by linearization of the rg flow equation in the vicinity of the origin . if we request the effective potential to be bounded by polynomials then the linearization of eq .
( [ eq4 ] ) identifies with the differential equation of hermite s polynomials of degree @xmath107 for the set of discrete values of @xmath108 satisfying @xcite : @xmath109 from which it follows that * for @xmath9 : @xmath110 @xmath111 , there are two non - negative eigenvalues : @xmath112 and @xmath113 * for @xmath10 : @xmath114 @xmath111 , there are three non - negative eigenvalues : @xmath112 , @xmath115 et @xmath116 if we denote by @xmath117 the eigenfunctions associated to the eigenvalue @xmath118 , it comes : * @xmath119 , @xmath120 , @xmath121 , @xmath122 , whatever the spatial dimensionality @xmath34 .
the upperscript `` @xmath123 '' is just a reminder of the fact that the eigenfunctions are defined up to a global factor and thus the functions @xmath124 are also eigenfunctions with the same eigenvalue @xmath118 . similarly to @xmath125 ,
the direction provided by @xmath126 in @xmath90 is a direction of instability of the gaussian fixed point .
now @xmath127 is associated with the well known renormalized trajectory t@xmath60 on which is defined the usual ( massless ) @xmath128-field theory @xcite , for the same reasons a renormalized trajectory t@xmath129 locally tangent to @xmath126 in the vicinity of the origin of @xmath11 exists with the same properties as t@xmath60 ( see @xcite ) .
the difference is that t@xmath19 lies entirely in the sector @xmath21 and is endless ( not ended by a fixed point ) .
this endless renormalized trajectory is associated with systems undergoing a first - order phase transition .
this is due to the absence of fixed point @xcite , in which case the correlation length @xmath130 can not be made infinite although for some systems lying close to t@xmath19 and attracted to it ( i.e. at the transition temperature ) , @xmath130 may be very large ( because t@xmath19 is endless ) , in which cases one may say that the transition is almost of second order @xcite . of course , a domain of first - order phase transition in @xmath11 was expected from the usual arguments @xcite , we only specify better the conditions of realization , in @xmath11 , of the first - order transition .
[ fig4 ] shows the attractive trajectory
t@xmath19 together with the attractive tricritical line approaching the gaussian fixed point . the tricritical surface @xmath4 separates the first - order surface @xmath2 from the critical surface @xmath1 . fig .
[ fig4 ] shows also that systems lying close to the tricritical surface may still be attracted to the wilson - fisher fixed point . in this case
the effective exponents may undergo a very retarded crossover to the asymptotic ising values compared to usual systems corresponding to initial points chosen in the sector @xmath20 of @xmath11 .
[ fig5 ] illustrates how minus the inverse of ( [ eq - omegaeff ] ) provides us with different evolutions [ calculated from ( [ eq4 ] ) ] of the effective exponent @xmath131 [ with @xmath132 $ ] according to the initial point chosen in @xmath11 .
it is worth to explain how we have defined @xmath133 .
we have seen at the end of section [ lpa - erge ] that the quantity ( [ eq - omegaeff ] ) undergoes a flat extremum ( or a flat inflexion point ) at an rg eigenvalue of @xmath42 each time the rg flow runs along an eigendirection in the vicinity of @xmath42 .
now it happens that this extremum is less and less flat as one chooses larger and larger values of @xmath134 ( for the eigenvalue @xmath135 ) but still exists .
this provides us with a way to express the evolution of an effective exponent @xmath136 when the rg - substitute to @xmath137 , namely @xmath138 , is varied .
[ fig6 ] shows such an evolution for some initial hamiltonian ( with @xmath139 ) .
notice that for such a hamiltonian , the extremum disappears before @xmath136 reaches the trivial value @xmath85 ( associated with the approach to the trivial high temperature fixed point and to a regular non critical behavior ) while in the case of a hamiltonian initialized close to the tricritical surface , the classical - to - ising crossover is complete ( see fig .
[ fig4 ] ) .
this is because in the latter case the rg trajectory comes close to the gaussian fixed point ( and @xmath140 has an extremum at @xmath85 ) before approaching to @xmath42 .
this reinforces the idea that the so - called classical - to - ising crossover actually exists only between the gaussian and wilson - fisher fixed points @xcite . the same configuration displayed by fig .
[ fig4 ] has been obtained also by tetradis and litim @xcite while studying analytical solutions of an exact rg equation in the local potential approximation for the @xmath141-symmetric scalar theory in the large @xmath142 limit .
but they were not able to determine `` _ the region in parameter space which results in first - order transitions _
'' @xcite . to decide whether the marginal operator ( associated with the eigenvalue equal to zero , i.e. @xmath143 in four dimensions , or @xmath144 in three dimensions ) is relevant or irrelevant
, one must go beyond the linear approximation .
the analysis is presented in @xcite for @xmath9 . if one considers a rg flow along @xmath125 such that @xmath145 , then one obtains , for small @xmath146 : @xmath147 $ ] with @xmath148 .
hence the marginal parameter decreases as @xmath24 grows .
as it is well known , in four dimensions the marginal parameter is irrelevant .
however , if one considers the direction opposite to @xmath125 ( i.e. @xmath126 ) then the evolution corresponds to changing @xmath149 .
this gives , for small values of @xmath146 : @xmath150 $ ] and the parameter becomes relevant .
the parameter @xmath146 is the renormalized @xmath151 coupling constant @xmath152 and it is known that in four dimensions the gaussian fixed point is ir stable for @xmath153 but ir unstable for @xmath154 @xcite .
we have verified that the trajectory t@xmath19 survives when @xmath9 ( contrary to t@xmath60 , see @xcite ) . that trajectory t@xmath129 is a renormalized trajectory on which we could define a continuum limit for the @xmath6-field theory and if the corresponding ( perfect ) action was positive for all @xmath7
, one could say that the @xmath155-field theory with a negative coupling is asymptotically free . unfortunately , because the @xmath5-term is dominant for large @xmath7 in the vicinity of the origin of @xmath11 ( due to the relevant direction provided by @xmath126 ) , the negative sign of the renormalized coupling prevents the path integral to be well defined .
however , because the action to which one refers in the continuum limit ( the perfect action ) is formal ( because it involves an infinite number of parameters and can not be written down , see @xcite ) we wonder whether the wrong sign of the action is actually a valid argument to reject the @xmath156-field theory with a negative renormalized coupling .
it is worth to mention that the asymptotically free scalar field theory which has recently been considered on a lattice @xcite could actually be the @xmath157-field theory with a negative coupling to which we refer here .
* acknowledgments * we dedicate this article to professor yukhnovskii in grateful recognition of his efficient and generous help in fostering the ukrainian - french symposium held in lviv in february 1993 , with the hope that in the future the contacts between our two communities will further develop .
for technical reasons , instead of studying eq .
( [ eq4 ] ) , we consider the differential equation satisfied by @xmath158 ( i.e. the second derivative of the potential with respect to the field ) : starting with a known initial function ( at `` time '' @xmath48 ) , we follow its evolution in @xmath11 by approximating the differential equation ( [ app4 ] ) by finite differences and a two dimensional grid with the uniform spacings @xmath160 and @xmath161 . the finite difference formulas for the derivatives
@xmath162 and @xmath163 have been chosen with the accuracy @xmath164 : @xmath165 + o(dy^{4 } ) \label{app ' }
\\ g^{\prime \prime } ( y ) & = & \frac{16}{12dy^{2}}\left [ g(y+dy)+g(y - dy)-30g(y)\right .
\nonumber \\ & & \left .
-g(y+2dy)-g(y-2dy)\right ] + o(dy^{4 } ) \label{app''}\end{aligned}\ ] ] the evolutionary function @xmath166 is known ( calculated ) at the discret set of points @xmath167 with ( @xmath168 ) and a maximum value @xmath169 .
( this value is large enough to study the approach to the wilson - fisher fixed point with a great accuracy but is too small for studying precisely the approach to the gaussian fixed point . ) at each time @xmath170 , the derivatives are estimated from @xmath171 by using eqs([app],[app ] ) which apply only for @xmath172 .
for the marginal points @xmath173 we use the parity of @xmath174 [ by inserting @xmath175 for @xmath176 in eqs([app],[app ] ) ] .
for the two other marginal points @xmath177 of the grid , there is no fixed solution and we shall used alternately the two following conditions [ using the obvious abreviation @xmath178 instead of @xmath179 : @xmath181 \label{app2 ' } \\
g^{\prime \prime } ( i ) & = & \frac{1}{dy^{2}}\left [ \frac{915}{244}g(i)-\frac{77% } { 6}g(i-1)+\frac{107}{6}g(i-2)-13g(i-3)\right .
\nonumber \\ & & \left .
+ \frac{61}{12}g(i-4)-\frac{5}{6}g(i-5)\right ] + o(dy^{4 } ) \label{app2''}\end{aligned}\ ] ] condition 2 is more accurate than condition 1 but leads sometimes to strong unstabilities which do not appear when we first use condition 1 and then condition 2 after some finite `` time '' @xmath182 .
the validity of the procedure is tested by , for example , trying to approach a given fixed point ( see the main part of the paper ) .
[ [ appearance - of - spurious - fixed - points - in - the - approach - to - the - gausian - fixed - pointspurious ] ] appearance of spurious fixed points in the approach to the gausian fixed point[spurious ] ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ in trying to determine the attractive tricritical trajectory ( approaching the gaussian fixed point ) , we have encountered a spurious twice unstable fixed point lying at some finite and non negligible distance to the gaussian fixed point .
to understand the origin of this undesirable numerical effect , it is necessary to discuss a bit the solution of the fixed point equation @xmath183 . from ( [ eq4 ] ) or ( [ app4 ] ) , one sees that the fixed point equation is a second order non linear differential equation and a solution would be parametrized by two arbitrary constants .
one of these two constants may easily be determined : since @xmath184 is expected to be an even function of @xmath7 [ o(1 ) symmetry ] then @xmath185 may be imposed .
it remains one free parameter , thus a one - parameter family of ( nontrivial ) fixed points are solutions to the differential equation .
but there is not an infinity of physically acceptable fixed points ; all but a finite number of the solutions in the family are singular at some @xmath186 @xcite .
formally , by requiring the physical fixed point to be defined for all @xmath7 then the acceptable fixed points are limited to the gaussian fixed point and ( for @xmath10 ) to the wilson - fisher fixed point .
however , in our study , because we numerically consider the function @xmath187 in some finite range of values of @xmath7 ( see above : @xmath188 ) , it appears that in approaching the origin of @xmath11 , infinitely many pseudo - fixed points exist which have there @xmath186-singularity located outside the finite range explicitly considered and there is at least one of them which looks like a tricritical fixed point . when we enlarge the range of @xmath7 to @xmath189 , the previously observed nontrivial tricritical fixed point disappeared to the benefit of another one located closer to the origin . in conclusion , a larger and larger number of grid - points must be considered as one tries to come closer and closer to the gaussian fixed point .
this particularity together with the slowness of the approach along a marginal direction makes it excessively difficult to come very close to the gaussian fixed point .
j. f. nicoll , t. s. chang and h. e. stanley , `` _ approximate renormalization group based on the wegner -houghton differential generator _ '' , phys .
33 * , 540 ( 1974 ) ; `` _ exact and approximate differential renormalization - group generators _ '' , phys . rev . * a13 * , 1251 ( 1976 ) . g. felder , `` _ non - trivial renormalization group fixed points _ '' , in `` _
@xmath191 international congress on mathematical physics , marseille _ '' ( 1986 ) ; `` _ renormalization group in the local potential approximation _ '' , com .
phys . * 111 * , 101 ( 1987 ) . f. j. wegner , `` _ corrections to scaling laws _ '' , phys .
rev . * b5 * , 4529 ( 1972 ) ; `` _ the critical state , general aspects _ '' , in phase transitions and critical phenomena vol .
* vi * , p. 7 , _ ed . by _ c. domb and m.s .
green ( acad .
press , n .- y . , 1976 ) .
toledano , l. michel , p. toledano and e. brzin , `` _ renormalization - group study of the fixed points and of their stability for phase transitions with four - component order parameters _ '' , phys . rev . * b11 * , 7171 ( 1985 ) .
k. gawedski and a. kupiainen , `` _ non - trivial continuum limit of a @xmath194 model with negative coupling constant _ '' , nucl . phys . * b257 [ fs14 ] * , 474 ( 1985 ) ; `` _ asymptotic freedom beyond perturbation theory _ '' , in `` _ critical phenomena , random systems , gauge theories _ '' , p. 185 , _ ed . by _ k. osterwalder and r. stora ( elsevier science pub .
b.v . , 1986 ) . 1 .
evolutions for @xmath10 of the first four hamiltonian parameters @xmath195 , @xmath196 , @xmath197 , @xmath198 in a close approach to the wilson - fisher fixed point @xmath42 along t@xmath199 or t@xmath78 .
the effective inverse eigenvalue @xmath200 is given by eq .
( [ eq - omegaeff ] ) for @xmath201 .
all these quantities reach the same universal value @xmath56 characteristic of the least irrelevant eigendirection of @xmath42 . to get this close approach to @xmath42 from eq .
( [ eq4 ] ) , the initial critical value @xmath69 corresponding to @xmath202 , has been determined with more than twenty digits.[fig1 ] 2 . when a second condition is imposed on the initial hamiltonian parameters , the approach to @xmath42
may be adjusted such as to asymptotically take the second least irrelevant eigendirection . here
@xmath203 is given by eq .
( [ eq - omegaeff ] ) for @xmath79 it clearly undergoes ( full line ) a flat inflexion point at the value @xmath204 corresponding to an approach to @xmath42 along t@xmath77 , the greater the critical parameter @xmath67 is accurately determined , the longer is the flat extremum . because @xmath67 is not completely determined [ within the available accuracy in solving eq .
( [ eq4 ] ) ] the trajectory leaves the direction of t@xmath77 to take one of the two directions of approach associated to the least irrelevant inverse eigenvalue @xmath56 ( corresponding to t@xmath60 or t@xmath78 as indicated by dashed curves ) . here
, the trajectory corresponding to the full line goes along t@xmath205 .
again a flat extemum of @xmath206 indicates the approach along an eigenvector of @xmath42 and requires an accurate determination of the critical value @xmath69 .
because this determination is not complete , the trajectory ends up going away from @xmath42 as indicated by the sudden departure of @xmath207 from @xmath56 for the large values of @xmath24.fig2 3 .
this figure is a continuation of fig .
it shows the various plateaux that @xmath81 undergoes along a rg trajectory first adjusted to approach @xmath42 along the second irrelevant direction ( plateau at @xmath204 ) .
because it is not possible to determine exactly the initial conditions , the trajectory always ends up going away the fixed point towards the trivial high temperature fixed point characterized by the classical value @xmath85 ( for minus the inverse of @xmath81 , thus the final plateau at @xmath208 ) . in - between ,
the rg flow has been influenced by the close vicinity of the least irrelevant eigenvector ( plateau at @xmath56 ) and that of the relevant eigenvector ( plateau at @xmath209 ) .
the various regimes of the rg flows are indicated by the vertical arrows on the left ( direction of the flow with respect to the fixed point ) and on the right of the figure ( distance to the fixed point).[fig3 ] 4 .
domains of attraction and repulsion of the gaussian fixed point .
the figure represents projections onto the plane @xmath210 of various rg trajectories running in the space @xmath11 minus one dimension .
the flows have been obtained by solving eq .
( [ eq4 ] ) .
black circles represent the gaussian and the wilson - fisher ( w - f fp ) fixed points .
the arrows indicate the directions of the rg flows on the trajectories .
the ideal trajectory ( dot line ) which interpolates between these two fixed points represents the usual renormalized trajectory t@xmath60 corresponding to the so - called @xmath211 renormalized field theory in three dimensions ( usual rt for @xmath20 ) .
white circles represent the projections onto the plane of initial critical hamiltonians . for @xmath212 ,
the effective hamiltonians run toward the ising fixed point asymptotically along t@xmath60 ( simple fluid ) . instead , for @xmath213 and according to the values of hamiltonian coefficients of higher order ( @xmath214 , @xmath215 , etc . ) , the rg trajectories either ( a ) meet an endless rt emerging from the gaussian fp(dashed curve ) and lying entirely in the sector @xmath21 or ( b ) meet the usual rt t@xmath60 to reach the ising fixed point .
the frontier which separates these two very different cases ( a and b ) corresponds to initial hamiltonians lying on the tri - critical subspace @xmath4 ( white square c ) .
this is a source of rg trajectories flowing asymptotically toward the gaussian fpalong the tricritical rt .
notice that the coincidence of the initial point b with the rg trajectory starting at point a is not real ( it is accidental , due to a projection onto a plane of trajectories lying in a space of infinite dimension ) .
the points a or b could correspond to the restricted primitive model of ionic systems ( see @xcite).[fig4 ] 5 .
evolutions of an effective exponent @xmath140 [ with @xmath216 $ ] along three different families of rg trajectories ( see text for additional details ) .
the full squares indicate the evolution of @xmath140 for a family of trajectories initialized in the sector @xmath20 with @xmath202 and for various values of @xmath217 ( the same system at criticality corresponds to the white circle `` simple fluid '' of fig .
[ fig4 ] ) . when @xmath218 the effective exponent approaches the critical exponent value @xmath219 compatible with the present study .
one observes that the crossover towards the classical value @xmath85 is not complete because @xmath133 ceases to make sense before @xmath137 becomes large .
this is not the case of the evolution represented by the full circles which corresponds to trajectories initialized close to the gaussian fixed point . in this case
the complete crossover reproduces the interpolation between the gaussian and the wilson - fisher fixed points and typically corresponds to the usual answer given by field theory @xcite .
the third evolution ( full triangles ) corresponds to a family of hamiltonian initialized close to the tricritical surface but still attracted to the wilson - fisher fixed point .
one sees that the classical - to - ising crossover is complete but highly retarded compared to the two other cases . this is because at criticality , the rg flow is first attracted to the gaussian fixed point ( showing then an apparent classical value of @xmath135 ) before interpolating between the gaussian and the wilson - fisher fixed point.[fig5 ] 6 .
illustration of the evolution of the extrema @xmath220 [ minus the inverse of eq .
( [ eq - omegaeff ] ) ] for various values of @xmath221 and for the family of rg flows initialized at @xmath202 .
the extremum ( grey triangle ) disappears at some not very large value of @xmath137 ( about @xmath222 ) and does not reach the classical value @xmath85 .
this induces the partial ising - to - classical crossover drawn on fig .
[ fig5 ] ( squares).[fig6 ] | using the local potential approximation of the exact renormalization group ( rg ) equation , we show the various domains of values of the parameters of the @xmath0-symmetric scalar hamiltonian . in three dimensions , in addition to the usual critical surface @xmath1 ( attraction domain of the wilson - fisher fixed point ) , we explicitly show the existence of a first - order phase transition domain @xmath2 separated from @xmath3 by the tricritical surface @xmath4 ( attraction domain of the gaussian fixed point ) . @xmath2 and @xmath1 are two distinct domains of repulsion for the gaussian fixed point , but @xmath2 is not the basin of attraction of a fixed point .
@xmath2 is characterized by an endless renormalized trajectory lying entirely in the domain of negative values of the @xmath5-coupling .
this renormalized trajectory exists also in four dimensions making the gaussian fixed point ultra - violet stable ( and the @xmath6 renormalized field theory asymptotically free but with a wrong sign of the perfect action ) .
we also show that very retarded classical - to - ising crossover may exist in three dimensions ( in fact below four dimensions ) .
this could be an explanation of the unexpected classical critical behavior observed in some ionic systems . _
pacs 05.10.cc , 05.70.jk , 11.10.hi_ -0.7 cm -2 cm |
a finite set @xmath6 of points is in a _ general position _ if no three points of @xmath7 are collinear .
the _ complete geometric graph on @xmath7 _ , denoted by @xmath8 , is the complete graph with vertex set @xmath7 , whose edges are drawn as the straight - line segments between pairs of points of @xmath7 .
the set of points @xmath7 is in _ convex position _ if @xmath7 is the set of vertices of a convex polygon .
if @xmath7 is in convex position , we say that @xmath8 is a _ convex complete geometric graph_. krolyi , pach and tth @xcite introduced the concept of ramsey numbers for geometric graphs as follows . given a graph @xmath9 ,
the _ geometric ramsey number _ of @xmath9 , denoted by @xmath10 , is the smallest integer @xmath5 such that every complete geometric graph @xmath11 on @xmath5 vertices with edges arbitrarily coloured by two colours contains a monochromatic non - crossing copy of @xmath9 .
the _ convex geometric ramsey number _ of @xmath9 , @xmath12 , is defined the same way except that @xmath11 is restricted to the convex complete geometric graph . a graph @xmath9
is said to be _ outerplanar _ if @xmath9 can be drawn in the plane without any edge crossing and with all vertices of @xmath9 incident to the unbounded face .
apparently , the numbers @xmath10 and @xmath12 are finite only if @xmath9 is outerplanar : consider a planar but not outerplanar graph @xmath9 , then it is easy to see that one can not find a non - crossing monochromatic copy of @xmath9 in a convex complete graph .
also , it follows immediately from the definitions that @xmath13 for every outerplanar graph @xmath9 .
the ramsey numbers of outerplanar graphs , as well as of all planar graphs , are bounded by a function linear in the number of vertices by a result of chen and schelp @xcite .
in contrast , the only known general upper bound on the geometric ramsey numbers of outerplanar graphs is exponential in the number of vertices .
this bound follows from the exponential upper bound on the ramsey numbers for cliques since a monochromatic clique on @xmath5 points implies a monochromatic non - crossing occurrence of every outerplanar graph on @xmath5 vertices by the result of gritzmann et al .
@xcite ( see lemma [ l : gritzmannetal ] ) .
the geometric ramsey numbers of some outerplanar graphs are known to be both larger than linear and smaller than exponential , and it remains open whether there is a general polynomial bound for all outerplanar graphs . by a simple constructive proof , it is easy to see that for every @xmath14 , the cycle graph @xmath15 on @xmath5 vertices satisfies @xmath16 .
balko and krl @xcite constructed colourings that improve this bound to @xmath17 .
this bound is tight both in the convex and general geometric setting by an earlier result @xmath18 of krolyi , pach , tth and valtr @xcite , this shows that one can not have geometric ramsey numbers for general outerplanar graphs asymptotically smaller than @xmath19 .
krolyi et al .
@xcite found the exact value @xmath20 and the upper bound @xmath21 , where @xmath22 is the path on @xmath23 vertices .
the bounds @xmath24 remain the best known bounds on the geometric ramsey number of paths .
further results and open problems on the geometric ramsey numbers can be found in the survey of krolyi @xcite .
the ladder graphs are defined as follows .
[ defladder ] for any integer @xmath25 , the _ ladder graph _ on @xmath1 vertices , denoted by @xmath26 , is the graph composed of two paths @xmath27 and @xmath28 , together with the set of edges @xmath29\}$ ] .
see an example in fig .
[ fig : exladder ] . in this paper , we contribute to this subject by showing polynomial upper bounds on the geometric ramsey numbers of the ladder graphs , and their generalisation . in section [ s :
ladder ] , we show that the geometric ramsey numbers of the ladder graph on @xmath1 vertices are bounded by @xmath30 and @xmath3 in the convex and general case , respectively . in section
[ s : generalisation ] , we generalise the polynomial upper bounds to the class of all subgraphs of pathwidth-2 outerplanar triangulations , see definition [ def : pw2 ] .
these bounds are @xmath31 and @xmath32 in the convex and general case , respectively in section [ s : ladder : ordered ] , we consider the closely related area of the ordered ramsey theory .
the ordered ramsey theory recently gained a lot of attention @xcite , mainly in the more general hypergraph setting .
an _ ordered graph _
@xmath9 is a graph with a total order @xmath33 on the vertices of @xmath9 .
we say that an ordered graph @xmath9 is a subgraph of an ordered graph @xmath34 if the vertices of @xmath9 can be injectively mapped to the vertices of @xmath34 while preserving both the ordering and the edges of @xmath9 .
the _ ordered ramsey number _ @xmath35 of ordered graphs @xmath36 and @xmath9 is the smallest number @xmath37 such that every @xmath0-colouring of the edges of the ordered complete graph @xmath38 on @xmath37 vertices either contains a blue copy of @xmath36 or a red copy of @xmath9 .
the proof of the upper bound on the convex geometric ramsey number of the ladder graph in section [ s : ladder : conv ] can be extended to show that the ordered ramsey number @xmath39 of the ladder graph @xmath26 with specifically ordered vertices is at most @xmath30 .
the ideas of the proof are applied in section [ s : ladder : ordered ] to give an @xmath40 upper bound on the ordered ramsey number @xmath41 , where @xmath42 is the ordered complete graph on @xmath5 vertices and @xmath43 is an arbitrarily ordered path on @xmath44 vertices .
we note here that all colourings in this paper , unless specified , refer to edge colourings . as a convention , in any @xmath0-colouring
, we assume that the colours used are blue and red .
when @xmath45 is a colour , we say that @xmath46 is a _
@xmath45-neighbour _ of @xmath47 if the edge @xmath48 has colour @xmath45 .
let @xmath49 be the set of @xmath45-neighbours of a vertex @xmath46 .
we abbreviate the set @xmath50 with @xmath51 $ ] and @xmath52 with @xmath53 $ ] .
we write @xmath54 for the sequence @xmath55 .
the sequence of vertices @xmath56 is a _ path of colour @xmath45 _ and length @xmath57 if every pair @xmath58 , @xmath59 $ ] is an edge and has colour @xmath45 .
a sequence @xmath60 is said to be a _ partition _ of @xmath61 if @xmath62 are pairwise disjoint and @xmath63 .
[ s : ladder ] in subsections [ s : ladder : conv ] and [ s : ladder : general ] , we prove upper bounds on the convex and geometric ramsey numbers @xmath64 and @xmath65 of ladder graphs @xmath26 .
both proofs use the following lemma due to gritzmann et al .
@xcite .
[ l : gritzmannetal ] let @xmath9 be an outerplanar graph on @xmath5 vertices and let @xmath7 be a set of @xmath5 points in general position .
then @xmath8 contains a non - crossing copy of @xmath9 .
in subsection [ s : ladder : ordered ] , a small change to the proof of the upper bound on the convex ramsey number is shown to give an upper bound on the ordered ramsey number of paths .
[ s : ladder : conv ] [ t : ladder ] for every @xmath25 , @xmath66 . in this section ,
let @xmath67 denote a set of @xmath30 points in convex position .
that is , @xmath67 is the set of vertices of some convex polygon .
we label the vertices @xmath68 in the clockwise order starting at an arbitrarily chosen vertex @xmath69 .
we write @xmath70 if and only if @xmath71 .
let @xmath72 .
we say that @xmath61 precedes @xmath73 and write @xmath74 if and only if for every @xmath75 and every @xmath76 , @xmath77 .
notice that if @xmath78 , then the sets @xmath61 and @xmath73 can be separated by a line . for a pair of disjoint vertex sets @xmath79 , @xmath80 , @xmath81 ,
the complete bipartite graph on @xmath79 , denoted by @xmath82 , is the set of edges @xmath48 , where @xmath83 and @xmath84 .
a complete bipartite graph @xmath82 is said to be _ well - split _ if @xmath85 or @xmath86 .
a well - split @xmath87 is a well - split @xmath82 , for some @xmath88 and @xmath89 such that @xmath90 , @xmath91 . the following lemma and its generalisation ( stated in corollary [ c : bippw ] in the next section )
are used frequently in later proofs .
[ l : completebipartite ] if a @xmath0-colouring of @xmath92 contains a monochromatic well - split @xmath93 , then it contains a monochromatic non - crossing copy of @xmath26 .
let @xmath94 and @xmath95 be the two vertex parts of the monochromatic well - split @xmath93 . without loss of generality , we assume that all the edges between @xmath94 and @xmath95 are coloured blue .
we use an idea that was used to prove a quadratic upper bound on @xmath96 and other results on geometric ramsey numbers @xcite .
we define partial orders @xmath97 on @xmath94 and @xmath98 on @xmath95 as follows .
a path @xmath99 on the vertices of @xmath62 is an _ increasing path _ if @xmath100 .
let @xmath101 for @xmath102 if and only if there exists an increasing blue path starting in @xmath47 and ending at @xmath46 .
since @xmath103 for @xmath104 , by a lemma of dilworth @xcite , each of @xmath105 has either a chain on @xmath5 elements or an antichain on @xmath1 elements . by the definition of @xmath105 ,
any two vertices that are incomparable in @xmath62 are connected by a red edge .
therefore , if @xmath105 contains an antichain with @xmath1 elements , then there exists a red convex complete geometric graph on @xmath1 vertices . by lemma [ l : gritzmannetal ]
, @xmath92 contains a red non - crossing copy of @xmath26 .
thus , we may assume that none of @xmath105 contains an antichain with @xmath1 elements .
then both @xmath105 contain a chain with @xmath5 elements , implying that each of @xmath62 contains an increasing blue path with @xmath5 vertices .
let @xmath27 and @xmath28 be the increasing blue paths on @xmath94 and @xmath95 , respectively .
these two paths together with the blue edges @xmath106 , @xmath107 $ ] form a blue non - crossing copy of @xmath26 .
[ l : longpath ] let @xmath37 and @xmath5 be positive integers .
let @xmath9 be the complete graph on a set @xmath61 of at least @xmath108 vertices and let @xmath109 be a partition of @xmath61 with @xmath110 for every @xmath111 $ ] .
then for any 2-colouring of the edges of @xmath9 , either there is a red path @xmath27 with @xmath112 for each @xmath111 $ ] or for some @xmath113 $ ] there exists a blue @xmath114 with @xmath115 , @xmath116 and @xmath117 .
assume that there is no red path @xmath28 with @xmath118 for each @xmath111 $ ] .
we call a vertex @xmath119 _ good _ if there is a red path @xmath120 with @xmath121 and @xmath118 for every @xmath122 $ ] .
every vertex in @xmath94 is good and all vertices in @xmath123 are bad .
let @xmath124 be the largest integer such that at least half of the vertices of @xmath62 are good .
then by the choice of @xmath124 , at least half of the vertices of @xmath125 are bad .
let @xmath126 denote the set of good vertices in @xmath62 and @xmath127 the set of bad vertices in @xmath125 .
it follows that both @xmath128 and @xmath127 have size at least @xmath129 and all the edges between @xmath128 and @xmath127 are blue .
let @xmath67 denote a set of @xmath130 points in convex position . arbitrarily choose a line that partitions @xmath67 into @xmath131 and @xmath132 each containing exactly @xmath133 points .
further , partition @xmath131 into @xmath134 with @xmath135 and @xmath136 for each @xmath137 $ ] .
partition @xmath132 into @xmath138 with @xmath139 and @xmath140 for each @xmath137 $ ] .
colour each vertex @xmath141 red if it is adjacent to at least half of the vertices in @xmath126 by a red edge .
otherwise , colour it blue .
we say that @xmath62 is red if at least half of the vertices in @xmath62 are coloured red .
otherwise , we say that @xmath62 is blue . without loss of generality ,
at least half of the sets @xmath62 are red .
let @xmath142 be an increasing sequence of indices such that each @xmath143 is red .
let @xmath144 be the set of red vertices of @xmath143 .
thus @xmath145 and for every vertex @xmath46 from @xmath146 , @xmath147 .
if for some @xmath113 $ ] , there exists a blue @xmath148 with @xmath149 , @xmath150 and @xmath151 , then @xmath92 contains a blue non - crossing copy of @xmath26 by lemma [ l : completebipartite ] .
thus , by lemma [ l : longpath ] , we can assume that we have a red path @xmath28 with @xmath152 for every @xmath111 $ ] .
so @xmath153 .
for each @xmath111 $ ] , let @xmath154 .
so @xmath155 . if there exists a blue @xmath148 with @xmath156 , @xmath157 and @xmath151 , then the proof is complete by lemma [ l : completebipartite ] .
thus by lemma [ l : longpath ] , we only need to consider the case when there is a red path @xmath158 with @xmath159 for every @xmath111 $ ] .
we have @xmath160 and so the two paths @xmath28 and @xmath158 together with the edges @xmath161 , @xmath111 $ ] form a red non - crossing copy of @xmath26 .
[ s : ladder : ordered ] the proof of theorem [ t : ladder ] shows that the ordered ramsey number @xmath39 of the ladder graph @xmath26 with vertices ordered @xmath162 is at most @xmath30 .
the ideas used in the proof of theorem [ t : ladder ] can be applied to give a subexponential upper bound on the ordered ramsey numbers of arbitrarily ordered paths .
[ t : orderedramsey ] let @xmath42 be the ordered complete graph on @xmath5 vertices and let @xmath43 be an arbitrarily ordered path on @xmath44 vertices . then @xmath163 .
if @xmath164 , the claim holds trivially .
we proceed by induction on @xmath5 while @xmath44 remains fixed .
it is enough to show that @xmath165 for values @xmath5 and @xmath44 of the form @xmath166 and @xmath167 for some integers @xmath168 and @xmath57 .
let @xmath169 .
let @xmath170 be the complete ordered graph on @xmath89 vertices with @xmath0-coloured edges .
we split the vertices of @xmath170 into @xmath44 intervals @xmath171 , ,
@xmath172 , each containing @xmath173 consecutive vertices .
let @xmath174 be the vertices of @xmath43 .
then the edges of @xmath43 are @xmath175 for some permutation @xmath176\rightarrow [ m]$ ] .
we let @xmath177 for every @xmath111 $ ] . by lemma [ l : longpath ] , we either find a red copy of @xmath43 , in which case the claim holds , or we find a pair of intervals @xmath178 , @xmath125 satisfying the following .
there are sets @xmath179 and @xmath180 of size @xmath181 such that all the edges between @xmath88 and @xmath89 are blue . by the induction hypothesis , @xmath182 .
thus in each of @xmath88 and @xmath89 , we either find a red copy of @xmath43 , or a blue copy of @xmath183 . if either @xmath88 or @xmath89 contains a red copy of @xmath43 , the claim holds .
otherwise both @xmath88 and @xmath89 contain a blue copy of @xmath183 and so @xmath184 is a blue copy of @xmath42 .
[ c : orderedramsey ] let @xmath22 be a path on @xmath5 arbitrarily ordered vertices
. then @xmath185 .
[ s : ladder : general ] [ gent : ladder ] the geometric ramsey number of the ladder graph @xmath26 satisfies @xmath186 . two sets of points @xmath61 and @xmath73 in the plane are _ mutually avoiding _
if @xmath187 and no line subtended by a pair of points in @xmath61 intersects the convex hull of @xmath73 , and vice versa . see fig .
[ fig : muavsets ] .
a simple example of a pair of mutually avoiding sets are sets @xmath61 and @xmath73 such that @xmath188 is in convex position and @xmath61 and @xmath73 can be separated by a straight line .
observe that for any mutually avoiding pair @xmath189 , every point in @xmath61 `` sees '' all the vertices in @xmath73 in the same order and vice versa .
that is , there are unique total orders @xmath190 of the points in @xmath61 and @xmath191 of the the points in @xmath73 such that every point in @xmath73 `` sees '' @xmath192 consecutively in a clockwise order before seeing any vertex in @xmath73 , whereas every point in @xmath61 `` sees '' @xmath193 consecutively in a counterclockwise order before seeing any vertex in @xmath61 .
a path @xmath99 in either @xmath61 or @xmath73 is an _ increasing path _ if @xmath194 .
for any two sets of vertices @xmath195 both contained in @xmath61 ( or @xmath73 ) , we write @xmath196 if and only if for every @xmath197 and @xmath198 , @xmath199 .
let @xmath200 be @xmath61 or @xmath73 .
a sequence @xmath201 of subsets of @xmath200 is an _ increasing sequence _ if @xmath202 .
an increasing sequence @xmath201 of subsets of @xmath200 is an _ increasing partition _ of @xmath200 if @xmath203 } u_i = u$ ] .
the following proposition follows from the definition of a pair of mutually avoiding sets .
[ prop : muav ] assume @xmath61 and @xmath73 are mutually avoiding
. then 1 .
an increasing path @xmath99 does not cross itself .
2 . let @xmath204 with @xmath205 and let @xmath206 .
then the two edges @xmath207 and @xmath208 cross if and only if @xmath209 .
3 . let @xmath204 with @xmath205 and let @xmath210 such that @xmath211 or @xmath212 or @xmath213 .
let @xmath76 .
then the two edges @xmath214 and @xmath215 do not cross .
the following corollary follows directly from proposition [ prop : muav ] .
[ cor : muavladder ] assume @xmath61 and @xmath73 are mutually avoiding .
let @xmath216 be an increasing path in @xmath61 and let @xmath217 be an increasing path in @xmath73 .
then the ladder graph composed of the paths @xmath218 and @xmath219 and edges @xmath220\}$ ] is non - crossing . given a set of points in general position , the following theorem guarantees the existence of two mutually avoiding subsets of relatively large sizes .
[ t : avoiding ] let @xmath221 and @xmath222 be two sets of points separated by a line , each of size @xmath223 . then there exist mutually avoiding sets @xmath224 and @xmath225 such that @xmath61 and @xmath73 are both of size @xmath5 .
an embedding of the complete bipartite graph @xmath87 on a set of points in general position is _ well - split _ if the two sets of points representing the two vertex parts are mutually avoiding . by lemma [ l : gritzmannetal ] and corollary [ cor : muavladder ] , we have the following generalisation of lemma [ l : completebipartite ] .
[ genl : completebipartite ] if a @xmath0-colouring of @xmath11 contains a monochromatic well - split @xmath93 , then it contains a monochromatic non - crossing @xmath26 .
let @xmath94 and @xmath95 be the two vertex parts of the monochromatic well - split @xmath93 . without loss of generality ,
assume @xmath226 is blue .
by applying the dilworth s lemma @xcite in the same way as in the proof of lemma [ l : completebipartite ] , we either find a red @xmath227 or blue increasing paths @xmath27 in @xmath94 and @xmath28 in @xmath95 .
in the first case , we get a red @xmath26 by lemma [ l : gritzmannetal ] and in the second case a blue @xmath26 by corollary [ cor : muavladder ] .
a complete geometric bipartite graph @xmath82 is said to be _ separable _ if @xmath88 and @xmath89 can be separated by a line .
notice that if @xmath184 is in convex position , then @xmath82 is separable if and only if it is well - split .
obviously , every complete bipartite geometric graph @xmath82 contains a separable complete bipartite graph with parts of sizes @xmath228 and @xmath229 .
however , all complete bipartite geometric graphs that we encounter in subsequent proofs are separable , so we state the following corollary of theorem [ t : avoiding ] and lemma [ genl : completebipartite ] for separable complete bipartite graphs only .
[ c : completebipartite ] every @xmath0-colouring of @xmath11 containing a monochromatic separable @xmath230 contains a monochromatic non - crossing copy of @xmath26 .
let @xmath9 be the complete geometric graph on vertex set @xmath7 , where @xmath7 is a set of @xmath231 points in general position , where @xmath45 is some sufficiently large absolute constant . by theorem [ t : avoiding ] ,
there exist two subsets @xmath232 , such that @xmath233 and @xmath234 are mutually avoiding and @xmath235 for some @xmath236 .
the proof of theorem [ gent : ladder ] is analogous to that of theorem [ t : ladder ] with @xmath233 and @xmath234 having the role of @xmath131 and @xmath132 .
let @xmath134 be the increasing partition of @xmath233 with @xmath237 for each @xmath137 $ ] .
let @xmath138 be the increasing partition of @xmath234 with @xmath238 for each @xmath137 $ ] .
as in the proof of theorem [ t : ladder ] , we find one colour , that we assume to be red , an increasing sequence @xmath239 of subsets of @xmath234 and an increasing sequence @xmath142 of integers from @xmath240 $ ] satisfying the following
. for every @xmath111 $ ] , @xmath241 and every vertex in @xmath146 is adjacent to at least half of the vertices of @xmath242 by a red edge . by lemma [ l : longpath ] either there is a blue copy of @xmath243 , where @xmath244 or there is a red path @xmath28 with @xmath152 for every @xmath111 $ ] . in the first case the proof is complete by corollary [ c : completebipartite ] . in the second case
we let @xmath245 .
now we apply lemma [ l : longpath ] on @xmath246 and either find a blue @xmath243 or a red path @xmath247 with @xmath248 for every @xmath111 $ ] . in the first case the proof is complete by corollary [ c : completebipartite ] and in the second by corollary [ cor : muavladder ] using @xmath28 and @xmath158 as the two paths of the ladder graph .
[ s : generalisation ] an _ outerplanar triangulation _
@xmath9 is a planar graph that can be drawn in the plane in such a way that the outer face is incident with all the vertices of @xmath9 and every other face is incident with exactly three vertices .
the pathwidth of a graph was first defined by robertson and seymour @xcite as follows .
a _ path decomposition _ of a graph @xmath9 is a sequence @xmath249 of subgraphs of @xmath9 such that each edge of @xmath9 is in at least one of @xmath250 and for every vertex @xmath46 of @xmath9 , the set of graphs @xmath250 containing @xmath46 forms a contiguous subsequence of @xmath249 .
the _ pathwidth _ of a graph @xmath9 is the smallest @xmath168 such that @xmath9 has a path decomposition in which every @xmath250 has at most @xmath251 vertices .
let @xmath252 denote the pathwidth of @xmath9 .
a _ pathwidth-@xmath168 _
graph is a graph of pathwidth at most @xmath168 . for every @xmath168 ,
the class of graphs of pathwidth at most @xmath168 is a minor - closed class .
every such class can be characterised by a finite list of forbidden minors by the graph minor theorem @xcite .
a characterisation of the class of pathwidth-@xmath0 graphs with @xmath253 forbidden minors was provided by kinnersley and langston @xcite .
simplified characterisations of pathwidth-@xmath0 graphs were obtained recently by bart , hajnal , lin and young @xcite and bir , keller and young @xcite .
we use these characterisations to provide an equivalent definition of pathwidth-@xmath0 outerplanar triangulations that will be used in our proofs .
[ def : pw2 ] let @xmath254 be the class of outerplanar triangulations @xmath9 on @xmath5 vertices whose vertices can be decomposed into two disjoint sets @xmath255 such that the subgraphs induced by the two sets , @xmath256 $ ] and @xmath257 $ ] , are paths .
see an example in fig .
[ fig : expw ] .
[ p : pathwidth ] a graph @xmath9 is a pathwidth-@xmath0 outerplanar triangulation if and only if @xmath258 .
a _ track _ is a graph composed of two _ rails _ and several _ cross - ties_. the two rails are paths @xmath259 and @xmath260 .
a cross - tie is a path of length one or two that connects @xmath261 with @xmath262 , for some @xmath263 $ ] and @xmath264 $ ] .
the cross - ties further satisfy that for every @xmath265 with @xmath266 , whenever one cross - tie connects @xmath261 to @xmath262 and another connects @xmath267 to @xmath268 , then @xmath269 .
the middle vertex of a cross - tie of length two has no neighbours other than @xmath261 and @xmath262 .
additionally , there always is a cross - tie of length one connecting @xmath270 to @xmath271 and another connecting @xmath272 to @xmath273 .
bart et al .
@xcite prove that a graph is a @xmath0-connected pathwidth-@xmath0 graph if and only if it is a track .
notice that every outerplanar triangulation is hamiltonian and thus @xmath0-connected .
observe also that every @xmath258 satisfies the definition of a track .
it remains to show that if a track @xmath9 on @xmath5 vertices is an outerplanar triangulation then @xmath258 .
if the track @xmath9 has a cross - tie of length two between @xmath270 and @xmath271 , then the middle vertex @xmath274 of the cross - tie can be added to one of the tracks to form , for example , the track @xmath275 .
thus we can assume , that the track @xmath9 has no cross - tie of length two connecting @xmath270 to @xmath271 or @xmath272 to @xmath273 .
then , since @xmath9 is outerplanar , it has no cross - tie of length two .
therefore the outerplanar triangulation @xmath9 satisfies the definition of graphs from @xmath254 .
the following is a corollary of property [ prop : muav ] .
[ cor : muavpw2 ] let @xmath276 and let @xmath9 be composed of induced paths @xmath277 , @xmath278 and edges between vertices of @xmath218 and vertices of @xmath219 .
let @xmath279 be a pair of mutually avoiding sets .
let @xmath280 be an increasing path in @xmath61 and @xmath281 an increasing path in @xmath73 .
then by mapping every @xmath261 on @xmath282 and every @xmath283 on @xmath284 we obtain a non - crossing embedding of @xmath9 . by corollary [ cor : muavpw2 ] ,
lemma [ l : completebipartite ] generalises to an arbitrary graph @xmath258 .
[ c : wellsplitbippw ] if @xmath11 with @xmath0-coloured edges contains a monochromatic well - split @xmath285 , then it contains a monochromatic non - crossing copy of every @xmath9 from @xmath254 .
then , by theorem [ t : avoiding ] , we also generalise corollary [ c : completebipartite ] . [ c : bippw ] if @xmath11 with @xmath0-coloured edges contains a monochromatic separable @xmath286 , then it contains a monochromatic non - crossing copy of every @xmath9 from @xmath254 .
we obtain upper bounds for the geometric ramsey numbers of graphs @xmath287 , both in the convex case and in the general case .
these two upper bounds follow directly from the following key lemma .
[ l : pathwidthboth ] let @xmath9 be a subgraph of a graph @xmath288 .
let @xmath289 and let @xmath233 and @xmath234 be two mutually avoiding sets of @xmath290 points each .
then every @xmath0-colouring of the complete geometric graph on @xmath291 either contains a monochromatic @xmath9 or a monochromatic separable @xmath292 .
we leave the technical proof of lemma [ l : pathwidthboth ] to the next section .
[ t : pathwidthconvex ] for any @xmath293 , @xmath294 .
let @xmath295 be a set of @xmath31 points in convex position .
we cut the set @xmath295 by a line into sets @xmath233 and @xmath234 of size @xmath296 each .
then @xmath233 and @xmath234 are mutually avoiding .
moreover if either @xmath233 or @xmath234 contains a monochromatic separable and thus well - split @xmath285 , then @xmath295 contains a monochromatic non - crossing @xmath9 by corollary [ c : wellsplitbippw ] .
therefore , by lemma [ l : pathwidthboth ] with @xmath297 , @xmath295 contains a monochromatic non - crossing copy of @xmath9 .
[ t : pathwidthgeneral ] for any @xmath293 , @xmath298 . let @xmath295 be a set of @xmath299 points in general position . by theorem [ t : avoiding ] , @xmath295 contains mutually avoiding sets @xmath233 and @xmath234 of size @xmath300 each .
if @xmath295 contains a monochromatic separable @xmath301 , then it contains a monochromatic non - crossing copy of @xmath9 by corollary [ c : bippw ] . therefore , by lemma [ l : pathwidthboth ] with @xmath302 , @xmath295 contains a monochromatic non - crossing copy of @xmath9 .
notice that not every pathwidth-@xmath0 outerplanar graph is a subgraph of a pathwidth-@xmath0 outerplanar triangulation .
see fig .
[ fig : nottriang ] .
without loss of generality , we only need to consider @xmath287 . if @xmath303 then the result is trivial , so we assume @xmath304 , which will be used in several estimates .
let @xmath305 and @xmath306 be sequences of vertices such that @xmath307 and @xmath308 for every @xmath59 $ ] , @xmath309 , @xmath310 and the alternating sequence of vertices @xmath311 forms a path in @xmath9 .
note that @xmath312 and that such a path is unique and can be constructed by starting at @xmath271 and always continuing to the largest neighbour in the other of the sets @xmath313 until reaching @xmath273 .
vertices @xmath314 are called the _ stem vertices _ and all the other vertices of @xmath9 are the _ leaf vertices_. notice that every leaf vertex of @xmath218 has exactly one neighbour in @xmath219 and vice versa .
refer to fig .
[ fig : pw - start ] .
we cut @xmath218 into a sequence of subpaths @xmath315 , where @xmath316 contains @xmath317 and all the vertices preceding @xmath317 on @xmath218 , @xmath318 contains all the vertices after @xmath319 and for @xmath320 $ ] , @xmath321 contains @xmath322 and the vertices strictly between @xmath323 and @xmath282 .
let @xmath324 be the sequence of subpaths created by removing vertices @xmath284 from @xmath219 .
that is , for every @xmath325 $ ] , @xmath326 contains the leaf vertices in between @xmath284 and @xmath327 .
let @xmath328 .
[ cl : initialpartition ] there exists a colour @xmath329 and sequences of vertex sets @xmath330 , @xmath331 and @xmath332 with @xmath333 that satisfy the following conditions .
1 . @xmath334 : |b_i| = 8 mn^3 $ ] ; 2 .
@xmath335 : |a'_i| = 4 mn^2 $ ] ; 3 .
@xmath334 : |m_i| = 9 m^2n^2 $ ] ; 4 .
[ it : halfneigh ] @xmath334 : \forall v \in a'_i:$ ] @xmath336 .
let @xmath337 be the increasing partition of @xmath233 with parts of size @xmath338 .
then we take sequences @xmath339 , @xmath340 of subsets of @xmath234 satisfying @xmath341 with @xmath342 for every @xmath343 $ ] and @xmath344 whenever @xmath345 $ ] .
the _ colour of a vertex @xmath346 _ is the colour of the majority of the edges between @xmath46 and the vertices of @xmath347 and it is red in case of a tie .
the _ colour of @xmath348 _ is the colour of the majority of the vertices @xmath346 and it is red in case of a tie .
we fix @xmath45 to be the colour such that at least half of the sets @xmath348 have colour @xmath45 .
let @xmath349 be an increasing sequence of indices such that for every @xmath59 $ ] , @xmath350 has colour @xmath45 .
for each @xmath59 $ ] let @xmath351 be the set of vertices of @xmath350 with colour @xmath45 .
then @xmath352 . let @xmath353 for every @xmath325 $ ] and @xmath354 for every @xmath325 $ ] .
it is easy to verify that the sets @xmath351 , @xmath126 and @xmath355 satisfy the requirements .
[ o : intersectingsets ] let @xmath37 , @xmath168 and @xmath274 be positive integers .
let @xmath295 be a set of size @xmath37 and let @xmath356 be sets such that @xmath357 for every @xmath358 $ ] .
then @xmath359 the observation is applied several times for some colour @xmath45 and a set @xmath360 of vertices in the following way .
we set @xmath361 where @xmath362 are the vertices of @xmath360 with the fewest @xmath45-neighbours in @xmath295 .
the observation says that if @xmath168 is large and every @xmath284 has few @xmath45-neighbours in @xmath295 , then we find a large complete bipartite graph in the other colour . without loss of generality , we assume that claim [ cl : initialpartition ] holds with @xmath363 .
let @xmath364 , @xmath365 and @xmath366 be the sequences that satisfy the conditions of the claim . the rest of the proof proceeds in several phases .
in each phase we either immediately find a blue well - split @xmath367 implying a monochromatic @xmath9 , a blue separable @xmath292 , or we move closer to finding a non - crossing embedding @xmath368 of @xmath9 with all edges red .
the mapping @xmath369 maps @xmath284 on some point of @xmath351 for each @xmath59 $ ] and for each @xmath325 $ ] , the vertices of @xmath321 are mapped on some points of @xmath126 and vertices of @xmath326 on some points of @xmath355 . in some phases , the embedding of some vertices of @xmath9 is selected and this will then remain fixed for the rest of the proof .
[ cl : finerpartition ] either the complete geometric graph on @xmath291 contains a monochromatic noncrossing @xmath9 or there is a sequence of sets @xmath370 with @xmath371 for every @xmath59 $ ] that satisfies the following conditions .
1 . @xmath372 ; 2 .
[ it : inters]@xmath334 , \forall u \in a_i , \forall v \in a_{i+1}:$ ] @xmath47 and @xmath46 have at least @xmath373 common red neighbours in @xmath128 ; to find the sets @xmath62 , we proceed in @xmath57 steps , unless we find a red @xmath9 earlier . in the first step ,
we let the set @xmath374 be an arbitrary subset of @xmath375 of size @xmath376 . at the beginning of step @xmath377 , @xmath378
, we have sets @xmath379 each of size @xmath376 and such that the requirement [ it : inters ] of the claim is satisfied for all @xmath380 . a vertex @xmath381 is _ compatible _ with @xmath382 if @xmath47 and @xmath46 have at least @xmath373 common red neighbours in @xmath383 .
we distinguish two cases .
the first case occurs when there is a vertex @xmath382 and a set @xmath384 of vertices of @xmath385 incompatible with @xmath46 .
let @xmath386 and for every @xmath387 $ ] , let @xmath388 .
let @xmath389 .
since the vertices of @xmath390 are incompatible with @xmath46 , we can apply observation [ o : intersectingsets ] on @xmath295 and @xmath391 with @xmath392 , @xmath393 and @xmath394 and obtain @xmath395 .
all edges between @xmath390 and @xmath67 are blue , thus @xmath396 forms a blue well - split @xmath285 and so @xmath397 contains a monochromatic noncrossing @xmath9 by corollary [ c : wellsplitbippw ] . in the second case , for every vertex @xmath398 at most @xmath399 vertices of @xmath385
are incompatible .
thus the number of vertices of @xmath385 compatible with every @xmath398 is at least @xmath400 .
we can thus let @xmath401 be the set of some @xmath376 vertices of @xmath385 compatible with every @xmath382 .
let @xmath370 be the sequence of sets satisfying the conditions of claim [ cl : finerpartition ] . to provide an exposition of the rest of the proof , we first prove lemma [ l : pathwidthboth ] for the case when there is no leaf vertex on @xmath219 .
[ cl : noleaf ] assume @xmath219 contains no leaf vertices , then there exists a blue separable @xmath292 or a monochromatic non - crossing @xmath9 .
we assume that neither @xmath233 nor @xmath234 contains a blue separable @xmath292 . by lemma [ l : longpath ] , we find a red path @xmath402 , where each @xmath403 .
then for every @xmath325 $ ] we take the set @xmath404 of @xmath373 common red neighbours of @xmath405 and @xmath406 . for
every @xmath325 $ ] , let @xmath407 be an increasing partition of @xmath408 with @xmath409 parts of size at least @xmath376 each .
by lemma [ l : longpath ] we find an increasing red path @xmath410 with exactly one vertex in each set in @xmath411 .
then we map every @xmath282 on @xmath412 and every @xmath261 on @xmath405 to obtain a red copy of @xmath9 .
the rest of this section deals with the leaf vertices on @xmath219 . for each @xmath325 $ ] such that @xmath413 , we take an increasing partition @xmath414 of @xmath355 with @xmath415 and @xmath416 for every @xmath417 $ ] .
let @xmath418 be the colouring of the edges of @xmath397 .
we define a new edge colouring @xmath419 of the edges of @xmath397 according to the following cases . 1 . the edge @xmath420 connects a vertex @xmath421 and a vertex @xmath422 such that @xmath423 .
we colour @xmath420 red if and only if @xmath46 and @xmath424 have at least @xmath399 common red neighbours in @xmath355 .
2 . the edge @xmath420 connects a vertex @xmath421 and a vertex @xmath422 such that @xmath425 .
we colour @xmath420 red if and only if @xmath426 and @xmath427 .
3 . otherwise @xmath428 .
[ cl : atleast1leaf ] if there exist sets @xmath429 and @xmath430 with @xmath431 and all edges between @xmath88 and @xmath89 blue under @xmath419 , then there exists a blue separable @xmath292 in @xmath234 under @xmath418 .
we distinguish three cases . 1
. we have @xmath423 .
if every @xmath432 has fewer than @xmath433 red neighbours in @xmath355 under @xmath418 , then there are at least @xmath434 vertices in @xmath355 that are connected by blue edges to every vertex in @xmath88 under @xmath418 .
this implies the existence of a blue separable @xmath292 in @xmath234 .
otherwise , there exists a vertex @xmath432 with at least @xmath435 red neighbours in @xmath355 . let @xmath436 denote the set of these neighbours of @xmath46 .
since every edge between @xmath88 and @xmath89 is blue under @xmath419 , each @xmath437 is connected by red edges to at most @xmath399 vertices in @xmath37 .
thus there are at least @xmath438 vertices of @xmath37 that are connected by blue edges to each vertex in @xmath89 .
thus we have a blue separable @xmath292 in @xmath234 under @xmath418 .
we have @xmath425 .
either each point of @xmath88 has fewer than @xmath439 red neighbours in @xmath440 or each point of @xmath89 has fewer than @xmath441 red neighbours in @xmath442 . without loss of generality ,
the first case occurs and then there are at least @xmath443 points in @xmath440 connected by blue edges to every point of @xmath88 .
we have @xmath444 .
then @xmath419 is equal to @xmath418 on all the edges between @xmath62 and @xmath125 .
this implies the existence of a blue separable @xmath292 . by claim [ cl : atleast1leaf ]
, we may assume that under @xmath419 , there exists no @xmath59 $ ] for which some two sets @xmath445 and @xmath430 with @xmath431 would form a blue @xmath292 .
then by lemma [ l : longpath ] , we can map each vertex @xmath284 on some point @xmath446 in such a way that @xmath447 is a red path under @xmath419 .
for every @xmath325 $ ] , let @xmath448 be a set of @xmath449 common red neighbours of @xmath450 and @xmath451 and let @xmath452 .
in what follows , for each @xmath325 $ ] we define a vertex set @xmath453 in which we then embed @xmath321 . if @xmath454 we define a vertex set @xmath455 in which we embed the only vertex of @xmath326 .
if @xmath456 we define a sequence of vertex sets @xmath457 , and on each one of these sets , we embed one of the leaf vertices from @xmath326 . refer to fig .
[ fig : pw - sets ] . if @xmath458 , we let @xmath459 .
if @xmath454 , then let @xmath460 be a set of @xmath399 common red neighbours of @xmath450 and @xmath451 .
we let @xmath461 be the subset of @xmath462 formed by the vertices connected by at least one red edge to a vertex in @xmath463 .
if @xmath464 , then we have a blue well - split @xmath367 with parts @xmath465 and @xmath463 . by corollary
[ c : wellsplitbippw ] , this implies a monochromatic non - crossing @xmath9 .
otherwise we have @xmath466 . if @xmath425 , let @xmath467 and @xmath468 . for every @xmath417 $ ] let @xmath469
thus , @xmath470 for every @xmath471 $ ] .
[ cl : starfromtop ] when @xmath425 , then either there is a monochromatic non - crossing @xmath9 or there exists a set @xmath472 of size @xmath473 such that for every @xmath471 $ ] , every point of @xmath461 has @xmath474 red neighbours in @xmath475 .
we call a vertex in @xmath462 _
good _ , if it has at least @xmath376 red neighbours in @xmath475 for every @xmath471 $ ] and _ bad _ otherwise .
we assume that the number of good vertices is smaller than @xmath476 .
the claim will be proven by finding a monochromatic non - crossing copy of @xmath9 .
the number of bad vertices in @xmath462 is at least @xmath477 .
for each bad vertex @xmath478 , label @xmath478 by @xmath377 if @xmath479 . since @xmath480
, there exists an index @xmath471 $ ] such that the number of bad vertices labelled @xmath377 is at least @xmath481 .
consider the set @xmath475 and a set @xmath390 of @xmath399 bad vertices in @xmath461 labelled @xmath377 .
for each @xmath482 , we have @xmath483 and we also have @xmath484
. then by observation [ o : intersectingsets ] , @xmath485 .
all the edges between @xmath486 and @xmath390 are blue .
this implies the occurrence of a well - split blue @xmath285 and a monochromatic non - crossing @xmath9 by corollary [ c : wellsplitbippw ] .
hence we can assume that for every @xmath124 with @xmath425 there is a set @xmath487 of @xmath476 points each having @xmath474 red neighbours in each of @xmath488 .
this completes the definition of @xmath489 for every @xmath325 $ ] .
next we take an increasing partition of each @xmath489 into @xmath409 parts , each of size at least @xmath376 . by lemma
[ l : longpath ] , we either find a blue separable @xmath292 or embed the red path @xmath490 on @xmath233 in such a way that @xmath491 is a red increasing path with every vertex of @xmath321 mapped on some point of @xmath489 .
we consider every star centred at some @xmath282 . if @xmath454 , then @xmath492 has a red neighbour in @xmath463 and we map the only vertex of @xmath326 on this red neighbour .
if @xmath425 , let @xmath493 be the @xmath376 red neighbours of @xmath492 in @xmath475 , for every @xmath471 $ ] .
we assume that there is no blue separable @xmath292 .
recall that red edges connect all the vertices of @xmath494 to @xmath495 and all the vertices of @xmath496 to @xmath451 .
this fact and lemma [ l : longpath ] imply the existence of a red increasing path between @xmath495 and @xmath451 visiting every @xmath497 exactly once .
this completes the embedding of the vertices of @xmath9 on @xmath291 that yields a monochromatic non - crossing graph isomorphic to @xmath9 .
see fig .
[ fig : pw - found ] .
the authors would like to thank to gyula krolyi for introduction to the geometric ramsey theory and to jan kynl and martin balko for discussions about the ramsey theory of ordered graphs . | we prove polynomial upper bounds of geometric ramsey numbers of pathwidth-@xmath0 outerplanar triangulations in both convex and general cases .
we also prove that the geometric ramsey numbers of the ladder graph on @xmath1 vertices are bounded by @xmath2 and @xmath3 , in the convex and general case , respectively .
we then apply similar methods to prove an @xmath4 upper bound on the ramsey number of a path with @xmath5 ordered vertices .
department of applied mathematics , charles university , faculty of mathematics and physics , + malostransk nm . 25 , 118 00 praha 1 , czech republic .
+ ` { cibulka,krcal}@kam.mff.cuni.cz ` + + max - planck - institut fr informatik , + saarbrcken , saarland , germany .
+ ` [email protected] ` + + czech technical university , faculty of information technology , + prague , czech republic .
+ ` [email protected] ` * msc codes : * 52c35 , 05c55 , 05c10 * keywords : * geometric ramsey theory , outerplanar graph , ordered ramsey theory , pathwidth |
in the past decade there has been much interest in the hydrodynamics of relativistic conformal field theories ( cfts ) , largely due to the ads / cft correspondence between ( quantum ) gravitational theories on asymptotically anti - de - sitter ( ads ) spacetimes in @xmath0 dimensions and cfts in @xmath1 dimensions @xcite . in @xcite it was shown explicitly that the @xmath1-dimensional cft hydrodynamics equations are dual to the gravitational field equations describing the evolution of large scale perturbations of the @xmath0-dimensional black brane .
this has been dubbed the fluid - gravity correspondence .
it has been noticed that this correspondence naturally completes the picture of the old membrane paradigm @xcite of the 1980 s , where black hole horizons are thought of as being analogous to fictitious viscous fluids . in this case , however , the dynamics of the entire spacetime is encoded holographically into the fluid living at the boundary of the spacetime and not at the event horizon @xcite . on the other hand , at least in some circumstances
, the horizon membrane paradigm can be modernized within the context of hydrodynamics as an effective field theory and the fluid - gravity correspondence .
the starting point is an equilibrium solution containing a timelike killing vector field and a stationary causal horizon .
this solution is associated with a thermal state at uniform temperature .
a general thermal state out of equilibrium is described by an inhomogeneous black hole , which is not an exact solution . however , when a hydrodynamic limit exists , we can work in a derivative expansion , assuming that there is no singularity at the horizon . in @xcite
we applied this approach to the event horizon of a boosted einstein - yang - mills black brane in an asymptotically ads spacetime . at lowest orders in derivatives ,
the set of einstein and maxwell equations projected into the horizon surface describe the hydrodynamics of a cft with anomalous global non - abelian charges . here
we will briefly describe the formalism used in @xcite and show how it can be applied to the simplest case of an uncharged black brane @xcite .
in the bulk spacetime , we choose coordinates of the form @xmath2 .
the @xmath3 are coordinates on the horizon @xmath4 .
@xmath5 is a transverse coordinate , with @xmath6 on the horizon .
@xmath7 is a null covector tangent to the @xmath4 .
when raised with the metric , it gives a vector field @xmath8 which is both normal and tangent to the horizon , and tangent to its null generators . in components
, we have @xmath9 .
the pullback of @xmath10 into @xmath4 is the degenerate horizon metric @xmath11 .
its null directions are the generating light - rays of @xmath4 , i.e. @xmath12 . the lie derivative of @xmath11 along @xmath13 gives us the shear / expansion tensor , or `` second fundamental form '' : @xmath14 we can write a decomposition of @xmath15 into a shear tensor @xmath16 and an expansion coefficient @xmath17 : @xmath18 we found it is convenient raise indices with @xmath19 , which is the inverse of any matrix @xmath20 of the form @xmath21 here we introduced the superfluous scalar field @xmath22 for later convenience : it will turn out that a matrix of the form coincides at leading order with the metric @xmath23 associated with the hydrodynamic dual .
since @xmath11 is degenerate , one can not use it to define an intrinsic connection on the null horizon , as could be done for spacelike or timelike hypersurfaces .
the bulk spacetime s connection does induce a notion of parallel transport in @xmath4 , but only along its null generators .
this structure is not fully captured by @xmath11 ; instead , it is encoded by the extrinsic curvature , or
weingarten map @xmath24 , which is the horizon restriction of @xmath25 : @xmath26 one can show that given an arbitrary @xmath20 of the form , @xmath24 can be written as : @xmath27 @xmath28 is the surface gravity " , which measures the non - affinity of @xmath13 .
the covector @xmath29 encodes the degrees of freedom in @xmath24 which are independent of @xmath11 . in the hydrodynamics
, these degrees of freedom will roughly correspond to the velocity and temperature fields . in their usual form ,
the gauss - codazzi equations involve the divergence of a hypersurface s extrinsic curvature .
our null horizon , however , does not possess an intrinsic connection , and a covariant divergence of @xmath24 can not be defined .
the solution is to define a tensor _ density _ constructed out of @xmath24 in the following manner @xcite : @xmath30 where @xmath31 is a scalar density equal to the horizon area density .
it turns out that the horizon - intrinsic covariant divergence @xmath32 is uniquely defined . using this divergence ,
the null gauss - codazzi equation can be written as @xmath33 } - v\partial_\mu\theta , \label{eq : gc_basic}\end{aligned}\ ] ] where we have defined the area entropy current @xmath34 .
this is the form we we will use for calculations .
the vacuum einstein equations with negative cosmological constant can be expressed as @xmath35 these equations have the homogeneous ( boosted ) black brane solution @xmath36 where @xmath37 .
@xmath13 is normalized @xmath38 and @xmath39 .
the horizon quantities defined in the above section can be expressed in terms of the bekenstein - hawking entropy density @xmath40 and the hawking temperature @xmath41 @xmath42 where the superscript zero anticipates inhomogeneous corrections .
consider again the configuration , but with @xmath43 slowly varying functions of @xmath3 rather than constants .
we will interpret this @xmath3-dependence by treating these quantities as fields on the brane s horizon @xmath4 .
various quantities and equations can be expanded order by order in powers of the small @xmath44 derivatives .
we use the symbolic small parameter @xmath45 to count these powers .
we will refer to the power of @xmath45 involved as the `` order '' of a quantity or an equation .
in general , with @xmath3-dependent parameters will not be a solution of the field equations .
our analysis will take place on the event horizon @xmath4 of the exact inhomogeneous solution . in the configuration with inhomogeneous @xmath43 ,
the hypersurface @xmath46 is null , and does nt intersect the singularity .
it is therefore an event horizon for the zeroth - order solution , and a zeroth - order approximation for the horizon of the corresponding exact solution . for the corrected solution
, we will still use coordinates so that @xmath6 on @xmath4 .
the gauss - codazzi equation reads at first order , @xmath47 } = o(\varepsilon^2 ) .
\label{eq : gc_1}\end{aligned}\ ] ] plugging in the zeroth - order values , and using the thermodynamic identities @xmath48 and @xmath49 , we rearrange the equation to get the conservation of a perfect fluid stress tensor @xmath50 for the black brane one can show that @xmath51 , so the fluid is conformal , as expected .
now we want to consider the gauss - codazzi equations at second order .
obviously , before we write down these equations , we must consider the possible first - order corrections to the horizon ansatz .
first of all , at viscous order in hydrodynamics there is naturally an ambiguity in definition of the fluid four - velocity .
we make the ( somewhat non - standard , but here natural ) choice that the four - velocity is the direction of the entropy velocity ; that is we fix the direction of the entropy velocity @xmath13 and the magnitude of the entropy density @xmath52 by requiring that they have the values in eq . without corrections .
due to this condition , the correction to @xmath11 must be transverse to @xmath13 and traceless with respect to @xmath53 .
we will find that the precise form of @xmath54 is otherwise irrelevant to the constraint equations .
let us now turn to the @xmath29 component of the extrinsic curvature . for our uncharged black brane , the most general correction to @xmath29 reads @xmath55 where @xmath56 and @xmath57 are functions of @xmath52 .
the requirement for the constraint equations to take the form of conservation laws will place restrictions on these functions . to start , we note that the horizon s first - order shear / expansion tensor can be derived directly from the zeroth - order metric via . using the ideal equations
, one finds that @xmath58 and @xmath59 . where @xmath60 is the fluid shear tensor @xmath61 .
let us now evaluate the second order pieces of the gauss - codazzi equation . plugging in , and using the ideal order hydrodynamics equations we find @xmath62 the first two terms can contribute to a relativistic conservation law equation .
however , the last does not and therefore we must set @xmath63 . as a result , at second - order ,
the gauss - codazzi equations have the form @xmath64 which are the relativistic hydrodynamics equations for a conformal fluid with shear viscosity @xmath65 .
it is remarkable that in our analysis the dynamics of the null horizon does not encode an arbitrary hydrodynamic system : once the choices discussed above are made , the transport coefficients are uniquely determined . specifically , the shear viscosity , the bulk viscosity ( and in general , other quantities like conductivity , see @xcite ) are fixed by the null hypersurface equations to specific functions of state , while for a general fluid they may be arbitrary ( under the restriction of positive semi - definiteness ) .
note however that in relation to membrane paradigm picture of @xcite , our construction can not capture all the dynamics of the theory at higher orders in the hydrodynamic @xmath45 expansion . nevertheless , to viscous order
we are able to obtain the hydrodynamics of the boundary fluid by working with the equations for the null horizon .
the reason is that we parametrized the membrane equations in terms of the thermodynamic variables associated with the boundary fluid and required the metric @xmath66 to match the boundary metric ( in this case @xmath53 ) . in principle
, one can also write the equations in terms of variables associated with a generic @xmath67 hypersurface , thereby obtaining the hydrodynamics of those surfaces as described in @xcite .
99 j. m. maldacena , adv .
phys . * 2 * , 231 ( 1998 ) [ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] ; o. aharony , s. s. gubser , j. m. maldacena , h. ooguri and y. oz , [ arxiv : hep - th/9905111 ] . c. eling , y. neiman and y. oz , arxiv:1010.1290 [ hep - th ] . c. eling and y. oz , jhep * 1002 * , 069 ( 2010 ) [ arxiv:0906.4999 [ hep - th ] ] ; c. eling , i. fouxon , y. oz , phys .
lett . * b680 * , 496 - 499 ( 2009 ) .
[ arxiv:0905.3638 [ hep - th ] ] .
j. jezierski , j. kijowski and e. czuchry , phys .
d * 65 * , 064036 ( 2002 ) [ arxiv : gr - qc/0110018 ] .
i. bredberg , c. keeler , v. lysov and a. strominger , arxiv:1006.1902 [ hep - th ] ; t. faulkner , h. liu , m. rangamani , [ arxiv:1010.4036 [ hep - th ] ] . | we discuss recent work showing that in certain cases the membrane paradigm equations governing the dynamics of black hole horizons can be recast as relativistic conservation law equations . in the context of gauge / gravity dualities ,
these equations are interpreted as defining the viscous hydrodynamics of a holographically dual relativistic field theory . using this approach
, one can derive the viscous transport coefficients and the form of the entropy current for field theories dual to gravity plus matter fields . |
over the past several years , low dimensional spin systems , belonging to family of cuprates@xcite , vanadates@xcite , titanates@xcite have emerged as a field of active research in the condensed matter community . in this context , the magnetic properties of the layered vanadium compounds @xmath6 ( @xmath7 ) has drawn interest as they are found to be quasi two dimensional ( 2d ) in nature with in - plane spin - frustration that corresponds well to a @xmath1 heisenberg antiferromagnetic ( af ) model@xcite .
early transition metal vanadium ( @xmath83d^34s^2 $ ] ) is in its nominal 4 + valences in the compounds giving rise to the possibility of realizing the spin 1/2 2d af .
the compound @xmath2 , in particular , has attracted attention due to interesting and unusual magnetic properties@xcite , showing weak superexchange ( @xmath9 mev ) mediated low dimensional magnetic behavior of v spins . susceptibility and specific heat measurements@xcite and neutron diffraction results@xcite
have confirmed the quasi 2d spin-@xmath0 af nature for the spin lattice of the system . in spite of existence of a 3d neel temperature of @xmath10
k , the compound shows signatures of short range 2d magnetic correlations well up to 6.95 k@xcite .
the broad peaks observed in magnetic - diffraction pattern of the compound , arising from the diffuse scattering at low temperatures ( @xmath11 ) have been attributed partly to spin waves ( for @xmath12 ) and partly ( for @xmath13 ) to short range 2d magnetic correlations . in this paper
, we represent a spin - wave analysis of the problem at low temperature taking into account the spin fluctuations within the tyablikov s decoupling approximation ( tda)@xcite on top of the mean - field linear spin wave ( lsw ) results and study the temperature dependence of the magnetic properties such as magnon dispersion , magnetization and spin susceptibility .
our calculation at mean - field+tda level shows the spin waves to exist up to a temperature of @xmath14 k. a recent density functional theory ( dft ) based study@xcite reported partial substitution of vanadium ions by titanium ions in @xmath2 , resulting in interesting modification of the crystal structure especially for the ti concentration of @xmath15 .
the underlying spin lattice for such compound , @xmath3 , turns out to be quasi-1d af with weak couplings in both the transverse directions . in our present study
we also carry out spin wave analysis for this ti - substituted compound and present a comparative study for the spin - wave modes obtained for the pristine and substituted compounds .
our calculation on the substituted compound produces three different magnon modes .
the content of our paper is arranged as follows . in section 2
we discuss the general formulation for the non - linear spin wave analysis of a quasi-2d system with frustration . in section 3 ,
we report our numerical results for @xmath2 and compare with the experimental findings .
section 4 is devoted to the compound @xmath16 which includes a linear and tda - level spin wave study of this quasi 1d spin 1/2 system .
section 5 discusses the comparison of magnon spectra of the two systems and section 6 gives the summary of our work .
@xmath2 has a tetragonal crystal structure of space group i4 cm with lattice parameters a = 8.93 @xmath17 and c = 9.04 @xmath17@xcite . the spin model for the compound
@xmath2 is a quasi-2d spin-1/2 heisenberg quantum antiferromagnet in a cubic lattice with magnetic @xmath18 ions sitting at the lattice sites .
the @xmath19 plane contains the prominent nearest neighbor ( nn ) antiferromagnetic coupling along with relatively weaker next nearest - neighbor ( nnn ) af coupling which are stacked along the @xmath20-direction coupled via weak ferromagnetic ( fm ) coupling@xcite ( fig.[fig1 ] , left panel ) . starting from this af reference state , we carry out a spin wave study for the system and examine the magnetization , susceptibility and the extent of neel ordering .
holstein - primakoff transformation is used to obtain the magnon modes in lsw approximation . in the next step ,
we extend our calculation beyond this approximation accounting for the additional terms of the hamiltonian that are quadratic in spin - wave operators . in order to do that we use decoupling approximation@xcite of the non - linear terms and
solve the equation of motion of the magnon green s function yielding the improved expression for the spin - wave modes .
.3 in the overall hamiltonian is , thus , given by @xmath21 with @xmath22 and @xmath23 .
@xmath24 ( @xmath25 ) and @xmath26 ( @xmath27 ) denote nn and nnn lattice positions in @xmath19 plane respectively whereas @xmath28 gives nn lattice positions along @xmath29 or @xmath20 directions . rewriting the hamiltonian in terms of spin wave operators ( about a reference state which is af in @xmath19 plane and fm along @xmath20 direction ) and retaining terms up to quadratic in bosonic spin wave operators
the fourier transformed hamiltonian is given by @xmath30 \label{eq2}\end{aligned}\ ] ] where @xmath31 are fourier transformed bosonic spin - wave operators@xcite , @xmath32 , @xmath33=(cos @xmath34+cos @xmath35)/2 , @xmath36=cos(@xmath34)cos(@xmath35 ) and @xmath37 is the vacuum or ground state energy .
the sum is over the first brillouin zone ( bz ) .
next we consider the non - linear correction to our results where additional terms that are quadratic in @xmath38 or quartic in bosonic spin wave operators are considered as well .
the overall correction becomes @xmath39+\frac{4j_2}{n}\gamma_{k - k^\prime}[a_{k}^\dagger a_{k^\prime}a_{q - k}^\dagger a_{q - k^\prime}+b_{k}^\dagger b_{k^\prime}b_{q - k}^\dagger b_{q - k^\prime}]\nonumber\\&-\frac{2j_2}{n}\gamma_{q+k^\prime}[a_k^\dagger a_{k^\prime}a_{k+q}a_{k^\prime+q}^\dagger + b_k^\dagger b_{k^\prime } b_{k+q}b_{k^\prime+q}^\dagger+h.c.]\nonumber\\&\frac{-2j_1}{n}[a_{k^\prime -q}^\dagger a_{k^\prime}a_{k}(\gamma_{k+q } b_{k+q}+\frac{\alpha~cos(k_{z}+q_{z})}{2}a_{k+q}^\dagger)\nonumber\\&+(\gamma_{k}a_{k}+\frac{\alpha~cos(k_{z})}{2}b_{k}^\dagger)b_{k^\prime}^\dagger b_{k+q}b_{k^\prime -q } + h.c . ]
\label{eq3}\end{aligned}\ ] ] where sum over @xmath40 and @xmath41 are implied . in order to find the spin wave dispersions and susceptibility , we first obtain the equation of motion@xcite of the bosonic green s functions , @xmath42 where @xmath43 denotes the fourier transform of the green s function involving operators a and b. fluctuation ( nonlinear ) terms
are averaged using tda and the resulting equations of motion become @xmath44 where @xmath45 $ ] , @xmath46 $ ] , @xmath47 $ ] and @xmath48 $ ] and @xmath49 , @xmath50 are the expectation values @xcite . from the solution of eq.[eq4 ]
we get@xcite @xmath51 where @xmath52 , being the pole of the green s function , is the spin wave dispersion which is given by @xmath53\sqrt{1-f(k)^2 } \label{eq6}\ ] ] @xmath54 and @xmath55 .
the spectral representation of correlation functions then leads to @xmath56\nonumber\\ & & < \xi_k>=-1/2f(k)a~coth(\beta\omega_k/2 ) \label{eq7}\end{aligned}\ ] ] which are actually thermal averages in a canonical ensemble ( expectation value ) .
+ .15 in the sublattice magnetization is given by @xmath57 .
\label{eq8}\ ] ] the k - vector summation runs over the af bz only .
calculations were carried out on a @xmath58 lattice , which is justified by the quasi 2d nature of the system .
we used the parameter values @xmath59 and @xmath60 to represent the compound @xmath2 .
the strength of the strongest coupling used is @xmath61@xcite .
fig.[fig2 ] shows the spin wave spectra with and without the non - linear correction .
the difference is found to be small and is maximum at the zone - boundaries . in fig.[fig3 ] , we show the calculated temperature dependences of magnon dispersion and sublattice magnetization . within the regime where lsw approximation is valid
, magnon energies do not alter much but the sublattice magnetization ( @xmath62 ) reduces rather sharply with temperature . @xmath63
curve in the bottom panel of fig.[fig3 ] points to a temperature around 4.31 k where the af order ceases to exist .
it indicates the upper cut - off beyond which a spin wave theory around the af reference state is not possible@xcite .
this mean field result falls between the experimental estimate of the neel temperature @xmath64 and the temperature @xmath65 up to which short range af correlation is observed in this compound@xcite .
the system size used for numerical computation is large enough to capture any short range af correlation and thus the difference of the calculated value of @xmath66=4.31 k from the experimental finding ( @xmath66=3.75 k@xcite ) can be due to the usual over - estimation that occurs in any mean - field calculation . in order to calculate the magnetic susceptibility , we need to consider the coupling to the external field @xmath67 and the term @xmath68 is added to the hamiltonian .
the resulting zero - field molar susceptibility , obtained using kubo s method of linear response function@xcite is given by @xmath69 where @xmath70 this gives the low temperature limit of the susceptibility while at high temperature , curie - weiss behavior is followed and the susceptibility is given by , @xmath71 where @xmath72 is the curie constant and the other parameters being@xcite @xmath73 , @xmath74x@xmath75 emu / mol , @xmath76 is avogadro s number and @xmath77 , the curie - weiss temperature ( see fig.[fig4 ] ) .
we should mention here that the susceptibility we calculate is actually the parallel susceptibility @xmath78 where the external magnetic field is along @xmath79 , the easy axis direction .
the susceptibility obtained experimentally from a powdered sample has contributions from perpendicular susceptibility @xmath80 as well ( @xmath81)@xcite and thus it takes non - zero value even at zero temperature@xcite .
spin wave calculation , being a mean field one , has little finite size effect as compared to more accurate quantum monte carlo ( qmc ) results as far as the low temperature behavior is concerned@xcite .
low temperature susceptibility behavior can thus be easily complimented by simple spin wave calculations without the expense of going to large size lattices .
the inset in fig.[fig4 ] shows the spin - spin correlation results obtained from combined spin wave calculations at low temperature and qmc results at high temperatures using qmc stochastic series expansion ( sse ) directed loop - update method , as used in ref.@xcite .
substitution of some of the @xmath5 ions of the compound @xmath2 with non - magnetic @xmath4 ions results in the new compound @xmath82 .
this newly formed compound shows interesting structural modification especially for @xmath83 substitution where one @xmath5 out of four @xmath5 ions in the square plaquette is substituted by @xmath4@xcite .
ab - initio calculation@xcite found the new structure to have a monoclinic a2 symmetry with lattice parameters a = 12.851 @xmath17 , b = 9.180 @xmath17 and c = 12.852 @xmath17 .
this computer - designed structure gives a spin model with interesting magnetic properties@xcite .
fig.[fig1 ] ( right panel ) schematically shows the underlying spin lattice containing 3 parallel layers of @xmath84 lattice planes of the substituted compound .
the lattice appears to be a repetition of two consecutive layers , only one of which has vacancies ( @xmath85 , non - magnetic @xmath4 ion ) in its alternate sites .
the black and red ( gray ) balls in fig.[fig1 ] represent the crystal positions of magnetic @xmath5 ions and the vacancies are due to the presence of non - magnetic @xmath4 ions .
overall the system can be viewed as a sum of six different superlattices yielding three different sets of magnon modes each with degeneracy 2 ( because of the up - down symmetry ) .
the strongest exchange interaction is found to be of af type and along one bond direction within the @xmath19 plane and it is much larger in value compared to the other intra - plane or inter - plane coupling@xcite .
the hamiltonian for the new system in a @xmath86 cubic lattice ( including both @xmath5 and @xmath4 ) is given by @xmath87 within the lsw approximation and fourier transformation , the hamiltonian transforms to @xmath88 \label{eq12}\end{aligned}\ ] ] where @xmath89 represents the fourier transformed bosonic spin - wave operators@xcite for the @xmath90-th superlattice , @xmath91 , @xmath92 , @xmath93 , @xmath94 , @xmath95 and @xmath37 is the vacuum or ground state energy .
the wave - vector sums are within the 1st brillouin zone corresponding to 3d unit cell ( 2a , 2b , 2c ) . with extended bogoliubov transformation@xcite , applicable for this six sublattice system , the spin wave dispersions @xmath52 are obtained as a solution of the characteristic determinant equation : det[@xmath96=0 , where @xmath97 represent coefficients or elements of the @xmath98 hamiltonian matrix in the basis of @xmath99 s . beyond the linear approximation , when quartic terms ( within tda ) of the hamiltonian are considered as well ,
the coefficients @xmath97 gets modified to @xmath100 with @xmath101 the rest of the coefficients being zero .
the correlation functions are obtained as @xmath102 while green s functions @xmath103 s and consecutively , coefficients @xmath100 are obtained self - consistently by solving the matrix equations .
small parameter @xmath104 is taken to be @xmath105 , where @xmath106 is the numerical resolution along the energy axis @xmath107 .
we find the spin wave modes by solving the modified characteristic equation .
this could have been obtained identifying the singularities of @xmath103 s as well .
+ .1 in + + fig.[fig6 ] shows the spin wave dispersions along @xmath34 directions for @xmath108 in a @xmath58 lattice .
though we figure out the magnetic behavior of the compound to be of a quasi-1d af , we keep the lattice size same as that of the pristine compound in order to compare the distinctive features between them . the different neighborhood of the three sublattices results in three different modes for this lattice .
weak coupling for the spins at sublattice e and f yields the magnon mode of small energy as compared to the energies of the other two modes .
further look at the magnon modes for the pristine and substituted compound make the quasi-2d and quasi-1d feature in them very evident .
fig.[fig7 ] shows the spin wave modes along different directions in k - space for both the compounds .
the pristine compound shows large - scale variation in magnon energies along @xmath109 and @xmath110 ( and not @xmath20 ) directions ( fig.[fig7 ] top panels ) .
but in the substituted compound , dispersion is visible mainly along @xmath109 direction ( fig.[fig7 ] , bottom panels ) .
the inelastic neutron scattering cross - section of a system can be measured from the the magnetic structure factor @xmath111@xcite which , within the approximation of one - magnon excitation ( creation or annihilation ) , can be written as @xmath112 here @xmath113 denotes @xmath114 and @xmath115 , the spin components transverse to the easy @xmath20 direction . @xmath116 represents the momentum transfer of the inelastic scattering and is related to the bz vector @xmath117 as @xmath118 , @xmath119 being a reciprocal lattice vector .
@xmath120 is the energy of the eigenstate @xmath121 and @xmath122 represents the magnetic form factor .
@xmath122 value is taken from an analytic expression ( obtained using dipole approximation ) with coefficients computed for the ion @xmath18@xcite .
fig.[fig8 ] shows the plot of the powder averaged structure factor @xmath111 vs. @xmath123 for our pristine ( top panel ) and substituted ( bottom panel ) compounds .
these structure factor plots can be compared with powder - neutron diffraction data .
for the pristine compound ( @xmath124 ) , the intensity plot shows high values for scattering cross - section spanned over almost the whole energy range at @xmath125 and @xmath126 .
they indeed correspond to magnetic bragg peak positions ( 100 ) and ( 120 ) respectively , as observed in neutron diffraction measurements of powder @xmath2@xcite . to make the comparison with experiment more quantitative , in fig.[fig9 ] we show the integrated powder averaged structure factor @xmath127 for the spin model of @xmath2 , which can be obtained from integrating @xmath111 ( fig[fig8 ] ) over @xmath128 .
it indicates peaks in the scattering cross - section at @xmath123 = 0.7 and 1.57 @xmath129 corresponding to the magnetic bragg peak positions ( 100 ) and ( 120 ) respectively .
same positions have been found by rietveld refined powder neutron diffraction ( pnd ) results on @xmath2 , as shown in fig.1c of ref.@xcite . there
the magnetic pattern at a low temperature of 1.5 k ( which is below the experimental @xmath10 k ) is obtained by subtracting the corresponding pnd pattern from the pnd pattern at a high temperature of 300 k ( which has only nuclear and no magnetic contributions ) . in the compound @xmath16 ,
three magnon modes give three terms in the expression for @xmath111 .
the low energy mode is almost dispersion - less compared to the other two almost degenerate modes and thus gives featureless large scattering cross - section at low energies ( fig.[fig8 ] , bottom panel ) . on the other hand ,
the higher energy nearly degenerate two other magnon modes are responsible for the diffused magnetic scattering peaks at values of q which are smaller compared to that for the pristine compound .
this is because of the larger direct - space unit cell in the substituted compound as compared to the pristine one .
the overall @xmath111 distribution of the substituted compound , unlike the pristine one , lacks sharpness in peaks indicating strengthening of the quantum fluctuation effect due to the reduced dimensionality of the spin hamiltonian .
to summarize , we have presented a linear and non - linear spin wave study for a quasi-2d af with parameters representing the compound @xmath2 .
the damping of the spin wave modes with temperature is observed and a decay in sublattice magnetization leading to a @xmath130 is witnessed .
this temperature is high compared to the experimentally obtained @xmath64 .
this difference can be a result of the approximation used in our calculation and can be improved using 1/n expansion of the nonlinear-@xmath131 model as shown for quasi-2d antiferromagnet by irkhin et al.@xcite .
the parallel susceptibility at zero field is obtained at low temperature which can well complement qmc data that shows enough finite size effect at low t. further a spin wave study for the @xmath4-substituted compound @xmath16 shows the multiple magnon modes obtained as a result of structural modification .
the six superlattice structure creates small size bz in the system which is also evident from the structure factor plot ( fig.[fig8 ] ) .
the magnon dispersions also confirm the quasi-1d nature of the underlying spin lattice .
we should mention here that our computer - designed compound @xmath16 has an ordered sublattice structure between v and ti .
our model , therefore , represents an idealistic situation where any possible interchange of the ti and v ion occupied sites , that may occur in actual synthesis of such compound , is not considered .
the possibility of such inter - mixing , however , can not be ruled out . in future
, we would like to address this aspect in our theoretical study .
s. kar thanks dr . anup bera for fruitful discussion on neutron powder diffraction calculation . t.s.d . would like to acknowledge support from dst .
99 h. das @xmath132 , phys . rev .
b 77 , 224437 ( 2008 ) .
r. valenti @xmath132 , phys . rev .
b 68 , 024411 ( 2003 ) .
t. saha - dasgupta @xmath132 , phys . rev .
b 71 , 153108 ( 2005 ) .
t. thalmeier @xmath132 , acta physica polinica a. 115 , 56 ( 2009 ) .
k. h. lii @xmath132 , j. solid state chem .
90 , 291 ( 1991 ) .
n. s. kini @xmath132 , j. phys .
cond . mat .
18,1303 ( 2006 ) . s. m. yusuf @xmath132 , phys .
b 82,094412 ( 2010 ) .
s. v. tyablikov , ukr . mat .
11,287 ( 1959 ) . m. e. lines , phys .
135,a1336 ( 1964 ) .
j. d. patterson et al , am . j. phys . 36,343 ( 1968 ) .
r. a. t - kheli , d. ter haar , phys .
127,88 ( 1962 ) . s. kanungo @xmath132 , phys .
b 87 , 054431 ( 2013 ) .
h. b. callen , phys .
130,890 ( 1963 ) .
a. pratap et al , phys .
( b ) 197 , 453 ( 1996 ) .
e. manousakis , rev .
phys . 63,1 ( 1991 ) .
s. kar @xmath132 , phys .
b 84,205107 ( 2011 ) .
r. kubo , j. phy . soc .
japan 12,570 ( 1957 ) .
k. h. hoglund @xmath132 , phys .
91 , 077204 ( 2003 ) .
j. m. rowson @xmath132 , j. phy . and chem . of solids 65,727 ( 2004 ) .
zhang zhi - dong , phys .
b 53,2569 ( 1995 )
. s. w. lovesay , `` theory of neutron scattering '' , clarendon press , oxford ( 1984 ) .
y. zhu , `` modern techniques for characterizing magnetic materials '' , springer ( 2005 ) ; http://www.neutron.ethz.ch/research/resources/ + formfactor . v. yu .
irkhin @xmath133 , phys . rev .
b 55,12318 ( 1997 ) . | in this study , we present non - linear spin wave analysis of a quasi 2d spin-@xmath0 @xmath1 antiferromagnet at the parameter regime relevant for the recently studied compound @xmath2 .
we obtain the temperature dependence of the spin wave energy , susceptibility and magnetization using green s function technique and hartree - fock factorization or tyablikov s decoupling approximation .
the comparison of our numerical results with the experimental findings is discussed . magnetic structure factor is calculated and compared with powder neutron diffraction data .
we also study the spin wave behavior of the compound @xmath3 obtained by partial chemical substitution of @xmath4 at @xmath5 sites of the compound @xmath2 ( phys .
rev .
b87 , 054431 ) . due to the superlattice structure of the spin lattice ,
the substituted compound possess multiple spin wave modes .
the spin wave analysis confirms the quasi-1d nature of the substituted system . |
the history of star formation in dusty galaxies was recently discussed by blain et al .
( 1999c ) , who assumed that the distant galaxies recently detected using the 450/850-@xmath0 m submillimetre common - user bolometer array ( scuba ) camera ( holland et al .
1999 ) were the high - redshift counterparts of local ultraluminous _ iras _ galaxies .
the global star formation rate ( sfr ) in dust obscured galaxies was inferred to be significantly greater than that of optically selected high - redshift galaxies ( steidel et al .
1996a , b , 1999 ) , subject to the uncertain fraction of the luminosity of the submillimetre - selected samples of galaxies ( smail , ivison & blain 1997 ; barger et al .
1998 ; hughes et al . 1998 ; barger et al . 1999a ; blain et al .
1999b ; eales et al .
1999 ) that is produced by accretion processes in active galactic nuclei ( agn ) . a fraction of at most 30per cent , and more likely 1020per cent , is suggested by both follow - up observations ( frayer et al .
1998 ; ivison et al . 1998
; smail et al .
1998 ; barger et al . 1999b ; frayer et al . 1999 ; lilly et al . 1999 ) , and information derived in other wavebands ; see section5.4 of blain et al .
( 1999c ) , almaini , lawrence & boyle ( 1999 ) and gunn & shanks ( 1999 ) . using a different approach , in which the high - redshift scuba population is decoupled from the local infrared - luminous galaxies , trentham ,
blain & goldader ( 1999 ) were able to reconcile the scuba counts with a less dramatic amount of obscured star - formation activity .
use another empirical approach , tan , silk & balland ( 1999 ) derived results somewhere between the two .
a summary of the existing data on the history of star formation is presented in fig.1 .
although well constrained , and in accord with the available observational data , the models in blain et al .
( 1999c ) and trentham et al .
( 1999 ) included few details of the physical origin of the large luminosity of scuba galaxies .
semi - analytic models of hierarchical galaxy formation , in which galaxies assemble by the merger of progressively larger subunits ( cole et al .
1994 ; baugh et al .
1998 ; kauffmann & charlot 1998 ; somerville , primack & faber 1999 ) have been used to account for a wide range of observations in the optical and near - infrared wavebands , and have been extended into the far - infrared and submillimetre wavebands by guiderdoni et al .
these models involve a large number of free parameters , and the interplay between them can make it difficult to identify the most important physics responsible for a particular observation . in this paper
we develop a model of infrared - luminous galaxies in a simple version of such a scenario ( blain & longair 1993a , b ; jameson , longair & blain 1999 ) , which includes many fewer parameters and hopefully makes the astrophysics more transparent .
we attempt to reproduce the scuba counts by invoking bright dust - enshrouded bursts of either star formation activity or agn fuelling at the epochs of mergers .
motivation for considering the scuba galaxies as luminous mergers is provided by both the optical identifications of the smail et al .
( 1998 ) sample , which appear to contain a large fraction of interacting galaxies , and the gas consumption rate that is inferred from observations of co emission of two submillimetre - selected galaxies made using the owens valley millimeter array ( frayer et al .
1998 , 1999 ) , which can not be sustained for more than a few @xmath3yr .
even the faint and compact counterparts listed in smail et al .
( 1998 ) could be merging galaxies , but too faint to identify as such ; see the simulations of the appearance of high - redshift mergers in bekki , shioya & tanaka ( 1999 ) . if the scuba galaxies are the high - redshift counterparts of the low - redshift ultraluminous infrared galaxies , which are predominantly merging systems , then this also offers support for modeling the scuba galaxies as mergers . using our simple model , we emphasise the most important features and the underlying physics of the evolution of submillimetre - selected galaxies and their relationship to the population of quiescent galaxies
in section2 we describe the details of the model , and investigate the constraints imposed by the intensity of the far - infrared and submillimetre - wave background radiation and the counts of low - redshift _ iras _ galaxies .
we discuss the evolution of the luminosity density in the model and compare the models with observations in the same way as the evolving _ iras _
luminosity function models discussed by blain et al .
( 1999c ) . in section3
we discuss the predictions in the context of source counts in the submillimetre and far - infrared wavebands , and investigate whether the scuba galaxies can easily be explained in an hierarchical picture of galaxy formation and evolution . in section 4
the corresponding background radiation intensities and galaxy counts in the near - infrared and optical wavebands are discussed . in section 5
we review the parameters we have introduced to describe the models .
we present our main conclusions in section 6 .
a value of hubble s constant @xmath4kms@xmath5mpc@xmath5 , with @xmath6 , a density parameter @xmath7 and a cosmological constant @xmath8 are assumed .
the evolution of galaxy - scale structures under gravity according to hierarchical clustering models can be analysed using the press
schechter formalism ( press & schechter 1974 ) , which describes the time - dependent mass spectrum of bound objects . the analytic results of the press schechter formalism are in quite acceptable agreement with those of n - body simulations ( brainerd & villumsen 1992 ) .
the formalism can be extended to yield a very straightforward semi - analytic merger rate , under the single assumption that the process of halo mergers is independent of mass ( blain & longair 1993a , b ) . according to the press
schechter prescription , the mass spectrum of bound objects with masses between @xmath9 and @xmath10 is @xmath11,\ ] ] in which @xmath12 is the smoothed comoving density of the universe , dominated by dark matter , @xmath13 , where @xmath14 is the power - law index of primordial density fluctuations , and @xmath15 is a parameter which describes the evolution of density fluctuations as a function of redshift @xmath1 : @xmath16^{2/\gamma}.\ ] ] @xmath17 is the function describing the growth of perturbations in a general cosmology , and is derived from the equation , @xmath18 in which @xmath19 is the scalefactor of the universe . in an einstein de - sitter model , the growing mode has @xmath20 .
@xmath21 is the typical mass of bound objects at @xmath22 .
inhomogeneities do not grow if @xmath23 , that is if @xmath24 . for scale - invariant density fluctuations , @xmath25 or @xmath26 .
this is close to the value observed on the largest scales from the cosmic microwave background radiation ( cmbr ) .
observations of large - scale structure indicate that @xmath27 , or @xmath28 , on the smallest scales ( peacock & dodds 1994 ) , which can be associated with the transfer function between a primordial @xmath25 spectrum and the spectrum after recombination . using the tully fisher relation ( hudson et al .
1998 ) , blain , mller & maller ( 1999 ) obtained a value @xmath29m@xmath30 .
the exact value of @xmath21 is not very important here , as a mass - to - light ratio is introduced to convert the mass spectrum into a luminosity function .
working from the mass spectrum @xmath31 , blain & longair ( 1993a , b ) showed that a formation rate of bound objects @xmath32 in galaxy halo mergers can be constructed if the mass distribution of the components involved in a statistical sample of merger events is assumed to be independent of mass . in this case
, the merger rate can be represented accurately by the function @xmath33,\ ] ] where @xmath34.\ ] ] @xmath35 and @xmath36 are numerical constants , typically about 1.7 and 1.4 ; their exact values depend on the assumed mass distribution of merging components ( blain & longair 1993a ) , but have little effect on the results .
the values of both @xmath35 and @xmath36 are weak functions of @xmath37 and depend on the world model parameters , but the form @xmath38 that appears in the calculations of the background radiation intensity and metal abundance is almost independent of the value of @xmath37 .
the merger rate as a function of mass @xmath39 can be readily used to estimate a number of observable quantities , starting with the luminosity density ( or volume emissivity ) , @xmath40 in which @xmath41 is the ratio of the mass of baryonic matter converted into metals by nucleosynthesis in a merger - induced starburst to the total dark mass involved in the merger .
the rationale behind this form of relation is given by longair ( 1998 ) .
the factor of 0.007 is the approximate efficiency of conversion of mass into energy in stellar nucleosynthesis .
the parameter @xmath42 describes the fraction of the total luminosity of merging galaxies that is attributable to accretion in agn , and is expected to lie in the range @xmath43 ( genzel et al .
1998 ; lutz et al .
1998 ; almaini et al . 1999 ; barger et al . 1999b
; gunn & shanks 1999 ) .
the parameter @xmath41 is expected to vary with redshift @xmath1 .
blain & longair ( 1993b ) predicted a flat background spectrum in the submillimetre and far - infrared wavebands , assuming a constant value of @xmath44 .
subsequent observations ( e.g. fixsen et al . 1998 ) demand a redshift - dependent form of @xmath41 , as discussed by blain et al .
( 1999c ) . by evaluating the integral in equation ( 6 ) ,
the luminosity density can be expressed as @xmath45 interestingly , @xmath46 and so , because the density contrast @xmath17 is not a function of the perturbation spectral index @xmath37 , the @xmath37 dependence in this term is just a simple scaling .
thus the effect of the value of @xmath37 on the background radiation intensity can be studied or removed very easily . in an einstein de sitter model the luminosity density @xmath47 .
the comoving density of metals produced in starbursts between a redshift @xmath48 , at which star formation activity begins , and @xmath1 , is @xmath49 where @xmath50 is the radial comoving distance coordinate .
note that this result depends on the merger efficiency parameter @xmath44 but not on the agn fraction @xmath51 , as metals are only generated in merger - induced starbursts and not in agn fuelling events .
similarly , the background radiation intensity per unit solid angle emitted by these galaxies , which have a spectral energy distribution ( sed ) @xmath52 , is @xmath53 if the form of @xmath54 ( equation 7 ) is included explicitly , then @xmath55 details of the assumed dust sed can be found in blain et al .
( 1999c ) .
the mid - infrared sed is assumed to take the form @xmath56 .
none of the quantities listed above are affected by the time dependence of the release of energy during merging events ; however , the source count requires the time profile of the merger induced starburst / agn to be included . for simplicity
, this profile is assumed to have a top - hat form with duration @xmath57 .
the time profile of the luminosity generated in a detailed simulation of the merger of gas - rich galaxies is discussed by mihos & hernquist ( 1996 ) , bekki et al .
( 1999 ) and mihos ( 1999 ) .
the typical duration of agn fuelling events and starbursts may differ ; for example , a lower limit to the duration of a starburst is set by the lifetime of the highest mass stars , but there is no lower limit to the duration of an agn fuelling event .
however , to avoid introducing an unnecessarily complicated model , the time - scale of a merger induced luminous phase is assumed to be independent of its origin .
in addition , because not all the mergers of dark matter haloes that take place at each epoch need induce a starburst / agn , a fraction @xmath58 is assumed . again , this fraction could differ for starbursts and agn , but for simplicity it is assumed not to .
the luminosity of a typical merger induced starburst / agn of mass @xmath9 is thus @xmath59 the source count @xmath60 of galaxies per unit solid angle brighter than a flux density @xmath61 is @xmath62 where @xmath63 is the comoving distance parameter .
the minimum mass merger visible at a flux density @xmath61 and redshift @xmath1 is @xmath64 the time - scale and bursting fraction parameters , @xmath57 and @xmath65 , always appear together in calculations , and thus the single parameter @xmath66 is constrained by observations .
we define @xmath67 to be the activity parameter , which is large in violent starbursts / agn and free to vary as a function of redshift .
the rate of energy release within each individual starburst / agn is controlled by the value of the activity parameter . within a representative cosmological volume
the presence of a population of either rare long - lived or common short - lived starbursts / agn can not be distinguished .
this is why the time - scale and bursting fraction parameters @xmath57 and @xmath65 are bound together in the activity parameter .
incorporating redshift evolution of the activity parameter introduces another degree of freedom into the count model , in addition to that provided by the redshift evolution of the star formation / agn fuelling efficiency parameter @xmath44 . of course ,
@xmath44 , @xmath65 and the time - scale @xmath57 are also free to vary as a function of mass . at present , we find no compelling reason to incorporate this additional complication into the model .
\(a ) 53.5 mm ( b ) 53.5 mm ( c ) -2 mm the background radiation intensities and source counts calculated from the equations derived above depend on a range of parameters : the world model , defined by @xmath68 , @xmath69 and @xmath70 ; the perturbation spectral index @xmath14 and the value of @xmath21 ; the constants @xmath35 and @xmath36 in the merger rate ; the merger efficiency @xmath41 ; the agn fraction @xmath51 ; the fraction of mergers that lead to a starburst / agn @xmath65 ; their duration @xmath57 ; and their sed @xmath52 . 0.75 mm @xmath71 blain et al .
( 1999c ) used the low - redshift 60-@xmath0 m _ iras _ source count and the 175-@xmath0 m _ iso _ counts to constrain their models ; here , however , we use the bright 60-@xmath0 m counts and the form of the far - infrared / submillimetre background spectrum , the two best determined observables , to constrain the parameters that describe the merger efficiency @xmath41 . if the sed does not depend on the mass of the merging galaxies , then the background spectrum ( equation 10 ) is determined entirely by the form of the merger efficiency @xmath41 , @xmath72 note that the dependence of @xmath73 on the perturbation index @xmath14 through @xmath37 is completely separate from the dependence on the world model .
thus the background spectrum determined by puget et al .
( 1996 ) , guiderdoni et al .
( 1997 ) , dwek et al .
( 1998 ) , fixsen et al .
( 1998 ) , hauser et al .
( 1998 ) , schlegel , finkbeiner & davis ( 1998 ) and lagache et al .
( 1999 ) can always be used to constrain the form of the merger efficiency @xmath41 .
we adopt a form of @xmath41 identical to the ` peak model ' described in blain et al .
( 1999c ) : . ]
@xmath74^{-1 } ( 1+z)^{p+(2 z_{\rm max})^{-1}}. } \ ] ] this is not a uniquely appropriate functional form of @xmath41 .
it was originally chosen to allow the star - formation history derived by madau et al .
( 1996 ) to be fitted .
its three parameters can be manipulated to produce a wide range of plausible star formation histories .
the three parameters are : @xmath75 , the asymptotic low - redshift slope of the merger efficiency @xmath41 in @xmath76 ; @xmath77 , the redshift above which the high - redshift exponential cut - off starts to take effect ; and @xmath78 the value of @xmath79 .
the epoch of most intense star - formation / agn - fuelling corresponds to a redshift @xmath80 in these models . in figs2(a ) and ( b ) the probabilities of fitting the background radiation spectrum and the slope of the 60-@xmath0 m counts predicted from the merger efficiency @xmath41 , defined in equation ( 16 ) , to observations are shown as a function of the key parameters @xmath75 and @xmath77 , as an example for a dust temperature of 45k . in fig.2(c ) the joint probability of fitting both sets of data is shown .
note that a constant dust temperature is assumed .
the value of @xmath51 does not affect the results .
if different dust temperatures are assumed , then the position of maximum probability moves around the @xmath75@xmath77 plane .
however , when the form of the evolution of luminosity density is calculated for each temperature , the curve has a similar form .
the best - fitting values of @xmath75 and @xmath77 , and the corresponding values of the merger efficiency @xmath78 , the activity parameter @xmath81 and the density parameter in metals at @xmath22 , @xmath82 , are presented in table1 for four plausible values of the dust temperature : 35 , 40 , 45 and 50k .
note that the best fitting values of @xmath75 and @xmath77 depend only weakly on the world model assumed .
the physical processes that demand a form of luminosity density which rises steeply with increasing redshift before turning over have not been considered here in any detail .
it seems likely , however , that the steep decline in the star formation rate to the present day is related to the declining gas content of galaxies at @xmath2 , and that the behaviour at high redshifts could be attributable to relatively inefficient cooling of gas and thus of star formation in the mergers of metal - poor high - redshift systems ( see pei & fall 1995 and pei , fall & hauser 1999 for discussions of gas and dust evolution in the universe ) .
we discuss these issues further in jameson , blain & longair ( 2000 ) .
the bright low - redshift 60-@xmath0 m count , @xmath83 is independent of the cosmological model . at @xmath84jy , @xmath85sr@xmath5 ( saunders et al .
a value of @xmath86gyr@xmath87 provides a good fit for any dust temperature between 30 and 60k .
the values of the normalization of the luminosity density @xmath88 and the activity parameter @xmath81 at @xmath89 that are required to fit the background spectrum and 60-@xmath0 m counts depend on the fluctuation index @xmath37 as @xmath90 and @xmath37 respectively .
thus the mass - to - light ratio of merging galaxies ( equation 12 ) is expected to be independent of the value of the perturbation index .
the values of @xmath66 listed in table1 are lower limits to the time - scale of the starburst / agn @xmath57 , as @xmath58 .
they are generally consistent with the starburst time - scales of order @xmath91yr derived by mihos & hernquist ( 1996 ) from hydrodynamical simulations of galaxy mergers .
the background radiation spectra and 60-@xmath0 m counts corresponding to the models listed in table1 are shown in figs3 and 4(a ) respectively .
note that the inclusion of positive evolution in the efficiency parameter @xmath41 overcomes the underprediction of the 60-@xmath0 m counts in the constant-@xmath44 model of blain & longair ( 1996 ) ; see their fig.6 .
the slope of the 60-@xmath0 m counts in the new model also more adequately represents the data than the predictions of guiderdoni et al .
the luminosity function of nearby dusty galaxies at 60-@xmath0 m ( saunders et al .
1990 ; soifer & neugebauer 1991 ) predicted in the 40-k model , which is calculated at @xmath22 by evaluating @xmath92 at the mass corresponding to a luminosity @xmath93 in equation ( 12 ) , is shown in fig.4(b ) .
the form of the luminosity function depends on the value of the perturbation index @xmath37 , even though the 60-@xmath0 m counts are independent of @xmath37 .
the best representation of the observed function is provided by a value of @xmath94 , or @xmath95 .
a similar high - luminosity slope could be achieved by modifying the form of equation ( 12 ) so that @xmath96 , where @xmath97 .
however , to match the observations with a scale - independent value of @xmath25 , a rather extreme value of @xmath98 is required . in the work that follows @xmath14
is assumed to take the value @xmath99 , similar to the value found for this range of masses by peacock & dodds ( 1994 ) . in this case , the faint - end slope of the low - redshift 60-@xmath0 m luminosity function is equivalent to a schechter function parameter @xmath100 ( schechter 1976 ) .
this is the same as the faint - end slope @xmath101 of the optically selected luminosity function derived by steidel et al .
( 1999 ) at @xmath102 , which describes the most numerous population of high - redshift galaxies that are actively forming stars . at @xmath103 ,
the luminosity function of the blue star - forming galaxies in the cfrs survey also has a faint end slope @xmath104 ( lilly et al .
steep faint - end slopes with indices of about @xmath105 are expected for the luminosity functions of dwarf , irregular and infrared - luminous galaxies , as discussed by hogg & phinney ( 1997 ) .
\(a ) ( b ) in section4 of blain et al .
( 1999c ) the self - consistency of models of dust - obscured galaxy formation was discussed .
we demanded that a sufficiently large mass of metals , and associated dust , was required to be generated by nucleosynthesis at each epoch to account for the far - infrared emission predicted by the model .
as the mass of dust required to generate a given far - infrared luminosity depends strongly on the dust temperature , this consistency condition is most easily expressed as a minimum dust temperature at each redshift . in the cases of the models listed in table1 , this lower limit to the dust temperature is presented in fig.5 , both with and without an assumed high - redshift population iii to generate dust .
if 2per cent of the total star formation activity takes place in a high - redshift population iii , then the self - consistency limits are always satisfied .
the details of the calculations can be found in blain et al .
( 1999c ) .
it is assumed that all the dust that is generated prior to a particular redshift still existed at that redshift and was available to absorb and reprocess the light from young hot stars and agn .
however , in an hierarchical model , only a small fraction of all the dust generated by any epoch is found in galaxies actively involved in a merger at that epoch ; this fraction increases from about 8 to 25per cent progressively from the 35-k to 50-k model listed in table1 .
thus , because most of the dust will be found in quiescent objects , the condition in fig.5 is less severe than it might be .
however , even if only 10per cent of all dust is involved in a luminous dust enshrouded merger , then the lower limit to the temperature increases only by about 50per cent , and so this consistency requirement is not difficult to satisfy . the same correction is required if 90per cent of the energy generated in merging galaxies is attributable to accretion onto agn , that is , if the agn fraction @xmath106 .
hence , the models are readily self - consistent if high - redshift population - iii stars exist to generate early metals . more sophisticated models of the coupled evolution of dust , gas and stars have been presented by pei & fall ( 1995 ) , eales & edmunds ( 1996 ) and pei et al .
( 1999 ) .
\(a ) ( b ) -2 mm at the present epoch the density of metals formed by nucleosynthesis in stars @xmath107 ( equation 9 ) is expected to be about @xmath108 in the models presented in table1 .
if solar metallicity , about 2.5per cent by mass ( savage & sembach 1996 ) , is typical of the universe as a whole , and the density parameter in baryons @xmath109 ( burles & tytler 1998 ) , then @xmath110 if @xmath6 .
thus all the models listed in table1 are consistent with this limit , even if the agn fraction @xmath111 and all the luminosity of merging galaxies is due to star formation activity that generates heavy elements .
the density parameter in the form of stars at the present epoch @xmath112 ( gnedin & ostriker 1992 ) .
observations of lyman-@xmath36 absorbers along the line of sight to distant quasars allow the evolution of the mass of neutral gas and the typical metallicity in the universe to be traced as a function of epoch ( storrie - lombardi , mcmahon & irwin 1996 ; pettini et al .
1997 ) . in fig.6(a ) the mass of material that has been processed into stars is derived as a function of epoch in each of the star - formation histories listed in table1 , assuming that the agn fraction @xmath111 and a salpeter initial mass function ( imf ) with a lower mass limit of 0.07m@xmath30 . in this case
about 6570per cent of all stars formed are still burning at the present epoch .
the values of @xmath113 predicted are thus about 3 times larger than the observed value , but are comparable with the values derived in our earlier models ( blain et al .
in order to account for this difference , either a lower mass limit to the imf of about 1m@xmath30 or a value of the agn fraction @xmath114 is required .
this high - mass imf would be compatible with the inferred lower limit to the imf of 3m@xmath30 required by zepf & silk ( 1996 ) to explain the mass - to - light ratios of elliptical galaxies , and by rieke et al .
( 1993 ) to interpret observations of m82 .
stars with masses less than 3m@xmath30 appear to be less numerous than expected from a salpeter imf in recent observations of the low - redshift starburst galaxy r136 ( nota et al .
goldader et al .
( 1997 ) report that the results of near - infrared spectroscopy of nearby _ iras _ galaxies with luminosities between 10@xmath115 and 10@xmath116l@xmath30 support a deficit of stars with masses less than 1m@xmath30 in these systems .
more details about variations in the high - redshift imf are discussed by larson ( 1998 ) .
metals appear to be overproduced by about a factor of 5 at redshifts of 2 and 3 in the hierarchical models , as shown in fig.6(b ) , but again these results can be reconciled with the observations if a significant fraction of the luminosity of dusty galaxies is being powered by accretion on to agn .
note , however , that the observations of metallicity could be biased against metal - rich regions of the universe , either because of their small physical size ( ferguson , gallagher & wyse 1998 ) or because of the complete obscuration of a fraction of background qsos ( fall & pei 1993 ) .
_ asca _ x - ray observations of significant enrichment in intracluster gas ( mushotzky & loewenstein 1997 ; gibson , loewenstein & mushotzky 1998 ) , could indicate that this is the case , at least in high - density environments .
the observed turn - over in the neutral gas fraction and the maximum rate of star formation shown in fig.6(a ) are approximately coincident in redshift , and the rate of enrichment in the hierarchical models is broadly consistent with the slope interpolated between the three highest redshift data points plotted in fig.6(b ) . in fig.7 the form of evolution of the luminosity density
is shown as a function of redshift in the models listed in table1 .
all the models predict curves with a rather similar form at @xmath117 .
the transformation between luminosity and star formation rate is the same as that assumed by blain et al .
( 1999c ) : that is , a sfr of 1m@xmath30yr@xmath5 is equivalent to a luminosity of @xmath118l@xmath30 . at low redshifts
the evolution of luminosity density is consistent with optical and near - infrared observations , and with the results presented in our earlier paper .
\(a ) ( b )
the models presented in table1 were constrained using the properties of the counts of the low - redshift 60-@xmath0 m _ iras _ galaxies .
the same formalism can be used to determine the counts of more distant dusty galaxies in the mid-/far - infrared and millimetre / submillimetre wavebands , where a large amount of additional information about the surface density of more distant dusty galaxies is available .
there is an upper limit to the surface density of sources at 2.8 mm ( wilner & wright 1997 ) ; counts at 850@xmath0 m ( smail et al .
1997 ; barger et al . 1998 ; holland et al . 1998 ; hughes et al . 1998 ; barger et al .
1999a ; blain et al .
1999b ; eales et al .
1999 ) ; upper limits ( smail et al . 1997 ; barger et al .
1998 ) , and a new count ( blain et al .
2000 ) at 450@xmath0 m ; 175-@xmath0 m _ iso _ counts from kawara et al .
( 1998 ) and puget et al .
( 1999 ) ; 95-@xmath0 m counts from kawara et al .
( 1998 ) ; and 7- and 15-@xmath0 m counts from an extremely deep _ iso _ image of abell2390 ( altieri et al . 1999 ) , which yields counts that are even deeper than those determined in blank - field surveys by oliver et al . (
1997 ) , aussel et al .
( 1999 ) and flores et al .
( 1999 ) .
if the values of the activity parameter at redshift zero , @xmath81 , listed in table1 are used to estimate the counts of galaxies at 850 and 175@xmath0 m , then the results underpredict the observed counts by a large factor .
the form of evolution of the merger efficiency @xmath41 is fixed by the observed background radiation intensity , and so , keeping within the framework of our well - constrained models , the value of the activity parameter @xmath67 at high redshift must be allowed to increase above its value at redshift zero in order to account for the observations .
this has the effect of increasing the luminosity of high - redshift mergers , thus increasing the 175- and 850-@xmath0 m counts .
however , the background radiation intensity and the low - redshift 60-@xmath0 m counts remain unchanged . the form of evolution of the activity parameter @xmath67 that is required to explain the data is illustrated in fig.8 . in fig.8(a )
the ratio of the model predictions and the observed counts at wavelengths of 175 and 850@xmath0 m ( kawara et al . 1998 ; blain et al .
1999b respectively ) are compared as a function of the activity parameter in the four models listed in table1 .
the same value of the activity parameter can not account for the observed counts at both wavelengths simultaneously , and the value required to explain the low - redshift 60-@xmath0 m counts is different from either .
the value of the activity parameter required to fit the 60- , 175- and 850-@xmath0 m counts increases monotonically . because the median redshift of the galaxies contributing to the counts at these redshifts is expected to increase monotonically , in fig.8(b ) we present the ratio of the model predictions and the observed counts as a function of a parameter @xmath119 that describes a simple form of exponential redshift evolution of the activity parameter ,
@xmath120 the exponential form provides a reasonable fit to the data , but is only one example of a whole family of potential functions .
the important feature is that the function chosen to represent the activity parameter @xmath67 increases rapidly with increasing redshift .
the zero - redshift value of the activity parameter @xmath121 is fixed by requiring that the low - redshift 60-@xmath0 m count prediction is in agreement with observations ; see table1 .
the values of the evolution parameter @xmath119 that correspond to the most reasonable fit for assumed single dust temperatures of 35 , 40 , 45 and 50k are about 1.5 , 1.5 , 2.0 and 2.3 respectively .
if the specific form of the redshift evolution of the activity parameter @xmath67 shown in equation ( 18 ) is assumed , then a dust temperature of 35 or 40k is most consistent with the data , the same temperature that was required for consistency by both blain et al .
( 1999c ) and trentham et al .
( 1999 ) , and is in agreement with the dust temperatures derived for high - redshift qsos by benford et al .
( 1999 ) .
the increase in the value of the activity parameter @xmath67 as a function of redshift can be interpreted in terms of two extreme scenarios , or as a combination of both . in the first scenario ,
the fraction of dark halo mergers that lead to a luminous phase in a dusty galaxy @xmath65 is fixed , but that the duration of the luminous phase @xmath57 is less at high redshifts .
this is plausible , based on the results of simulations of galaxy mergers ( e.g. mihos 1999 ; bekki et al .
1999 ) ; on average , the typical mass of a merging pair of galaxies is expected to be less at high redshifts in an hierarchical scenario of galaxy evolution , and the gas content of the galaxies is expected to be greater . as a result , the dynamical time of a merger
would be expected to decrease with increasing redshift , and the viscosity of the ism would be expected to increase .
both of these factors might be expected to increase the star formation efficiency of a merger with increasing redshift . in the second scenario ,
the duration of the luminous phase associated with a merger @xmath57 is independent of redshift , but the fraction of mergers that induce such a phase @xmath65 is reduced as redshift increases .
it is perhaps more plausible that the second of these scenarios could produce the large change in the activity parameter @xmath67 , by a factor of about 100 from @xmath22 to @xmath122 that is required to fit the data .
this is because the duration of the luminous phase of a merger - induced starburst @xmath57 must exceed the lifetime of a reasonably massive star , i.e. @xmath123yr . if star formation activity powers a significant fraction of the scuba galaxies , as seems reasonable , then this limit to the value of the merger duration @xmath57 is constrained to be greater than about 10@xmath124yr , only a few times less than the values of @xmath57 at @xmath22 listed in table1 .
thus it seems likely that a large fraction of the increase in the value of the activity parameter @xmath67 , which is required at high redshifts to account for the observed counts , should be attributed to a reduction in the fraction @xmath65 of dark halo mergers that generate a luminous galaxy .
we speculate that this may be connected with the lower typical metallicity expected at higher redshifts . in a lower metallicity system the cooling of dense gas
would be expected to be less efficient , and so a large amount of high - mass star formation may be unable to take place during the short merger process .
counts predicted by the four models listed in table1 , employing the values of @xmath119 listed above , are compared with observations at wavelengths of 15 , 60 , 175 , 450 , 850 , 1300 and 2800@xmath0 m in fig.9 .
while these models do not present unique solutions , fewer parameters are involved in the model than the number of separate pieces of constraining data . in future , by comparing the predictions of the models with observations , especially with the redshift distributions of the scuba galaxies ( barger et al .
1999b ; lilly et al .
1999 ; smail et al . 1999 , in preparation )
, the models can be developed to account more accurately for the increasing amount of available data .
\(a ) ( b ) -4 mm \(c ) ( d ) -4 mm there is a tight correlation between the flux densities of low - redshift galaxies in the radio and far - infrared wavebands ( see the review by condon 1992 ) .
thus the counts of faint galaxies observed in the radio waveband should not be overproduced when the seds of the galaxies in the 35- , 40- , 45- and 50-k models presented here are extended into the radio waveband using this correlation .
it is permissible to underpredict the counts , as there will be a contribution from agn to the faint counts , which need not be associated with powerful restframe far - infrared emission from dust .
partridge et al .
( 1997 ) report a 8.4-ghz galaxy count of @xmath125arcmin@xmath126 brighter than a flux density of 10@xmath0jy .
the correspondings count predicted by the 35- , 40- , 45- and 50-k models are 0.6 , 0.7 , 1.0 and 0.6arcmin@xmath126 respectively . thus all the models discussed here are consistent with the deep radio observations . for comparison , a count of 0.8arcmin@xmath126 is predicted by the modified gaussian model discussed in barger et al .
( 1999b ) , which is modified from the results of the simple luminosity evolution models presented by blain et al .
( 1999c ) .
the redshift distributions of submillimetre - selected sources at , or just below , the flux density limits of current surveys have been discussed by blain et al .
( 1999c ) in the context of models of a strongly evolving population of distant dusty galaxies , based on the low - redshift _ iras _ galaxy luminosity function .
the first spectroscopic observations of a large fraction of the potential optical counterparts to scuba galaxies identified in deep multicolour optical images ( smail et al .
1998 ) have been made by barger et al .
( 1999b ) ( see fig.10 ) .
this redshift distribution is consistent with the optical identifications made by lilly et al .
( 1999 ) in a scuba survey of canada france redshift survey ( cfrs ) fields .
the distribution shown in fig.10 is , however , subject to potential misidentifications of scuba galaxies .
for example , recent deep near - infrared images show that two of the smail et al .
( 1998 ) scuba galaxies , which were originally identified with low - redshift spiral galaxies , can more plausibly be associated with extremely red objects ( eros ) that were unidentified in optical images ( smail et al .
1999 ) . in two other cases , at @xmath127 ( ivison et al .
1999 ) and @xmath128 ( ivison et al .
1998 ) , the identifications have been confirmed by detections of redshifted co emission ( frayer et al . 1998 ; 1999 ) , and in another case spectroscopy and _ iso _ observations ( soucail et al .
1999 ) strongly support the identification of a ring galaxy at @xmath129 .
the preliminary redshift distribution of scuba galaxies , shown in fig.10 , is broadly consistent with the predictions of the gaussian model of blain et al .
( 1999c ) .
a modified gaussian model , as shown in fig.1 , was described by barger et al .
( 1999b ) ; the values of the evolution parameters in the modified gaussian model were explicitly fitted both to the background radiation intensity and count data and to the observed median redshift . in fig.10
the observed redshift distribution is compared with the redshift distributions predicted in the gaussian and modified gaussian models , and with the predictions of the hierarchical models developed here ( see table1 ) .
median redshifts of about 2.2 , 2.7 , 3.2 and 3.5 are expected in the 35- , 40- , 45- and 50-k hierarchical models respectively .
the redshift distributions predicted by the hierarchical models have median redshifts greater than that in the modified gaussian model , but less than those in either the other models presented by blain et al . ( 1999c ) or the hierarchical model e from guiderdoni et al .
( 1998 ) , all of which provide a reasonable fit to both the background radiation intensity and source counts in the far - infrared / submillimetre waveband .
based on these results , the coolest 35-k model seems to be in best agreement with the available data .
a model in which the single - temperature dust clouds discussed here are replaced by a temperature distribution will probably be required to account for the redshift distribution of the scuba galaxies . when the two spiral galaxies at @xmath130 are replaced by eros at @xmath131 ( smail et al .
1999 ) , the agreement between the 35-k hierarchical prediction and the observed redshift distribution is rather satisfactory . in all the hierarchical models , despite strong negative evolution of the mass - to - light ratio of mergers with increasing redshift , most of the detected galaxies are expected to lie at redshifts less than 5 , and so will be accessible to multi - waveband study using 8-m class telescopes .
when final reliable identifications and redshifts for submillimetre - selected galaxies are available , this information will be crucial for refining the hierarchical model .
the discussion has so far centred on the properties of merging galaxies as observed through their dust emission in the mid - infrared , far - infrared and millimetre / submillimetre wavebands .
here we assume the same forms of evolution of both the merger efficiency parameter @xmath44 and the activity parameter @xmath67 that were required to account for the data in the far - infrared and submillimetre wavebands in the previous section , but make predictions in the near - infrared , optical and ultraviolet wavebands .
in particular , we investigate the 35-k model , in which the redshift distribution of scuba galaxies is in best agreement with observations .
subject to the uncertain fraction of the luminosity of these galaxies that is assumed to be powered by star formation activity , we predict the integrated background radiation intensity from the near - infrared to ionizing ultraviolet wavebands , and the counts of galaxies with seds that are dominated by evolved stars in the near - infrared @xmath132-band , and by young stars in the optical @xmath133-band . by requiring that the @xmath132- and @xmath133-band counts are reproduced accurately , we estimate both the fraction of all energy released in mergers that is reprocessed by dust @xmath134 and the normalization of the activity parameter at @xmath22 , @xmath81 in the optical waveband . for a discussion of the evolution of faint galaxies and their stellar populations see ellis ( 1997 ) . the counts of galaxies in the @xmath132-band at a flux density @xmath135 can be predicted by assuming the forms of the merger efficiency @xmath41 , as listed in table1 , an sed typical of evolved stars @xmath136 ( charlot , worthey & bressan 1996 ) , the press
schechter mass function ( equation 1 ) and the mass - to - light ratio of evolved stellar populations @xmath137 .
the sed was calculated using a 9.25-gyr old bruzual
charlot instantaneous burst model with a salpeter imf .
upper and lower mass limits of 0.1 and 125m@xmath30 were assumed for the imf .
note that the form of the evolved stellar spectrum derived is almost independent of the exact values of the upper and lower mass limits assumed .
the @xmath132-band count @xmath138 with @xmath139 by ensuring that the predicted counts match the observed @xmath132-band counts , a suitable form of the mass - to - light ratio @xmath137 is determined as a function of redshift .
the mass in this ratio is defined as the mass of the dark matter haloes of galaxies , taken from the press schechter function ( equation 1 ) , and the luminosity is the bolometric luminosity of the evolved stellar population in the galaxies .
\(a ) ( b ) in order to reproduce the observed @xmath133-band counts , the counts derived for the evolved and merging components are added together , as shown in fig.11(b ) . the redshift dependence of the mass - to - light ratio is the same as that of the luminosity density of evolved stars , @xmath140 which depends on the sfr at all earlier epochs . a factor of @xmath141 is included in the denominator to mimic the effects of passive stellar evolution ( longair 1998 ) .
at @xmath22 a form of the mass - to - light ratio , @xmath142 where @xmath143l@xmath30 , is required to match the observed @xmath132-band counts ( fig.11a ) and the faint - end slope of the observed @xmath132-band luminosity function ( gardner et al .
1996 ; szokoly et al .
1998 ) .
the well fitting @xmath132-band count that is derived from the model with this form of the mass - to - light ratio @xmath137 is shown , along with the observational data , in fig.11(a ) .
both passive evolved galaxies and luminous merging galaxies make a contribution to the @xmath133-band counts .
the evolved contribution is predicted by evaluating the function that produces the @xmath132-band count at the frequency of the @xmath133-band .
the extrapolation is made using the model sed described above .
an additional population of merging galaxies is also included .
their counts are determined using equation ( 13 ) directly , with a value of @xmath144 @xmath134 is the fraction of the total energy released in a merger that is reprocessed into the far - infrared waveband , and @xmath145 is the sed of a flat star - forming young stellar spectrum at frequencies less than the lyman limit frequency @xmath146hz . @xmath147 if @xmath148 and zero otherwise .
the blue power - law sed expected from an agn will be described reasonably well by this sed at @xmath149 .
the faint counts at @xmath150 , which are dominated by merging galaxies , can be reproduced in the model only if the dust absorption fraction @xmath151 and the zero - redshift activity parameter @xmath152gyr@xmath5 .
the activity parameter incorporated in the model evolves with redshift as shown in equation ( 18 ) , with the value of @xmath153 that is appropriate in the 35-k model .
note that this value of the activity parameter is less than that required to account for the observed submillimetre - wave counts , and that the ratio of energy emitted in the restframe ultraviolet and far - infrared wavebands is 1:4 .
this ratio is equivalent to 1.75 magnitudes of extinction when integrated over the optical and ultraviolet wavebands .
the redshift distribution of faint galaxies at @xmath154 derived in the hierarchical model is in good agreement with that observed for galaxies with @xmath155 ( cowie , hu & songaila 1995 ) ; these details are discussed more extensively elsewhere ( jameson et al . 1999 , in preparation ) .
note that the dependence of the submillimetre and faint @xmath133-band counts on the merging efficiency parameter @xmath41 and the agn fraction @xmath51 is identical , and so the value of the agn fraction does not affect the resulting counts .
the most numerous population of faint high - redshift optically selected galaxies known are the lyman - break galaxies ( steidel et al . 1996a , 1999 ) at @xmath156 , which have apparent sfrs of a few 10m@xmath30yr@xmath5 .
the surface density of detected lyman - break galaxies is about an order of magnitude greater than that of submillimetre - selected galaxies , while their sfrs are typically about an order of magnitude less .
the violence of a typical merger - induced starburst / agn , and thus its detectability , is determined by the product of the merging efficiency and activity parameters , @xmath157 , in the hierarchical model .
the value of this composite parameter that is required to fit the observed submillimetre - wave counts is about 40 times greater than that required to fit the faint @xmath133-band counts .
the value of the activity parameter @xmath67 that is required to fit the submillimetre - wave counts is a factor of about 10 times greater than that required to fit the @xmath133-band counts .
these differences are thus comparable with the observed ratios of the surface densities and luminosities of typical galaxies in the submillimetre - selected and lyman - break samples . based on these differences
, we suggest a scenario in which the optically selected lyman - break galaxies and the submillimetre - selected scuba galaxies are drawn from the same underlying population of luminous galaxy merger events , but are distinguished by being observed during two distinct phases of the merger process .
we associate one phase , which is very luminous , short - lived and heavily dust enshrouded , with the scuba galaxies , and the other , which is less luminous and relatively lightly obscured , with the lyman - break galaxies . during the first phase , which is about 40 times more luminous , but
only about a tenth of the duration as compared with the second , most of the activity in the merger will be almost completely obscured from view in the optical waveband , but extremely bright in the submillimetre waveband .
this phase is consistent with the extremely compact nuclear starburst / agn activity observed by downes & solomon ( 1998 ) on sub - kpc scales in nearby ultraluminous _ iras _ galaxies . during the second phase , less intense star formation activity
would probably be distributed throughout the ism of both merging galaxies .
a short lived ultraluminous phase and a longer - lived less intense burst of star formation activity during a merger are consistent with the results of hydrodynamic models of galaxy mergers by mihos & hernquist ( 1996 ) and bekki et al .
( 1999 ) . however , while plausible , this scenario is not necessarily correct .
the faint counts in the submillimetre and optical waveband could simply be drawn from two distinct populations .
the questions of whether and how ultraluminous dust - enshrouded mergers are connected with the lyman - break galaxies can only be answered by making multiwaveband observations of large samples of submillimetre - selected galaxies in order to observe a time sequence of merging galaxies and to investigate the merger process in detail . observations of the lyman - break population in the submillimetre waveband ( chapman et al . 1999 ) will also help to address these questions .
the global luminosity density predicted by the hierarchical models is based on the evolution of the merging efficiency parameter @xmath41 . by making minor modifications to the formalism presented in section 2.3
, the models listed in table1 can be used to predict the background radiation intensity . the near - infrared / optical / ultraviolet background radiation intensity produced by merging galaxies
can be calculated as shown in equation ( 11 ) , if the fraction of energy generated by a merger that is absorbed by dust @xmath134 is included .
thus @xmath158 the background radiation intensity due to the evolved stellar population is not given directly by the expression for the far - infrared background in equation ( 11 ) . in that case
the volume emissivity is given by equations ( 21 ) and ( 22 ) . by integrating this emissivity of evolved objects over redshift , assuming the sed @xmath136 introduced above , the background radiation intensity produced by evolved non - merging galaxies
can be calculated . when it is added to the background radiation spectrum produced in merging galaxies , a complete prediction of the background intensity between the near - infrared and near - ultraviolet wavebands
is obtained .
the background radiation intensity predicted using the 35-k model , which can explain the @xmath132- and @xmath133-band galaxy counts , is shown across the millimetre to the ultraviolet wavebands in fig.12 .
the background intensity is in agreement with almost all the observed limits and detections .
the only spectral region in which the background radiation intensity is still not very well defined is between about 3 and 7@xmath0 m . at these wavelengths , the dominant source of background radiation switches from dust emission to starlight .
there is almost certainly an additional , and perhaps dominant , contribution to the background radiation intensity in the mid - infrared waveband from very hot dust grains in the central regions of agn , which are not modelled here .
the model curves shown fig.12 are not extrapolated into this region from the well - determined populations of galaxies at 60-@xmath0 m and in the @xmath132 band .
we have introduced a series of parameters and functions to account for different observations section by section through the paper . excluding the world model parameters @xmath68 , @xmath69 and @xmath70 , five parameters are required to define the merger rate of dark matter haloes .
five further independent parameters [ @xmath78 , @xmath75 , @xmath159 , @xmath81 and @xmath119 ] and a sed for dusty galaxies , are required to fit the counts at wavelengths of 15 , 60 , 175 , 450 and 850@xmath0 m and the submillimetre - wave / far - infrared background radiation spectrum .
the most important element of the models are the two functions that describe the merger efficiency parameter @xmath41 and the activity parameter @xmath67 .
we present appropriate forms for these functions in equations ( 16 ) and ( 18 ) respectively , but stress that these forms are not unique .
as more data becomes available , other functional forms or free - form fitting functions may be more appropriate . by modifying the activity parameter @xmath81 , introducing the total fraction of luminosity absorbed by dust @xmath134 , and including a template sed for star - forming galaxies in the rest frame ultraviolet waveband
, the @xmath133-band counts can also be reproduced . in table2
we summarize these parameters and the most important pieces of data used to constrain them .
the values of the parameters required to fit the data in the 35-k model are also listed .
0.75 mm ' '' '' 1.2 mm [ cols= " < , < , < , < " , ] 1.2 mm ' '' ''
1 . we have presented a simple model of hierarchical galaxy formation which incorporates the effects of obscuration by dust , in which the galaxies that are detected in submillimetre - wave surveys are observed during a merger - induced episode of star formation or agn fuelling .
the aim of this model is to elucidate the most important physical processes that could be at work in luminous dusty galaxies , rather than to provide a detailed quantitative description .
the model is constrained primarily by the intensity of background radiation in the far - infrared / submillimetre waveband . from these data alone , the luminosity density from high - redshift galaxies is inferred to exceed that deduced from observations in the rest frame ultraviolet and optical wavebands by up to an order of magnitude . the source counts and background radiation intensity in the submillimetre , far- , mid- and near - infrared , and optical wavebands are reproduced adequately in the model without introducing a large number of parameters .
the counts of galaxies detected in the far - infrared / submillimetre and optical wavebands , and the associated background radiation intensities in these wavebands are consistent if about 4 times more energy is emitted by galaxies after being reprocessed into the far - infrared waveband by interstellar dust as is radiated directly in the optical / ultraviolet waveband .
4 . in order to account for the observed abundance of distant galaxies detected at 175 and 850@xmath0 m using _ iso
_ and scuba , the mass - to - light ratio of a typical galaxy merger must decrease with redshift , by a factor of about 10 and 200 at @xmath161 and 3 respectively .
thus high - redshift mergers must be typically more violent as compared with their low - redshift counterparts .
we suggest two possible physical explanations .
first , that gas is converted into stars / feeds an agn uniformly more efficiently and rapidly in all merging galaxies as redshift increases , perhaps due to a lower bulge - to - disk ratio , which makes disk instabilities grow more quickly ( mihos & hernquist 1996 ) . secondly , that a decreasing fraction of dark matter halo mergers are associated with an efficient mode of star formation / agn fuelling as redshift increases .
5 . in the context of galaxy formation within merging dark matter haloes ,
we have described how the physical processes that convert merging mass into visible radiation must evolve with redshift in order to account for the data in the far - infrared and submillimetre wavebands .
this has previously been discussed by guiderdoni et al .
( 1998 ) , in the conventional context of semi - analytic models , where gas is assumed to cool into dark matter haloes and form stars on galactic scales . in order to account for the observations , an additional population of ultraluminous galaxies
was incorporated arbitrarily into their models .
we have improved our previous models ( blain et al .
1999c ) significantly , by including some astrophysics and not simply invoking an empirical form of the evolution of a low - redshift luminosity function to fit the data . by assuming only a single population of luminous merging galaxies we are able to account for all the data in the far - infrared and submillimetre wavebands .
clear forms of evolution of both the efficiency with which luminosity is generated by a galaxy merger as a function of redshift , and of a function that connects the duration of the luminous phase and the fraction of dark matter halo mergers that generate a luminous event are required to reproduce the results of observations .
the way in which gas is processed in the sub - kpc core regions of galaxy mergers to reproduce the necessary high efficiency and short time - scale of luminous events must be investigated in future work .
we find that the observed counts of both submillimetre - selected galaxies and lyman - break galaxies can be accounted for in terms of merger events in an hierarchical model of galaxy formation , which include identical forms of evolution with redshift , but with different absolute normalisations .
we find that 80per cent of the total amount of energy generated in merger - induced starbursts / agn is liberated in the far - infrared waveband .
it is plausible that the submillimetre - selected galaxies and the lyman - break galaxies are associated with temporally distinct phases of a common population of merging dark matter haloes .
a scenario in which a short - lived , highly obscured far - infrared starburst / agn phase dominates the integrated luminosity of the merger and is surrounded in time by a less luminous , more lightly obscured phase that lasts about 10 times longer is consistent with the data . in this scenario ,
the merger would be classified as a scuba galaxy if it was observed during the short - lived phase , and as a lyman - break galaxy during the long - lived phase .
the results presented here provide excellent opportunities for further study .
two key scientific questions remain unanswered .
first , what are the physical processes that are responsible for the evolution of both the star - formation / agn - fuelling efficiency and the activity parameter in galaxy mergers as a function of redshift ?
secondly , what is the relationship between samples of faint galaxies selected in the optical waveband and submillimetre - selected galaxies
? larger samples of submillimetre - selected galaxies and more comprehensive multiwaveband follow - up observations will allow these questions to be answered .
we thank nigel metcalfe for providing a comprehensive list of optical count data , and chris mihos , priya natarajan , kate quirk , chuck steidel and neil trentham for providing useful comments on the manuscript .
thanks are also due to an anonymous referee for helpful suggestions and prompt reading of the manuscript .
awb , aj and rji acknowledge pparc , is thanks the royal society , and jpk thanks the cnrs for support .
in addition , awb thanks menrt for support while in toulouse , and the caltech ay visitors program for support while this work was completed .
jameson a. , longair m.s . ,
blain a.w .
, in gibson b.k .
, axelrod t.s . , putman m.e .
, the third stromlo symposium : the galactic halo .
pacific conf . series .
165 , astron .
, san francisco , p.346 pettini m. , steidel c.c .
, adelberger k.l . , kellogg m. , dickinson m. , giavalisco m. , 1998b , in shull j.m . , woodward c.e .
, thronson h.a .
eds , cosmic origins : evolution of galaxies , stars , planets and life .
, san francisco , p.67 | a large fraction of the luminous distant submillimetre - wave galaxies recently detected using the submillimetre common - user bolometer array ( scuba ) camera on the james clerk maxwell telescope appear to be associated with interacting optical counterparts .
we investigate the nature of these systems using a simple hierarchical clustering model of galaxy evolution , in which the large luminosity of the scuba galaxies is assumed to be generated at the epoch of galaxy mergers in a burst of either star formation activity or the fuelling of an active galactic nucleus ( agn ) .
the models are well constrained by the observed spectrum of the far - infrared / submillimetre - wave background radiation and the 60-@xmath0 m counts of low - redshift _ iras _ galaxies .
the ratio between the total amount of energy released during mergers and the mass of dark matter involved must increase sharply with redshift @xmath1 at @xmath2 , and then decrease at greater redshifts .
this result is independent of the fraction of the luminosity of mergers that is produced by starbursts and agn .
one additional parameter the reciprocal of the product of the duration of the enhanced luminosity produced by the merger and the fraction of mergers that induce an enhanced luminosity , which we call the activity parameter is introduced , to allow the relationship between merging dark matter haloes and the observed counts of distant dusty galaxies to be investigated .
the observed counts can only be reproduced if the activity parameter is greater by a factor of about 5 and 100 at redshifts of 1 and 3 respectively , as compared with the present epoch .
hence , if merging galaxies account for the population of scuba galaxies , then the merger process must have been much more violent at high redshifts . we discuss the counts of galaxies and the intensity of background radiation in the optical / near - infrared wavebands in the context of these hierarchical models , and thus investigate the relationship between the populations of submillimetre - selected and lyman - break galaxies .
galaxies : evolution galaxies : formation cosmology : observations cosmology : theory diffuse radiation infrared : galaxies |
quantum and statistical field theory investigate the influence of field fluctuations on the @xmath1-point functions .
interactions lead to an infinite hierarchy of schwinger - dyson equations for the @xmath1-point functions .
these integral equations can only be closed approximately , for instance , by the well - established the self - consistent method of kadanoff and baym .
+ recently , it has been shown that the schwinger - dyson equations of qed can be closed in a certain functional - analytic sense . using functional derivatives with respect to the free propagators and
the interaction two closed sets of equations were derived .
the first one involves the connected electron and two - point function as well as the connected three - point function , whereas the second one determines the electron and photon self - energy as well as the one - particle irreducible three - point function .
their conversion to graphical recursion relations leads to a systematic graphical generation of all connected and one - particle irreducible feynman diagrams in qed , respectively .
+ the purpose of the present paper is to apply this functional - analytic approach to the @xmath0-theory of second - order phase transitions in the disordered , symmetric phase .
a short ouline of this program was already published in ref .
@xcite . to this end
we derive in section [ phi ] a closed set of equations for the connected two- and four - point function .
analogously , we determine in section [ irred ] a closed set of schwinger - dyson equations for the self - energy and the one - particle irreducible four - point function . in both cases ,
the closed set of schwinger - dyson equations can be converted into graphical recursion relations for the corresponding connected and one - particle irreducible feynman diagrams in @xmath0-theory . from these the respective connected vacuum diagrams
follow by short - circuiting external legs .
thus our present approach is complementary to ref . which was based on the observation that the complete knowledge of the vacuum energy implies the knowledge of the entire theory ( `` the vacuum is the world '' ) @xcite . in that paper
the vacuum diagrams were constructed in a first step , together with their weights , as solutions of a graphical recursion relation derived from a nonlinear functional differential equation . in a second step
, all diagrams with external lines are obtained from functional derivatives of the connected vacuum diagrams with respect to the free correlation function .
euclidean @xmath0-theories in @xmath2 dimensions are useful models for a large family of universality classes of continuous phase transitions . in particular , the o@xmath3-symmetric @xmath0-theory serves to describe the critical phenomena in dilute polymer solutions ( @xmath4 ) , ising- and heisenberg - like magnets ( @xmath5 ) , and superfluids ( @xmath6 ) . in all these systems , the thermal fluctuations of a self - interacting scalar order parameter field
@xmath7 with @xmath8 components are controlled by the ginzburg - landau energy functional e [ ] = d^d x \ { _ = 1^n _ ( x ) ( -^2_x + m^2 ) _ ( x ) + ^2 } , where the mass @xmath9 is proportional to the temperature deviation from the critical point , and @xmath10 denotes the coupling constant . in the following
it turns out to be advantageous to rewrite the ginzburg - landau energy functional ( [ gl ] ) as e [ ] = _ 12 g^-1_12 _ 1 _ 2 + _
1234 v_1234 _ 1 _ 2 _ 3 _ 4 . in this short - hand notation ,
the spatial and tensorial arguments of the order parameter field @xmath7 , the bilocal kernel @xmath11 , and the quartic interaction @xmath12 are indicated by simple number indices , i.e. , 1 \ { x_1 , _ 1 } , _ 1 _ _ 1=1^n d^d x_1 , _ 1 _ _ 1 ( x_1 ) .
the kernel @xmath11 represents the functional matrix g^-1_12 g__1 , _ 2 ^ -1 ( x_1 , x_2 ) = _ _ 1 , _ 2 ( - _ x_1 ^ 2 + m^2 ) ( x_1 - x_2 ) , while the interaction @xmath12 is given by the functional tensor v_1234 v__1,_2,_3,_4 ( x_1 , x_2 , x_3 , x_4 ) = ( _ _ 1 , _ 2 _ _ 3 , _
4 + _ _ 1 , _ 3 _ _ 2 , _
4 + _ _ 1 , _ 4 _ _ 2 , _ 3 ) ( x_1 - x_2 ) ( x_1 - x_3 ) ( x_1 - x_4 ) , both being symmetric in their indices . for the purpose of this paper
we shall leave the kernel @xmath11 in the energy functional ( [ ef ] ) completely general , except for the symmetry with respect to its indices . by doing so ,
we regard the energy functional ( [ ef ] ) as a functional of the kernel @xmath11 : e [ ] = e[,g^-1 ] . as a consequence , all global and local statistical quantities derived from ( [ eff ] )
are also functionals of the bilocal kernel @xmath11 .
in particular , we are interested in studying the functional dependence of the partition function , defined by a functional integral over a boltzmann weight in natural units z [ g^-1]= e^- e [ * * , g^-1 ] , and the ( negative ) vacuum energy w [ g^-1]= z [ g^-1 ] . for the sake of simplicity we restrict ourselves in the present paper to study the disordered , symmetric phase of the @xmath0-theory where the @xmath1-point functions _
1 n [ g^-1 ] = _ 1 _ n e^- e [ * * , g^-1 ] , with odd @xmath1 vanish .
thus the first nonvanishing @xmath1-point functions are the two - point function _ 12 [ g^-1]= _ 1 _ 2 e^- e [ , g^-1 ] and the four - point function _ 1234 [ g^-1]= _ 1 _ 2 _ 3 _ 4 e^- e [ , g^-1 ] .
further important statistical quantities are the correlation functions , i.e. the connected @xmath1-point functions . in the disordered , symmetric phase ,
the connected two - point function coincides with the two - point function _ 12^c [ g^-1 ] = _ 12 [ g^-1 ] , whereas the connected four - point function is defined by _
1234^c[g^-1 ] _ 1234 [ g^-1]- _ 12 [ g^-1]_34[g^-1 ] - _ 13 [ g^-1]_24[g^-1]- _ 14[g^-1 ] _
23[g^-1 ] . by expanding the functional integrals ( [ pf ] ) and ( [ npo ] ) in powers of the coupling constant @xmath10 , the expansion coefficients of the partition function and the @xmath1-point functions consist of free - field expectation values .
these are evaluated with the help of wick s rule as a sum of feynman integrals , which are pictured as diagrams constructed from lines and vertices .
thereby the free correlation function @xmath13 , which is the functional inverse of the kernel @xmath11 in the energy functional ( [ ef ] ) _ 2 g_12 g^-1_23 = _ 13 with @xmath14 , is represented by a line in a feynman diagram g_12 , and the interaction @xmath12 is pictured as a vertex - v_1234 .
+ the graphical elements ( [ pro ] ) and ( [ ve ] ) are combined by an integral which graphically corresponds to the gluing prescription - _ 4 v_1234 g_45 .
[ 5 ] in this paper we analyze the resulting diagrams for the statistical quantities ( [ pf])([c4p ] ) by using their functional dependence on the kernel @xmath15 . to this end
we introduce the functional derivative with respect to the kernel @xmath15 whose basic rule reflects the symmetry of its indices : = ( _ 13 _ 42 + _ 14 _ 32 ) . from the identity ( [ fp ] ) and the functional product rule we find the effect of this derivative on the free correlation function - = ( g_13 g_42 + g_14 g_32 ) , which has the graphical representation - = ( + ) .
thus a functional derivative with respect to the kernel @xmath15 is represented by a graphical operation which cuts a line of a feynman diagram in all possible ways . for practical purposes
it is convenient to use also functional derivatives with respect to the free correlation function @xmath13 whose basic rule reads = ( _ 13 _ 42 + _ 14 _ 32 ) .
such functional derivatives are represented graphically by removing a line of a feynman diagram in all possible ways . the functional derivatives with respect to the kernel @xmath15 and
the correlation function @xmath13 are related via the functional chain rule = - _ 34 g_13 g_24 . [ 13 ] these functional derivatives are used in subsection [ cse ] to derive a closed set of schwinger - dyson equations for the connected two- and four - point functions . in subsection [ grc ]
they are converted into graphical recursion relations for the corresponding connected feynman diagrams .
finally , the connected vacuum diagrams contributing to the vacuum energy are constructed in a graphical way in subsection [ vaa ] . in this subsection
we apply the functional derivatives introduced so far to a functional identity which immediately follows from the definition of the functional integral . by doing so ,
we derive a closed set of schwinger - dyson equations determining the connected two- and four - point function . in order to derive an equation for the connected two - point function ,
we start with the trivial identity ( _ 2 e^- e [ ] ) = 0 , which follows via direct functional integration from the vanishing of the exponential at infinite fields . taking into account the explicit form of the energy functional ( [ ef ] ) , we perform the functional derivative with respect to the field and obtain ( _ 12 - _ 3 g^-1_13 _ 2 _ 3 - _ 345 v_1345 _ 2 _ 3 _ 4 _ 5 ) e^- e [ ] = 0 . applying the definitions ( [ pf])([4p ] ) and ( [ c4p ] )
, this equation can be expressed in terms of the connected two- and four - point function as follows + _ 12 - _ 3 g^-1_13 _ 23 = _
345 v_1345 _ 34 _
52 + _ 345 v_1345 _ 2345^c .
multiplying this equation with @xmath16 and integrating with respect to the index @xmath17 finally leads to the schwinger - dyson equation which determines the connected two - point function : _ 12 = g_12 - _ 3456 g_13 v_3456 _ 45 _ 62 - _ 3456 g_13 v_3456 _ 2456^c . when the connected two - point function is graphically represented in feynman diagrams by a double line [ fs ] _ 12 , and the connected four - point function
is pictured by a vertex with an open dot with four legs + _ 1234^c , + this schwinger - dyson equation reads graphically : = + + .
it represents an integral equation for the connected two - point function on the left - hand side which turns out to appear also on the right - hand side .
iteratively solving the integral equation ( [ sdg1 ] ) for the connected two - point function @xmath18 necessitates , however , the knowledge of the connected four - point function @xmath19 . in principle , one could determine the connected four - point function ( [ c4p ] ) from the connected two - point function ( [ 2p ] ) as follows .
we obtain from ( [ ef ] ) and ( [ dr1 ] ) [ sr ] _ 1 _ 2 = 2 , so that we yield from ( [ pf ] ) , ( [ 4p ] ) , and ( [ c4p ] ) [ gde ] _ 1234^c = - 2 - _ 13 _ 24- _ 14 _ 23 .
this result reads graphically = - 2 - - .
+ however , such a procedure would have the disadvantage that cutting a line in the diagrams of the connected two - point function @xmath18 would also lead to disconnected diagrams which are later on removed by the second and the third term on the right - hand side of ( [ gdeb ] ) . as the number of undesired disconnected diagrams occuring at an intermediate step of the calculation increases with the loop order ,
this procedure is quite inefficient to determine the connected four - point function @xmath19 . therefore we aim at deriving another equation for @xmath19 whose iterative solution only involves connected diagrams . to this end
we insert the schwinger - dyson equation ( [ sd1 ] ) into relation ( [ gde ] ) and obtain _ 1234^c & = & _ 5678 g_15 v_5678 - _ 5678 g_15 v_5678 _ 62 _ 73 _ 84 - _ 5678 g_15 v_5678 _ 6734^c _ 82 + & & - _ 5678 g_15 v_5678 _ 8234^c _ 67 + _
5678 g_15 v_5678 _ 6784^c _ 23 + _ 5678 g_15 v_5678 _ 6783^c _ 24 .
[ 59 ] the last two terms in ( [ zw ] ) are still disconnected and cancel the disconnected diagrams which are generated by the functional derivative of the connected four - point function with respect to the kernel in the first term . in order to eliminate all disconnected terms from ( [ zw ] )
we use the commutator relation - & = & ( _ 13 _ 2456^c + _ 14 _ 2356^c + _ 23 _ 1456^c + _ 24 _ 1356^c .
- _ 15 _ 2346^c - _ 16 _ 2345^c - _ 25 _ 1346^c - _ 26 _ 1345^c ) , which directly follows from ( [ gde ] ) because of the identity that mixed second functional derivatives with respect to the kernel @xmath11 can be interchanged : = .
the commutator relation ( [ ko1 ] ) states that within the functional derivative of the connected four - point function with respect to the kernel indices might be interchanged at the expense of the additional terms on the right - hand side . inserting ( [ ko1 ] ) in ( [ zw ] ) by taking into account the schwinger - dyson equation ( [ sd1 ] ) and the functional chain rule ( [ 13 ] ) , we yield the following functional integrodifferential equation for the connected four - point function : _
1234^c & = & - _ 5678 g_15 v_5678 _ 62 _ 73 _ 84 - _ 5678 g_15 v_5678 g_69 g_70 - _ 5678 g_15 v_5678 _ 8234^c _ 67
+ & & - _ 5678 g_15 v_5678 _ 6723^c _ 84 - _ 5678 g_15 v_5678 _ 6724^c _ 83 - _ 5678 g_15 v_5678 _ 6734^c _ 82 .
although eq .
( [ sd2 ] ) has the disadvantage of being more complex than eq .
( [ gde ] ) , it has the advantage that it does not lead to disconnected diagrams at an intermediate stage of the calculation .
this can be immediately seen in its graphical representiation which reads = + + + + + .
[ sdg2 ] thus the closed set of schwinger - dyson equations for the connected two- and four - point function is given by ( [ sdg1 ] ) and ( [ sdg2 ] ) .
now we demonstrate how the diagrams of the connected two- and four - point function are recursively generated in a graphical way . to this end
we perform for both quantities a perturbative expansion & & _ p=0^ , + & & _ p=0^ , [ 64 ] + & & where @xmath20 denotes the number of interactions @xmath12 which contribute . with these
we obtain from ( [ sdg1 ] ) and ( [ sdg2 ] ) the following closed set of graphical recursion relations : = _ q=0^p + , = _
q=0^p _ r=0^q + + _ q=1^p + _ q=1^p + _ q=1^p + _ q=1^p . + this is solved starting from & = & , + [ 4 mm ] & = & 0 .
+ note that these graphical recursion relations ( [ rek1])([in2 ] ) allow to prove via complete induction that all diagrams contributing to the connected two- and four - point function are , indeed , connected .
+ the first few perturbative contributions to @xmath18 and @xmath19 are determined as follows .
inserting ( [ in1 ] ) and ( [ in2 ] ) in ( [ rek1 ] ) and ( [ rek2 ] ) , we obtain for @xmath21 the connected two - point function = [ 69 ] and the connected four - point function = . [ 70 ] + with this we get from ( [ rek1 ] ) the second - order contribution to the connected two - point function + = + + .
[ 73 ] amputating one line from ( [ 70 ] ) , & = & _ 26 + _ 36 + _ 46 + _ 27 + & & + _ 37 + _ 47 + _ 56 + _ 57 , [ 71 ] + & & we find the second - order contribution to the connected four - point function from ( [ rek2 ] ) : + & = & ( + + + ) + & & + ( + + ) . [ 74 ] + & & the results ( [ 69])([73 ] ) , and ( [ 74 ] ) are listed in table [ tab1 ] and [ tab2 ] which show all diagrams of the connected two- and four - point function up to the forth perturbative order irrespective of their spatial indices . however , to evaluate the recursion relations in higher orders , it becomes necessary to reassign the spatial indices to the end points of the diagrams of the connected two- and four - point function in table [ tab1 ] and [ tab2 ] .
this involves a decomposition of the weights shown in table [ tab1 ] and [ tab2 ] due to symmetry considerations . to this end
we characterize a diagram by its symmetry degree @xmath8 .
thus reassigning the spatial indices to the end points of the diagrams of the connected two- and four - point function leads to @xmath22 different diagrams . as an example for determining the symmetry degree @xmath8 we consider diagram # 4.5 in table [ tab2 ] . successively assigning the indices @xmath23 to the respective end points leads to @xmath24 : + & & 24 24 8 + 8 + 8 + & & 4 + 4 + 4 + 4 + 4 + 4 .
+ note that the weights of the diagrams of the connected two- and four - point function in table [ tab1 ] and [ tab2 ] have been determined by solving the graphical recursion relations .
however , as a cross - check , they also follow from the formula w^(n ) = , where @xmath1 stands for the number of external legs and @xmath25 denote the number of self- , double , triple connections between vertices .
furthermore , @xmath26 stands for the number of vertex permutations leaving the diagrams unchanged .
the connected vacuum diagrams of @xmath0-theory can be generated order by order together with their weights in two ways . in this section
we show that they follow from short - circuiting the external legs of the diagrams of the connected two- and four - point function , respectively .
thus our present approach is complementary to ref .
@xcite , where a nonlinear functional integrodifferential equation for the vacuum energy was recursively solved in a graphical way in order to directly generate the connected vacuum diagrams .
our first approach is based on the connected two - point function @xmath18 .
we start with concluding _
12 = -2 , which follows from ( [ pf ] ) , ( [ vas ] ) , and ( [ 2p ] ) by taking into account ( [ sr ] ) .
we can read off from ( [ git ] ) that cutting a line of the connected diagrams of the vacuum energy in all possible ways leads to all diagrams of the connected two - point function . here , however , we want to regard ( [ git ] ) as a functional differential equation for the vacuum energy @xmath27 .
if the interaction @xmath12 vanishes , eq . ( [ git ] )
is solved by the free contribution of the vacuum energy w^(0 ) = - g^-1 . [ 28 ] here the trace of the logarithm of the kernel @xmath11 is defined by the series g^-1 - _ n=1^ _ 1 ... n ( _ 12 - g^-1_12 ) ...
( _ n-1,n - g^-1_n-1,n ) ( _ n1 - g^-1_n1 ) , [ 29 ] so that we get [ 0id ] - 2 = g_12 . for a non - vanishing interaction @xmath12 , the functional differential equation ( [ git ] ) produces corrections to ( [ 28 ] ) , which we shall denote with @xmath28 .
thus the vacuum energy decomposes according to w = w^(0 ) + w^(int ) , [ 34 ] and we obtain together with ( [ git ] ) and ( [ 0id ] ) [ gitb ] - _ 12 g^-1_12 = _ 12 g^-1_12 ( _ 12 - g_12 ) . in the following , we aim at recursively determining @xmath28 in a graphical way . to this end
we perform a perturbative expansion of the interaction part of the vacuum energy w^(int ) = _ p=1^ w^(p ) , [ 39 ] and a corresponding one of the connected two - point function _ 12 = g_12 + _ p=1^_12^(p ) .
[ 34b ] inserting ( [ 39 ] ) and ( [ 34b ] ) into ( [ gitb ] ) and using the functional chain rule ( [ 13 ] ) , we obtain for @xmath29 [ 34c ] _ 12 g_12 = _ 12 g^-1_12 ^(p)_12 .
the contributions @xmath30 of the vacuum energy obey the following eigenvalue problem for @xmath29 [ eig ] _ 12 g_12 = 2 p w^(p ) . indeed , in the @xmath20th perturbative order the functional derivative @xmath31 generates diagrams in each of which one of the @xmath32 lines of the original vacuum diagram is removed and , subsequently , the removed line is again reinserted .
thus the left - hand side of ( [ eig ] ) results in counting the lines of the vacuum diagrams in @xmath30 , so that we obtain the eigenvalue @xmath32 . with ( [ eig ] )
we can explicitly solve ( [ 34c ] ) for the respective perturbative contributions of the vacuum energy and obtain for @xmath29 w^(p ) = _ 12 g^-1_12 ^(p)_12 . [ 83 ] now we supplement the above - mentioned feynman rules with a graphical representation for the contributions of the vacuum energy w^(p ) and for the kernel g^-1_12 .
[ 2vert ] + the latter graphical element serves for gluing two lines together according to _ 12 g_31 g^-1_12 g_24 = , [ 84 ] which follows from _ 12 g_31 g^-1_12 g_24 = g_34 .
thus our result ( [ 83 ] ) can be depicted graphically as follows : = .
[ gr83 ] + it states that closing the perturbative contributions of the connected two - point function yields the corresponding contributions of the vacuum energy . from ( [ 69 ] ) and ( [ gr83 ] ) we yield for @xmath21 = .
[ 85 ] correspondingly , the second - order result follows from ( [ 73 ] ) and ( [ gr83 ] ) : = + . [ 86 ] now we elaborate a second approach which is based on closing the diagrams of the connected four - point function @xmath33 . to this end
we insert ( [ git ] ) into the left - hand side of the schwinger - dyson equation ( [ idy ] ) and use ( [ 13 ] ) , ( [ 0id ] ) as well as the decomposition ( [ 34 ] ) to obtain a functional differential equation for the interaction part of the vacuum energy : _ 12 g_12 = - _ 1234 v_1234 _ 1234 - _ 1234 v_1234 _ 12 _ 34 . [ 81 ] combining the perturbation expansions ( [ 39 ] ) , ( [ 34b ] ) for @xmath28 , @xmath18 with the corresponding one for @xmath33 , i.e. _ 1234 = _ p=1^ ^c,(p)_1234 , we obtain together with the eigenvalue problem ( [ eig ] ) from ( [ 81 ] ) w^(p+1 ) = - _ 1234 v_1234 ^c,(p)_1234 - _ q=0^p _ 1234 v_1234 _
12^(q ) _ 34^(p - q ) .
[ 81b ] the graphical representation of this result reads = + _ q=0^p .
+ inserting the diagrams of the connected two- and four - point function from tab .
[ tab1 ] and [ tab2 ] , we obtain the corresponding vacuum diagrams .
for instance , eq .
( [ vre ] ) reduces for @xmath34 = + + [ 82.1 ] with the respective terms & = & + 2 , + & = & + + , + & = & .
thus we obtain the connected vacuum diagrams shown in tab .
[ tab3 ] for @xmath35 : = + + + . [ 82.2 ] the results ( [ 85 ] ) , ( [ 86 ] ) , and ( [ 82.2 ] ) are listed in table [ tab3 ] which shows also one further perturbative order .
note that the weights of the connected vacuum diagrams obey a formula similar to ( [ weig ] ) w = , where @xmath36 denote the number of self- , double , triple , fourfold connections and @xmath26 stands for the number of vertex permutations leaving the vacuum diagram unchanged .
so far , we have explained how to generate connected feynman diagrams of the @xmath0-theory .
we shall now eliminate from them the reducible contributions . to this end
we derive in subsection [ nanu ] a closed set of schwinger - dyson equations for the one - particle irreducible two- and four - point function . in subsection [ nana ]
they are converted into graphical recursion relations for the corresponding one - particle irreducible feynman diagrams needed for renormalizing the @xmath0-theory .
subsection [ ivva ] discusses how these feynman diagrams of the one - particle irreducible two- and four - point functions are related to the connected vacuum diagrams which are also one - particle - irreducible . in this subsection
we revisit the functional identity ( [ idy ] ) which immediately followed from the definition of the functional integral .
we show that it can be also used to derive a closed set of schwinger - dyson equations for the one - particle irreducible two- and four - point functions . from table [ tab1 ]
we can distinguish two classes of diagrams contributing to the connected two - point function @xmath18 .
the first one contains one - particle irreducible diagrams which remain connected after amputating one arbitrary line .
the second one consists of the remaining diagrams which are called reducible .
the diagrams # 2.1 and # 2.2 , for instance , are one - particle irreducible , whereas # 2.3 is reducible . when considering the self - energy _
12 = g^-1_12 - ^-1_12 , [ sigma ] where @xmath37 is the functional inverse of @xmath38 : _ 3 _ 13 ^-1_32 _ 12 , [ 101 ] we will see later on that this quantity is graphically represented by all one - particle irreducible diagrams of the connected two - point function @xmath18 where the external legs are omitted .
reversely , all connected diagrams of the connected two - point function @xmath18 are reconstructed from the one - particle irreducible ones of the self - energy @xmath39 according to the dyson equation _
12 = g_12 + _ 34 g_13 _ 34 _ 42 , [ dyson ] which immediately follows from ( [ sigma ] ) by taking into account ( [ fp ] ) and ( [ 101 ] ) . when the self - energy is graphically represented in feynman diagrams by a big open dot with two legs
_ 12 , the dyson equation ( [ dyson ] ) reads graphically = + .
[ voll ] in a similar way , table [ tab2 ] shows that the diagrams of the connected four - point function @xmath40 are either one - particle irreducible , as for instance diagram # 2.1 , or reducible such as # 2.2 .
1234 = - _ 5678 _ 5678^c ^-1_51 ^-1_62 ^-1_73 ^-1_84 , [ gamma ] we will see later on that this quantity consists of all one - particle irreducible diagrams of the connected four - point function @xmath40 where the external legs are omitted .
therefore this quantity @xmath41 is called the one - particle irreducible four - point function . inverting relation ( [ gamma ] ) yields _ 1234^c = - _ 5678 _ 5678 _ 51 _ 62 _ 73 _ 84 .
[ 100 ] when the one - particle four - point function is depicted by using a big open dot with four legs + -_1234 , the identity ( [ 100 ] ) reads graphically = .
[ 150 ] + + in the following we determine a closed set of schwinger - dyson equations for the self - energy and the one - particle irreducible four - point function . multiplying the identity ( [ idy ] ) with @xmath42 and integrating with respect to the index @xmath43
, we take into account the self - energy ( [ sigma ] ) as well as the connected four - point function ( [ 100 ] ) .
thus we yield the schwinger - dyson equation for the self - energy : _ 12 = - _ 34 v_1234 _ 34 + _ 345678 v_1345 _ 36 _ 47 _ 58 _ 6782 .
[ 103 ] its graphical representation reads = + . [ 106 ] + it contains on the right - hand side the connected two - point function which is determined by the dyson equation ( [ voll ] ) .
iteratively solving the integral equations ( [ voll ] ) and ( [ 106 ] ) necessitates , however , the knowledge of the one - particle irreducible four - point function @xmath41 . in principle , one could determine @xmath41 from the self - energy @xmath39 . to this end
we insert relation ( [ gde ] ) into the definition ( [ gamma ] ) of the one - particle irreducible four - point function : _ 1234 = 2 _ 5678 ^-1_51 ^-1_62 ^-1_73 ^-1_84
+ ^-1_13 ^-1_24 + ^-1_14 ^-1_23 .
[ 107 ] the functional derivative on the right - hand side follows from applying a functional derivative with respect to the kernel @xmath11 to the identity ( [ 101 ] ) : = - _ 90 _ 59 _ 06 . using the definition of the self - energy ( [ sigma ] ) and eq . ( [ dr1 ] ) , the one - particle irreducible four - point function ( [ 107 ] )
reduces to _ 1234 = 2 _ 56 ^-1_35 ^-1_64 . [ 109 ] applying again ( [ sigma ] ) and the functional chain rule ( [ 13 ] ) , this yields -_1234 = 2 - 2 _ 56 _ 35 g_56 - 2 _ 56 g_56 _ 64 + 2
_ 5678 _ 35 g_57 g_86 _ 64 , [ 109 * ] which is represented graphically as + & = & 2 + 2 + & & - 2 - 2 .
[ 111 ] thus amputating a line from the self energy @xmath39 leads to the one - particle irreducible diagrams of @xmath41 . however , this procedure has the disadvantage that the first two terms yield reducible diagrams which are later on removed in the last two terms in ( [ 111 ] ) .
as the number of undesired one - particle reducible diagrams occuring at an intermediate step of the calculation increases with the loop order , the procedure of determining @xmath41 via relation ( [ 111 ] ) is quite inefficient .
therefore we aim at deriving another equation for @xmath41 whose iterative solution only involves one - particle irreducible diagrams .
+ going back to eq .
( [ 109 ] ) , we insert the schwinger - dyson equation ( [ 103 ] ) for the self - energy and obtain _ 1234 & = & v_1234 - _ 5678 v_1256 _ 57 _ 68 _ 7834 - _ 5678 v_1356 _ 57 _ 68 _ 7824 + & & - _ 5678 v_1456 _ 57 _ 68 _ 7823 + _ 567890|1|2 v_5167 _ 69 _ 70 _ 902|1 _ |1|2 _ |2348 _ 85 + & & + _ 567890|1|2 v_1567 _ 58 _ 69 _ 70 ^-1_|13 ^-1_|24 . [ 113 ] the last term is still problematic , as inserting the definition ( [ sigma ] ) of the self - energy would yield again an inefficient equation .
this time , however , we can circumvent the inefficiency problem by deriving from ( [ 109 ] ) the commutator relation _
78 ^-1_75 ^-1_86 - _ 78 ^-1_73 ^-1_84 & = & 2 _
7890|1|2 ^-1_57 ^-1_68 ( ^-1_39 ^-1_40 ) _ 9|1 _ 0|2
_ |1|212 + & & - 2 _
7890|1|2 ^-1_37 ^-1_48 ( ^-1_59 ^-1_60 ) _ 9|1 _ 0|2 _ |1|212 .
[ 115 ] applying then = ( _ 13 _ 24 + _ 14 _ 23 ) - _ 56 _ 35 _ 5126 _ 64 , [ 116 ] which follows from inserting ( [ 109 ] ) in ( [ sigma ] ) , the intermediate commutator relation ( [ 115 ] ) is converted to the final one _ 78 ^-1_75 ^-1_86 - _ 78 ^-1_73 ^-1_84 & = & _ 78 _ 3457 _ 78 _ 8126 + _ 78 _ 3467 _ 78 _ 8125 + & & - _ 78 _ 5637 _ 78 _ 8124 - _ 78 _ 5647 _ 78 _ 8123 .
[ 117 ] note that the present commutator relation ( [ 117 ] ) is equivalent to the former one ( [ ko1 ] ) in section [ phi ] . indeed , eq . (
[ 117 ] ) follows also from inserting ( [ 100 ] ) into ( [ ko1 ] ) after a lengthy but straight - forward calculation .
now we can treat the last problematic term in ( [ 113 ] ) with the commutator relation ( [ 117 ] ) and thus yield a nonlinear functional integrodifferential equation for the one - particle irreducible four - point function : - _ 1234 & = & - v_1234 + _ 567890 v_1567 _
58 g_69 g_70 + _ 5678 v_1256 _ 57 _ 68 _ 7834 + & & + _ 5678 v_1356 _ 57 _ 68 _ 7824 + _ 5678 v_1456 _ 57 _ 68 _ 7823 - _ 567890|1|2 v_5167 _ 69 _ 70 _ 902|1 _ |1|2 _ |2348 _ 85 + & & - _ 567890|1|2 v_5167 _ 69 _ 70 _ 903|1 _ |1|2 _ |2248 _ 85 - _
567890|1|2 v_5167 _ 69 _ 70 _ 904|1 _ |1|2 _ |2238 _ 85 . [ 119 ] its graphical representation reads = + + + + + + + . [
120 ] thus the closed set of schwinger - dyson equations for the connected two - point function , the self - energy and the one - particle irreducible four - point function is given by ( [ voll ] ) , ( [ 106 ] ) , and ( [ 120 ] ) .
in the subsequent section we show that its recursive graphical solution leads to all one - particle irreducible feynman diagrams needed for renormalizing the @xmath0-theory . to this end
we supplement the perturbative expansion ( [ g2p ] ) of the connected two - point function with corresponding ones for the self - energy = _ p=1 ^ and for the one - particle irreducible four - point function = _ p=1 ^ .
[ 122 ] as a result we obtain the following closed set of graphical recursion relations : + & = & + _
s=1 ^r , + & = & _ q=1 ^p , = _
r=1 ^q & & + _ q=1 ^p _
q=2 ^p _ r=2 ^q _
t=1 ^s-1 _ k=0 ^t-1 + & & + _
t=1 ^s-1 _ k=0 ^t-1 + _ q=2 ^p
+ & & this is to be solved starting from & = & , + & & + & = & .
+ & & note that these graphical recursion relations ( [ rek4])([gam1 ] ) allow to prove via complete induction that all diagrams which contribute to the self - energy and the one - particle irreducible four - point function are , indeed , one - particle irreducible .
+ the first few perturbative contributions to @xmath39 and @xmath41 are determined as follows .
inserting ( [ g0 ] ) in ( [ rek4 ] ) and ( [ rek5 ] ) yield for @xmath21 the self - energy = , [ 125 ] and the connected two - point function = = . [ 126 ] as ( [ gam1 ] ) represent a bare vertex with no line , we read off = 0 . [ 127 ] the second - order contribution to the one - particle irreducible four - point function follows from ( [ rek6 ] ) to be = + + .[128 ] + the results in ( [ 125 ] ) , ( [ 126 ] ) , and ( [ 128 ] ) are then used to determine for @xmath34 the self - energy ( [ rek4 ] ) = + [ 129 ] and the connected two - point function ( [ rek5 ] ) = + = + + .
[ 130 ] amputating one line from ( [ 128 ] ) , = ( + + ) = ( + + + + + ) , [ 132 ] + & & we find the third - order contribution to the one - particle irreducible four - point function from ( [ rek6 ] ) : + & & = ( + + ) + & & + ( + + + + + ) + & & + ( + + ) .
[ 136 ] as expected , the diagrams ( [ 126 ] ) and ( [ 130 ] ) for the connected two - point function coincide with the ones from section [ grc ] which are shown in table [ tab1 ] .
furthermore , the diagrams ( [ 125 ] ) , ( [ 129 ] ) and ( [ gam1 ] ) , ( [ 128 ] ) , ( [ 136 ] ) for the self - energy and the one - particle irreducible four - point function are listed in table [ tab4 ] and [ tab5 ] which show all one - particle irreducible diagrams up to the order @xmath44 irrespective of their spatial indices ( compare the discussion before eq .
( [ symmm ] ) ) .
note that these one - particle irreducible feynman diagrams have to be evaluated in order to determine the critical exponents of the @xmath0-theory from renormalizing the field @xmath7 , the coupling constant @xmath10 and the mass @xmath9 .
so far , the results are available up to six and partly to seven loops in @xmath45 @xcite and up to five loops in @xmath46 dimensions whithin the minimal subtraction scheme @xcite .
the vacuum diagrams of @xmath0-theory are not only connected but also one - particle irreducible . in this section
we elaborate how they can be generated from short - circuiting the external legs of the diagrams of the self - energy and the one - particle irreducible four - point function , respectively . at first
, we consider the approach which is based on the self - energy .
combining the dyson equation ( [ dyson ] ) with the perturbative expansion of the connected two - point function ( [ 34b ] ) and the corresponding one for the self - energy _ 12 = _ p=1^ _
12^(p ) , we obtain _ 12 ^(p)_12 g^-1_12 = _ q=1^p _ 12 ^(q)_12 ^(p - q)_12 .
[ 161 ] thus our previous equation ( [ 83 ] ) for determining the contributions of the vacuum energy @xmath30 from the connected two - point function is converted to w^(p ) = _
q=1^p_12 ^(q)_12 ^(p - q)_12 .
[ 162 ] this result reads graphically = _ q=1^p . + we consider one example how the diagrams of the self - energy lead to the corresponding diagrams of the vacuum energy . for @xmath35
( [ sev ] ) reduces to = + + [ 163 ] with the respective terms & = & + + , [ 164 ] + & = & + , [ 165 ] + & = & + + + .
[ 166 ] + thus we reobtain the connected vacuum diagrams ( [ 82.2 ] ) shown in tab . [ tab3 ] for @xmath35 . for the sake of completeness we also mention that the vacuum diagrams can be generated from closing the diagrams of the one - particle irreducible four - point function .
to this end we insert in the equation ( [ sev ] ) determining the contributions of the vacuum energy the recursion relation ( [ rek4 ] ) for the self - energy and yield + = _ q=0^p + _ q=1^p _ r=1^q _ s=1^r _ t=1^s .
[ newv ] + note that this equation follows also directly from the previous recursion relation ( [ vre ] ) by taking into account the identity ( [ 150 ] ) and the perturbative expansions ( [ g2p ] ) , ( [ 122 ] ) .
in this paper we have derived a closed set of schwinger - dyson equations in the disordered , symmetric phase of the @xmath0-theory . in particular , we supplemented the well - known integral equations ( [ sdg1 ] ) and ( [ 106 ] ) for the connected two - point function and the self - energy by the new functional integrodifferential equations ( [ sdg2 ] ) and ( [ 120 ] ) for the connected and one - particle irreducible four - point function .
their conversion to graphical recursion relations has allowed us to systematically generate the corresponding connected and one - particle irreducible feynman diagrams .
furthermore , we have discussed how the short - circuiting of their external legs leads to the associated connected vacuum diagrams . in the subsequent paper
we shall elaborate how tadpoles and , more generally , corrections to the connected two - point function as well as the vertex can be successively eliminated by introducing higher legendre transformations .
this will lead to graphical recursion relations for the skeleton feynman diagrams in @xmath0-theory .
+ the recursive graphical solution of our closed set of schwinger - dyson equations in @xmath0-theory is straightforward and has been carried out in this paper up to the forth perturbative order by hand .
our iterative procedure can easily be automatized by computer algebra . in ref .
it was demonstrated that the basic graphical operations as amputating a line or gluing two lines together can be formulated with the help of a unique matrix notation for feynman diagrams .
it would be interesting to compare the efficiency of our schwinger - dyson approach of generating feynman diagrams of @xmath1-point functions together with their weights with already existing computer programs such as feynarts and qgraf .
some of them are based on a combinatorial enumeration of all possible ways of connecting vertices by lines according to feynman s rules .
others use a systematic generation of homeomorphically irreducible star graphs @xcite .
the latter approach is quite efficient and popular at higher orders .
it has , however , the conceptual disadvantage that it renders at an intermediate stage numerous superfluous diagrams with different vertex degrees which have to be discarded at the end .
further promising methods have been proposed in refs . .
whereas the first one is based on a bootstrap equation that uses only the free field value of the energy as an input , the second one combines schwinger - dyson equations with the two - particle irreducible ( `` skeleton '' ) expansion .
+ we believe that our closed set of schwinger - dyson equations for the connected and one - particle irreducible two- and four - point function will turn out to be useful for developing nonperturbative approximations .
this may proceed , for instance , as in ref . which proposes a self - consistent solution of the dyson equation in @xmath0-theory by using a scaling ansatz for the connected two - point function near the phase transition .
a similar consideration in ginzburg - landau theory has allowed ref . to analyze the influence of the thermal fluctuations of the order parameter and the vector potential on the superconducting phase transition .
a self - consistent solution of our closed schwinger - dyson equation may be found via a phenomenological ansatz for the connected and one - particle irreducible two- and four - point function , respectively .
within such an ansatz one has to find how the functional derivative with respect to the free correlation function is approximated , as this operation is crucial in the new schwinger - dyson equations ( [ sdg2 ] ) and ( [ 120 ] ) .
we thank hagen kleinert for stimulating discussions and for reading the manuscript .
bjorken and s.d .
drell , vol .
i _ relativistic quantum mechanics _ , vol .
ii _ relativistic quantum fields _ ( mcgraw - hill , new york , 1965 ) .
amit , _ field theory , the renormalization group and critical phenomena _ ( mcgraw - hill , new york , 1978 ) .
m. le bellac , _ quantum and statistical field theory _
( oxford science publications , oxford , 1991 ) . c. itzykson and j .- b .
zuber , _ quantum field theory _
( mcgraw - hill , new york , 1985 ) .
j. zinn - justin , _ quantum field theory and critical phenomena _ , third edition ( oxford university press , oxford , 1996 ) .
m.e . peskin and d.v .
schroeder , _ introduction to quantum field theory _ ( addison - wesley , reading , 1995 ) .
kadanoff and g. baym , _ quantum statistical mechanics _ ( benjamin , menlo park , 1962 ) .
a. pelster , h. kleinert , and m. bachmann , ann . of phys .
( n.y . ) * 297 * , 363 ( 2002 ) .
m. bachmann , h. kleinert , and a. pelster , phys .
d 61 * , 085017 ( 2000 ) .
h. kleinert , a. pelster , b. kastening , and m. bachmann , phys . rev . * e 62 * , 1537 ( 2000 ) .
b. kastening , phys . rev . *
e 61 * , 3501 ( 2000 ) .
a. pelster and h. kleinert , physica * a * ( in press ) ; e - print : hep - th/0006153 . h. kleinert , a. pelster , and b. van den bossche , physica * a 312 * , 141 ( 2002 ) .
a. pelster and k. glaum , phys .
* b * ( in press ) , eprint : cond - mat/0211361 .
a. pelster and k. glaum , _ recursive graphical construction of tadpole - free feynman diagrams and their weights in @xmath0-theory _ ; in w. janke , a. pelster , h .- j .
schmidt , and m. bachmann ( editors ) : _ fluctuating paths and fields dedicated to hagen kleinert on the occasion of his 60th birthday _ ( world scientific , singapore , 2001 ) , p. 269 ; eprint : hep - th/0105193 .
r.f . streater and a.s .
wightman , _ pct , spin and statistics , and all that _
( w.a . benjamin , reading , massachusetts , 1964 ) . j. schwinger , _
particles , sources , and fields _ , vols .
i and ii ( addison - wesley , reading , 1973 ) . h. kleinert and v. schulte - frohlinde , _ critical properties of @xmath0-theories _ ( world scientific , singapore , 2001 ) .
k. glaum , ms thesis ( in german ) , fu - berlin ( 2001 ) .
j. neu , ms thesis ( in german ) , fu - berlin ( 1990 ) .
a. pelster , k. glaum , and h. kleinert , to be published .
h. kleinert , _ gauge fields in condensed matter _ , vol .
i , _ superflow and vortex lines _ ( world scientific , singapore , 1989 ) . b.g .
nickel , d.i .
meiron , and g.b .
baker jr . , university of guelph , preprint ( 1977 ) , http://www.physik.fu-berlin.de/~kleinert/kleiner_reb8/programs/programs.html .
s.a . antonenko and a.i .
sokolov , phys .
e 51 * , 1894 ( 1995 ) .
murray and b.g . nickel , university of guelph , preprint ( 1998 ) .
h. kleinert , j. neu , v. schulte - frohlinde , k.g .
chetyrkin , and s.a .
larin , phys . lett .
* b 272 * , 39 ( 1991 ) ; * 319 * , 545 ( e ) ( 1993 ) .
h. kleinert , fortschr . phys . * 30 * , 187 ( 1982 ) .
h. kleinert , fortschr . phys .
* 30 * , 351 ( 1982 ) .
vasiliev , _ functional methods in quantum field theory and statistical physics _
( gordon and breach science publishers , new york , 1998 ) ; translation from the russian edition ( st . petersburg university press , st .
petersburg , 1976 ) .
j. klbeck , m. bhm , and a. denner , comp .
phys . comm .
* 60 * , 165 ( 1991 ) .
t. hahn , comp .
140 * , 418 ( 2001 ) . .
p. nogueira , j. comput .
105 * , 279 ( 1993 ) . .
b.r . heap , j. math .
* 7 * , 1582 ( 1966 ) .
nagle , j. math .
* 7 * , 1588 ( 1966 ) .
s. schelstraete and h. verschelde , z. phys . *
c 67 * , 343 ( 1995 ) . k. kajantie , m. laine , and y. schrder , phys . rev .
* d 65 * , 045008 ( 2002 ) .
bray , phys .
lett . * 32 * , 1413 ( 1974 ) .
l. radzikovsky , europhys . lett .
* 29 * , 227 ( 1995 ) . | using functional derivatives with respect to the free correlation function we derive a closed set of schwinger - dyson equations in @xmath0-theory . its conversion to graphical recursion relations
allows us to systematically generate all connected and one - particle irreducible feynman diagrams for the two- and four - point function together with their weights . |
the next several years should be very exciting for cosmologists : microwave anisotropy probe ( map ; @xcite ) and planck ( @xcite ) will make high resolution maps of the microwave background sky ; while the sloan digital sky survey ( sdss ; @xcite ; see also http://www-sdss.fnal.gov:8000/ ) will measure redshifts of 10@xmath5 galaxies and positions of 10@xmath6 galaxies . in this paper , we explore the _ direct _ connection between these two measurements through gravitational lensing : the path of a cosmic microwave background ( cmb ) photon is distorted by inhomogeneities in the matter distribution ; galaxy surveys detect these inhomogeneities as fluctuations in galaxy number counts .
the effect of the gravitational lensing on the cmb anisotropies has been studied by many authors .
the uncomfortably low upper limits ( @xcite ) provoked a great deal of controversy ( @xcite ) about the possibility that gravitational lensing washes out the intrinsic fluctuation .
after the detection by the cosmic background explorer ( cobe ; @xcite ) , there has been renewal of interest ( @xcite , b , @xcite , b , @xcite ) in investigating how the cmb power spectrum is redistributed owing to gravitational lensing .
for example , seljak ( 1996 ) recently presented detailed calculations of gravitationally deflected cmb power spectra , including the effect of the nonlinear evolution of matter density fluctuations .
his result shows , however , that the modification of the cmb power spectrum is a second - order effect of the photon bending angle and less than a few percent on angular scales greater than ten arcminutes .
hence , the lensing effect on the cmb spectrum itself is extremely difficult to detect , even with observations such as the map project .
linder ( 1997 ) has also studied the effects of lensing on the correlation function and has introduced a cross - correlation function similar to the one that we study here . in this paper
, we introduce a cross - correlation function that is sensitive to the gravitational lensing correlations between the temperature fluctuations and matter density fluctuations .
we show that the cross - correlation is first - order in the bending angle so it should be easier to detect if we have both accurate cmb maps and redshift surveys .
we quantitatively estimate its magnitude and its cosmic variance in cold dark matter ( cdm ) universes .
the rest of the paper is organized as follows .
we review the formalism developed by seljak ( 1996 ) for computing the angular excursion of the cmb photon paths on celestial sphere in section 2 . in section 3
, we formulate the cross - correlation between matter density inhomogeneities and cmb temperature fluctuations .
section 4 concludes .
in this section , we review gravitational lensing by density fluctuations . we follow the power spectrum approach of seljak ( 1994 , 1996 ) .
we focus on the angular excursions produced by matter fluctuations at low redshifts , where they can be most easily inferred from redshift surveys .
fluctuations in matter density , @xmath7 , generate variations in the gravitational potential , @xmath8 where @xmath9 is the gravitational constant , and @xmath10 is the mean background mass density .
conventionally , the matter density fluctuations are related to the fluctuations in galaxy counts by a linear biasing parameter , @xmath11 : @xmath12 since most of the lensing effects will be produced by fluctuations on large physical scales ( @xmath13mpc ) , the linear biasing model will hopefully be valid .
it is important to note that detailed nonlinear and/or time - dependent biasing may somewhat change the statistics we present in this paper .
a photon emitted at some angular position @xmath14 has been deflected by gravitational lensing during its long travel , with the result that it is observed at different angular position , @xmath15 .
the photon angular excursion on celestial sphere is given by seljak ( 1994 ) : @xmath16 where @xmath17 is transverse component of the potential gradient with respect to the photon path , @xmath18 is a projection operator on celestial sphere , and @xmath19 is unperturbed comoving radial distance corresponding to redshift @xmath20 at decoupling time . in equation ( [ eqn : pro ] ) , @xmath21 and sinh(u ) in
a closed , flat , and open universe , respectively .
next , we consider the relative angular excursion @xmath22 of a photon pair emitted from angular positions @xmath23 and @xmath24 .
we restrict our calculation to the small angular separation limit , @xmath25 , and assume that the relative angular excursion @xmath22 obeys gaussian statistics .
lensing is primarily due to scattering events from mass fluctuations on the 10 - 100 mpc scale .
as there are 30 - 300 of these fluctuations between the surface of last scatter and the present along each photon path , the central limit theorem implies that this is a good approximation .
following seljak ( 1994 ) , we characterize the statistics of the lensing fluctuations by its root - mean - square dispersion : @xmath26 ^ 2\right\rangle _
\xi ^{1/2 } = \left [ c_{{\rm gl}}(0;z)-c_{{\rm gl}}(\xi;z)\right ] ^{1/2 } , \label{eqn : cgf } \\ c_{{\rm gl}}(\xi ; z ) & \equiv & \frac{2}{\pi}\int_0^\infty k^3dk\,\int_0^{\chi ( z)}d\chi ^{\prime}\ , p_\phi ( k,\tau _ 0-\chi ^{\prime } ) w^2(\chi^{\prime } , \chi _ { \rm dec } ) j_0(k\xi \sin _
k\chi ^{\prime } ) , \nonumber\end{aligned}\ ] ] where @xmath27 denotes the averaging over pairs observed with fixed angular separation @xmath28 , @xmath29 is the bessel function of order @xmath30 , and @xmath31 is the gravitational potential power spectrum .
the power spectrum of the potential fluctuations are related to the power spectrum of the density fluctuations through , @xmath32 where @xmath33 is the mass density parameter given by @xmath34 , and @xmath35 is the hubble parameter .
figures 1a shows @xmath36 as functions of @xmath37 for several redshift values .
we consider throughout this paper two cosmological models : one is the standard cdm ( scdm ) model with @xmath38 , @xmath39 , @xmath40 , and @xmath41 , and the other is a low - density , cosmological constant dominated cdm ( @xmath42cdm ) model with @xmath43 , @xmath44 , @xmath45 , and @xmath46 the best fit model of ostriker and steinhardt ( 1995 ) . here ,
@xmath47 and @xmath48 are the present mass density and the present vacuum energy density normalized by the critical density ; @xmath49 is the present hubble parameter in units of @xmath50 ; @xmath51 is the mass fluctuations within a sphere of radius @xmath52 .
we use the cobe normalized value ( @xcite ) for @xmath51 . in numerical calculation of @xmath53 ,
we have used the fitting formula for cdm linear transfer function given in bardeen et al .
( 1986 ) .
the sdss project measures the redshift of galaxies up to @xmath54 within solid angle @xmath55 .
furthermore , we expect to obtain photometric redshifts ( @xcite ) of galaxies up to @xmath56 with fairly small uncertainties .
then we will obtain the galaxy number density perturbation , @xmath57 , within solid angle @xmath58 for @xmath59 , and for @xmath60 with some uncertainties due to error in photometric redshifts .
we can see from figures 1b that the matter within @xmath59 , @xmath61 , and @xmath60 contributes 15% , 35% , and 55% , respectively , to the angular excursion from the last scattering surface .
hereafter , the values with subscript @xmath62 denote the integral from the observer to some redshift @xmath63 , the edge of the survey data , and those with subscript @xmath64 denote the integral over remaining part ; for example : @xmath65 the total lensing deviation , @xmath66 its variance is the sum of the contribution from the two regions : @xmath67 for our purposes , the distant lensing is unimportant ; its only effect is to slightly reduce the amplitude of the temperature fluctuations .
gravitational lensing distorts the microwave background sky : @xmath68 = t\left ( \mbox{\boldmath $ \psi $ } + \delta \mbox{\boldmath $ \theta $ } \right ) = t(\mbox{\boldmath $ \psi $ } ) + \delta \mbox{\boldmath $ \theta \cdot \nabla $ } t ( \mbox{\boldmath $ \psi $ } ) + \frac{1}{2}\left [ \delta \mbox{\boldmath $ \theta \cdot \nabla $ } \right]^2 t\left ( \mbox{\boldmath $ \psi $ } \right ) + \cdots .\ ] ] here @xmath69 denotes the measured temperature map and @xmath1 denotes the unlensed temperature map .
this distortion alters the statistics of the microwave background by smearing out temperature correlations ( seljak 1996 ) : @xmath70 \left [ \delta \mbox{\boldmath $ \theta $ } ^{\prime } \mbox{\boldmath $ \cdot \nabla$ } t(\mbox{\boldmath $ \psi $ } ^{\prime } ) \right ] \right\rangle \nonumber \\ & & \qquad + \left\langle t(\mbox{\boldmath $ \psi $ } ) ( \delta \mbox{\boldmath $ \theta $ } ^{\prime } \mbox{\boldmath $ \cdot \nabla$})^2 t ( \mbox{\boldmath $ \psi $ } ^{\prime } ) \right\rangle .\end{aligned}\ ] ] this can be alternatively written in terms of the correlation function : @xmath71 where @xmath72 if we have a redshift survey , then the effects of gravitational lensing of the microwave background can be observed more easily . here , we introduce a new quantity , @xmath73 which measures the cross - correlation between the temperature map and the predicted lensing : @xmath74 \mbox{\boldmath $ \cdot \nabla $ } \widetilde{t}(\mbox{\boldmath $ \psi $ } ^{\prime } ) \right ) \right\rangle _ \xi , \ ] ] where @xmath75 is determined from the redshift survey , and @xmath76 is a normalization factor given by @xmath77 at a typical value of angular separation @xmath78
. we can evaluate this by expanding the temperature map in a fourier series . since the lensing bending angles are small , we simplify the equations by making the plane parallel approximation and expanding out the unlensed temperature map : @xmath79 where @xmath80 denotes multipole moments .
we can now rewrite the effects of lensing in the multipole expansion : @xmath81 , \nonumber\end{aligned}\ ] ] and @xmath82 .\ ] ] we can now evaluate the lensing statistic : @xmath83 \exp \left ( -i\mbox{\boldmath $ l \cdot \psi $ } + i \mbox{\boldmath $ l$}^{\prime } \mbox{\boldmath $ \cdot \psi $ } ^{\prime } \right ) \bigr \rangle \\ & = & { \cal n } \sum_{\mbox{\boldmath $ l$ } } \left\langle a_{\mbox{\boldmath$l$ } } ^2 \right\rangle \exp \left [ -i\mbox{\boldmath $ l \,\cdot $ } \left ( \mbox{\boldmath $ \psi $ } -\mbox{\boldmath $ \psi $ } ^{\prime } \right ) \right ] \nonumber \\ & & \times \left\langle i\mbox{\boldmath $ l \,\cdot $ } \left ( \delta \mbox{\boldmath $ \theta $ } _ a-\delta \mbox{\boldmath
$ \theta $ } _ a^{\prime } \right ) \exp \left [ -i \mbox{\boldmath $ l \,\cdot $ } \left ( \delta \mbox{\boldmath $ \theta $ } _ a-
\delta \mbox{\boldmath $ \theta $ } _ a^{\prime } \right ) \right ] \right\rangle \nonumber \\ & & \times \left\langle \exp \left [ -i\mbox{\boldmath $ l \,\cdot$ } \left ( \delta \mbox{\boldmath $ \theta $ } _
b- \delta \mbox{\boldmath $ \theta $ } _ b^{\prime } \right ) \right ] \right\rangle .
\nonumber\end{aligned}\ ] ] since the gravitational deflections are the sum of many small scattering due to superclusters and voids along the line - of - sight , we can treat @xmath84 as a gaussian random variable .
the dispersion of @xmath85 is @xmath86 , where we have kept only the main isotropic term .
the anisotropic term makes a subdominant contribution to the gravitational lensing effect on two - point auto - correlation function of cmb ( @xcite ) .
though the anisotropic term is not too small in this statistics , the more rigorous analysis will be given in a subsequent paper ( @xcite ) .
then @xmath87 combining equations ( [ eqn : heq ] ) and ( [ eqn : gauss ] ) , @xmath88 \exp \left [ - i\mbox{\boldmath $ l \,\cdot$ } \left ( \mbox{\boldmath $ \psi $ } - \mbox{\boldmath $ \psi $ } ^{\prime } \right ) \right ] \nonumber \\ & & \times \exp \left [ -\frac{l^2\sigma _ b^2(\xi ) } 2\right ] .\end{aligned}\ ] ] the final term represents the smearing of the microwave fluctuations due to density fluctuations beyond the edge of the redshift survey .
this effect is a small correction term , which we will ignore for the remainder of the paper .
we average equation ( [ eqn : hxilvec ] ) over angle and rewrite it on the celestial sphere , then @xmath89 p_l(\cos \xi ) , \label{eqn : signal}\ ] ] where @xmath90 is the angular separation on the sky , @xmath91 is the usual multipole moment , @xmath92 is the window function , and @xmath93 is the beam of the detector ( @xcite )
. the cosmic variance in the cross - correlation statistic can also be estimated by taking the leading order term in the expansion : @xmath94 \mbox{\boldmath $ \cdot \nabla $ } \widetilde{t } ( \mbox{\boldmath $ \psi $ } ^{\prime } ) \right ) \right ] ^2\right\rangle \\ & \simeq & { \cal n}^2 \sigma _ a^2\left ( \xi \right ) \left\langle t^2\right\rangle \left\langle \left ( \nabla t\right ) ^2\right\rangle . \nonumber\end{aligned}\ ] ] note that the signal - to - cosmic variance scales as @xmath95 @xmath96 which is proportional to @xmath97 as we claimed earlier , this is a first - order effect .
we can estimate the signal - to - cosmic variance ratio by computing the predicted signal per multipole : @xmath98 \left [ i \mbox{\boldmath $ l \,\cdot $ } \left ( \delta \mbox{\boldmath $ \theta $ } ( \mbox{\boldmath $ \psi $ } ) -\delta \mbox{\boldmath $ \theta$}^{\prime}(\mbox{\boldmath $ \psi $ } ^{\prime } ) \right ) \right ] \nonumber \\ & & \times a_{\mbox{\boldmath $ l$}}^{\ast } \exp \left [ i\mbox{\boldmath $ l \,\cdot$ } \left ( \mbox{\boldmath $ \psi $ } ^{\prime } + \delta\mbox{\boldmath $ \theta $ } ^{\prime } \right ) \right ] \\ & \simeq & c_lw_l\,l^2\sigma _ l^2\exp ( -l^2\sigma _ l^2/2 ) , \nonumber\end{aligned}\ ] ] where @xmath99 if we include detector noise , then @xmath100 where @xmath101 is the measure of the detector noise ( @xcite ) .
map s highest frequency channel has a full width at half maximum of @xmath102 and a system noise of @xmath103 . by combining the three highest frequency channel , the map system noise drop to @xmath104 .
this implies that the noise plus cosmic variance in @xmath105 is @xmath106 where the factor of @xmath107 comes from averaging over all multipoles with the same @xmath108 note that there is no factor of 2 in the numerator as the variance is proportional to the product of two uncorrelated fields , @xmath109 , rather than the more familiar @xmath110 @xmath111 @xmath112 summing over all multipoles yields the expected signal in the microwave maps : @xmath113 ^2}. \nonumber\end{aligned}\ ] ] note that an additional cross - correlation function that can be computed from the microwave background fluctuations and the lensing maps : @xmath114 \mbox{\boldmath $ \cdot \,$}\mbox{\boldmath $ \widehat{n}$ } \widetilde{t}(\mbox{\boldmath $ \psi $ } ^{\prime } ) \right ) \right\rangle _ \xi , \ ] ] where @xmath115 the statistics has the simplest form among possible inclusions of gravitational lensing angular excursion .
the cross - correlation function is reduced to @xmath116 however we confirmed that the expected signal - to - noise ratio , @xmath117 , in @xmath9 is lower than in @xmath118 .
this is because , unlike in @xmath118 , the cosmic variance in @xmath9 contains a term proportional to @xmath119 .
we have computed the expected cross - correlation between the temperature fluctuations and the lensing bending angle ( equation ( [ eqn : signal ] ) ) for the standard cdm model and the `` best fit '' vacuum dominated model ( @xcite ) . in estimating the lensing bending angle , we have assumed that we have a redshift survey that extends to a characteristic redshift @xmath120 figure 2 shows the results for large scale surveys of varying depths . in this figure , we have assumed the standard parameters for the map and sdss projects .
the characteristic depth of the sdss redshift survey is @xmath121 sdsshas photometric redshifts for 10@xmath6 galaxies that should extend the survey to @xmath122 these photometric redshifts are accurate to @xmath123 in redshift ; certainly accurate enough to compute the projected surface densities needed to predict lensing .
figure 3 shows the cumulative signal - to - noise , where we have summed over all of the multipoles .
we have divided @xmath124 in equation ( [ eqn : snr ] ) by a factor of @xmath125 taking into account the limited sky coverage , @xmath126 . the predicted signal - to - noise is quite large ( 15 and 35 ) for the vacuum - dominated and standard cdm models .
this cross - correlation , if detected , directly probes the gravitational potential fluctuations at low redshift . in principle
, it should yield an accurate determination of the biasing factor .
armed with this measurement , we should be able to directly compare the gravitational potential fluctuations at decoupling with the gravitational potential fluctuations in the local universe .
we thank seljak and zaldarriaga for providing their code to generate the intrinsic cmb power spectrum .
dns acknowledges the map / midex project for support .
ms and ts acknowledge support from research fellowships of the japan society for the promotion of science . 99 bardeen , j. m. , bond , j. r. , kaiser , n. , and szalay , a. s. 1986 , , 304 , 15 bouchet , f. r. , gispert , r. , and puget , j .- l .
1995 , clustering in the universe , s. maurogordato et al .
, sigapore : frontieres , 537 bunn , e. f. , and white , m. 1997 , , 480 , 6 cayn , l. , martnez - gonzlez , e. , and sanz , j. l. 1993a , , 403 , 471 cayn , l. , martnez - gonzlez , e. , and sanz , j. l. 1993b , , 413 , 10 connolly , a. j. , csabai , i. , szalay , a. s. , koo , d. c. , kron , r. g. , munn , j. a. 1995 , , 110 , 2655 gunn , j. e. , and weinberg , d. h. 1995 , wide field spectroscopy and the distant universe , s. j. maddox and a. aragn - salamanca , singapore : world scientific , 3 kashlinsky , a. 1988 , , 331 , l1 knox , l. 1995 , , 52 , 4307 linder , e. v. 1990a , , 243 , 353 linder , e. v. 1990b , , 243 , 362 linder , e. v. 1997 , , 323 , 305 martnez - gonzlez , e. , sanz , j. l. , and cayn , l. 1997 , , 484 , 1 ostriker , j. p. , and steinhardt , p. , j. 1995 , , 377 , 600 readhead , a. c. s. , lawrence , c. r. , myers , s. t. , sargent , w. l. w. , hardebeck , h. e. , and moffet , a. t. 1989 , , 346 , 566 sasaki , m. 1989 , , 240 , 415 seljak , u. 1994 , , 436 , 509 seljak , u. 1996 , , 463 , 1 seljak , u. , and zaldarriaga , m. 1996 , , 469 , 437 smoot , g. f. et al .
1992 , , 396 , l1 suginohara , m. , suginohara , t. , and spergel , d. n. 1997 , in preparation tomita , k. 1988 , , 40 , 751 uson , j. m. , and wilkinson , d. t. 1984 , , 283 , 471 watanabe , k. , and tomita , k. 1991 , , 370 , 481 wright , e. l. , hinshaw , g. , and bennett , c. l. 1996 , , 458 , l53 | density inhomogeneities along the line - of - sight distort fluctuations in the cosmic microwave background .
usually , this effect is thought of as a small second - order effect that mildly alters the statistics of the microwave background fluctuations .
we show that there is a first - order effect that is potentially observable if we combine microwave background maps with large redshift surveys .
we introduce a new quantity that measures this lensing effect , @xmath0 where @xmath1 is the microwave background temperature and @xmath2 is the lensing due to matter in the region probed by the redshift survey .
we show that the expected signal is first order in the gravitational lensing bending angle , @xmath3 and find that it should be easily detectable , ( s / n)@xmath4 if we combine the microwave anisotropy probe satellite and sloan digital sky survey data . measurements of this cross - correlation will directly probe the `` bias '' factor , the relationship between fluctuations in mass and fluctuations in galaxy counts . |
supershells are large gaseous shells of radius greater than a few hundred parsecs .
they are distinct from other shell - like structures in their extraordinarily large energy requirement , i.e. , @xmath3 erg , which corresponds to @xmath4 supernova ( sn ) explosions @xcite .
about twenty supershells have been found in the milky way , and numerous neutral atomic hydrogen ( ) holes corresponding to supershells have been discovered in nearby dwarfs and spiral galaxies @xcite .
these gigantic structures are generally thought to arise from multiple sn explosions in stellar ob associations .
but most supershells are missing a stellar association in their interior , and the number of supershells and their energies are usually incompatible with the level of star formation in those galaxies @xcite .
therefore , several alternative scenarios have been proposed , the most popular of which is the collision of high - velocity clouds ( hvcs ) with the disk @xcite .
hvcs are clouds with radial velocities very different from the disk material in the milky way , e.g. , with a deviation more than 50 from the range of permitted velocities in a simple model of the distribution and rotation of the gas in the galaxy @xcite .
some large hvc complexes are known to be gas streams tidally stripped from satellite galaxies of the milky way , but the origin of isolated compact hvcs ( chvcs ) remain controversial : they could be clouds formed from galactic fountain or intergalactic accreting flows , part of the large hvc complexes , or condensations in the multi - phase circumgalactic medium @xcite .
the hvc origin has been proposed for a few supershells @xcite , but there has been no clear example showing a direct link between the two , particularly for chvcs . here
we report the detection of a galactic supershell with an associated hvc , gs040.2 + 00.6@xmath170 ( hereafter gs040 ) .
gs040 was first identified as a faint , forbidden - velocity wing feature ( fvw 40.0 + 0.5 ) in the low - resolution , large - scale longitude - velocity study of @xcite .
we have found that gs040 appears to be a complete circular ring with complicated structures inside in our high - resolution i - galfa ( inner - galaxy alfa ) 21-cm line survey data .
the i - galfa survey is a survey of the first galactic quadrant visible to arecibo ( @xmath5 to 77 and @xmath6 ) done by using the 7-beam arecibo l - band feed array ( alfa ) receiver on the arecibo 305 m telescope , and it provides sensitive ( @xmath7 k ) and fully - sampled maps at spatial and spectral resolutions of 4 and 0.184 , respectively @xcite .
the i - galfa survey data further reveal that there is a chvc at the very center of gs040 .
this chvc , named hvc 040 + 01@xmath1282 ( hereafter chvc040 ) , was first identified in the leiden / dwingeloo survey @xcite and was later classified as an isolated chvc by @xcite and @xcite .
@xcite presented a higher - resolution ( 9 ) image obtained from the effelsberg telescope , which showed that chvc040 has a pronounced head - tail structure .
our arecibo images reveal detailed spatial and velocity structure of chvc040 strongly suggesting its association with the supershell gs040 .
we describe two structures in section [ sec : targets ] , and discuss their physical characteristics and their association together with some implications on the disruption of hvcs in section [ sec : disc ] .
the supershell gs040 is centered at @xmath8 and clearly visible from @xmath9 to @xmath10 . in an integrated intensity map ( figure [ fig : itg_gs040 ] )
, gs040 appears as a complete circular ring of radius @xmath11 with complicated structures inside .
more detailed structures can be seen in figure [ fig : chmap_gs040 ] , which presents velocity channel maps . at the most negative velocities ( @xmath12 )
, we see diffuse emission with embedded knotty filaments near the center and an extended filament in the south .
the features are observable at even more negative velocities ( @xmath13 @xmath14 ) but they are extremely weak and hard to detect in velocity - channel maps . as the velocity increases ( @xmath15 ) , the nebulosity fades out , and a larger ( @xmath16 ) ring structure appears , which resembles a cartwheel with a bright central `` hub '' and several `` spokes '' ( see also figure [ fig : itg_gs040 ] ) .
the size of the ring increases slightly with velocity indicating that the ring structure is an approaching portion of an expanding shell . at velocities greater than about @xmath10 ,
the galactic background emission becomes dominant , and the emission associated with gs040 is less clear . at the very center of gs040
is the hvc chvc040 .
the positional coincidence of chvc040 with the gs040 s central hub is striking as can be seen in figure [ fig : pvmap ] , which is the position - velocity map crossing the center of gs040 .
the morphological agreement between the two is also noticeable .
our high - resolution arecibo image reveals that chvc040 has a blunt cone shape with a steep southwestern boundary and a faint envelope flaring out northeast ( figure [ fig : hvc ] ) .
this morphology of chvc040 matches well with that of the central hub of gs040 , e.g. , see figure [ fig : itg_gs040 ] and also the channel map at @xmath17 in figure [ fig : chmap_gs040 ] .
figure [ fig : hvc ] shows the detailed spatial and velocity structures of chvc040 .
the integrated intensity map in the left frame shows that chvc040 has a bright , @xmath18- sized `` core '' elongated along the northeast - southwest direction .
the core appears to be composed of several clumps with a sharp boundary at southwest , whereas the diffuse envelope appears to be slightly more extended toward southeast .
the mean column density , assuming that the emission is optically thin , is 1.5@xmath19 while the peak column density is about two times higher .
the velocity centroid map in the middle frame shows that systemic velocity attains its most negative ( @xmath20 ) at the southwestern boundary and most positive ( @xmath21 ) along the southeastern boundary and also in the middle of the northwestern boundary .
the mean systematic velocity is @xmath22 .
if the most positive feature in the northwestern boundary were not present , the velocity structure could have been suggestive of a rotation with respect to the northeast - southwest symmetry axis at speed of @xmath23 .
the velocity width ( full width at half maximum ; fwhm ) ranges 2545 over most parts of the cloud with a median of 36 .
the southwesternmost thin layer has relatively narrow width ( @xmath241928 ) while parts of the northern area has width @xmath25 .
the region with the narrowest line width is very thin ( @xmath26 ) , and therefore @xcite could not spatially resolve it .
some representative line profiles are shown in figure [ fig : hvc_prf ] together with the average line profile of the cloud .
most profiles are well described with a single gaussian component , while some profiles , e.g. , c and e - g , show clear double peaks or a narrow component superposed on a broad component .
the velocity width of the mean profile is 45 , which is considerably larger than the median value ( 36 ) of individual profiles , presumably due to the dispersion of central velocities .
the median width corresponds to a kinetic temperature of @xmath27 k. at several positions , narrow velocity components are detected , but still their widths are @xmath28 , which implies a kinetic temperature @xmath29 k. this indicates that chvc040 is mostly composed of warm neutral gas , which is not unusual for chvcs @xcite .
but it is different from those head - tail hvcs with a narrow - line head of undisturbed cold gas and a wide - line tail of disturbed warm gas ( e.g. , hvc125 + 41 - 207 ; * ? ? ?
the location of chvc040 at the geometrical center of gs040 suggests that their physical association is very likely .
the centroids of the chvc emission and that of the gs040 s hub emission overlap within @xmath30 degrees .
the probability of this being a random alignment is @xmath31 , and multiplying this by @xmath32 chvcs yields an overall probability of @xmath33 for any chvc aligning this well with the gs040 s hub .
no intermediate - velocity connecting the two is apparent ( figure [ fig : pvmap ] ) , but this could be because the gas is ionized .
we searched for a warm ionized gas associated with gs040 using the wisconsin h@xmath34 mapper northern sky survey ( wham - nss ) data @xcite .
the survey has an angular resolution of 1 , and , around the gs040 area , it provides h@xmath34 spectra covering @xmath35 from @xmath36 to @xmath37 at spectral resolutions of 12 .
we have examined the h@xmath34 intensity map integrated over the velocity range of gs040 ( @xmath36 to @xmath38 ) , but could not detect an associated emission ( @xmath39 0.05 @xmath40 where @xmath41 ) .
note that this area is bright in the total h@xmath34 intensity map with a mean intensity of about 3 @xmath40 , so that we do not expect to see the faint emission associated with either gs040 or chvc040 in the all - sky h@xmath34 maps @xcite .
we also searched for an associated hot ionized gas using the 0.1 - 2.4 kev image of the rosat all - sky x - ray survey ( 1 pixel scale = 44 ; * ? ? ? * ) , but could nt detect any emission .
chvc040 belongs to the `` galactic center negative '' ( gcn ) hvc complex , which is a collection of small discrete hvcs sparsely distributed over a @xmath42 area within @xmath43 to @xmath44 and @xmath45 to 10@xmath46 @xcite .
a kinematic model has been proposed where gcn is a smooth gas flow starting at @xmath45 at a heliocentric distance of 35 kpc and crossing the galactic plane obliquely at 15 kpc @xcite . for comparison
, gs040 is probably located near the scutum - centaurus ( sct - cen ) arm at a distance of @xmath47 kpc @xcite because the disrupted interstellar medium ( ism ) is seen only at velocities below that of the sct - cen arm ( @xmath48 ) , not at higher velocities ( see figures [ fig : chmap_gs040 ] and [ fig : pvmap ] ) .
we examined lists of known stellar objects , regions , ob stars , and sn remnants ( snrs ) , but no known sources are likely to be associated with gs040 .
there is one region ( g@xmath49 ) in the _ wise _ catalog of galactic regions @xcite that is located at @xmath50 west of the hub . but
this region has a systematic velocity of @xmath51 @xcite and is enclosed by a small ( 68 ) dust shell , so it can not be responsible for the hi shell . instead the distance of @xmath47 kpc is not unreasonable for chvc040 , because gcn does not have a smooth extended envelope like other hvc complexes , and it appears to be composed of several subpopulations that do not share a common origin @xcite .
note that the geometrical center of gs040 is well below the galactic plane ( @xmath52 pc where @xmath53 ) , i.e. , at @xmath54 while the midplane there is at @xmath55 because the galactic plane is warped in the outer galaxy @xcite .
it is difficult to imagine the sn origin for a supershell at such height , and chvc040 is most likely the energy / momentum source for gs040 .
the total energy deposited ( @xmath56 ) in the galactic disk by chvc040 can be inferred from the parameters of the gs040 supershell .
the radius of gs040 is @xmath57 pc , while its mass at @xmath58 is @xmath59 including the cosmic abundance of helium .
if we account for the mass unobservable due to galactic background emission , the total mass of gs040 would be considerably greater .
adopting @xmath60 as the expansion speed of the shell , its kinetic energy is @xmath61 erg .
the collision should have occurred @xmath62 yr ago , where the numerical factor @xmath63 accounts for the deceleration of the shell . note that @xmath64 is a small fraction of the total energy deposited ( @xmath56 ) ,
most of which should have been radiated away . if gs040 was produced by multiple sne , then , assuming instantaneous energy injection @xcite , @xmath65 erg , where @xmath66 is ambient hydrogen density . at the position of gs040 , the mean density in the midplane is @xmath67 , and the scale height is about 720 pc @xcite .
so @xmath68 , and we have @xmath69 erg . for comparison , the total extent of chvc040 is @xmath70 while its mass is @xmath71 .
the area used to derive this mass has a geometrical mean radius of 150 pc , so that the mean hydrogen density of chvc040 is 0.017 @xmath72 .
if chvc040 collided with the rotating disk with the mean ` deviation ' speed ( absolute difference in velocity from the disk gas there ) of hvcs , i.e. , @xmath73 @xcite , the kinetic energy and momentum would be @xmath74 ergs and @xmath75 including the he abundance , respectively .
this implies that the mass of chvc040 when it collided with the disk should have been an order of magnitude greater and that the hvc that we see ( chvc040 ) might be the remains of the original hvc .
the spatial morphology and velocity structure of chvc040 suggest that it is moving southwest in the plane of the sky and is approaching us .
the steep southwestern boundary of the cloud might represent the region compressed by the interaction with the ambient medium , whereas the diffuse envelope might be the material stripped off from the cloud due to the interaction or by the ram pressure of the surrounding medium @xcite .
the small velocity width along the southwestern boundary and large velocity width beyond appear to be consistent with such speculation .
recently , @xcite suggested that , for a chvc with head - tail structure , the inclination angle , i.e. , the angle between the chvc s trajectory and the line - of - sight , can be derived from the asymmetry in position - velocity diagram along the head - tail line crossing the center of mass .
the morphology of chvc040 is different from typical head - tail chvcs , i.e. , chvc040 has a wide flaring ` tail ' not a narrow elongated tail ( see also 2 ) , and its core is fragmented , so that it is not obvious if their model can be applied .
nevertheless if we assume that the center of mass is near the southwestern boundary and apply their model ( their equations 14 ) to the position - velocity diagram along the red - dashed line in figure 1 , we obtain an inclination angle of @xmath76 which is consistent with our expectation .
the shape of chvc040 appears to be pointed away from the galactic midplane as if it already went through the midplane .
but the geometrical center of the supershell coincides with the current location of chvc040 , not being located in the midplane .
perhaps chvc040 is approaching at an angle to the warped galactic disk and has not yet fully penetrated the disk to the midplane .
it is worth to note that , according to the @xcite s model , the gcn hvc stream is colliding to the galactic plane almost perpendicularly _ from below _ , while the head - tail directions of individual hvcs seem to indicate that there is no preferential direction in their motions @xcite .
therefore , the orbit of chvc040 is uncertain . the colliding geometry and the origin of the complex structures such as the hub and spokes in gs040 need to be explored .
there are about three hundred known chvcs in the milky way @xcite .
a considerable fraction of chvcs has a head - tail structure indicating a ram pressure interaction with the diffuse galactic halo gas @xcite .
an important question is whether they are totally dissipated in the galactic halo to feed the multi - phase circumgalactic medium or they can survive their trip through the halo ( e.g. , * ? ? ?
* ) . since chvc040 is located in the far outer galaxy , it may be of extragalactic origin rather than originating from a galactic fountain , although it is not clear whether chvc040 was originally a fragment of a nearby tidally disrupted galaxy or a cold cloud formed in a larger accreting flow of ionized , low - metallicity intergalactic gas .
our result then directly shows that at least some chvcs of extragalactic origin do survive and collide with the galactic disk .
according to numerical studies , chvcs with masses @xmath77 would be totally disrupted in the galactic halo unless they are embedded in dark matter @xcite .
but dynamical shielding by an extended diffuse gaseous component can significantly extend their lifetime @xcite .
we have checked whether there are additional sources like the chvc040-gs040 system in the i - galfa data using the hvc catalog of @xcite .
there are twelve chvcs in their table 2 including chvc040 in the i - galfa survey area , most of which are at relatively high latitudes .
they are isolated hvcs and have @xmath78 , so they belong to the gcn hvc complex .
we see low - velocity features around some chvcs , but none of them appear associated .
a systematic study against all chvcs may reveal other chvc - supershell systems .
anderson , l. d. , bania , t. m. , balser , d. s. , & rood , r. t. 2011 , , 194 , 32 anderson , l. d. , bania , t. m. , balser , d. s. , et al .
2014 , , 212 , 1 bagetakos , i. , brinks , e. , walter , f. , et al .
2011 , , 141 , 23 braun , r. , & burton , w. b. 1999 , , 341 , 437 brns , c. , kerp , j. , & pagels , a. 2001 , , 370 , l26 dame , t. m. , & thaddeus , p. 2011
, , 734 , l24 de heij , v. , braun , r. , & burton , w. b. 2002 , , 391 , 159 dennison , b. , simonetti , j. h. , & topasna , g. a. 1998 , , 15 , 147 draine , b. t. 2011 , _ physics of the interstellar and intergalactic medium , ed . b. t. draine _ ( princeton univ . press ) faridani , s. , fler , l. , kerp , j. , & westmeier , t. 2014 , , 563 , a99 finkbeiner , d. p. 2003
, , 146 , 407 gibson , s. j. , koo , b. , douglas , k. a. , et al . 2012 , american astronomical society meeting abstracts , 219 , # 349.29 haffner , l. m. , reynolds , r. j. , tufte , s. l. , et al .
2003 , , 149 , 405 heiles , c. 1979 , , 229 , 533 heiles , c. 1984 , , 55 , 585 heitsch , f. , & putman , m. e. 2009 , , 698 , 1485 heitsch , f. , bartell , b. , clark , s. e. , et al .
2016 , arxiv:1606.06689 jin , s. 2010 , , 408 , l85 kang , j .- h . , & koo , b .-
2007 , , 173 , 85 kamphuis , j. , sancisi , r. , & van der hulst , t. 1991 , , 244 , l29 koo , b .- c . , heiles , c. , stanimirovi , s. , & troland , t. 2010 , , 140 , 262 kwak , k. , shelton , r. l. , & raley , e. a. 2009 , , 699 , 1775 levine , e. s. , blitz , l. , & heiles , c. 2006 , , 643 , 881 mcclure - griffiths , n. m. , dickey , j. m. , gaensler , b. m. , & green , a. j. 2002 , , 578 , 176 mirabel , i. f. , & morras , r. 1990 , , 356 , 130 plckinger , s. , & hensler , g. 2012 , , 547 , aa43 putman , m. e. , de heij , v. , staveley - smith , l. , et al .
2002 , , 123 , 873 putman , m. e. , saul , d. r. , & mets , e. 2011 , , 418 , 1575 putman , m. e. , peek , j. e. g. , & joung , m. r. 2012 , , 50 , 491 rhode , k. l. , salzer , j. j. , westpfahl , d. j. , & radice , l. a. 1999 , , 118 , 323 santilln , a. , franco , j. , martos , m. , & kim , j. 1999 , , 515 , 657 tamanaha , c. m. 1997 , , 109 , 139 tenorio - tagle , g. 1980 , , 88 , 61 tenorio - tagle , g. , franco , j. , bodenheimer , p. , & rozyczka , m. 1987 , , 179 , 219 voges , w. , aschenbach , b. , boller , t. , et al . 1999 , , 349 , 389 wakker , b. p. 1991
, , 250 , 499 wakker , b. p. , & van woerden , h. 1991 , , 250 , 509 wakker , b. p. 2004
, , 289 , 381 westmeier , t. , brns , c. , & kerp , j. 2005 , , 432 , 937 winkel , b. , ben bekhti , n. , darmstdter , v. , et al .
2011 , , 533 , a105 | neutral atomic hydrogen ( ) gas in interstellar space is largely organized into filaments , loops , and shells , the most prominent of which are `` supershells '' .
these gigantic structures requiring @xmath0 erg to form are generally thought to be produced by either the explosion of multiple supernovae ( sne ) in ob associations or alternatively by the impact of high - velocity clouds ( hvcs ) falling to the galactic disk .
here we report the detection of a kiloparsec ( kpc)-size supershell in the outskirts of the milky way with the compact hvc 040 + 01@xmath1282 ( hereafter chvc040 ) at its geometrical center using the `` inner - galaxy arecibo l - band feed array '' 21-cm survey data .
the morphological and physical properties of both objects suggest that chvc040 , which is either a fragment of a nearby disrupted galaxy or a cloud originated from an intergalactic accreting flow , collided with the disk @xmath2 myrs ago to form the supershell .
our result shows that some compact hvcs can survive their trip through the galactic halo and inject energy and momentum into the milky way disk . |
if we fix a finite field @xmath0 , we can ask about the distribution of the number of ( @xmath0-rational ) points on a random curve over @xmath0 . there has been a surge of recent activity on this question including definitive answers of kurlberg and rudnick for hyperelliptic curves @xcite , of bucur , david , feigon and laln for cyclic @xmath2-fold covers of @xmath3 @xcite , of bucur , david , feigon and laln for plane curves @xcite , and of bucur and kedlaya on curves that are complete intersections in smooth quasiprojective subschemes of @xmath8 @xcite . in the first three cases ,
the average number of points on a curve in the family is @xmath4 .
in contrast , for curves that are complete intersections in @xmath8 , the average number of points is @xmath5 , despite , as pointed out by bucur and kedlaya , the abundance of @xmath0 points lying around in @xmath8 . in this paper , we give the distribution of the number of points on trigonal curves over @xmath0 ( i.e. curves with a degree @xmath9 map to @xmath3 ) , and in particular show that the average number of points is greater that @xmath4 .
let @xmath10 our main theorem is the following .
[ t : main ] let @xmath0 have characteristic @xmath11 .
we have @xmath12 where the @xmath13 are independent identically distributed random variables and @xmath14 moreover , we give a rigorous explanation of the random variables @xmath13 .
theorem [ t : main ] is a corollary of the following , which gives the distribution of the number of points in the fiber over a given @xmath0-rational point of @xmath3 .
[ t : bypoint ] let @xmath0 have characteristic @xmath11 . given a point @xmath15 , @xmath16 moreover , these probabilities ( of various size fibers ) are independent at the @xmath0 points @xmath17 of @xmath3 .
the average number of points of a random trigonal curve over @xmath0 ( in the @xmath18 limit as above ) is @xmath1 .
the method used in this paper will be to relate trigonal curves to cubic extensions of function fields , and then to use the work of datskovsky and wright @xcite to count cubic extensions with every possible fiberwise behavior above each rational point of the base curve .
in fact , our methods work with any smooth curve @xmath19 over @xmath0 replacing @xmath3 .
let @xmath20 be the moduli space of genus @xmath21 curves with a specified degree @xmath9 map to @xmath19 .
[ t : ebypoint ] let @xmath0 have characteristic @xmath11 . given a point @xmath22 , @xmath23 moreover , these probabilities ( of various size fibers ) are independent at the @xmath0 points of @xmath19 .
in particular , the average number of points of a random curve over @xmath0 with a degree @xmath9 map to @xmath19 ( in the @xmath18 limit ) is @xmath24 . studying the distribution of the number of points on a curve is equivalent to studying the distribution of the trace of frobenius on the @xmath25-adic cohomology group @xmath26 .
bucur , david , feigon and laln , for cyclic @xmath2-fold covers of @xmath3 , in fact give the finer information of the distribution of the trace of frobenius on each subspace of @xmath26 invariant under the cyclic action .
their methods , under appropriate interpretation , use kummer theory to enumerate the cyclic @xmath2-fold covers of @xmath3 , and thus require the hypothesis that @xmath27 .
the hardest part of the method is sieving for @xmath2-power free polynomials .
the work of bucur , david , feigon and laln for plane curves @xcite and of bucur and kedlaya on curves that are complete intersections @xcite is also given `` fiberwise , '' ( as in theorems [ t : bypoint ] and [ t : ebypoint ] of this paper ) in that it computes the probability that any point is the ambient space is in a random smooth curve in the specified family and shows these probabilities are independent .
( here we think of the embedding of a curve @xmath28 , and @xcite computes for each point @xmath29 the distribution of the size of the fiber @xmath30 for random @xmath31 .
note that since @xmath32 is an embedding , in this case the fiber has either 0 or 1 points , so these are bernoulli random variables . )
kurlberg and wigman @xcite have also studied the distribution of the number of @xmath0 points in families of curves in which the average number of points is infinite . in section [ s :
further ] of this paper , we discuss distribution of points on other families of @xmath6-gonal curves , and in particular give a conjecture for the expected number of points on a random @xmath6-gonal curve with full @xmath7 monodromy , based on function field analogs of bhargava s heuristics for counting number fields .
the conjecture in fact predicts the expected number of points in a fixed fiber over @xmath3 .
when one instead fixes a genus @xmath21 and lets @xmath33 tend to infinity , the philosophy of katz and sarnark @xcite predicts how the average number of points behaves , and in particular it should be governed by statistics of random matrices in a group depending on the monodromy of the moduli space of curves under consideration .
let @xmath0 have characteristic @xmath34 .
let @xmath19 be a smooth curve over @xmath0 , and let @xmath35 be the function field of @xmath19 .
then smooth , integral curves over @xmath0 with finite , degree 3 maps to @xmath36 are in one - to - one correspondence with cubic extensions @xmath37 of @xmath35
. however , these smooth , integral curves may not be geometrically integral . in this case
, the only such possibility is given by a cubic extension of the constant field of @xmath19 , and since we will take a @xmath18 limit , we can ignore this extension altogether .
we now consider all the possible completions of a cubic extension @xmath37 of @xmath35 . for a place @xmath38 of @xmath35 , the absolute tame galois group of the local field @xmath39 is topologically generated by @xmath40 with the relation @xmath41 where @xmath42 is a generator of the inertia subgroup and @xmath43 is a frobenius element . for a place @xmath38 of @xmath35 , the following is a chart of all possible cubic tale @xmath39 algebras @xmath44 ( and thus of all possible isomorphism types for @xmath45 ) . to each isomorphism type
, we note the images of @xmath40 from the absolute galois group of @xmath39 in the associated representation to @xmath46 ( given by the action of galois on the three homomorphisms @xmath47 ) , the number of @xmath0 rational points in the fiber above @xmath38 , and a constant @xmath48 that will be important later .
( it turns out that @xmath49 , but this fact will not be used . ) [ cols= " < , < , < ,
< , < " , ] let @xmath50 be the set of places of @xmath35 corresponding to the @xmath0-rational points of @xmath19 .
let @xmath51 be a choice @xmath52 of a cubic tale @xmath39 algebra for each @xmath53 .
we define @xmath54 , where the constants @xmath55 are defined by the chart above .
we also choose a @xmath56 similarly .
let @xmath57 be the number of isomorphism classes of cubic extensions @xmath37 of @xmath35 such that for all @xmath53 , we have @xmath58 , and such that the norm of the relative discriminant @xmath59 now we can state the result of datskovsky and wright and cubic extensions of function fields . [ t : dw ] in the above notation @xmath60 theorem [ t : ebypoint ] now follows because the above chart gives all the possibilities for the completions of cubic extensions @xmath37 of @xmath35 at each @xmath0-rational place of @xmath35 ( and the thus determined number of @xmath0-rational points in that fiber ) , and theorem [ t : dw ] gives their relative probabilities . for a triple cover @xmath61
corresponding to an extension @xmath62 , we have @xmath63 , where @xmath64 , and thus we can replace the limit in @xmath6 with a limit in @xmath21 .
the above computations of the distribution of points of trigonal curves suggests that similar questions could be attacked for @xmath6-fold covers of @xmath3 ( or an arbitrary curve ) using methods from the study of counting extensions of global fields by their discriminants , which has been more heavily studied in the case of number fields . in analog to this work , one would naturally only study @xmath6-gonal curves with a specified monodromy group ( i.e. the galois group of the galois closure of the field extension ) .
however , for this method to apply , one needs information on the densities of local behaviors of those number fields . in the number field case ,
such results are only currently available for cubic extensions ( the work of datskovsky and wright @xcite used above , or going back to davenport and heilbronn @xcite over @xmath65 ) , for @xmath66 galois extensions ( @xcite ) , and for degree @xmath6 extensions with galois closure group @xmath7 and @xmath67 ) ( by work of bhargava @xcite ) .
the first two cases above also have results in the function field case , which have been applied in this paper and forthcoming work of the author , respectively .
it is intriguing to ask if a function field version of bhargava s counting results @xcite could give the distribution of the number of points of @xmath68-gonal and @xmath69-gonal curves . in analogy with the number field case ,
one expects such methods would only count @xmath68-gonal curves with @xmath70 monodromy , and that @xmath71-monodromy curves would be a non - negligible portion of all @xmath68-gonal curves . since the methods for counting @xmath71 quartic extensions have not given any results on densities of local behaviors or independence of those behaviors , understanding the distribution of points of @xmath68-gonal curves with @xmath71 monodromy would be very interesting . in the case of @xmath6-gonal curves with full @xmath7-monodromy , we can at least predict what the distribution of the number of points in each fiber should be using the heuristics of bhargava ( * ? ? ? * conjecture 5.1 ) . in particular , we give the prediction for the average .
[ c ] let @xmath19 be a smooth curve over a finite field @xmath0 of characteristic @xmath72 and fix @xmath22 .
the average number of points in the fiber over @xmath73 of a random curve over @xmath0 with a degree @xmath6 map to @xmath19 and full monodromy ( in the @xmath18 limit ) is @xmath74 where @xmath75 is the number of partitions of @xmath6 into @xmath76 parts such that the parts take exactly @xmath35 values and @xmath77 is the number of partitions of @xmath6 into @xmath76 parts .
it seems within reach to prove this conjecture for @xmath67 by proving function field analogs of @xcite . for @xmath78
the expected fiber size is ( conjecturally ) @xmath79 and for @xmath80 the expected fiber size is ( conjecturally ) @xmath81 in particular , we now show that conjecture [ c ] follows from the function field analog of ( * ? ? ?
* conjecture 5.1 ) .
let @xmath35 be the function field of @xmath19 and @xmath38 place of @xmath19 corresponding to a @xmath39-rational point .
then ( * ? ? ?
* conjecture 5.1).gives conjectures for local densties for @xmath39 algebras among @xmath82 while @xmath37 ranges over degree @xmath6 @xmath7-extensions of @xmath35 .
in particular , ( * ? ? ? * conjecture 5.1 ) predicts , as we range over degree @xmath6 tale @xmath39 algebras @xmath44 , that @xmath44 appears with relative density @xmath83 ( recall @xmath84 is the absolute norm of the relative discriminant of @xmath85 ) .
recall that isomorphism classes of degree @xmath6 tale @xmath39 algebras exactly correspond to continuous homomorphisms @xmath86 , and we can rephrase the above to say that @xmath44 appears with relative density @xmath87 continuous homomorphisms @xmath86 correspond to choices @xmath88 such that @xmath89 , and @xmath90 .
the number of @xmath0 rational points above @xmath38 in the curve corresponding to the extension @xmath37 of @xmath35 is given by the number of @xmath91-orbits on @xmath92 that are also @xmath93-orbits .
let @xmath94 have @xmath95 orbits of size @xmath96 , for @xmath97 .
then the number of @xmath98 such that @xmath89 is @xmath99 .
if we consider a single cycle @xmath100 of @xmath42 of length @xmath96 , then the number of @xmath98 such that @xmath100 remains a @xmath91-orbit is @xmath101 .
so given the choice of @xmath42 , the expected contribution of @xmath100 to rational points is @xmath102 .
thus , given a choice of @xmath42 , the expected number of rational points is @xmath35 , the number of distinct cycle lengths of @xmath42 .
note that @xmath99 is the size of the centralizer of @xmath42 , so choosing a permutation @xmath94 with relative probability @xmath103 is the same as choosing it with relative probability @xmath104 or @xmath105 .
this is equivalent to choosing a conjugacy class in @xmath7 , i.e. a partition of @xmath6 , with relative probability @xmath106 .
since a partition with @xmath35 distinct part sizes , gives an expected @xmath35 parts , conjecture [ c ] follows from the function field analog of ( * ? ? ?
* conjecture 5.1 ) .
a. bucur , c. david , b. feigon and m. laln , statistics for traces of cyclic trigonal curves over finite fields , _ int .
res . not .
advance access published on october 27 , 2009 , doi:10.1093/imrn / rnp162 .
a. bucur , c. david , b. feigon and m. laln , biased statistics for traces of cyclic p - fold covers over finite fields , to appear in proceedings of 2008 banff workshop " women in numbers ( in the fields institute communications series ) .
n. m. katz and p. sarnak , _ random matrices , frobenius eigenvalues , and monodromy_. american mathematical society colloquium publications , 45
. american mathematical society , providence , ri , 1999 .
xii+419 pp . | we give a short determination of the distribution of the number of @xmath0-rational points on a random trigonal curve over @xmath0 , in the limit as the genus of the curve goes to infinity .
in particular , the expected number of points is @xmath1 , contrasting with recent analogous results for cyclic @xmath2-fold covers of @xmath3 and plane curves which have an expected number of points of @xmath4 ( by work of kurlberg , rudnick , bucur , david , feigon and laln ) and curves which are complete intersections which have an expected number of points @xmath5 ( by work of bucur and kedlaya ) .
we also give a conjecture for the expected number of points on a random @xmath6-gonal curve with full @xmath7 monodromy based on function field analogs of bhargava s heuristics for counting number fields . |
the _ ab initio _ theoretical description of light nuclei is based on direct solution of the nuclear many - body problem given realistic nucleon - nucleon interactions . in no - core
configuration interaction ( ncci ) calculations @xcite , the nuclear many - body problem is formulated as a matrix eigenproblem .
the hamiltonian is represented in terms of basis states which are antisymmetrized products of single - particle states for the full @xmath2-body system of nucleons , _
i.e. _ , with no assumption of an inert core .
in practice , the nuclear many - body calculation must be carried out in a truncated space .
the dimension of the problem grows combinatorially with the size of the included single - particle space and with the number of nucleons in the system .
computational restrictions therefore limit the extent to which converged results can be obtained , for energies or for other properties of the wave functions . except for the very lightest systems ( @xmath3 )
, convergence is generally beyond reach .
instead , we seek to approach convergence as closely as possible . based on the still - unconverged calculations which are computationally feasible
, we would then ideally be able to obtain a reliable estimate of the true values of observables which would be obtained in the full , untruncated space .
therefore , progress may be pursued both by seeking accelerated convergence , _
e.g. _ , through the choice of basis , as considered here , and by developing means by which robust extrapolations can be made @xcite .
ncci calculations have so far been based almost exclusively upon bases constructed from harmonic oscillator single - particle wave functions .
the harmonic oscillator radial functions have the significant limitation that their asymptotic behavior is gaussian , _
i.e. _ , falling as @xmath4 for large @xmath5 .
the actual asymptotics for nucleons bound by a finite - range force are instead expected to be exponential , _
i.e. _ , falling as @xmath6 .
is shown on a logarithmic scale , so that exponential asymptotics would appear as a straight line .
the woods - saxon and oscillator potentials are shown in the inset .
( solutions are for the woods - saxon @xmath7 function , with potential parameters appropriate to neutrons in @xmath8{o}$ ] @xcite , with maximal basis radial quantum numbers @xmath9 as indicated . ) , scaledwidth=67.0% ] the problem encountered in using an oscillator basis to describe a system with exponential asymptotics may be illustrated through the simple one - dimensional example of the schrdinger equation with a woods - saxon potential . in fig .
[ fig - ws - soln ] , we see the results of solving for a particular eigenfunction in terms of successively larger bases of oscillator radial functions . in the classically forbidden region , where the potential is nearly flat , the tail of the wave function should be exponential .
it should thus appear as a straight line on the logarithmic scale in fig .
[ fig - ws - soln ] .
inclusion of each additional basis function yields a small extension to the region in which the expected straight - line behavior is reproduced , but , for any finite number of oscillator functions , there is a radius beyond which the calculated tail is seen to sharply fall below the true asymptotics .
observables which are sensitive to the large - radius asymptotic portions of the nuclear wave function therefore present a special challenge to convergence in ncci calculations with a conventional oscillator basis .
such `` long - range '' observables include the rms radius and @xmath10 moments and transitions , since the @xmath11 dependence of the relevant operators in both cases preferentially weight the larger-@xmath5 portions of the wave - function .
the results for these observables in ncci calculations are in general highly basis - dependent @xcite .
furthermore , a prominent feature in light nuclei is the emergence of halo structure @xcite , in which one or more loosely - bound nucleons surround a compact core , spending much of their time in the classically - forbidden region .
a realistic treatment of the long - range properties of the wave function is essential for an accurate reproduction of the halo structure @xcite .
we are therefore motivated to consider alternative bases which might be better suited for expanding the nuclear wave function in its asymptotic region .
the framework for carrying out ncci calculations with a general radial basis is developed in ref .
we explore the use of the coulomb - sturmian functions @xcite , which form a complete set of square - integrable functions and have exponential asymptotics . in the present work ,
we apply the colomb - sturmian basis to ncci calculations for the neutron halo nuclei @xmath12{he}$ ] as well as to the baseline case @xmath13{he}$ ] , for which converged results can be obtained .
we examine the possibility of extracting rms radii for the proton and neutron distributions based on a relatively straightforward estimate , the `` crossover point '' @xcite , pending further development of more sophisticated extrapolation schemes @xcite .
motivated by the disparity between proton and neutron radial distributions in the neutron - rich halo nuclei , we also explore the use of proton - neutron asymmetric bases , with different length scales for the proton and neutron radial basis functions .
the basis and methods are first reviewed ( sec .
[ sec - methods ] ) , after which the results for @xmath14{he}$ ] are discussed ( sec . [ sec - results ] ) .
the harmonic oscillator basis functions , as used in conventional ncci calculations , constitute a complete , discrete , orthogonal set of square - integrable functions and are given by @xmath15 , with radial wave functions @xmath16 e^{-\tfrac12(r / b)^2},\ ] ] where the @xmath17 are generalized laguerre polynomials , the @xmath18 are spherical harmonics , @xmath9 is the radial quantum number , @xmath19 and @xmath20 are the orbital angular momentum and its @xmath21-projection , and @xmath22 is the oscillator length .
the coulomb - sturmian functions likewise constitute a complete , discrete , orthogonal set of square - integrable functions , while also possessing exponential asymptotics more appropriate to the nuclear problem .
they are given by @xmath23 , with radial wave functions @xmath24 where @xmath22 again represents a length scale .
further details may be found in ref .
@xcite . both sets of radial functions
are shown in fig .
[ fig - scheme - radial ] , for comparison . for either basis , the single - particle basis states @xmath25 are then defined by coupling of the orbital angular momentum with the spin , to give total angular momentum @xmath26 , and the many - body basis is defined by taking antisymmetrized products of these single - particle states .
thus , the structure of the many - body calculation is independent of the details of the radial basis .
the choice of radial basis only enters the calculation through the values of the hamiltonian two - body matrix elements ( or higher - body matrix elements , if present ) , which we must first generate as the input to the many - body calculation .
the nuclear hamiltonian for ncci calculations has the form @xmath27 , where @xmath28 is the galilean - invariant , two - body relative kinetic energy operator , and @xmath29 is the nucleon - nucleon interaction . of center - of - mass oscillator quanta can also be included , to shift center - of - mass excitations out of the low - lying spectrum , but it is not essential for the ground - state properties considered here .
the implications of center - of - mass dynamics for general bases are addressed in ref .
the relative kinetic energy decomposes into one - body and two - body terms as @xmath30.\ ] ] since the two - body term is separable , matrix elements of @xmath28 may be calculated in a straightforward fashion for any radial basis , in terms of one - dimensional radial integrals of the operators @xmath31 and @xmath32 @xcite .
calculation of the interaction two - body matrix elements becomes more involved if one moves to a general radial basis .
the nucleon - nucleon interaction is defined in relative coordinates .
the oscillator basis is special , in that matrix elements in a relative oscillator basis , consisting of functions @xmath33 , can readily be transformed to the two - body oscillator basis , consisting of functions @xmath34 , by the moshinsky transformation .
we therefore still begin by carrying out the transformation to two - body matrix elements @xmath35 with respect to the oscillator basis , and only then carry out a change of basis to the coulomb - sturmian basis in the two - body space , as @xcite @xmath36 where we label single - particle orbitals for the oscillator basis by unbarred symbols @xmath37 and those for the coulomb - sturmian basis by barred symbols @xmath38 .
the coefficients @xmath39 required for the transformation are obtained from straightforward one - dimensional overlaps of the harmonic oscillator and coulomb - sturmian radial functions , @xmath40 .
the oscillator length @xmath41 with respect to which the interaction two - body matrix elements are defined and the length scale @xmath22 of the final coulomb - sturmian basis functions may in general be different .
the change - of - basis transformation in ( [ eqn - tbme - xform ] ) is , in practice , limited to a finite sum , _
e.g. _ , with a shell cutoff @xmath42 .
the cutoff @xmath43 must be chosen high enough to insure that the results of the subsequent many - body calculation are cutoff - independent , which may in general depend upon the oscillator and coulomb - sturmian length parameters , interaction , nucleus , and observable at hand .
of the harmonic oscillator basis and ( b ) @xmath44 of the coulomb - sturmian basis , with @xmath45 given by the node - matching prescription ( see text ) .
these functions are shown arranged according to the harmonic oscillator principal quantum number @xmath46 , and are labeled by @xmath19 .
the dotted curves show the same functions dilated outward by a factor of @xmath47 . ]
any single particle basis , including ( [ eqn - ho - r ] ) or ( [ eqn - cs - s ] ) , has a free length scale @xmath22 . for the oscillator basis
, this is traditionally quoted as the oscillator energy @xmath48 , where @xmath49^{1/2}}.\ ] ] in deference to the convention of presenting ncci results as a function of the basis `` @xmath48 '' , we nominally carry over this relation to define an @xmath50 parameter for general radial bases , although @xmath48 no longer has any direct physical meaning as an energy scale .
regardless , the inverse square - root dependence remains , so that a factor of two change in @xmath48 describes a factor of @xmath51 change in radial scale , as illustrated for both harmonic oscillator and coulomb - sturmian bases by the dotted curves in fig .
[ fig - scheme - radial ] .
furthermore , there is much additional freedom in the basis , since the many - body basis states ( antisymmetrized product states ) constructed from a single - particle basis are orthonormal so long as the single - particle states are orthonormal .
orthogonality for single - particle states of different @xmath19 or @xmath26 follows entirely from the angular and spin parts of the wave function
. only orthogonality _ within _ the space of a given @xmath19 and @xmath26 follows from the radial functions , _
e.g. _ , for the coulomb - sturmian functions , @xmath52\,\delta_{l'l}\delta_{j'j}$ ] .
we are therefore free to choose @xmath22 independently , firstly , for each @xmath19 space ( or @xmath26 space ) , as @xmath45 ( or @xmath53 ) , and , secondly , for protons and neutrons , as @xmath54 and @xmath55 .
the first observation raises the possibility , still to be explored , of obtaining significant improvements in the efficacy of the basis by optimizing the @xmath19-dependence of the length parameter . in ref .
@xcite , the radial scale of the coulomb - sturmian functions , for each @xmath19 , was fixed by matching the first node of the @xmath56 coulomb - sturmian function to the first node of the @xmath56 oscillator function , at that @xmath19 , yielding the prescription @xmath57^{1/2}b({{\hbar\omega}})$ ] @xcite .
the second observation raises the possibility of proton - neutron asymmetric length scales , which might be advantageous for nuclei with significant disparities between the proton and neutron distributions , in particular , halo nuclei .
therefore , in the present work , we adopt @xmath58 where @xmath59 sets an overall relative scale @xmath60 . for example , if the solid and dotted curves in fig . [ fig - scheme - radial](b ) are taken to represent the proton and neutron radial functions , respectively , then the figure illustrates the case in which @xmath61 .
{ he}$ ] ground state energy ( top ) and rms point - proton radius @xmath63 ( bottom ) , using the conventional oscillator ( left ) and coulomb - sturmian ( right ) bases .
these are shown as functions of the basis @xmath48 parameter , for @xmath64 to @xmath65 ( as labeled ) , and for transformation cutoffs @xmath66 , @xmath67 , and @xmath68 ( coulomb - sturmian basis only , indicated by dashing , curves nearly indistinguishable ) .
the converged energy is indicated by the horizontal line ( at top ) , the crossover radii by dashed horizontal lines ( at bottom ) , and the spread in radius values by vertical bars ( again at bottom ) .
, scaledwidth=90.0% ] { he}$ ] ground state rms point - proton radius @xmath63 , as estimated from the crossover point ( see text ) , calculated for the harmonic oscillator and coulomb - sturmian bases .
the experimental value is from ref .
, scaledwidth=70.0% ] we carry out calculations for the isotopes @xmath14{he}$ ] using both the harmonic oscillator and coulomb - sturmian bases .
these calculations are based on the jisp16 nucleon - nucleon interaction @xcite , plus coulomb interaction .
the bare interaction is used , _
i.e. _ , without renormalization .
the proton - neutron @xmath69-scheme code mfdn @xcite is employed for the many - body calculations .
results are calculated with basis truncations up to @xmath70 for @xmath13{he}$ ] , @xmath71 for @xmath72{he}$ ] , and @xmath73 for @xmath74{he}$ ] .
scheme , based on the total number of oscillator quanta .
that is , the many - body basis states are characterized by a total number of oscillator quanta @xmath75 , where @xmath76 . if @xmath77 is written as @xmath78 , where @xmath79 in the lowest pauli - allowed number of quanta , then the basis is subject to the restriction @xmath80 .
we formally carry this truncation over to the coulomb - sturmian basis , although @xmath46 no longer has significance as an oscillator principal quantum number . ]
the last neutrons in @xmath72{he}$ ] and @xmath74{he}$ ] are only weakly bound , with two - neutron separation energies of @xmath81 and @xmath82 , respectively .
these isotopes are interpreted as consisting of neutron halos surrounding an @xmath83 core @xcite .
the basic observables indicating halo properties are the rms radii of the proton and neutron distributions , @xmath63 and @xmath84 , respectively . and @xmath84 are the rms radii of the point - proton and point - neutron distributions , measured relative to the center of mass .
see ref .
@xcite for definitions , and ref .
@xcite for evaluation of the two - body relative rms radius observable with a general radial basis . from the analysis of experimental charge and matter radii in ref .
@xcite , @xmath13{he}$ ] has @xmath85 , @xmath72{he}$ ] has @xmath86 and @xmath87 , and @xmath74{he}$ ] has @xmath88 and @xmath89 . ]
moving from @xmath13{he}$ ] to @xmath72{he}$ ] , @xmath63 increases by @xmath90 .
this may be understood as resulting from the recoil of the @xmath83 core against the halo neutrons , and potentially core polarization , as well . in turn
, @xmath84 is larger than @xmath63 by @xmath91 , reflecting the extended halo neutron distribution .
the radii for @xmath74{he}$ ] are comparable to those for @xmath72{he}$ ] .
we first consider calculations for @xmath13{he}$ ] as a baseline .
results are shown over two doublings in @xmath50 , _
i.e. _ , representing a doubling in basis length scale , in fig .
[ fig-4he - scan ] .
energy convergence is reached for the harmonic oscillator basis , as evidenced by approximate @xmath92 and @xmath50 independence of the higher @xmath92 results over a range of @xmath50 values , in fig .
[ fig-4he - scan](a , b ) .
convergence is obtained at the @xmath93 level by @xmath70 .
the binding energies for @xmath13{he}$ ] computed with the coulomb - sturmian basis lag significantly behind those obtained with the oscillator basis , by about two steps in @xmath92 .
this should perhaps not be surprising , given that @xmath13{he}$ ] is tightly bound , and the structure can thus be expected to be driven by short - range correlations rather than asymptotic properties .
incidentally , it may be seen from fig .
[ fig-4he - scan](b , d ) that stability with respect to the cutoff in the change - of - basis transformation ( [ eqn - tbme - xform ] ) has been obtained calculations with @xmath66 , @xmath67 , and @xmath68 are virtually indistinguishable ( the transformation has been carried out from oscillator basis interaction matrix elements at @xmath94 ) .
convergence of the computed rms radii , for both the oscillator and coulomb - sturmian bases , is again indicated by approximate @xmath92 and @xmath50 independence over a range of @xmath50 values , which appears as a shoulder in the curves of fig .
[ fig-4he - scan](c , d ) .
the @xmath50 dependence for the coulomb - sturmian calculations appears to be moderately shallower , over the range ( two doublings ) of @xmath50 shown , than for the harmonic oscillator calculations [ see vertical bars in fig .
[ fig-4he - scan](c , d ) ] .
it was proposed in refs .
@xcite that the radius can be estimated even before convergence is well - developed by the crossover point between the curves obtained for successive @xmath92 values .
this is an admittedly _ ad hoc _ prescription , rather than a theoretically motivated extrapolation .
however , we can test it for both oscillator and coulomb - sturmian bases in this case of @xmath13{he}$ ] , where the final converged value is known .
the crossover radii are shown as a function of @xmath92 , for both bases , in fig .
[ fig-4he - nmax ] .
the curves used in deducing these crossovers are computed by cubic interpolation of the calculated data points at different @xmath50 .
the crossovers already serve to estimate the final converged value to within @xmath95 at @xmath96 .
it may be noted , from fig .
[ fig-4he - nmax ] , that the converged radius obtained with the jisp16 interaction agrees with experiment to within @xmath97 .
let us now consider the calculations for the halo nuclei @xmath12{he}$ ] .
the computed ground state energies , proton radii , and neutron radii are shown in figs .
[ fig-6he - scan ] and [ fig-8he - scan ] .
results are included ( at right in each figure ) for a coulomb - sturmian basis with proton - neutron asymmetric length scales in the ratio @xmath98 , which is comparable to the ratio @xmath99 of neutron and proton distribution radii for these nuclei .
energy convergence in the coulomb - sturmian basis lags that of the harmonic oscillator basis , but less dramatically than seen above for @xmath13{he}$ ] . a basic three - point exponential extrapolation of the energy with respect to @xmath92 , at each @xmath50 value ,
is indicated by the dashed curves in figs .
[ fig-6he - scan ] and [ fig-8he - scan ] .
the extrapolated energy is remarkably @xmath50-independent in the @xmath98 calculations , still with some @xmath92 dependence .
it appears to be approximately consistent with the harmonic oscillator extrapolations as well .
however , such extrapolations must be viewed with caution , as both theoretical arguments and empirical studies suggest that other functional forms may be more appropriate , over at least portions of the @xmath50 range @xcite . comparing the results for radii obtained with the various bases , for @xmath12{he}$ ]
, we see that the coulomb - sturmian results ( for either @xmath100 or @xmath98 ) again have a moderately shallower @xmath50 dependence than obtained with the harmonic oscillator basis . well - defined and stable crossover points are visible in figs .
[ fig-6he - scan ] and [ fig-8he - scan ] , especially for the @xmath98 calculations ( at right ) . the extracted crossover radii are shown , as functions of @xmath92 , in figs .
[ fig-6he - nmax ] and [ fig-8he - nmax ] .
the radii obtained for the coulomb - sturmian calculations with different ratios of neutron and proton length scales ( @xmath100 , @xmath101 , and @xmath102 ) track each other closely from @xmath103 onward , agreeing with each other to within @xmath104 . for @xmath63 ,
the values are stable with respect to @xmath92 and agree with the values obtained from the harmonic oscillator basis crossover as well . for @xmath84
, it appears that the values might be drifting systematically with @xmath92 , although they do remain within an @xmath105 range from @xmath96 onward .
( the crossover radii obtained from the harmonic oscillator calculations are fluctuating over a wider range . )
therefore , it is not possible to give a definitive value , but an estimate of @xmath106 can reasonably be made , for both @xmath12{he}$ ] .
thus , _ ab initio _ ncci calculations for @xmath12{he}$ ] with the jisp16 interaction , using both conventional and coulomb - sturmian bases , yield consistent estimates of the rms point - proton and point - neutron radii , when these are extracted by the crossover prescription .
the results qualitatively reproduce the trend in proton and neutron radii across the @xmath62 isotopes , while quantitatively suggesting that the jisp16 interaction may yield radii which are smaller than experimentally observed , by as much as @xmath107 for the @xmath12{he}$ ] neutron radii .
{ he}$ ] ground state energy ( top ) and rms point - proton radius @xmath63 and point - neutron radius @xmath84 ( bottom ) , using the conventional oscillator basis ( left ) , coulomb - sturmian basis ( center ) , and coulomb - sturmian basis with @xmath98 ( right ) .
exponentially extrapolated energies are indicated by dashed curves ( at top ) , and crossover radii by dashed horizontal lines ( at bottom ) .
, scaledwidth=100.0% ] { he}$ ] ground state rms point - proton radius @xmath63 ( lower curves ) and point - neutron radius @xmath84 ( upper curves ) , as estimated from the crossover point ( see text ) , calculated for the harmonic oscillator basis and for coulomb - sturmian bases with @xmath100 , @xmath101 , and @xmath102 ( as indicated ) .
experimental values are from ref .
, scaledwidth=70.0% ] { he}$ ] ground state energy ( top ) and rms point - proton radius @xmath63 and point - neutron radius @xmath84 ( bottom ) , using the conventional oscillator basis ( left ) , coulomb - sturmian basis ( center ) , and coulomb - sturmian basis with @xmath98 ( right ) .
exponentially extrapolated energies are indicated by dashed curves ( at top ) , and crossover radii by dashed horizontal lines ( at bottom ) .
, scaledwidth=100.0% ] { he}$ ] ground state rms point - proton radius @xmath63 ( lower curves ) and point - neutron radius @xmath84 ( upper curves ) , as estimated from the crossover point ( see text ) , calculated for the harmonic oscillator basis and for coulomb - sturmian bases with @xmath100 , @xmath101 , and @xmath102 ( as indicated ) .
experimental values are from ref .
, scaledwidth=70.0% ]
we thank s. quaglioni , s. bacca , and m. brodeur for valuable discussions and a. e. mccoy for comments on the manuscript .
this work was supported by the research corporation for science advancement through the cottrell scholar program , by the us department of energy under grants no .
de - fg02 - 95er-40934 , desc0008485 ( scidac / nuclei ) , and de - fg02 - 87er40371 , and by the us national science foundation under grant no .
computational resources were provided by the national energy research supercomputer center ( nersc ) , which is supported by the office of science of the u.s .
department of energy under contract no .
de - ac02 - 05ch11231 . | the rapid falloff of the oscillator functions at large radius ( gaussian asymptotics ) makes them poorly suited for the description of the asymptotic properties of the nuclear wave function , a problem which becomes particularly acute for halo nuclei .
we consider an alternative basis for _ ab initio _ no - core configuration interaction ( ncci ) calculations , built from coulomb - sturmian radial functions , allowing for realistic ( exponential ) radial falloff .
ncci calculations are carried out for the neutron - rich isotopes , and estimates are made for the rms radii of the proton and neutron distributions .
+ * keywords : * _ no - core configuration interaction ; coulomb - sturmian basis ; neutron halo ; nuclear radii _ * * halo nuclei with the coulomb - sturmian basis + @xmath0__department of physics , university of notre dame , notre dame , indiana 46556 , usa _ _
+ @xmath1__department of physics and astronomy , iowa state university , ames , iowa 50011 , usa _ _ |
in the recent years , the model of light vector particles with kinetic mixing to the standard model photon has received tremendous attention , theoretically as well as experimentally . whereas @xmath3 is mainly being probed in medium - to - high energy collider experiments , masses in the sub - mev regime are subject to severe astrophysical and cosmological constraints .
below @xmath4 ev , those limits are complemented by direct laboratory searches for dark photons in non - accelerator type experiments . among the most prominent
are the `` light - shining - through - wall '' experiments ( lsw ) @xcite and the conversion experiments from the solar dark photon flux , `` helioscopes '' @xcite ; a collection of low - energy constraints on dark photons can _ e.g. _ be found in the recent review @xcite .
helioscopes derive their sensitivity from the fact that such light vectors are easily produced in astrophysical environments , such as in the solar interior , covering a wide range of masses up to @xmath5 few kev . in general , stellar astrophysics provides stringent constraints on any type of light , weakly - interacting particles once the state becomes kinematically accessible @xcite . only in a handful of examples
does the sensitivity of terrestrial experiments match the stellar energy loss constraints . here
we review our works @xcite in which we have identified a new stellar energy loss mechanism originating from the resonant production of longitudinally polarized dark photons and derived ensuing constraints from underground rare event searches .
limits on dark photons were improved to the extent that previously derived constraints from all lsw and helioscope experiments are now superseded by the revised astrophysical and new experimental limits .
and @xmath6 .
the solid / dotted line shows the longitudinal(l)/transverse(t ) contribution .
_ right : _
constraints on @xmath7 as a function of @xmath1 .
the black solid / dashed / dotted curves show the total / longitudinal / transverse energy loss limit of the sun by requiring that the dark photon luminosity does not exceed 10% of the standard solar luminosity @xcite .
the red line shows the constraint derived from the xenon10 data .
previous and future ( = proj .
) experimental bounds / sensitivities are shown by the shaded regions . from light to dark shading these are from the cast experiment @xcite considering the contributions from only the transverse modes @xcite , from the alps collaboration @xcite , and from tests of the inverse square law of the coulomb interaction @xcite.,title="fig:",scaledwidth=50.0% ] and @xmath6 .
the solid / dotted line shows the longitudinal(l)/transverse(t ) contribution .
_ right : _
constraints on @xmath7 as a function of @xmath1 .
the black solid / dashed / dotted curves show the total / longitudinal / transverse energy loss limit of the sun by requiring that the dark photon luminosity does not exceed 10% of the standard solar luminosity @xcite .
the red line shows the constraint derived from the xenon10 data .
previous and future ( = proj .
) experimental bounds / sensitivities are shown by the shaded regions . from light to dark shading these are from the cast experiment @xcite considering the contributions from only the transverse modes @xcite , from the alps collaboration @xcite , and from tests of the inverse square law of the coulomb interaction @xcite.,title="fig:",scaledwidth=50.0% ] the minimal extension of the sm gauge group by an additional @xmath8 gauge factor yields the following effective lagrangian well below the electroweak scale , @xmath9 where @xmath0 is the vector field associated with the abelian factor @xmath8 .
the field strengths of the photon @xmath10 and of the dark photon @xmath11 are connected via the kinetic mixing parameter @xmath7 where a dependence on the weak mixing angle was absorbed ; @xmath12 is the usual electromagnetic current with electric charge @xmath13 .
because of the u(1 ) nature of ( [ eq : l ] ) , we must distinguish two cases for the origin of @xmath1 : the stueckelberg case ( sc ) with non - dynamical mass , and the higgs case ( hc ) , where @xmath1 originates through the spontaneous breaking of @xmath14 by a new higgs field @xmath15 .
the crucial difference between the two cases comes in the small @xmath1 limit : while all processes of production or absorption of @xmath16 in sc are suppressed , @xmath17 , in hc there is no decoupling , and @xmath18 .
indeed , in the limit @xmath19 the interaction resembles one of a mini - charged scalar with the effective em charge of @xmath20 @xcite . in the following
we discuss the sc and refer the reader to our work @xcite as well as to @xcite and references therein for hc .
[ [ sec : flux ] ] solar flux + + + + + + + + + + the solar flux of dark photons in the sc is thoroughly calculated in ref .
@xcite ; for further discussion see also @xcite . in the small mass region , @xmath21 where @xmath22 is the plasma frequency ,
the emission of longitudinal modes of @xmath16 dominates the total flux , and the emission power of dark photons per volume can be approximated as @xmath23 this formula is most readily obtained by noting that a resonant conversion of longitudinal plasmons into dark photons is possible whenever @xmath24 .
the energy - differential flux of dark photons at the location of the earth is shown in the left panel of fig .
[ fig : dp ] .
resonant emission stops for @xmath25 since @xmath22 is limited by the temperature in the sun s core .
[ [ sec : dds ] ] absorption of dark photons + + + + + + + + + + + + + + + + + + + + + + + + + + in the sc , the ionization of an atom @xmath26 in the detector can then be schematically described as @xmath27 .
the total dark photon absorption rate is given by , @xmath28 @xmath29 are the effective mixings for the transverse ( t ) and longitudinal ( l ) modes respectively .
the polarization functions @xmath30 are found from the in - medium polarization tensor @xmath31 , @xmath32 where @xmath33 is the dark photon four momentum and @xmath34 are the polarization vectors for the transverse and longitudinal modes of the dark photon , @xmath35 . the first relation ( [ eq : gamma ] )
is a manifestation of the optical theorem .
the polarization functions @xmath30 are related to the complex index of refraction , @xmath36 or , equivalently , to the permittivity of the medium @xmath37 . for an isotropic , non - magnetic medium @xmath38 and @xmath39 , so that for an incoming on - shell dark photon with @xmath40 , @xmath41 indeed holds .
we obtain @xmath42 from its relation to the forward scattering amplitude @xmath43 where the atomic scattering factors @xmath44 are _
e.g. _ tabulated in @xcite .
close to the ionization threshold we make use of the kramers - kronig dispersion relations to relate @xmath45 and @xmath46 for estimating @xmath36 .
alternatively , one can solve an integral equation relating @xmath47 and @xmath48 in a self - consistent manner , an approach taken in @xcite .
[ [ sec : dds ] ] limits from direct detection + + + + + + + + + + + + + + + + + + + + + + + + + + + + with flux @xmath49 and absorption rate @xmath50 at hand , the expected number of signal events in a given experiment reads @xmath51 where @xmath16 and @xmath52 are the fiducial volume and live time of the experiment , respectively , and @xmath53 is the branching ratio of photoionization rate to total absorption rate . given the significant infrared enhancement of the solar dark photon spectrum , left panel of fig .
[ fig : dp ] , the low - energy ionization signals measured in the xenon10 @xcite dark matter experiment have the best sensitivity to constrain a dark photon flux that is also supported by the sun . with @xmath5412
ev ionization energy in xenon , the absorption of a dark photon with 300 ev energy can produce about 25 electrons . from @xcite
we estimate a 90% c.l upper limit on the detecting rate to be @xmath55 events kg@xmath56day@xmath56 ( similar to limits deduced in ref .
@xcite ) . in the region
@xmath57 ev the ionization process dominates the absorption , and therefore @xmath53 in this region can be set to unity .
the 90% c.l .
upper limit on @xmath7 as a function of @xmath1 is shown by the thick red curve in fig .
[ fig : dp ] . as can be seen it surpasses other current experimental limits as well as the solar energy loss bound in a mass interval from @xmath58 .
given the enormous amount of experimental progress in the field of direct dark matter detection , one can be optimistic that future sensitivity to dark photons , and other light particles is bound to be further improved .
the speaker would like to thank the conference organizers for financial support . | `` dark photons '' , light new vector particles @xmath0 kinetically mixed with the photon , are a frequently considered extension of the standard model . for masses below 10 kev
they are emitted from the solar interior . in the limit of small mass @xmath1
the dark photon flux is strongly peaked at low energies and we demonstrate that the constraint on the atomic ionization rate imposed by the results of the xenon10 dark matter experiment sets the to - date most stringent limit on the kinetic mixing parameter of this model : @xmath2 .
the result significantly improves previous experimental bounds and surpasses even the most stringent astrophysical and cosmological limits in a seven - decade - wide interval of @xmath1 . |
maximum entropy models are an important class of statistical models for biology .
for instance , they have been found to be a good model for protein folding @xcite , antibody diversity @xcite , neural population activity @xcite , and flock behavior @xcite . in this paper
we develop a general framework for studying maximum entropy distributions on weighted graphs , extending recent work of chatterjee , diaconis , and sly @xcite .
the development of this theory is partly motivated by the problem of sensory coding in neuroscience . in the brain , information is represented by discrete electrical pulses , called _ action potentials _ or _ spikes _ @xcite .
this includes neural representations of sensory stimuli which can take on a continuum of values .
for instance , large photoreceptor arrays in the retina respond to a range of light intensities in a visual environment , but the brain does not receive information from these photoreceptors directly . instead , retinal ganglion cells must convey this detailed input to the visual cortex using only a series of binary electrical signals .
continuous stimuli are therefore converted by networks of neurons to sequences of spike times .
an unresolved controversy in neuroscience is whether information is contained in the precise timings of these spikes or only in their rates " ( i.e. , counts of spikes in a window of time ) .
early theoretical studies @xcite suggest that information capacities of timing - based codes are superior to those that are rate - based ( also see @xcite for an implementation in a simple model ) .
moreover , a number of scientific articles have appeared suggesting that precise spike timing @xcite and synchrony @xcite are important for various computations in the brain . here
, we briefly explain a possible scheme for encoding continuous vectors with spiking neurons that takes advantage of precise spike timing and the mathematics of maximum entropy distributions .
consider a network of @xmath0 neurons in one region of the brain which transmits a continuous vector @xmath1 using sequences of spikes to a second receiver region .
we assume that this second region contains a number of coincidence detectors that measure the absolute difference in spike times between pairs of neurons projecting from the first region .
we imagine three scenarios for how information can be obtained by these detectors . in the first
, the detector is only measuring for synchrony between spikes ; that is , either the detector assigns a 0 to a nonzero timing difference or a 1 to a coincidence of spikes . in another scenario
, timing differences between projecting neurons can assume an infinite but countable number of possible values . finally ,
in the third scenario , we allow these differences to take on any nonnegative real values .
we further assume that neuronal output and thus spike times are stochastic variables .
a basic question now arises : how can the first region encode @xmath2 so that it can be recovered robustly by the second ?
we answer this question by first asking the one symmetric to this : how can the second region recover a real vector transmitted by an unknown sender region from spike timing measurements ?
we propose the following possible solution to this problem .
fix one of the detector mechanics as described above , and set @xmath3 to be the measurement of the absolute timing difference between spikes from projecting neurons @xmath4 and @xmath5 .
we assume that the receiver population can compute the ( local ) sums @xmath6 efficiently .
the values @xmath7 represent a weighted graph @xmath8 on @xmath0 vertices , and we assume that @xmath3 is randomly drawn from a distribution on timing measurements @xmath9 . making no further assumptions , a principle of jaynes @xcite suggests that the second region propose that the timing differences are drawn from the ( unique ) distribution over weighted graphs with the highest entropy @xcite having the vector @xmath10 for the expectations of the degree sums @xmath11 . depending on which of the three scenarios described above
is true for the coincidence detector , this prescription produces one of three different maximum entropy distributions .
consider the third scenario above ( the other cases are also subsumed by our results ) . as we shall see in section [ sec : cont ] , the distribution determined in this case is parameterized by a real vector @xmath12 , and finding the maximum likelihood estimator ( mle ) for these parameters using @xmath13 as sufficient statistics boils down to solving the following set of @xmath0 algebraic equations in the @xmath0 unknowns @xmath14 : @xmath15 given our motivation , we call the system of equations the _ retina equations _ for theoretical neuroscience , and note that they have been studied in a more general context by sanyal , sturmfels , and vinzant @xcite using matroid theory and algebraic geometry . remarkably , a solution @xmath16 to has the property that with high probability , it is arbitrarily close to the original parameters @xmath2 for sufficiently large network sizes @xmath0 ( in the scenario of binary measurements , this is a result of @xcite ) . in particular , it is possible for the receiver region to recover reliably a continuous vector @xmath2 from a _ single _ cycle of neuronal firing emanating from the sender region .
we now know how to answer our first question : _ the sender region should arrange spike timing differences to come from a maximum entropy distribution_. we remark that this conclusion is consistent with modern paradigms in theoretical neuroscience and artificial intelligence , such as the concept of the boltzmann machine @xcite , a stochastic version of its ( zero - temperature ) deterministic limit , the little - hopfield network @xcite . [ [ organization . ] ] organization .
+ + + + + + + + + + + + + the organization of this paper is as follows . in section [ sec : general ] , we lay out the general theory of maximum entropy distributions on weighted graphs . in section
[ sec : specific ] , we specialize the general theory to three classes of weighted graphs . for each class , we provide an explicit characterization of the maximum entropy distributions and prove a generalization of the erds - gallai criterion for weighted graphical sequences .
furthermore , we also present the consistency property of the mle of the vertex parameters from one graph sample .
section [ sec : proofs ] provides the proofs of the main technical results presented in section [ sec : specific ] .
finally , in section [ sec : discussion ] we discuss the results in this paper and some future research directions .
[ [ notation . ] ] notation .
+ + + + + + + + + in this paper we use the following notation .
let @xmath17 , @xmath18 , @xmath19 , and @xmath20 .
we write @xmath21 and @xmath22 for the summation and product , respectively , over all @xmath23 pairs @xmath24 with @xmath25 . for a subset @xmath26 , @xmath27 and
@xmath28 denote the interior and closure of @xmath29 in @xmath30 , respectively . for a vector @xmath31 , @xmath32 and
@xmath33 denote the @xmath34 and @xmath35 norms of @xmath36 . for an @xmath37 matrix @xmath38
, @xmath39 denotes the matrix norm induced by the @xmath40-norm on vectors in @xmath30 , that is , @xmath41
in this section we develop the general machinery of maximum entropy distributions on graphs via the theory of exponential family distributions @xcite , and in subsequent sections we specialize our analysis to some particular cases of weighted graphs .
consider an undirected graph @xmath8 on @xmath42 vertices with edge @xmath43 having weight @xmath44 , where @xmath45 is the set of possible weight values .
we will later consider the following specific cases : 1 .
_ finite discrete weighted graphs , _ with edge weights in @xmath46 , @xmath47 .
2 . _ infinite discrete weighted graphs , _ with edge weights in @xmath48 .
_ continuous weighted graphs , _ with edge weights in @xmath49 . a graph @xmath8 is fully specified by its _ adjacency matrix _ @xmath50 , which is an @xmath37 symmetric matrix with zeros along its diagonal . for fixed @xmath0 , a probability distribution over graphs @xmath8 corresponds to a distribution over adjacency matrices @xmath51 . given a graph with adjacency matrix @xmath7 ,
let @xmath52 be the degree of vertex @xmath4 , and let @xmath53 be the degree sequence of @xmath54 .
let @xmath55 be a @xmath56-algebra over the set of weight values @xmath57 , and assume there is a canonical @xmath56-finite probability measure @xmath58 on @xmath59 .
let @xmath60 be the product measure on @xmath61 , and let @xmath62 be the set of all probability distributions on @xmath61 that are absolutely continuous with respect to @xmath60 .
since @xmath60 is @xmath56-finite , these probability distributions can be characterized by their density functions , i.e. the radon - nikodym derivatives with respect to @xmath60 . given a sequence @xmath63 , let @xmath64 be the set of distributions in @xmath62 whose expected degree sequence is equal to @xmath13 , @xmath65 = \mathbf{d}\},\ ] ] where in the definition above , the random variable @xmath66 is drawn from the distribution @xmath67 .
then the distribution @xmath68 in @xmath64 with maximum entropy is precisely the exponential family distribution with the degree sequence as sufficient statistics ( * ? ? ?
* chapter 3 ) . specifically , the density of @xmath68 at @xmath51 is given by in the parameterization , instead of the canonical parameterization @xmath69 , because it simplifies the notation in our later presentation . ]
@xmath70 where @xmath71 is the _ log - partition function _ , @xmath72 and @xmath73 is a parameter that belongs to the _ natural parameter space _
@xmath74 we will also write @xmath75 if we need to emphasize the dependence of @xmath68 on the parameter @xmath2 . using the definition @xmath52
, we can write @xmath76 hence , we can express the log - partition function as @xmath77 in which @xmath78 is the marginal log - partition function @xmath79 consequently , the density in can be written as @xmath80 this means the edge weights @xmath81 are independent random variables , with @xmath82 having distribution @xmath83 with density @xmath84 in particular , the edge weights @xmath81 belong to the same exponential family distribution but with different parameters that depend on @xmath85 and @xmath86 ( or rather , on their sum @xmath87 ) .
the parameters @xmath88 can be interpreted as the potential at each vertex that determines how strongly the vertices are connected to each other .
furthermore , we can write the natural parameter space @xmath89 as @xmath90 using the characterization of @xmath68 as the maximum entropy distribution in @xmath64 , the condition @xmath91 means we need to choose the parameter @xmath2 for @xmath75 such that @xmath92 = \mathbf{d}$ ] . in place of @xmath93 to emphasize the dependence of the expectation on the parameter @xmath2 . ]
this is an instance of the moment - matching equation , which , in the case of exponential family distributions , is well - known to be equivalent to finding the maximum likelihood estimator ( mle ) of @xmath2 given an empirical degree sequence @xmath94 .
specifically , suppose we draw graph samples @xmath95 i.i.d . from the distribution @xmath68 with parameter @xmath96 , and
we want to find the mle @xmath97 of @xmath96 based on the observations @xmath95 . using the parametric form of the density ,
this is equivalent to solving the maximization problem @xmath98 where @xmath13 is the average of the degree sequences of @xmath95 . setting the gradient of @xmath99 to zero
reveals that the mle @xmath97 satisfies @xmath100 recall that the gradient of the log - partition function in an exponential family distribution is equal to the expected sufficient statistics . in our case , we have @xmath101 $ ] , so the mle equation recovers the moment - matching equation @xmath102 = \mathbf{d}$ ] .
in section [ sec : specific ] we study the properties of the mle of @xmath2 from a _ single _ sample @xmath103 . in the remainder of this section
, we address the question of the existence and uniqueness of the mle with a given empirical degree sequence @xmath13 .
define the _
mean parameter space _
@xmath104 to be the set of expected degree sequences from all distributions on @xmath61 that are absolutely continuous with respect to @xmath60 , @xmath105 \colon { \mathbb{p}}\in \mathfrak{p } \}.\ ] ] the set @xmath104 is necessarily convex , since a convex combination of probability distributions in @xmath62 is also a probability distribution in @xmath62 .
recall that an exponential family distribution is _ minimal _ if there is no linear combination of the sufficient statistics that is constant almost surely with respect to the base distribution .
this minimality property clearly holds for @xmath68 , for which the sufficient statistics are the degree sequence .
we say that @xmath68 is _ regular _ if the natural parameter space @xmath89 is open . by the general theory of exponential family distributions (
* theorem 3.3 ) , in a regular and minimal exponential family distribution , the gradient of the log - partition function maps the natural parameter space @xmath89 to the interior of the mean parameter space @xmath104 , and this mapping , instead of @xmath106 , because of our choice of the parameterization in using @xmath107 . ]
@xmath108 is bijective .
we summarize the preceding discussion in the following result .
[ prop : regularminimal ] assume @xmath89 is open .
then there exists a solution @xmath109 to the mle equation @xmath110 = \mathbf{d}$ ] if and only if @xmath111 , and if such a solution exists then it is unique .
we now characterize the mean parameter space @xmath104 .
we say that a sequence @xmath112 is _ graphic _ ( or a _ graphical sequence _ ) if @xmath13 is the degree sequence of a graph @xmath8 with edge weights in @xmath57 , and in this case we say that @xmath8 _ realizes _ @xmath13 .
it is important to note that whether a sequence @xmath13 is graphic depends on the weight set @xmath57 , which we consider fixed for now .
[ prop : mconvw ] let @xmath113 be the set of all graphical sequences , and let @xmath114 be the convex hull of @xmath113 . then @xmath115
furthermore , if @xmath62 contains the dirac delta measures , then @xmath116 .
the inclusion @xmath115 is clear , since any element of @xmath104 is of the form @xmath117 $ ] for some distribution @xmath67 and @xmath118 for every realization of the random variable @xmath119 .
now suppose @xmath62 contains the dirac delta measures @xmath120 for each @xmath121 .
given @xmath122 , let @xmath123 be the adjacency matrix of the graph that realizes @xmath13 .
then @xmath124 \in { \mathcal{m}}$ ] , which means @xmath125 , and hence @xmath126 since @xmath104 is convex .
as we shall see in section [ sec : specific ] , the result above allows us to conclude that @xmath116 for the case of discrete weighted graphs . on the other hand , for the case of continuous weighted graphs we need to prove @xmath116 directly since @xmath62 in this case does not contain the dirac measures .
we emphasize the distinction between a _ valid _
solution @xmath109 and a _ general
_ solution @xmath1 to the mle equation @xmath92 = \mathbf{d}$ ] . as we saw from proposition [ prop : regularminimal ] , we have a precise characterization of the existence and uniqueness of the valid solution @xmath109 , but in general , there are multiple solutions @xmath2 to the mle equation . in this paper
we shall be concerned only with the valid solution ; sanyal , sturmfels , and vinzant study some algebraic properties of general solutions @xcite .
we close this section by discussing the symmetry of the valid solution to the mle equation .
recall the decomposition of the log - partition function @xmath71 into the marginal log - partition functions @xmath127 .
let @xmath128 , and let @xmath129 denote the ( marginal ) _ mean function _
@xmath130 observing that we can write @xmath131 = \int_s a \
: \exp \big ( -(\theta_i + \theta_j ) a - z_1(\theta_i + \theta_j ) \big ) \ : \nu(da ) = \mu(\theta_i+\theta_j),\ ] ] the mle equation @xmath92 = \mathbf{d}$ ] then becomes @xmath132 in the statement below , @xmath133 denotes the sign function : @xmath134 if @xmath135 , and @xmath136 .
[ prop : symsoln ] let @xmath111 , and let @xmath109 be the unique solution to the system of equations . if @xmath137 is strictly increasing , then @xmath138 and similarly , if @xmath137 is strictly decreasing , then @xmath139 given @xmath25 , @xmath140 if @xmath137 is strictly increasing , then @xmath141 has the same sign as @xmath142 for each @xmath143 , and thus @xmath144 also has the same sign as @xmath142 .
similarly , if @xmath137 is strictly decreasing , then @xmath141 has the opposite sign of @xmath142 , and thus @xmath144 also has the opposite sign of @xmath142 .
in this section we analyze the maximum entropy random graph distributions for several specific choices of the weight set @xmath57 . for each case , we specify the distribution of the edge weights @xmath81 , the mean function @xmath137 , the natural parameter space @xmath89 , and characterize the mean parameter space @xmath104 .
we also study the problem of finding the mle @xmath97 of @xmath2 from one graph sample @xmath103 and prove the existence , uniqueness , and consistency of the mle . along the way
, we derive analogues of the erds - gallai criterion of graphical sequences for weighted graphs .
we defer the proofs of the results presented here to section [ sec : proofs ] .
we first study weighted graphs with edge weights in the finite discrete set @xmath46 , where @xmath47 .
the case @xmath145 corresponds to unweighted graphs , and our analysis in this section recovers the results of chatterjee , diaconis , and sly @xcite .
the proofs of the results in this section are provided in section [ sec : prooffinitedisc ] .
we take @xmath58 to be the counting measure on @xmath57 .
following the development in section [ sec : general ] , the edge weights @xmath82 are independent random variables with density @xmath146 where the marginal log - partition function @xmath147 is given by @xmath148 since @xmath149 for all @xmath150 , the natural parameter space @xmath151 is given by @xmath152 .
the mean function is given by @xmath153 at @xmath154 the mean function takes the value @xmath155 while for @xmath135 , the mean function simplifies to @xmath156 figure [ fig : finitediscmean ] shows the behavior of the mean function @xmath157 and its derivative @xmath158 as @xmath159 varies . [ cols="^,^ " , ] figure [ fig : finitedisc ] demonstrates the performance of the algorithm presented above .
we set @xmath160 and sample @xmath161^n$ ] uniformly at random .
then for each @xmath162 , we sample a graph from the distribution @xmath75 , compute the empirical degree sequence @xmath13 , and run the iterative algorithm starting with @xmath163 until convergence .
the left panel ( figure [ fig : finitedisc](a ) ) shows the rate of convergence ( on a logarithmic scale ) of the algorithm for various values of @xmath159 .
we observe that the iterates @xmath164 indeed converge geometrically fast to the mle @xmath97 , but the rate of convergence decreases as @xmath159 increases . by examining the proof of theorem [ thm : mlealgfinitedisc ] in section [ sec :
prooffinitedisc ] , we see that the term @xmath165 has the expression @xmath166 where @xmath167 .
this shows that @xmath165 is an increasing function of @xmath159 , which explains the empirical decrease in the rate of convergence as @xmath159 increases .
figures [ fig : finitedisc](b ) and ( c ) show the plots of the estimate @xmath97 versus the true @xmath2 .
notice that the points lie close to the diagonal line , which suggests that the mle @xmath97 is very close to the true parameter @xmath2 .
indeed , the following result shows that @xmath97 is a consistent estimator of @xmath2 .
recall that @xmath97 is _ consistent _ if @xmath97 converges in probability to @xmath2 as @xmath168 .
[ thm : consistencyfinitedisc ] let @xmath169 and @xmath170 be fixed
. given @xmath1 with @xmath171 , consider the problem of finding the mle @xmath97 of @xmath2 based on one graph sample @xmath103 .
then for sufficiently large @xmath0 , with probability at least @xmath172 the mle @xmath97 exists and satisfies @xmath173 where @xmath29 is a constant that only depends on @xmath174 .
in this section we study weighted graphs with edge weights in @xmath175 .
the proofs of the results presented here are provided in section [ sec : proofcont ] .
we take @xmath58 to be the lebesgue measure on @xmath175 .
the marginal log - partition function is @xmath176 thus @xmath177 , and the natural parameter space is @xmath178 for @xmath109 , the edge weights @xmath81 are independent exponential random variables with density @xmath179 and mean parameter @xmath180 = 1/(\theta_i+\theta_j)$ ] .
the corresponding mean function is given by @xmath181 we now consider the problem of finding the mle of @xmath2 from one graph sample @xmath103 . as we saw previously ,
the mle @xmath182 satisfies the moment - matching equation @xmath102 = \mathbf{d}$ ] , where @xmath13 is the degree sequence of the sample graph @xmath8 .
equivalently , @xmath182 is a solution to the system of equations @xmath183 the system is a special case of a general class that sanyal , sturmfels , and vinzant @xcite study using algebraic geometry and matroid theory ( extending the work of proudfoot and speyer @xcite ) .
define @xmath184 in which @xmath185 is the stirling number of the second kind and @xmath186 is a generalized falling factorial .
then , there is a polynomial @xmath187 in the @xmath188 such that for @xmath189 with @xmath190 , the number of solutions @xmath191 to is @xmath192 .
moreover , the polynomial @xmath187 has degree @xmath193 and characterizes those @xmath13 for which the equations above have multiple roots .
we refer to @xcite for more details . since the natural parameter space @xmath89 is open , proposition [ prop : regularminimal ] tells us that the mle @xmath97 exists and is unique if and only if the empirical degree sequence @xmath13 belongs to the interior @xmath194 of the mean parameter space @xmath104 .
we characterize the set of graphical sequences @xmath113 and determine its relation to the mean parameter space @xmath104 .
we say @xmath112 is a _ ( continuous ) graphical sequence _ if there is a graph @xmath8 with edge weights in @xmath175 that realizes @xmath13 .
the finite discrete graphical sequences from section [ sec : finitediscrete ] have combinatorial constraints because there are only finitely many possible edge weights between any pair of vertices , and these constraints translate into a set of inequalities in the generalized erds - gallai criterion in theorem [ thm : graphicalfinitedisc ] . in the case of continuous weighted graphs , however , we do not have these constraints because every edge can have as much weight as possible . therefore , the criterion for a continuous graphical sequence should be simpler than in theorem [ thm : graphicalfinitedisc ] , as the following result shows .
[ thm : graphicalcont ] a sequence @xmath195 is graphic if and only if @xmath196 we note that condition is implied by the case @xmath197 in the conditions .
this is to be expected , since any finite discrete weighted graph is also a continuous weighted graph , so finite discrete graphical sequences are also continuous graphical sequences .
given the criterion in theorem [ thm : graphicalcont ] , we can write the set @xmath113 of graphical sequences as @xmath198 moreover , we can also show that the set of graphical sequences coincide with the mean parameter space .
[ lem : w - convex ] the set @xmath113 is convex , and @xmath199 .
the result above , together with the result of proposition [ prop : regularminimal ] , implies that the mle @xmath97 exists and is unique if and only if the empirical degree sequence @xmath13 belongs to the interior of the mean parameter space , which can be written explicitly as @xmath200 let @xmath201 and @xmath202 with @xmath203 .
it is easy to see that the system of equations gives us @xmath204 from which we obtain a unique solution @xmath205 .
recall that @xmath206 means @xmath207 , @xmath208 , and @xmath209 , so the equations above tell us that @xmath206 if and only if @xmath210 .
in particular , this also implies @xmath211 , so @xmath212 .
hence , there is a unique solution @xmath206 to the system of equations if and only if @xmath111 , as claimed above . finally , we prove that the mle @xmath97 is a consistent estimator of @xmath2 .
[ thm : consistencycont ] let @xmath213 and @xmath214 be fixed . given @xmath109 with @xmath215 , @xmath25 ,
consider the problem of finding the mle @xmath182 of @xmath2 from one graph sample @xmath103 . then for sufficiently large @xmath0 , with probability at least @xmath172 the mle @xmath182 exists and satisfies @xmath216 where @xmath217 is a universal constant .
we now turn our focus to weighted graphs with edge weights in @xmath218 .
the proofs of the results presented here can be found in section [ sec : proofinfinitedisc ] .
we take @xmath58 to be the counting measure on @xmath218 . in this case
the marginal log - partition function is given by @xmath219 thus , the domain of @xmath147 is @xmath220 , and the natural parameter space is @xmath221 which is the same natural parameter space as in the case of continuous weighted graphs in the preceding section . given @xmath109 , the edge weights @xmath81 are independent geometric random variables with probability mass function @xmath222 the mean parameters are @xmath223 = \frac{\exp(-\theta_i-\theta_j)}{1-\exp(-\theta_i-\theta_j ) } = \frac{1}{\exp(\theta_i+\theta_j)-1},\ ] ] induced by the mean function @xmath224 consider the problem of finding the mle of @xmath2 from one graph sample @xmath103 .
let @xmath13 denote the degree sequence of @xmath8 .
then the mle @xmath182 , which satisfies the moment - matching equation @xmath102 = \mathbf{d}$ ] , is a solution to the system of equations @xmath225 we note that the natural parameter space @xmath89 is open , so by proposition [ prop : regularminimal ] , the mle @xmath97 exists and is unique if and only if @xmath111 , where @xmath104 is the mean parameter space .
since @xmath60 is the counting measure on @xmath226 , the set @xmath62 contains all the dirac measures , so we know @xmath116 from proposition [ prop : mconvw ] . here
@xmath113 is the set of all _ ( infinite discrete ) graphical sequences _
, namely , the set of degree sequences of weighted graphs with edge weights in @xmath218 .
the following result provides a precise criterion for such graphical sequences .
note that condition below is implied by the limit @xmath227 in theorem [ thm : graphicalfinitedisc ] .
[ thm : graphicalinfinitedisc ] a sequence @xmath228 is graphic if and only if @xmath229 is even and @xmath230 the criterion in theorem [ thm : graphicalinfinitedisc ] allows us to write an explicit form for the set of graphical sequences @xmath113 , @xmath231 now we need to characterize @xmath114 .
let @xmath232 denote the set of all continuous graphical sequences from theorem [ thm : graphicalcont ] , when the edge weights are in @xmath175 , @xmath233 it turns out that when we take the convex hull of @xmath113 , we essentially recover @xmath232 .
[ lem : convwinfinitedisc ] @xmath234 . recalling that a convex set and its closure
have the same interior points , the result above gives us @xmath235 let @xmath201 and @xmath202 with @xmath203 .
it can be easily verified that the system of equations gives us @xmath236 from which we can obtain a unique solution @xmath205 .
recall that @xmath206 means @xmath237 , @xmath238 , and @xmath239 , so the equations above tell us that @xmath240 if and only if @xmath241 , or equivalently , @xmath210
. this also implies @xmath211 , so @xmath212 .
thus , the system of equations has a unique solution @xmath206 if and only if @xmath111 , as claimed above .
finally , we prove that with high probability the mle @xmath97 exists and converges to @xmath2 .
[ thm : consistencyinfinitedisc ] let @xmath213 and @xmath214 be fixed . given @xmath109 with @xmath215 , @xmath25 ,
consider the problem of finding the mle @xmath182 of @xmath2 from one graph sample @xmath103 .
then for sufficiently large @xmath0 , with probability at least @xmath242 the mle @xmath182 exists and satisfies @xmath243 where @xmath217 is a universal constant .
in this section we provide proofs for the technical results presented in section [ sec : specific ] .
the proofs of the characterization of weighted graphical sequences ( theorems [ thm : graphicalfinitedisc ] , [ thm : graphicalcont ] , and [ thm : graphicalinfinitedisc ] ) are inspired by the constructive proof of the classical erds - gallai criterion by choudum @xcite .
we begin by presenting several results that we will use in this section .
we use the definition of sub - exponential random variables and the concentration inequality presented in @xcite .
we say that a real - valued random variable @xmath244 is _ sub - exponential _ with parameter @xmath245 if @xmath246^{1/p }
\leq \kappa p \quad \text { for all } p \geq 1.\ ] ] note that if @xmath244 is a @xmath247-sub - exponential random variable with finite first moment , then the centered random variable @xmath248 $ ] is also sub - exponential with parameter @xmath249 .
this follows from the triangle inequality applied to the @xmath250-norm , followed by jensen s inequality for @xmath251 : @xmath252\big|^p\big]^{1/p } & \leq { \mathbb{e}}[|x|^p]^{1/p } + \big|{\mathbb{e}}[x]\big| \leq 2{\mathbb{e}}[|x|^p]^{1/p}. \end{split}\ ] ] sub - exponential random variables satisfy the following concentration inequality .
[ thm : concineqsubexp ] let @xmath253 be independent centered random variables , and suppose each @xmath254 is sub - exponential with parameter @xmath255 .
let @xmath256 .
then for every @xmath257 , @xmath258,\ ] ] where @xmath217 is an absolute constant .
we will apply the concentration inequality above to exponential and geometric random variables , which are the distributions of the edge weights of continuous weighted graphs ( from section [ sec : cont ] ) and infinite discrete weighted graphs ( from section [ sec : infinitedisc ] ) .
[ lem : subexp - exp ] let @xmath244 be an exponential random variable with @xmath259 = 1/\lambda$ ] . then @xmath244 is sub - exponential with parameter @xmath260 , and the centered random variable @xmath261 is sub - exponential with parameter @xmath262 . for any @xmath251 , we can evaluate the moment of @xmath244 directly : @xmath246 = \int_0^\infty x^p \cdot \lambda \ , \exp(-\lambda x ) \ : dx = \frac{1}{\lambda^p } \ : \int_0^\infty y^p \ :
\exp(-y ) \ : dy = \frac{\gamma(p+1)}{\lambda^p},\ ] ] where @xmath263 is the gamma function , and in the computation above we have used the substitution @xmath264 .
it can be easily verified that @xmath265 for @xmath251 , so @xmath246^{1/p } = \frac{\big(\gamma(p+1)\big)^{1/p}}{\lambda } \leq \frac{p}{\lambda}.\ ] ] this shows that @xmath244 is sub - exponential with parameter @xmath260 .
[ lem : subexp - geo ] let @xmath244 be a geometric random variable with parameter @xmath266 , so @xmath267 then @xmath244 is sub - exponential with parameter @xmath268 , and the centered random variable @xmath269 is sub - exponential with parameter @xmath270 .
fix @xmath251 , and consider the function @xmath271 , @xmath272 .
one can easily verify that @xmath273 is increasing for @xmath274 and decreasing on @xmath275 , where @xmath276 .
in particular , for all @xmath277 we have @xmath278 , and @xmath279 now note that for @xmath280 we have @xmath281 , and for @xmath282 we have @xmath283 .
thus , we can bound @xmath284 using the substitution @xmath285 , we can evaluate the integral to be @xmath286 where in the last step we have again used the relation @xmath265 .
we use the result above , along with the expression of @xmath287 , to bound the moment of @xmath244 : @xmath288 & = \sum_{a=0}^\infty a^p \cdot ( 1-q)^a \ : q = q \ : \sum_{a=0}^\infty f(a ) \\ & \leq q \ :
\int_0^\infty f(x ) \ : dx + 2q \ : f(\lambda ) \\ & \leq \left(\frac{q^{1/p } \ : p}{(-\log(1-q))^{1 + 1/p } } \right)^p + \left ( \frac{(2q)^{1/p } \ : p}{-e \cdot \log ( 1-q ) } \right)^p \\ & \leq \left ( \frac{q^{1/p } \ : p}{(-\log(1-q))^{1 + 1/p } } + \frac{(2q)^{1/p } \ : p}{-e \cdot \log ( 1-q ) } \right)^p , \end{split}\ ] ] where in the last step we have used the fact that @xmath289 for @xmath290 and @xmath251 .
this gives us @xmath291^{1/p } & \leq \frac{q^{1/p}}{(-\log(1-q))^{1 + 1/p } } + \frac{2^{1/p } \ : q^{1/p}}{-e \cdot \log ( 1-q ) } \\ & = \frac{1}{-\log(1-q ) } \ : \left(\left(\frac{q}{-\log(1-q)}\right)^{1/p } + \frac{2^{1/p } \ : q^{1/p}}{e}\right ) .
\end{split}\ ] ] now note that @xmath292 for @xmath293 , so @xmath294 .
moreover , @xmath295 .
therefore , for any @xmath251 , we have @xmath291^{1/p } & \leq \frac{1}{-\log(1-q ) } \ : \left(1 + \frac{2}{e}\right ) < \frac{2}{-\log(1-q)}. \end{split}\ ] ] thus , we conclude that @xmath244 is sub - exponential with parameter @xmath268 .
an @xmath296 real matrix @xmath297 is _ diagonally dominant _ if @xmath298 we say that @xmath297 is _ diagonally balanced _ if @xmath299 for @xmath300 .
we have the following bound from @xcite on the inverses of diagonally dominant matrices .
this bound is independent of @xmath301 , so it is also applicable to diagonally balanced matrices .
we will use this result in the proofs of theorems [ thm : consistencycont ] and [ thm : consistencyinfinitedisc ] .
[ thm : main ] let @xmath42 .
for any symmetric diagonally dominant matrix @xmath297 with @xmath302 , we have @xmath303 in this section we present the proofs of the results presented in section [ sec : finitediscrete ] .
we first prove the necessity of .
suppose @xmath304 is the degree sequence of a graph @xmath8 with edge weights @xmath44 .
then @xmath305 is even .
moreover , for each @xmath306 , @xmath307 counts the total edge weights coming out from the vertices @xmath308 .
the total edge weights from these @xmath309 vertices to themselves is at most @xmath310 , and for each vertex @xmath311 , the total edge weights from these @xmath309 vertices to vertex @xmath5 is at most @xmath312 , so by summing over @xmath311 we get . to prove the sufficiency of we use induction on @xmath313 .
the base case @xmath314 is trivial .
assume the statement holds for @xmath315 , and suppose we have a sequence @xmath13 with @xmath316 satisfying with @xmath317 . without loss of generality we may assume @xmath318 , for
otherwise we can proceed with only the nonzero elements of @xmath13 .
let @xmath319 be the smallest index such that @xmath320 , with @xmath321 if @xmath322 .
define @xmath323 , so we have @xmath324 and @xmath325 .
we will show that @xmath326 satisfies . by the inductive hypothesis , this means
@xmath327 is the degree sequence of a graph @xmath328 with edge weights @xmath329 .
we now attempt to modify @xmath328 to obtain a graph @xmath8 whose degree sequence is equal to @xmath13 .
if the weight @xmath330 of the edge @xmath331 is less than @xmath332 , then we can obtain @xmath8 by increasing @xmath330 by @xmath333 , since the degree of vertex @xmath334 is now @xmath335 , and the degree of vertex @xmath0 is now @xmath336
. otherwise , suppose @xmath337 .
since @xmath338 , there exists a vertex @xmath339 such that @xmath340 . since @xmath341
, there exists another vertex @xmath342 such that @xmath343 .
then we can obtain the graph @xmath8 by increasing @xmath344 and @xmath345 by @xmath333 and reducing @xmath346 by @xmath333 , so that now the degrees of vertices @xmath334 and @xmath0 are each increased by @xmath333 , and the degrees of vertices @xmath347 and @xmath342 are preserved .
it now remains to show that @xmath326 satisfies .
we divide the proof into several cases for different values of @xmath309 .
we will repeatedly use the fact that @xmath13 satisfies , as well as the inequality @xmath348 . 1 . for @xmath349
: @xmath350 2 . for @xmath351 : @xmath352 3 . for @xmath353
: first suppose @xmath354 .
then for all @xmath5 we have @xmath355 so @xmath356 4 . for @xmath353
: suppose @xmath357 , and @xmath358 .
we claim that @xmath13 satisfies at @xmath309 with a strict inequality .
if this claim is true , then , since @xmath359 and @xmath360 , @xmath361 now to prove the claim , suppose the contrary that @xmath13 satisfies at @xmath309 with equality . let @xmath362 be the smallest integer such that @xmath363 .
then , from our assumption , @xmath364 therefore , since @xmath365 , @xmath366 which contradicts the fact that @xmath13 satisfies at @xmath367 .
thus , we have proved that @xmath13 satisfies at @xmath309 with a strict inequality .
5 . for @xmath353
: suppose @xmath368 . in particular , we have @xmath369 and @xmath370 for all @xmath5 .
first , if we have @xmath371 then we are done , since @xmath372 condition is obvious if @xmath373 or @xmath374 ( since there are @xmath375 terms in the summation and each term is at least @xmath333 ) .
otherwise , assume @xmath376 and @xmath377 , so in particular , we have @xmath378 ( since @xmath379 ) , @xmath321 , and @xmath380 .
note that we can not have @xmath381 , for then @xmath382 would be odd , so we must have @xmath383 .
similarly , @xmath0 must be even , for otherwise @xmath384 would be odd .
thus , since @xmath385 we must have @xmath386 .
therefore , @xmath387 this shows that @xmath326 satisfies and finishes the proof of theorem [ thm : graphicalfinitedisc ] .
we follow the outline of the proof of ( * ? ? ?
* theorem 1.5 ) .
we first present the following properties of the mean function @xmath157 and the jacobian matrix of the function @xmath388 .
we then combine these results at the end of this section into a proof of theorem [ thm : mlealgfinitedisc ] .
[ lem : propertymeanfunction ] the mean function @xmath157 is positive and strictly decreasing , with @xmath389 for all @xmath150 , and @xmath390 as @xmath391 .
its derivative @xmath158 is increasing for @xmath392 , with the properties that @xmath393 , @xmath394 for all @xmath150 , and @xmath395 .
it is clear from that @xmath157 is positive . from the alternative representation it is easy to see that @xmath389 , and @xmath390 as @xmath391 .
differentiating expression yields the formula @xmath396 and substituting @xmath154 gives us @xmath395 .
the cauchy - schwarz inequality applied to the expression above tells us that @xmath393 , where the inequality is strict because the vectors @xmath397 and @xmath398 are not linearly dependent .
thus , @xmath157 is strictly decreasing for all @xmath150 .
the relation @xmath389 gives us @xmath399 .
furthermore , by differentiating the expression twice , one can verify that @xmath400 for @xmath392 , which means @xmath158 is increasing for @xmath392 .
see also figure [ fig : finitediscmean ] for the behavior of @xmath157 and @xmath158 for different values of @xmath159 .
[ lem : lowerboundratio ] for all @xmath150 , we have @xmath401 multiplying the numerator and denominator of by @xmath402 , we can write @xmath403 therefore , @xmath404 we recall the following definition and result from @xcite . given @xmath405 , let @xmath406 denote the set of @xmath37 matrices @xmath407 with @xmath408 , @xmath409 , and @xmath410 , for each @xmath411 .
[ lem : lndelta ] if @xmath412 , then @xmath413 in particular , for @xmath42 , @xmath414 given @xmath415 , let @xmath416 denote the @xmath37 matrix whose @xmath43-entry is @xmath417 [ lem : normjinfinity ] for all @xmath415 , we have @xmath418 .
the partial derivatives of @xmath388 are @xmath419 and for @xmath25 , @xmath420 where the last inequality follows from @xmath421 .
using the result of lemma [ lem : lowerboundratio ] and the fact that @xmath137 is positive , we also see that @xmath422 setting @xmath423 and integrating over @xmath424 , we also get that @xmath425 for @xmath25 , and @xmath426 .
this implies @xmath418 , as desired .
[ lem : jinlndelta ] let @xmath415 with @xmath427 and @xmath428 for some @xmath429 .
then @xmath430 , where @xmath431 from lemma [ lem : normjinfinity ] we already know that @xmath432 satisfies @xmath433 , so to show that @xmath434 it remains to show that @xmath435 and @xmath436 for @xmath25 . in particular , it suffices to show that for each @xmath424 we have @xmath437 and @xmath438 , where @xmath439 .
fix @xmath424 .
since @xmath427 and @xmath428 , we also know that @xmath440 , so @xmath441 for all @xmath442 . using the properties of @xmath137 and @xmath443 from lemma [ lem : propertymeanfunction ] , we have @xmath444 and @xmath445 then from and using the definition of @xmath446 , @xmath447 furthermore , by lemma [ lem : lowerboundratio ] we have @xmath448 so from , we also get @xmath449 as required .
we are now ready to prove theorem [ thm : mlealgfinitedisc ] .
theorem [ thm : mlealgfinitedisc ] by the mean - value theorem for vector - valued functions @xcite , for any @xmath415 we can write @xmath450 where @xmath416 is the jacobian matrix defined in . since @xmath418 ( lemma [ lem : normjinfinity ] ) , this gives us @xmath451 first suppose there is a solution @xmath97 to the system of equations , so @xmath97 is a fixed point of @xmath388 .
then by setting @xmath452 and @xmath453 to the inequality above , we obtain @xmath454 in particular , this shows that @xmath455 for all @xmath456 , where @xmath457 . by lemma [ lem : jinlndelta ] , this implies @xmath458 for all @xmath456 , where @xmath446 is given by . another application of the mean - value theorem gives us @xmath459 so by lemma [ lem : lndelta ] , @xmath460 unrolling the recursive bound above and using gives us @xmath461 which proves with @xmath462 .
now suppose the system of equations does not have a solution , and suppose the contrary that the sequence @xmath463 does not have a divergent subsequence .
this means @xmath463 is a bounded sequence , so there exists @xmath429 such that @xmath455 for all @xmath456 .
then by lemma [ lem : jinlndelta ] , @xmath464 for all @xmath456 , where @xmath446 is given by .
in particular , by the mean value theorem and lemma [ lem : normjinfinity ] , we get for all @xmath456 , @xmath465 this implies @xmath466 , which means @xmath463 is a cauchy sequence .
thus , the sequence @xmath463 converges to a limit , say @xmath97 , as @xmath467 .
this limit @xmath97 is necessarily a fixed point of @xmath388 , as well as a solution to the system of equations , contradicting our assumption .
hence we conclude that @xmath463 must have a divergent subsequence .
a little computation based on the proof above gives us the following result , which will be useful in the proof of theorem [ thm : consistencyfinitedisc ] .
[ prop : finitediscboundfirstiter ] assume the same setting as in theorem [ thm : mlealgfinitedisc ] , and assume the mle equation [ eq : mleeqfinitedisc ] has a unique solution @xmath97 .
then @xmath468 where @xmath446 is given by with @xmath167 . with the same notation as in the proof of theorem [ thm : mlealgfinitedisc ] , by applying the mean - value theorem twice and using the bound in lemma [ lem : normjinfinity ] , for each @xmath469 we have @xmath465 therefore , since @xmath463 converges to @xmath97 , @xmath470 where the last inequality follows from .
our proof of theorem [ thm : consistencyfinitedisc ] follows the outline of the proof of theorem 1.3 in @xcite . recall that @xmath113 is the set of graphical sequences , and the mle equation has a unique solution @xmath471 if and only if @xmath472 .
we first present a few preliminary results .
we will also use the properties of the mean function @xmath137 as described in lemma [ lem : propertymeanfunction ] .
the following property is based on ( * ? ? ?
* lemma 4.1 ) .
[ lem : finitediscboundmle ] let @xmath473 with the properties that @xmath474 and @xmath475 where @xmath476 and @xmath477 are constants .
then the mle equation has a solution @xmath97 with the property that @xmath478 , where @xmath479 is a constant that only depends on @xmath480 .
first assume @xmath97 exists , so @xmath97 and @xmath13 satisfy @xmath481 let @xmath482 and let @xmath483 be such that @xmath484 and @xmath485 .
we begin by observing that since @xmath137 is a decreasing function and we have the assumption , @xmath486 so @xmath487 thus , if we have a lower bound on @xmath488 by a constant , then we also get a constant upper bound on @xmath489 and we are done .
we now proceed to prove the lower bound @xmath490 . if @xmath491 , then there is nothing to prove , so let us assume that @xmath492 .
we claim the following property .
[ [ claim . ] ] claim .
+ + + + + + _ if @xmath488 satisfies @xmath493 and @xmath494 , then the set @xmath495 has @xmath496 . _
+ claim let @xmath497 and @xmath498 .
note that @xmath499 since @xmath500 , so @xmath501 . then using the property that @xmath137 is a decreasing function and the assumption on @xmath502
, we obtain @xmath503 this implies @xmath504 , which means there exists @xmath505 , so @xmath506 .
let @xmath507 , and let @xmath508 . then , using the properties that @xmath137 is decreasing and bounded above by @xmath332 , and using the assumption on @xmath509 , we get @xmath510 rearranging the last inequality above gives us @xmath511 .
note that for every @xmath512 , @xmath513 , we have @xmath514 .
therefore , if @xmath515 , then we see that @xmath516 , so @xmath517 as desired .
now assume @xmath518 for otherwise we are done .
then @xmath493 and @xmath494 , so by the claim above , the size of the set @xmath495 is at least @xmath519 .
let @xmath520 and for integers @xmath521 , define the set @xmath522 since the sets @xmath523 are disjoint , by the pigeonhole principle we can find an index @xmath524 such that @xmath525 fix @xmath526 , and consider the set @xmath527 note that @xmath528 , which implies @xmath529 , so @xmath530 . for @xmath531 , define @xmath532 and observe that @xmath533 we note that for @xmath534 we have @xmath535 , so @xmath536 where in the last inequality we have used the definition @xmath537 .
now take @xmath538 , so @xmath539 .
we consider three cases : 1 . if @xmath540 , then for every @xmath541 , we have @xmath542 , so @xmath543 2 . if @xmath544 , then for every @xmath534 , we have @xmath545 , so @xmath546 3 .
if @xmath547 , then @xmath548 , and in this case @xmath549 there are at most @xmath0 such indices @xmath5 in both the first and second cases above , and there are at most @xmath550 such indices @xmath5 in the third case .
therefore , @xmath551 combining this bound with and using give us @xmath552 assumption tells us that the left hand side of the inequality above is bounded below by @xmath553 , so we obtain @xmath554 the left hand side is a decreasing function of @xmath555 , so the bound above tells us that @xmath556 for a constant @xmath557 that only depends on @xmath558 ( and @xmath159 ) , and so @xmath559 , as desired . [
[ showing - existence - of - hat - theta . ] ] showing existence of @xmath97 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + now let @xmath473 satisfy and .
let @xmath560 be a sequence of points in @xmath561 converging to @xmath13 , so by proposition [ prop : regularminimal ] , for each @xmath469 there exists a solution @xmath562 to the mle equation with @xmath563 in place of @xmath13 .
since @xmath13 satisfy , , and @xmath564 , for all sufficiently large @xmath309 , @xmath563 also satisfy and with some constants @xmath565 depending on @xmath566 .
the preceding analysis then shows that @xmath567 for all sufficiently large @xmath309 , where @xmath568 is a constant depending on @xmath566 .
this means @xmath569 is a bounded sequence , so it contains a convergent subsequence @xmath570 , say @xmath571 .
then @xmath572 , and since @xmath573 is a solution to the mle equation for @xmath574 , @xmath97 is necessarily a solution to for @xmath13 , and we are done .
we are now ready to prove theorem [ thm : consistencyfinitedisc ] .
theorem [ thm : consistencyfinitedisc ] let @xmath575 denote the expected degree sequence under @xmath576 , so @xmath577 .
since @xmath578 and @xmath137 is a decreasing function , we see that @xmath579 for @xmath580 , let @xmath581 and similarly for @xmath582 . using the notation @xmath583 as introduced in the proof of lemma [ lem : finitediscboundmle ] , we notice that for @xmath538 , @xmath584 and similarly , @xmath585 therefore , using the relation , we see that @xmath586 we now recall that the edge weights @xmath9 are independent random variables taking values in @xmath587 , with @xmath180 = \mu(\theta_i + \theta_j)$ ] . by
hoeffding s inequality @xcite , for each @xmath300 we have @xmath588 therefore , by union bound , with probability at least @xmath589 we have @xmath590 .
assume we are in this situation .
then from we see that for all @xmath300 , @xmath591 thus , for sufficiently large @xmath0 , we have @xmath592 with @xmath593 moreover , it is easy to see that for every @xmath580 we have @xmath594 .
since we already know that @xmath595 , this gives us @xmath596 thus , for @xmath597 and for sufficiently large @xmath0 , we have @xmath598 with @xmath599 .
we have shown that @xmath13 satisfies the properties and , so by lemma [ lem : finitediscboundmle ] , the mle @xmath97 exists and satisfies @xmath572 , where the constant @xmath29 only depends on @xmath174 ( and @xmath159 ) .
assume further that @xmath600 , so @xmath601 as well .
to bound the deviation of @xmath97 from @xmath2 , we use the convergence rate in the iterative algorithm to compute @xmath97 . set @xmath602 in the algorithm in theorem [ thm : mlealgfinitedisc ] ,
so by proposition [ prop : finitediscboundfirstiter ] , we have @xmath603 where @xmath446 is given by with @xmath604 .
from the definition of @xmath388 , we see that for each @xmath531 , @xmath605 noting that @xmath606 for @xmath607 , we have @xmath608 . using the bounds on @xmath609 and @xmath610 that we have developed above , we get @xmath611 plugging this bound to gives us the desired result . in this section
we present the proofs of the results presented in section [ sec : cont ] .
clearly if @xmath195 is a graphical sequence , then so is @xmath612 , for any permutation @xmath613 of @xmath614 .
thus , without loss of generality we can assume @xmath615 , and in this case condition reduces to @xmath616 first suppose @xmath195 is graphic , so it is the degree sequence of a graph with adjacency matrix @xmath7 .
then condition is satisfied since @xmath617 for the converse direction , we first note the following easy properties of weighted graphical sequences : a. [ p : cha ] the sequence @xmath618 is graphic for any @xmath619 , realized by the `` cycle graph '' with weights @xmath620 for @xmath621 , @xmath622 , and @xmath623 otherwise .
b. [ p : eq ] a sequence @xmath624 satisfying with an equality is graphic , realized by the `` star graph '' with weights @xmath625 for @xmath626 and @xmath623 otherwise .
c. [ p : ext ] if @xmath624 is graphic , then so is @xmath627 for any @xmath628 , realized by inserting @xmath629 isolated vertices to the graph that realizes @xmath13 .
d. [ p : sum ] if @xmath630 are graphic , then so is @xmath631 , realized by the graph whose edge weights are the sum of the edge weights of the graphs realizing @xmath632 and @xmath633 .
we now prove the converse direction by induction on @xmath0 . for the base case @xmath201 ,
it is easy to verify that @xmath634 with @xmath635 and @xmath636 is the degree sequence of the graph @xmath8 with edge weights @xmath637 assume that the claim holds for @xmath638 ; we will prove it also holds for @xmath0 .
so suppose we have a sequence @xmath112 with @xmath639 satisfying , and let @xmath640 if @xmath641 then is satisfied with an equality , and by property we know that @xmath13 is graphic .
now assume @xmath429 .
we consider two possibilities . 1 .
suppose @xmath642 .
then we can write @xmath643 , where @xmath644 and @xmath645 the assumption @xmath642 implies @xmath646 , so @xmath647 is a graphical sequence by induction hypothesis .
thus , @xmath632 is also graphic by property .
furthermore , @xmath633 is graphic by property , so @xmath643 is also a graphical sequence by property .
2 . suppose @xmath648 .
then write @xmath649 , where @xmath650 and @xmath651 by construction , @xmath652 satisfies @xmath653 , so @xmath652 is a graphical sequence by property . since @xmath654 is also graphic by property ,
we conclude that @xmath649 is graphic by property .
this completes the induction step and finishes the proof of theorem [ thm : graphicalcont ] .
we first prove that @xmath113 is convex .
given @xmath112 and @xmath655 in @xmath113 , and given @xmath424 , we note that @xmath656 which means @xmath657 .
next , recall that we already have @xmath658 from proposition [ prop : mconvw ] , so to conclude @xmath199 it remains to show that @xmath125 .
given @xmath122 , let @xmath8 be a graph that realizes @xmath13 and let @xmath659 be the edge weights of @xmath8 , so that @xmath660 for all @xmath300 . consider a distribution @xmath67 on @xmath661 that assigns each edge weight @xmath81 to be an independent exponential random variable with mean parameter @xmath662 , so @xmath67 has density @xmath663 then by construction , we have @xmath664 = w_{ij}$ ] and @xmath665 = \sum_{j \neq i } { \mathbb{e}}_{\mathbb{p}}[a_{ij } ] = \sum_{j \neq i } w_{ij } = d_i , \quad i = 1 , \dots , n.\ ] ] this shows that @xmath666 , as desired .
we first prove that the mle @xmath97 exists almost surely .
recall from the discussion in section [ sec : cont ] that @xmath97 exists if and only if @xmath111 .
clearly @xmath122 since @xmath13 is the degree sequence of the sampled graph @xmath8 .
since @xmath199 ( lemma [ lem : w - convex ] ) , we see that the mle @xmath97 does not exist if and only if @xmath667 , where @xmath668 in particular , note that @xmath669 has lebesgue measure @xmath670 .
since the distribution @xmath68 on the edge weights @xmath671 is continuous ( being a product of exponential distributions ) and @xmath13 is a continuous function of @xmath119 , we conclude that @xmath672 , as desired .
we now prove the consistency of @xmath97 .
recall that @xmath2 is the true parameter that we wish to estimate , and that the mle @xmath97 satisfies @xmath673 .
let @xmath674 denote the expected degree sequence of the maximum entropy distribution @xmath75 . by the mean value theorem for vector - valued functions @xcite ,
we can write @xmath675 here @xmath297 is a matrix obtained by integrating ( element - wise ) the hessian @xmath676 of the log - partition function on intermediate points between @xmath2 and @xmath97 : @xmath677 recalling that @xmath678 $ ] , at any intermediate point @xmath679 , we have @xmath680 therefore , the hessian @xmath676 is given by @xmath681 and @xmath682 since @xmath683 and we assume @xmath684 , it follows that for @xmath25 , @xmath685 therefore , the hessian @xmath676 is a _ diagonally balanced _
matrix with off - diagonal entries bounded below by @xmath686 .
in particular , @xmath297 is also a symmetric , diagonally balanced matrix with off - diagonal entries bounded below by @xmath686 , being an average of such matrices . by theorem [ thm : main ] , @xmath297 is invertible and its inverse satisfies the bound @xmath687 where the last inequality holds for @xmath688 .
inverting @xmath297 in and applying the bound on @xmath689 gives @xmath690 let @xmath671 denote the edge weights of the sampled graph @xmath103 , so @xmath691 for @xmath300 .
moreover , since @xmath692 is the expected degree sequence from the distribution @xmath75 , we also have @xmath693 . recall that @xmath81 is an exponential random variable with rate @xmath694 , so by lemma [ lem : subexp - exp ] , @xmath695 is sub - exponential with parameter @xmath696 . for each @xmath300 ,
the random variables @xmath697 are independent sub - exponential random variables , so we can apply the concentration inequality in theorem [ thm : concineqsubexp ] with @xmath698 and @xmath699 assume @xmath0 is sufficiently large such that @xmath700 .
then by theorem [ thm : concineqsubexp ] , for each @xmath300 we have @xmath701 by the union bound , @xmath702 assume for the rest of this proof that @xmath703 , which happens with probability at least @xmath589 . from and using the triangle inequality ,
we get @xmath704
what we have shown is that for sufficiently large @xmath0 , @xmath705 satisfies the inequality @xmath706 , where @xmath707 is the quadratic function @xmath708 it is easy to see that for sufficiently large @xmath0 we have @xmath709 and @xmath710 .
thus , @xmath706 means either @xmath711 or @xmath712 .
we claim that for sufficiently large @xmath0 we always have @xmath711 .
suppose the contrary that there are infinitely many @xmath0 for which @xmath712 , and consider one such @xmath0 . since @xmath182
we know that @xmath713 for each @xmath25 , so there can be at most one index @xmath4 with @xmath714 .
we consider the following two cases : 1 .
* case 1 : * suppose @xmath715 for all @xmath300 .
let @xmath716 be an index with @xmath717 . then , using the fact that @xmath97 satisfies the system of equations and @xmath718 for @xmath719 , we see that @xmath720 which can not hold for sufficiently large @xmath0 , as the last expression tends to @xmath670 as @xmath168 .
* case 2 : * suppose @xmath714 for some @xmath300 , so @xmath721 for @xmath513 since @xmath182 . without loss of generality assume @xmath722 , so @xmath723 . following the same chain of inequalities as in the previous case ( with @xmath724 ) , we obtain @xmath725 so for sufficiently large @xmath0 , @xmath726 and thus @xmath727 for each @xmath728 .
however , then @xmath729 which can not hold for sufficiently large @xmath0 , as the right hand side of the last expression tends to @xmath730 faster than the left hand side .
the analysis above shows that @xmath711 for all sufficiently large @xmath0 . plugging in this result to
, we conclude that for sufficiently large @xmath0 , with probability at least @xmath172 we have the bound @xmath731 as desired . in this section
we prove the results presented in section [ sec : infinitedisc ] . without loss of generality
we may assume @xmath316 , so condition becomes @xmath732 .
the necessary part is easy : if @xmath733 is a degree sequence of a graph @xmath8 with edge weights @xmath734 , then @xmath735 is even , and the total weight coming out of vertex @xmath333 is at most @xmath736 . for the converse direction ,
we proceed by induction on @xmath737 .
the statement is clearly true for @xmath314 and @xmath738 .
assume the statement is true for some even @xmath739 , and suppose we are given @xmath740 with @xmath741 , @xmath742 , and @xmath732 . without loss of generality we may assume @xmath318 , for
otherwise we can proceed with only the nonzero elements of @xmath13 .
let @xmath319 be the smallest index such that @xmath320 , with @xmath321 if @xmath743 , and let @xmath744 .
we will show that @xmath326 is graphic .
this will imply that @xmath13 is graphic , because if @xmath326 is realized by the graph @xmath328 with edge weights @xmath745 , then @xmath13 is realized by the graph @xmath8 with edge weights @xmath746 and @xmath747 otherwise .
now for @xmath655 given above , we have @xmath748 and @xmath749 is even .
so it suffices to show that @xmath750 , for then we can apply the induction hypothesis to conclude that @xmath326 is graphic . if @xmath751 , then @xmath752 . if @xmath753 then @xmath754 , so @xmath755 since @xmath318 . in particular , since @xmath229 is even , @xmath756 is also even , hence @xmath757 .
therefore , @xmath758 .
this finishes the proof of theorem [ thm : graphicalinfinitedisc ] .
clearly @xmath759 , so @xmath760 since @xmath232 is closed and convex , by lemma [ lem : w - convex ] .
conversely , let @xmath761 denote the set of rational numbers
. we will first show that @xmath762 and then proceed by a limit argument .
let @xmath763 , so @xmath764 with @xmath765 and @xmath766 .
choose @xmath767 large enough such that @xmath768 for all @xmath300 .
observe that @xmath769 has the property that @xmath770 is even and @xmath771 , so @xmath772 by definition .
since @xmath773 as well , all elements along the segment joining @xmath670 and @xmath774 lie in @xmath114 , so in particular , @xmath775 .
this shows that @xmath762 , and hence @xmath776 .
to finish the proof it remains to show that @xmath777 . on the one hand we have @xmath778 for the other direction , given @xmath779 , choose @xmath780 such that @xmath781 are in general position , so that the convex hull @xmath29 of @xmath782 is full dimensional
this can be done , for instance , by noting that the following @xmath783 points in @xmath232 are in general position : @xmath784 where @xmath785 are the standard basis of @xmath30 . for each @xmath786 and @xmath787 ,
choose @xmath788 on the line segment between @xmath13 and @xmath789 such that the convex hull @xmath790 of @xmath791 is full dimensional and has diameter at most @xmath792 .
since @xmath790 is full dimensional we can choose a rational point @xmath793 .
thus we have constructed a sequence of rational points @xmath794 in @xmath232 converging to @xmath13 , which shows that @xmath795 .
we first address the issue of the existence of @xmath97 .
recall from the discussion in section [ sec : infinitedisc ] that the mle @xmath182 exists if and only if @xmath111 .
clearly @xmath122 since @xmath13 is the degree sequence of the sampled graph @xmath8 , and @xmath796 from proposition [ prop : mconvw ] .
therefore , the mle @xmath97 does not exist if and only if @xmath667 , where the boundary @xmath669 is explicitly given by @xmath668 using union bound and the fact that the edge weights @xmath81 are independent geometric random variables , we have @xmath797 furthermore , again by union bound , @xmath798 note that we have @xmath799 for some @xmath4 if and only if the edge weights @xmath800 for all @xmath801 .
this occurs with probability @xmath802 therefore , @xmath803 where the last inequality holds for sufficiently large @xmath0 .
this shows that for sufficiently large @xmath0 , the mle @xmath97 exists with probability at least @xmath804 .
we now turn to proving the consistency of @xmath97 . for the rest of this proof , assume that the mle @xmath182 exists , which occurs with probability at least @xmath804 .
the proof of the consistency of @xmath97 follows the same outline as in the proof of theorem [ thm : consistencycont ] .
let @xmath674 denote the expected degree sequence of the distribution @xmath75 , and recall that the mle @xmath97 satisfies @xmath805 . by the mean value theorem @xcite ,
we can write @xmath806 where @xmath297 is the matrix obtained by integrating the hessian of @xmath807 between @xmath2 and @xmath97 , @xmath808 let @xmath424 , and note that at the point @xmath809 the gradient @xmath106 is given by @xmath810 thus , the hessian @xmath676 is @xmath811 and @xmath812 since @xmath813 and we assume @xmath684 , for @xmath25 we have @xmath814 this means @xmath297 is a symmetric , diagonally dominant matrix with off - diagonal entries bounded below by @xmath815 , being an average of such matrices . then by theorem [ thm : main ] , we have the bound @xmath816 where the second inequality holds for @xmath688 . by inverting @xmath297 in and applying the bound on @xmath817 above , we obtain @xmath818 let @xmath671 denote the edge weights of the sampled graph @xmath103 , so @xmath691 for @xmath300 . since @xmath692 is the expected degree sequence from the distribution @xmath75 , we also have @xmath819 . recall that @xmath81 is a geometric random variable with emission probability @xmath820 so by lemma [ lem : subexp - geo ] , @xmath821 is sub - exponential with parameter @xmath822 . for each @xmath300 ,
the random variables @xmath823 are independent sub - exponential random variables , so we can apply the concentration inequality in theorem [ thm : concineqsubexp ] with @xmath824 and @xmath825 assume @xmath0 is sufficiently large such that @xmath826 .
then by theorem [ thm : concineqsubexp ] , for each @xmath300 we have @xmath827 the union bound then gives us @xmath828 assume now that @xmath829 , which happens with probability at least @xmath589 . from and using the triangle inequality ,
we get @xmath830 this means @xmath705 satisfies the inequality @xmath831 , where @xmath832 is the function @xmath833 one can easily verify that @xmath834 is a convex function , so @xmath834 assumes the value @xmath670 at most twice , and moreover , @xmath835 as @xmath836 .
it is also easy to see that for all sufficiently large @xmath0 , we have @xmath837 and @xmath838 .
therefore , @xmath831 implies either @xmath711 or @xmath839 .
we claim that for sufficiently large @xmath0 we always have @xmath711 .
suppose the contrary that there are infinitely many @xmath0 for which @xmath839 , and consider one such @xmath0 .
since @xmath713 for each @xmath25 , there can be at most one index @xmath4 with @xmath714 .
we consider the following two cases : 1 .
* case 1 : * suppose @xmath715 for all @xmath300 .
let @xmath716 be an index with @xmath840 . then , since @xmath718 for @xmath719 , @xmath841 which can not hold for sufficiently large @xmath0 , as the last expression tends to @xmath670 as @xmath168 .
* case 2 : * suppose @xmath714 for some @xmath300 , so @xmath721 for @xmath513
. without loss of generality assume @xmath722 , so @xmath842 .
following the same chain of inequalities as in the previous case ( with @xmath724 ) , we obtain @xmath843 this implies @xmath844 where the last inequality assumes @xmath0 is sufficiently large . therefore , for @xmath728 , @xmath845 however , this implies @xmath846 which can not hold for sufficiently large @xmath0 , as the right hand side in the last expression grows faster than the left hand side on the first line .
the analysis above shows that we have @xmath711 for all sufficiently large @xmath0 .
plugging in this result to gives us @xmath847 finally , taking into account the issue of the existence of the mle , we conclude that for sufficiently large @xmath0 , with probability at least @xmath848 the mle @xmath182 exists and satisfies @xmath849 as desired .
this finishes the proof of theorem [ thm : consistencyinfinitedisc ] .
in this paper , we have studied the maximum entropy distribution on weighted graphs with a given expected degree sequence . in particular
, we focused our study on three classes of weighted graphs : the finite discrete weighted graphs ( with edge weights in the set @xmath587 , @xmath47 ) , the infinite discrete weighted graphs ( with edge weights in the set @xmath218 ) , and the continuous weighted graphs ( with edge weights in the set @xmath175 ) .
we have shown that the maximum entropy distributions are characterized by the edge weights being independent random variables having exponential family distributions parameterized by the vertex potentials .
we also studied the problem of finding the mle of the vertex potentials , and we proved the remarkable consistency property of the mle from only one graph sample . in the case of finite discrete weighted graphs , we also provided a fast , iterative algorithm for finding the mle with a geometric rate of convergence . finding the mle in the case of continuous or infinite discrete weighted graphs
can be performed via standard gradient - based methods , and the bounds that we proved on the inverse hessian of the log - partition function can also be used to provide a rate of convergence for these methods .
however , it would be interesting if we can develop an efficient iterative algorithm for computing the mle , similar to the case of finite discrete weighted graphs .
another interesting research direction is to explore the theory of maximum entropy distributions when we impose additional structures on the underlying graph
. we can start with an arbitrary graph @xmath850 on @xmath0 vertices , for instance a lattice graph or a sparse graph , and consider the maximum entropy distributions on the subgraphs @xmath8 of @xmath850 . by choosing different types of the underlying graphs
@xmath850 , we can incorporate additional prior information from the specific applications we are considering .
finally , given our initial motivation for this project , we would also like to apply the theory that we developed in this paper to applications in neuroscience , in particular , in modeling the early - stage computations that occur in the retina .
there are also other problem domains where our theory are potentially useful , including applications in clustering , image segmentation , and modularity analysis .
w. bialek , a. cavagna , i. giardina , t. mora , e. silvestri , m. viale , and a. walczak .
_ statistical mechanics for natural flocks of birds_. proceedings of the national academy of sciences , * 109 * ( 2012 ) , 47864791 .
g. desbordes , j. jin , c. weng , n. a. lesica , g. b. stanley , and j. m. alonso .
_ timing precision in population coding of natural scenes in the early visual system
. _ plos biology , * 6*(12 ) , december 2008 .
b. e. kilavik , s. roux , a. ponce - alvarez , j. confais , s. grn , and a. riehle .
_ long - term modifications in motor cortical dynamics induced by intensive practice_. the journal of neuroscience , * 29 * ( 2009 ) , 1265312663 . p. maldonado , c. babul , w. singer , e. rodriguez , d. berger , and s. grn . _
synchronization of neuronal responses in primary visual cortex of monkeys viewing natural images_. journal of neurophysiology , * 100 * ( 2008 ) , 15231532 . j. shlens , g. d. field , j. l. gauthier , m. i. grivich , d. petrusca , a. sher , a. m. litke , and e. j. chichilnisky . _ the structure of multi - neuron firing patterns in primate retina_. journal of neuroscience , * 26*(32):82548266 , 2006 .
j. shlens , g. d. field , j. l. gauthier , m. greschner , a. sher , a. m. litke , and e. j. chichilnisky .
_ the structure of large - scale synchronized firing in primate retina_. journal of neuroscience , * 29*(15):50225031 , 2009 .
a. tang , d. jackson , j. hobbs , w. chen , j. smith , and et al .
_ a maximum entropy model applied to spatial and temporal correlations from cortical networks in vitro_. journal of neuroscience , * 28 * ( 2008 ) , 505518 .
j. uhlhaas , g. pipa , g. b. lima , l. melloni , s. neuenschwander , d. nikoli , and w. singer .
_ neural synchrony in cortical networks : history , concept and current status_. frontiers in integrative neuroscience , * 3 * ( 2009 ) , 119 . | inspired by the problem of sensory coding in neuroscience , we study the maximum entropy distribution on weighted graphs with a given expected degree sequence . this distribution on graphs is characterized by independent edge weights parameterized by vertex potentials at each node . using the general theory of exponential family distributions , we prove the existence and uniqueness of the maximum likelihood estimator ( mle ) of the vertex parameters .
we also prove the consistency of the mle from a single graph sample , extending the results of chatterjee , diaconis , and sly for unweighted ( binary ) graphs .
interestingly , our extensions require an intricate study of the inverses of diagonally dominant positive matrices . along the way
, we derive analogues of the erds - gallai criterion of graphical sequences for weighted graphs . |
one of the exciting developments in statistics during the last decade has been the development of the theory and methodologies for dealing with high - dimensional data . the term _ high dimension _ is primarily interpreted as meaning that the dimensionality of the observed multivariate data is comparable to the available number of replicates or subjects on which the measurements on the different variables are taken .
this is often expressed in the asymptotic framework as @xmath3 , where @xmath0 denotes the dimension of the observation vectors ( forming a triangular array ) and @xmath2 the sample size .
much of this development centered on understanding the behavior of the sample covariance matrix and especially its eigenvalues and eigenvectors , due to their role in dimension reduction , in estimation of population covariances and as building block of numerous inferential procedures for multivariate data .
comprehensive reviews of this topic can be found in johnstone @xcite and paul and aue @xcite .
one most notable high - dimensional phenomena associated with sample covariance matrices is that the sample eigenvalues do not converge to their population counterparts if dimension and sample sizes remain comparable even as the sample size increases . a formal way to express
this phenomenon is through the use of the _ empirical spectral distribution _ ( esd ) , that is , the empirical distribution of the eigenvalues of the sample covariance matrix . the celebrated work of marenko and pastur @xcite shows that if one studies a triangular array of random vectors @xmath4 , whose components form independent , identically distributed ( i.i.d . ) random variables with zero mean , unit variance and finite fourth moment , then as @xmath5 such that @xmath6 , the esd of @xmath7 converges almost surely to a nonrandom probability distribution known as the marenko pastur distribution . since this highly influential discovery
a large body of literature under the banner of random matrix theory ( rmt ) has been developed to explore the properties of the eigenvalues and eigenvectors of large random matrices .
one may refer to anderson et al .
@xcite , bai and silverstein @xcite and tao @xcite to study various aspects of this literature .
many important classes of high - dimensional data , particularly those arising in signal processing , economics and finance , have the feature that in addition to the dimensional correlation , the observations are correlated in time .
classical models for time series often assume a stationary correlation structure and use spectral analysis methods or methods built on the behavior of the sample autocovariance matrices for inference and prediction purposes . in spite of this , to our knowledge , no work exists that analyzes the behavior of the sample autocovariance matrices of a time series from a random matrix perspective , even though jin et al .
@xcite have dealt recently covered autocovariance matrices in the independent case .
a striking observation is that , in the high - dimensional scenario , the distribution of the eigenvalues of the symmetrized sample autocovariance of a given lag order tends to stabilize to a nondegenerate distribution even in the setting where the observations are i.i.d .
this raises questions about the applicability of sample autocovariance matrices as diagnostic tools for determining the nature of temporal dependence in high - dimensional settings .
thus a detailed study of the phenomena associated with the behavior of the esd of the sample autocovariance matrices when the observations have both dimensional and temporal correlation is of importance to gain a better understanding of the ways in which the dimensionality affects the inference for high - dimensional time series .
all the existing work on high - dimensional time series dealing with the limiting behavior of the esd focuses on the sample covariance matrix of the data when @xmath4 are @xmath0-dimensional observations recorded in time and @xmath5 such that @xmath8 .
this includes the works of jin et al .
@xcite , who assume the process @xmath9 has i.i.d .
rows with each row following a causal arma process .
pfaffel and schlemm @xcite and yao @xcite extend this framework to the setting where the rows are arbitrary i.i.d .
stationary processes with short - range dependence .
zhang @xcite , paul and silverstein @xcite and el karoui @xcite , under slightly different assumptions , consider the limiting behavior of the esd of the sample covariance when the data matrices are of the form @xmath10 where @xmath11 and @xmath12 are positive semidefinite matrices , and @xmath13 has i.i.d .
entries with zero mean , unit variance and finite fourth moment .
this model is known as the separable covariance model , since the covariance of the data matrix is the kronecker product of @xmath11 and @xmath12 .
if the rows indicate spatial coordinates and columns indicate time instances , then this model implies that spatial ( dimensional ) and temporal dependencies in the data are independent of each other .
the work of this paper is also partly related to the results of hachem et al .
@xcite , who prove the existence of the limiting esd for sample covariance of data matrices that are rectangular slices from a bistationary gaussian process on @xmath14 . in this paper
, the focus is on a class of time series known as linear processes [ or @xmath15 processes ] .
the assumptions to be imposed in section [ secmain ] imply that , up to an unknown rotation , the coordinates of the linear process , say @xmath9 , are uncorrelated stationary linear processes with short range dependence . extending the work of jin et al .
@xcite to the time series case , the goal is to relate the behavior of the esd of the lag-@xmath16 symmetrized sample autocovariances , defined as @xmath17 , with @xmath18 denoting complex conjugation , to that of the spectra of the coefficient matrices of the linear process when @xmath5 such that @xmath8 .
this requires assuming certain stability conditions on the joint distribution of the eigenvalues of the coefficient matrices which are described later .
the class of models under study here includes the class of causal autoregressive moving average ( arma ) processes of finite orders satisfying the requirement that the coefficient matrices are simultaneously diagonalizable and the joint empirical distribution of their eigenvalues ( when diagonalized in the common orthogonal or unitary basis ) , converges to a finite - dimensional distribution .
the results are expressed in terms of the _ stieltjes transform _ of the esd of the sample autocovariances .
specifically , it is shown that the esd of the symmetrized sample autocovariance matrix of any lag order converges to a nonrandom probability distribution on the real line whose stieltjes transform can be expressed in terms a unique _ stieltjes kernel_. the definition of the stieltjes kernel involves integration with respect to the limiting joint empirical distribution of the eigenvalues of the coefficient matrices as well as the spectral density functions of the one - dimensional processes that correspond to the coordinates of the process @xmath9 , after rotation in the common unitary or orthogonal matrix that simultaneously diagonalizes the coefficient matrices .
thus this result neatly ties the dimensional correlation , captured by the eigenvalues of the coefficient matrices , with the temporal correlation , captured by the spectral density of the coordinate processes .
the main contributions of this paper are the following : ( i ) a framework is provided for analyzing the behavior of symmetrized autocovariance matrices of linear processes ; ( ii ) for linear processes satisfying appropriate regularity conditions , a concrete description of the limiting stieltjes transform is given in terms of the limiting joint esd of the coefficient matrices and the spectral density of the coordinate processes after a rotation of the coordinates of the observation .
extensions to these main results are ( iii ) the characterization of the behavior of the esd of autocovariances of linear filters applied to the observed process ; ( iv ) the description of the esds of a class of tapered estimates of the spectral density operator of the observed process that can be used to analyze the long - run variance and spectral coherence of the process .
these contributions surpass the work in the existing literature dealing with high - dimensionality effects for time series in two different ways .
first , the class of time series models that are analyzed in detail encompasses the setting of stationary i.i.d .
rows studied by jin et al .
@xcite , pfaffel and schlemm @xcite and yao @xcite , as well as the setting of separable covariance structure studied by zhang @xcite , paul and silverstein @xcite and el karoui @xcite .
the proofs of the main results also require more involved arguments .
they are partly related to the constructions in hachem et al .
@xcite , but additional technical arguments are needed to go beyond gaussanity .
the results are also related to the work of hachem et al .
@xcite , who studies limiting spectral distributions of covariance matrices for data with a given variance profile .
the connection is through the fact that after an approximation of lag operators by circulant shift matrices , and appropriate row and column rotations , the data matrix in our setting can be equivalently expressed as a matrix with independent entries and with a variance profile related to the spectral densities of the different coordinates of the time series .
second , the framework allows for a unified analysis of the esd of symmetrized autocovariance matrices of all lag orders as well as that of the tapered spectral density operator .
none of the existing works deals with the behavior of autocovariances for time series ( note again that jin et al .
@xcite treat the i.i.d .
case ) , and this analysis requires a nontrivial variation of the arguments used for dealing with the stieltjes transform of the sample covariance matrix .
moreover , even though we stick to the setting where the coefficient matrices are hermitian and simultaneously diagonalizable , the main steps in the derivation , especially the construction of a `` deterministic equivalent '' of the resolvent of the symmetrized autocovariance matrix , is very general and can be applied to linear processes with structures that go beyond the settings studied in this paper , for example , when the simultaneous diagonalizability of the coefficient matrices is replaced by a form of simultaneous block diagonalizability , even though the latter is not pursued in this paper due to lack of clear statistical motivation .
the existence and uniqueness of the limits of the resulting equations and their solutions is the key to establishing the existence of liming esds of the autocovariances .
this step requires certain regularity conditions on the coefficient matrices and is not pursued beyond the setting described in section [ secmain ] . a number of potential applications , for example
, to problems in signal processing , and dynamic and static factor models , are discussed in section [ secexandapp ] .
the remaining sections of the paper are organized as follows .
extensions of the main results in section [ secmain ] are discussed in section [ secex ] .
the outcomes of a small simulation study are reported in section [ secsimulations ] , while the proofs of the main results are provided in sections [ secstructure][secpreal ] .
several technical lemmas are collected in the online supplemental material ( sm ) @xcite .
let @xmath19 denote the set of integers .
a sequence of random vectors @xmath9 with values in @xmath20 is called a linear process or moving average process of order infinity , abbreviated by the acronym ma(@xmath21 ) , if it has the representation @xmath22 where @xmath23 denotes a sequence of independent , identically distributed dimensional random vectors whose entries are independent and satisfy @xmath24=0 $ ] , @xmath25=1 $ ] and @xmath26<\infty$ ] , where @xmath27 denotes the @xmath28th coordinate of @xmath29 . in the complex - valued case
this is meant as @xmath30=\mathbb{e}[\operatorname { im}(z_{jt})^2]=1/2 $ ] .
it is also assumed that real and imaginary parts are independent .
let further @xmath31 , the identity matrix . to ensure finite fourth moments for @xmath32 and a sufficiently fast decaying weak dependence structure , assumption [ asmainfty ] below lists several additional conditions imposed on the coefficient matrices @xmath33 .
the results presented in this paper are concerned with the behavior of the symmetrized lag-@xmath16 sample autocovariances @xmath34 assuming observations for @xmath4 are available . for @xmath35 ,
this definition gives the covariance matrix @xmath36 discussed in the .
note that in order to make predictions in the linear process setting , it is imperative to understand the second - order dynamics which are captured in the population autocovariance matrices @xmath37 $ ] , @xmath38 , as all of the popular prediction algorithms such as the durbin levinson and innovations algorithms are starting from there ; see , for example , ltkepohl @xcite .
the set - up in ( [ eqmainfty ] ) provides a ( strictly ) stationary process and consequently the definition of @xmath39 does not depend on the value of @xmath40 .
the main goal of this paper is to analyze the behavior of the matrices @xmath41 , which can be viewed as a special sample counterpart to the corresponding @xmath42 , in the high - dimensional setting for which @xmath43 is a function of the sample size such that @xmath44 thereby extending the above mentioned marenko pastur - type results to more general time series models and to autocovariance matrices
. we can weaken requirement ( [ eqpn ] ) to `` @xmath45 bounded away from zero and infinity , '' in which case , the asymptotic results hold for subsequences @xmath46 satisfying @xmath47 converging to a positive constant @xmath48 , provided that the structural assumptions on the model continue to hold .
let then @xmath49 denote the empirical spectral distribution ( esd ) of @xmath50 given by @xmath51 where @xmath52 are the eigenvalues of @xmath41 .
the proof techniques for establishing large - sample results about @xmath53 are based on exploiting convergence properties of stieltjes transforms , which continue to play an important role in verifying theoretical results in rmt ; see , for example , paul and aue @xcite for a recent summary .
the stieltjes transform of a distribution function @xmath54 on the real line is the function @xmath55 where @xmath56 denotes the upper complex half plane .
it can be shown that @xmath57 is analytic on @xmath58 and that the distribution function @xmath54 can be reconstructed from @xmath57 using an inversion formula ; see @xcite . in order to make statements about @xmath49
, the following additional assumptions on the coefficient matrices @xmath33 are needed .
let @xmath59 and @xmath60 denote the positive and nonnegative integers , respectively .
[ asmainfty ] ( a ) the matrices @xmath61 are simultaneously diagonalizable random hermitian matrices , independent of @xmath23 and satisfying @xmath62 for all @xmath63 and large @xmath0 with @xmath64 note that one can set @xmath65 .
\(b ) there are continuous functions @xmath66 , @xmath63 , such that , for every @xmath0 , there is a set of points @xmath67 , not necessarily distinct , and a unitary @xmath68 matrix @xmath69 such that @xmath70 and @xmath71 .
[ note that the functions @xmath72 are allowed to depend on @xmath43 as long as they converge to continuous functions as @xmath73 uniformly . ]
\(c ) with probability one , @xmath74 , the esd of @xmath75 , converges weakly to a nonrandom probability distribution function @xmath76 on @xmath77 as @xmath78 .
let @xmath79 $ ] denote the matrix collecting the coefficient matrices of the linear process @xmath9 .
define the transfer functions @xmath80 as well as the power transfer functions @xmath81 note that the contribution of the temporal dependence of the underlying time series on the asymptotic behavior of @xmath53 is quantified through @xmath82 . specifically ,
@xmath83 with @xmath84 as in part ( b ) of assumption [ asmainfty ] is ( up to normalization ) the spectral density of the @xmath28th coordinate of the process rotated with the help of the unitary matrix @xmath69 . with these definitions ,
the main results of this paper can be stated as follows .
[ thesdmainfty ] if a complex - valued linear process @xmath32 with independent , identically distributed @xmath27 , @xmath85=0 $ ] , @xmath30=\mathbb { e}[\operatorname{im}(z_{jt})^2]=1/2 $ ] , @xmath86 and @xmath87 are independent , and @xmath88<\infty$ ] , satisfies assumption [ asmainfty ] , then , with probability one and in the high - dimensional setting ( [ eqpn ] ) , @xmath53 converges to a nonrandom probability distribution @xmath89 with stieltjes transform @xmath90 determined by the equation @xmath91^{-1}\,df^{\mathbf{a}}(\lambda),\ ] ] where @xmath92\to\mathbb{c}^+$ ] is a stieltjes kernel ; that is , @xmath93 is the stieltjes transform of a measure with total mass @xmath94 for every fixed @xmath95 $ ] , whenever @xmath96 .
moreover , @xmath97 is the unique solution of @xmath98\\[-12pt]\nonumber & & \qquad = \int\biggl[\frac{1}{2\pi}\int _ 0^{2\pi } \frac{\cos(\tau\nu^\prime)h(\lambda,\nu^\prime)}{1+c\cos(\tau \nu^\prime)k_\tau(z,\nu^\prime)}\,d\nu^\prime - z \biggr]^{-1}h(\lambda,\nu ) \,df^{\mathbf{a}}(\lambda),\hspace*{-20pt}\end{aligned}\ ] ] subject to the restriction that @xmath97 is a stieltjes kernel . otherwise , if @xmath99 , then @xmath100 is identically zero on @xmath58 and so still satisfies ( [ eqktau ] ) .
[ coesdmainfty ] if a real - valued linear process @xmath32 with independent , identically distributed real - valued @xmath27 , @xmath24=0 $ ] , @xmath101=1 $ ] and @xmath102<\infty$ ] , satisfies assumption [ asmainfty ] with real symmetric coefficient matrices @xmath61 , then the result of theorem [ thesdmainfty ] is retained . when each coefficient matrix @xmath33 is a multiple of the identity matrix ,
that is , @xmath103 where @xmath104 is a sequence of real numbers satisfying @xmath105 , the result of theorem [ thesdmainfty ] reduces to the results obtained in pfaffel and schlemm @xcite and yao @xcite .
one can relax the assumption of simultaneous diagonalizability of the coefficient matrices of the linear process to certain forms of near - simultaneous diagonalizability , so that the conclusions of theorem [ thesdmainfty ] continue to hold for linear processes where the ma coefficients are toeplitz matrices whose entries decay away from the diagonal at an appropriate rate .
specifically , if @xmath106 is the toeplitz matrix with @xmath28th row equaling @xmath107 , for the bi - infinite sequence @xmath108 satisfying the condition @xmath109 for some @xmath110 , which in particular implies assumption [ asmainfty](a ) by the gershgorin theorem , then the existence of the limiting esd of symmetrized autocovariance matrices can be proved . for brevity , instead of giving a thorough technical argument
, we only provide the main idea of proof .
first , the @xmath15 series is approximated by an @xmath111 series with @xmath112 , using arguments along the line of section [ secintlin ] .
second , banding with bandwidth @xmath113 is applied to the coefficient matrices @xmath106 .
it can be shown through an application of norm inequality , that the limiting spectral behavior is unchanged under the banding so long as @xmath114 under ( [ eqpn ] ) .
third , circulant matrices are constructed from the banded toeplitz matrices by periodization .
the resulting matrices are therefore simultaneously diagonalizable , and the eigenvalues of the @xmath115th approximate coefficient matrix approximate the transfer function of the sequence @xmath116 .
the limiting spectral behavior is seen to be unchanged after the use of the rank inequality so long as @xmath117 under ( [ eqpn ] ) .
the rest of the derivations follow the arguments in the proof of theorem [ thesdmainfty ] .
while this particular result is related to the work of hachem et al .
@xcite , who study the convergence of the empirical distribution of the sample covariance matrix of rectangular slices of bistationary gaussian random fields , @xcite does not cover the transition to non - gaussian processes or the spectral behavior of sample autocovariance matrices .
several extensions of theorem [ thesdmainfty ] are discussed in section [ secex ] below .
the proof steps needed in order to verify the main result are outlined in section [ secstructure ] , and the details are in sections [ secgauss][secpreal ] .
the online sm @xcite contains additional technical lemmas .
in this section , let @xmath9 be the causal arma(@xmath119 ) process given by the stochastic difference equations @xmath120 where @xmath121 and @xmath122 are , respectively , the matrix - valued autoregressive and moving average polynomials in the lag operator @xmath123 for which it is assumed that @xmath124 and @xmath125 .
moreover , @xmath126 with entries possessing finite fourth moments . under these conditions
@xmath9 admits the ma(@xmath21 ) representation @xmath127 with @xmath128 .
assume now further that @xmath129 and @xmath130 are simultaneously diagonalizable
. then @xmath131 and @xmath132 , where @xmath133 and @xmath134 such that @xmath135 and @xmath136 .
with regard to assumption [ asmainfty ] , let @xmath137 .
part ( c ) of the assumption then requires almost sure weak convergence of the esd of @xmath138 to a nonrandom probability distribution function on @xmath139 .
moreover , using that for each coordinate , @xmath140 it follows that @xmath141 with @xmath142 and @xmath143 for @xmath144 .
this illustrates part ( b ) of assumption [ asmainfty ] .
the summability conditions stated in part ( a ) are clearly satisfied .
generalization to arbitrary causal arma models follows in a similar fashion . in this section ,
the situation of time series with independent rows is considered .
our results describe the limiting esd of the symmetrized sample autocovariances in the setting where the @xmath28th row of the time series , denoted @xmath145 , is given by @xmath146 where : ( i ) the @xmath27 s are independent , identically distributed real- or complex - valued random variables with mean zero , unit variance and finite fourth moments ; ( ii ) the @xmath72 s are continuous functions from @xmath147 satisfying @xmath148 and the summability condition @xmath149 ; ( iii ) the @xmath150 s are i.i.d .
realizations from an @xmath151-dimensional probability distribution denoted by @xmath152 . if the supremum in condition ( ii ) is taken over @xmath77 , condition ( iii ) can be weakened to require that the empirical distribution of the @xmath150 s converges almost surely to a nonrandom distribution @xmath153 .
let @xmath154 .
then the empirical distributions of the eigenvalues of the lag-@xmath16 symmetrized autocovariance matrices converge almost surely to a nonrandom probability distribution @xmath89 with stieltjes transform @xmath90 determined by equations ( [ eqstau ] ) and ( [ eqktau ] ) , where @xmath100 is as in theorem [ thesdmainfty ] .
this is in the spirit of the works of jin et al .
@xcite , pfaffel and schlemm @xcite and yao @xcite , who studied the sample covariance case with @xmath155 for all @xmath156 , hachem et al .
@xcite , who considered the sample covariance case for stationary gaussian fields and jin et al .
@xcite , who studied the symmetrized sample autocovariance case with @xmath157 and @xmath158 for @xmath159 ( i.e. , when the @xmath160 s are i.i.d . with zero mean and unit variance ) .
the results derived here can be useful in dealing with a number of important statistical questions .
signal detection in a noisy background is one of the most important problems in signal processing and communications theory .
often the observations are taken in time , and the standard assumption is that the noise is i.i.d . in time , referred to as white noise . however , in spatio - temporal signal processing , it is quite apt to formulate the noise as `` colored '' or correlated in time , as well as in the spatial dimension . the proposed model for the time series is a good prototype for such a noise structure
. thus the problem of detecting a low - dimensional signal embedded in dimensional noise , for example , through a factor model framework , can be effectively addressed by making use of the behavior of the esds of autocovariances of the noise .
another potential application of the results is in building diagnostic tools for high - dimensional time series . by focusing on the esds of the autocovariances for various lag orders , or that of a tapered estimate of the spectral density operator
, one can infer about the nature of dependence , provided the model assumptions hold .
the proposed model also provides a broad class of alternatives for the hypothesis of independence of observations in settings where those observations are measured in time .
finally , in practical applications , it is of interest whether the spectrum of the coefficient matrices of the linear process can be estimated from the data .
the equations for the limiting stieltjes kernel and its relation to the stieltjes transform of the autocovariance matrices provide a tool for attacking this problem .
this aspect has been explored in the ph.d .
thesis of the first author @xcite and the methodology will be reported elsewhere .
forni and lippi @xcite describe a class of time series models that captures the subject specific variations of microeconomic activities .
this class of models , referred to as dynamic factor models ( dfm ) , has proved immensely popular in the econometrics community and beyond .
dfms have , for example , been used for describing the stock returns in @xcite , forecasting national accounts in @xcite , modeling portfolio allocation in @xcite and modeling psychological development in @xcite , as well as in many other applications .
important theoretical and inferential questions regarding dfms have been investigated in a series of papers by forni and lippi @xcite , forni et al .
@xcite and stock and watson @xcite , to name a few .
dfms have also shown early promise for applications to other interesting multivariate time series problems such as the study of fmri data .
a dfm can be described as follows . as in @xcite , let @xmath161 be the response corresponding to the @xmath28th individual / agent at time @xmath40 , modeled as @xmath162 the model specifies that @xmath161 is determined by a small , fixed number of underlying common factors @xmath163 and their lags , determined by the polynomials @xmath164 in the lag - operator @xmath123 , plus an idiosyncratic component @xmath165 assumed independent across individuals .
typically , @xmath166 is taken to be a stationary linear processes , independent across @xmath28 .
one of the key questions pertaining to dfm is the determination of the number of dynamic factors .
this question has been investigated by bai and ng @xcite , stock and watson @xcite and hallin and lika @xcite . unlike in pca , here
one has to deal with the additional problem of detecting the lag orders of the dynamic factors .
this can be approached through the study of the behavior of the extreme eigenvalues of the sample autocovariance matrices as in jin et al . @xcite .
the issue becomes even more challenging when the dimensionality of the problem increases .
in such settings , one expects that a form of phase transition phenomenon , well known in the context of a high - dimensional static factor model ( or spiked covariance model ) with i.i.d .
observations ( see , e.g. , baik and silverstein @xcite ) , will set in .
in particular , as jin et al .
@xcite argue , a dynamic factor will be detectable from the data only if the corresponding total signal intensity , as measured , for example , by the sum of the variances of the factor loadings , is above a threshold . moreover , the number of eigenvalues that lie outside the bulk of the eigenvalues of the symmetrized sample autocovariance of a certain lag order provide information about the lag order of the dfm . driven by the analogy with the static factor model with i.i.d .
observations , it is expected that the detection thresholds will depend on the dimension - to - sample size ratio , as well as the behavior of the bulk spectrum of the autocovariances of the idiosyncratic terms at specific lag orders , including the support of the limiting esd . equation ( [ eqidiot ] ) in section [ secind ] constitutes the `` null '' model for the dfm in which the dynamic common factors are absent . therefore follow - up studies on the different aspects of the esd of the symmetrized sample autocovariances of such processes will be helpful in determining the detection thresholds and estimation characteristics of high - dimensional dfms .
onatski @xcite describes a model for production @xmath167 , at time @xmath40 , involving @xmath0 different industries in an economy that is given by the equations @xmath168 model ( [ eqproductionmodel ] ) is a static factor model in which @xmath169 vectors @xmath170 denote the ( unobserved ) common static factors , @xmath171 denote the ( unobserved ) factor scores consisting of independent time series corresponding to different factors @xmath172 and @xmath173 denote the @xmath169 vectors of idiosyncratic components .
the entries of the matrix @xmath174 indicate the interactions among the different industries . in the following an enhanced version of the model
is considered where the economy is thought to be divided into a finite number of distinct sectors for which the interaction across the sectors is assumed `` weak '' in a suitable sense to be described .
in addition , the assumption of separable covariance structure of the @xmath173 made in @xcite is relaxed by requiring instead that the temporal variation in @xmath173 for all the industrial units within a sector is the same and is stationary in time .
this assumption means that the component of the vector @xmath173 corresponding to a particular sector has a separable covariance structure with stationary time variation , and the components corresponding to different sectors are independent .
specifically , if there are @xmath175 sectors , we can divide @xmath174 into @xmath176 block matrices @xmath177 = \widetilde{\mathbf{w } } + \delta,\ ] ] where @xmath178 . if the sectors have no interaction at all , that is , if @xmath179 for all @xmath180 , then the corresponding data model is an instance of a blockwise separable covariance model .
`` weak interaction '' means that the norms of the off - diagonal blocks in the matrix @xmath174 are small .
more precisely , if @xmath181 as @xmath182 , then the limiting esd of the symmetrized autocovariances for the data matrix @xmath183 is the same as that of @xmath184 obtained by replacing @xmath174 by @xmath185 in ( [ eqproductionmodel ] ) . under the assumption of a linear process structure on the different components of @xmath173 , and a natural requirement on the stability of the singular values of @xmath185 , the existence and characterization of the limiting esds of the symmetrized autocovariances of @xmath186 can be dealt within the framework studied in section [ secmain ] .
these limiting esds will help in determining the detection thresholds for the static factors , or even dynamic factors , if the model were to be enhanced further .
this section discusses three different extensions of the main result .
the arguments for the proof are similar to that of the proof of theorem [ thesdmainfty ] and hence only a brief outline is provided .
moreover , the results stated here apply to both real- and complex - valued cases , the only difference being that , in the former case , the relevant matrices are real symmetric while , in the latter case , they are hermitian .
the first extension involves a rescaling of the process defined in ( [ eqmainfty ] ) .
thus it is assumed that @xmath187 where the processes @xmath23 and matrices @xmath188 satisfy assumption [ asmainfty ] , and the matrix @xmath189 is the square root of a @xmath68 positive semidefinite hermitian ( or symmetric ) matrix @xmath12 satisfying the following assumption .
[ asscaling ] let @xmath190 be as in assumption [ asmainfty ] .
then @xmath191 where @xmath192 is continuous and bounded on @xmath77 , and @xmath193 is nonempty .
as before , @xmath53 is defined to be the esd of the symmetrized autocovariance @xmath194 of lag order @xmath16 . if the linear process @xmath9 defined through ( [ eqxtscaled ] ) satisfies assumptions [ asmainfty ] and [ asscaling ] , then the statement of theorem [ thesdmainfty ] holds with the function @xmath195 replaced by @xmath196 .
the second extension is about the existence and description of the limiting esd of the autocovariances of linear filters of the process @xmath9 defined through ( [ eqxtscaled ] ) .
a linear filter of this process is of the form @xmath197 where @xmath198 is a sequence of real numbers for which the following summability condition is needed .
the sequence @xmath199 satisfies @xmath200 .
if the linear process @xmath9 defined through ( [ eqxtscaled ] ) satisfies assumptions [ asmainfty ] and [ asscaling ] , then the statement of theorem [ thesdmainfty ] holds with the function @xmath82 replaced by @xmath201 , where @xmath202 , @xmath95 $ ] .
this result follows using the properties of convolution and fourier transform .
the third extension is about estimation of the spectral density operator @xmath203 , \qquad\eta\in[0,2\pi].\ ] ] it is well known from classical multivariate time series analysis ( see , e.g. , chapter 10 of hamilton @xcite ) that the `` natural '' estimator that replaces the population autocovariance @xmath204 $ ] by the corresponding sample autocovariance may not be positive definite . in order to obtain positive definite estimators and a better bias - variance trade - off , it is therefore standard in the literature to consider certain tapered estimators with standard choices given , for example , by the bartlett and parzen kernels as described in @xcite . in the following , the behavior of a class of tapered estimators of @xmath205 , which are given by @xmath206,\ ] ]
is studied , where @xmath207 is a sequence of even functions and the quantities @xmath208 are known as tapering weights for which the following restriction is imposed . [ astapering ] ( i ) the even functions @xmath207 are such that @xmath209 for @xmath210 ; ( ii ) there exists an even function @xmath211 such that @xmath212 as @xmath213 and @xmath214 for some @xmath215 , for all @xmath216 ; ( iii ) @xmath217 .
an implication of this assumption is that the function @xmath218 defined by @xmath219 is well defined and is uniformly lipschitz , and @xmath220 converges to @xmath221 uniformly in @xmath222 .
examples of kernels @xmath223 are @xmath224 for @xmath225 and @xmath226 for @xmath227 .
it can be seen from assumption [ astapering ] that in the high - dimensional setting under consideration here , standard choices for tapering weights , such as those given by the bartlett and parzen kernels , are ruled out .
now the following generalization of theorem [ thesdmainfty ] is obtained .
[ thesdmainftyspectraldensity ] suppose that the linear process @xmath9 defined through ( [ eqxtscaled ] ) satisfies assumptions [ asmainfty ] and [ asscaling ] , and that the estimated spectral density operators @xmath228)$ ] are defined by ( [ eqtaperedspectraldensity ] ) with tapering weights satisfying assumption [ astapering ] .
then , with probability one and in the high - dimensional setting ( [ eqpn ] ) , for every @xmath229 $ ] , the esd of @xmath230 converges weakly to a probability distribution @xmath231 with stieltjes transform @xmath232 determined by the equation @xmath233^{-1 } \,df^{\mathbf{a}}(\lambda),\ ] ] where @xmath234\to\mathbb { c}^+$ ] is a stieltjes kernel ; that is , @xmath235 is the stieltjes transform of a measure with total mass @xmath236 for every fixed @xmath95 $ ] , whenever @xmath96 .
moreover , @xmath237 is the unique solution to @xmath238^{-1 } \\ & & \quad\qquad{}\hspace*{7pt}\times g_{\mathbf{b}}(\lambda)h(\lambda,\nu ) \,df^{\mathbf{a}}(\lambda),\nonumber\end{aligned}\ ] ] subject to the restriction that @xmath237 is a stieltjes kernel .
else , if @xmath99 , @xmath239 is identically zero on @xmath58 and so still satisfies the latter equation .
section [ secspecdens ] outlines the main argument needed to prove theorem [ thesdmainftyspectraldensity ] .
observations . left panel : @xmath240 ; right panel : @xmath241 .
top panel : @xmath242 ; middle panel : @xmath243 ; bottom panel : @xmath244 .
red curve : lsd for i.i.d . ; black curve : lsd for @xmath245 .
esds are corresponding to @xmath246 [ deep blue for i.i.d . , light green for @xmath245 ] and @xmath248 [ light blue for i.i.d .
, orange for @xmath245 ] . ] in this section , a small simulation study is conducted to illustrate the behavior of the lsd ( limiting esd ) of the symmetrized sample autocovariances @xmath41 for different lag orders @xmath16 when the observations are i.i.d .
( @xmath249 ) versus when they come from an @xmath245 process ( @xmath250 ) with a symmetric coefficient matrix @xmath251 . in this study ,
two different sets of @xmath252 are chosen such that @xmath253 and @xmath241 , respectively , for @xmath246 and 50 .
also , for the @xmath245 case , the esd of @xmath254 is chosen to be @xmath255 , where @xmath256 denotes the degenerate distribution at @xmath257 .
the distribution of the @xmath27 s is chosen to be i.i.d .
@xmath258 . in figure
[ figsimul2050 ] , the esd of @xmath41 is plotted for the lags @xmath259 , for one random realization corresponding to each setting . the theoretical limits ( c.d.f . ) for all cases are plotted as solid smooth curves [ red for i.i.d . and
black for @xmath245 ] .
these c.d.f.s are obtained through numerically inverting the stieltjes transform @xmath90 of the lsd of @xmath41 using the inversion formula for stieltjes transforms ; cf .
@xcite , @xcite . here
@xmath90 is obtained from equations ( [ eqstau ] ) and ( [ eqktau ] ) where @xmath260 and @xmath261 .
given @xmath262 , the stieltjes kernel @xmath263 for the @xmath245 case is solved numerically by using calculus of residues .
the computational algorithm can be found in the first author s ph.d .
thesis @xcite .
the graphs clearly show the distinction in behavior of esd of symmetrized sample autocovariances between the i.i.d . and the @xmath245 processes .
the lsds of @xmath41 for the i.i.d .
case for @xmath243 and 2 are the same , which follows immediately from equations ( [ eqstau ] ) and ( [ eqktau ] ) that reduce to a single equation since the i.i.d .
case corresponds to @xmath264 .
the behavior of the lsds in the @xmath245 case is distinctly different .
it can be shown that lsds of @xmath41 in the @xmath245 case converge to the lsd for @xmath265 in the i.i.d .
case as @xmath16 becomes larger . owing to space constraints ,
the graphical displays for higher lags are omitted .
another important feature is that the lsds approximate the esds quite well even for @xmath0 as small as 20 , indicating a fast convergence .
in this section , the overall proof strategy is briefly outlined , and the intuition behind the individual steps is developed for simpler first - order moving average , @xmath245 , time series .
the key ideas in the proof of theorem [ thesdmainfty ] consist of showing that : * there is a unique stieltjes kernel solution @xmath97 to equation ( [ eqktau ] ) ; * almost surely , the stieltjes transform of @xmath53 , say , @xmath266 converges pointwise to a stieltjes transform @xmath90 which will be identified with @xmath89 ; * @xmath53 is tight . to achieve the second item
, one can argue as follows .
first , replace the original linear process observations @xmath4 with transformed vectors @xmath267 that are serially independent .
second , replace the symmetrized lag-@xmath16 autocovariance matrix @xmath41 by a transformed version @xmath268 built from @xmath267 . a heuristic formulation for the simpler gaussian @xmath245 case is given below in some detail .
once these two steps have been achieved , the proof proceeds by verifying some technical conditions with the help of classical rmt results , available , for example , in the monograph bai and silverstein @xcite . in the following ,
let @xmath32 denote the @xmath0-dimensional @xmath245 process given by the equations @xmath269 where @xmath23 are assumed complex gaussian in addition to the requirements of section [ secmain ] .
for this time series , the conditions imposed through assumption [ asmainfty ] simplify considerably with part ( a ) reducing to the condition that the eigenvalues of @xmath270 be uniformly bounded and part ( b ) being satisfied by choosing @xmath271 as the identity . moreover , @xmath272 $ ] , @xmath273 , @xmath274 , @xmath275 and @xmath276 , implying for each @xmath277 , @xmath82 is the spectral density ( up to normalization ) of a univariate @xmath245 process with parameter @xmath277 .
the transformation to serial independence requires two steps , the first consisting of an approximation of the lag operator by a circulant matrix and the second of a rotation using the complex fourier basis to achieve independence .
accordingly , let @xmath278 \quad\mbox{and}\quad{\tilde{\mathbf{l}}}=[e_n\dvtx e_1\dvtx\cdots\dvtx e_{n-1}]\ ] ] be the @xmath279 lag operator and its approximating circulant matrix , respectively , where @xmath280 denotes the @xmath2-dimensional zero vector and @xmath281 the @xmath28th canonical unit vector in @xmath282 taking the value 1 in the @xmath28th component and 0 elsewhere . since @xmath283 is a circulant matrix , its spectral decomposition is @xmath284 , @xmath285 , where @xmath286 , @xmath287 and @xmath288 the vector whose @xmath28th entry is @xmath289 .
it follows that @xmath283 diagonalizes in the complex fourier basis with the usual fourier frequencies .
let @xmath290 and @xmath291 $ ] denote the corresponding eigenvalue and eigenvector matrices , so that @xmath292 .
using @xmath293 $ ] and @xmath294 $ ] , the @xmath245 process ( [ eqma1 ] ) can be transformed into @xmath295 where @xmath296 constitutes a redefinition of @xmath297 such that the first column is changed to @xmath298 , while all other columns are as in the original data matrix @xmath297 .
rotating in the complex fourier basis , the observations are transformed again into the vectors @xmath267 given by @xmath299={\mathbf{x}}_1 { \mathbf{u}}_{{\tilde{\mathbf{l}}}}. \ ] ] observe that @xmath300 has independent columns . to see this ,
note first that @xmath301 possesses the same distribution as @xmath302 , since @xmath302 has ( complex ) gaussian entries and @xmath303 is a unitary matrix .
write then @xmath304,\ ] ] where @xmath305 $ ] , and independence of the columns ( and thus serial independence ) follows .
note also that @xmath306 and consequently @xmath307 , using the transfer function @xmath308 defined in ( [ eqtransferpsi ] ) .
the vectors @xmath267 give rise to approximations @xmath268 to the lag-@xmath16 symmetrized autocovariance matrices @xmath41 ; in particular @xmath268 and @xmath41 will be shown to have the same large - sample spectral behavior , irrespective of the distribution of the entries .
so let @xmath309^\tau\bigr){\mathbf{x}}_1^ * = \frac{1}n \sum_{t=1}^n\cos(\tau \nu_t){\tilde{x}}_t{\tilde{x}}_t^*,\ ] ] where the latter equality follows from several small computations using the quantities introduced in the preceding paragraph .
now , @xmath310^\tau\bigr){\mathbf{x}}_1^ * \biggr ) \\ & & { } + \biggl(\frac { 1}{2n}{\mathbf{x}}_1 \bigl({\mathbf{l}}^\tau+\bigl [ { \mathbf{l}}^*\bigr]^\tau\bigr){\mathbf{x}}_1^*-\frac { 1}{2n } { \mathbf{x}}_1 \bigl({\tilde{\mathbf{l}}}^\tau+\bigl[{\tilde{\mathbf{l}}}^*\bigr]^\tau \bigr){\mathbf{x}}_1^ * \biggr).\end{aligned}\ ] ] the rank of the first difference on the right - hand side of the last display is at most @xmath311 , since @xmath297 and @xmath296 differ only in the first column .
the rank of the second difference is at most @xmath312 .
the rank of @xmath313 is therefore at most @xmath314 . defining the resolvents
@xmath315 the stieltjes transforms corresponding to the esds of @xmath41 and @xmath268 are , respectively , given by @xmath316 \quad\mbox{and}\quad\tilde s_{\tau , p}(z)=\frac{1}p \operatorname{tr}\bigl[{\tilde{\mathbf{r}}}_\tau(z)\bigr].\ ] ] it follows from lemma s.2 that with probability one , the esds of @xmath41 and @xmath268 converge to the same limit , provided the limits exist .
we conclude from lemma s.3 that the esd of @xmath268 converges a.s . to a nonrandom distribution by showing that @xmath317 converge pointwise a.s . to the stieltjes transform of a probability measure , thus establishing a.s .
convergence of the esd of @xmath41 . in order to derive limiting equation ( [ eqktau ] ) , a finite sample counterpart is needed .
this can be derived as follows .
the transformed data @xmath300 gives rise to the transformed stieltjes kernel @xmath318\to\mathbb{c}^+$ ] given by @xmath319.\ ] ] following arguments typically used to establish the deterministic equivalent of a resolvent matrix , @xmath68 matrix - valued function solutions @xmath320 are needed such that , for sufficiently large @xmath0 , @xmath321\approx0\ ] ] for all @xmath322 and all @xmath68 hermitian matrix sequences @xmath323 with uniformly bounded norms @xmath324 . if one uses @xmath325 and the definition of @xmath326 , the latter approximate equation becomes @xmath327.\ ] ] section [ secgauss ] below is devoted to making precise the use of @xmath328 in the above equations and to showing that choosing @xmath329 with @xmath330 and @xmath331 , is appropriate . in order to verify the statement of theorem [ thesdmainfty ] for non - gaussian innovations
@xmath23 , two key ideas are invoked , namely showing that : * for any @xmath322 , the stieltjes transform @xmath332 concentrates around @xmath333 $ ] regardless of the underlying distributional assumption ; * the difference between the expectations @xmath333 $ ] under the gaussian model and the non - gaussian model is asymptotically negligible . to establish the concentration property of the first item ,
mcdiarmid s inequality is used to bound probabilities of the type @xmath334{\vert}\geq\varepsilon)$ ] for arbitrary .
these probabilities are then shown to converge to zero exponentially fast under ( [ eqpn ] ) . to establish the second item ,
the generalized lindeberg principle of chatterjee @xcite is applied . to this end , the argument @xmath335 is viewed as a parameter and @xmath332 as a function of the real parts @xmath336 and the imaginary parts @xmath337 of the innovation entries @xmath27 , for @xmath338 and @xmath285 .
the difference between gaussian @xmath339 and non - gaussian @xmath339 can then be analyzed by consecutively changing one pair @xmath340 from gaussian to non - gaussian , thereby expressing the respective differences in the expected stieltjes transforms as a sum of these entrywise changes .
these differences will be evaluated through a taylor series expansion , bounding certain third - order partial derivatives of @xmath341 .
details are given in section [ secpnon - gauss ] .
while the arguments established so far work in the same fashion also for ma(@xmath342 ) processes , certain difficulties arise when making the transition to the ma(@xmath21 ) or linear process case .
first , if one constructs the data matrix @xmath297 not from @xmath245 observations as above but from the linear process @xmath343 , then every column of @xmath297 is different from the corresponding column in the transformed matrix @xmath344 and not only the first column [ or the first @xmath342 columns for the ma(@xmath342 ) case ] .
second , for the @xmath245 case one can write the stieltjes transform @xmath332 as a function of @xmath345 variables @xmath336 and @xmath337 [ or @xmath346 variables for the ma(@xmath342 ) case ] , but for linear processes , even for finite @xmath0 , @xmath332 is a function of infinitely many @xmath336 and @xmath337 .
this makes their study substantially harder .
linear processes are thus , for the purposes of this paper , approached through truncation , that is , by approximation through finite - order ma processes @xmath347 whose order @xmath348 is a function of the dimension @xmath0 and therefore grows with the sample size under ( [ eqpn ] ) .
obviously @xmath349 is a necessary condition to make this approximation work .
however , @xmath348 can not grow too fast ( leading to the same difficulties in transitioning from the gaussian to the non - gaussian case as for the linear process itself ) or too slow ( showing that the lsds of the linear process and its truncated version are identical becomes an issue ) .
it turns out in section [ secplin ] that @xmath350 , with @xmath351 denoting the ceiling function , is an appropriate choice . to address the statements of theorem [ coesdmainfty ] , the arguments presented thus far have to be adjusted for real - valued innovations @xmath352 .
this is done using the eigen - decomposition of the coefficient matrices in the real fourier basis , after which arguments already developed for the complex case apply .
detailed steps are given in section [ secpreal ] .
the key step toward proving theorem [ thesdmainftyspectraldensity ] is to express @xmath353 as @xmath354 and then noticing that the matrix @xmath355 diagonalizes in the ( real or complex ) fourier basis with eigenvalues @xmath356 , for @xmath285 , so that the esd of @xmath230 can be approximated by the esd of the matrix @xmath357 , where @xmath358 is the @xmath40th column of the matrix @xmath359 , and @xmath360 denotes the @xmath361 fourier rotation matrix .
we give the main steps of the arguments leading to this result .
first , suppose that @xmath362 such that @xmath363 as @xmath73
. then we can write @xmath364 where @xmath365 with @xmath366 , @xmath367 and @xmath368 now , the following facts together with representation ( [ eqhatgammanetarepresentation ] ) and theorem a.43 of bai and silverstein @xcite ( rank inequality ) and lemma s.1 ( norm inequality ) prove the assertion : @xmath369}\max\{{\vert}\mathfrak{f}_{t_n , m_n}(\theta ) - \mathfrak{f}_{t_n}(\theta){\vert},{\vert}\mathfrak{f}_{t_n}(\theta ) -\mathfrak { f}_t(\theta){\vert}\ } \to0 $ ] , as @xmath73 ; rank@xmath370 ; @xmath371 and @xmath372
throughout , @xmath374 is treated as a sequence of nonrandom matrices , and all the arguments are valid conditionally on this sequence . in this section ,
the result of theorem [ thesdmainfty ] is first verified for the ma(@xmath342 ) process @xmath375 when @xmath27 s are i.i.d .
standard complex gaussian ; that is , real and imaginary parts of @xmath27 are independent normals with mean zero and variance one half . following the outline in section [ subsecintu ] , the data matrix @xmath376
$ ] is transformed into the matrix @xmath377 $ ] , with each column satisfying @xmath378 .
then , since the rank of @xmath379 is at most @xmath342 , by lemma s.2 , it follows that the esds of @xmath194 and @xmath380 have the same limit , if the latter exists . for simplicity ,
let @xmath381 . to keep notation more compact , the extra subscripts @xmath0 and @xmath16 ( indicating the lag of the autocovariance matrix under consideration )
are often suppressed when no confusion can arise .
for example , in ( [ eqapprox ] ) the notation @xmath382 will be preferred over the more complex @xmath383 . the proof is given in several steps .
first , a bound on the approximation error is derived if the stieltjes kernel @xmath175 in ( [ eqktau ] ) is replaced with its finite sample counterpart @xmath384 .
second , existence , convergence and continuity of the solution to ( [ eqapprox ] ) are verified .
third , tightness of the esds @xmath385 and convergence of the corresponding stieltjes transforms @xmath341 is shown .
the goal is to provide a rigorous formulation of ( [ eqapprox ] ) and a bound on the resulting approximation error .
the first step consists of giving a heuristic argument for the definition of @xmath382 in ( [ eqhtaup ] ) . to this end , note that @xmath386 it follows that , to achieve ( [ eqapprox ] ) , it is sufficient to solve @xmath387 .
( the use of @xmath328 will be clarified below . )
let @xmath388 and define the rank - one perturbation and its corresponding resolvent , respectively , given by @xmath389 using @xmath390+\frac{1}n\gamma_{\tau , t}x_tx_t^*)^{-1}$ ] and defining @xmath391 the sherman morrison formula and some matrix algebra lead to @xmath392 } \nonumber \\ & = & { \tilde{\mathbf{r}}}(z)\frac{1}n\sum_{t=1}^n \frac{\gamma_{\tau , t}{\mathcal{h}}_t}{1+c_n\gamma_{\tau , t}{\tilde{k}}(z,\nu_t ) } , \nonumber\end{aligned}\ ] ] thus validating the choice of @xmath382 as given in ( [ eqhtaup ] ) .
the @xmath328 sign is due to substituting @xmath393 , @xmath394 , @xmath395 with @xmath396 , @xmath397 , @xmath398 $ ] , respectively .
the arguments are made precise in the following theorem .
[ thapperr ] let @xmath399 be a hermitian matrix such that @xmath400 and @xmath322 .
if the assumptions of theorem [ thesdmainfty ] are satisfied , then @xmath401 \to0 \qquad\mbox{a.s.}\ ] ] under ( [ eqpn ] ) , where @xmath402 is defined in ( [ eqhtaup ] ) , and @xmath403 is the resolvent of the symmetrized autocvariance matrix @xmath404 with @xmath405 . observe that using ( [ eq * ] ) , ( [ eq * * ] ) and the definition of @xmath402 in ( [ eqhtaup ] ) , the a.s
convergence in ( [ eq * * * ] ) is shown to be equivalent to @xmath406 where @xmath407,\ ] ] with @xmath408 , @xmath409\\[-8pt]\nonumber \beta_{\tau , t } & = & \biggl(1+\frac{1}n\gamma_{\tau , t}\operatorname{tr}\bigl[{\tilde{\mathbf{r}}}(z ) { \mathcal{h}}_t\bigr ] \biggr)^{-1}.\end{aligned}\ ] ] decomposing further , write next @xmath410 , where @xmath411 , \\ d_{\tau , t}^{(2 ) } & = & \frac{1}{zp}\operatorname{tr } \bigl [ \gamma_{\tau , t}\tilde{\beta}_{\tau , t}{\tilde{\mathbf{r}}}_t(z ) ( { \tilde{{\mathcal{h}}}}_t-{\mathcal{h}}_t ) \bigl({\mathbf{i}}+{\mathbf{h}}_t(z ) \bigr)^{-1}{\mathbf{d}}\bigr ] , \\ d_{\tau , t}^{(3 ) } & = & \frac{1}{zp}\operatorname{tr } \bigl [ \gamma_{\tau , t}\tilde{\beta}_{\tau , t } { \tilde{\mathbf{r}}}_t(z ) { \mathcal{h}}_t \bigl ( \bigl({\mathbf{i}}+{\mathbf{h}}_t(z ) \bigr)^{-1}- \bigl({\mathbf{i}}+{\mathbf{h}}(z ) \bigr)^{-1 } \bigr){\mathbf{d}}\bigr ] , \\ d_{\tau , t}^{(4 ) } & = & \frac{1}{zp}\operatorname{tr } \bigl [ \gamma_{\tau , t}\tilde\beta_{\tau , t } \bigl({\tilde{\mathbf{r}}}_t(z)- { \tilde{\mathbf{r}}}(z ) \bigr){\mathcal{h}}_t \bigl({\mathbf{i}}+{\mathbf{h}}(z ) \bigr)^{-1}{\mathbf{d}}\bigr ] , \\ d_{\tau , t}^{(5 ) } & = & \frac{1}{zp}\operatorname{tr } \bigl [ \gamma_{\tau , t } ( \tilde\beta_{\tau , t}-\beta_{\tau , t } ) { \tilde{\mathbf{r}}}(z){\mathcal{h}}_t \bigl({\mathbf{i}}+{\mathbf{h}}(z ) \bigr)^{-1}{\mathbf{d}}\bigr],\end{aligned}\ ] ] with @xmath412 \biggr)^{-1},\ ] ] thus exhibiting the various approximations being made in the proof .
the borel
cantelli lemma provides that @xmath413 almost surely is implied if @xmath414 faster than @xmath415 for all @xmath416 . since @xmath417 , in order to verify ( [ eq * * * ] ) , it is sufficient to show that @xmath418 goes to zero faster than @xmath415 for all @xmath416 and @xmath419 .
the corresponding arguments are detailed in section s1.1 of the online sm . in this section ,
the proof of theorem [ thesdmainfty ] is completed for the complex gaussian innovation model . in
what follows , @xmath420 is without loss of generality assumed to be nonrandom , thereby restricting randomness to the innovations @xmath294 $ ] . for a fixed @xmath421 in the underlying sample space @xmath422 ,
notation such as @xmath423 will be utilized to indicate realizations of the respective random quantities .
noticing first that theorem [ thesdmainfty ] makes an almost sure convergence statement , a suitable subset @xmath424 with @xmath425 is determined .
this subset is used for all subsequent arguments . to this end , observe that since the matrix @xmath305 $ ] has i.i.d .
entries with zero mean and unit variance the norm of @xmath426 converges almost surely to a number not exceeding @xmath427 .
let @xmath428 denote the esd of @xmath426 , and define @xmath429 with suitably chosen @xmath430
. then @xmath431 .
define next @xmath432 , \\
\bar{d}_{\mathbf{i}}(z,\omega ) & = & \frac{1}{zp}\operatorname{tr } \bigl [ \bigl ( { \mathbf{i}}+{\mathbf{h}}(z ) \bigr)^{-1}+z{\tilde{\mathbf{r}}}(z,\omega ) \bigr].\end{aligned}\ ] ] let @xmath433 denote the set of complex numbers with rational real part and positive rational imaginary part and @xmath434_\mathbb{q}=[0,2\pi]\cap\mathbb{q}$ ] .
define the set @xmath435_\mathbb{q } \bigr\}.\ ] ] in view of theorem [ thapperr ] , it follows that @xmath436 a.s . and @xmath437 a.s . for all fixed @xmath438 and @xmath95_\mathbb{q}$ ] .
thus @xmath439 .
henceforth only @xmath440 , so that @xmath441 , are considered .
recall that @xmath442 $ ] .
the following theorem establishes existence of a stieltjes kernel solution to the equations in ( [ eqktau ] ) along a subsequence .
[ thexist ] suppose that the assumptions of theorem [ thesdmainfty ] are satisfied : for all @xmath443 and for all subsequences of @xmath444 , there exists another subsequence @xmath445 along which @xmath446 converges pointwise in @xmath322 and uniformly in @xmath95 $ ] to a limit @xmath447 analytic in @xmath335 and continuous in @xmath448 . for every subsequence
@xmath445 satisfying , @xmath449 satisfies ( [ eqktau ] ) for any @xmath322 .
moreover , @xmath447 is the stieltjes transform of a measure with mass @xmath94 , provided that @xmath96 .
\(a ) let @xmath450 .
then for a compact subset @xmath451 , @xmath452 for all @xmath95 $ ] and @xmath453 .
enumerate @xmath434_\mathbb{q}=\ { \nu_\ell\dvtx \ell\in\mathbb{n}\}$ ] .
let @xmath454 mean that @xmath455 is a further subsequence of @xmath456 , and let @xmath444 denote the original sequence .
an application of lemma 3 in geronimo and hill @xcite yields that , for any fixed @xmath443 , there is a sequence of subsequences @xmath457 so that @xmath458 converges to an analytic function of @xmath335 on @xmath459 .
a standard diagonal argument implies that @xmath460 converges to an analytic function of @xmath335 on @xmath434_\mathbb{q}$ ] . to simplify notation , write @xmath461 in place of @xmath460 .
observe that the thus obtained limit , which will be denoted by @xmath447 , is so far defined only on @xmath462_\mathbb{q}$ ] .
it remains to obtain the extension of the limit to @xmath462 $ ] .
note that lemma s.10 implies that , for any @xmath322 , @xmath463 are equicontinuous in @xmath448 and converge pointwise to @xmath464 on the dense subset @xmath434_\mathbb{q}$ ] of @xmath434 $ ] . by the arzela
ascoli theorem , @xmath465 converges therefore uniformly to a continuous function of @xmath448 on @xmath462 $ ] .
this limit , denoted again by @xmath447 , is also analytic on @xmath466 . to see this , pick @xmath467\setminus\mathbb{q}$ ] and a sequence @xmath468_\mathbb{q}$ ] such that @xmath469 .
then @xmath470 satisfies the conditions of lemma 3 in @xcite , and consequently there is a subsequence of @xmath470 that converges to an analytic function .
the limit of this subsequence has to coincide with @xmath471 by continuity on @xmath434 $ ] .
it follows that @xmath447 analytic .
\(b ) by lemma s.11 and the definition of @xmath472 , for all @xmath443 , @xmath447 satisfies ( [ eqktau ] ) for all @xmath473 .
so , by analyticity in @xmath335 , it holds for all @xmath466 .
suppose first that @xmath99
. then @xmath474 .
thus @xmath475 for all @xmath322 , and the claim is verified . for the remainder ,
suppose that @xmath96 .
showing that @xmath447 is a stieltjes transform of a measure with mass @xmath94 is equivalent to showing that @xmath476 is stieltjes transform of a borel probability measure .
let @xmath477 and @xmath478 denote the eigenvalue and eigenvector matrices of @xmath268 .
then @xmath479 $ ] is the stieltjes transform of a measure with mass @xmath480 $ ] . by the weak convergence of @xmath481 to @xmath76 , @xmath482 as @xmath483
this shows that @xmath484 for all @xmath485 .
since the diagonal entries of @xmath486 are bounded from above by @xmath487 , it follows that @xmath488 is the stieltjes transform of a measure @xmath489 , say , such that , for all real @xmath216 , @xmath490 , where @xmath491 denotes the esd of @xmath268 .
it follows from lemma s.12 that @xmath492 is a tight sequence .
therefore @xmath493 are the stieltjes transforms of a tight sequence of borel measures .
an application of lemma s.13 yields that @xmath494 is the stieltjes transform of a measure with mass @xmath495 , completing the proof . [ thunique ] if the assumptions of theorem [ thesdmainfty ] are satisfied , then there is a unique solution @xmath100 to ( [ eqktau ] ) that is analytic in @xmath322 and continuous in @xmath95 $ ] with @xmath263 being a stieltjes transform of a measure with mass @xmath496 .
suppose there are two solutions @xmath497 and @xmath498 to ( [ eqktau ] ) .
let @xmath499 and @xmath500 .
define then @xmath501 , @xmath502 .
note that@xmath503 and @xmath504 have nonpositive imaginary parts .
now @xmath505 and thus @xmath506 using the fact that @xmath507 is a stieltjes transform with mass bounded from above by @xmath487 , it follows that @xmath508 is bounded by @xmath509 .
thus @xmath510 if @xmath511 , then @xmath512 and thus @xmath513 which by continuity in @xmath448 implies that @xmath514 for any fixed @xmath322 .
since both solutions are analytic , the equality holds indeed for all @xmath322 .
this proves uniqueness . in the remainder of this section , the proof of theorem [ thesdmainfty ]
is completed for the gaussian ma(@xmath342 ) case .
this is done by establishing that ( a ) the convergence along subsequences as stated in theorem [ thexist ] holds indeed for the whole sequence and ( b ) the relevant esds converge . toward ( a )
, it is necessary to prove that , for every @xmath443 , @xmath515 converges to @xmath100 pointwise in @xmath322 and uniformly for @xmath95 $ ] under ( [ eqpn ] ) .
assume the contrary , and suppose that there are @xmath516 , @xmath467 $ ] and @xmath517 such that @xmath518 does not converge to @xmath519 .
by boundedness of @xmath518 , there is a subsequence @xmath445 along which @xmath518 converges to a limit different from @xmath520 . invoking theorems [ thexist ] and [ thunique ]
, there is a further subsequence @xmath521 of @xmath445 along which @xmath522 converges to @xmath523 uniformly in @xmath95 $ ] .
this is a contradiction .
it follows that for every @xmath443 , @xmath515 converges to @xmath100 pointwise in @xmath322 and @xmath524 $ ] .
an application of theorem [ thexist ] and the arzela
ascoli theorem shows that the convergence is uniform on @xmath434 $ ] .
note that , for any @xmath322 , @xmath525 $ ] converges to @xmath100 uniformly on @xmath434 $ ] .
since we have @xmath526 , assertion ( a ) follows . toward ( b ) , let @xmath443 .
it needs to be shown that @xmath527 on @xmath466 . by arguments as in the proof of lemma s.11
, it is already established that @xmath528 on @xmath529 .
now , for any compact @xmath451 and @xmath530 , @xmath531\bigr{\vert}\\ & = & \frac{1}{p}\bigl{\vert}\operatorname{tr } \bigl[(z_1-z_2 ) { \tilde{\mathbf{r}}}(z_1,\omega){\tilde{\mathbf{r}}}(z_2,\omega ) \bigr]\bigr{\vert}\\ & \leq & { \vert}z_1-z_2{\vert}\bigl(\min _ { z\in s}v \bigr)^2.\end{aligned}\ ] ] thus @xmath532 are equicontinuous in @xmath335 ( with @xmath0 and @xmath421 as parameters ) on @xmath533 . by arzela
ascoli , @xmath534 thus converges uniformly to @xmath535 on @xmath533 .
consequently @xmath528 on @xmath466 .
since @xmath536 , the esd of @xmath268 , is tight ( by lemma s.12 ) , it follows from lemmas s.13 and s.3 that @xmath535 is a stieltjes transform of a ( nonrandom ) probability measure , and @xmath537 converges a.s . to the distribution
whose stieltjes transform is given by @xmath90 .
since , by lemma s.2 , @xmath538 a.s .
, it follows that @xmath49 converges a.s . to the same limit , and
hence @xmath539 converges a.s . to @xmath535 .
the proof for the gaussian ma(@xmath342 ) case is complete . it can be checked that all the statements remain valid even if @xmath540 sufficiently slowly under ( [ eqpn ] ) , for example , if @xmath541 .
the extension of the result to non - gaussian innovations requires in its first step , a truncation argument , followed by a centering and rescaling of the innovations .
this section justifies that the symmetrized autocovariance matrices obtained from gaussian innovations and from their truncated , centered and rescaled counterparts have the same lsd .
the extension to the non - gaussian case is then completed in section [ secpnon - gauss ] .
since the underlying process is an ma(@xmath342 ) series , the observations @xmath4 are functions of the innovations @xmath542 . for @xmath338 and @xmath543 , define the quantities @xmath544 , \\
\bar{z}_{jt}^i&=&z_{jt}^i\mathbb{i } \bigl(\bigl{\vert}z_{jt}^i\bigr{\vert}<\sqrt{p}\bigr ) , \qquad\check{z}_{jt}^i=\bar z_{jt}^r- \mathbb{e}\bigl[\bar z_{jt}^i\bigr],\end{aligned}\ ] ] where @xmath545 and @xmath546 the indicator function .
correspondingly define @xmath547 and @xmath548 to be the autocovariance matrices obtained from @xmath549 and @xmath550 , @xmath551 , respectively .
if the assumptions of theorem [ thesdmainfty ] are satisfied , then a.s . under ( [ eqpn ] ) ,
@xmath552 where @xmath553 , @xmath554 and @xmath555 denote the esds of @xmath41 , @xmath547 and @xmath556 , respectively .
let @xmath557 and @xmath558 .
let further @xmath559 and @xmath560 , and note that these quantities are independent due to the assumed i.i.d .
structure on @xmath27 .
since the fourth moments of the latter random variables are assumed finite , it also follows that @xmath561<\infty$ ] and @xmath562<\infty$ ] .
observe next that the rank of a matrix does not exceed the number of its nonzero columns and that each nonzero @xmath27 causes at most @xmath563 nonzero columns .
recalling that @xmath564 , theorem a.44 of @xcite ( using @xmath565 and @xmath566 ) implies that , for any @xmath416 , @xmath567 let @xmath568^{-1}$ ] .
for @xmath0 large enough so that @xmath569 , hoeffding s inequality yields @xmath570 as well as @xmath571 the borel
cantelli lemma now implies that @xmath572 a.s .
, which is the first claim of the proposition . to verify the second , note that the equality @xmath573+i\mathbb{e}\bigl[\bar{z}_{11}^i\bigr ] \bigr ) \mathbf{1},\ ] ] with @xmath574 being the vector with all entries equal to 1 , shows that @xmath575 is independent of @xmath40 .
thus an application of lemma s.1 leads to @xmath576 which converges to 0 a.s . under ( [ eqpn ] ) .
this is the second assertion .
after truncation and centering , it does not necessarily follow that @xmath577=\mathbb{e}[{\vert}\check { z}_{11}^r+i\check{z}_{11}^i{\vert}^2]$ ] is equal to 1 . however , rescaling @xmath578 by dividing with @xmath579)^{1/2}$ ] ( in order to obtain unit variance ) does not affect the lsd because @xmath577\to1 $ ] under ( [ eqpn ] ) .
the detailed arguments follow as in section 3.1.1 of bai and silverstein @xcite .
this shows that the symmetrized autocovariances from @xmath27 and their truncated , centered and rescaled counterparts have the same lsd .
thus to simplify the argument , it can be assumed that the recentered process has variance one .
in this section , the results for the gaussian ma(@xmath342 ) case are extended to general innovation sequences @xmath580 satisfying the same moment conditions as their gaussian counterparts @xmath581 .
the processes of interest are then the two ma(@xmath342 ) processes @xmath582 define the symmetrized autocovariance matrix @xmath583 , the resolvent @xmath584 and the stieltjes transform @xmath585 $ ] . in the following , it will be shown that the esds of @xmath41 and @xmath586 converge to the same limit .
this is done via verifying that , for all @xmath322 , @xmath587 and @xmath332 converge to the same limit @xmath588 under ( [ eqpn ] ) , which in turn requires us to show that : @xmath589\to0 $ ] under ( [ eqpn ] ) for all @xmath322 ; @xmath590{\vert}\geq \varepsilon]\to0 $ ] under ( [ eqpn ] ) for all @xmath322 and @xmath416 .
part ( a ) requires the use of the lindeberg principle , and part ( b ) is achieved via an application of mcdiarmid s inequality . for the use in this section , redefine @xmath591 $ ] , define @xmath592 $ ] and let @xmath593 , @xmath594 and @xmath595 , @xmath596 be the corresponding matrices of real and imaginary parts .
claim ( a ) will be verified via the lindeberg principle developed in chatterjee @xcite .
this involves successive replacements of gaussian variables with non - gaussian counterparts in a telescoping sum . to this end , define an order on the index set @xmath597 by letting @xmath598 if either ( 1 ) @xmath599 or ( 2 ) @xmath600 and @xmath601 , so that one successively moves columnwise through the entries of a matrix .
moreover , let @xmath602 let @xmath603 denote the @xmath604 matrix given by the entries @xmath605 further let the @xmath604 matrix @xmath606 be equal to @xmath607 for all entries but the @xmath608th one , which is set to equal 0 , and define analogously the matrices @xmath609 and @xmath610 . these matrices determine how many of the original gaussian @xmath27 s have been replaced by the non - gaussian @xmath611 . in the following
, @xmath332 will be viewed as a function of @xmath593 and @xmath595 , fixing @xmath335 and @xmath0 as parameters , that is , @xmath612 .
similarly , let @xmath613 .
utilizing this notation , the quantity to be bounded in expectation can be written as @xmath614 \\ & & \quad\qquad { } + \sum_{(j , t)=(1,1-q)}^{(p , n ) } \bigl[s_p^z\bigl({\mathbf{v}}_{0,-q}^r,{\mathbf{v}}_{jt}^i\bigr)-s_p^z\bigl ( { \mathbf{v}}_{0,-q}^r,{\mathbf{v}}_{(j , t)-1}^i\bigr ) \bigr ] \\ & & \qquad = \delta_1+\delta_2,\end{aligned}\ ] ] where @xmath615 and @xmath616 are the real and imaginary part of the difference with @xmath617 and @xmath618 . in the following only the telescoping real parts will be discussed , as the imaginary parts can be estimated along the same lines . inserting @xmath606 , one obtains @xmath619 + \bigl[s_p^z \bigl({\bar{{\mathbf{v}}}}_{j ,
t}^r,{\mathbf{v}}_{p , n}^i \bigr)-s_p^z\bigl({\mathbf{v}}_{(j , t)-1}^r,{\mathbf{v}}_{p , n}^i\bigr ) \bigr ] \\ & = & \delta_{j , t}^{(1,1)}+\delta_{j , t}^{(1,2)}.\end{aligned}\ ] ] let @xmath620 denote the @xmath172th - order partial derivative of @xmath621 with respect to @xmath336 .
a taylor series expansion gives @xmath622}\bigl{\vert}\partial_{j , t,1}^{(3)}s_p^z \bigl(\alpha{\mathbf{v}}_{jt}^r+(1-\alpha){\bar{{\mathbf{v}}}}_{jt}^r , { \mathbf{v}}_{pn}^i\bigr)\bigr{\vert}\end{aligned}\ ] ] and @xmath623}\bigl{\vert}\partial_{j , t,1}^{(3)}s_p^z \bigl(\alpha{\mathbf{v}}_{(j , t)-1}^r+(1-\alpha){\bar{{\mathbf{v}}}}_{jt}^r , { \mathbf{v}}_{pn}^i\bigr)\bigr{\vert}.\end{aligned}\ ] ] the entries of the matrices @xmath593 , @xmath595 , @xmath594 and @xmath596 are all independent of each other and the first and second moments of the various real parts ( and imaginary parts ) coincide , so that the bound in the last two inequalities also hold for higher - order terms ( hot )
. this leads to @xmath624\bigr { \vert}&=&\biggl{\vert}\mathbb{e } \biggl [ \bigl(z_{jt}^r - w_{jt}^r \bigr)\partial_{j , t,1}^{(1)}s_p^z \bigl({\bar{{\mathbf{v}}}}^r_{jt},{\mathbf{v}}_{pn}^i\bigr ) \nonumber \\ & & \hspace*{15pt}{}+ \frac{1}2 \bigl [ \bigl(z_{jt}^r \bigr)^2-\bigl(w_{jt}^r \bigr)^2 \bigr]\partial_{j , t,1}^{(2)}s_p^z\bigl ( { \bar{{\mathbf{v}}}}^r_{jt},{\mathbf{v}}_{pn}^i\bigr)\pm \mathrm{hot } \biggr]\biggr{\vert}\nonumber\\[-8pt]\\[-8pt]\nonumber & \leq&\frac{1}6\mathbb{e } \bigl[\bigl{\vert}z_{jt}^r \bigr{\vert}^3 \max_{\alpha\in[0,1]}\bigl{\vert}\partial_{j , t,1}^{(3)}s_p^z\bigl(\alpha { \mathbf{v}}_{jt}^r+(1-\alpha){\bar{{\mathbf{v}}}}_{jt}^r , { \mathbf{v}}_{pn}^i\bigr)\bigr{\vert}\bigr ] \\ & & { } + \frac{1}6\mathbb{e } \bigl[\bigl{\vert}w_{jt}^r\bigr{\vert}^3 \max _ { \alpha\in[0,1]}\bigl{\vert}\partial_{j , t,1}^{(3)}s_p^z \bigl(\alpha{\mathbf{v}}_{(j , t)-1}^r+(1-\alpha){\bar{{\mathbf{v}}}}_{jt}^r , { \mathbf{v}}_{pn}^i\bigr)\bigr{\vert}\bigr].\nonumber\hspace*{-25pt}\end{aligned}\ ] ] dealing with the right - hand side of ( [ eqtaylor ] ) requires the computation and estimation of the third - order derivatives @xmath625 $ ] . focusing only on the first term of the right - hand side of ( [ eqtaylor ] ) ( the second can be handled similarly ) , lemma s.14 shows that this term converges to zero under ( [ eqpn ] ) if , almost surely under ( [ eqpn ] ) , @xmath626 } \biggl(\sum _ { \ell\in\mathcal{i}_+(t)}{\vert}{\mathbf{a}}_\ell{\vert}\bigl{\vert}{\tilde{x}}_{t+\tau + \ell}^\alpha\bigr{\vert}\nonumber\\[-8pt ] \label{eqremainder1 } \\[-8pt ] \nonumber & & \hspace*{165pt } { } + \sum_{\ell\in\mathcal{i}_-(t)}\bigl{\vert}{\tilde{x}}_{t-\tau+\ell}^\alpha \bigr{\vert}{\vert}{\mathbf{a}}_\ell{\vert}\biggr)^3 \biggr]\to 0,\hspace*{-20pt } \nonumber \\ & & \frac{q+1}{n^2p}\sum_{(j , t)=(1,1-q)}^{(p , n ) } \mathbb{e } \biggl[\bigl{\vert}z_{t , j}^r\bigr{\vert}^3\max_{\alpha\in[0,1 ] } \biggl(\sum _ { \ell\in\mathcal{i}_+(t)}{\vert}{\mathbf{a}}_\ell{\vert}\bigl{\vert}{\tilde{x}}_{t+\tau+\ell}^\alpha\bigr{\vert}\nonumber\\[-8pt ] \label{eqremainder2 } \\[-8pt]\nonumber & & \hspace*{165pt } { } + \sum_{\ell\in\mathcal{i}_-(t)}\bigl{\vert}{\tilde{x}}_{t-\tau+\ell}^\alpha
\bigr{\vert}{\vert}{\mathbf{a}}_\ell{\vert}\biggr ) \biggr]\to0,\hspace*{-20pt } \nonumber\end{aligned}\ ] ] where @xmath627 , @xmath628 , @xmath629=\sum_{\ell = 0}^q{\mathbf{a}}_\ell{\mathbf{v}}^\alpha{\mathbf{l}}^\ell{\mathbf{f}}$ ] , @xmath630 and @xmath631^\prime$ ] being an @xmath632 matrix .
the choice of @xmath633 $ ] only affects the value of the @xmath28th entry of @xmath634 . for @xmath635
, the notation @xmath636 is therefore preferred . by definition
, @xmath637 is the vector whose @xmath28th entry has a real part of zero .
let @xmath638 .
then @xmath639}\bigl{\vert}{\tilde{x}}^\alpha_{t\pm\tau+\ell } \bigr{\vert}\\ & & \qquad \leq\sum_{\ell^\prime\in\mathcal{j}_\pm(t,\ell)}\bar { \lambda } _ { { \mathbf{a}}_{\ell^\prime}}{\vert}z_{t\pm\tau+\ell-\ell^\prime}{\vert}+\bar { \lambda}_{{\mathbf{a}}_{\ell\pm\tau } } \bigl(\bigl{\vert}v_t^0\bigr{\vert}+\bigl{\vert}z_{j , t}^r\bigr{\vert}\bigr).\end{aligned}\ ] ] hence , setting @xmath640 , @xmath639}\sum_{\ell\in\mathcal{j}_\pm(t ) } { \vert}{\mathbf{a}}_\ell{\vert}\bigl{\vert}{\tilde{x}}^\alpha_{t\pm\tau+\ell}\bigr{\vert}\\ & & \qquad \leq\bar{\lambda}_{\mathbf{a}}\sum_{\ell =- q,\ell\neq\mp\tau}^q \bar{\lambda}_\ell^{\mathrm{rem}}{\vert}z_{t\pm\tau+\ell}{\vert}+ \bar{\lambda}_{\mathbf{a}}\bar{\lambda}_{\pm\tau}^{\mathrm{rem } } \bigl ( \bigl{\vert}v_t^0\bigr{\vert}+\bigl{\vert}z_{j , t}^r\bigr{\vert}\bigr).\end{aligned}\ ] ] using this , it follows from lemma s.15 that the left - hand sides of ( [ eqremainder1 ] ) and ( [ eqremainder2 ] ) converge to zero a.s .
, thus establishing that @xmath641\to0 $ ] . for a fixed @xmath335 ,
@xmath587 is a function of the @xmath642 vectors @xmath643 .
letting @xmath644 , these are now segmented into the groups @xmath645 , @xmath646 , possibly adding additional vectors to the last group to ensure all groups have the same length [ even though the value @xmath587 does not depend on the additions ] . to satisfy the conditions needed in order to apply mcdiarmid s inequality ,
note that a change of the values @xmath647 in one group to , say , @xmath648 , causes the values of @xmath649 to change to , say , @xmath650 . in the following ,
the focus is on changes applied to the first group of innovations @xmath651 .
consider the case @xmath652 and let @xmath653 @xmath654 and @xmath655 $ ] .
the goal is now to represent @xmath656 as a finite rank perturbation of @xmath547 in the form @xmath657 with appropriate @xmath658 and @xmath659 .
write @xmath660.\end{aligned}\ ] ] choosing @xmath661 and repeatedly utilizing ( s.3 ) with @xmath662 replaced by @xmath663 , it follows that @xmath664 for some appropriately chosen constant @xmath665 .
this bound holds for any of the @xmath151 groups of innovations .
mcdiarmid s inequality consequently implies , for any @xmath416 and a suitable constant @xmath666 , @xmath667\bigr{\vert}\geq\varepsilon\bigr ) \leq4\exp \biggl(-c_2\frac{\varepsilon^2v^2p^2}{mq^2 } \biggr ) \sim4\exp\biggl(-c_2 \frac{\varepsilon^2v^2c}{q}p \biggr).\ ] ] the right - hand side converges to zero at a rate faster than @xmath668 and concentration of the stieltjes transform around its mean is established , since the case @xmath669 can be handled in a similar fashion . note that the last argument remains valid if @xmath670 as @xmath78 at a sufficiently slow rate , for example , if @xmath671
let @xmath9 now denote a linear process . to complete the proof of theorem [ thesdmainfty ] , a truncation argument is invoked .
let @xmath672 denote the truncated process given by @xmath673 , @xmath674 , where @xmath348 depends on the dimension .
let @xmath675 be the process given by @xmath676 , @xmath677 , and denote by @xmath678 the lvy distance between distribution functions @xmath54 and @xmath679 .
[ lemlinproclevy ] if the assumptions of theorem [ thesdmainfty ] are satisfied and if @xmath680 , then @xmath681 under ( [ eqpn ] ) , where @xmath682 . by lemma s.1
, it suffices to show that @xmath683 \to0 $ ] a.s .
write @xmath684 and @xmath685
. then @xmath686 say . from repeated applications of the cauchy
schwarz inequality , we have @xmath687 & = & 2\operatorname{tr}\bigl({\mathbf{p}}^2\bigr ) + \operatorname{tr } \bigl({\mathbf{q}}^2\bigr ) + 2 \operatorname{tr}\bigl({\mathbf{p}}{\mathbf{p}}^*\bigr ) + 4 \operatorname{tr}({\mathbf{p}}{\mathbf{q } } ) \\ & \leq & 4 \operatorname{tr}\bigl({\mathbf{p}}{\mathbf{p}}^*\bigr ) + \operatorname{tr}\bigl({\mathbf{q}}^2\bigr ) + 4 \sqrt{\operatorname{tr}\bigl({\mathbf{p}}{\mathbf{p}}^*\bigr ) } \sqrt{\operatorname{tr}\bigl({\mathbf{q}}^2 \bigr)}.\end{aligned}\ ] ] since @xmath688 , @xmath689 $ ] , and by another application of cauchy schwarz , @xmath690 \biggr)^{1/2 } \biggl(\frac{1}{n^2}\operatorname{tr}\bigl[\bigl(\check{\mathbf{x}}\check{\mathbf{x}}^*\bigr)^2\bigr ] \biggr)^{1/2}.\ ] ] since it is easy to see that @xmath691 $ ] is stochastically bounded ( e.g. , by showing that the expectation is finite ) , it is enough to show that @xmath692 \to0 $ ] a.s .
this is established by showing that the sum @xmath693 ] < \infty$ ] , and then applying the borel
cantelli lemma .
to this end , note that @xmath694 \bigr ] \\ & & \qquad = \sum_{t=1}^n \sum _ { s=1}^n \mathbb{e } \bigl[\bigl{\vert}\check x_t^ * \check x_s\bigr{\vert}^2 \bigr ] \\ & & \qquad = \sum_{t=1}^n \sum _ { s=1}^n \sum_{\ell = q+1}^\infty \sum_{\ell'=q+1}^\infty\sum _ { m = q+1}^\infty\sum_{m'=q+1}^\infty \mathbb{e } \bigl[\operatorname{tr}\bigl(z_{t-\ell'}z_{t-\ell}^ * { \mathbf{a}}_\ell{\mathbf{a}}_m \\ & & \hspace*{204pt}{}\times z_{s - m}z_{s - m ' } { \mathbf{a}}_{\ell ' } { \mathbf{a}}_{m'}\bigr ) \bigr].\end{aligned}\ ] ] it is clear from the independence of @xmath29 s that the summands are nonzero only if the indices of @xmath695 s pair up .
direct calculation shows that the total contribution of all four types of pairings : ( i ) @xmath696 ; ( ii ) @xmath697 ; ( iii ) @xmath698 and ( iv ) @xmath699 can be bounded by @xmath700 for some @xmath215 , using the fact that @xmath0 and @xmath2 are of the same order .
thus since @xmath701 , @xmath702 this proves the result .
using gaussian innovations @xmath23 , let @xmath703=\break \sum_{\ell=0}^q{\mathbf{a}}_\ell{\mathbf{z}}{\tilde{\mathbf{l}}}^\ell{\mathbf{u}}_{{\tilde{\mathbf{l}}}}$ ] be the corresponding transformed data matrix .
then define @xmath704 , @xmath705 , @xmath706,@xmath707 , @xmath708 $ ] , @xmath709 $ ] , @xmath710 and @xmath711 .
then one verifies in a similar vein , as in the proofs of theorems [ thexist ] and [ thunique ] , that for all @xmath443 with @xmath472 defined in the beginning of section [ secgaussexconcon ] , under ( [ eqpn ] ) , @xmath712^{-1}h^{\mathrm{tr}}(\lambda,\nu)\,df_{p_\ell}^{\mathbf{a } } ( \lambda ) \to0,\ ] ] where @xmath713^{-1}h^{\mathrm{tr}}(\lambda,\nu _ { tj})$ ] .
this is done by exploiting the convergence @xmath714 , which is uniform in @xmath448 and @xmath277 , and @xmath715 , which is uniform in @xmath277 .
therein , @xmath716^{-1}$ ] . from these facts
it follows that the limiting version of the truncated version satisfies the defining equations for the stieltjes kernel ( [ eqktau ] ) .
therefore the results for the complex gaussian innovation model with fixed order @xmath342 are , subject to minor modifications , still applicable when orders @xmath348 grow at a suitable rate .
the idea of the proof of theorem [ coesdmainfty ] is motivated by focusing on the @xmath245 case .
the derivation for the @xmath373 and finally @xmath15 cases follows the from the corresponding transformations and subsequent constructions analogous to the complex case .
so , let @xmath717 be the data matrix obtained from a gaussian @xmath245 time series , and suppose that @xmath270 possesses an eigendecomposition @xmath718 with @xmath719 orthogonal .
let @xmath69 denote the real fourier basis ( see , e.g. , chapter 10 of @xcite ) , and let for the real case @xmath720 .
since @xmath721 and @xmath722 have independent columns , it follows that @xmath300 has independent columns .
moreover , @xmath300 has also independent rows . to see this , note that the transpose of the @xmath28th row of @xmath300 is @xmath723 where @xmath724 is the @xmath28th column of @xmath725 and @xmath84 the @xmath28th eigenvalue of @xmath270 .
the covariance of the @xmath28th column is @xmath726 = { \mathbf{u}}^t \bigl({\mathbf{i}}+\lambda_j \bigl({\tilde{\mathbf{l}}}+{\tilde{\mathbf{l}}}^t\bigr)+\lambda_j^2{\tilde{\mathbf{l}}}{\tilde{\mathbf{l}}}^t \bigr){\mathbf{u}}.\ ] ] since @xmath727 is a symmetric circulant matrix , it diagonalizes in the real fourier basis , and hence the covariance matrix of the last display is diagonal . from the same display it follows also that the variance of the @xmath608th entry is @xmath728 , so that the rest of the proof follows as in the complex case ( theorem [ thesdmainfty ] ) . | this paper is concerned with extensions of the classical marenko pastur law to time series .
specifically , @xmath0-dimensional linear processes are considered which are built from innovation vectors with independent , identically distributed ( real- or complex - valued ) entries possessing zero mean , unit variance and finite fourth moments .
the coefficient matrices of the linear process are assumed to be simultaneously diagonalizable . in this
setting , the limiting behavior of the empirical spectral distribution of both sample covariance and symmetrized sample autocovariance matrices is determined in the high - dimensional setting @xmath1 for which dimension @xmath0 and sample size @xmath2 diverge to infinity at the same rate .
the results extend existing contributions available in the literature for the covariance case and are one of the first of their kind for the autocovariance case . , |
after the discovery of a higgs boson in july 2012 @xcite , an intense period of analysis of its properties has begun and is bearing fruits .
we now know that this object is very consistent with the spinless scalar state embedded in the sm .
following the precision measurement of its mass , around 125 gev , its couplings to all other states of the sm can be derived and compared with experimental data .
agreement between sm and experimental results is presently within a few tens of percent at worse , thus leaving some scope for a beyond the sm ( bsm ) higgs sector . by bearing in mind that the discovered higgs state has a doublet nature , amongst the many new physics scenarios available embedding such structures , those easier to deal with are clearly 2-higgs doublet models ( 2hdms ) .
in fact , these scenarios always include a neutral scalar higgs state that can play the role of the discovered one , which as intimated is very sm - like . furthermore , they are also easily compliant with past collider data ( from lep / slc and tevatron ) as well as present ones ( from the lhc ) while still offering a wealth of new higgs states and corresponding signals that can be searched for by the atlas and cms collaborations .
in fact , a significant amount of experimental effort at the lhc is presently being spared on direct searches for new higgs bosons , in parallel with the one of extracting their possible presence indirectly from the aforementioned precision measurements . however , 2hdms _ per se _ do not have the ability to solve the so - called hierarchy problem of the sm . an elegant way to do so though
, is to presume that the higgs boson discovered in 2012 and its possible 2hdm companions are not fundamental particles .
this approach is not unreasonable as any other ( pseudo)scalar state found in nature eventually revealed itself to be a ( fermion ) composite state , i.e. , a mesonic state of the now standard theory of strong interactions ( qcd ) .
specifically , one can construct 2hdms in which all higgs bosons , both neutral and charged , both scalar or pseudoscalar , are not fundamental , rather composite .
a phenomenologically viable possibility , wherein the mass of the lightest higgs state is kept naturally lighter than a new strong scale ( of compositeness , @xmath6 , in the @xmath7 tev region ) is , in particular , the one of assigning to them a pngb nature .
in essence , we have in mind those composite higgs models ( chms ) with an approximate global symmetry @xcite broken by sm interactions through the _ partial compositeness _ paradigm of refs . @xcite . in the minimal chm @xcite , the composite version of the sm higgs doublet ,
the only light scalar in the spectrum is indeed a pngb ( surrounded by various composite resonances , both spin-1/2 and spin-1 , generally heavier ) .
hence , it is natural to assume that the new ( pseudo)scalar higgs states of a c2hdm are also pngbs .
in fact , even in the case in which they are eventually found to be heavier than the sm - like higgs state , compositeness could provide a mechanism to explain their mass differences with respect to the latter .
finally , in the case of extra higgs doublets with no vacuum expectation value ( vev ) nor couplings to quark and leptons , one could also have neutral light states as possible composite dark matter candidates @xcite .
c2hdms embedding pngbs arising from a new strong dynamics at the tev scale , ultimately driving electro - weak symmetry breaking ( ewsb ) , can be constructed or by adopting an effective lagrangian description ( see example @xcite ) invariant under sm symmetries for light composite @xmath8 higgses ; or by explicitly imposing a specific symmetry breaking structure containing multiple pngbs .
we take here the second approach . in detail
, we will analyse 2hdms based on the spontaneous global symmetry breaking of an @xmath0 symmetry @xcite . within this construct , which we have tackled in a previous paper @xcite , one
can then study both the deviations of c2hdm couplings from those of a generic renormalizable e2hdm @xcite as well as pursue searches for new non - sm - like higgs signals different from the elementary case . in the @xmath9
limit the pngb states are in fact identified with the physical higgs states of doublet scalar fields of the e2hdm and deviations from the e2hdm are parametrized by @xmath10 , with @xmath11 the sm higgs vev .
once the new strong sector is integrated out , the pngb higgses , independently of their microscopic origin , are described by a non - linear @xmath12-model associated to the coset . in ref .
@xcite , we have constructed their effective low - energy lagrangian according to the prescription developed by callan , coleman , wess and zumino ( ccwz ) @xcite , which makes only few specific assumptions about the strong sector , namely , the global symmetries , their pattern of spontaneous breaking and the sources of explicit breaking ( in our case they come from the couplings of the new strong sector with the sm fields ) .
the scalar potential is in the end generated by loop effects and , at the lowest order , is mainly determined by the free parameters associated to the top sector @xcite .
however , both in ref .
@xcite and here , we will not calculate the ensuing higgs potential _ a la _ coleman - weinberg ( cw ) @xcite generated by such radiative corrections , instead
, we will assume the same general form as in the e2hdm with a @xmath2 symmetry , the latter imposed in order to avoid flavor changing neutral currents ( fcncs ) at the tree level @xcite .
we do so in order to study the phenomenology of c2hdms in a rather model independent way , as this approach in fact allows for the most general 2hdm higgs potential .
it is our intention to eventually construct the true version of the latter through the proper cw mechanism @xcite .
however , first we intend to infer guidance in approaching this task from the study of theoretical ( i.e. , perturbativity , unitarity , vacuum stability , triviality , etc .
the subject of ref .
@xcite ) and experimental ( one of the subjects of the present paper ) constraints , specifically , by highlighting the parameter space regions where differences can be found between the e2hdm and c2hdm .
this will inform the choice of how to construct a phenomenologically viable and different ( from the e2hdm ) realization of a c2hdm in terms of underlying gauge symmetries , their breaking patterns and the ensuing new bosonic and fermionic spectrum , that is , indeed , to settle on a specific model dependence .
the paper is organized as follows . in section
ii we describe the c2hdm based on @xmath13 . in section
iii , the lhc phenomenology is discussed in presence of both theoretical and experimental constraints .
conclusions are drawn in section iv . in appendix
a , relevant feynman rules for the phenomenological study are presented .
we construct the lagrangian of the c2hdm based on the spontaneous breaking of the global symmetry @xmath0 at a scale @xmath6 . in this model ,
two isospin doublet ( pseudo)scalar fields emerge as pngbs from such a breaking pattern . in our approach , we do not specify the physics at any scale above a ( large ) cutoff @xmath14 which is expected to be @xmath15 from a nave dimensional analysis @xcite , i.e. , we do not fix the concrete structure of the gauge and matter contents . even in this setup ,
the kinetic term of the pngbs is uniquely determined by the structure of the global symmetry breaking . for the yukawa sector
though , we need to assume an embedding scheme for sm fermions into @xmath1 multiplets to build the lagrangian at low energy .
although in this framework the scalar potential is generated via the cw mechanism at loop level @xcite , as intimated , we assume here its renormalisable form of the e2hdm .
this gives a sort of more generic approach to the potential , namely , once the cw potential is calculated in a fixed configuration , all the potential terms can be translated into a few parameters describing the new strong sector .
we therefore adopt the same setup of @xcite , to which we refer the reader for further details of the model construction .
we construct the @xmath16 pngb matrix @xmath17 using the eight broken generators generators given in ref .
] of @xmath1 @xmath18 ( @xmath19 and @xmath20 - 4 ) as @xmath21 the eight real spinless fields @xmath22 associated with the broken generators can be expressed through two complex doublets as @xmath23 where the @xmath24 s acquire the non - zero vacuum expectation values ( vevs ) : @xmath25 .
the ratio of the vevs is expressed as @xmath26 and we define @xmath27 .
( the relation between @xmath28 and the fermi constant @xmath29 is described in the next subsection . )
we here introduce the higgs basis in which the physical higgs states are separated from the ng boson states @xmath30 and @xmath31 which are absorbed into the longitudinal components of the @xmath32 and @xmath33 bosons , respectively , as @xmath34 where @xmath35 the doublet @xmath36 contains the physical cp - odd higgs boson @xmath37 and a pair of charged higgs bosons @xmath38 . as noted in ref .
@xcite , in the higgs basis the @xmath31 , @xmath30 and @xmath39 fields do not yield the kinetic terms in canonical form hence we shift these fields so as to render them to canonical up to @xmath40 by @xcite @xmath41 in general , the two cp - even scalar states @xmath42 and @xmath39 can mix with each other .
the mass eigenstates of the cp - even scalar states can be defined by introducing the mixing angle @xmath43 by @xmath44 where @xmath45 .
we identify the mass eigenstate @xmath3 as the higgs boson with a mass of 125 gev discovered at the lhc .
the matrix @xmath17 is transformed under @xmath1 non - linearly , i.e. , @xmath46 with @xmath47 and @xmath3 being the transformation matrices for @xmath1 and @xmath48 , respectively .
it is useful to define the linear representation of the pngb fields from @xmath17 to construct the @xmath1 invariant lagrangian . in the following ,
we use the @xmath1 adjoint representation @xmath49 , i.e. , * 15 * -plet , which is reducible under the @xmath48 subgroup as @xmath50 .
namely , @xmath51 where @xmath52 is the @xmath48 invariant vev parametrized as @xmath53 then , the field @xmath49 is transformed linearly under @xmath1 , i.e. , @xmath54 .
the kinetic terms of the eight pngb fields are determined by using the ccwz method @xcite as @xmath55 the covariant derivative @xmath56 is given in terms of the unbroken @xmath57 generators @xmath58 ( @xmath59 - 3 ) by @xmath60 where @xmath61 , @xmath62 and @xmath63 with @xmath64 being the weak mixing angle . from eq .
( [ ccwz1 ] ) , the term with two gauge bosons is extracted by @xmath65 having rewritten @xmath24 as @xmath66 following ewsb .
the masses of the weak bosons are obtained by @xmath67 where the vev @xmath68 is given by @xmath69 the kinetic terms can also be expressed by using the * 15 * -plet representation @xmath49 given in eq .
( [ sigma ] ) as follows @xmath70 , \label{kin15}\end{aligned}\ ] ] where @xmath71 .
\end{aligned}\ ] ] the expressions given in eqs .
( [ ccwz1 ] ) and ( [ kin15 ] ) are in fact equivalent . in this subsection
, we construct the low - energy ( below the scale @xmath6 ) yukawa lagrangian . in order to do this ,
we need to determine the embedding scheme of sm the fermions into @xmath1 multiplets .
this embedding can be justified via the mechanism based on the partial compositeness assumption @xcite , where elementary sm fermions mix with composite fermions in the invariant form under the sm @xmath72 gauge symmetry but not under the global @xmath1 symmetry . through the mixing ,
the @xmath1 invariant yukawa lagrangian given in terms of @xmath49 and composite fermions turns out to be the sm - like yukawa lagrangian after integrating out the ( heavy ) composite fermions .
next , we discuss fcncs , which are generally induced at tree level via neutral higgs boson mediation due to the existence of two isospin doublets as seen in general e2hdms .
we first show the generic form of the yukawa lagrangian which introduces fcncs at tree level and then we give its discrete @xmath73 symmetric version without fcncs .
we discuss the embeddings of sm quarks and leptons using * 6 * -plet representations of @xmath1 . in order to reproduce the correct electric charge of sm fermions
, we introduce an additional @xmath74 symmetry and assign its appropriate charge to * 6 * -plets .
the electric charge @xmath75 is thus given by charge for the higgs doublets @xmath76 must be zero to have a neutral component . ]
@xmath77 . in the @xmath1 basis ,
the * 6 * -plet fermion @xmath78 , with the @xmath74 charge @xmath79 expressed as a mixture of the states in the @xmath80 basis , is obtained as follows : @xmath81_x^t , \end{aligned}\ ] ] where @xmath82 , @xmath83 , @xmath84 and @xmath85 denote the ( @xmath86 ) , ( @xmath87 ) , ( @xmath88 ) and ( @xmath89 ) state for ( @xmath90 , @xmath91 ) , while @xmath92 and @xmath93 are singlets under @xmath94 and @xmath95 , respectively . from this relation
, we can embed the sm quarks and leptons into the * 6 * -plet representation @xmath78 as follows : @xmath96 note that in general the right - handed fermions are expressed as @xmath97 and that physics should not change because of the choice of the angle @xmath98 owing to the @xmath99 invariance . in the above , we simply take @xmath100 without any loss of generality in the following discussion
. the yukawa lagrangian at low energy is then given in terms of the * 15 * -plet of pngb fields @xmath49 and the * 6 * -plet of fermions defined in the previous subsection : @xmath101 + \text{h.c .
} \end{aligned}\ ] ] we note that the @xmath102 term is equivalent to the @xmath103 term , thus the terms with the cubic and more than cubic power of @xmath49 do not give any additional independent contributions to the yukawa lagrangian . the parameters @xmath104 and @xmath105 should be understood as @xmath106 complex matrices in flavor space .
this lagrangian is rewritten using the complex doublet form of the higgs fields defined in eq .
( [ cdoublet ] ) as @xmath107u_r^{}\notag\\ & + \sqrt{2}b_u\overline{q_l}\left[\tilde{\phi}_2-\frac{2}{3f^2}\left(\tilde{\phi}_1(\phi_1^\dagger \phi_2 + \text{h.c.})+2\tilde{\phi}_2(\phi_2^\dagger\phi_2 ) \right ) \right ] u_r^{}\notag\\ % % & + \sqrt{2}a_d\overline{q_l } \left[\phi_1 -\frac{1}{f^2}\phi_1\left(\frac{1}{3}\phi_1^\dagger \phi_1 + \phi_2^\dagger \phi_2 \right ) + \frac{1}{3f^2}\phi_2(\phi_1^\dagger \phi_2 + \text{h.c . } ) \right]d_r^{}\notag\\ & + \sqrt{2}b_d\overline{q_l}\left[\phi_2-\frac{2}{3f^2}\left(\phi_1(\phi_1^\dagger \phi_2 + \text{h.c.})+2\phi_2(\phi_2^\dagger\phi_2 ) \right ) \right ] d_r^{}\notag\\ % % & + \sqrt{2}a_e\overline{l_l } \left[\phi_1 -\frac{1}{f^2}\phi_1\left(\frac{1}{3}\phi_1^\dagger \phi_1 + \phi_2^\dagger \phi_2 \right ) + \frac{1}{3f^2}\phi_2(\phi_1^\dagger \phi_2 + \text{h.c . } ) \right]e_r^{}\notag\\ & + \sqrt{2}b_e\overline{l_l}\left[\phi_2-\frac{2}{3f^2}\left(\phi_1(\phi_1^\dagger \phi_2 + \text{h.c.})+2\phi_2(\phi_2^\dagger\phi_2 ) \right ) \right ] e_r^ { } , \label{gen_yuk}\end{aligned}\ ] ] where @xmath108 .
the masses of fermions are extracted as ( we work at the first order in @xmath109 : @xmath110 . \end{aligned}\ ] ] the interaction terms for @xmath3 , @xmath111 , @xmath112 and @xmath113 are expressed by @xmath114 where @xmath115 for @xmath116 and @xmath117c_\theta + \left[-a_fs_\beta + b_fc_\beta\left(1 -\frac{\xi}{2}\right)\right]s_\theta , \\
x_f^h & = -\left[a_fc_\beta\left(1 - \frac{\xi}{2 } \right)+ b_fs_\beta\left(1 -2\xi\right)\right]s_\theta + \left[-a_fs_\beta + b_fc_\beta\left(1 -\frac{\xi}{2}\right)\right]c_\theta , \\
x_f^a & = -a_fs_\beta\left(1 - \frac{\xi}{2 } \right ) + b_fc_\beta ~~(f = u , d , e ) .
\end{aligned}\ ] ] again , notice that the above @xmath118 ( @xmath119 ) couplings are @xmath106 matrices in flavor space . in general , these matrices are non - diagonal in the mass eigenbasis of fermions , so that there appear fcncs at tree level because of the existence of two independent @xmath106 yukawa couplings @xmath104 and @xmath105 .
in other words , the fcnc problem is induced by the fact that both doublets @xmath120 and @xmath121 couple to each fermion types , i.e. , up - type quarks , down - type quarks and charged leptons .
this property is kept even if we take the e2hdm limit @xmath9 .
therefore , our yukawa lagrangian given in eq .
( [ gen_yuk ] ) reproduces the so - called type - iii e2hdm . in the next subsection
, we discuss the case without tree level fcncs .
.charge assignment for right - handed fermions under the @xmath73 symmetry in the c2hdm .
all the left - handed fermions @xmath122 , @xmath123 and @xmath124 are transformed as even under @xmath73 . in the third column ,
the symbol @xmath125 means non - zero @xmath104 or @xmath105 . [
cols="^,^,^,^,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] we present the trilinear couplings of the higgs bosons which are relevant to the discussion of the phenomenology given in sec .
first , the gauge - gauge - scalar type interactions are given by @xmath126 second , the coefficients of the scalar - scalar - gauge type interactions are extracted as given in table [ fr1 ] , where we introduce @xmath127 finally , the scalar trilinear @xmath128 and @xmath129 couplings defined by @xmath130 are extracted by @xmath131 \notag\\ & + \frac{\xi}{12v_{\text{sm}}}s_\theta \left[m_h^2 - 2m_h^2 + ( 1 + 3c_{2\theta}+6\cot 2\beta s_{2\theta})m^2\right ] , \\ \lambda_{h^+h^-h } & = \frac{c_\theta}{v_{\text{sm}}}\left[-(1 + 2\cot2\beta\tan\theta)m_h^2 - 2m_{h^\pm}^2+\frac{2}{s_{2\beta}c_\theta}s_{2\beta + \theta } m^2\right]\notag\\ & + \frac{\xi}{6v_{\text{sm}}}c_{\theta}\left[(1 + 4\cot2\beta\tan\theta)m_h^2 + 2(m_{h^\pm}^2-m^2 ) \right ] .
\end{aligned}\ ] ] r. barbieri , l. j. hall and v. s. rychkov , phys . rev .
d * 74 * , 015007 ( 2006 ) [ hep - ph/0603188 ] ; + m. frigerio , a. pomarol , f. riva and a. urbano , jhep * 1207 * , 015 ( 2012 ) [ hep - ph/1204.2808 ] ; + n. fonseca , r.z .
funchal , a. lessa and l. lopez - honorez , jhep * 1506 * , 154 ( 2015 ) [ hep - ph/1501.05957 ] ; + a. carmona and m. chala , jhep * 1506 * , 105 ( 2015 ) [ hep - ph/1504.00332 ] .
v. d. barger , j. l. hewett and r. j. n. phillips , phys .
d * 41 * , 3421 ( 1990 ) .
y. grossman , nucl .
b * 426 * , 355 ( 1994 ) .
m. aoki , s. kanemura , k. tsumura and k. yagyu , phys .
d * 80 * , 015017 ( 2009 ) . | we investigate the phenomenology of composite 2-higgs doublet models ( c2hdms ) of various yukawa types based on the global symmetry breaking @xmath0 .
the kinetic term and the yukawa lagrangian are constructed in terms of the pseudo nambu - goldstone boson ( pngb ) matrix and a * 6 * -plet of fermions under @xmath1 .
the scalar potential is assumed to be the same as that of the elementary 2-higgs doublet model ( e2hdm ) with a softly - broken discrete @xmath2 symmetry .
we then discuss the phenomenological differences between the e2hdm and c2hdm by focusing on the deviations from standard model ( sm ) couplings of the discovered higgs state ( @xmath3 ) as well as on the production cross sections and branching ratios ( brs ) at the large hadron collider ( lhc ) of extra higgs bosons .
we find that , even if the same deviation in the @xmath4 ( @xmath5 ) coupling is assumed in both scenarios , there appear significant differences between e2hdm and c2hdm from the structure of the yukawa couplings , so that production and decay features of extra higgs bosons can be used to distinguish between the two scenarios . |
mode coupling and conversion are commonly known and primarily important phenomena in fiber and integrated optics , either hindering or fueling the operation of numerous devices and elements , e.g. , electro - optical and acousto - optical switches and modulators , waveguide couples , power splitters , wavelength filters , and others @xcite . when a mode of certain configuration transverses the structure and converts into some other mode , say , due to the externally induced perturbation in optical properties or geometrical configuration of the structure or due to the imperfectness of the materials used , or when the exchange of energy between the guided modes of adjacent waveguides occurs anyway we face the phenomena of mode coupling and conversion .
the crucial regime for the mode coupling is in the vicinity of the points of virtual phase matching ( otherwise referred to as `` anticrossings '' or `` avoided crossings '' a terminology brought about from quantum mechanics and solid - state physics ) . at anticrossings ,
the complete mode conversion can be achieved , which makes tailoring this regime particularly important for the applications . for a positive example , let me mention a promising method for controllable dispersion compensation in photonic bandgap fibers based on weak interactions at anticrossings between the core - guided mode and a mode localized in an intentionally introduced defect of the crystal @xcite .
an opposing example of technology strains due to the undesirable intermodal interactions is radiation losses caused by the coupling of the guided mode to radiation modes in a two - dimensional photonic crystal etched into a planar waveguide @xcite .
a dominating theoretical tool for the whole of mode coupling and conversion phenomena ( except a few cases when rigorous analytical treatment is possible @xcite ) is the coupled - mode , or the coupled - wave , formalism @xcite that was first proposed more than fifty years ago and since then became commonplace in both optical engineering research and textbook literature .
the mathematical foundation of all the modifications of the coupled - mode formalism is the method of variation of independent coefficients one of the fruitful methods in perturbation theory @xcite
. nevertheless , the formalism grounds upon the differing amount of approximations e.g. , that of slowly - varying dielectric function @xmath1 , so that the electric field @xmath2 is assumed to satisfy @xmath3 ; the so - called parabolic approximation ; and others which certainly limits the usefulness of the model .
moreover , the resultant `` coupled modes '' often fail to satisfy orthogonality relations and boundary conditions in the actual ( perturbed ) structure .
finally , the coupled - mode formalism ignores all of the explicit mathematical parallels to the classical quantum - mechanics problems that also exploit perturbation techniques , in particular the widely known perturbation method for stationary levels @xcite a modification of the lindstedt - poincar method of strained parameters sometimes called the rayleigh - schrdinger method ( see refs .
@xcite for the bibliographical details ) .
meanwhile , establishing and tracking the analogies between classical quantum mechanics and optics tends to be a really stimulating approach both through adopting the formalism of electromagnetic - wave propagation in dielectric materials to collisionless propagation of effective - mass electron waves in semiconductor crystals @xcite , and on the other hand through applying the formalism of quantum - mechanical solid - state electronics to electromagnetic propagation in periodic dielectric structures which proved to be a very successful programme during the last decade @xcite . in this article ,
i translate the classical stationary perturbation technique i.e. , the method of strained parameters to the case of mode coupling and conversion in optical waveguiding systems .
the problem of source - free light propagation is formulated in as an ordinary eigenvalue problem in the squared free - space wavenumber @xmath4 ( or , actually , in the angular frequency @xmath5 ) for the magnetic field @xmath6 , and perturbations in dielectric index profile of the waveguide are considered , embracing the cases of optical inhomogeneity , magnetization of arbitrary orientation , and nonlinearity .
both the eigenvalues and the eigenvectors are expanded into the series in perturbation parameter , giving rise to an electromagnetic counterpart of the quantum - mechanics stationary perturbation method as an alternative to the coupled - mode formalism .
although physical interpretation of the former in the case of nondegenerate spectrum of eigenvalues looks rather questionable from the point of experiment , the method however provides promptly solvable and naturally interpretable treatment of the modal behavior at the anticrossings , which is an issue in the mode conversion analysis . to demonstrate the efficiency of the method ( to distinguish from the coupled - mode theory , the qualifying term `` stationary '' is retained here with the primary meaning that the coefficients of an expansion into which an input wave is decomposed
are not varying with coordinates ) , i refer in to the two practice - targeted examples from integrated optics concerning @xmath7 mode coupling : a multilayer gyrotropic waveguide subject to constant , arbitrarily oriented magnetic field , and an optically - inhomogeneous planar dielectric waveguide an example which falls out of the scope of the standard coulped - mode theory .
the unperturbed basis for the both cases is a multilayer waveguide composed of linear isotropic dielectric materials a simple structure known to exhibit under certain circumstances the perfect phase matching of guided te and tm modes @xcite , thus it is natural to implement here the developed theory for the mode conversion at anticrossings . for both systems ,
i pursue the formulated way to derive closed - form first - order analytic expressions for the minimum value of mode mismatch and hence for the length of complete @xmath0 mode conversion ( the coupling length , or the beat length ) .
finally , i conclude in with some claims concerning further possible applications and modifications of the stationary perturbation technique for the optical waveguiding theory .
let me start with the wave equation for the magnetic field @xmath8 where @xmath9 is the dielectric permittivity distribution , @xmath10 the squared free - space wave number , @xmath5 the angular frequency , @xmath11 the vacuum speed of light , and the field @xmath12 satisfies additionally @xmath13 eq . ( [ h - eigenproblem ] ) can be treated as an ordinary eigenvalue problem in @xmath14 ( or , actually , in @xmath5 ) for the field @xmath15 , with maxwellian operator @xmath16 defined by @xmath17 in fact , a similar equation can be written for , say , the @xmath18 field : @xmath19 but there are two sound reasons to restrict oneself to the @xmath15-eigenproblem ( [ h - eigenproblem ] ) , but not to its @xmath20-counterpart ( [ e - eigenproblem ] ) : first , an accompanying to eq .
( [ h - eigenproblem ] ) divergence equation ( [ divh ] ) is clearly simpler than @xmath21 a satellite equation for the @xmath18-eigenproblem ; second , given the dielectric index a real scalar function of coordinates i.e. , a real scalar field or , more generally , a hermitian dyadic field , the maxwellian ( [ maxwellian ] ) becomes hermitian too @xcite the fact that is though not crucial in macroscopic electrodynamics ( non - hermitian maxwellian would generete a set of nonorthogonal eigenmodes and complex - valued eigenfrequencies a situation normally faced in optics of lossy qyrotropic media @xcite ) nor pertinent to the present treatment , but aesthetically pleasant and legitimates many of the cross - references between eq .
( [ h - eigenproblem ] ) and the stationary schrdinger eigenproblem with hermitian hamiltonian as well . for a large class of waveguiding systems ,
namely , for those exhibiting continuous translation symmetry along the direction of light propagation ( say , the @xmath22 direction ) we can assign harmonic dependence of the fields on that direction ; in particular , the magnetic field reads @xmath23 where @xmath24 is the propagation constant , so that for a given @xmath22-independent dielectric index profile @xmath25 eq . ( [ h - eigenproblem ] ) reduces to an eigenvalue problem for @xmath26 ; the reduced eigenproblem operator depends parametrically on @xmath24 then and at a fixed @xmath24 spawns a set of eigenvalues @xmath27 , thus yielding dispersion structure @xmath28 of a perfect waveguide . if an actual dielectric permittivity distribution @xmath29 of the waveguide happens to differ from @xmath25 for which eq .
( [ expbeta ] ) holds , that is @xmath30 ( we do not need to specify at this stage whether @xmath31 is a scalar or a tensor field ) and @xmath32 where @xmath33^i/[\epsilon_{(0)}(x , y)]^{i+1}$ ] , @xmath34 , then in the spirit of the lindstedt - poincar perturbation method of strained parameters ( * ? ? ?
3 ) one can rewrite eq .
( [ h - eigenproblem ] ) as @xmath35 in terms of the unperturbed maxwellian @xmath36 and the higher - order perturbation operators , and expand then the eigenvectors @xmath37 of the perturbed problem into a series in @xmath38 : @xmath39 and similarly for the eigenvalues : @xmath40 the procedure for further solving eq .
( [ h - eigenproblem - perturbed ] ) using expansions ( [ h - expansion ] ) and ( [ k - expansion ] ) , i.e. , for finding the unknown functions @xmath41 and @xmath42 , is quite well - developed in perturbation theory . substituting eqs .
( [ h - expansion ] ) , ( [ k - expansion ] ) into eq .
( [ h - eigenproblem - perturbed ] ) yields for the zeroth - order and linear in @xmath38 terms : @xmath43 with the homogeneous boundary conditions @xmath44 .
the solution to the unperturbed problem ( [ zeroth ] ) is assumed to be known ; to say more and this is an important point i further assume all the @xmath45 vectors to be of the form ( [ expbeta ] ) which is likely to embrace all the cases of practical interest .
the subsequent steps of the method depend essentially on whether the spectrum of @xmath36 is degenerate or not .
if there are no degenerate eigenvalues among those of @xmath36 spectrum , then we customarily obtain up to the first - order correcting terms @xcite : @xmath46 and @xmath47 where the first - order coupling matrix @xmath48 the integration volume @xmath49 embracing the waveguide is formally infinite , @xmath50 , the normalizing factor @xmath51 equals either @xmath52 the distance of light propagation within a waveguide in the @xmath22 direction ( i.e. , the length of the waveguide ) in the case of cylindrical ( two - dimensional ) geometry , or @xmath53 the @xmath54 square of a waveguiding cell for planar ( one - dimensional ) systems .
this normalization is due to specific orthogonality condition assigned to the @xmath55 modes : @xmath56 i explicitly designated the ( parametric ) dependence of all the quantities in eqs .
( [ h - expansion - nondeg ] ) , ( [ k - expansion - nondeg ] ) , and ( [ m1 ] ) on @xmath24 entailed by the previous assumption that the unperturbed modes have the form as per eq .
( [ expbeta ] ) .
now , the restriction @xmath57 for the set of nondegenerate modes @xmath58 into which the _ _ n__th perturbed eigenmode @xmath37 is expanded via eq .
( [ h - expansion - nondeg ] ) modes characterized thus by essentially different eigenwavenumbers @xmath59 and differing normalized propagation constants @xmath60 seems to correspond to an utterly odd situation from the point of experiment ( see ) . to aid this shortcoming somehow
, one might consider eq .
( [ h - eigenproblem ] ) as a generalized eigenproblem in , e.g. , @xmath24 or @xmath61 , parametrically dependent on @xmath14 , which would lead to expansions like ( [ h - expansion - nondeg ] ) but into the @xmath62 set .
of course , this severely complicates the formalism and probably here is an issue why the coupled - mode theory has been exclusively dominating for decades in the field .
fortunately , the case of quasi - degenerate eigenvalues allows promptly solvable and easily interpretable treatment in terms of the ordinary @xmath12-eigenproblem in @xmath14 .
lines ( ) for the case of ( a ) nondegenerate spectrum of @xmath36 and ( b ) the two modes , @xmath63 and @xmath64 , exhibiting degeneration . here
@xmath65 is a characteristic dimension of the structure ; the grey circles mark the modes related to the expansions ( [ h - expansion - nondeg ] ) and ( [ h - expansion - deg ] ) . ]
let us focus on the case of two - degenerate or nearly degenerate eigenmodes @xmath66 and @xmath67 .
borrowing the known result from quantum mechanics @xcite , we see that in this case both the @xmath66 and @xmath67 modes dominate the expansion ( [ h - expansion ] ) of the related fundamental modes @xmath68 with the constant ( in particular , @xmath22-independent ) coefficients @xmath69^{1\over 2},\ ] ] @xmath70^{1\over 2}.\ ] ] here and below in the article i pursue only the first - order approximations and hence omit the overly superscript @xmath71 over the coupling matrix elements @xmath72 , now @xmath24-independent , defined by eq .
( [ m1 ] ) .
the first - order correction to the corresponding eigenvalues leads to @xcite @xmath73 ^ 2 + \varepsilon^2\mathcal{m}_{12}\mathcal{m}_{21}}.\ ] ] in quantum mechanics , this situation corresponds to quasi - degenerate energy levels of a quantum system and is associated with enormous variety of effects and treatments take the landau - zener model for example @xcite , nourishing nearly 100 publications a year . in the realm of optics , eq .
( [ h - expansion - deg ] ) with coefficients given by eqs .
( [ c1 ] ) , ( [ c2 ] ) is fully consistent with the general statement that complete mode conversion can be achieved in any system exhibiting anticrossings .
indeed , instead of the @xmath74 mode and the @xmath67 mode crossing one another on the dispersion structure diagram ( ) , at the anticrossings the @xmath74 mode `` transforms '' continuously into the @xmath75 one , and vice - versa , over a certain frequency range , so it is intuitive to expect that somewhere in the vicinity of the virtual crossing point the @xmath66 and @xmath75 modes would be equally unsuitable to play the role of the fundamental modes @xmath76 and @xmath77 , and if the original ( input ) wave is @xmath66 or @xmath75 , it would nt `` feel '' which of the fundamental modes , @xmath76 or @xmath77 , is more suited for it , and thus would oscillate between @xmath66 and @xmath75 .
mathematically , these heuristic speculations are expressed via the requirement @xmath78 ( under appropriate normalization ) necessary for the complete mode conversion to occur , which is equivalent , as immediately follows from the comparison of eqs .
( [ c1 ] ) and ( [ c2 ] ) , to canceling out the @xmath79 and @xmath80 elements under the root in eqs .
( [ c1 ] ) , ( [ c2 ] ) , and ( [ k - deg ] ) due to @xmath81 this condition does not necessarily correspond to the case of stringently degenerate eigenvalues @xmath82 and @xmath83 , but with a good deal of reason we may assume @xmath84 to estimate the minimum mode mismatch @xmath85 from eq .
( [ k - deg ] ) . with this assumption , it is under condition ( [ m11=m22 ] ) that the quantity @xmath86 where @xmath87 , @xmath88 is the frequency of a virtual mode crossing ( phase matching ) , reaches its minimum value of @xmath89 , thus the minimum eigenmode mismatch @xmath85 along the @xmath90 curve corresponds to the complete mode conversion and reads @xmath91 in a nonabsorbing waveguide ( in sec .
[ sec3 ] we will deal with the two ) @xmath72 is a hermitian matrix , hence @xmath92 and eq .
( [ minmodemismatch - gen ] ) is simplified to @xmath93 complete mode conversion occurs when the @xmath66 and @xmath67 modes propagating in the @xmath22 direction accumulate a phase difference of @xmath94 : @xmath95 ; substituting here eq .
( [ minmodemismatch ] ) , we obtain for the corresponding coupling length : @xmath96 where , as before [ see eq .
( [ m1 ] ) ] , the coupling matrix element @xmath97\cdot\mathbf h^{*(0)}_2 dv.\ ] ] in contrast to the coupled - mode theory , the @xmath98 element might and , in general , does depend on the length of a waveguide along the @xmath22 direction because of the possible @xmath22 dependence of the @xmath99 function . in what follows
i calculate @xmath98 and hence @xmath100 explicitly for the two specific systems relevant to integrated optics : for a multilayer gyrotropic stack and for an optically - inhomogeneous waveguide .
let the unperturbed waveguide be described by the scalar dielectric permittivity distribution @xmath101 . as argued by symmetry reasoning , relativity considerations , and energy conservation @xcite , the most general form to allow for the electromagnetic disturbance induced by the @xmath102 field applied to ( initially ) isotropic medium
is @xmath103 where @xmath104 is a unit three - dimensional dyadic ; @xmath105 and @xmath106 are real functions of @xmath107 ; the superscript @xmath108 over the @xmath109 vector denotes the antisymmetric dyadic dual to @xmath109 , so that @xmath110 gives conventional vector product of @xmath109 and @xmath111 ; the circled cross @xmath112 denotes the outer product the dyad : @xmath113 , @xmath114 .
if @xmath38 is small , we may assume @xmath105 and @xmath106 to be constant ( this assumption leads to the introduction of gyration vector @xcite ) and write down for the inverse to the @xmath115 dyadic @xcite @xmath116 in terms of @xmath117 , so that the coupling matrix element ( [ m12 ] ) can be written as @xmath118\cdot\mathbf h^{*(0)}_2 dv,\ ] ] where i used the frequency - domain maxwell s equation for @xmath119 in the source - free region , @xmath120 .
equation ( [ m12gyr ] ) can be significantly simplified by applying the vector identity @xcite @xmath121 valid for any three - vectors @xmath122 and @xmath123 .
the first term on the right vanishes immediately due to @xmath124 ; then , in a specific but physically reasonable and normally considered case of @xmath122 being a constant ( or piecewise - constant ) vector in a given volume @xmath125 and equaling zero outside of it , the second and third terms of the sum in eq .
( [ vectoridentity ] ) being multiplied by @xmath15 and integrated as per eq .
( [ m12gyr ] ) lead to the surface integrals instead of the volume ones . for the second term one obtains : @xmath126 , where @xmath127 defines the boundary @xmath128 of @xmath125 , @xmath129 is the _ outer _
normal to @xmath128 ; hence @xmath130\cdot\mathbf h \ , dv = - \int_\sigma [ \mathbf d(\mathbf{\hat n}\cdot\mathbf u)]\cdot\mathbf h \ , d\sigma.\ ] ] for the third term we similarly have : @xmath131 , hence for an integral @xmath132 \cdot\mathbf h \,dv = -\int_\sigma [ ( \mathbf d \cdot \mathbf{\hat n})\mathbf u ] \cdot\mathbf h \ , d\sigma.\ ] ] thus for the case of constantly magnetized waveguide ( up to this point , no assumptions have been made regarding its actual geometry ) the integral in eq .
( [ m12gyr ] ) is reduced via eqs .
( [ vectoridentity ] ) , ( [ secondterm ] ) , and ( [ thirdterm ] ) to @xmath133\cdot\mathbf h \ , dv = \int_\sigma [ \mathbf d(\mathbf{\hat n}\cdot\mathbf u)]\cdot\mathbf h \ , d\sigma - \int_\sigma [ ( \mathbf d \cdot \mathbf{\hat n})\mathbf u ] \cdot\mathbf h \ ,
d\sigma - \int_v [ ( \mathbf u \cdot \bm\nabla)\mathbf d ] \cdot\mathbf h \,dv.\ ] ] now let us switch to the planar structures . in a planar unperturbed
waveguide the eigenmodes are classified in terms of te , or _
s_-polarized , and tm , or _
p_-polarized waves ( for an elegant derivation of this common fact via symmetry considerations see ref .
@xcite ) . in a coordinate system
in which the @xmath134 axes is normal to the bimedium interfaces and the light energy is guided in the @xmath22 direction , we can write down , to evaluate the @xmath135 element as per eq .
( [ m12gyr ] ) : @xmath136 for the @xmath137-polarized mode , and @xmath138 for the @xmath139-polarized one . here
@xmath140 and @xmath141 are the lateral distributions of @xmath142 and @xmath143 respectively , obtained from the unperturbed maxwell s equations for the given virtually - phase - matched modes .
we see immediately from eq .
( [ ds ] ) that the second surface integral in eq .
( [ threeintegrals ] ) vanishes at planar geometry , since @xmath144 holds identically in this case .
finally we arrive at @xmath145\cdot\mathbf h^{*}_p \ , d\sigma + ks^{-1}\!\!\int_v [ ( \mathbf u \cdot \bm\nabla)\mathbf d_s]\cdot\mathbf h^{*}_p \,dv \nonumber\\ & = & - ks^{-1}\!\!\int_v u_x\left(\frac{d\phi_s}{dx}\psi_p + \phi_s\frac{d\psi_p}{dx}\right)\,dv + ks^{-1}\!\!\int_v \left(u_x\frac{d\phi_s}{dx}\psi_p + i\beta u_z\phi_s\psi_p\right)\,dv
\nonumber\\ & = & -k\left[u_x^\alpha j^\alpha\left(\frac{\phi_s\,d\psi_p}{dx}\right ) - iu_z^\alpha j^\alpha\left(\frac{\phi_s\psi_p}{\beta^{-1}}\right ) \right],\end{aligned}\ ] ] where @xmath146 is the value of vector @xmath122 in the @xmath147th layer bounded between the @xmath148 and @xmath149 planes , and the integrals @xmath150 @xmath151 ( @xmath152 is the number of layers ) .
( [ m12gyr - planar ] ) accounts for no mode conversion at transversal magnetization , in agreement with the long ago established result @xcite .
a few more implicit tips could also be deduced thereof : \1 . in a symmetrically sandwiched stack , the _ polar _ magnetization @xmath153 virtually does nt couple the modes exhibiting similar symmetry in the @xmath154 and @xmath155 functions ( that is , when both @xmath154 and @xmath155 are either even or odd ) ; on the contrary , the modes of opposite symmetry in @xmath154 and @xmath155 are virtually not sensitive to the _ longitudinal _ magnetization @xmath156 .
asymmetric sandwiching impairs this behavior .
the integral ( [ jbeta-1 ] ) contains scaling parameter @xmath157 that defines characteristic thickness of the waveguide corresponding to the comparable values of the both integrals , eqs .
( [ jdx ] ) and ( [ jbeta-1 ] ) , in eq .
( [ m12gyr - planar ] ) , and hence to the comparable shares of polar and longitudinal polarizations in the mode conversion efficiency . for optical frequencies @xmath158 nm
; if the thickness of the waveguide considerably exceeds @xmath159 , than @xmath155would appear to be a too slowly varying function and therefore @xmath160 for the ultrathin layers the inverse inequality holds , but in this regime at most one mode is guided in the structure , which is apparently out of the scope here .
if we assume @xmath161 in eq .
( [ m12 ] ) to be a scalar function within the waveguide volume @xmath125 , which conveys the cases of optical inhomogeneity and `` isotropic '' nonlinearity , then it is advantageous to simplify eq .
( [ m12 ] ) using the identity @xmath162 where @xmath163 is taken into account . in planar geometry , we are concerned as before with the te and tm modes , but now expressed exclusively through the magnetic field vectors . for the tm ( @xmath139-polarized )
mode we can readily apply eq . (
[ hp ] ) ; for the te ( @xmath137-polarized ) mode we have for the electric field , in parallel with eq .
( [ ds ] ) , @xmath164 hence @xmath165 where we once again encounter @xmath24 ( in fact , @xmath157 ) in the role of scaling parameter .
if we put , say , @xmath166 in eq .
( [ vectoridentity-2 ] ) and multiply the result by @xmath167 , then only the first and second terms of the sum ( [ vectoridentity-2 ] ) will produce nonzero results , as it immediately follows from eqs .
( [ hs ] ) and ( [ hp ] )
. equation ( [ m12 ] ) thus gives @xmath168\cdot\mathbf h_p^ { * } \ , dv - s^{-1}\!\!\int_w [ ( \mathbf h_s\cdot\bm\nabla)\bm\nabla\xi ] \cdot\mathbf h_p^ { * } \ , dv.\ ] ] now care should be taken when dealing with @xmath169 function which is discontinuous in the vicinity of material boundaries [ otherwise if @xmath169 is assumed to be all - continuous eq .
( [ m12-diel ] ) gives @xmath170 .
the safest way is to introduce a new vector field @xmath171 instead of @xmath169 , with a requirement @xmath172 everywhere except at the boundaries : @xmath173 and with this in mind to perform the brute - force evaluation of integrals in eq .
( [ m12-diel ] ) : @xmath174\cdot\mathbf h_p^ { * } \ , dv = \beta k^{-1}\!\ !
\int_w\frac{\partial\varsigma_x}{\partial y}\varphi_s(x)\psi_p(x ) \ , dv - ik^{-1}\!\!\int_w\frac{\partial\varsigma_z}{\partial y}\frac{d\varphi_s(x)}{dx}\psi_p(x ) \ , dv , \\ \int_w [ ( \mathbf h_s\cdot\bm\nabla)\bm\nabla\xi ] \cdot\mathbf h_p^ { * } \ , dv = \beta k^{-1}\!\ !
\int_w\frac{\partial\varsigma_y}{\partial x}\varphi_s(x)\psi_p(x ) \ , dv - ik^{-1}\!\!\int_w\frac{\partial\varsigma_y}{\partial z}\frac{d\varphi_s(x)}{dx}\psi_p(x ) \ , dv,\end{aligned}\ ] ] hence @xmath175 we see that for the nonzero mode conversion , the @xmath176 function should exhibit explicit dependence on the @xmath177 coordinate .
for this reason @xmath178 would depend on the transversal dimension @xmath179 of the waveguide and on the values of @xmath176 at the @xmath180 and @xmath181 boundaries .
another important consequence relates the effect of the @xmath22 modulation of @xmath176 on the coupling length .
if @xmath176 does not depend on @xmath22 , then @xmath178 would in turn be independent on the waveguide length @xmath52 ; but on the contrary , whenever @xmath176 is a stochastically oscillating or periodic function of @xmath22 , the coupling matrix @xmath178 and hence the coupling length @xmath100 would become @xmath52-dependent .
say , for the @xmath182 dependence we have @xmath183 which is of the order of @xmath184 and brings about an unexpected scaling rule @xmath185 for the coupling length in a planar waveguide with periodic modulation of @xmath186 in the @xmath22 direction .
finally , i would like to remind here that conventional expression for the off - diagonal coupling matrix element given by the standard coupled - mode theory @xcite , @xmath187 gives an identical zero for the @xmath188 mode conversion , being the dielectric perturbation @xmath189 a scalar function .
mode conversion displays breaking the ( initial ) symmetry of the hamiltonian of a system by some perturbation that distorts the mode spectrum , i.e. , shifts the eigenvalues and alters polarization of the eigenmodes , hence a natural tool to treat the mode coupling and conversion phenomena is the perturbation technique . for the mode conversion at anticrossings , i presented in this article an electromagnetic counterpart of quantum - mechanical perturbation theory for quasi - degenerate levels based on the lindstedt - poincar method of strained parameters as a specific alternative to the entrenched coupled - mode formalism grounded upon the method of variation of independent coefficients .
the general expressions derived for the minimum mode mismatch at anticrossings , eq .
( [ minmodemismatch ] ) , and for the coupling length , eq .
( [ couplinglength ] ) , are compact , transparent and premise on the calculation of just one element of the coupling matrix , eq .
( [ m12 ] ) . that matrix element was calculated explicitly for the two cases of interest in integrated optics : for a multilayer gyrotropic waveguide under piecewise - constant , arbitrarily oriented magnetization [ eq . ( [ m12gyr - planar ] ) ] , and for an optically - inhomogeneous planar dielectric waveguide [ eq . ( [ m12diel - planar ] ) ] . in a similar way a large variety of optical systems can be analyzed ,
including bragg s fibers , photonic crystals with broken periodicity , etc . finally , i should note that by means of an appropriate coordinate mapping , the problem of perturbation due to shifted material boundaries so to say , geometrical perturbation known to spur difficulties when treated via conventional perturbation techniques @xcite can be reduced to the problem of perturbation in permittivity and permeability profiles of a waveguide exhibiting perfect geometry in those ( curvilinear ) coordinates and governed by `` cartesian - looking '' maxwell s equations .
such a trick that recently aided the fdtd modeling of high index - contrast photonic crystals @xcite would also significantly augment both the standard coupled - mode theory and the stationary perturbation technique described here . | intermodal interactions displayed through the phenomena of mode coupling and conversion in optical systems are treated by means of the lindstedt - poincar perturbation method of strained parameters more widely known in classical quantum mechanics and quantum chemistry as the stationary perturbation technique .
the focus here is on the mode conversion at the points of virtual phase matching ( otherwise called anticrossings or avoided crossings ) associated with the maximum conversion efficiency .
the method is shown to provide a convenient tool to deal with intermodal interactions at anticrossings interactions induced by any kind of perturbation in dielectric index profile of the waveguide , embracing optical inhomogeneity , magnetization of arbitrary orientation , and nonlinearity .
closed - form analytic expressions are derived for the minimum value of mode mismatch and for the length of complete mode conversion ( the coupling length , or the beat length ) in generic waveguiding systems exhibiting anticrossings .
demonstrating the effectiveness of the method , these general expressions are further applied to the case of @xmath0 mode conversion in ( i ) a multilayer gyrotropic waveguide under piecewise - constant , arbitrarily oriented magnetization , and ( ii ) an optically - inhomogeneous planar dielectric waveguide an example which the standard coupled - mode theory fails to describe . |
since the pioneering work of e. yablonovitch @xcite and s. john @xcite , research on photonic crystals(pcs ) has enjoyed a nearly exponential increase . the manufacture of pcs at the optical regime has become a reality @xcite .
manufacturing brings with it the practical reality of random errors introduced during the manufacturing process and it is the effect of these random errors on the desirable features of pcs , namely photonic band gaps , that we wish to address in this paper .
bandgaps in pcs depend on two crucial properties : an infinite _ and _ perfect translational symmetry .
clearly , in real life no crystal is infinite in size or perfectly periodic .
when randomness is introduced in the geometry of the pc , one quantity of interest is the size of the bandgaps as the level of randomness is increased , and whether the bandgaps of the bulk perfect pc will survive the randomness .
same considerations apply for a _ finite _ pc .
in fact , even for a perfect but finite pc , one needs to give up the notion of a bandgap and has to be content with severe depressions in transmittance instead . in this paper
, we will consider both finite imperfect pcs by examining the dependence of their transmittance on randomness , and bulk imperfect pcs by determining their density of states ( dos ) under varying degrees of randomness , using the supercell method .
although much has been done @xcite regarding imperfect two- and three - dimensional pcs , we feel that a study of the problem for one - dimensional pcs is warranted because of the inherent simplicity of the geometry and because a variety of extremely accurate mathematical tools are readily available which allow a detailed study of the problem without having to compromise accuracy .
for instance , because the electric field and its first derivative are continuous across the interface , and because of the low dimensionality of the pc , the convergence problem that plagued band structure calculations for many 3d pcs @xcite is essentially non - existent for 1d structures .
thus , we were able to use the old trusted plane wave ( pw ) method to find the band structure and the dos for supercell sizes not even imaginable in three- or even two - dimensional supercell calculations @xcite .
one can obtain better than 0.1% convergence with as few as @xmath130 plane waves per unit cell in the supercell .
the transmission coefficient , too , can be calculated for nearly arbitrary supercell sizes .
finally , one can calculate the band structure and the imaginary part of the wave vector using a semi - analytical approach for very large supercells .
the 1d pc is , in many ways , the infinite square - well " problem of photonic crystals .
it contains the essential features of its bigger cousins in two and three dimensions without the mathematical complexities and the accompanying numerical uncertainties @xcite that can sometimes overshadow the essentials .
for example , with 3d face centered cubic structures , it becomes practically impossible , due to convergence problems , to increase the supercell size beyond @xmath2 ( or at most @xmath3 ) conventional cubic unit cells which contain only 32 primitive cells per supercell ( or 108 ) , since typically at least @xmath4 terms per primitive cell are necessary to ensure sufficient convergence for inverse opal structures .
it s not obvious from the start whether a randomness analysis with such small supercell sizes would yield results that are physically meaningful .
artifacts due to the small supercell size are bound to be inextricably intertwined with the physically significant bulk features of the imperfect pc .
it s important to realize that , with the supercell method one still calculates the bands of an infinite _ perfect _ pc .
the randomness is only _ within _ the supercell , but on a global scale , it s still a perfectly periodic structure ! in order to be able to resolve the supercell artifacts from the physical features brought about by randomness , one needs to ensure that the interaction between neighboring supercells , which , to a good approximation , is proportional to the surface area of the supercell , be small compared to the bulk properties of the imperfect pc , which can be taken to be proportional to the volume of the supercell .
hence , on purely dimensional grounds , one can argue that the surface to volume ratio of the supercell @xmath5 , where @xmath6 is the linear size of the supercell , should be small compared to the typical length scale of the problem , namely the wavelength of the bandgap .
we allowed the supercell size @xmath7 to vary from @xmath8 to @xmath9 , and one can clearly see the supercell artifacts gradually diminishing while the bulk features become more prominent in the limit as @xmath10 . on the other hand , 1d structures can have features , such as a bandgap for any geometry and any refractive index contrast , that are certainly not shared by 2d or 3d pcs .
the precise distribution of randomness in the geometry of a pc would surely depend on the details of the specific manufacturing process . in the interest of simplicity , we chose the simplest distribution , the uniform distribution , in our study .
as the unit cell , we chose a unit supercell " that consisted of up to @xmath11 unit cells .
the thicknesses of the layers were perturbed by a given percent , by adding random numbers chosen from a uniform distribution . as the unperturbed structure , we chose the quarter - wave stack that has , for a given dielectric contrast , the largest relative gap between the first and the second bands , as can be seen in fig .
[ gapvsf ] . in
what follows , we will consider this structure with a dielectric contrast of 13 as our perfect pc . the relative gap width vs the filling ratio @xmath12 for a 1d pc made of slabs of alternating dielectric constant of @xmath13 and @xmath14 . the lowest gap has a maximum for @xmath15 , which is the quarter - wave stack value . for this value of @xmath12 , the even numbered gaps , the 2nd , 4th , etc , which are n general nonzero for an arbitrary value of @xmath12 are all closed.,width=226 ] for 1d pcs , one further has the luxury of calculating the bandgaps using an analytical method@xcite
this approach also permits the calculation of the imaginary part of the wave vector in the forbidden gap region and allows a reliable assessment of the accuracy of the plane wave method for the problem at hand .
we also investigated the transmission coefficient for a @xmath16 unit cell quarter - wave stack structure .
the transmission coefficient was calculated by simply matching the boundary conditions for the electric and the magnetic fields at each interface between the slabs in the multilayer structure .
maxwell s equations for waves propagating in the @xmath17 direction in a medium with a dielectric constant @xmath18 that depends only on @xmath17 , can be reduced to @xmath19 where @xmath20 is parallel to the slabs . with @xmath18 periodic along @xmath17 with lattice constant @xmath21 , and translationally invariant along @xmath22 and @xmath23 , @xmath24 where @xmath25 , is a reciprocal lattice vector with @xmath26 , and @xmath27
can be written as @xmath28 where @xmath29 . for a given @xmath30 ,
this yields an @xmath31-dimensional generalized eigenproblem @xmath32 or by multiplying both sides from the left by @xmath33 , one obtains the ordinary eigenproblem @xmath34 where @xmath35 , @xmath36 , and @xmath37 is the inverse of the matrix @xmath38 . for a given value of @xmath30 , a truncation of this @xmath31-dimensional ordinary eigenvalue problem yields , by retaining only the @xmath39-vectors with @xmath40 , the band structure @xmath41 and the modes @xmath42 .
we choose a structure where the dielectric constant alternates between two values @xmath43 and @xmath44 each with thickness @xmath45 and @xmath46 , respectively .
the choice of the lattice constant @xmath47 is not unique .
although the choice @xmath47 is the most obvious and the most convenient , the lattice constant can be chosen as any integer multiple of @xmath48 , @xmath49 . with a choice for @xmath50 with @xmath51 , and following the same formalism
one can write @xmath52 and @xmath53 where @xmath54 , with @xmath26 and @xmath55 .
clearly , to get results with the same level of accuracy as before , _
i.e. _ with @xmath56 , one would now need to include @xmath7 times as many plane waves in the expansion , which simply increases the computational burden , both in terms of storage and computing time . the band structure for @xmath57 and @xmath16
are displayed in fig [ folding ] for a perfect pc .
i the folding of the bands in the first brilloin zone for each @xmath7 , makes the appearance of the bands rather different for each case , although the dos and the eigenfunctions @xmath20 would be independent of the choice of the supercell size .
the frequency is plotted in units of @xmath58 for all cases , so the frequency scale is not affected with the result that the bandgaps are at the same frequency , as would be expected . to calculate the dos
, we choose a uniform mesh in @xmath30-space to calculate the bands and then choose a small frequency window , @xmath59 , and count the number of modes whose frequencies fall within that window .
we add random perturbations to the thicknesses of the layers in the supercell such that @xmath60\ ] ] where @xmath61 are the unperturbed values of the thicknesses of the layers , _
i.e. _ the quarter - wave stack values , @xmath62 is a uniformly distributed random number in the interval @xmath63 .
we control the amount of disorder by varying the percent randomness parameter @xmath64 between 0 and 1 .
@xmath65 corresponds to perfectly periodic structures , and @xmath66 corresponds to @xmath67 fluctuation where @xmath45 , @xmath46 can range between @xmath68 and twice their unperturbed values .
when disorder is introduced , gaps appear between every fold for @xmath69 . the upper and lower band edges for the lowest three gaps calculated with a supercell of size @xmath70 , as a function of the disorder parameter @xmath64.,width=264 ] in fig .
[ gapedgevsp ] we plot the upper and lower limits for the lowest three bandgaps as a function of @xmath64 , the percent randomness with a supercell size of @xmath70 .
note that since for quarter - wave stack structures the even numbered gaps are closed , the bandgaps in this figure are in fact the first , third and the fifth bandgaps of a 1d pc with arbitrary values for the layer thicknesses .
the third gap centered at @xmath71 closes around @xmath72 , the second gap centered at @xmath73 closes around @xmath74 , and the lowest gap centered at @xmath75 closes around @xmath76 .
the ratios of the critical values of randomness @xmath77 agree well with the ratios of the corresponding center gap frequencies , @xmath78 .
this can be understood using the simple argument that when the random fluctuations in the thicknesses of the layers become comparable to the wavelength of the gap center , the bandgap disappears since the destructive interference responsible for the existence of the forbidden band depends on the long range periodicity at that scale . as discussed in detail in ref@xcite , for @xmath79 dielectric layers with thicknesses @xmath80 , with dieletric constants @xmath81 , for a given @xmath82
, one can obtain the transfer matrix , defined by @xmath83 = \left[\begin{array}{cc } m_{11 } & m_{12 } \\
m_{21 } & m_{22 } \end{array } \right ] \left[\begin{array}{c } e_{2n } \\
e_{2n+1 } \end{array } \right]\ ] ] where @xmath84 with @xmath85 \ \ \ { \rm and}\ \ \ p_l=\left[\begin{array}{cc } e^{i\sqrt{\epsilon_l}\omega d_l / c } & 0\\ 0&e^{-i\sqrt{\epsilon_l}\omega d_l / c } \end{array } \right]\ ] ] imposing the bloch condition on the @xmath20-field , one obtains , @xmath86 = e^{ika } \left[\begin{array}{c } e_{2n } \\
e_{2n+1 } \end{array } \right].\ ] ] comparing with eq .
[ meq ] , the eigenvalues of the transfer matrix are seen to be @xmath87 . then , for @xmath88 , @xmath89 is real and is given by @xmath90 , while for @xmath91 , @xmath89 is complex with @xmath92 and @xmath93 .
@xmath94 is the decay length of the evanescent mode , and is a measure of the strength of the bandgap . for finite pcs ,
it s desirable to have @xmath95 to have a significant drop in transmittance .
the advantage of the exact method is that the supercell size @xmath7 can be increased to values that are practically impossible using the pw method .
while with the pw method , using 30 plane waves per unit cell of the supercell , the memory requirements scale as @xmath96 , and the time requirements scale as @xmath97 , the exact method requires a very small amount of memory .
fields used in transmittance and analytic badgap calculations.,width=302 ] the only disadvantage of the analytical method over the pw method is that , while in the pw method one chooses a real @xmath89 and calculates the frequencies corresponding to that value of @xmath89 , in the analytical method , one chooses the frequency @xmath82 and calculates the real and imaginary parts of @xmath89 , @xmath98 and @xmath99 , corresponding to that value of @xmath82 .
if the bands are nearly flat , as is the case for very large supercell sizes , then one needs to sample the frequency interval of interest in very tiny increments in @xmath82 in order to catch " a propagating mode .
thus the computation time can become very large . also for large values of @xmath7
, the transfer matrix @xmath100 can have very large elements so one requires very high precision in order to calculate the transmission resonance frequencies .
we used quadruple precision ( 128-bit ) floating point variables and functions in the intel fortran compiler in order to be able to resolve the transmission resonances for supercell sizes up to @xmath101 . for large values of @xmath7 , even 128-bit precision is not sufficient , with the result that @xmath99 can not be made to completely vanish due to insufficient precision . nevertheless , the propagating modes appear as sharp cusps in the @xmath99 vs @xmath82 graph which can easily be identified ( fig .
[ imbnd ] ) . for @xmath102 , one needs more than 128-bit precision to even see the cusps in ( fig .
[ imbnd ] ) . for larger values of @xmath7
, we used mathematica for its arbitrary precision capabilities .
however , compared to compiled code , mathematica is slower by several orders of magnitude , so we had to stop at around @xmath103 . the dependence of @xmath104 on @xmath7 for different values of the randomness parameter @xmath64 for the center gap frequency of the first gap , @xmath105 .
, width=196 ] to understand the _ bulk _ features of imperfect pcs using the supercell method , one would need a large supercell , in fact the larger the better .
as the supercell size is increased , small bandgaps begin to appear over regions that used to have propagating modes .
one would normally expect the gaps of the perfect crystal to gradually shrink in size , rather than have _ more _ gaps , so this result seems somewhat puzzling at first sight .
however , as the supercell size is increased , the statistical fluctuations decrease , and the pass bands become increasingly more densely populated . in fig [ imbnd ]
we display the behavior of @xmath99 as @xmath7 is increased . as @xmath7 becomes larger what used to be a photonic bandgap becomes more and more populated with transmission resonances , and the forbidden gap edges gradually approach each other , narrowing the gap .
it s possible that as @xmath106 , the whole bandgap region will be populated , albeit extremely sparsely , and instead of the bandgap , we will have a region where the dos is extremely small but non - zero nevertheless .
we were able to increase up to @xmath107 and the bandgap was reduced as @xmath7 became larger , although the decrease for very large @xmath7 values was very small . to actually see the gap narrow even more
would require an impractically large @xmath7 .
( upper left ) @xmath104 and @xmath99 vs @xmath82 for a supercell of size @xmath108 for a randomness level of 10% .
( upper right ) a closeup for @xmath109 which contains the lower edge of the second gap @xmath110 of the perfect pc .
( lower ) scatterplot of @xmath104 vs @xmath99 for the same structure.,title="fig:",width=157 ] ( upper left ) @xmath104 and @xmath99 vs @xmath82 for a supercell of size @xmath108 for a randomness level of 10% .
( upper right ) a closeup for @xmath109 which contains the lower edge of the second gap @xmath110 of the perfect pc .
( lower ) scatterplot of @xmath104 vs @xmath99 for the same structure.,title="fig:",width=163 ] ( upper left ) @xmath104 and @xmath99 vs @xmath82 for a supercell of size @xmath108 for a randomness level of 10% .
( upper right ) a closeup for @xmath109 which contains the lower edge of the second gap @xmath110 of the perfect pc .
( lower ) scatterplot of @xmath104 vs @xmath99 for the same structure.,title="fig:",width=226 ] the number of transmission resonances in any fixed frequency interval @xmath111 is proportional to @xmath7 , so in the bulk limit with @xmath106 , any wave packet with a small , but nonzero frequency spread @xmath111 would contain many transmission resonances and thus would be partly transmitted and partly reflected .
clearly in the bandgap regions of the perfect pc , the density of these transmission resonances is extremely small , and these regions still appear to be bandgaps with a large , but still finite , supercell .
hence , it seems plausible to conclude that one can not speak of a true bandgap " for imperfect pcs , but only of large depressions in the dos , which , in practice , would serve the same purpose as _ bona fide _ bandgaps .
for instance , for a cavity made of an impurity " embedded in a pc , localized cavity modes would eventually leak out through the pc walls " of finite thickness , regardless of how perfect the pc walls are , bacause of the finite thickness of the walls .
for such an application , what is important is that the lifetime of the cavity mode be much larger than the relevant time scale .
since the lifetime of the cavity mode is a function of the transmittance , for a given value of transmittance , one would simply need to use thicker walls as the random perturbations are increased .
practical applications must necessarily use finite sized pcs , and for such structures , a quantity of more relevance is the transmittance .
we calculate the transmittance by considering a pc of @xmath7 unit cells , and each layer is perturbed as described earlier .
the transmission coefficient is calculated by imposing the boundary conditions for @xmath20 @xmath112 , @xmath113 , and @xmath114 at each layer boundary ( figure [ transfig ] ) .
this yields a set of @xmath115 linear equations for the unknowns @xmath116 . setting the incident field @xmath117 , and assuming vacuum dielectric values for the incident and transmitted fields , @xmath118 , one obtains , @xmath119 and @xmath120 for the reflection and transmission coefficients , respectively .
alternatively , one could also obtain @xmath121 and @xmath122 from the transfer matrix as detailed in ref .
@xcite but with our approach , we can also obtain the @xmath123 within each layer .
for a given frequency in the gap region , the dependence of @xmath104 on the number of layers @xmath7 is approximately linear for all values of the randomness parameter @xmath64 , as is the case @xcite for the perfectly periodic finite crystal ( fig .
[ lntvsn ] ) .
however , depending on the level of randomness @xmath104 can change by many orders of magnitude . in practice
, this would mean that in order to obtain a given value of transmittance using an imperfect pc , one would now have to use a thicker pc .
although there is a strong relationship between the transmittance of a finite imperfect pc , and the modes of the superlattice formed by choosing the same exact finite pc as the unit supercell , this relationship is not perfect in the sense that the existence of a propagating mode does not necessarily imply a large value for @xmath122 .
conversely , the existence of a bandgap does not necessarily imply a low value for the transmittance @xmath122 .
a close examination of the @xmath104 and @xmath99 vs @xmath82 plots in fig [ lntkivsw ] will reveal that , on a large scale , the transmission has a dip where @xmath99 is large , and when @xmath99 is small the transmittance is nearly unity .
however , a closer look at a finer scale in fig .
[ lntkivsw ] , one sees that @xmath104 can still be not as large as what one might expect from @xmath99
. looking at the plots of @xmath104 vs @xmath99 fig .
[ lntkivsw ] , one could argue that @xmath99 sets an upper limit for @xmath104 .
this upper limit seems to be a linearly decreasing function of @xmath99 . that the actual value of @xmath104 can fluctuate below a particular value for a given @xmath99
can be understood if one considers the coupling of the incident plane wave to the modes inside the pc .
if , owing to the randomness , the propagating mode happens to have a small amplitude near @xmath124 , it will not be transmitted effectively .
similar coupling problems were reported by robertson _
@xcite for 2d pcs .
we studied the behavior of the photonic band gap and the transmittance for an imperfect pc using the supercell method combined with both the plane wave method and the analytical method .
we also studied the transmittance for an imperfect finite 1d photonic crystal and have shown that the results we obtain in all cases are consistent .
the bandgaps of the perfect pc are replaced by a dos that is extremely small to be detected even with extremely large supercells that consist of over 32000 unit cells .
the higher frequency bandgaps disappear first with the lowest gap closing at around a randomness level of @xmath125 .
this work was supported by a grant from the research fund at izmir institute of technology .
the band structure computations were performed on the 128-node beowulf cluster at tubitak , the scientific and technological research council of turkey .
99 e. yablonovitch , , 2059 ( 1987 ) .
s. john , , 2486 ( 1987 ) .
lin , j.g .
fleming , d.l .
hetherington , b.k .
smith , r. biswas , k.m . ho , m.m .
sigalas , w. zubrzycki , s.r .
kurtz , and j. bur , nature ( london ) * 394 * , 251 ( 1998 ) .
pochi yeh ,
optical waves in layered media " john wiley & sons , ( 1988 ) .
szer , j.w .
haus , and r. inguva , , 13962 ( 1992 ) .
a. moroz , , 115109 ( 2002 ) .
z.y . li and z.q .
zhang , , 1516 ( 2000 ) .
li , x. zhang and z.q .
zhang , , 15738 ( 2000 ) .
s. fan , p. r. villeneuve , and j.d .
joannopoulos , j. appl . phys .
* 78 * , 1415 ( 1995 ) .
m. m. sigalas , c. m. soukoulis , c. t. chan , r. biswas , and k. m. ho , phys .
b 59 , 12 767 ( 1999 ) . m.a .
kaliteevski , j.m .
martinez , d. cassagne , and j.p .
albert , , 113101 ( 2002 ) ; a.a .
asatryan , p.a .
robinson , l.c .
botten , r.c .
mcphedran , n.a .
nicorovici , and c. martijndesterke , phys .
e * 62 * , 5711 ( 2000 ) .
w. m. robertson , g. arjavalingam , r. d. meade , k. d. brommer , a. m. rappe , and j. d. joannopoulos , , 2023 ( 1992 ) . | the supercell method is used to study the variation of the photonic bandgaps in one - dimensional photonic crystals under random perturbations to thicknesses of the layers .
the results of both plane wave and analytical band structure and density of states calculations are presented along with the transmission coefficient as the level of randomness and the supercell size is increased .
it is found that higher bandgaps disappear first as the randomness is gradually increased .
the lowest bandgap is found to persist up to a randomness level of @xmath0 percent . |
the gross properties of a star , such as broad - band colours and flux distributions , are significantly affected by the effects of convection in stars later than mid a - type .
consequently , our modelling of convection in stellar atmosphere models can significantly alter our interpretation of observed phenomena . by comparison with stars of known @xmath0 and/or @xmath1 ( the fundamental stars ) , we can evaluate different treatments of convection in model atmosphere calculations .
photometric indices are a fast and efficient method for determining approximate atmospheric parameters of stars . for the commonly - used strmgren @xmath2 system a vast body of observational data exists which can be used to estimate parameters using calibrated model grids ( e.g. ( * ? ? ?
* moon & dworetsky 1985 ) , ( * ? ? ?
* smalley & dworetsky 1995 ) ) .
conversely , knowing atmospheric parameters from other methods , allows observed colours to be compared to model predictions .
this method has been used to compare various treatments of stellar convection .
the effects of convection on the theoretical @xmath2 colours of a , f , and g stars was discussed by @xcite , who compared the predicted colours for the @xcite ( cm ) model with that from the standard @xcite mixing - length theory ( mlt ) models with and without `` approximate overshooting '' .
comparison against fundamental @xmath0 and @xmath1 stars revealed that the cm models gave better agreement than mlt without overshooting .
models with overshooting were clearly discrepant .
this result was further supported by stars with @xmath0 obtained from the infrared flux method ( irfm ) and @xmath1 from stellar evolutionary models .
the observed stellar flux distribution is influenced by the effects of convection on the atmospheric structure of the star . as we have seen with photometric colours ,
these effects have a clearly observable signature ( see fig .
[ smalley - fig ] ) . in their discussion of convection
@xcite presented model stellar atmospheres using a modified mixing - length theory .
they found small , systematic differences in the optical fluxes .
their figures also demonstrate that convection can have a measurable effect on stellar fluxes .
hence , high precision stellar flux measurements will provide significant and useful information on convection .
= 7000k , @xmath1 = 4 models with cm and mlt ( @xmath3 = 0.5 and 1.25 ) , compared to that for a model with zero convection .
note that the region 4000 @xmath4 5000 is especially sensitive and the effect of overshooting is considerable . ]
unfortunately , very little high - precision stellar spectrophotometry exists .
this situation will be rectified , once the astra spectrophotometer ( see below ) begins operation .
this will allow spectrophotometry to be added to our observational diagnostic toolkit .
the temperature sensitivity of balmer lines makes them an excellent diagnostic tool for late a - type stars and cooler . the @xmath5 and @xmath6 profiles behave differently due to convection :
@xmath5 is significantly less sensitive to mixing - length than @xmath6 ( ( * ? ? ?
* vant veer & mgessier 1996 ) ) .
both profiles are affected by the presence of overshooting .
since @xmath5 is formed higher in the atmosphere than @xmath6 , balmer lines profiles are a very good depth probe of stellar atmospheres .
balmer profiles are also affected by microturbulence , metallicity and , for hotter stars , surface gravity ( ( * ? ? ?
* heiter 2002 ) ) . in their comparison of balmer line profiles
, @xcite found that both cm and mlt without overshooting gave satisfactory agreement with fundamental stars .
overshooting was again found to be discrepant .
in addition , @xcite found evidence for significant disagreement between all treatments of convection for stars with @xmath0 around 8000 @xmath4 9000 k. subsequently , @xcite reviewed this region using binary systems with known @xmath1 values and their revised fundamental @xmath0 values of the component stars .
they found that the discrepancy found was no longer as evident .
however , this region was relatively devoid of stars with fundamental values of both @xmath0 and @xmath1 .
further fundamental stars are clearly required in this region .
the automated spectrophotometric telescope research associates ( astra ) have developed a cassegrain spectrophotometer and its automated 0.5-m f/16 telescope .
there are being integrated at the fairborn observatory near nogales , arizona .
scientific observations are expected to begin in 2007 ( ( * ? ? ?
* ; * ? ? ?
* adelman 2007 , smalley 2007 ) ) . in an hour
the system will obtain s / n = 200 ( after correction for instrumental errors ) observations of stars as faint as 9.5 mag .
the spectrograph uses both a grating and a cross - dispersing prism to produce spectra from both the first and the second orders simultaneously .
the square 30 arc second sky fields for each order do not overlap .
the resolution is 7 in second and 14 in first order .
the wavelength range is of approximately @xmath73300 - 9000 .
the effects of convection on the stellar atmospheric structure can be successfully probed using a variety of observational diagnostics ( ( * ? ? ?
* smalley 2004 ) ) .
the combination of photometric colours and balmer - line profiles has given us a valuable insight into the nature of convection in stars .
high quality observations that are currently available and those that will be in the near future , will enable further refinements in our theoretical models of convection and turbulence in stellar atmospheres . | convection and turbulence in stellar atmospheres have a significant effect on the emergent flux from late - type stars .
the theoretical advancements in convection modelling over recent years have proved challenging for the observers to obtain measurements with sufficient precision and accuracy to allow discrimination between the various predictions .
an overview of the current observational techniques used to evaluate various convection theories is presented , including photometry , spectrophotometry , and spectroscopy .
the results from these techniques are discussed , along with their successes and limitations .
the prospects for improved observations of stellar fluxes are also given . |
observation of the kondo effect in a quantum dot ( qd ) @xcite has stimulated researches in the field of quantum transport , and recent experimental developments enable one to examine the kondo physics in a variety of systems , such as an aharonov - bohm ( ab ) ring with a qd and double quantum dots ( dqd ) . in these systems
multiple paths for electron propagation also affect the tunneling currents , and the interference causes fano - type asymmetric line shapes .
superconductivity also brings rich and interesting features into the quantum transport .
competition between superconductivity and the kondo effect has been reported to be observed in carbon nanotube qd and in semiconductor nanowires .
@xcite furthermore , interplay between the andreev scattering and the kondo effect has been studied intensively for a qd coupled to a normal ( n ) lead and superconductor ( s ) , theoretically @xcite and experimentally .
@xcite so far , however , the andreev - kondo physics has been discussed mainly for a single dot . in this paper
, we consider a dqd system with a t - shape geometry as shown in fig .
[ modelndqdscom ] , and study how multiple paths affect the interplay at low temperatures , using the numerical renormalization group ( nrg ) method .
@xcite golub and avishai calculated first , to our knowledge , the andreev transport through an ab ring with a qd,@xcite in which a similar interference effect is expected .
however , the underlying kondo physics in such a combination with superconductivity and interference is still not fully understood , and is needed to be clarified precisely , as measurements are being not impossible .
@xcite we find that the coulomb interaction in the side dot ( qd2 in fig .
[ modelndqdscom ] ) suppresses destructive interference typical of the t - shape geometry , and it enhances substantially the tunneling current between the normal and superconducting ( sc ) leads in the kondo regime .
this is quite different from the behavior seen in the normal transport in the same configuration fig .
[ modelndqdscom ] ( b ) , for which the conductance is suppressed , and shows a wide minimum called a kondo valley as a result of strong interference by the kondo resonance in the side dot .
@xcite the sc proximity penetrating into the interfacial dot ( qd1 in fig . [ modelndqdscom ] ) causes this stark contrast between the andreev and normal transports .
it also changes the fano line shape in the gate - voltage dependence of the conductance .
furthermore , we show that the proximity deforms the kondo cloud to make a singlet bond long , and it can be deduced from the fermi - liquid properties of the ground state . in sec .
[ sec : model ] , we introduce the model and describe the effective hamiltonian in a large gap limit . in sec .
[ sec : result ] , we show the numerical results and clarify the transport properties using the renormalized parameters .
the fano structures in the gate - voltage dependence of the conductance are also discussed .
a brief summary is given in the last section .
we start with an anderson impurity connected to sc and normal leads , @xmath0 where @xmath1 @xmath2 describes the interfacial ( @xmath3 ) and side ( @xmath4 ) dots : @xmath5 the energy level , @xmath6 the coulomb interaction , @xmath7 , and @xmath8 the inter - dot hopping matrix element . @xmath9 describes the sc / normal lead , and @xmath10 is a @xmath11-wave bcs gap .
@xmath12 is the tunneling matrix element between qd1 and the sc / normal lead .
we assume that @xmath13 is a constant independent of the energy @xmath14 , where @xmath15 is the total number of @xmath16 s in the leads . throughout the work , we concentrate on a large gap limit @xmath17 .
then the starting hamiltonian @xmath18 can be mapped exactly onto a single - channel model , which still captures the essential physics of the andreev reflection and makes nrg approach efficient , @xcite @xmath19 note that at @xmath17 the real and virtual excitations towards the continuum states outside the gap in the sc lead are prohibited .
nevertheless , the proximity from the sc lead to the dot remains finite , and it induces a local static pair potential @xmath20 ( @xmath21 ) at qd1 .
furthermore , the current can flow between the sc lead and the qd1 via @xmath22 .
for ( a ) @xmath24 and ( b ) @xmath25 , for several side - dot repulsions @xmath26 . the parameters for qd1 are chosen to be @xmath27 , and @xmath28 which is the local sc gap defined by @xmath29 .
] we can calculate the conductance at zero temperature as a function of the level position @xmath23 of qd2 for different values of @xmath26 , using the kubo formula.@xcite in this paper we focus on the coulomb interaction in the side dot ( qd2 ) , assuming that @xmath30 in the following .
the results of the conductance are shown in fig .
[ con - ed2 ] for @xmath27 and @xmath31 .
the coupling to the normal lead is chosen to be ( a ) @xmath24 and ( b ) @xmath25 .
the conductance is enhanced for the kondo regime @xmath32 , where the wide kondo valley appears in the case of the normal transport .
this is a novel phenomenon caused by the interplay between superconductivity and the kondo effect ; the local gap @xmath20 due to the proximity into qd1 leads to the andreev transport with destructive interference , but the introduction of @xmath26 suppresses the sc interference via qd2 , which in turn enhances the conductance .
note that the couplings are symmetric @xmath33 for fig .
[ con - ed2 ] ( a ) , and in this particular case the conductance increases with @xmath26 for any @xmath23 , except for the values @xmath34 and @xmath35 . outside of the kondo regime , the side dot is empty or doubly occupied , and the interference becomes no longer important . when the coupling to the normal lead is small @xmath36 as fig .
[ con - ed2 ] ( b ) , the conductance in the kondo regime decreases after the peak reaches the unitary limit @xmath37 .
this behavior can be related to a crossover from short - range to long - range kondo screening as illustrated in fig .
[ modelndqdscom ] , and is discussed later again .
we examine the behavior at the middle point @xmath38 of fig .
[ con - ed2 ] in detail .
the low - lying energy states show the fermi - liquid properties , @xcite and the conductance can be deduced from the renormalized parameters for the quasi - particles ( see eq .
( [ pcstg ] ) ) . at @xmath39 and @xmath27 for several values of @xmath26 .
( a ) @xmath24 and ( b ) @xmath25 .
the dashed line is the conductance of a single dot ( @xmath40 ) , for which the horizontal axis should be interpreted as ( a ) @xmath41 and ( b ) @xmath42 . ]
the conductance is plotted as a function of @xmath43 in fig .
[ con - gs ] for ( a ) @xmath24 and ( b ) @xmath25 .
we see that the peak shifts towards smaller @xmath22 as @xmath26 increases and will coincide in the limit of @xmath44 with the dashed line , which corresponds to the conductance without the side dot .
it means that the interference caused by the side dot is suppressed completely for large @xmath26 , and in this limit the conductance reaches the unitary limit value for the symmetric couplings @xmath45 .
the difference in the line shape of fig .
[ con - ed2 ] ( a ) and that of fig .
[ con - ed2 ] ( b ) at fixed @xmath22 reflects the position of the unitary - limit peak in fig .
[ con - gs ] . in order to clarify the properties of the the ground state
precisely , we consider a special case @xmath38 .
then the hamiltonian @xmath46 can be written in terms of the interacting bogoliubov particles , which conserve the total charge , as shown in appendix.@xcite consequently , the low - energy states can be described by a local fermi liquid , the fixed - point hamiltonian@xcite of which can be written in the form @xmath47 here , @xmath48 is a local sc gap that emerges in qd2 via the self - energy correction due to @xmath26 , while @xmath29 as defined in eq .
( [ delta_d1 ] ) is caused by the direct proximity from the sc lead .
@xmath49 is the renormalized value of the inter - dot hopping matrix element .
we calculate these parameters from the fixed - point of nrg.@xcite then , the conductance @xmath50 and a staggered sum @xmath51 of the pair correlation are deduced from the phase shift , @xmath52 , of the bogoliubov particles , @xmath53 : ( a ) conductance , ( b ) @xmath49 , @xmath48 , ( c ) @xmath54 , @xmath55 , and @xmath56 .
we choose @xmath27 , @xmath57 , and for filled ( open ) marks @xmath24 ( @xmath58 ) .
inset of ( b ) : phase boundary , between singlet and doublet ground states , for an isolated dqd ( @xmath59 ) with a finite local sc gap @xmath29 . ] in fig .
[ all ] , we show the @xmath26 dependence of the ground - state properties at @xmath38 for @xmath27 and @xmath57 .
the coupling is chosen to be @xmath60 , and @xmath61 .
the conductance for @xmath62 shows a peak as a function of @xmath26 , while for @xmath63 it increases simply towards the unitary limit .
this corresponds to the difference that we see in fig .
[ con - ed2](a ) and ( b ) at @xmath38 .
figure [ all](b ) shows the renormalized parameters ( @xmath64 ) @xmath49 and ( @xmath65,@xmath66 ) @xmath67 .
the ratio @xmath68 , which equals to the square root of the wavefunction renormalization factor @xmath69 ( see appendix ) , decreases monotonically from @xmath70 to @xmath35 with increasing @xmath26 , while the local sc gap @xmath48 becomes large for intermediate values of @xmath26 .
the behavior of these fermi - liquid parameters implies that a crossover from weak to strong correlation regimes occurs around @xmath71 .
the nature of the crossover can be related to a level crossing taking place in a _ molecule _
limit @xmath59 , where qd1 is decoupled from the normal lead . in this limit
, the isolated dqd is described by a hamiltonian @xmath72 , which includes the local sc gap @xmath29 at qd1 .
the ground state of the _ molecule _ is a singlet or doublet depending on @xmath73 and @xmath43 , as shown in the inset of fig .
[ all](b ) .
the ground state is a spin - singlet , if either @xmath73 or @xmath43 is small . in the opposite case ,
a spin - doublet becomes the ground state .
the local moment in this doublet state emerges mainly at qd2 , because the correlation at qd1 is small in the present situation @xmath27 .
we see in the phase diagram in fig .
[ all](b ) that the transition takes place in this _ molecule _ limit at @xmath71 for @xmath57 , and it agrees well with the position where the conductance peak appears in fig .
[ all](a ) . for finite @xmath74
, conduction electrons can tunnel from the normal lead to qd2 via qd1 .
however , the sc correlation @xmath29 tends to make the local state at qd1 a singlet , which consists of a linear combination of the empty and doubly occupied states .
thus , for large @xmath22 , the electrons at qd1 can not contribute to the screening of the moment at qd2 .
in this situation , the kondo screening is achieved mainly by the conduction electrons tunneling to qd2 virtually via qd1 .
this process is analogous to a superexchange mechanism , which can also be expected from the hamiltonian written in terms of the bogoliubov particles in appendix . from these observations we see that the conductance peak at @xmath71 in fig .
[ all](a ) reflects the crossover from the short - range singlet to long - range one due to the superexchange screening process ( see also fig . [ modelndqdscom ] ) for the bogoliubov particles . note that the peak structure of the conductance vanishes for @xmath75 .
the deformation of the kondo cloud can also be deduced from the results shown in fig .
[ all](c ) .
this is because the staggered pair correlation @xmath51 is related directly to the scattering phase shift @xmath52 of the bogoliubov particles , by the friedel sum rule eq . given in appendix
. therefore , the value of @xmath51 reflects the changes occurring in the kondo clouds . particularly , a sudden change observed in @xmath51 around @xmath71 shows that the phase of the wavefunction shifts by @xmath76 during this change .
this also explains the occurring of the crossover from the short - range to long - range screening .
we can also calculate each correlation function @xmath77 directly with nrg based on the definition .
the local sc correlations @xmath54 and @xmath55 have the same value in the noninteracting case @xmath78 , and thus in this particular limit there is no reduction in the amplitude of the proximity from qd1 to qd2 .
the coulomb interaction @xmath26 causes the reduction , as both @xmath54 and @xmath55 decrease for small @xmath26 where the ground state is a singlet with a _ molecule _ character . for large @xmath26 ,
the sc correlation @xmath79 almost vanishes in qd2 , while @xmath54 shows an upturn and approaches the value expected for @xmath40 .
therefore , the sc proximity into qd1 is enhanced when the kondo cloud expands to form a long - range singlet .
then the tunneling current is not interfered much by the local moment at qd2 , and flows almost directly without using the path to the side dot . for several values of @xmath81 , where @xmath82 and @xmath57 .
other parameters : ( a ) @xmath24 , @xmath83 .
( b ) @xmath25 , @xmath84 . ]
so far , we have chosen the level of qd1 to be @xmath85 .
the result obtained for different values of @xmath81 is plotted as a function of @xmath23 in fig .
[ con - ed2_extra ] .
we see that two asymmetric fano structures , each of which consists of a pair of peak and dip , emerge at @xmath86 and @xmath87 for @xmath80 , as the fermi level crosses the energy corresponding to the upper and lower levels of the atomic limit .
the conductance peaks become sharper for @xmath62 as @xmath88 increases .
et al _ studied the fano structure in the side - coupled dqd system with two normal leads,@xcite and showed that at low temperatures the conductance has only one pair of the peak and dip , which are separated widely by a fano - kondo plateau at @xmath89 .
this type of the plateau was known earlier to appear in a qd embedded in an ab ring.@xcite in contrast , our result shows that the fano - kondo plateau vanishes , when one of the leads is a superconductor .
this is because the kondo screening in this case is achieved by the long - range singlet bond due to the superexchange process , as a result of the competition between the sc proximity into qd1 and coulomb interaction at qd2 .
we have studied andreev transport through the side - coupled dqd with nrg approach .
we have found that the coulomb interaction in the side dot suppresses the destructive interference effect typical of the t - shape geometry , and enhances the tunneling current between the normal and sc leads .
this novel phenomenon is caused by the interplay between the sc correlation and the kondo effect ; the sc proximity into qd1 pushes the kondo cloud towards the normal lead , and the conductance shows a peak of the unitary limit as the nature of the singlet changes from a short - range to long - range one .
we have also clarified that two asymmetric fano structures appear in the gate - voltage dependence of the andreev transport , instead of a reduced single fano - kondo plateau which appears in the kondo regime of the normal transport .
we thank k. inaba for valuable discussions .
is also grateful to j. bauer and a. c. hewson for discussions . the work is partly supported by a grant - in - aid from mext japan ( grant no.19540338 ) .
y.t . is supported by jsps research fellowships for young scientists .
is supported by jsps grant - in - aid for scientific research ( c ) .
the hamiltonian @xmath46 defined in eq . can be transformed , at @xmath38 , into the interacting bogoliubov particles , which conserve the total charge . for describing this property briefly ,
we rewrite @xmath46 using the logarithmic discretization of nrg,@xcite @xmath90 for @xmath91 , the operator @xmath92 describes the conduction electron in the normal lead , and @xmath93 is given by @xmath94 here , @xmath95 is the half - width of the conduction band . for the double - dot part , we use a notation @xmath96 for @xmath97 .
correspondingly , @xmath98 and @xmath99 with @xmath100 at @xmath101 , the system has a uniaxial symmetry in the nambu pseudo - spin space,@xcite and the hamiltonian can be simplified by the transformation @xmath102 = \left [ \begin{array}{cc } u & -v \\ v & \ u \rule{0cm}{0.5 cm } \end{array } \right ] \left [ \begin{array}{c } f_{n\uparrow}^{\phantom{\dagger } } \\ ( -1)^{n-1 } f_{n\downarrow}^{\dagger } \end{array } \right]\ ; , \label{eq : bogo_b } \\ & u=\sqrt{\frac{1}{2}\left(1+\frac{\xi_{d1}}{e_{d1}}\right ) } , \quad v=\sqrt{\frac{1}{2}\left(1-\frac{\xi_{d1}}{e_{d1}}\right ) } \label{eq : bogo_factor_b } \;.\end{aligned}\ ] ] here , @xmath103 , @xmath104 , and @xmath29 as defined in eq . .
then @xmath105 is transformed into a normal two - impurity anderson model for the bogoliubov particles @xmath106 .
\label{eq : nrg_h_nt_bogo}\end{aligned}\ ] ] here , @xmath107 , and the total number of the bogoliubov particles , @xmath108 , is conserved .
the equation implies that the low - energy excited states can be described by a local fermi - liquid theory .
this is true also for the original hamiltonian @xmath46 , and it does not depend on the discretization procedure of nrg.@xcite to be specific , we assume that @xmath30 in the following . in this case , the bogoliubov particles feel a normal impurity potential @xmath109 at qd1 , and this potential causes the superexchange mechanism that makes the singlet - bond long range as discussed in sec .
[ subsec : fl ] . the retarded green s function for
the bogoliubov particle @xmath110 at qd2 takes the form @xmath111 where @xmath112 is the self - energy caused by the interaction @xmath113 . at zero temperature ,
the asymptotic form of the green s function for small @xmath114 is given by @xmath115 note that @xmath48 has a finite value even though @xmath116 , because @xmath117 .
the value of the parameters @xmath48 and @xmath49 can be deduced from the fixed point of nrg.@xcite then , using the friedel sum rule for eq . , the local charge at the double dot can be calculated from the phase shift @xmath52 of the bogoliubov particles , @xmath118 the charge of the bogoliubov particles corresponds to the sc pair correlation for the original electrons @xmath92 . specifically for @xmath119 ,
it is transformed into the staggered sum @xmath51 given in eq . , by the inverse transformation of eq . .
similarly , the conductance @xmath50 can be expressed in terms of the phase shift @xmath52 .
furthermore , the free quasi - particles corresponding to the green s function given in eq . can be described by a hamiltonian , which is rewritten in terms of the original electron operators in eq . by the inverse transformation .
d. goldhaber - gordon , h. shtrikman , d. mahalu , d. abusch - magder , u. meirav , and m. a. kanster , nature ( london ) * 391 * , 156 ( 1998 ) ; s. m. cronenwett , t. h. oosterkamp , and l. p. kouwenhoven , science * 281 * , 540 ( 1998 ) . | we study the transport through side - coupled double quantum dots , connected to normal and superconducting ( sc ) leads with a t - shape configuration .
we find , using the numerical renormalization group , that the coulomb interaction suppresses sc interference in the side dot , and enhances the conductance substantially in the kondo regime .
this behavior stands in total contrast to a wide kondo valley seen in the normal transport . the sc proximity penetrating into the interfacial dot pushes the kondo clouds , which screens the local moment in the side dot , towards the normal lead to make the singlet bond long .
the conductance shows a peak of unitary limit as the cloud expands .
furthermore , two separate fano structures appear in the gate - voltage dependence of the andreev transport , where a single reduced plateau appears in the normal transport . |
motivated by the beliefs that one can directly identify a black hole existence by comparing qnms with the gravitational waves observed in the universe , and the study of qnms may lead to a deeper understanding of the thermodynamic properties of black holes in loop quantum gravity @xcite @xcite , as well as the qnms of anti - de sitter black holes have a direction interpretation in terms of the dual conformal field theory @xcite @xcite @xcite , the investigation of qnms in black hole spacetimes becomes appealing recent years .
since regge and wheeler @xcite first presented the idea of the qnms and chandrasekhar and detweller @xcite worked out the numerically calculation of the qnms for the schwarzschild black hole , a great deal of effort has been contributed to compute the qnms of the black holes @xcite-@xcite for fields of integer spin , such as the scalar , electromagnetic and gravitational perturbations . on the contrary , the study of the qnms of the dirac field is very limited @xcite@xcite@xcite .
cho @xcite studied the qnms of the dirac field of the schwarzschild black hole with the help of the third - order wkb method , and zhidenko @xcite extended the study to the case of the schwarzschild - de sitter black hole by using the sixth - order wkb method .
we @xcite investigated the dirac qnms of the reissner - nordstrm de sitter black hole using the pshl - teller potential approximation .
nevertheless , in these papers the study of the dirac qnms were limited by using wkb or pshl - teller potential approximative methods .
the reason to obstruct study of the dirac qnms is that for static black holes the standard wave - equations @xmath2 possesses the special potentials @xmath3 which are the functions of @xmath4 @xcite@xcite@xcite , where @xmath5 is a function related to metric , say @xmath6 for the schwarzschild black hole .
so , it is very hard to calculate dirac qnms by using the known numerical methods ( except wkb or pshl - teller approximative methods ) since we have to expand the potentials as a series at boundary of the field , such as at the event horizon .
however , castello - branco _ et al _ @xcite evaluated the dirac qnms of the schwarzschild black hole by using the continued fraction approach in which they expanded the wave function as a series of @xmath7 , and they found that the spacing for imaginary part is not @xmath8 , as it takes place for scalar , electromagnetic , and gravitational perturbations , but @xmath9 .
recently , we @xcite@xcite found that the wave functions and potentials of the dirac field in the static spacetimes can be expressed as new forms , and the new wave functions are related to the standard wave functions @xmath10 in eq .
( [ sta ] ) in a simple way @xcite .
starting from the new wave functions and potentials , we @xcite@xcite showed that the dirac qnms of the schwarzschild anti - de sitter and reissner - norstrm anti - de sitter black holes can easily be worked out by using horowitz - hubeny approach @xcite .
we have checked that the new potentials also present correct quasinormal frequencies for the scalar and gravitational perturbations in the schwarzschild black hole .
the main purpose of this paper is to study dirac qnms of the schwarzschild black hole by using the new kinds of wave functions and potentials with the help of the continued fraction and hill - determinant approaches .
the results obtained here are different from castello - branco s one @xcite ( if taking @xmath11 , we can only obtain results @xmath12 in ref .
@xcite . that is to say we do not find specific modes " . ) .
we are sure that we do not miss any quasinormal frequency because the hill - determinant approach gives all roots for @xmath13 .
the organization of this paper is as follows . in sec .
2 the decoupled dirac equations and corresponding wave equations in the schwarzschild spacetime are obtained by using newman - penrose formalism . in sec .
3 the numerical approach to computing the dirac qnms is introduced . in sec .
4 the numerical results for the dirac qnms in the schwarzschild black hole are presented .
the last section is devoted to a summary .
the dirac equations @xcite are @xmath14 where @xmath15 is covariant differentiation , @xmath16 and @xmath17 are the two - component spinors representing the wave function , @xmath18 is the complex conjugate of @xmath19 , and @xmath20 is the particle mass . in the newman - penrose formalism
@xcite the equations become @xmath21 for the schwarzschild spacetime @xmath22 where m represents the black hole mass , the null tetrad can be taken as @xmath23 with @xmath24 then , if we set @xmath25 where @xmath26 and @xmath27 are the energy and angular momentum of the dirac particles , after tedious calculation eq .
( [ np ] ) can be simplified as @xmath28 with @xmath29 we can eliminate @xmath30 ( or @xmath31 ) from eqs .
( [ aa1 ] ) and ( [ aa2 ] ) and obtain @xmath32s_s=0,\end{aligned}\ ] ] here and hereafter we take @xmath33 for the case @xmath30 ( @xmath34 ) and @xmath35 for @xmath31 ( @xmath36 ) , and @xmath37 and @xmath38 . the angular equation ( [ ang ] ) can be solved exactly and @xmath39 , where @xmath40 is the quantum number characterizing the angular distribution .
so , for both cases @xmath33 and @xmath35 we have @xmath41 we will focus our attention on the massless case in this paper .
therefore , we can eliminate @xmath42 ( or @xmath43 ) from eqs .
( [ dd2 ] ) and ( [ dd3 ] ) to obtain a radial decoupled dirac equation for @xmath44 ( or @xmath42 ) . then , introducing an usual tortoise coordinate @xmath45 and resolving the equation in the form @xmath46 we obtain the wave equation @xmath47 where @xmath48+\frac{s^2 r^4}{4}\left[\frac{d}{d r}\left(\frac{\delta}{r^4}\right)\right]^2+is \omega r^2\frac{d}{d r}\left(\frac{\delta}{r^4}\right)+\frac{\lambda^2 \delta}{r^4}.\end{aligned}\ ] ]
although the form of the potential is more complicated than standard wave - equations @xmath49 with the potentials @xmath50 @xcite@xcite@xcite , we will see that we can easily work out the dirac quasinormal frequencies of the schwarzschild black hole from eqs .
( [ wave ] ) and ( [ poten ] ) by using continued fraction and hill - determinant approaches .
for the schwarzschild black hole , the qnms are defined to be modes with purely ingoing waves at the event horizon and purely outgoing waves at infinity @xcite .
then , the boundary conditions on wave function @xmath51 at the horizon @xmath52 and infinity @xmath53 can be expressed as @xmath54 where @xmath55 is the surface gravity of the black hole . a solution to eq .
( [ wave ] ) that has the desired behaviour at the boundary can be written in the form @xmath56 if we take @xmath57 , the sequence of the expansion coefficient @xmath58 is determined by a three - term recurrence relation staring with @xmath59 : @xmath60 the recurrence coefficient @xmath61 , @xmath62 , and @xmath63 are given in terms of @xmath64 and the black hole parameters by @xmath65 with @xmath66 the three terms recurrence relation ( [ rec ] ) can be used to find quasinormal frequencies by using both the continued fraction and hill - determinant methods which are introduced in the following . *
[ the continued fraction approach ] : * the ratio of the series coefficients @xmath67 is finite and can be determined in two ways : @xmath68 using eqs .
( [ rec ] ) and ( [ an ] ) we can obtain an equality between two continued fractions , one of infinite length , and the other finite : @xmath69= \left[\frac{\alpha_m\gamma_{m+1}}{\beta_{m+1}- } \frac{\alpha_{m+1}\gamma_{m+2}}{\beta_{m+2}- } \frac{\alpha_{m+2}\gamma_{m+3}}{\beta_{m+3}-} ... \right ] , ~~(m=1,2 ... ) .
\end{aligned}\ ] ] this leads to a simple method to find quasinormal frequencies define a function which returns the value of the continued fraction for an initial guess at the frequency .
then , use a root finding routine to find the zeros of this function in the complex plane .
the frequency for which this happens is a quasinormal frequency . the @xmath70th quasinormal frequency is usually found to be numerically the most stable root of the @xmath70th inversion @xcite .
the success of the root - finding technique depends greatly on the initial guess given for the root .
however , which initial approximation we shall take is a difficult work , especially at the beginning .
fortunately , we can easily find the low overtone quasinormal frequencies by means of the hill - determinant method .
* [ the hill - determinant method ] : * the condition that nontrivial solutions of eq .
( [ rec ] ) exist is given by the requirement that the hill - determinant vanish @xcite@xcite .
the determinant can be easily seen in this case to be of the form @xmath71 if @xmath72 stand for the @xmath73 determinant of @xmath74 , then the @xmath72 can be concisely expressed in terms of a difference relation @xcite @xmath75 these equations are polynomials in @xmath26 and can be solved at each order . however , the further requirement that the waves be purely outgoing at infinity and obey the stability conditions restrict the frequencies to complex plane with negative imaginary parts which are just the quasinormal frequencies we look for .
the method is conceptually simpler than the continued fraction method of leaver .
however , the method looses accuracy for the higher overtones since one must evaluate the determinant at values of order @xmath76 ( for giving @xmath76 , one require solving polynomials of the order of @xmath77 in @xmath26 ) very much higher than the overtone number @xmath70 of the quasinormal modes .
also , the method does not have any control over the stability of the root search algorithm .
we have checked that both the methods present the same results for @xmath78 to @xmath79 and @xmath80 to @xmath81 .
in this section we represent the numerical results of the dirac quasinormal frequencies of the schwarzschild black hole obtained by using the numerical approaches just outlined in the previous section .
the results will be organized into two subsections : the fundamental quasinormal modes and high overtones .
the fundamental quasinormal frequencies ( n=0 ) for @xmath82 to @xmath83 are listed in the table ( [ table1 ] ) .
( [ fig1 ] ) shows that @xmath84 as a function of @xmath85 . from the table and figure
we know that the fundamental quasinormal frequencies become evenly spaced for large @xmath85 and the spacing is given by @xmath86 .[table1 ] the fundamental dirac quasinormal frequencies of the schwarzschild black hole for @xmath82 to @xmath83 . [ cols="^,^,^,^,^,^,^,^",options="header " , ] first 21 dirac quasi - normal frequencies of the schwarzschild black hole for @xmath82 to @xmath79 . ]
dependence of the real parts of the quasinormal frequencies on @xmath85 for the schwarzschild black hole , which shows that @xmath85 has the surprising effect of increasing @xmath87 . ]
dependence of the imaginary parts of the quasinormal frequencies on @xmath85 for the schwarzschild black hole , which shows that @xmath85 almost does not affect imaginary part , @xmath88 , especially for the lower overtones . ] graphs of @xmath89 versus overtone number @xmath70 for @xmath90 .
the top figure is drawn for @xmath91 which shows that @xmath92 tends to zero for large @xmath70 .
the bottom one is for @xmath93 which shows that @xmath94 for large @xmath70.,title="fig : " ] + graphs of @xmath89 versus overtone number @xmath70 for @xmath90 .
the top figure is drawn for @xmath91 which shows that @xmath92 tends to zero for large @xmath70 .
the bottom one is for @xmath93 which shows that @xmath94 for large @xmath70.,title="fig : " ]
the wave equations for the dirac fields in the schwarzschild black hole spacetime are obtained by means of the newman - penrose formulism .
then , the quasinormal frequencies corresponding to the dirac field perturbation in the black hole spacetime are evaluated by using continued fraction and hill - determinant approaches .
three interesting results obtained in this paper are listed in the following : ( i ) the fundamental quasinormal frequencies become evenly spaced for large @xmath85 and the spacing is given by @xmath95 .
( ii ) @xmath85 has the surprising effect of increasing real part of the quasinormal frequencies , but it almost does not affect imaginary part , especially for the low overtones .
( iii ) the quasinormal frequencies also become evenly spaced for large @xmath70 and the spacing for imaginary part is @xmath8 , which is same as that of the scalar , electromagnetic , and gravitational perturbations . in future papers in this series
the new wave functions and potentials will be applied to study of dirac qnms in other black hole spacetimes , such as the charged reissner - norstrm black hole and the rotating kerr black hole .
this work was supported by the national natural science foundation of china under grant no .
10275024 and under grant no .
10473004 ; the fanedd under grant no .
200317 ; and the srfdp under grant no . | the quasinormal modes ( qnms ) associated with the decay of dirac field perturbation around a schwarzschild black hole is investigated by using continued fraction and hill - determinant approaches .
it is shown that the fundamental quasinormal frequencies become evenly spaced for large angular quantum number and the spacing is given by @xmath0 .
the angular quantum number has the surprising effect of increasing real part of the quasinormal frequencies , but it almost does not affect imaginary part , especially for low overtones .
in addition , the quasinormal frequencies also become evenly spaced for large overtone number and the spacing for imaginary part is @xmath1 which is same as that of the scalar , electromagnetic , and gravitational perturbations .
= 0.65 cm |
since the first extrasolar planet was discovered in 1995 @xcite , more than 800 exoplanets have been detected as of march 2013 , owing to improvements in both observational instruments and the methods of analysis .
although most known exoplanets are gas giants , estimates based on both theory and observation indicates that terrestrial planets are also common @xcite .
supporting these estimates is the fact that earth - like planets have indeed been discovered .
moreover , space telescopes ( e.g. , _ kepler _ ) have now released observational data about many terrestrial - planet candidates . whether terrestrial planets with liquid water exist is an important question to consider because it lays the groundwork for the consideration of habitability .
the orbital range around a star for which liquid water can exist on a planetary surface is called the habitable zone ( hz ) ( hart 1979 ; kasting et al .
the inner edge of the hz is determined by the runaway greenhouse limit @xcite , and the outer edge is estimated from the effect of co@xmath0 clouds @xcite .
the region between these edges is generally called the hz for terrestrial planets with plentiful liquid water on the surface ( ocean planets ) .
planets with plentiful water on the surface but outside the outer edge of the hz would be globally covered with ice , and no liquid water would exist on the surface .
these are called `` snowball planets '' ( tajika 2008 ) .
moreover , an ocean planet could be ice - covered even within the hz because multiple climate modes are possible , including ice - free , partially ice - covered , and globally ice - covered states @xcite .
although such planets would be globally ice - covered , liquid water could exist beneath the surface - ice shell if sufficient geothermal heat flows up from the planetary interior to melt the interior ice . in this scenario ,
only a few kilometers of ice would form at the surface of the ocean @xcite , and life could exist in the liquid water under the surface - ice shell @xcite . another possibility is presented by planets that float in space without being gravitationally bound to a star ( free - floating planets ) , as have been found thanks to recent advances in observational techniques @xcite . although such planets receive no energy from a central star , even a free - floating earth - sized planet with considerable geothermal heat could have liquid water under an ice - covered surface
. considering geothermal heat from the planetary interior , @xcite discusses the theoretical restrictions for ice - covered extrasolar terrestrial planets that , on the timescale of planetary evolution , have an internal ocean .
@xcite shows that an internal ocean can exist if the water abundance and planetary mass are comparable to those of earth .
a planet with a mass less than @xmath2 can not maintain an internal ocean .
for a planet with mass @xmath3 , liquid water would be stable either on the planetary surface or under the ice , regardless of the luminosity of the central star and of the planetary orbit .
these are important conclusions and have important implications for habitable planets . in this paper
, we extend the analysis of tajika ( 2008 ) and vary the parameter values such as abundance of radiogenic heat sources and h@xmath0o abundance on the surface . although @xcite assumed that the mass ratio of h@xmath0o on the planetary surface is the same as that on earth ( 0.023 wt% ) , the origin of water on the earth is not apparent @xcite so it is possible that extrasolar terrestrial planets have some order of h@xmath0o abundance .
we investigate this possibility by varying the h@xmath0o abundance in our simulation , and also check whether ice appears under h@xmath0o layers under high - pressure conditions ( see section 2.2 ) .
therefore , in this work , we consider the effect of high - pressure ice under an internal ocean and discuss its implications for habitability ( see section 4.2 ) . with these considerations , we discuss the conditions required for bound and unbound terrestrial planets to have an internal ocean on the timescale of planetary evolution ( owing to geothermal heat flux from the planetary interior ) .
our discussion further considers various planetary masses , distances from the central star , water abundances , and the abundances of radiogenic heat sources . finally , taking into account the effects of high - pressure ice
, we investigate the structure of surface - h@xmath0o layers of ice - covered planets .
to calculate the mass - radius relationships for planets with masses in the range 0.1 @xmath4 - 10 @xmath4 , we adjust the planetary parameters .
we assume @xmath5 as per @xcite , where @xmath6 is the planetary radius and @xmath7 is the planetary mass .
the subscript @xmath8 denotes values for earth . the mantle thickness , core size , amount of h@xmath0o , average density , and other planetary properties
are scaled according to this equation .
the planetary surfaces are assumed to consist of frozen h@xmath0o and to have no continental crust .
we define the planetary radius as @xmath6 @xmath9 @xmath10 @xmath11 @xmath12 , where @xmath10 is the h@xmath0o thickness and @xmath12 is the mantle - core radius ( see fig .
the mass of h@xmath13o on the planetary surface is given by @xmath14,\ ] ] where @xmath15 is the density of h@xmath0o .
we vary @xmath16 from @xmath17 to @xmath18 , where @xmath19 @xmath9 @xmath20 with the prefactor being the h@xmath0o abundance of earth ( 0.023 wt.% ) . assuming that the heat flux @xmath21 is transferred from the planetary interior through the surface ice shell by thermal conduction
, the ice thickness @xmath22 can be obtained as @xmath23 where @xmath24 is the thermal conductivity of ice , @xmath25 is the temperature at the bottom of the ice , and @xmath26 is the temperature at the surface .
we assume that the surface ice is hexagonal ice ( ice ih ) . between 0.5 k and 273 k , the thermal conductivity of ice ih
is known @xcite . for temperatures greater than @xmath27 25 k ,
it is given by @xcite as @xmath28}}{t}.\ ] ] to estimate @xmath25 , we assume that the melting line of h@xmath0o is a straight line connecting ( 0 bar , 273 k ) to ( 2072 bar , 251 k ) in the linear pressure - temperature phase diagram . the temperature @xmath25 can be estimated using @xmath29}}{2072 \mathrm{[bar]}}p_{ib } \nonumber \\ & = & 273 - \frac{22 \mathrm{[k]}}{2072 \mathrm{[bar]}}dh\rho_w g \times 10^{-5},\end{aligned}\ ] ] where @xmath30 ( bar ) is the pressure at the bottom of the ice and @xmath31 is the gravitational acceleration on earth . considering energy balance on the planetary surface ,
the planetary surface temperature @xmath26 is @xmath32 where @xmath33 is the planetary albedo , @xmath34 is the distance from the central star in au , @xmath35 is the luminosity of the central star , @xmath36 m is a distance of 1 au in meters , @xmath37 is the emissivity of the planet , and @xmath38 wm@xmath39k@xmath40 is the stefan - boltzmann constant .
we assume @xmath41 and @xmath42 ( i.e. , the planetary atmosphere contains no greenhouse gases , which yields an upper estimate of the ice thickness ) .
the increase in luminosity due to the evolution of the central star as a main sequence star @xcite is considered using @xmath43^{-1}l_\odot,\ ] ] where @xmath44 years , and @xmath45 w. from these models , we can obtain the h@xmath0o thickness @xmath10 and the ice thickness @xmath22 .
the condition for terrestrial planets to have an internal ocean is @xmath46 to estimate the geothermal heat flux @xmath21 through planetary evolution , we investigate the thermal evolution of terrestrial planets by using a parameterized convection model ( tajika matsui 1992 ; mcgovern
schubert 1989 ; franck bounama 1995 ; von blow et al .
2008 ; see appendix for details ) .
we assume @xmath47 , which is the initial heat generation per unit time and volume , is 0.1@xmath48 to 10@xmath48 , where the constant @xmath48 is the initial heat generation estimated from the present heat flux of the earth ( see appendix for details ) .
ice undergoes a phase transition at high pressure ( fig .
2 ) . unlike ice
i , the other phases are more dense than liquid h@xmath0o .
we call the denser ice `` high - pressure ice . ''
because @xcite assumes that the amount of h@xmath0o on the planetary surface is the same as that on the earth s surface @xmath19 ( @xmath49 ) , the only possible conditions on the planetary surface are those labeled 1 , 2 , and 3 in fig .
3 . however , because we consider herein that h@xmath0o amounts may range from @xmath17 to @xmath18 , the h@xmath0o - rock boundary could move to higher pressure , so we should account for the effect of high - pressure ice ( fig .
therefore , types 4 , 5 , and 6 of fig . 3b are added as possible surface conditions .
types-2 and type-5 planets both have an internal ocean , but high - pressure ice exists in type-5 planets between the internal ocean and the underlying rock .
we approximate the melting curve by straight lines connecting the triple points in the linear pressure - temperature phase diagram .
we also assume that the amount of heat flux from the planetary interior that is transferred through the internal ocean by thermal convection is the same as that which is transferred through the surface ice . here
, we presume that temperature gradient in liquid - water part of phase diagram is isothermal ( figs .
3a and 3c ) , although a gradient for a deeper internal ocean than what is considered in this study should be carefully discussed .
the condition for high - pressure ice to exist under the internal ocean is @xmath50 where @xmath51 is the pressure at the h@xmath0o - rock boundary and @xmath52 is the pressure on the phase diagram where the temperature gradient and the high - pressure melting line cross ( fig .
3c ) . as a representative value
, we assume that high - pressure ice has a density of 1.2 g/@xmath53 . because the characteristic features of high - pressure ice are poorly understood , we simplified the model ; and , in particular , the thermal conductivities of high - pressure ice ( see below ) .
when thermal conductivity of high - pressure ice is relatively high and the temperature gradient in the high - pressure - ice part of the phase diagram is less than the gradient of the melting lines , the layer of high - pressure ice continues to the h@xmath0o - rock boundary [ ( i ) in fig .
however , when the thermal conductivity is comparatively low and the temperature gradient is greater than that of the melting lines , the temperature gradient joins the melting line and goes along with the melting lines to the h@xmath0o - rock boundary [ dashed arrows ( ii ) and ( iii ) in fig .
although little is known about the conditions on the dashed arrows , we assume that layer to be high - pressure ice in this study . therefore , from the point where the temperature - pressure line crosses into the high - pressure - ice part , the high - pressure layer continues to the h@xmath0o - rock boundary .
figures 4a and 4b show the surface conditions for planets with masses from @xmath54 to @xmath55 at 4.6 billion years after planetary formation , with varying h@xmath0o masses on the surface , with initial radiogenic heat sources , and at 1au from our sun .
we assumed @xmath56 for fig . 4a and @xmath57 for fig . 4b .
because larger planets have larger geothermal heat flux and thicker h@xmath0o layers , they could have an internal ocean with less h@xmath0o mass on the planetary surface ( fig .
4a ) and a weaker initial radiogenic heat source ( fig .
4b ) . however , larger planets also have larger gravitational acceleration .
thus , on those planets , high - pressure ice tends to appear under the internal ocean with smaller h@xmath0o mass on the surface ( fig .
4a ) . for example , if a planet of mass of @xmath1 has an h@xmath58o mass of @xmath59 to @xmath60 , it could have an internal ocean .
however , if a planet has an h@xmath58o mass @xmath61 , high - pressure ice should exist under the ocean ( fig .
4a ) . note , however , that an internal ocean can exist on a planet having a mass of @xmath1 if the initial radiogenic heat source exceeds @xmath62 ( fig .
figures 4c and 4d give the temperature profiles of surface - h@xmath0o layers .
figure 4c gives the conditions of a planet parameterized by @xmath1 and @xmath63 , whereas fig .
4d is for @xmath1 and @xmath64 .
given the conditions of fig .
4c , the surface h@xmath0o layers consist of a conductive - ice - i layer and a convective - liquid layer ( i.e. , an internal ocean ) .
when the planet has a greater h@xmath0o mass , high - pressure ice could appear under the internal ocean ( fig .
4d ) . here
, the planetary surface temperature seems very low , so we assume that the planet is covered by ice that has a higher albedo ( @xmath270.6 ) than ocean / land ( @xmath270.3 ) . figures 5a and 5b show the surface conditions for free - floating planets ( @xmath65 ) with masses from @xmath54 to @xmath55 at 4.6 billion years after planetary formation .
the incident flux from the central star affects the surface temperature , thereby affecting the condition on the surface .
therefore , the conditions , and in particular those shown in fig . 5a , are different from those shown in figs . 4a and 4b .
the results of fig .
5a show that , regardless of the amount of h@xmath0o a @xmath1 planet has , an internal ocean can not exist under the ice shell .
an internal ocean could exist on free - floating planets under certain conditions , but the planetary size and water abundance strongly constrain these conditions ( see fig .
5a ) . for instance , if a free - floating planet has an initial radiogenic heat source greater than @xmath66 , it can have an internal ocean ( fig .
figures 5c and 5d give the temperature gradient of surface - h@xmath0o layers for free - floating planets .
the parameters are set to the same values as those for figs . 4c and 4b , except that @xmath65 .
the temperature on the planetary surface , approximately 35 k , is calculated by eq .
( 6 ) on the assumption that @xmath65 , and the ice - i layer is thicker than that for the conditions of figs . 4c and 4d .
given the conditions of fig .
5c , the surface - h@xmath0o layers consist only of a conductive - ice - i layer
. however , for grater h@xmath0o mass , high - pressure ice could appear under the ice - i layer ( fig .
figure 6 shows the surface conditions for planets with masses from @xmath54 to @xmath55 at varying distances from a central star , and at 4.6 billion years after planetary formation .
the runaway greenhouse limit @xcite indicating the inner edge of the hz is not considered .
the effect of the incident flux from the central star on the surface conditions is estimated in each graph of fig .
6 as a function of distance from the central star .
we find that the existence of an internal ocean on planets far from a central star depends on the planetary mass and surface - h@xmath0o mass . for the conditions of fig .
6b on which a planet has two times the h@xmath0o mass ( i.e. , 2@xmath67 ) , for example , a @xmath68 planet can have an internal ocean under which there is no high - pressure ice only out to approximately 7 au from the central star . for a planet with five times
the h@xmath0o mass ( i.e. , 5@xmath67 ; fig .
6c ) , an internal ocean with no underlying high - pressure ice can exist out to approximately 30 au .
however , if the planet has ten times the h@xmath0o mass ( i.e. , 10@xmath67 ; fig .
6d ) , an internal ocean without underlying high - pressure ice could exist to only approximately 5 au .
the planetary mass and surface - h@xmath0o mass strongly constrain the conditions under which an extrasolar terrestrial planet far from its central star can have an internal ocean with no underlying high - pressure ice .
we expanded on the models of tajika ( 2008 ) by ( 1 ) invoking the mass - radius relationship [ eq . ( 1 ) ] to consider planetary compression by gravity , ( 2 ) considering the temperature dependence of the thermal conductivity of ice ih [ eq .
( 4 ) ] , and varying the ( 3 ) abundance of radiogenic heat sources and ( 4 ) h@xmath0o abundance on the surface , both of which were held constant in tajika ( 2008 ) .
we used eq .
( 1 ) to determine the mass - radius relationships because we expected more accurate results , although this change makes little quantitative difference in the results .
below , we discuss items ( 2 ) , ( 3 ) , and ( 4 ) to analyze the models of this study . it is known from experiment that the thermal conductivity of ice ih depends on temperature [ @xmath69 @xcite ] . in the present study ,
the surface - ice shell is thicker than that considered by tajika ( 2008 ) .
for example , the surface - ice shell considered herein is about 1.1 times thicker at 1 au and about 3.5 to 4.4 times thicker on the free - floating planets ( @xmath65 ) .
this increased thickness is due to the use of eq .
( 4 ) to describe the thermal conductivity of ice ih in contrast with tajika ( 2008 ) , where a constant thermal conductivity ( @xmath70 ) was used .
tajika ( 2008 ) treated the abundance of h@xmath0o and heat source elements ( a ratio of mass against a planetary mass ) to be constant , i.e. , @xmath57 ( @xmath19 @xmath9 @xmath20 ) and @xmath56 , although the amounts should vary with a planetary mass , and showed that liquid water would be stable either on the surface or beneath ice for a planet with a mass exceeding @xmath71 , regardless of planetary orbit and luminosity of the central star . in this paper , however , when we consider a planet with parameters @xmath57 and @xmath56 , the results shown in fig .
5 are not consistent with those of tajika ( 2008 ) .
our results indicate that free - floating planets with masses between @xmath54 and @xmath55 could not have liquid water on the planetary surface and also under the ice for planetary parameters of @xmath57 and @xmath56 . as just explained , when using eq .
( 4 ) to describe the thermal conductivity of ice ih , the planetary surface temperature becomes more sensitive to the thickness of the surface - ice layer , which leads to a several - times increase in the thickness of the surface ice .
these results thus differ from those of tajika ( 2008 ) . in this study , we assume that the high - pressure layer continues to the h@xmath0o - rock boundary from the point where the temperature - pressure line crosses into the high - pressure - ice part ( see section 2.2 ) .
the results of this paper could depend on this assumption .
when the thermal conductivity of high - pressure ice is comparatively low ( similar to that of ice ih ) , the heat can not be transported through the high - pressure ice effectively , thus the bottom of the high - pressure ice could be thermally unstable and be melted enough to form internal ocean . the result of the present study indicate that the high - pressure layers under the internal ocean could be from @xmath721 km to @xmath27100 km thick because a super - earth planet with a few wt.% h@xmath0o could have @xmath27100 km - thick high - pressure ice layers . in high - pressure ice that is @xmath27100 km thick ,
it is possible that convective ice layers appear . to estimate whether or not such convection layers arise
, we use the rayleigh number @xmath73 , which is given by @xmath74 where @xmath75 is the coefficient of thermal expansion , @xmath76 is the density , @xmath77 is the thickness of the high - pressure - ice layer , @xmath78 is the temperature difference between top and bottom part of the layer , @xmath79 is the thermal diffusivity , and @xmath80 is the viscosity coefficient . when the rayleigh number @xmath73 exceeds the critical value for the onset of convection ( @xmath81 @xmath82 ) , convective motion spontaneously begins .
if we use the parameters from kubo ( 2008 ) , i.e. , @xmath83 k@xmath84 , @xmath85 kgm@xmath86 , @xmath87 k , @xmath88 m@xmath89s@xmath84 , and @xmath90 pa s , and use the typical values of this study , @xmath91 ms@xmath39 and @xmath92 km , the rayleigh number is @xmath93 , which indicates that the convective motion is possible . in this study , we consider @xmath16 from 0.1@xmath19 ( 0.0023 wt.% ) to 100@xmath19 ( 2.3 wt.% )
. if the planet has significantly more h@xmath0o on the surface , the water layer would be very thick and our model would not apply to that planet . if the planet has significantly less h@xmath0o on the surface , the regassing flux of water would change because of insufficient water on the surface , and the planet s thermal evolution would differ from that of earth . in other words , we consider only those planets that are relatively active geothermally and have an adequate amount of water as earth - like planets . improving our model so that it applies to other types of terrestrial planets
is an important problem that we leave for future work . for genesis and sustenance of life , we need at least ( 1 ) liquid water and ( 2 ) nutrient salts because these substances are required to synthesize the body of life ( maruyama et al
because nutrient salts are supplied from rocks , it is necessary that liquid water should be in contact with rock to liberate the salts .
a type-5 planet ( fig .
3b ) is thus not likely to be habitable because the internal ocean does not come in contract with rocks .
however , it is possible for a type-2 planet to meet this requirement .
therefore , we presume that only type-2 planets have an internal ocean that is possibly habitable . therefore , the results of this study indicate that planetary mass and h@xmath0o mass constitute two more conditions to add to the previous conditions for an extrasolar planet to have an internal ocean without high - pressure ice . in other words , these considerations indicate that only a planet with the appropriate planetary mass and h@xmath0o mass can have an internal ocean that is possibly habitable .
however , it is possible that hydrothermal activities within the rocky crust may transfer nutrient salts to the internal ocean through cracks in the high - pressure ice .
large terrestrial planets such as those considered in this study are likely to have areas of high geothermal activity along mid - oceanic ridges and subduction zones as well as large submarine volcanos whose tops might emerge into the internal ocean from the high - pressure ice . furthermore , for planets with thick high - pressure - ice layers ( section 4.1 ) , the convective ice layer could transfer nutrient salts from the rock to the internal ocean . in these cases ,
high - pressure ice might not prevent nutrient flux from the rock floor from reaching the internal ocean . as shown in figs . 4b and 5b
, an appropriate initial radiogenic heat source is an important factor in determining whether or not a planet has an internal ocean .
because the variation in the amount of initial heat generation in planets throughout space is not known , we assume in this study that the initial heat generation @xmath47 per unit time and volume ranges from @xmath94 to @xmath95 . thus , in order to resolve this issue , the general amount of radiogenic heat sources for extrasolar terrestrial planets should be estimated .
an earth - size planet ( @xmath57 , @xmath56 ) orbiting at 1au around the sun for 4.6 billion years with no greenhouse gases might be globally covered by ice ( fig .
4 ) . in contrast
, earth currently is partially covered with ice but retains liquid water on its surface .
therefore , it is almost certain that greenhouse gases play a major role in keeping the surface warm .
even free - floating , earth - sized planets with atmospheres rich in molecular hydrogen could have liquid water on the surface because geothermal heat from the interior would be retained by the greenhouse - gas effect of h@xmath0 @xcite . by applying this model here ,
we can account for the greenhouse - gas effect for various values of emissivity @xmath37 in eq.(6 ) . in future presentations
, we will thus discuss how greenhouse gases modify the conditions necessary for the development of a terrestrial ocean planet or ice - covered planet with an internal ocean . in the present study
, we assumed pure h@xmath0o on the planetary surface .
however , even if an extrasolar terrestrial planet has surface h@xmath0o , the h@xmath0o might not be pure because it might contain dissolved nutrient salts such as those found in earth s oceans . in this case , the melting point of the solution would differ from that of pure h@xmath0o .
moreover , the phase diagram would become more complicated , and the properties of h@xmath0o ( e.g. , thermal conductivity ) could be transmuted as a result of the appearance of phases in which h@xmath0o ice contains nutrient salts .
a very important work would thus be to analyze multicomponent water ( e.g. , seawater on earth ) to see what qualitative changes such a modified phase diagram would bring to our results .
consider the examples , europa and ganymede , which are the two satellites of jupiter and are thought to have internal oceans .
ganymede is thought to have high - pressure ice under its internal ocean ( see , e.g. , lupo 1982 ) .
therefore , determining whether the internal oceans of europa and ganymede are suitable for life would be pertinent to the discussion of the habitability of internal oceans with or without high - pressure ice .
in addition , a more circumstantial discussion of the habitability of internal oceans would require considering the redox gradient within the internal ocean [ as exemplified by the discussions of europa ( gaidos et al . 1999 ) ] , and the effects of the riverine flux of nutrient salts ( maruyama et al .
herein , we discuss the conditions that must be satisfied for ice - covered bound and unbound terrestrial planets to have an internal ocean on the timescale of planetary evolution .
geothermal heat flow from the planetary interior is considered as the heat source at the origin of the internal ocean . by applying and improving the model of @xcite
, we also examine how the amount of radiogenic heat and h@xmath0o mass affect these conditions .
moreover , we investigate the structures of surface - h@xmath0o layers of snowball planets by considering the effects of high - pressure ice .
the results indicate that planetary mass and surface - h@xmath0o mass strongly constrain the conditions under which an extrasolar terrestrial planet might have an internal ocean without a high - pressure ice existing under the internal ocean .
+ we thank the reviewer eiichi tajika for the constructive comments .
we also thank s. ida and m. ikoma for valuable discussions .
this research was supported by a grant for the global coe program `` from the earth to earths '' from the ministry of education , culture , sports , science and technology of japan .
t. s. was supported by a grant - in - aid for young scientists ( b ) , jsps kakenhi grant number 24740120 .
by applying conservation of energy , we obtain @xmath96 where @xmath76 is the density of the mantle , @xmath97 is the specific heat at constant pressure , @xmath98 and @xmath99 are the outer and inner radii of the mantle , respectively , @xmath100 is the average mantle temperature , and @xmath101 is the rate of energy production by decay of radiogenic heat sources in the mantle per unit volume . the mantle heat flux is parameterized in terms of the rayleigh number @xmath73 as @xmath102 where @xmath103 is the thermal conductivity of the mantle , @xmath104 is the temperature at the surface of the mantle , @xmath81 is the critical value of @xmath73 for convection onset , and @xmath105 is an empirical constant .
the radiogenic heat source is parameterized as @xmath106 where @xmath107 is the decay constant of the radiogenic heat source and @xmath47 is the initial heat generation per unit time and volume .
we assume @xmath47 is 0.1@xmath48 to 10@xmath48 , where the constant @xmath48 is the initial heat generation estimated from the present heat flux of earth .
we obtain @xmath73 as @xmath108 where @xmath75 is the coefficient of thermal expansion , @xmath79 is the thermal diffusivity , and @xmath109 is the water - dependent kinematic viscosity .
the viscosity @xmath109 strongly depends on the evolution of the mass @xmath110 of mantle water and the mantle temperature @xmath100 and is parameterized as @xmath111 @xmath112 @xmath113 where @xmath114 , @xmath115 , and @xmath116 are constants , @xmath117 is the activation temperature for solid - state creep , and @xmath118 is the mass of the mantle .
the evolution of the mantle water can be described by the regassing flux @xmath119 and outgassing flux @xmath120 as @xmath121 where @xmath122 is the water content in the basalt layer , @xmath123 is the average density , @xmath124 is the average thickness of the basalt layer before subduction , @xmath125 is the areal spreading rate , @xmath126 is the regassing ratio of water , @xmath127 is the melting generation depth , and @xmath128 is the outgassing fraction of water
. the regassing ratio of water linearly depends on the mean mantle temperature @xmath100 via @xmath129 as given by von bloh et al .
( 2008 ) , where @xmath130 is the temperature dependence of the regassing ratio , and @xmath131 is the initial regassing ratio .
the areal spreading rate @xmath125 is @xmath132 where @xmath133 is the ocean - basin area . the area @xmath133 can be parameterized as @xmath134 where @xmath135 is the present ocean - basin area on earth . by using eq .
( 1 ) , we obtain @xmath133 as @xmath136 30 von blow , w. et al .
2008 , , 476 , 1365 .
budyko , m. i. 1969 , tellus , 21 , 611 .
dillard , d. s. , & timmerhaus , k. d. 1965 , bull .
froid annexe , part 2 , 35 .
franck , s. , & bounama , c. 1995 , phys .
earth planet .
inter . , 92 , 57 .
gaidos , e. j. , nealson , k. h. , & kirschvink , j. l. 1999 , science , 284 , 1631 .
genda , h. , & ikoma , m. 2008 , icarus , 194 , 42 .
gough , d. o. 1981 , , 74 , 21 .
hart , m. h. 1979 , , 37 , 351 .
hoffman , p. f. , & schrag , d. p. 2002
, terra nova , 14 , 129 .
hoffman , p. f. et al .
1998 , science , 281 , 1342 .
howard , a. w. et al .
2010 , science , 330 , 653 .
jackson , m. j. , & pollack , h. n. 1984 , j. geophys .
res . , 89 , 10103 .
kamb , b. 1973 , physics and chemistry of ice,28 .
kasting , j. f. 1988 , , 74 , 472 .
kasting , j. f. et al .
1993 , , 101 , 108 .
klinger , j. 1975 , j. glaciol .
, 14 , 517 .
klinger , j. 1980 , science , 209 , 11 .
kubo , t. 2008 , low temperature science , 66 , 123 .
lupo , m. j. 1982 , , 52 , 40 .
maruyama , s. et al .
2013 , geoscience frontiers , 4 , 141 .
mayor , m. , & queloz , d. 1995 , , 378 , 355 .
mcgovern , p. j. , & schubert , g. 1989 , earth planet .
lett . , 96 , 27 .
mischna , m. a. , kasting , j. f. , pavlov , a. , & freedman , r. 2000 , , 145 , 546 .
nakajima , s. , hayashi , y .- y . , & abe , y. 1992 , j. atmos .
, 49 , 2256 . sellers , w. d. 1969 , j. appl .
meteorol . ,
stevenson , d. j. 1999 , , 400 , 32 .
sumi , t. et al .
2011 , , 473 , 349 .
tajika , e. 2008 , , 680 , l53 .
tajika , e. , & matsui , t. 1992 , earth planet .
, 113 , 251 .
valencia , d. , oconnell , r. j. , & sasselov , d. 2006 , , 181 , 545 .
varrot , m. , rochas , g. , & klinger , j. 1978 , j. glaciol . , 21 , 241 . | many extrasolar ( bound ) terrestrial planets and free - floating ( unbound ) planets have been discovered .
the existence of bound and unbound terrestrial planets with liquid water is an important question , and of particular importance is the question of their habitability . even for a globally ice - covered planet , geothermal heat from the planetary interior may melt the interior ice , creating an internal ocean covered by an ice shell . in this paper
, we discuss the conditions that terrestrial planets must satisfy for such an internal ocean to exist on the timescale of planetary evolution .
the question is addressed in terms of planetary mass , distance from a central star , water abundance , and abundance of radiogenic heat sources .
in addition , we investigate the structures of the surface - h@xmath0o layers of ice - covered planets by considering the effects of ice under high pressure ( high - pressure ice ) . as a fiducial case , @xmath1 planet at 1 au from its central star and with 0.6 to 25 times
the h@xmath0o mass of earth could have an internal ocean .
we find that high - pressure ice layers may appear between the internal ocean and the rock portion on a planet with an h@xmath0o mass over 25 times that of earth .
the planetary mass and abundance of surface water strongly restrict the conditions under which an extrasolar terrestrial planet may have an internal ocean with no high - pressure ice under the ocean .
such high - pressure - ice layers underlying the internal ocean are likely to affect the habitability of the planet . |
ss 433 is a peculiar binary system , consisting of a black hole ( as proposed by lopez et al . , 2005 ) and a massive companion .
this system is accreting at a super - eddington rate , and is expelling two - sided relativistic jets at a velocity of 0.26c .
these jets precess in a cone of half - opening angle of 20@xmath0 @xcite .
ss 433 is near the center of w50 , a large 2@xmath11@xmath0 nebula stretched in the east - west direction , and catalogued as an snr @xcite.the ss 433/w50 system is the only galactic object known of its kind , giving rise to a unique laboratory to study the association between snrs and black holes as well as the interaction between relativistic jets and the surrounding medium .
this system has been studied extensively in radio continuum and hi @xcite , millimetre wavelengths @xcite , and in x - rays with and ( * ? ? ? * and references therein ) and with _ rxte _ ( safi - harb & kotani , 2002 , safi - harb & petre , 1999 ) . from this multi - wavelength study , it was concluded that the morphology and energetics of w50 are consistent with the picture of the jets interacting with an inhomogeneous medium and likely hitting a denser cloud in the west .
the observation presented here provides the highest resolution x - ray image obtained to date of the bright region of the western lobe of w50 .
this region was chosen because it coincides with ir emission and can probe the jet - cloud interaction site .
we performed a spatially resolved spectroscopy of this region to primarily determine the nature of the emission and correlate the x - ray emission with radio and ir observations .
the paper is organized as follows . in 2 , we summarize the observation imaging and spectral results and compare them to the and data . in 3 , we study the x - ray emission in correlation with the infrared and radio emission , and finally present our conclusions in 4 .
the western lobe of w50 was observed with the acis - i chips on board on 2003 august 21 at a ccd temperature of -120@xmath0c .
the charge transfer inefficiency was corrected using the apply_cti tool on the level 1 raw data .
a new level 2 file was then obtained using the standard ciao 3.0 routines .
the final exposure time was 71 ksec . to illustrate the w50 region covered by , we show in fig .
1 the the radio image of w50 ( grey scale ) , and the regions covered by observations in infrared ( large box ) and x - ray ( small box ) . the projection on the sky of the precession cone axes of the ss 433 jets is also overlayed .
the radio image shows that the eastern wing of w50 exhibits a corkscrew pattern , which mimics the precession of the arcseconds - scale jets from ss 433 ( dubner et al . , 1998 , hjellming & johnston , 1981 ) .
interestingly , there is a hint of a corkscrew pattern visible in the chandra image ( fig . 2 and 3 ) , supporting the conclusion that the ss 433 subarcsecond - scale relativistic jets are affecting the large scale radio and x - ray emission from w50 . in fig .
2 , we show the energy image in which red corresponds to the soft energy band ( 0.3 - 2.4 kev ) and blue corresponds to the hard energy band ( 2.4 - 10 kev ) . in fig .
3 , we show the intensity image in the 0.3 - 10 kev energy range .
we resolve many point sources in the field ( a list of which will be provided elsewhere ) and note the knotty structure of the nebula .
the x - ray emission peaks at @xmath2 ( j2000 ) = 19@xmath3 09@xmath4 [email protected] , @xmath6 ( j2000 ) = 05@xmath0 03@xmath7 [email protected] .
( measured from n to e ) , and half - opening angle of 20@xmath0,width=384 ] .
+ to perform spatially resolved spectroscopy of the remnant , we excluded the point sources in the field , and extracted spectra from the diffuse emission for 11 regions shown in fig .
the w2 and irknot2 regions correspond to the x - ray w2 region presented in @xcite and the infrared knot2 region presented by @xcite , respectively .
these regions will be the focus of this paper and are selected in order to compare the results with those found in x - rays with and and in infrared with _
the proximity of the western lobe to the galactic plane complicates the spectral analysis because of contamination by the galactic ridge . to minimize this contamination
, we extracted several background regions from source - free regions around the diffuse emission from w50 and from the same acis chip .
we subsequently determined the spectral parameters using the resulting average background .
spectra were extracted in the 0.5 - 10.0 kev range .
the background subtracted count rate for the w2 and irknot2 regions are @xmath9 counts s@xmath10 and @xmath11 counts s@xmath10 respectively .
to determine whether the emission is thermal or not , we fitted the spectra with thermal bremsstrahlung and power - law models ( following * ? ? ? * ) .
the bremsstrahlung model is characterized by the shock temperature , @xmath12 , and the power - law model is characterized by the photon index , @xmath13 .
[ h ] = [email protected] . * fig .
3 . ( right ) : 0.3 - 10 kev image of w50 showing regions used for spectroscopy ( see 2.2 ) . the dots hint to a corkscrew pattern.*,title="fig:",width=528 ] * * 3 for details.,width=384 ]
both models give adequate fits in each region .
however , we find that the power - law models give slightly lower reduced @xmath14 values , and that the temperatures derived from the thermal bremsstrahlung models are high ( unrealistically high for even the youngest snrs ) .
this , together with the absence of line emission in the spectra , leads us to favor the non - thermal interpretation for the x - ray emission .
table [ tab1 ] summarizes the results for the w2 region in comparison to the and results .
a distance of 3 kpc ( scaled by @xmath15 ) is used in the luminosity calculations ( as in * ? ? ?
* ) , and the errors are at the 90% confidence level .
the spectroscopic results of the other regions are beyond the scope of this paper and will be presented elsewhere ; we note here that the spectrum softens with increasing distance from ss 433 except for regions 3 , 6 and 8 .
this is consistent with the energy - color image shown in fig 2 .
.thermal bremsstrahlung and power - law model results for the w2 region [ cols="<,^,^,^,^ " , ] we subsequently use the power - law fits to derive the synchrotron emission parameters .
assuming equipartition between particles and fields and integrating from radio to x - ray frequencies , the equipartition magnetic field ( @xmath16 ) , the magnetic energy density ( @xmath17 ) , the total synchrotron electron energy ( @xmath18 ) and the lifetime of the electrons ( @xmath19 ) can be determined . for the w2 region ,
we derive @xmath20 g , @xmath21 erg @xmath22 , @xmath23 ergs , and @xmath24 years .
the range of values corresponds to a=1 - 100 ( the ratio of baryon energy to electron energy ) .
the derived values of the synchrotron parameters as well as @xmath25 ( table [ tab1 ] ) agree with those found using and for the w2 region , within error . however , the spectra appear harder with the observation .
to probe the interaction between the western jet of ss 433 and the ambient medium , we study the x - ray emission in correlation with radio continuum and hi data obtained with the nrao vla and green - bank radio telescope and infrared data obtained with _ isocam_. fig .
1 shows the radio , infrared , and x - ray regions and fig .
4 shows the x - ray emission with the infrared contours .
the average value of @xmath25 found on the basis of the hi observations is @xmath26 @xmath27 .
this is slightly lower than the average found using the data , which is to be expected .
the energetics in the western lobe found with the x - ray data can then be compared to that found in @xcite .
we found the total synchrotron electron energy in x - rays to be @xmath28 ergs , which is in good agreement with the energy found from radio observations .
as seen in fig .
4 , there is no correlation between the infrared emission and the peak of x - ray emission .
this , along with the high value of @xmath25 ( n@xmath29@xmath302@xmath3110@xmath32 @xmath27 ) derived for the irknot2 region , suggests that the infrared emission is not associated with the western lobe of w50 .
the derived value of @xmath33 kev is higher than the expected temperatures for snrs , indicating that the x - ray emission in irknot2 is non - thermal .
we favor a non - thermal interpretation for the emission of the western lobe of w50 .
the derived values of @xmath25 , equipartition magnetic field , synchrotron electron energies and lifetimes agree with those derived previously with and _ asca . _ the infrared emission is not correlated with the peak of x - ray emission .
this , in addition to the high value of @xmath25 derived for this region , suggests that the infrared emission is not originating from w50 , and could be associated with a star forming region ( work in progress ) .
the corkscrew pattern seen in both the radio and x - ray images provides strong support to the hypothesis that the relativistic jets from ss 433 are causing the morphology of the w50 nebula .
s. safi - harb acknowledges support by an nserc ufa fellowship and an nserc discovery grant ( canada ) .
y. fuchs is supported by a cnes ( france ) fellowship .
g. dubner is a member of the carrera del investigador cientfico , conicet ( argentina ) . | w50 remains the only supernova remnant ( snr ) confirmed to harbor a microquasar : the powerful enigmatic source ss 433 .
our past study of this fascinating snr revealed two x - ray lobes distorting the radio shell as well as non - thermal x - rays at the site of interaction between the ss 433 eastern jet and the eastern lobe of w50 . in this paper
we present the results of a 75 ksec acis - i observation of the peak of w50-west targeted to 1 ) determine the nature of the x - ray emission and 2 ) correlate the x - ray emission with that in the radio and infrared domains .
we have confirmed that at the site of interaction between the western jet of ss 433 and dense interstellar gas the x - ray emission is non - thermal in nature . the helical pattern observed in radio is also seen with .
no correlation was found between the infrared and x - ray emission . , ,
, binaries : close , ism : individual ( w50 ) , x - rays : stars , stars : individual ( ss 433 ) , radiation mechanisms : non - thermal , supernova remnants |
in three spatial dimensions , in the unitary limit of a quantum bose or fermi gas with point - like interactions , the scattering length diverges .
`` unitary '' here refers to the limit on the cross section imposed by unitarity .
these systems provide intriguing examples of interacting , scaling invariant theories with dynamical exponent @xmath0 , i.e. non - relativistic .
the infinite scattering length occurs at a fixed point of the renormalization group in the zero temperature theory , thus the models are quantum critical .
the only energy scales in the problem are the temperature and chemical potential , and thermodynamic properties are expected to reveal universal behavior .
they can be realized experimentally by tuning the scattering length to @xmath8 using a feshbach resonance .
( see for instance @xcite and references therein . )
they are also thought to occur at the surface of neutron stars .
the systems have attracted much theoretical interest , and remain challenging problems due to the lack of small parameter for a perturbative expansion , such as @xmath9 or @xmath10 where @xmath11 is the scattering length , and @xmath12 the range of the 2-body potential .
early works were done by leggett and nozires and schmitt - rink@xcite .
the universal scaling behavior was studied in @xcite . in 3 dimensions ,
this is the physics of the bcs / bec crossover : since the fixed point occurs for an attractive coupling , the fermions may form a bosonic bound state which can subsequently undergo bec .
this cross - over was studied analytically in @xcite by considering a model of 2-component fermions coupled to the bosonic bound state .
monte - carlo studies were performed in @xcite .
the models can be studied in spatial dimension @xmath13@xcite and an epsilon expansion carried out@xcite .
there has also been some attempts to apply the ads / cft correspondence to these non - relativistic systems@xcite . in the present work ,
we describe a new analytic approach to studying the unitary limit based on our treatment of quantum gases in @xcite , which appears to be well suited to the problem since it allows an expansion in the inverse coupling .
let us motivate this approach as follows .
in one spatial dimension , the fixed point occurs for repulsive interactions .
the model is integrable@xcite and its thermodynamics determined exactly by the so - called thermodynamic bethe ansatz ( tba)@xcite . in the tba , the free energy is expressed in terms of a pseudo - energy which is a solution to an integral equation with a kernel related to the logarithm of the s - matrix . in the unitary limit the coupling goes to @xmath14 and the s - matrix @xmath1 .
the tba is then identical to a free gas of fermionic particles .
the formalism developed in @xcite was modeled after the tba : the free energy is expressed as a sum of diagrams where the vertices represent matrix elements of the logarithm of the ( zero temperature ) s - matrix .
however since generally the n - body s - matrix does not factorize into 2-body scattering , the series can not be summed exactly as in the tba . nevertheless , a consistent resummation of an infinite number of diagrams involving only 2-body scattering , the so - called foam diagrams , can serve as a useful approximation if the gas is not too dense .
the result of summing these diagrams leads to an integral equation for a pseudo - energy , as in the tba ; in fact in 1 spatial dimension the tba is recovered to lowest order in the kernel . since the formalism is based on the s - matrix , it can be very useful for studying the unitary limit where @xmath1 .
in this paper we present the main formulas for the 3-dimensional case , however we mainly analyze the 2-dimensional case ; analysis of the 3d case will be published separately@xcite .
phase transitions in two - dimensional fermi gases were studied in e.g. @xcite .
the fixed point separating the bec and bcs regimes goes to zero coupling when @xmath15 , thus it is not obvious whether a unitary limit exists at strong coupling .
as we will argue , there is a scale - invariant limit at infinite coupling @xmath16 where the s - matrix @xmath1 .
this is a meaningful unitary limit at very low ( @xmath17 ) or very high ( @xmath18 ) energy , although it does not correspond to a fixed point in the usual sense of a zero of the beta function .
the scattering length indeed diverges in this limit .
the possibility of this kind of unitary limit in two dimensions has not been considered before in the literature . in the next section
we describe the unitary limit in 1,2 and 3 dimensions and its relation to the renormalization group . in section iii
, we define the interesting scaling functions for the free energy and single - particle energies by normalizing with respect to free theories . in section
iv we describe the unitary limit of the formalism in @xcite in both two and three dimensions , where the integral equation becomes scale invariant .
the @xmath15 case is especially simple since the kernel reduces to a constant and the integral equation becomes algebraic .
analysis of these equations in @xmath19 is carried out for both infinitely repulsive or attractive fermions and bosons in sections v - viii .
the extension of our formalism to multiple species of possibly mixed bosonic and fermionic particles is considered in section ix .
kovtun et .
conjectured that there is a universal lower bound to the ratio of the shear viscosity to the entropy density , [ boundzero ] /s where @xmath20 is boltzmann s constant@xcite .
this was based on the ads / cft correspondence for _ relativistic _ theories in 3 spatial dimensions , and the bound is saturated for certain supersymmetric gauge theories .
counterexamples to the @xmath3 bound were suggested to be non - relativistic@xcite , however no known fluid violates the bound .
it thus interesting to study this ratio for non - relativistic theories , and in particular for @xmath19 theories where no conjecture have been put forward .
it has also been suggested that the gases in the unitary limit may represent the most perfect fluid , i.e. with the lowest value of @xmath3 . for the spacetimes
considered thus far for a non - relativistic ads / cft correspondence in 2d , the result found is that @xmath3 is exactly @xmath21@xcite .
we analyze @xmath3 for the attractive fermionic case as a function of @xmath22 in section vi .
if one disregards unphysical , potentially metastable regions , one finds @xmath23 .
the other cases of repulsive bosons or fermions also satisfy the bound
. however the attractive boson is below it : @xmath24 .
the models considered in this paper are the simplest models of non - relativistic bosons and fermions with quartic interactions .
the bosonic model is defined by the action for a complex scalar field @xmath25 .
[ bosonaction ] s = d^d dt @xmath26 ( throughout this paper , @xmath27 refers to the number of spatial dimensions ) . for fermions , due to the fermionic statistics , one needs at least a 2-component field @xmath28 : [ fermionaction ] s = d^d dt @xmath29 in both cases , positive @xmath30 corresponds to repulsive interactions .
the bosonic theory only has a @xmath31 symmetry .
the fermionic theory on the other hand has the much larger so(5 ) symmetry .
this is evident from the work@xcite which considered a relativistic version , since the same arguments apply to a non - relativistic kinetic term .
this is also clear from the work@xcite which considered an @xmath32-component version with sp(2n ) symmetry , and noting that sp(4 ) = so(5 ) .
the free versions of the above theories have a scale invariance with dynamical exponent @xmath0 , i.e. are invariant under @xmath33 at a renormalization group fixed point , i.e. quantum critical point , the model is expected to have the same scale invariance .
it is natural to define scaling dimensions @xmath34 in units of inverse length or wave - vector @xmath35 , i.e. @xmath36 , @xmath37 , and @xmath38 .
requiring the action to have zero scaling dimension gives @xmath39 and @xmath40 .
the interaction is thus relevant for @xmath41 .
the renormalization group behavior can be inferred from the coupling constant dependence of the s - matrix , or equivalently the kernels in section iv . for completeness
we compute the beta function for arbitary @xmath27 in appendix a using conventional methods .
consider first a single boson in @xmath42 dimensions .
the kernel in eq .
( [ g3d ] ) depends on the renormalized coupling @xmath43 given in eq .
( [ gr ] ) , where @xmath44 is a high momentum cutoff .
the kernel @xmath45 is related to the logarithm of the 2-body s - matrix as in eq .
( [ g2_def ] ) , and from this we can deduce the s - matrix function : [ 3ds ] s(|- | ) = galilean invariance , for equal mass particles , implies @xmath46 is only a function of the relative momenta @xmath47 of the two in - coming particles .
unitarity of the s - matrix amounts to @xmath48 .
defining @xmath49 , and requiring @xmath43 to be independent of @xmath44 gives the beta - function : [ betafun ] = - - ^2 where @xmath50 is the logarithm of a length scale .
the above result agrees with the calculation in appendix a. with the normalization of @xmath30 in the fermionic case as given in ( [ fermionaction ] ) , the resulting beta function is the same as above . as discussed in @xcite , the beta - function in @xmath27 dimensions ( see appendix a ) has the following implications
. see figure [ rg ] . for @xmath41
, there is an infra - red stable fixed point at the positive value [ ghatstar ] _ * = ( 2-d ) ^d/2 ( d/2 ) 2^d / m corresponding to repulsive interactions . when @xmath51 , this fixed point is approached from @xmath52 when the cutoff @xmath44 is infinite . in the limit of @xmath53 the s - matrix , as given in eq .
( [ 1ds ] ) becomes @xmath1 , which is the unitary limit . in the 1-dimensional thermodynamic bethe ansatz ,
the thermodynamics is simply that of a free , one - component , fermionic gas , even though the original particle was a boson . for @xmath54 , there is an _ ultra - violet _ stable fixed point at the _ negative _ coupling @xmath55 , i.e. an attractive interaction .
a bosonic gas with attractive interactions can be unstable .
however for a fermionic gas , the attractive interaction can be balanced by the fermi pressure , and the gas can be stable against collapse .
we now describe the implications of the above fixed point on the scattering length .
consider a single boson .
a straightforward calculation of the differential cross section in @xmath27 spatial dimensions gives [ cross ] = | ( k ) |^2 where @xmath56 is the scattering amplitude . in the above formula
@xmath57 is the momentum of one of the particles in the center of mass frame , i.e. @xmath58 . the @xmath59 are an important ingredient of the kernels of the integral equations in section iv , and were computed to all orders in @xcite for instance . in @xmath42 , we equate @xmath60 .
this gives [ a3d ] a ( k ) = where @xmath43 is the renormalized coupling ( [ gr ] ) .
if @xmath61 is measured at very small momentum transfer @xmath62 , this leads to the definition of the scattering length [ a3d.2 ] a = = where @xmath44 is the ultra - violet cutoff . in order for @xmath11 to diverge , @xmath30 must be negative .
in fact , at precisely the fixed point @xmath63 , @xmath64 , depending on from which side @xmath65 is approached . when @xmath66 , i.e. just less than @xmath65 , then @xmath67 , whereas when @xmath68 , @xmath69 .
the case @xmath70 is on the bcs side of the crossover , whereas @xmath71 is on the bec side .
see figure [ rg ] .
[ rg ] the s - matrix ( [ 3ds ] ) has a pole at @xmath72 .
since physical bound states correspond to poles at @xmath73 , the bound state exists only on the bec side of the critical point , with energy [ ebound ] e_bound - state = - in the bec to bcs crossover from @xmath74 to @xmath75 the binding energy goes to zero and the bound state disappears .
nevertheless , the crossover is expected to be smooth , if on the bec side one includes the bound state in the thermodynamics .
the @xmath15 case is somewhat more subtle due to the marginality of the coupling @xmath30 .
the fixed point @xmath76 and the rg flows are depicted in figure [ rg ] .
the fixed point at @xmath76 is just the free field theory , thus there is no analog of the bec / bcs cross - over .
one can nevertheless formally define the unitary limit as @xmath1 , as in @xmath77 and @xmath78 . in this subsection
we explore this possibility and interpret it using the renormalization group .
as we ll see , this limit occurs at @xmath16 , and the scattering length diverges .
first begin with the beta - function in 2d , eq .
( [ a.4 ] ) : [ 2dfp.1 ] = let @xmath79 at some arbitrary scale @xmath80 .
integrating the beta function one finds : [ 2dfp.2 ] g ( ) = note that @xmath30 diverges at the scale : [ lambdastar ] _ * = _ 0 e^4/mg_0 this is the familiar landau pole . whereas in for example the relativistic @xmath81 theory in @xmath42 where the landau pole is unphysical due to higher order corrections , here
the beta function ( [ 2dfp.1 ] ) is exact@xcite , thus this divergence is physical .
there are two cases to consider : if @xmath82 is negative , @xmath83 , and thus @xmath17 occurs in the infra - red .
if @xmath82 is positive , @xmath84 , and @xmath85 occurs in the ultra - violet .
both cases are consistent with the flows depicted in figure [ rg ] .
in fact , the scale @xmath86 is an rg invariant : [ 2dfp.3 ] = 0 and is the natural coupling constant in this problem that physical quantities are expressed in terms of .
the s - matrix can be inferred from the kernel in eq .
( [ kernel.2d ] ) : s ( || ) & = & + & & [ 2ds ] + & = & where @xmath87 is the relative momentum of the two incoming particles . near the fixed point @xmath88 , @xmath89 ( @xmath90 ) and the s - matrix @xmath91 , consistent with the free theory .
the same applies to the other side of the fixed point @xmath92 where @xmath90 . on the other hand , consider the low energy limit of the attractive case where @xmath17 .
here @xmath93 , and this scale is effectively an infra - red cutoff .
thus at low energies @xmath94 and @xmath1 .
similar arguments apply to the high energy limit of the repulsive case @xmath18 .
it is clear that , unlike in @xmath78 , this definition of the unitary limit does not correspond to a renormalization group fixed point in the usual sense of a zero of the beta function and is somewhat engineered ; nevertheless it defines a scale - invariant theory .
let us turn now to the scattering length .
using the scattering amplitude @xmath59 computed in @xcite , the cross section for relative momenta @xmath95 is : [ 2dunit.2 ] ( k ) & = & + & = & near the scale @xmath96 : [ sigstar ] ( _ * ) = @xmath97 ^ 2 equating @xmath98 with a scattering length @xmath11 , as appropriate to @xmath19 , one sees that @xmath11 diverges in the unitary limit @xmath53 .
because of the logarithmic dependence on @xmath99 , the s - matrix does not have a pole .
however the denominator is zero when @xmath100 , which is a remanant of the bound state pole in @xmath78 .
the energy of this quasi - bound state is @xmath101 .
note that this quasi - bound state disappears as @xmath102 , since in this limit @xmath103 , and this is analagous to the situation in @xmath78 .
we thus expect that the `` attractive '' case of the unitary limit should be better behaved since this bound state disappears , in contrast to the repulsive case where @xmath104 as @xmath105 , and this will be born out of our subsequent analysis .
at a quantum critical point , the only length scales of the quantum gas are the thermal wavelength @xmath106 and the length scale @xmath107 set by the density @xmath108 .
equivalently , one can express physical properties in terms of the only two energy scales , the temperature @xmath109 and the chemical potential @xmath110 , since the density is a function of @xmath111 . in order to fix normalizations in a meaningful way , it is useful to consider the simplest theories with @xmath0 scale invariance : free , non - relativistic bosons and fermions .
we set @xmath112 , except in a few select formulas .
the free energy density is given by the well - known expression for each particle : = @xmath113 where @xmath114 , @xmath115 and @xmath116 corresponds to bosons , fermions respectively .
the integrals over wave - vectors can be expressed in terms of polylogarithms @xmath117 , where @xmath118 is a fugacity , using @xmath119 and the integrals _ 0^dx & = & ( ) _ ( z ) + [ liint ] _ 0^dx & = & - ( ) _ ( -z ) valid for @xmath120 .
the result , as expected , is proportional to @xmath121 : = - s t @xmath122^d/2 _ ( d+2)/2 ( s z ) there are two important limits to consider . for bose gases near bose - einstein condensation ,
physically the interesting limit is @xmath123 . since @xmath124 , where @xmath125 is riemann s zeta function , this leads us to define the scaling functions
@xmath126 : & = & - c_2 ( /t ) ( d=2 ) [ freec ] + & = & - c_3 ( /t ) ( d=3 ) where we have used @xmath127 . with the above normalizations , @xmath128 for a single free boson when @xmath129 .
the above formulas are well - defined for fermions at zero chemical potential .
using [ lifermion ] - _ ( -1 ) = @xmath130 ( ) one finds as @xmath131 : [ cfermion ] c_2 = , c_3 = 1 - ( free fermions ) it should be pointed out that the coefficients @xmath132 are analogous to the virasoro central charge for relativistic systems in @xmath51 , as discussed in @xcite .
the other interesting limit is @xmath133 , i.e. @xmath134 or @xmath135 , depending on the sign of the chemical potential . here
the scaling forms are naturally based on the zero temperature degenerate free fermion gas , where @xmath136 is the fermi energy .
in fact , the function @xmath117 has a branch cut along the real axis from @xmath137 to @xmath14 and the bosonic free energy is ill - defined at @xmath138 , in contrast with fermions . using the analytic continuation of the asymptotic behavior [ largez ] - _ ( -z ) as z from positive integer @xmath139 to half - integer values , we define the scaling functions @xmath140 as follows : & = & - b_2 ( t/ ) ( d=2 ) [ tzero ] + & = & - b_3 ( t/ ) ( d=3 ) the above normalizations are defined such that @xmath141 for a single free fermion .
one can verify that the above @xmath42 expression is the standard result for a zero temperature , single component , degenerate fermion gas where @xmath110 is the fermi energy .
other thermodynamic quantities follow from the free energy .
the pressure @xmath142 .
the density is @xmath143 and the entropy density is @xmath144 . using the scaling form in eq .
( [ tzero ] ) , one obtains for @xmath42 : [ density3d ] n & = & ( m ) ^3/2 @xmath145 + s & = & ( m ) ^3/2 b_3 ( t/ ) where @xmath146 is the derivative with respect to its argument @xmath147 .
( henceforth , @xmath148 will always refer to the derivative of the function @xmath30 with respect to it s argument @xmath2 or @xmath149 as defined above . )
the analogous formulas in 2 dimensions are : [ density2d ] n & = & @xmath150 + s & = & b_2 the energy density @xmath151 follows from the relation @xmath152 : [ ener ] = - it is interesting to note that the above result , which in terms of the pressure is simply @xmath153 , is a consequence of the mechanics of _ free _ gases .
this shows that this relation continues to hold for interacting gases at a quantum critical point , as pointed out by ho@xcite
. also of interest is the energy per particle @xmath154 : [ enpart.1 ] _ 1 = - consider first the limit @xmath133 , with @xmath110 positive . in this limit ,
if @xmath155 is a smooth function of @xmath149 as @xmath156 , then the @xmath146 terms in the density vanish and one simply obtains [ enpart.2 ] _ t / 0^+ _ 1 = which is the same result as for a free gas , where for fermions @xmath110 is equal to the fermi energy @xmath157 .
the above formula is usually not appropriate to the @xmath158 limit when @xmath110 is negative . for the interacting gas @xmath159 , so this leads us to define the scaling functions @xmath160 : [ enpart.3 ] _ d ( t/ ) = as @xmath161 , the functions @xmath162 should become universal constants , and for free fermions @xmath163 . the fermi energy @xmath157 can be defined based on its relation to density in the zero temperature free fermion gas . for bosons , one can formally use the same definition . for @xmath15 , @xmath164 , whereas for @xmath42 , @xmath165 .
this leads to the definitions : [ xidefs ] _ 2 & = & @xmath166 ^ 2 b_2 ( t/ ) + _ 3 & = & @xmath167 ^ 5/3 ( m ) ^5/2 b_3 ( t/ ) next consider the energy per particle in the limit @xmath168 . here
it is more appropriate to use the form in eq .
( [ freec ] ) , which gives [ enpart.3b ] _ 1 = t the expression for @xmath169 for free fermions in the limit @xmath170 in @xmath42 leads us now to define @xmath171 : [ enpart.4 ] _ 1 = @xmath172 t _ 3 ( /t ) with this normalization , for free fermions , as @xmath173 , @xmath174 . in two dimensions
, @xmath175 goes to zero for free bosons as @xmath168 since it is proportional to @xmath176 .
however @xmath169 is finite for fermions in this limit .
using @xmath177 , we define [ enpart.5 ] _ 1 = _ 2 ( /t ) ( d=2 ) with the above normalization , @xmath178 for free fermions .
henceforth we drop the subscripts @xmath179 indicating the spatial dimension on the functions @xmath180 etc .
, since in the sequel we will carry out the analysis of the @xmath15 case only .
in the formalism developed in @xcite , the contributions to the free energy can be expressed as vacuum diagrams where the vertices are matrix elements of the logarithm of the s - matrix , and the lines are the filling fractions
. there are vertices with @xmath181 lines representing @xmath32-body scattering for any @xmath32 .
the main result obtained in @xcite , and reviewed below , is the formula for the free energy density in a self - consistent resummation of the 2-body scattering `` foam diagrams '' .
the 2-body scattering can be computed to all orders in the coupling in a standard zero - temperature calculation .
clearly this result is still an approximation in that higher n - body scattering , and 2-body scattering contributions not of the foam diagram type , are neglected .
nevertheless , this approximation is expected to be valid when the gas is sufficiently dilute .
the main ingredients of the formalism are as follows .
define the filling fractions as a function of a pseudo - energy @xmath182 [ fill ] f ( ) = which determine the density : [ dens ] n = the consistent summation of 2-body scattering leads to an integral equation for the pseudo - energy @xmath183 : [ pseudo.energy ] ( ) = _ - - ( 1 + ( d ) g ( , ) ) the kernel @xmath45 in this equation is related to the logarithm of the 2-body s - matrix @xmath184 as follows : @xmath185 where @xmath186 is the spatial volume and @xmath187 .
the kernel has the following structure : [ gstructure ] g = - ( 1 + i ) where @xmath59 is the scattering amplitude and @xmath188 represents the available phase space for two - body scattering .
the argument of the @xmath189 can be identified as the s - matrix function .
finally , the free energy density then takes the simple form : [ freefoam ] = - ( d ) @xmath190 where @xmath191 is the filling fraction of the free theory : [ fzero ] f_0 ( ) = consider first the 1-dimensional bosonic case .
the model is integrable , which implies that the n - body s - matrix factorizes into 2-body s - matrices , and the exact free energy is given by the thermodynamic bethe ansatz@xcite .
the 2-body s - matrix is [ 1ds ] s(k - k ) = in the unitary limit @xmath53 , @xmath1 , and the thermodynamic bethe ansatz reduces to that of a free gas of fermionic particles . in 3 spatial dimensions the exact kernel for a single component boson is the following@xcite : [ g3d ] g ( |-| ) & = - ( ) + & = - ( - ) where @xmath43 is a renormalized coupling : [ gr ] = + . with @xmath44 an ultra - violet momentum cutoff . as in the 1-dimensional case , as discussed in section ii , the unitary limit corresponds to @xmath192 where @xmath1 .
thus in the unitary limit the kernel becomes [ unitaryg3d ] g ( , ) = ( d=3 ) where @xmath193 corresponds to @xmath30 being just below , above @xmath65 , where the scattering length @xmath194 ( bec side ) and @xmath195 ( bcs side ) respectively .
it should be kept in mind that the underlying interactions are attractive in both cases since the fixed point occurs at negative @xmath30 . in two spatial dimensions
the single boson kernel obtained in @xcite is @xmath196 where @xmath197 is defined in eq .
( [ lambdastar ] ) , and @xmath198 is the relative momentum . in the unitary limit @xmath53 ,
the theory is at the scale @xmath197 and one should consider @xmath94 .
the result is that @xmath45 becomes a constant in this unitary limit : [ unitaryg2d ] g ( || ) = ( d=2 ) in the attractive case , @xmath199 approaches
@xmath200 from above as @xmath201 , and thus corresponds to the @xmath202 sign above .
the @xmath203 sign then corresponds to the repulsive case where @xmath200 is approached from below . for two - component fermions , the phase space factors
@xmath188 in @xcite are doubled , and since @xmath204 , the kernels have an extra @xmath205 in the fermionic case : [ fermg ] g_fermi = g_bose the above unitary limit of the kernels leads to a scale - invariant integral equation for the pseudo - energy , which in turn leads to the scaling forms of the previous section .
this will be described in detail for the @xmath15 case in subsequent sections .
note that the kernel has a well - defined expansion in @xmath206 ; in the sequel we only consider the above leading terms . solving the integral equation for intermediate values of @xmath30 between @xmath207 and @xmath8 and inputting the solution into the expression for the free energy should track the rg flow between @xmath88 and @xmath208 , or between @xmath209 and @xmath210 , however we will not study this in the present work .
the two - dimensional case is considerably simpler to analyze since the kernel @xmath45 is a constant . in any dimension
it is convenient to define : [ 2d.1 ] ( ) ( ) - _ + ; ( ) e^- ( ) the expression ( [ freefoam ] ) for the free energy can be simplified to the following : [ freefoam2 ] = - ( d ) @xmath211 consider first a hypothetical single component fermion . in order to have a point - like local interaction one needs at least two components , and this will be treated in section viii .
as explained there , due to the su(2 ) symmetry , the two - component fermion reduces to two identical copies of the following 1-component results . since the kernel is independent if @xmath35
, @xmath212 is a constant .
the integral equation then becomes the transcendental algebraic equation : [ 2d.2 ] = 1 ( 1 + z ) ( we have used @xmath213 . ) the @xmath203 ( @xmath202 ) sign corresponds to the repulsive ( attractive ) case .
the free energy takes the form eq .
( [ freec ] ) where the scaling function , expressed in terms of the fugacity @xmath214 , is [ 2d.3 ] c = - @xmath215 the scaling function @xmath155 in eq .
( [ tzero ] ) is [ 2d.3b ] b = c and the density is [ 2d.4 ] n = ( 1 + z ) the energy per particle scaling function @xmath171 is also expressed in terms of @xmath216 : [ 2d.5 ] = c finally @xmath160 takes the form : [ 2d.5b ] = @xmath217 ^ 2 b we will mainly plot these quantities against @xmath2 or its inverse ; these functions are implicitly functions of the more physical quantity @xmath218 , where @xmath219 is the fermi temperature .
the relation with @xmath2 is [ ttf ] = there exists a solution to eq .
( [ 2d.2 ] ) for any @xmath214 , i.e. for all @xmath220 .
the density is shown in figure [ atfermdensity ] and takes on all positive values and approaches @xmath14 as @xmath221 . as @xmath161 , the density approaches the free field value @xmath222 .
however at high temperatures there is a departure from the free field value : [ denzero ] _ /
t 0 n = 0.9546 the filling fraction @xmath223 is plotted in figure [ fatfermi ] .
the scaling function @xmath216 is shown in figure [ atfermc2 ] .
one finds : [ atferm.1 ] _
/ t 0 c = 0.624816 which is higher than the free field value @xmath224 . from these results ,
one obtains the equation of state : [ state.1 ] p = 1.077 n t , ( /t 0 ) one needs to make sure that all regions of @xmath2 are physical , and not for instance metastable .
in particular , the entropy must increase with temperature , otherwise the specific heat is negative . in figure [ atfermentropy ]
we plot the entropy density @xmath225 , and one sees that @xmath226 when @xmath227 .
one must bear in mind that our formalism , though non - perturbative in some respects , is still an approximation , and such a feature could be an artifact that disappears once the corrections are incorporated .
since it is beyond the scope of this work to systematically explore these corrections , we will instead try and interpret our results as they are , for this and subsequent cases .
the above feature of the entropy density could signify a phase transition at @xmath228 , where the density and temperature are related by : [ 117 ] n = 13.1 t_c = 0.076 t_f where @xmath229 is the fermi temperature @xmath230 , where @xmath157 is defined in section iii , and @xmath231 is the critical temperature for this hypothetical phase transition .
the above value is comparable to the prediction @xmath232 in @xcite for quasi 2d ( trapped ) systems . in all other regions , and for all other cases considered below , @xmath233 .
.,width=377 ] [ atfermdensity ] for attractive fermions at two different values of @xmath234 .
, width=377 ] [ fatfermi ] for attractive fermions as a function of @xmath2.,width=377 ] [ atfermc2 ] divided by @xmath235 for attractive fermions as a function of @xmath236.,width=377 ] [ atfermentropy ] the scaling function @xmath155 is shown in figure [ atfermb2 ] .
it has a well defined limit as @xmath237 , however the limiting value depends on whether @xmath110 is positive or negative : [ atferm.2 ] _
t/0 ^ - b = 0 , _
t/0^+ b = 1 recall @xmath238 is the free fermion value .
the discontinuity at @xmath239 is traced to the fact that @xmath240 , and has no physical significance .
since the density approaches @xmath241 , the equation of state is [ state.2 ] p = 2 n^2 /m , ( /t ) for attractive fermions as a function of @xmath242.,width=377 ] [ atfermb2 ] the energy per particle scaling function @xmath171 is shown in figure [ atfermxiprime ] .
it has the limiting values : [ atfermxiprime ] _
/t 0 = 0.907412 , _ /t - = 0.842766 , _ /t
= the above value for @xmath171 as @xmath243 was determined numerically , however it turns out to equal @xmath244 .
equation ( [ enpart.5 ] ) then shows that the energy per particle @xmath245 , which simply means that in the limit @xmath243 the gas is effectively classical . this feature will be encountered in other cases below . for attractive fermions as a function of @xmath2.,width=377 ] [ atfermxiprime ]
the single particle energy scaling function @xmath160 has the behaviour shown in figure [ atfermxitmu ] .
it has the limiting behavior [ atfermxi ] _
t/ 0^+ = 1 , _ t/ = 2.25593 the value @xmath246 is consistent with the arguments in @xcite .
it diverges as @xmath247 due to the vanishing density at @xmath248 . for attractive fermions as a function of @xmath242.,width=377 ] [ atfermxitmu ]
the ratio of the shear viscosity @xmath249 to the entropy density @xmath225 has units of @xmath250 in any dimension .
a lower bound was conjectured for 3d relativistic systems@xcite : [ bound ] based on the ads / cft correspondence .
the bound is saturated in certain strongly coupled supersymmetric gauge theories .
it is therefore interesting to study this ratio for non - relativistic , strongly interacting systems , in 3 and lower dimensions , since it is unknown whether there really is a lower bound .
this ratio was studied for @xmath78 unitary fermi gases in @xcite .
the attractive fermion case is the most interesting and well - behaved case in our formalism , as will be evident in the subsequent sections , so we study the issue in this case first .
consider first a gas with a single species of particle of mass @xmath251 .
for a non - relativistic system in 2 spatial dimensions : [ eta.1 ] = n |v m where @xmath252 is the average speed and @xmath253 the mean free path .
the mean free path is @xmath254 where @xmath98 is the total cross - section .
the @xmath255 comes from the ratio of the mean speed to the mean relative speed@xcite . in the unitary limit , formally @xmath256 , which implies the scattering amplitude @xmath257 .
( see section iv . )
the cross - section in eq .
( [ cross ] ) gives @xmath258 in 2 dimensions , where @xmath259@xcite .
thus , in the unitary limit [ eta.2 ] = where @xmath35 is a single particle momentum .
this gives [ eta.3 ] = @xmath260 since the pressure is determined by the average kinetic energy , and the ideal gas relation eq .
( [ ener ] ) still holds for a unitary gas , the average kinetic energy per particle is just @xmath261 , eq .
( [ enpart.1 ] ) .
finally we obtain : [ eta.4 ] = @xmath262 ^ -1 where all the quantities in the above formula are the single component values of the last section . in terms of the scaling functions : [ eta.5 ] = = for two - component fermions
the available phase space @xmath188 is doubled .
the cross - section is halved since spin up particles only scatter with spin down .
finally the entropy is doubled .
this gives [ etasbosfer ] @xmath263 _ fermi = 4 @xmath264_bose the above expression is easily evaluated numerically using the expressions of the last section .
the result is displayed for small values of @xmath2 in figure [ etassmall ] . in this regime
, @xmath3 is well above the conjectured bound , and comparable to the @xmath78 values extracted from the experimental data@xcite .
we find [ etaslim.1 ] _
/t 0 = 7.311 this is well - below the values for common substances like water , however it is comparable to values for liquid helium , which is about 9 times the bound@xcite . in the region shown : [ etasmin ] 6.07 in figure [ etasttf ] we plot @xmath3 as a function of @xmath218 , and one observes a behavior quite similar both qualitatively and quantitatively to the 3d data summarized in@xcite , where the minimal value is about 6 times the bound . as a function of @xmath2 .
the horizontal line is @xmath21.,width=377 ] [ etassmall ] as a function of @xmath218 for attractive fermions.,width=377 ] [ etasttf ] recall that in the last section , it was argued that the region @xmath265 is unphysical , or perhaps metastable , since the entropy increases with decreasing temperature there .
thus for the regions that are surely physical , the bound ( [ etasmin ] ) holds .
it is nevertheless interesting to study the viscosity in this unphysical region . in the zero temperature limit ,
i.e. @xmath156 , @xmath3 slowly decreases and seems to approach the bound .
see figure [ etasmiddle ] .
however it eventually dips below it .
see figure [ etaslarge ] .
this behavior can be understood analytically as follows . for @xmath266
very large , the solution to the equation ( [ 2d.2 ] ) is approximately : [ yap ] y(x ) + 1/2 + ( x - 1 ) /4 this leads to the asymptotic expansions : [ cnap ] c & & 3 ( x^2 + xx -x ) /^2 + n & & ( x + ( x)/2 ) and [ etasap ] @xmath267 in terms of the density : [ etaslim ] _ /t
= it is important to note that although the energy per particle scaling function @xmath160 approaches the free field value at low temperatures , the above behavior is very different from the free field case . in the latter , the scaling function @xmath268 , which gives the diverging behavior at very low temperature : [ etasfree ] , ( free fermions ) finally , as @xmath269 , @xmath270 , and @xmath271 and @xmath272 .
this leads to exponential growth : [ etaslimmin ] _
/t - = - e^-/t = - as a function of @xmath2 as @xmath2 gets very large .. the horizontal line is @xmath21.,width=377 ] [ etasmiddle ] as a function of @xmath2 in the limit of very low temperatures ( top curve ) .
the horizontal line is @xmath21 .
the bottom curve is the approximation eq .
( [ etasap]).,width=377 ] [ etaslarge ]
using the same definitions as for fermions in eq .
( [ 2d.1 ] ) , the integral equation for bosons is [ bos.1 ] y = 1 ( 1- z y ) where the @xmath202 ( @xmath203 ) sign refers to repulsive ( attractive ) interactions .
the scaling function @xmath216 is now [ bos.c2 ] c = @xmath273 the scaling function @xmath155 has the same expression as in eq .
( [ 2d.3b ] ) , with the above @xmath216 .
the density now is [ bos.density ] n = - ( 1 - z y ) the energy per particle scaling functions now take the form : [ bos.xiprime ] = - and [ xibosdef ] = @xmath274 ^ 2 b for this bosonic case , there is only a solution to eq .
( [ bos.1 ] ) for @xmath275 , or @xmath276 .
the density is shown in figure [ atbosdensity ] , and note that it has a maximum .
the filling fractions are shown for several @xmath22 up to @xmath277 . from these plots
, it appears that @xmath223 is diverging at @xmath278 as @xmath214 approaches @xmath279 , suggestive of condensation to a superfluid .
let us refer to the maximum density then as the critical density [ atbos.den ] n_c 1.24 t_c 0.81 t_f . the origin of the axes is at @xmath280.,width=377 ] [ atbosdensity ] for attractive bosons for @xmath281 .
, width=377 ] [ fatbos ] the scaling function @xmath216 is shown in figure [ atbosc2 ] .
the limiting value is [ atbos.c2 ] c ( ) 0.35 which is considerably less than for a free boson with @xmath282 .
the function @xmath155 decreases to zero at @xmath133 , and the limiting value is [ atbos.b2 ] b ( ) 0.94 for attractive bosons as a function of @xmath2 .
the origin of the axes is at @xmath283.,width=377 ] [ atbosc2 ] the energy per particle scaling functions @xmath160 and @xmath171 are shown in figures [ atbosxiprime ] , [ atbosxitmu ] .
the limiting values are [ atbos.xip ] _ /t - = 0.842766 , ( ) 0.39 and [ atbos.xi ] _
t/0 ^ - = , ( ) 2.3 as for the fermionic case , the value @xmath284 implies the energy per particle @xmath245 , which means the gas is in the classical limit . for attractive bosons as a function of @xmath2.,width=377 ] [ atbosxiprime ] for attractive bosons as a function of @xmath242.,width=377 ] [ atbosxitmu ] the ratio @xmath3 given by the expressions eq .
( [ eta.5 ] ) with the appropriate bosonic @xmath216 and @xmath155 .
we find that @xmath3 has its minimum value at @xmath285 , where it is actually below the bound [ etasboszc ] 0.4
there are solutions to the eq .
( [ 2d.2 ] ) for all @xmath286 , and the density is positive in this range .
a plot of the density is shown in figure [ repfermdensity ] .
the density maximizes at @xmath287 , where @xmath288 .
this seems physically reasonable given the strong repulsion in the unitary limit .
in contrast , for small coupling @xmath30 , the kernel @xmath289 , and there are solutions for positive chemical potential . in the bosonic case
, our formalism leads to a critical density of @xmath290 for the kosterlitz - thouless transition@xcite .
the filling fractions are shown in figure [ frepferm ] .
note that they are considerably smaller than in the attractive case , as expected .
the scaling function @xmath216 is shown in figure [ repfermc2 ] .
it has the limiting value [ crepferm ] _ /
t 0 c = 0.303964 , which is considerably less than the free field value @xmath291 .
the scaling function @xmath155 has the limiting values @xmath292 , and @xmath293 .
since the pressure is proportional to @xmath216 , the same density occurs at two different pressures .
one sees that for @xmath294 , the density increases with pressure as it should .
for @xmath295 the density decreases with increasing pressure , violating @xmath296 , and should thus be viewed as an unphysical region .
we interpret this as a phase transition occuring at @xmath297 , where the critical density [ repferm.1 ] n_c = 0.25 t_c 4.0 t_f at this critical point @xmath298 .
this perhaps corresponds to a transition to a crystaline phase .
we can not prove this , since we have not calculated the shear modulus for instance , and it could simply be an artifact of our approximation , but let us take it as a hypothesis . whereas a wigner crystal phase occurs at low density in three dimensions , it occurs at high density in two dimensions@xcite .
the possibility of crystal phases for repulsive bosons was studied in @xcite . for coulomb repulsion with strength @xmath299 ,
the thermodynamic state of a classical coulomb system is determined by @xmath300 , where @xmath108 is the density .
experimentally it was found that @xmath301@xcite , which gives a critical density [ gammacrystal ] n_c 2.15 10 ^ 9 t^2 on the other hand , since our model has point - like interactions , and is in the unitary limit , the conditions for a crystal phase are expected to be different . in particular our formula
( [ repferm.1 ] ) has no @xmath299 dependence , and this leads to a linear in @xmath109 dependence . for @xmath251
equal to the electron mass [ crystalnc ] n_c 4.5 10 ^ 9 t it is interesting to note that for @xmath109 of order 1 , which is where the data in @xcite was taken , the two densities ( [ gammacrystal ] ) and ( [ crystalnc ] ) are comparable .
.,width=377 ] [ repfermdensity ] for repulsive fermions for @xmath302 .
the top curve corresponds to the critical density in eq .
( [ repferm.1 ] ) .
, width=377 ] [ frepferm ] for repulsive fermions as a function of @xmath2 .
, width=377 ] [ repfermc2 ] the energy per particle function @xmath171 is shown in figure [ repfermxiprime ] .
the limiting values are : [ repferm.xiprime ] _ / t - = 0.842766 , ( z_c ) = 1.32 as in previous cases , the above value of @xmath171 implies @xmath303 as @xmath243 and the gas is thus in the classical limit . beyond the critical point where the density increases , @xmath304 .
note that @xmath171 starts to diverge around the proposed phase transition at @xmath305 .
finally the other energy per particle function @xmath160 is less interesting , as it diverges at both endpoints of the density range where the density goes to zero . at the critical point
it is still quite large : @xmath306 . for repulsive fermions as a function of @xmath2.,width=377 ] [ repfermxiprime ] the ratio @xmath3 reaches its minimum near @xmath285 , and
is quite large compared to the attractive case : [ etasrepf ] 22.6 the repulsive boson case is similar to the repulsive fermion , except that now there are only positive density solutions to the eq .
( [ bos.1 ] ) for @xmath307 .
figure [ bothdensity ] compares the density for bosons verses fermions .
the maximum density for bosons , which could again possibly signify a critical point at @xmath308 , is half that of the fermion case : [ ncbos ] n_c = 0.125 t_c 8.0 _ f at this point @xmath309 .
the ratio @xmath3 is shown in figure [ etasrepbos ] .
it has a minimum at @xmath310 near the critical point and in the physical region : [ etasbosrep ] 8.85 recall , the region @xmath311 is unphysical .
.,width=377 ] [ bothdensity ] as a function of @xmath2 for repulsive bosons.,width=377 ] [ etasrepbos ]
let us suppose the gas consists of multiple species of particles labeled by the index `` a '' , of mass @xmath312 , chemical potential @xmath313 , and statistical parameter @xmath314 . introduce a pseudo - energy @xmath315 for each type of particle and the filling fractions : [ multi.f ] f_a ( ) = as shown in appendix b , in our approximation , the pseudo - energies satisfy the coupled integral equations : [ multi.integral ] _ a ( ) = _ a ( ) - _ a - @xmath316 where @xmath317 , [ multi.zd ] y_a e^-_a , _ a _ a - _ a + _ a and @xmath318 is the 2-body scattering kernel . by galilean invariance ,
the kernel is a function of @xmath319 where @xmath320 is a velocity . for any function
@xmath30 we defined : [ multi.gstar ] @xmath321 ( ) ( d ) g_ab ( , ) g ( ) the free energy density then takes the simple form : [ multi.free ] = - _ a ( d ) @xmath322 for the two - component fermion defined by the action ( [ fermionaction ] ) , spin up particles scatter with spin down , and @xmath323 .
thus , if the chemical potentials are equal , @xmath324 , then @xmath325 , and the thermodynamics is just two copies of the single component fermion described above . in particular ,
the density and @xmath216 are doubled , but the energy per particle scaling functions are the same .
we have shown that the s - matrix based approach to quantum gases developed in @xcite leads to a new treatment of the scale - invariant unitary limit , where all of the thermodynamic scaling functions can be computed as a function of @xmath234 . though our methods are of course not exact , they are sufficiently novel to provide new insights into these systems .
it would be worthwhile to undertand the corrections to our results due to other types of diagrams involving n - body scattering for @xmath326 , and also other more complicated diagrams involving 2-body scattering ; this can be systematically explored using the full formalism in @xcite . in this paper
we have mainly analyzed the 2-dimensional case , deferring the analysis of the integral equations in the 3-dimensional case to a separate publication@xcite .
for the 2-dimensional case , this required us to define a meaningful unitary limit where the s - matrix equals @xmath327 , and such a limit has not been considered before .
we have calculated most of the interesting scaling functions for the free energy and energy per particle .
the ratio of the shear viscosity to entropy density @xmath3 was also analyzed , and for fermions and repulsive bosons , it is above the conjectured lower bound of @xmath4 . for attractive bosons it drops below it by a factor of @xmath328 , however this could be an artifact of our approximation . for attractive fermions , the conjectured lower bound
is reached at very large @xmath329 , however this was argued to occur in an unphysical or metastable region ; in the physical regions , @xmath330 .
we thank erich mueller for helpful discussions . this work is supported by the national science foundation under grant number nsf - phy-0757868 .
consider a single free boson with action eq .
( [ bosonaction ] ) . setting the external energy and momentum to zero
, the 4-point vertex at tree - level plus 1-loop is [ a.1 ] ^(4 ) = - i g - @xmath331 @xmath332 the @xmath333 integral can be done by closing the contour in the upper half plane picking up the pole at @xmath334 : [ a.2 ] 1-loop = - i introducing an ultraviolet cut - off @xmath44 : [ a.3 ] ^(4)= - i @xmath335 defining the dimensionless coupling @xmath336 as @xmath337 , and requiring the @xmath338 be independent of @xmath44 gives the beta function : [ a.4 ] = ( 2-d ) - ^2 where @xmath339 is the logarithm of a length scale ; increasing @xmath340 corresponds to flowing to lower energy . with the convention for the coupling @xmath30 in eq .
( [ fermionaction ] ) for the fermionic case , the beta - function is the same as above but to half the symmetry factor of the 1-loop diagram .
in this appendix , we extend the derivation in @xcite to the general case of multiple species of particles , of possibly mixed bose / fermi statistics . following the definitions in section vii ,
the 2-body foam diagram approximation is obtained by considering the free energy functional : [ b.1 ] & = & - ( d ) _ a @xmath341 + & & - ( d ) ( d ) _ a , b _ a ( ) g_ab ( , ) _ b ( ) where [ b.2 ] f_0,a = , _ a = f_a / y_a the integral equation for the pseudo - energy @xmath342 follows from the variational principle @xmath343 . using [ b.3 ] = _ a , b ( 2 ) ^d ( - ) , after some algebra one obtains the integral equation ( [ multi.integral ] ) .
( we have used @xmath344 . ) using the solution to the integral equation ( [ multi.integral ] ) in the expression for @xmath345 , one obtains the free energy density @xmath346 in eq .
( [ multi.free ] ) .
astrakharchik , j. boronat , j. casulleras and s. giorgini , _ equation of state of a fermi gas in the bec - bcs crossover : a quantum monte carlo study _ , phys .
* 93 * ( 2004 ) 200404 [ arxiv : cond - mat/0406113 ] .
e. burovski , n. prokofev , b. svistunov and m. troyer , _ the fermi - hubbard model at unitarity _ , new j.phys.*8 * ( 2006 ) 153
[ arxiv : cond - mat/0605350 ] ; _ critical temperature and thermodynamics of attractive fermions at unitarity _ , phys .
* 96 * ( 2006 ) 160402 .
p. nikoli and s. sachdev , _ renormalization group fixed points , universal phase diagram , and 1/n expansion for quantum liquids with interactions near the unitarity limit _ , phys
* a75 * ( 2007 ) 033608 [ arxiv : cond - mat/0609106 ] . | in three spatial dimensions , in the unitary limit of a non - relativistic quantum bose or fermi gas , the scattering length diverges .
this occurs at a renormalization group fixed point , thus these systems present interesting examples of interacting scale - invariant models with dynamical exponent @xmath0 .
we study this problem in two and three spatial dimensions using the s - matrix based approach to the thermodynamics we recently developed .
it is well suited to the unitary limit where the s - matrix @xmath1 , since it allows an expansion in the inverse coupling .
we define a meaningful scale - invariant , unitary limit in two spatial dimensions , where again the scattering length diverges . in the two - dimensional case ,
the integral equation for the pseudo - energy becomes transcendentally algebraic , and we can easily compute the various universal scaling functions as a function of @xmath2 , such as the energy per particle .
the ratio of the shear viscosity to the entropy density @xmath3 is above the conjectured lower bound of @xmath4 for all cases except attractive bosons . for attractive 2-component fermions , @xmath5 , whereas for attractive bosons @xmath6 .
\def\ ] ] ] @xmath7 ) # 1#2#1 # 2 # 11 # 1 # 1#1 # 1#2#1 # 2 # 1#1 # 1 | # 1 pi # 1e^^#1 # 1_#1 # 1#2 # 1 , # 2 pi # 1e^^#1
# 1_#1 = cmss12 = cmu10 scaled1 # 1 pid(2)^d # 1|#1 # 1#1 | # 1#1 0=0=0 0 by-1exa # 1dim[#1 ] |
one - dimensional ( 1d ) interacting fermionic systems show remarkable physical properties and are promising elements for future nanoelectronics .
the electron - electron interaction manifests itself in a particularly dramatic way in 1d systems , inducing a strongly correlated electronic state
luttinger liquid ( ll ) @xcite . a paradigmatic experimental realization of quantum wires are carbon nanotubes @xcite ; for a recent review see ref . .
further realizations encompass semiconductor , metallic and polymer nanowires , as well as quantum hall edges .
there is currently a growing interest in non - equilibrium phenomena on nanoscales .
a tunneling spectroscopy ( ts ) technique for non - equilibrium nanostructures was developed in ref . .
employing a superconducting tunneling electrode allows one to explore not only the tunneling density of states ( tdos ) but also the energy distribution function .
the energy relaxation found in this way provides information about inelastic scattering in the system . in a very recent experiment @xcite this ts method
was applied to a carbon nanotube under strongly non - equilibrium conditions . in this paper
, we develop a theory of ts of a ll out of equilibrium .
specifically , we consider a ll conductor connected , via non - interacting leads , to reservoirs with different electrochemical potentials , @xmath0 and different temperatures @xmath1 , @xmath2 ( where the indices @xmath3 , @xmath4 stand for left- and right - movers ) .
it is assumed that the coupling to the leads is adiabatic on the scale of the fermi wave length , so that no backscattering of electrons takes place .
we model the leads as non - interacting 1d wires , so that the electron - electron interaction is turned on at the vicinity of the points @xmath5 , see fig .
this model is quite generic to properly describe the problem at hand , independently of the actual geometry of the leads .
note also that the 1d setup with strongly non - uniform interaction may be experimentally realized by using external screening gates .
it is known that energy relaxation is absent in a uniform clean ll . within the golden - rule framework , the lack of energy relaxation for forward scattering processes results from 1d kinematic constraints that do not allow to satisfy the energy and momentum conservation laws simultaneously @xcite . on a more formal level
, the conservation of energies of individual particles in a spatially uniform ll is protected by the integrability of the system , which implies an infinite number of conservation laws @xcite .
inclusion of spatial dependence into the model violates these laws and leads to energy relaxation that takes place at the regions where the interaction varies in space @xcite .
the fact that inhomogeneous interaction induces energy relaxation of electrons has been pointed out for the first time in ref . in the context of interacting quantum hall edges but a detailed analysis of this effect
has been missing until now . on the other hand
, one may expect this to be a dominant effect on the electron distribution function in experiments done on modern high - quality quantum wires ( such as ultraclean carbon nanotubes @xcite ) , under non - equilibrium conditions .
there is thus a clear need in the theory of ts in non - equilibrium ll .
it is worth noting that we assume the absence of backscattering due to impurities in the wire .
when present , such impurities strongly affect the electronic properties of a ll wire : they induce diffusive dynamics at sufficiently high temperature @xmath6 and localization phenomena proliferating with lowering @xmath6 ( ref . ) , as well as inelastic processes @xcite .
we also neglect the nonlinearity of the electron dispersion whose influence on spectral and kinetic properties of 1d electrons was recently studied in refs . , .
the dashed line corresponds to the limit of a sharp variation of @xmath7 at the boundaries . ]
within the ll model , the electron field is decoupled in a sum of right- and left - moving terms , @xmath8 , where @xmath9 is the fermi momentum .
the hamiltonian of the system reads @xmath10 where @xmath11 is the electron velocity and @xmath12 is the spatially dependent electron - electron interaction constant .
we will proceed by following the lines of the functional bosonization approach @xcite in the non - equilibrium ( keldysh ) formulation @xcite@xmath13@xcite . performing the hubbard - stratonovich transformation , one decouples the interaction term via a bosonic field @xmath14 and gets the action @xmath15=i\sum_{\eta = r , l}\psi^\dagger_\eta ( \partial_\eta-\phi)\psi_\eta-\frac{1}{2}\phi g^{-1}\phi\ , , \end{aligned}\ ] ] where @xmath16 and the fields are defined on the keldysh time contour .
the information about physical observables is contained in keldysh green functions @xcite @xmath17 and @xmath18 ; see , in particular , appendix [ s9 ] where we express tunneling current in terms of functions @xmath19 and discuss how its measurement allows to determine @xmath19 experimentally .
the green functions @xmath19 can be presented in the form @xmath20 e^{-{i\over 2 } \phi g^{-1}\phi } \nonumber\\ & \times & g^\gtrless_\eta[\phi](x , t;x',t ' ) , \label{green}\end{aligned}\ ] ] where we introduced the green function in a given field configuration , @xmath21 $ ] , and the sum of vacuum loops , @xmath22 $ ] . in 1d geometry
the coupling between the fermionic and bosonic fields can be eliminated by a gauge transformation @xmath23 , if we require @xmath24 as a result , @xmath25 $ ] can be cast in the form @xmath26(x , t;x',t')&= & g^\gtrless_{\eta,0}(x - x';t - t ' ) e^{-i\eta ev(t - t')/2 } \nonumber \\ & \times & e^{\phi^\gtrless_\eta(x , t;x',t ' ) } \,.\end{aligned}\ ] ] here @xmath27 @xmath28 is the green function of free fermions , @xmath29 the coordinate @xmath30 labels the trajectory of a particle , and we use the convention that in formulas @xmath31 should be understood as @xmath32 for right / left moving electrons .
it is convenient to perform a rotation in keldysh space , thus decomposing fields into classical and quantum components , @xmath33 , where the indices @xmath34 and @xmath35 refer to the fields on two branches of the keldysh contour .
further , we introduce vector notations by combining @xmath36 and @xmath37 in a 2-vector @xmath38 .
to proceed further , we resolve eq .
( [ diff_eq ] ) and express @xmath39 through @xmath14 as @xmath40 where @xmath41 is the green function of free bosons , @xmath42 its retarded and advanced components are given by @xmath43 the keldysh component of @xmath44 is given by @xmath45 , where @xmath46 is determined by the temperature @xmath47 of the reservoir from which the electrons moving in direction @xmath31 emerge , @xmath48 using eqs .
( [ phi ] ) and ( [ theta ] ) and performing a transformation to the coordinate space , we express the exponent @xmath49(x , t , x',t')$ ] through the bosonic field @xmath50 , @xmath51(x , t , x',t')= \int { d\omega\over 2\pi } dy \bm{\phi}^t_{-\omega } ( y)\bm{j}^{\gtrless}_{\eta,\omega}(y;x , t , x',t').\ ] ] the components of @xmath52 are found as @xmath53e^{-i\omega\xi_\eta}\ !
-\!\theta[\eta(\!x'\!-\!y\!)]e^{-i\omega\xi_\eta'}\right)\,,\nonumber \\ & & j^\gtrless_{2,\eta,\omega}(y)=-\frac{e^{i\eta \frac{\omega}{v}y}}{\sqrt{2}v } \left(e^{i\omega\xi_\eta}-e^{i\omega\xi_\eta'}\right)b_\eta^{(0)}(\omega ) \nonumber \\ & & \mp \frac{e^{i\eta\frac{\omega}{v}y}}{\sqrt{2}v } \left(\theta[\eta(y - x)]e^{-i\omega\xi_\eta } + \theta[\eta(y - x')]e^{-i\omega\xi_\eta'}\right),\label{j12}\end{aligned}\ ] ] where @xmath54 is the heviside @xmath55-function .
the vacuum loop factor in eq .
( [ green ] ) is given by @xmath56 = \exp\left(-\frac{i}{2}\bm{\phi}^t \pi \bm{\phi}\right)\,,\end{aligned}\ ] ] where @xmath57 is the polarization operator , @xmath58 it can be decomposed into left and right moving parts , @xmath59 , with @xmath60 where @xmath61 .
performing the averaging over the auxiliary field @xmath14 , we get @xmath62 where the effect of the interaction is represented by the `` debye - waller factor '' @xmath63 with @xmath64 here @xmath65 is the screened electron - electron interaction potential .
its retarded component is given by @xmath66\,,\end{aligned}\ ] ] where the function @xmath67 is determined by the following differential equation @xmath68 which describes the plasmon propagation in a medium with spatially dependent sound velocity @xmath69 .
the keldysh component of the interaction propagator is obtained as @xmath70 where @xmath71 at equilibrium , @xmath72 , this reduces to @xmath73b(\omega)\,,\end{aligned}\ ] ] in agreement with the fluctuation - dissipation theorem .
so far we made no restriction on the way the interaction changes in space .
let us consider first the case when the interaction turns on and off sharply on the scale set by the temperatures , @xmath74 .
this limit can be modelled via a stepwise interaction as represented by the dashed line in fig .
[ fig1 ] . equation ( [ plazmon ] ) for @xmath67 can be then straightforwardly solved by using the fact that the velocity @xmath75 is constant in each of three regions and employing the proper boundary conditions [ continuity of @xmath76 and of @xmath77 at @xmath78 . in the ts context , we are interested in the green functions @xmath19 with coinciding spatial arguments , @xmath79 . assuming @xmath80 to be in the interacting part of the wire ( and not too close to the boundaries ) and setting @xmath81 , we find @xmath82\,,\end{aligned}\ ] ] where @xmath83 is the conventional dimensionless parameter characterizing the interaction strength in a ll and @xmath84 the integral in eq .
( [ eq_f ] ) and in analogous formulas below is logarithmically divergent at large frequencies and require an ultraviolet regularization .
specifically , these integrals are understood as regularized by a factor @xmath85 , where @xmath86 is an ultraviolet cutoff .
deriving eq .
( [ eq_f ] ) , we have neglected terms of the form @xmath87 ( with non - zero integer @xmath88 ) that arise due to the fabry - perot - type interference of plasmon modes reflected at the boundaries . keeping these terms would lead to an additional oscillatory structure in energy @xcite with the scale @xmath89 .
since we are interested in ts of long wires , we assume that this scale is much less than @xmath90 , so that oscillations are suppressed . substituting eq .
( [ eq_f ] ) into eq .
( [ eq_g ] ) , we finally get the green functions : @xmath91^{1+\alpha } \bigg[g^\gtrless_{l,0}(t)\bigg]^{\beta}e^{-i\eta evt/2}\,,\ ] ] where @xmath92 the green functions ( [ green_fun ] ) can be determined experimentally from ts measurements @xcite , see appendix [ s9 ] .
their difference determines the tdos @xmath93 , @xmath94 while each of them separately ( or their sum ) contains also information about the distribution function , as discussed below .
the results for the tdos have been found in ref . .
next we consider the non - interacting parts of the wire , and discuss , e.g. , the right moving electrons . in the region
i ( see fig . [ fig1 ] ) , @xmath95 , we find from eqs .
( [ f ] ) , ( [ j12 ] ) that @xmath96 , so that the green functions of the right movers are not modified by interaction .
physically this is quite transparent : the right - moving electrons in this part of the system are just coming from the reservoir and are not yet `` aware '' of the interaction with the left - movers .
the situation is distinctly different in the region iii , @xmath97 .
assuming @xmath79 , we find @xmath98\,.\end{aligned}\ ] ] substituting eq .
( [ region3 ] ) into eq .
( [ eq_g ] ) , one gets @xmath99^{\cal t}\bigg[g_{l,0}^\gtrless(t ) \bigg]^{\cal r } e^{-i\eta evt/2},\end{aligned}\ ] ] where @xmath100 since @xmath101 in eq .
( [ region3 ] ) is real , the tdos is not affected by the interaction , @xmath102 , as expected .
the modification of the functions @xmath103 as compared to that of incoming electrons , @xmath104 , implies therefore the change in the distribution function @xmath105 of right - movers .
indeed , for non - interacting particles @xmath106 and @xmath107 $ ] .
we thus see that the electrons ejected from the interacting part of the wire into the lead are affected by the interaction : their distribution function has changed .
the left - moving electrons can be analyzed in the same way ; the corresponding results are obtained by replacing @xmath108 in eqs .
( [ green_fun ] ) , ( [ eq2 ] ) .
clearly , the role of the regions i and iii is interchanged in this case .
it is also worth mentioning that in the non - interacting parts of the wire the green functions are both galilean and translationally invariant , depending on coordinates and times via @xmath109 only .
we turn now to generalization of these results for the case of an arbitrary shape of @xmath12 in the contact region between the interacting part of the wire and the non - interacting leads .
the contact regions are in general characterized by some reflection coefficients @xmath110 for the plasmon amplitude , yielding reflection coefficients @xmath111 for the plasmon intensity ( @xmath112 for the left and right contact , respectively ) .
the corresponding transmission coefficients are @xmath113 .
it is instructive in this context to compare our present approach with that developed in ref .
, where we analyzed the tunneling density of states and focussed on the case of smooth variation of @xmath12 in the contact regions . as we are going to show , the method of ref .
can be generalized to the case of arbitrary contacts ( this was briefly discussed at the end of ref . ) and is also useful for the analysis of the electron distribution function . within that approach , the propagator of bosons is calculated in momentum space ( rather than in real space as in the above calculation ) .
the keldysh component of the propagator is then characterized by distribution function functions @xmath114 and @xmath115 associated with poles at @xmath116 and @xmath117 and describing `` ghosts '' ( free electron - hole pairs ) and plasmons , respectively @xcite . while the distribution function of ghosts is simply determined by that of incoming electrons , the plasmons experience in general reflection at the boundaries .
we have for the left boundary ( see fig . [ fig2 ] ) @xmath118 and similarly at the right boundary . here
we have introduced the notation @xmath119 for plasmon distributions in the interacting region of the wire and @xmath120 for out - going channels . solving these equations
, we find the plasmon distribution functions of right - movers in the interacting part of the wire , as well as in the outgoing channel ( in the right lead ) : @xmath121 the corresponding results for left movers are obtained by exchanging the indices r@xmath122l and 1@xmath1222 . in different parts of the wire .
the distributions of incoming plasmons are determined by respective leads , @xmath123 ] the method of ref .
allows us to express the exponents @xmath124 in terms of these distribution functions . for the interacting part of the wire , we get @xmath125\,.\end{aligned}\ ] ] the result for the tunneling into the non - interacting region iii of fig .
[ fig1 ] can be obtained from eq .
( [ inf_theory_gen ] ) by using the distribution functions @xmath120 corresponding to this region and replacing the interaction constant @xmath126 by zero , @xmath127\,,\end{aligned}\ ] ] where @xmath128 is the total reflection coefficient on a double - step structure , @xmath129 . for the case of sharp boundaries
the reflection and transmission coefficients are given by the fresnel law , @xmath130 and @xmath131 , so that eqs .
( [ inf_theory_gen ] ) and ( [ inf_theory_region3_gen ] ) reduce to the earlier results ( [ eq_f ] ) , ( [ region3 ] ) .
the total reflection and transmission coefficients @xmath128 and @xmath132 take in this case the values ( [ tran_coeff ] ) ( which explains the notations introduced there ) . clearly , the general formulas ( [ inf_theory_gen ] ) and ( [ inf_theory_region3_gen ] ) can also be obtained in the framework of a real - space calculation that was presented above for sharp boundaries . to do this , one has to modify the boundary conditions for the green function in @xmath67 in eq .
( [ plazmon ] ) by including the appropriate reflection and transmission amplitudes @xmath110 and @xmath133 at two boundaries and then proceeding in the same way as in course of the derivation of eqs .
( [ eq_f ] ) and ( [ region3 ] ) .
the two methods ( real space and k space ) are thus in full agreement with each other .
the formal results obtained thus far can be implemented to obtain physical observables .
consider first the non - interacting part of the setup , region iii of fig .
the effect of the interaction there amounts to modification of the distribution function of outgoing particles ( right - movers ) , which has ( in time domain ) the form @xmath134 where @xmath135 is given by eq .
( [ inf_theory_region3_gen ] ) .
this yields @xmath136 the way in which the electron distribution function is modified depends on the kinetics of the plasmons inside the interacting region . for adiabatic switching of interaction , there is essentially no plasmon scattering .
therefore , the total reflection coefficient @xmath128 and , consequently , the exponent @xmath135 in the region iii vanish . in this case
the fermions retain their distribution function : the right - movers going out into the right lead have the same distribution as the right - movers injected into the interacting region from the left lead .
( the same applies to the left - movers , of course . )
let us now discuss the opposite limit of strong reflection , @xmath137 .
for a structure with a sharp boundary , this is the case provided the interaction is strong , @xmath138 .
alternatively , this limit may be realized if the boundary regions are sufficiently extended and characterized by random @xmath7 such that plasmons with relevant frequencies are localized .
regardless of the cause , in the limit @xmath137 the left- and right - moving electrons exchange their distribution functions , except for keeping their total flux ( i.e. the chemical potential ) .
next , we consider the interacting part of the wire . analyzing the result ( [ inf_theory_gen ] )
, we see that two terms in square brackets have distinctly different physical origin .
the second term , which is proportional to the local strength of the interaction @xmath126 at the measurement point is responsible for creation of the zero - bias anomaly ( zba ) as well as for its dephasing smearing , with the non - equilibrium dephasing rate @xcite @xmath139\,.\end{aligned}\ ] ] on the other hand , the first term in the integrand of ( [ inf_theory_gen ] ) , which is governed by the difference between the incoming and local distribution of plasmons , is fully analogous to the expression for @xmath140 in the non - interacting region , eq .
( [ inf_theory_region3_gen ] ) , and describes the modification of the distribution function inside the wire , @xmath141 ( 1-\cos\omega t)\right\ } \nonumber \\ & & = \frac{i}{2\pi}\frac{1}{t+i0 } \exp\left\{-\int_0^\infty { d\omega\over \omega } [ b_\eta^{\rm w}(\omega ) - 1 ] ( 1-\cos\omega t)\right\}\ , .
\label{distr_func}\end{aligned}\ ] ] as is clear from eq .
( [ distr_func ] ) , the `` ghost '' term with @xmath142 essentially serves to cancel the bare distribution function @xmath143 , so that the distribution function @xmath144 is determined only by the plasmonic distribution @xmath145 in the wire .
this is in fact a manifestation of a general relation between the functional and full bosonization approaches , as will be discussed in detail elsewhere @xcite .
( no interaction ) and @xmath146 ( with sharp boundaries ) .
temperatures of the leads are @xmath147 and @xmath148 ; the bias voltage is @xmath149 . ]
fourier transformation of our results into the energy representation can be done numerically ( for analytic calculation at equilibrium see appendix [ s8 ] ) ; representative results are shown in figs .
[ fig3 ] , [ fig4 ] . in fig .
[ fig3 ] we present distribution functions for non - interacting parts of the wire .
temperatures are set to @xmath147 and @xmath148 ( in arbitrary units ) , the applied voltage is @xmath149 , and a sharp variation of the interaction at the boundaries ( as in sec .
[ s3 ] ) is assumed .
the distribution function of free fermions ( @xmath150 ) , plotted by a dashed line , is the same on both ends of the wire . for interacting electrons ( we choose the interaction parameter to be @xmath146 , which is in the range of characteristic values reported for carbon nanotubes , see , e.g. , ref . )
the distribution functions in two leads are different .
in particular , the distribution function in the left lead ( region i in fig . [ fig1 ] ) has a sharp edge at the energy @xmath151 , which corresponds to cold right - moving electrons . in the right lead ( region iii )
, this edge is broadened due to interaction with hot left - moving electrons .
the situation is opposite for left - moving particles .
the distribution in the right lead has a broad edge at @xmath152 that corresponds to hot left - moving electrons . due to interaction inside the wire this edge in the region i sharpens . in fig .
[ fig4 ] we present the results for the distribution functions of left- and right - moving quasiparticles in the central ( interacting ) part of the wire , eq .
( [ distr_func ] ) . for @xmath146
the plasmon reflection at the boundaries is strong . in a symmetric structure
this leads to almost equal distribution functions of both types of carriers inside the wire . in the upper panel of fig .
[ fig5 ] we show the results for tdos for @xmath153 .
the minima of tdos are reached at energies @xmath154 .
the broadening of the zba dips has two origins : smearing of the distribution function and dephasing . while the dephasing broadening [ cf .
second term in eq .
( [ inf_theory_gen ] ) ] is the same for both chiral branches , the distribution functions [ cf .
first term in eq .
( [ inf_theory_gen ] ) ] are in general different . a deeper minimum at @xmath151 reflects the fact that right - moving electrons in the wire have a much narrower distribution function .
this is because at @xmath153 the energy relaxation at the boundaries is quite weak , so that the distribution functions of cold right - movers and hot left - movers are only slightly modified .
the situation is different for @xmath146 , when distribution functions @xmath155 and @xmath156 are nearly identical ( up to a shift by @xmath157 ) , see fig .
[ fig4 ] . as a result
, the structure of the tdos also becomes symmetric .
in fact , for the chosen value of the voltage , two broad zba dips merge together . ) , in the interacting part of the wire .
all parameters are the same as in fig .
] ( normalized to its non - interacting value @xmath158 ) in the interacting region for @xmath153 , 0.4 , and 0.2 .
the temperatures of leads and the voltage are the same as in fig .
[ fig3 ] . ]
we discuss now a relation between our results for the electron distribution function and previous findings on the electric and thermal conductance of a ll wire . in the absence of backscattering the number of left and right moving particles
is separately preserved . as a result ,
the electric current is linear in the voltage @xmath159 , @xmath160 with unrenormalized landauer conductance @xmath161 , ref . .
in our formalism , this relation immediately follows from eq .
( [ eq_n ] ) and the condition @xmath162 .
this ensures that the modification of the distribution function of right ( or left ) movers by a spatially varying interaction does not affect the integral of the distribution function over energy , i.e. the total number of carriers of each type .
we turn now to the thermal conductance .
the energy current is easily found from the green functions of electrons in non - interacting parts of the wire , @xmath163\right|_{t = t'},\ ] ] which can be rewritten in terms of the electron distribution functions , @xmath164\,.\end{aligned}\ ] ] substituting the result ( [ eq_n ] ) , ( [ inf_theory_region3_gen ] ) for the distribution functions , we get the expression of the thermal current in terms of distribution functions of incoming electron - hole pairs , @xmath165\ , .
\label{thermal_2}\ ] ] according to ( [ thermal_2 ] ) , the thermal conductance is affected by the interaction [ through the reflection coefficient @xmath166 , as was first found in ref . .
note that due to the particle - hole symmetry of ll model , the applied voltage drops out of eq .
( [ thermal_2 ] ) . for the case of sufficiently sharp boundaries , when @xmath167 can be considered as @xmath168-independent for relevant frequencies , eq .
( [ thermal_1 ] ) reduces to @xmath169 deviation of the transmission coefficient @xmath167 from unity leads to the violation of the wiedemann - franz law @xcite .
as is seen from our analysis , this deviation is a manifestation of a microscopic phenomenon : energy relaxation of electrons due to non - uniform interaction . the heat current ( [ thermal_2 ] )
can be equivalently represented in terms of plasmonic distributions in the wire @xmath170\ , .
\label{thermal_4}\ ] ] this implies that the presentation of the heat current in the form ( [ thermal_1 ] ) is also valid in the interacting part of the wire , with the electronic distribution functions @xmath171 given by ( [ distr_func ] ) .
thus , also in the interacting part of the wire , the energy current can be understood as carried by properly defined quasi - particle excitation .
this is a remarkable result , which demonstrates that the concept of fermionic quasiparticles remains meaningful in a strongly interacting 1d system ( ll ) despite its non - fermi - liquid features .
to summarize , we have developed a theory of tunneling spectroscopy of ll conductor connected to reservoirs away from equilibrium . in the specific setup considered here ,
each branch originates from a source which is at equilibrium .
however , the right and the left sources have different temperatures and different chemical potentials . we have modeled the system as a ll with spatially non - uniform interaction , and calculated the single - electron green functions @xmath19 that carry information about the tdos and the fermionic distribution functions in different parts of the wire .
the interaction affects the tunneling characteristics in three distinct ways .
first , it induces a power - law zba in the tdos @xmath93 ( with two dips split by the voltage ) in the interacting part of the wire .
second , it leads to broadening of zba singularities due to dephasing , with the dephasing rate governed by the interaction strength and the plasmon distribution inside the wire .
both the zba and the dephasing effects are encoded in the second term of eq .
( [ inf_theory_gen ] ) . the third effect of the interaction which is specifically at the focus of the present work
is the inelastic scattering of electrons , leading to their redistribution over energies .
this effect takes place in those regions where the interaction strength varies in space ( near the wire boundaries in our model ) , inducing backscattering of plasmons ( but not of electrons ) .
this leads to relaxation of the electron distribution functions : left and right moving fermions `` partly exchange '' their distributions , see eqs .
( [ n_r ] ) , ( [ distr_func ] ) and figs
. [ fig3 ] , [ fig4 ] , [ fig5 ] . for slowly varying interaction ,
when the plasmons with relevant frequencies go through essentially without reflection , the energy relaxation of electrons is negligible . in the opposite limit ,
when the plasmons are almost entirely reflected ( due to strong and sharply switched interaction or , else , due to disordered boundary regions inducing the plasmon localization ) , the left- and right - movers essentially exchange their distribution functions ( but not their total density ) .
we have also discussed a connection between these results and earlier findings on the thermal conductivity of ll structures .
our results are important for the analysis of ts experiments on strongly correlated 1d structures ( in particular , carbon nanotubes @xcite ) out of equilibrium . in this connection ,
let us emphasize the following important point . what can actually be measured in experiment are green functions , @xmath17 and @xmath18 .
the tdos @xmath93 in the interacting part of the wire , as well as the distribution function @xmath172 in the non - interacting regions are related to @xmath17 and @xmath18 in a simple way . on the other hand , in order to extract the distributions @xmath105 and @xmath173 from @xmath19 in the _ interacting _ part of the wire , a non - trivial deconvolution procedure is necessary .
the broadening of ( split ) fermi - edge structures in @xmath19 in the interacting part of the wire is governed by both the distribution function and the dephasing .
the dephasing contributes to the smearing of fermi - edge singularities also in higher - dimensional ( diffusive ) systems @xcite , and should be taken into account for the accurate interpretation of corresponding experiments@xcite . in the 1d case the role of dephasing becomes particularly dramatic ( if the interaction is sufficiently strong ) .
this is very well illustrated by fig .
[ fig5 ] : two fermi - edge singularities almost ( middle panel ) or even completely ( lower panel ) merge , despite the fact that the fermi edges in the distribution functions remain well separated ( fig . [ fig4 ] ) .
a comment of a more general nature is in order here .
our results illustrate the fact that there is no unique answer to the question : `` how much is a ll different from a fermi liquid ? '' on one hand , the strong , power - law zba in tdos of a ll clearly distinguishes it from the fermi liquid
. in more formal terms , the single - particle residue @xmath174 , which is finite in the fermi liquid , vanishes in a power - law fashion at the fermi level of the ll . also the dephasing rate determining the broadening of zba , eq .
( [ tauphi ] ) , is linear in temperature , contrary to the fermi - liquid @xmath175 behavior .
one could think that it makes little sense to speak about fermionic excitations in this situation , but this is not the case .
first , the power - law vanishing of tdos has little importance ( like the value of @xmath174 in the fermi liquid ) for kinetic properties of the system .
second , the dephasing rate ( [ tauphi ] ) is governed by processes with zero energy transfer and do not lead to any energy relaxation . as a result ,
the distribution function of fermionic excitations , @xmath171 , is a fully meaningful concept even in the case of a strong interaction .
it stays preserved as long as the interaction is spatially constant ( or varies adiabatically slow with @xmath80 ) .
furthermore , both the charge and the energy current in the interacting part of the wire can be understood as carried by these fermionic quasiparticles . from this point of view
, the ll is a _
perfect _ fermi liquid .
we conclude the paper by reviewing some future research prospects ; the work in those directions is currently underway .
first , one may consider a more general non - equilibrium situation where the distribution functions `` injected '' into the interacting part of the wire are of non - equilibrium ( e.g. , double - step ) form by themselves @xcite , see setups b , c in fig.1 of ref . ;
the first of these setups is close to the experimental situation of ref . .
this requires a generalization of the bosonization technique that will be presented elsewhere @xcite .
second , it is interesting to study correlations between outgoing left- and right - movers . in a general situation
, one finds that their density matrices are not decoupled , i.e. they are entangled , which manifests itself , in particular , in current cross - correlations .
third , one may study the effect of a random variation of the interaction strength @xmath7 in the wire .
if the wire is sufficiently long , plasmons with not too low frequencies get localized . using our general results ,
one concludes that in the left ( right ) half of the wire both distributions @xmath155 , @xmath156 are determined by that of the left ( respectively , right ) reservoir , with a transition region which extends over the localization length of the middle section . to refine this picture
, one has to include into consideration also plasmons with low frequencies ( that remain delocalized ) .
also , including the spectral curvature will induce plasmon decay processes .
( in the context of thermal conductivity , this physics was discussed in ref . .
) finally , our results can be generalized to the case of chiral ll , where both branches move in the same direction , which is the situation characteristic for quantum - hall edge - state devices @xcite .
we thank d. bagrets , n. birge , a. finkelstein , i. gornyi , d. maslov , y. nazarov , d. polyakov , and r. thomale for useful discussions .
we are particularly grateful to the late yehoshua levinson for numerous illuminating discussions on the physics of non - equilibrium systems .
this work was supported by us - israel bsf , isf of the israel academy of sciences , the minerva foundation , and dfg spp 1285 ( yg ) , ec transnational access program at the wis braun submicron center ( adm ) , german - israeli foundation under grant 965 , and einstein minerva center .
the tunneling current between a probe and a quantum wire can be expressed in terms of the functions @xmath19 as @xmath176\,,\end{aligned}\ ] ] where the subscripts `` tp '' and `` w '' refer to the tunnel probe and the wire respectively , @xmath177 is a voltage between the tunneling probe and the wire , and @xmath178 is a tunneling matrix element in the coordinate representation . if electron tunneling is local in space , we have @xmath179 , where @xmath80 is a position of tunneling probe .
since the tunneling probe is at equilibrium , one can use a standard relation between the green functions and distribution function @xmath180 of electrons in the probe , @xmath181.\end{aligned}\ ] ] differentiating the tunneling current with respect to voltage and substituting eq .
( [ app_a_eq3 ] ) into eq .
( [ app_a_eq1 ] ) , one finds @xmath182 \nonumber \\ & & -2\pi i\nu_{{\rm tp}}(\epsilon - eu)\nu_{{\rm w}}(\epsilon)\frac{\partial n_{{\rm tp}}(\epsilon - eu)}{\partial \epsilon}\bigg\ } \ , .
\end{aligned}\ ] ] for a ll wire the green functions @xmath183 and the tdos @xmath184 represent a sum of contributions of both chiral branches . if the density of states in the tunneling probe ( @xmath185 ) is a constant ( as in a normal metal ) , the first term in eq .
( [ app_eq_2 ] ) drops out . in this case
the result is proportional to the tdos in the wire . assuming that the tunneling probe is at zero temperature , one then finds @xmath186 on the other hand , if the density of states in the tunneling probe is strongly energy dependent ( as for superconducting electrodes ) , the first term in eq .
( [ app_eq_2 ] ) survives .
unlike tdos ( which is determined by the difference @xmath187 ) , this term contains also the information about @xmath188 .
therefore , measurement of the tunneling current with two different types of tunneling probes ( normal and superconducting ) allows one to find functions @xmath189 and @xmath190 separately .
the idea to use superconducting electrodes for the tunneling spectroscopy was introduced in ref . and
more recently employed in ref . .
at thermal equilibrium the green functions in the energy domain can be calculated explicitly .
using eq .
( [ inf_theory_gen ] ) and @xmath191 , we find the exponent @xmath192 for the green functions in interacting part of the wire , @xmath193 where we drop the chirality index @xmath31 , as it is immaterial for @xmath79 in equilibrium .
using eq .
( [ eq_g ] ) and performing a fourier transform from the time into the energy domain , one finds @xmath194 after calculating an auxiliary integral @xmath195\bigr|^2 \,,\nonumber\end{aligned}\ ] ] one obtains @xmath196\bigr|^2 , \ ] ] where @xmath197 .
similarly , one finds the function @xmath18 , @xmath198\bigr|^2.\ ] ] this yields the following asymptotic behavior of the green function at low temperatures ( @xmath199 ) , @xmath200 and high temperatures ( @xmath201 ) , @xmath202 \left(\frac{\pi t}{\lambda}\right)^\gamma\ , \,.\ ] ] using eqs .
( [ app_nu ] ) , ( [ app3 ] ) , and ( [ app4 ] ) , one obtains tdos at equilibrium , @xmath203\biggr|^2 \cosh\frac{\pi z}{2}.\end{aligned}\ ] ] equation ( [ app6 ] ) describes the well - known zba in tdos , @xmath204 , smeared at the scale @xmath205 .
this smearing results from a combined effect of ( i ) the thermal broadening of the distribution function and ( ii ) the dephasing rate@xcite @xmath206 .
m. bockrath , d.h .
cobden , j. lu , a.g .
rinzler , r.e .
smalley , l. balents , and p.l .
mceuen , nature ( london ) * 397 * , 598 ( 1999 ) ; z. yao , h.w.ch .
postma , l. balents , and c. dekker , nature ( london ) * 402 * , 273 ( 1999 ) . h. c. fogedby , j. phys .
c * 9 * , 3757 ( 1976 ) ; d. k. lee and y. chen , j. phys .
a * 21 * , 4155 ( 1988 ) ; c.m .
naon , m.c .
von reichenbach , and m.l .
trobo , nucl . phys .
b * 435 * , 567 ( 1995 ) ; c.m .
naon , m.j .
salvay , and m.l .
trobo , int . j. mod
a * 19 * , 4953 ( 2004 ) ; i. v. yurkevich , in _ strongly correlated fermions and bosons in low - dimensional disordered systems _ , edited by i.v .
lerner , b.l .
altshuler , v.i .
falko , and t. giamarchi ; a. grishin , i.v .
yurkevich and i.v .
lerner , phys .
b. * 69 * , 165108 ( 2004 ) .
for review of the keldysh technique see , e.g. , j. rammer and h. smith , rev .
phys . * 58 * , 323 ( 1986 ) ; a. kamenev , in _ nanophysics : coherence and transport_(elsevier , 2005 ) , edited by h. bouchiat , y. gefen , g. montambaux , and j. dalibard , p. 177 .
maslov and m. stone , phys .
b * 52 * , r5539 ( 1995 ) ; v. ponomarenko , phys . rev .
b * 52 * , r8666 ( 1995 ) ; i. safi and h.j .
schulz , phys .
b * 52 * , r17040 ( 1995 ) ; y. oreg and a.m. finkelstein , phys . rev .
b * 54 * , r14265 ( 1996 )
. y. ji , y. chung , d. sprinzak , m. heiblum , d. mahalu , and h. shtrikman , nature * 422 * , 415 ( 2003 ) ; i. neder , f. marquardt , m. heiblum , d. mahalu , and v. umansky , nature phys . * 3 * , 534 ( 2007 ) ; i.p .
levkivsky and e.v .
sukhorukov , phys .
b * 78 * , 045322 ( 2008 ) . | we develop a theory of tunneling spectroscopy of interacting electrons in a non - equilibrium quantum wire coupled to reservoirs .
the problem is modelled as an out - of - equilibrium luttinger liquid with spatially dependent interaction .
the interaction leads to the renormalization of the tunneling density of states , as well as to the redistribution of electrons over energies .
energy relaxation is controlled by plasmon scattering at the boundaries between regions with different interaction strength , and affects the distribution function of electrons in the wire as well as that of electrons emitted from the interacting regions into non - interacting electrodes .
# 1eq .
( [ # 1 ] ) # 1([#1 ] ) |
the two pulsars in the recently discovered double pulsar binary system , and ( hereafter simply `` a '' and `` b '' ) have periods of @xmath1 ms and @xmath2 s. they are in a 2.4-hr mildly - eccentric orbit that we view nearly edge - on , with an inclination angle of @xmath3 ( burgay et al . 2003 ,
lyne et al . 2004 ) .
the phenomenology exhibited by this system is extremely rich .
the flux density of the b pulsar varies dramatically and systematically around the orbit ( lyne et al .
2004 , ramachandran et al .
2004 , mclaughlin et al .
2004 ) , indicating significant interaction between the two pulsars .
while the flux density of the a pulsar is constant for most of the orbit , for @xmath4 30 s around its superior conjunction , a is eclipsed by the magnetosphere of b ( lyne et al .
this eclipse is asymmetric , with the flux density of a decreasing more slowly on ingress than it increases on egress .
the duration of the eclipse is mildly dependent on radio frequency , with eclipses lasting longer at lower frequencies ( kaspi et al .
2004 ) .
theoretical models by arons et al .
( 2004 ) and lyutikov ( 2004 ) explain these eclipse properties in the context of synchrotron absorption of a s radio emission by the shock - heated plasma surrounding and containing b s magnetosphere . because b s magnetosphere will respond differently to the pressure of the relativistic wind of a at different rotational phases of b , one prediction of these models is that the density of absorbing plasma , and hence the properties of a s eclipse , will depend on the rotational phase of b. in their early analysis of the exploratory observations performed at the 100-m green bank telescope ( gbt ) in 2003 december and 2004 january , kaspi et al .
( 2004 ) found no evidence for modulation of the eclipse properties with b s rotational phase .
we have analyzed this publically available dataset and , in this letter , show that the eclipse properties of a are in fact strongly dependent upon the rotational phase of b.
the double pulsar system j07373039 was observed at the gbt in 2003 december and 2004 january using receivers centered at at 427 , 820 and 1400 mhz . the 5-hr 820-mhz observation discussed here
was acquired with the gbt spectrometer spigot card with a sampling time of 40.96 @xmath5s and 1024 synthesized frequency channels covering a 50-mhz bandwidth . for further details of the observations and data acquisition system , see ransom et al .
( 2004 ) and references therein .
the data were dedispersed and folded using freely available software tools ( lorimer 2001 ) assuming the nominal dispersion measure of 48.9 @xmath6 pc ( burgay et al .
2003 ) and using ephemerides for a and b from lyne et al .
we folded the data with 512 phase bins and , for each a pulse , calculated the mean pulsed flux density in two on - pulse phase windows centered on the two pulse components ( burgay et al .
2003 ) . from these , we subtracted the baseline level calculated from two off - pulse windows .
the on- and off - pulse regions contained 95 and 320 phase bins , respectively .
for each orbit , we created a light curve of a s pulsed flux density with time by averaging every 12 pulses , for an effective time resolution of @xmath4 0.27 s. orbital phases were calculated using the modified julian dates of pulse arrival at the solar - system barycenter and the pulsar ephemeris .
analysis of these light curves reveals that , for the majority of the orbit , the pulsed flux density of a is constant within measurement uncertainties , exhibiting no obvious orbital phase - dependent variations .
however , close to superior conjunction ( i.e. orbital phase 90@xmath7 , where orbital phase is defined as the longitude from the ascending node ) , the pulsed flux density of a varies dramatically .
= 9truecm figure [ fig : eclipse1 ] shows the light curves of a for all three eclipses included in the 820-mhz observation . in the lower panel ,
we also show a composite light curve , averaged over all three eclipses and over 100 a pulses , for an effective time resolution of @xmath4 2.3 s. our measurements of the eclipse duration and of ingress and egress shapes from this composite light curve are consistent with the results of kaspi et al .
( 2004 ) , who averaged data in 2 s intervals .
they calculated an eclipse duration ( defined by the full width at half maximum of the light curve ) of 27 s , which was used to place a limit of 18,600 km on the size of the eclipsing region .
they also found that eclipse ingress takes roughly four times longer than egress and that the eclipse duration is frequency dependent , lasting longer at lower frequencies .
= 9truecm inspection of the top three panels of figure [ fig : eclipse1 ] shows that the rise and fall of the flux density during eclipse is not monotonic .
in fact , up until orbital phases approaching 90@xmath7 , the light curves of a show amplitude peaks consistent with pre - eclipse levels .
figure [ fig : eclipse2 ] shows an expanded view of figure [ fig : eclipse1 ] covering a smaller range of orbital phases centered on the eclipse . on this plot , we also indicate the measured barycentric arrival times of the pulses of the 2.8 s pulsar b. this demonstrates clearly that the pulsed flux density of a is modulated in synchronism with @xmath8 . in all three
eclipses , we see negative dips in pulsed flux density occurring first at a time about 0.5 @xmath9 out of phase with the b pulses , and later also at a time in phase with the b pulses . in order to determine
more sensitively how the flux density of a depends upon the rotational phase of b , we shifted the light curves of the second and third eclipses by up to @xmath10 so that the phases of the b pulses were identical to those of the first eclipse .
we then summed the light curves of all three eclipses to create the bottom panel of figure [ fig : eclipse2 ] . dividing each 2.8 s window of b s rotational phase into four equal regions
, we can calculate an average light curves for each region , as shown in figure [ fig : eclipse3 ] .
the light curves for each of the b pulse phase windows vary smoothly .
eclipse durations range from 20 s to 34 s for the four b phase windows , calculated , as in kaspi et al .
( 2004 ) , to be the full width at half maximum of the eclipse .
ingress occurs first at b phase 0.5 and then at 0.0 , when , assuming b is an orthogonal rotator , the magnetic axis of b is aligned with the line of sight to a. it occurs later and more gradually for the other two phase windows centered on phases 0.25 and 0.75 , when the magnetic axis of b is at right angles to the line of sight .
the orbital phases of egress are similar for all b pulse phases . for light curves at b phases centered on 0.0 and 0.5 ,
the eclipses are essentially symmetric , with minimum beginning at orbital phases near 89.4@xmath7 and ending at orbital phases near 90.6@xmath7 .
the light curves for b phases centered on 0.25 and 0.75 are less symmetric , with the flux density minimum occurring briefly , around orbital phases @xmath11 . only at around orbital phase 90.5@xmath7
is flux density of all phases of b consistent with being zero .
it is important to recognize that the features we see in the 820-mhz data can not be due to confusion with the single pulses of b ( e.g. mclaughlin et al .
2004 ) since the pulsed features seen in fig .
2 are not in phase with the radio pulses of b , and , in addition , the latter , which have width of @xmath12 ms , would be removed by the on - pulse / off - pulse subtraction procedure used to determine a s flux density .
our analysis of these data differs from that of kaspi et al .
( 2004 ) in one significant respect . while kaspi et al .
( 2004 ) calculated light curves with 2 s time resolution , we have integrated over only 12 a pulses , affording us @xmath4 7 times better time resolution .
kaspi et al .
( 2004 ) did detect some emission from a on short time scales during eclipse , but their 2 s integration time was not sufficient to detect the b modulation .
we have shown that the eclipse duration varies considerably with b phase and , hence , that the properties of the occulting medium are in fact different in different regions of b s magnetosphere .
the observed eclipse phenomenology can be understood in the context of a model invoking synchrotron absorption of the radio emission from a in a magnetosheath surrounding b s rotating magnetosphere ( arons et al .
2004 , lyutikov 2004 ) .
much of the rotational energy of the faster - spinning a pulsar is emitted in the form of a relativistic , magnetized wind composed primarily of electron - positron pairs .
the rotational energy loss rate of a ( @xmath13 erg s@xmath14 ) is @xmath4 3600 times greater than that of the slower b pulsar .
because the dynamic pressure of a s wind is balanced against b s magnetic pressure only at a point well inside the light cylinder of b , a s wind confines the b magnetosphere on the side facing a , with a long , magnetospheric tail extending behind b. the collision of a s wind with the magnetosphere of b leads to the formation of a bow shock consisting of hot , magnetized plasma .
this plasma will surround the cometary shaped magnetosphere of b , creating what arons et al .
( 2004 ) call a `` magnetosheath '' . in figure
[ fig : eclipse4 ] , we show a schematic of the interaction of a s relativistic wind with b s magnetosphere . the rotation of b inside the magnetosheath is expected to modulate the shape of the sheath and to produce cusps within it , leading to higher plasma densities in the cusps .
note that the plasma will not reach all the way to the polar caps of b , but is confined at the `` magnetopause '' , where the plasma pressure equals the magnetic pressure .
an excellent description of the shape of the magnetosphere and of the polar cusps , is given in arons & lea ( 1976 ) .
as already discussed qualitatively by kaspi et al .
( 2004 ) and quantitatively by arons et al .
( 2004 ) and lyutikov ( 2004 ) , synchrotron absorption in such a magnetosheath can explain many of the properties of the a eclipse , including the sharp edge and the mild frequency dependence .
if b is rotating prograde to the orbital velocity , it can also explain the asymmetry between eclipse ingress and egress .
the observed changes in eclipse properties with b pulse phase give further support to the model .
the pulses from a are first absorbed at the two b pulse phases ( i.e. 0.0 and 0.5 ) at which the magnetic axis of b is aligned with the line of sight to a. this can be attributed to the higher density of absorbing plasma in b s polar cusps .
note that the pulses at b pulse phase 0.50 ( i.e. when the radio - bright pole is pointed towards a ) are absorbed before those at b pulse phase 0.0 . when the line of sight to a is at a right angle to b s magnetic axis , the pulses of a travel through regions of higher plasma density later in the orbit and start fading closer to superior conjunction , where the opacity of the magnetosheath tail could also become predominant . as shown in figure [ fig : eclipse3 ] , the ingress phase and eclipse duration for these two b phases ( i.e. 0.25 and 0.75 ) differ , with absorption happening later at phase 0.75
. this may be due to the rotationally - induced asymmetry predicted by arons et al .
( 2004 ) .
the above discussion , and the simulations of arons et al . ( 2004 ) , assume that b is an orthogonal rotator .
as demorest et al .
( 2004 ) and arons et al .
( 2004 ) discuss , it is quite likely that this is the case as the wind torque from a should have aligned b s rotation axis with the orbital angular momentum .
furthermore , it is very difficult to explain the similarity between the light curves at b pulse phases 0.0 and 0.5 and phases 0.25 and 0.75 if b is not nearly orthogonal . from the measured eclipse durations ,
it is clear that the size of the eclipsing region is much smaller than the light cylinder of b. however , because the plasma density varies greatly depending on the rotational phase of b , determining the extent and density of the eclipsing region is not trivial and will require detailed modeling of the plasma density variations within b s magnetosphere . due to the effects of relativistic periastron advance and geodetic precession , the durations and morphologies of the eclipse and their dependence on b
rotational phase should change significantly with time .
forthcoming observations will allow us to sensitively probe the dynamics of the wind / magnetosphere interaction and of the geometry of the system .
these observations , at multiple frequencies , will also allow us to test the hypothesis of arons et al .
( 2004 ) that there should be no eclipse at frequencies above 5 ghz , already supported by the measurements of kaspi et al . (
2004 ) . because pulsars emit a large portion of their spin - down energy in the form of a relativistic wind , understanding these winds is crucial for forming a complete picture of the pulsar energy budget
however , because the winds can not be directly observed , progress in this field has been limited to studies of pulsar wind nebulae and bow shocks .
if the eclipses of a are indeed caused by synchrotron absorption in b s magnetosheath , both arons et al .
( 2004 ) and lyutikov ( 2004 ) reach the very interesting conclusion that the density of a s wind must be at least four orders of magnitude greater than is expected given currently accepted models of pair creation .
the arons et al .
( 2004 ) model also implies a very low wind magnetization .
however , in our paper describing the drifting features observed in the single pulses of b ( mclaughlin et al .
2004 ) , we concluded that that most of the spin - down energy is carried by the poynting flux rather than by energetic particles .
clearly , self - consistent models which can explain both of these phenomena are needed .
the rich phenomenology offered by the double pulsar system will offer us new ways to explore the physics of these winds on much smaller scales than has been possible with other systems .
eclipse measurements of the double pulsar system may also be important for constraining pulsar pair creation scenarios .
we thank a. spitkovsky for a very useful discussion which motivated us to look for this effect and d. melrose , f. graham smith and j. arons for helpful discussions .
we thank the national radio astronomy observatory for making these observations publically available .
the national radio astronomy observatory is facility of the national science foundation operated under cooperative agreement by associated universities , inc .
ihs holds an nserc ufa and is supported by a discovery grant .
drl is a university research fellow funded by the royal society .
nda , ap and mb received support from the italian ministry of university and research ( miur ) under the national program _
cofin 2003_. arons , j. & lea , s. m. 1976 , , 207 , 914 arons , j. , backer , d. c. , spitkovsky , a. , & kaspi , v. m. 2004 , proceedings of the 2004 aspen winter conf .
on astrophsics `` binary radio pulsars '' , asp conf .
series , ( eds . )
f. rasio & i. stairs ( astro - ph/0404159 ) burgay , m. et al .
2003 , nature , 426 , 531 demorest , p. et al , 2004 , apj , submitted ( astro - ph/0402025 ) kaspi , v. m. et al .
2004 , apj , 613 , 137 , d. r. 2001 , arecibo technical memo no .
200101 , see also ` http://www.jb.man.ac.uk/~drl/sigproc ` lyne , a. g. et al .
2004 , science , 303 , 1153 lyutikov , m. 2004 , mnras , 353 , 1095 mclaughlin , m. a. et al .
2004 , apj , 613 , 57 ramachandran , r. et al .
2004 , apj , accepted ( astro - ph/0404392 ) ransom , s. m. et al .
2004 , apj , 609 , l71 | we have investigated the eclipse of the 23 ms pulsar , psr j0737@xmath03039a , by its 2.8 s companion psr j0737@xmath03039b in the recently discovered double pulsar system using data taken with the green bank telescope at 820 mhz .
we find that the pulsed flux density at eclipse is strongly modulated with the periodicity of the 2.8 s pulsar .
the eclipse occurs earlier and is deeper at those rotational phases of b when its magnetic axis is aligned with the line of sight than at phases when its magnetic axis is at right angles to the line of sight .
this is consistent with the eclipse of a being due to synchrotron absorption by the shock - heated plasma surrounding b , the asymmetry arising from the higher plasma densities expected in the b magnetosphere s polar cusps . |
after many years of investigating hadron - hadron and heavy ion collisions , the study of hadron production remains an active and important field of research .
the lack of detailed knowledge of the microscopic mechanisms has led to the use of many different models , often from completely opposite directions .
thermal models , based on statistical weights for produced hadrons @xcite , are very successful in describing particle yields at different beam energies @xcite , especially in heavy ion collisions .
these models assume the formation of a system which is in thermal and chemical equilibrium in the hadronic phase and is characterised by a set of thermodynamic variables for the hadronic phase .
the deconfined period of the time evolution dominated by quarks and gluons remains hidden : full equilibration generally washes out and destroys large amounts of information about the early deconfined phase .
the success of statistical models implies the loss of such information , at least for certain properties , during hadronization .
it is a basic question as to which ones survive the hadronization and behave as messengers from the early ( quark dominated ) stages , especially if these are strongly interacting stages . in the case of full thermal and chemical equilibrium ,
relativistic statistical distributions can be used , leading to exponential spectra for the transverse momentum distribution of hadrons . on the other hand , experimental data at sps and rhic energies display non - exponential behaviours at high @xmath1 .
one explanation of this deviation is connected to the power - like hadron spectra obtained from perturbative qcd descriptions : the hadron yield from quark and gluon fragmentation overwhelms the thermal ( exponential ) hadron production . however
, this overlap is not trivial .
one can assume the appearance of near - thermal hadron distributions , which is similar to the thermal distribution at lower @xmath1 , but it has a non - exponential tail at higher @xmath1 .
a stationary distribution of strongly interacting hadron gas in a finite volume can be characterized by such a distribution ( or strongly interacting quark matter ) , which will hadronize into hadron matter .
tsallis distributions satisfy such criteria @xcite . in the next section
we will review the tsallis distribution and emphasize the properties most relevant to particle yields .
neglecting quantum statistics , the entropy of a particle of species @xmath2 is given by @xcite @xmath3 where the mean occupation numbers , @xmath4 , are given by @xmath5 with @xmath6 being the degeneracy factor of particle @xmath2 .
the total number of particles of species @xmath2 is given by an integral over phase space of eq .
( [ boltzmann ] ) : @xmath7 the transition to the tsallis distribution makes use of the following substitutions @xcite @xmath8^{1\over 1-q } , \label{subb}\end{aligned}\ ] ] which leads to the standard result @xcite @xmath9^{-{1\over q-1 } } , \label{tsallis}\ ] ] which is usually referred to as the tsallis distribution @xcite .
as these number densities are not normalized , we do not use the normalized @xmath10-probabilities which have been proposed in ref .
@xcite . in the limit
where @xmath11 this becomes the boltzmann distribution : @xmath12 the particle number is now given by @xmath13 note that @xmath14 is the maximum value that still leads to a convergent integral in eq .
( [ tsallis_number ] ) .
a derivation of the tsallis distribution , based on the boltzmann equation , has been given in ref .
@xcite . a comparison between the two distributions is shown in fig .
( [ tsallis_boltzmann ] ) , where it can be seen that , at fixed values of @xmath15 and @xmath16 , the tsallis distribution is always larger than the boltzmann one if @xmath17 .
taking into account the large @xmath1 results for particle production we will only consider this possibility in this paper . as a consequence , in order to keep the particle yields the same , the tsallis distribution always leads to smaller values of @xmath15 for the same set of particle yields . , keeping the temperature and chemical potential fixed , for various values of the tsallis parameter @xmath10 .
the value @xmath14 is the maximum value that still leads to a convergent integral in eq .
[ tsallis_number ] .
chemical potential @xmath16 , keeping the temperature and the energy fixed . ,
title="fig:",scaledwidth=47.0%,height=377 ] , keeping the temperature and chemical potential fixed , for various values of the tsallis parameter @xmath10 .
the value @xmath14 is the maximum value that still leads to a convergent integral in eq .
[ tsallis_number ] .
chemical potential @xmath16 , keeping the temperature and the energy fixed .
, title="fig:",scaledwidth=47.0%,height=377 ] the dependence on the chemical potential is also illustrated on the right of fig .
[ tsallis_boltzmann ] for a fixed temperature @xmath15 and a fixed energy @xmath18 . as one can see
, the tsallis distribution in this case increases with increasing @xmath10 .
the tsallis distribution for quantum statistics has been considered in ref .
the parameter @xmath10 plays a central role in the tsallis distribution and a physical interpretation is needed to appreciate its significance . to this end
we follow the analysis of ref .
@xcite and write the tsallis distribution as a superposition of boltzmann distributions @xmath19 where the detailed form of the function @xmath20 is given in @xcite .
the parameter @xmath21 is the standard temperature as it appears in the boltzmann distribution .
it is straightforward to show @xcite that the average value of @xmath22 is given by the tsallis temperature : @xmath23 while the fluctuation in the temperature is given by the deviation of the tsallis parameter @xmath10 from unity : @xmath24 which becomes zero in the boltzmann limit .
the above leads to the interpretation of the tsallis distribution as a superposition of boltzmann distributions with different temperatures .
the average value of these ( boltzmann ) temperatures is the temperature @xmath15 appearing in the tsallis distribution .
this is the interpretation of the tsallis temperature that we will follow .
the other parameter in the tsallis distribution , @xmath10 , describes the spread around the average value of the ( boltzmann ) temperature @xmath15 .
for @xmath25 we have an exact boltzmann distribution , for values of @xmath10 which deviate from 1 , we have a corresponding deviation . from this point of view
the tsallis distribution describes a distribution of ( boltzmann ) temperatures .
a deviation from @xmath25 means that a spread of temperatures is needed instead of a single value .
/d.o.f . of the fits as a function of the tsallis parameter @xmath10.,scaledwidth=80.0%,height=377 ] in order to identify the energy dependence of the deviation from ideal gas behaviour , thermal fits were performed on yields measured at the cern sps in central pb - pb collisions at 40 agev , 80 agev and 158 agev ( using the same data as analyzed in @xcite ) and yields measured at rhic in central au - au collisions at @xmath26 = 130 agev ( using the same data as analyzed in @xcite ) and at @xmath26 = 200 agev . in the cern
sps fits , the thermal parameters @xmath15 , @xmath27 , @xmath28 and @xmath29 were fit to the data , while @xmath30 and @xmath31 were fixed by the initial baryon - to - charge ratio and strangeness content in the colliding system , respectively . in the case of the rhic analysis
we again fit @xmath15 , @xmath27 , @xmath31 and @xmath28 to the data .
the use of mid - rapidity data here led to the relaxing of the constraints on @xmath31 and @xmath30 typical in analyzes of 4@xmath32 data .
instead , @xmath30 was set to zero as justified by the observed @xmath33 ratio .
the following expression was used to calculate primordial particle yields , @xmath34^{-1/(q-1)},\ ] ] where @xmath35 is the number of valence strange quarks and anti - quarks in species @xmath2 .
the value @xmath36 obviously corresponds to complete strangeness equilibration .
all calculations were done using the thermus package @xcite .
the most surprising result of our analysis is shown in fig .
( [ chi ] ) : the quality of the fits , as measured by the @xmath37 , improves at first as the tsallis parameter @xmath10 increases . it reaches a minimum value around @xmath38 1.07 for sps beam energy of 158 agev .
this behaviour is repeated at other sps energies with the minima at slightly different values of @xmath10 , i.e. 1.08 for 80 agev and about 1.05 at 40 agev beam energy .
this behaviour is not seen at rhic energies .
clearly , changes in the tsallis parameter @xmath10 have only a small negligible effect on the @xmath0 values at rhic energies , of course , this still leaves open the possibility for @xmath10 values larger than 1 @xcite .
however on the sps data the effect is substantial and changes the interpretation substantially .
one possible interpretation is that at sps energies fluctuations in the freeze - out temperature are substantial .
recently @xcite a coalescence model with a tsallis distribution for quarks was used to fit the transverse momenta spectra measured at rhic .
this fit does not include decays from resonances and therefore can not be compared directly to ours since decays can substantially modify the transverse momenta , also the emerging hadrons are not in a tsallis - type equilibrium gas , which is an assumption of the present analysis .
the authors obtained values for the tsallis parameter @xmath10 which are remarkably similar for all particle species considered , i.e. @xmath39 which can not be excluded by our analysis .
the freeze - out temperature @xmath15 decreases , as expected , with increasing values of @xmath10 .
this can be understood from the fact that the tsallis distribution is always larger than the boltzmann one ( as long as @xmath17 ) .
hence , in order to match the same particle yields one has to adjust @xmath15 to lower values .
this is seen at all energies in fig .
however , the drop in @xmath15 , turns out to be quite drastic numerically .
in fact , the decrease in particle numbers has to be compensated by increases in all other thermodynamic variables .
the ( modest ) increase in the baryon chemical potential is shown in fig .
[ temp ] on the right hand side .
the strangeness non - equilibrium factor @xmath40 as shown in fig .
[ gammas ] .
it is interesting to note that the tsallis distribution leads to a much better chemical equilibrium than the corresponding boltzmann distribution with @xmath41 . in all cases
considered the @xmath40 is very close to the chemical equilibrium value of 1 .
.,title="fig:",scaledwidth=47.0%,height=377 ] .,title="fig:",scaledwidth=47.0%,height=377 ] as a function of the tsallis parameter @xmath10.,scaledwidth=80.0%,height=377 ] clearly , the use of the tsallis distribution in relativistic heavy ion collisions calls for a reevaluation of the understanding gained from previous analyses @xcite .
the authors acknowledge support from the hungary - south africa scientific cooperation programme and hungarian otka grant nk62044 .
we acknowledge useful discussions with h.g .
miller , t.s .
bir , p. vn and g.g . barnafldi .
10 e. fermi , progress theor .
phys . , * 5 * ( 1950 ) 570 .
pomeranchuk , dokl .
78 ( 1951 ) 889 .
w. heisenberg , naturwissenschaften , * 39 * ( 1952 ) 69 .
r. hagedorn , nuovo cimento , * 35 * ( 1965 ) 395 .
j. cleymans , h. oeschler , k. redlich , s. wheaton , phys .
c * * 73 ( 2006 ) 034905 .
j. cleymans , p. braun - munzinger , h. oeschler and k. redlich , nucl .
phys . a * 697 * ( 2002 ) 902 .
a. andronic , p. braun - munzinger , j. stachel , nucl .
phys . a * 772 * ( 2006 ) 167 .
f. becattini , j. manninen , m. gadzicki phys .
c * 73 * ( 2006 ) 044905 .
see e.g. s.r .
de groot , w.a .
van leeuwen , ch.g .
van weert , _ relativistic kinetic theory _ , north holland 1980 . c. tsallis , j. stat
. phys . * 52 * ( 1988 ) 479 . c. tsallis , r.s .
mendes , a.r .
plastina , physics a * 261 * ( 1998 ) 534 .
bir , g. purcsel , phys .
* 95 * ( 2005 ) 162302 .
a.m. teweldeberhan , a.r .
plastino , h.g .
miller , phys .
a * 343 * ( 2005 ) 71 .
plastino , a. plastino , h.g .
miller , phys .
lett . a * 343 * ( 2005 ) 71 .
f. buyukkilic , d. demirhan , phys .
lett . a * 181 * ( 1993 ) 24 .
f. buyukkilic , d. demirhan , a. gulec , phys .
lett . a * 197 * ( 1995 ) 209 .
g. wilk and z. wlodarczyk , phys . rev
* 84 * ( 2000 ) 2770 .
s. wheaton , j. cleymans , m. hauer , computer physics communications * 180 * ( 2009 ) 84 .
j. cleymans , b. kmpfer , m. kaneta , s. wheaton and n. xu , phys .
c * 71 * ( 2005 ) 054901 .
bir , k. rmssy k and g.g .
barnafldi , j. phys .
g * 35 * 044012 t.s .
bir , k. rmssy , arxiv:0812.2985 [ hep - ph ] . | hadron yields in high energy heavy ion collisions have been fitted with thermal models using standard ( extensive ) statistical distributions .
these models give insight into the freeze - out conditions at varying beam energies and lead to a systematic consistent picture of freeze - out conditions at all beam energies . in this paper
we investigate changes to this analysis when the statistical distributions are replaced by non - extensive tsallis distributions for hadrons .
we investigate the particle yields at sps and rhic energies and obtain better fits with smaller @xmath0 for the same hadron data , as applied earlier in the thermal fits for sps energies but not for rhic energies . |
transport and diffusion of microscopic particles through periodic potentials is a rich field of research from both fundamental and technological points of view @xcite .
investigation of the particle motion along ordered @xcite or disordered @xcite energy landscapes helps to better understand the dynamics in more complex situations , such as abrikosov @xcite and josephson vortices in superconductors @xcite , cell migration @xcite , or transport of molecular motors @xcite .
moreover , a periodic potential can be used to perform precise particle sorting and fractionation @xcite , thus , being of significant impact in diverse fields in analytical science and engineering which make use of microfluidic devices .
colloidal systems provide an ideal opportunity to investigate different transport scenarios , because of having particle sizes in the visible wavelength range and dynamical time scales which are experimentally accessible . in order to force colloidal particles to move along periodic or random trajectories ,
static potentials can be readily realized by using optical @xcite , magnetic @xcite , or electric fields @xcite .
dynamic landscapes ( obtained by periodically or randomly modulating the potential ) are a subject of growing interest since a rich dynamics can be induced due to the presence of competing time scales . moreover , flashing potentials where static landscapes are modulated in time , are usually employed to study molecular systems @xcite or as an efficient way to transport and fractionate brownian species @xcite . here
we present a combined experimental and theoretical study focused on the dynamics of microscale particles driven above a flashing magnetic potential .
this potential is generated by periodically modulating the energy landscape created by an array of magnetic bubbles . as it was previously reported @xcite ,
this experimental system is able to generate different dynamical states depending on the applied field parameters , such as frequency or amplitude of the external field .
in particular we report the observation of enhanced diffusive dynamics at high field strength @xmath0 , while the motion at lower @xmath0 is subdiffusive with a crossover to normal diffusion at long times . at low values of @xmath0 , the lattice structure is slightly disordered during the switching of the magnetic field direction .
the resulting randomness is dynamic , i.e. not reproducible after a cycle of external drive , and enhances with decreasing @xmath0 . by means of numerical simulations
we verify that in the presence of lattice disorder , the particles frequently experience oscillatory motion in local traps .
the back and forth motion of particle in local traps happens more frequently as the dynamic disorder in the structure of the magnetic lattice increases .
the trapping events change the statistics of the turning angles of the particle from an isotropic distribution ( limited to the directions allowed by the lattice structure ) to an anisotropic one with a tendency towards backward directions . using a persistent random walk model ,
we show that anomalous diffusion arises when the turning - angle distribution of a random walker is asymmetric along the arrival direction . when the walker has the tendency towards backward directions , the resulting antipersistent motion is subdiffusive or even strongly oscillatory at short time scales .
however , the walker has a finite range memory of the successive step orientations , i.e. the direction gets randomized after long times and the asymptotic behavior is ordinary diffusion with a smaller long - term diffusion coefficient compared to an ordinary random walk .
we obtain good agreement between the analytical predictions , simulations , and experiments .
the paper is organized in the following manner : first we introduce the setup in sec . [
sec : setup ] .
section [ sec : expresults ] contains the experimental results obtained at different field parameters . in sec .
[ sec : simulation ] , the results of numerical simulations for transport in the presence of dynamic disorder are discussed and compared with the corresponding experimental data . the motion of particles is modeled at the level of individual steps via an antipersistent random walk approach in sec .
[ sec : prw ] , and finally sec .
[ sec : conclusions ] concludes the paper .
a schematic illustrating the experimental system is shown in fig .
[ fig:1](a ) .
the colloidal particles used are polystyrene paramagnetic microspheres ( dynabeads m-270 , invitrogen ) having diameter @xmath1 and magnetic volume susceptibility @xmath2 .
the particles are diluted in high - deionized water and let sediment above the periodic potential generated by a bismuth - substituted ferrite garnet film ( fgf ) .
the fgf has composition y@xmath3bi@xmath4fe@xmath5ga@xmath6o@xmath7 ( @xmath8 $ ] ) and was previously grown by dipping liquid phase epithaxy on a @xmath9 thick gadolinium gallium garnet substrate @xcite . the film has thickness @xmath10 and saturation magnetization @xmath11 . in the absence of external field , this fgf is characterized by a labyrinth of stripe domains with alternating magnetization and a spatial periodicity of @xmath12 .
this pattern is converted into a periodic lattice of cylindrical magnetic domains by using high frequency magnetic field pulses applied perpendicular to the film , with amplitude @xmath0 and oscillating at angular frequency @xmath13 . as shown in fig .
[ fig:1 ] , the cylindrical domains , also known as magnetic bubbles " @xcite , are ferromagnetic domains with radius @xmath14 , having the same magnetization direction and arranged into a triangular lattice with lattice constant @xmath15 .
we can visualize both the magnetic domains in the film and the particles using polarization microscopy , due to the polar faraday effect .
the external oscillating field is obtained by connecting a coil perpendicular to the film plane ( @xmath16-direction ) with a wave generator ( tt_i _ - tga1244 ) feeding a power amplifier ( img sta-800 ) .
the custom - made coil was mounted on the stage of a polarization light microscope ( nikon , e400 ) equipped with a @xmath17 , @xmath18na objective and a @xmath19 tv lens .
the movements of the particles are recorded at @xmath20 fps for @xmath21 min with a @xmath22 ccd camera inside an observation area of @xmath23 .
we then use particle tracking routines @xcite to extract the particle trajectories and later calculate correlation functions .
once deposited above the garnet film , the paramagnetic particles pin to the bloch walls , which are located at the boundary of the magnetic bubbles . to generate a two - state flashing potential
we apply to the fgf film a square - wave modulation of type @xmath24 where @xmath0 is the field amplitude , @xmath13 the angular frequency , and @xmath25 denotes the sign function .
the applied field periodically changes the radii of the bubbles in the fgf , increasing ( decreasing ) the size of the bubbles when it is parallel ( resp .
antiparallel ) to their magnetization direction , thus alternating between two distinct states .
one can understand the effect of the applied field on the energy landscape by calculating the magnetostatic potential at the particle elevation @xcite see fig .
[ fig:1](c ) .
in particular , when the field expands the bubbles ( _ state 1 _ ) , the energy displays a paraboloid - like minimum within the magnetic domains .
thus , the magnetic colloids are attracted towards the center of the bubble domains .
when the field has opposite polarity , it shrinks the size of the bubbles and enlarges the interstitial region ( _ state 2 _ ) .
in this situation , the magnetostatic potential features six regions of energy minima with triangular shape at the vertices of the wigner - seitz cell around each bubble . as a consequence , during the transition _
1_@xmath26_2 _ , a particle can jump to @xmath27 possible places [ fig .
[ fig:1](c ) , left ] , while in the transition _
2_@xmath26_1 _ the possibilities reduce to @xmath28 [ fig .
[ fig:1](c ) , right ] .
since we apply a field perpendicular to the sample plane ( no tilt ) , the potential preserves its spatial symmetry , i.e. there is no net drift motion as it was induced in ref .
[ 17 ] by using a precessing field .
we analyze the particle dynamics by measuring the mean squared displacement via a temporal moving average @xmath29 . here , @xmath30 denotes the position of the particle projected along one of the crystallographic axes and @xmath31 the exponent of the power - law which is used to distinguish the diffusive ( @xmath32 ) from subdiffusive ( @xmath33 ) dynamics . in our system
we observe both types of dynamics , which are discussed in the following subsections . in fig .
[ fig:2](a ) we compare the msd for a paramagnetic colloid freely diffusing on a glass plane in the absence of fgf film and the one which is strongly driven by the flashing potential .
both cases exhibit a normal diffusion but with different diffusion coefficients . from the experimental data of the msd
, the effective diffusion coefficient of the particle can be estimated as @xmath34 , with @xmath35 being the spatial dimension ( here @xmath36 since the data is projected along one of the crystallographic axes ) .
we find that @xmath37 in the presence of the flashing potential ( @xmath38 ) is enhanced by nearly two orders of magnitude with respect to the one on a glass plate ( @xmath39 ) . in this regime of motion , which is observed for field amplitudes @xmath40 \ , { \rm m_s}$ ] and frequency @xmath41 , the particle moves from one domain to the next without performing continuous oscillation around a single site , thus , it performs an ordinary random walk on a triangular lattice .
the step length @xmath42 of the walker can be estimated to be given by one side of the wigner - seitz cell , i.e. @xmath43 , and the component of the msd along a crystallographic axis equals @xmath44 .
each step takes a half - period of the magnetic field pulse , thus , the duration of each step is given by @xmath45 .
one thus obtain the diffusion coefficient as @xcite @xmath46 .
we confirm this relation in fig .
[ fig:2](b ) by measuring @xmath37 versus the frequency and amplitude of the applied field .
while @xmath37 increases linearly with frequency , it decrease rapidly beyond @xmath47 since the overdamped particle is unable to follow the fast vibrations of the potential , and reduces significantly its diffusive dynamics .
in contrast to frequency , @xmath37 is almost independent of the amplitude of the applied field , since the lattice constant of the magnetic bubble array does not change significantly for amplitude @xmath48 @xcite . beyond this value , however , the magnetic bubble lattice starts melting and transport of the particles is not possible anymore . our experimental setup allows us to independently vary both the amplitude @xmath0 and the frequency @xmath13 of the driving field . at strong fields ( @xmath49 ) , the particle jumps between nearest bubbles with an equal probability to choose any of the possible movements ,
thus , performs a normal random walk on a triangular lattice ( @xmath50 on all time scales ) .
in contrast , when the field is weak ( @xmath51 ) , the landscape deformations induced by the applied field are small and the particle is unable to leave the magnetic bubble .
the corresponding particle trajectory is composed of simple oscillations between the center of one bubble and one of its six surrounding energy minima .
in such a pure localization the msd saturates rapidly , leading to an exponent @xmath52 . in the intermediate regime of amplitudes ,
i.e. @xmath53 \ , { \rm m_s}$ ] , we observe a subdiffusive dynamics at short time scales with a crossover to normal diffusion at long times .
figure [ fig:3 ] shows several examples of the temporal evolution of msd for different values of @xmath0 and @xmath13 .
the initial anomalous exponent varies between @xmath54 and @xmath55 depending on the strength @xmath0 of the applied magnetic field .
while in a previous work @xcite it was found a stable subdiffusive exponent @xmath56 for @xmath57 and at different frequencies , the value of @xmath31 may vary by changing the amplitude of the switching field . to see this more clearly
, we rescale both axes in such a way that the asymptotic diffusive regimes of the curves collapse on a master curve ( see inset ) .
this can be achieved if the msd is scaled by @xmath58 ( @xmath59 being the long - term diffusion coefficient , @xmath60 the step size , and @xmath61 the velocity of the particle ) , and the time axis is scaled by @xmath62 to synchronize the oscillations . only the samples which reach the long - term diffusion limit within our experimental time window
are considered . with increasing @xmath0 ,
the curves are initially more steep and converge faster to the asymptotic limit . and frequency @xmath13 of the applied magnetic field .
inset : the dimensionless mean squared displacement @xmath63 versus the scaled time @xmath64 . ] indeed , the lattice structure is slightly disordered at weak fields , because of the non - uniform deformations of the magnetic bubbles during their periodic expansion / contraction .
these deformations arise from the presence of pinning sites in the film and other inhomogeneities ( such as dislocations or magnetic and non - magnetic inclusions present in the fgf crystal ) .
these defects exert an influence on the bloch walls analogous to the action of a frictional force against the motion of the walls @xcite .
furthermore , when the field is applied , the magnetic bubbles interact via long range dipolar forces @xcite and small deviations from the 2d projected circular shape induce a slight distortion on the triangular lattice . for the fields used in our experiments ,
these distortions are not strong enough to create permanent defects in the film ( such as disclinations or dislocations ) , but slightly vary the lattice spacing from place to place .
thus , the spatial structure of the lattice is not perfectly symmetric in practice . in the presence of disorder
, the particle may oscillate around each site before it moves to the next domain ( see fig .
[ fig:1](d ) for an example of such movements ) , which leads to subdiffusive dynamics at short time scales .
however , this bias decays with time and the directions of the particle motion become asymptotically randomized , leading to diffusive dynamics at long times . with decreasing @xmath0 , the lattice disorder and thus the frequency of trapping events enhances , which further decreases the initial anomalous exponent from one ( ordinary random walk ) towards zero ( pure localization ) .
the frequency @xmath13 of the driving field does not affect the mds behavior ; it only rescales the time because the time step is given by @xmath65 .
thus , it can be understood why the crossover time to asymptotic diffusion was found to follow a power - law @xmath66 with @xmath13 @xcite .
in the previous section we explained the origin of the disorder in the lattice structure during switching of the magnetic field . as a result , the particle spends part of the time in local traps , where it experiences an oscillatory movement . in a back and forth motion ,
the particle chooses nearly backward directions when it turns .
therefore , the turning - angle distribution @xmath67 changes from an isotropic one for ordinary random walk on the lattice ( limited to the allowed directions by the lattice structure ) to an anisotropic one with a tendency towards backward directions with respect to the current direction of motion . in the extreme limit of pure localization ( remaining permanently in a trap ) , @xmath67 is a delta function at @xmath68 .
such anisotropic turning - angle distributions slow down the spread of colloids on the lattice and cause subdiffusion at short times .
the structural irregularities in the magnetic bubble lattice are more pronounced at smaller values of @xmath0 .
hence , with decreasing @xmath0 the chance of trapping events increases and the resulting @xmath67 becomes more anisotropic , which decreases the initial anomalous exponent .
after @xmath69 steps in simulations for different values of the lattice disorder @xmath70 . ] in order to ensure that the lattice disorder causes asymmetric turning - angle distribution @xmath67 and subdiffusive dynamics , we perform molecular dynamics simulations of motion on a dynamic disordered triangular lattice of synchronous flashing magnetic poles . to obtain smooth particle trajectories and a detailed time evolution of the msd , in particular to monitor the oscillatory dynamics at short times , md simulations are advantageous compared to other possible methods such as monte carlo simulations .
the simulation cell consists of nearly 3000 magnetic poles and a colloidal particle which is initially released near the center of the system to avoid boundary effects .
we impose periodic boundary conditions , and consider in - plane magnetic interactions between the immobile poles and the magnetic colloid .
the time step of our in - house code is chosen to be @xmath71s , so that the time @xmath72 between two successive switching of the field direction is resolved into more than 1000 time steps .
an explicit euler update scheme is used for integration and the simulations run until the crossover to asymptotic diffusion occurs or the total number of time steps exceeds @xmath73 .
the structural disorder is effectively introduced by random displacements of magnetic poles from their ordered lattice positions .
the strength of disorder can be quantified by the parameter @xmath70 , denoting the maximum possible displacement of each magnetic pole from its ordered lattice position .
we randomly choose the position of each node within a circle of radius @xmath74 around the corresponding node of the ordered triangular lattice , where @xmath75 denotes the size of the wigner - seitz cell .
the dynamic disorder is generated by instantaneous random rearrangements of the poles at each switching event .
thus , the poles move to new positions where they stay immobile until the next switching event .
the new random position of each pole remains within the allowed range around the original position of the corresponding ordered lattice node . this way we keep the deviations dynamic but smaller than @xmath76 in all cases . )
( black solid lines ) .
the experimental data at @xmath77 and various amplitudes @xmath0 are compared to the best fits obtained by tunning a single fitting parameter ( i.e.the lattice disorder @xmath70 in simulations , or the turning - angle anisotropy @xmath78 in the analytical model ) .
the best - fitting parameters for each value of @xmath0 are given . ]
we analyze the particle trajectory for different values of @xmath70 . as shown in fig .
[ fig:4 ] , with increasing @xmath70 the distortion of the lattice is more pronounced , thus , the particle experiences oscillatory movements in traps more frequently and for longer times , which increases the asymmetry of the turning - angle distribution @xmath67 . by tunning @xmath70 , as a single free parameter in our simulations
, a remarkable agreement with the experimental msd data is achieved ( see fig . [ fig:5 ] ) . moreover , by smoothening the msd curves obtained from simulations , we fit the initial slope of the curves to @xmath79 to get the anomalous exponent at short times .
the results shown in fig . [ fig:6 ] reveal that with increasing the amount of structural disorder , the slope gradually decreases from @xmath32 for normal diffusion towards @xmath80 corresponding to pure localization .
the growth of msd is extremely slow for @xmath81 , in accord with the experimental data at weak fields @xmath82 .
so far , we have shown that for small amplitudes of the external field the turning - angle distribution @xmath67 is asymmetric , which is accompanied by a subdiffusive dynamics at short times . in this section ,
we theoretically consider the particle motion at the level of individual steps and show how the asymmetry of @xmath67 with a tendency towards backward directions leads to subdiffusion .
it was demonstrated in a previous work @xcite that the statistical properties of the subdiffusive motion could be captured by the
_ random walk on random walk _ " model @xcite ( rwrw ) , which is a simple example of stochastic motion in a complex environment . in the rwrw model
, the walker performs an ordinary random walk on an environment which is constructed by an ordinary random walk process as well . while this model showed a quantitative agreement with the experiments for certain field parameters @xcite
, the randomness of the environment indeed varies depending on the choice of @xmath0 .
thus a more general theoretical framework is needed in order to capture the particle dynamics in the subdiffusive regime for all field parameters . here
, we look at the individual steps of the random walker and present an antipersistent model which enables us to reproduce fine details of motion such as the oscillatory dynamics observed at smaller values of @xmath0 . persistent random walk models @xcite have been used to describe e.g. stochastic transport on cytoskeletal filaments @xcite , self - propulsion subject to fluctuations @xcite , or diffusive transport of light in foams and granular media @xcite . in the following
, the asymmetry of @xmath67 is quantified with @xmath83 , which varies from @xmath54 for an isotropic distribution @xmath84 to @xmath85 for an extremely asymmetric distribution @xmath86 .
we obtain analytical expressions for the time evolution of the msd , the crossover time to asymptotic diffusion , and the long - term diffusion coefficient in terms of the control parameter @xmath78 and compare the analytical predictions with the experimental data . at short times
( after smoothening the msd curves ) in terms of the structural disorder @xmath70 of the lattice . ]
we consider a random walk in 2d , with uncorrelated step sizes @xmath60 which are obtained from an arbitrary distribution @xmath87 .
the successive step orientations are however correlated such that a new orientation is obtained from a turning - angle distribution @xmath67 along the previous direction of motion ( see fig .
[ fig:7 ] ) .
while an isotropic @xmath67 leads to a normal diffusion , introducing anisotropy along the arrival direction induces persistency and results in an anomalous transport on short time scales .
the choices of @xmath67 which encourage the walker towards backward ( forward ) directions lead to subdiffusion ( superdiffusion ) .
such an approach had been used e.g. in the context of cell migration along surfaces @xcite , animal movement @xcite , and dynamics of polymer chains @xcite .
starting from the origin , let us assume that the first direction of motion is chosen randomly , i.e. the initial condition is @xmath88 .
the @xmath30-coordinate of walker after @xmath89 steps can be obtained by projecting each step along the @xmath30-axis @xmath90 where @xmath91 .
for simplicity , in the following we consider a constant step length @xmath92 and @xmath67 distributions with left - right symmetry .
the first moment of displacement @xmath93 then reads @xmath94 since the integral over @xmath95 vanishes . here
@xmath96 denotes an ensemble average .
the second moment of displacement is given by @xcite @xmath97 similar to eq .
( [ mean - x ] ) , the first term can be obtained as for different values of the asymmetry measure @xmath78 . ] versus time for different values of @xmath78 .
positive ( negative ) @xmath78 corresponds to persistent ( antipersistent ) motion .
@xmath98 denotes a ballistic motion without any turning . ]
@xmath99 the second term on the right hand side of eq .
( [ mean - x2 - 1 ] ) can be evaluated in the following way @xmath100\bigg)^{i{-}j } = \frac{\langle \ell \rangle^2}{2 } \eta^{i{-}j } , \label{mean - x2 - 3 } \end{aligned}\ ] ] where @xmath101 .
a few examples of the turning - angle distribution @xmath67 and the corresponding values of the asymmetry parameter @xmath78 are shown in fig .
[ fig:7](b ) . from eqs .
( [ mean - x2 - 1 ] ) to ( [ mean - x2 - 3 ] ) one can obtain the second moment @xmath102 .
similar conclusions for the moments of the @xmath103-coordinate can be drawn due to symmetry .
thus , one gets the following expression for the mean squared displacement @xmath104 the case @xmath105 corresponds to an isotropic distribution @xmath67 , for which eq .
( [ mean - r2 ] ) reduces to @xmath106 , i.e. a normal diffusion .
negative ( positive ) values of @xmath78 denote an increased probability for motion in the near backward ( forward ) directions , thus , leading to antipersistent ( persistent ) motion . for @xmath107
one obtains a superdiffusive short - time dynamics , while @xmath108 leads to subdiffusion or oscillatory dynamics . in fig .
[ fig:5 ] , the theoretical predictions for the msd via eq .
( [ mean - r2 ] ) are compared with the experimental and simulation results .
it can be seen that our theoretical approach remarkably reproduces the observed behavior by fitting the single free parameter of the model , i.e. @xmath78 .
notably , the overall behavior of the msd , the frequency of oscillations , the crossover time , and even the asymptotic diffusion coefficient are all captured by the theory .
the persistent random walker has a finite - range memory , beyond which the direction of motion gets completely randomized .
according to eq .
( [ mean - x2 - 3 ] ) , the auto - correlation between step orientations is given by @xmath109 which vanishes in the limit of @xmath110 since @xmath111 .
some examples are shown in fig .
[ fig:8 ] for various values of @xmath78 .
one can estimate the crossover time @xmath112 to asymptotic diffusion by numerically solving @xmath113 moreover , it can be seen from eq .
( [ mean - r2 ] ) that the asymptotic diffusion coefficient also depends on @xmath78 as @xmath114 with @xmath61 being the average particle velocity . for a constant step size
, @xmath59 ranges from @xmath54 for @xmath115 to @xmath116 for @xmath105 .
we combined experiment and theory to investigate the dynamics of paramagnetic colloids driven above a two - state flashing potential .
this potential was realized by periodically modulating the stray field generated at the surface of a magnetic bubble lattice in an uniaxial garnet film .
the particles experience either enhanced diffusive or anomalous sub - diffusive dynamics , depending on the strength of the external drive . applying a strong field
leads to an ordinary random walk on the magnetic bubble lattice , while weaker fields result in structural disorder in the lattice which slows the particle dynamics . by means of a persistent random walk approach and numerical simulations we verified that increasing lattice disorder sharpens the turning - angle distribution of the particle towards backward directions , which decreases the anomalous exponent , postpones the crossover time to asymptotic diffusion , and modifies the long - term diffusion coefficient .
* acknowledgment * + p. t. acknowledges support from the european research council project no . 335040 , from mineco ( grant no .
fis2013 - 41144-p and ryc-2011 - 07605 ) , and agaur ( grant no .
2014sgr878 ) .
m.r.s . acknowledges support by the deutsche forschungsgemeinschaft ( dfg ) through collaborative research centers sfb 1027 ( a7 ) . | we combine experiment and theory to investigate the diffusive and subdiffusive dynamics of paramagnetic colloids driven above a two - state flashing potential .
the magnetic potential was realized by periodically modulating the stray field of a magnetic bubble lattice in a uniaxial ferrite garnet film . at large amplitudes @xmath0 of the driving field ,
the dynamics of particles resembles an ordinary random walk with a frequency - dependent diffusion coefficient .
however , subdiffusive and oscillatory dynamics at short time scales is observed when decreasing @xmath0 .
we present a persistent random walk model to elucidate the underlying mechanism of motion , and perform numerical simulations to demonstrate that the anomalous motion originates from the dynamic disorder in the structure of the magnetic lattice , induced by slightly irregular shape of bubbles . |
neutron stars are thought to be born as rapidly rotating ( @xmath1 ms ) radio pulsars created during a type ii / ib supernova explosion involving a massive star .
their existence was postulated in 1934 by baade & zwicky ( 1934 ) based on theoretical arguments , but had to wait until the 1970s for observational support , provided by the remarkable discoveries of the crab and vela pulsars in their respective supernova remnants ( snrs ) .
the properties of these pulsars were found to be uniquely explained in the context of rapidly rotating , magnetized neutron stars emitting beamed non - thermal radiation .
their fast rotation rates and large magnetic fields ( @xmath2 g ) are consistent with those of a main - sequence star collapsed to ns dimension and density .
a fast period essentially precluded all but a ns hypothesis and thus provided direct evidence for the reality of nss ( see shapiro & teukolsky 1983 for a brief history and intro to ns physics ) .
furthermore , their inferred age and association with snrs provided strong evidence that nss are indeed born in supernova explosions .
these properties were considered typical of all young pulsars , but as we shall see , there is new evidence that suggests that this is unlikely to be the case .
most supernovae ( non type ia ) are expected to produce a ns , whose unpulsed emission should be easily discernible in the radio - band during the lifetime of a typical snr ( @xmath3 yrs ) as a radio - loud `` plerion '' ( weiler & sramek 1988 ) .
so it is quite remarkable that , despite detailed radio searches , few of the hundreds of known snrs have yielded a ns candidate . furthermore , comprehensive radio surveys suggest that most radio pulsars near snrs can be attributed to chance overlap ( e.g. lorimer 1998 ; gaensler & johnston 1995 ; see kaspi 1996 for a review ) . with the results of these new surveys , traditional arguments for the lack of observed radio pulsars associated with snr , such as those invoking beaming and large `` kick '' velocities ,
are less compelling .
it is now clear that this discrepancy is an important and vexing problem in current astrophysics .
progress in resolving this mystery is suggested by x - ray observations of young snrs .
these are revealing x - ray bright , but radio - quiet compact objects at their centers .
it is now understood that these objects form a distinct class of radio - quiet neutron stars ( caraveo et al .
1996 , gotthelf , petre , & hwang 1997 ; vasisht et al . 1997 ; and refs . therein ) , perhaps born or evolving in a fashion drastically different from than of the crab .
some of these sources have been found to be slowly rotating pulsars with unique properties .
their temporal signal is characterized by spin periods in the range of @xmath4 s , steady spin - down rates , and highly modulated sinusoidal pulse profiles ( @xmath5 ) .
they have steep x - ray spectra ( photon index @xmath6 ) with x - ray luminosities of @xmath7 erg @xmath8 s@xmath9 . as a class , these seemingly isolated pulsars
are currently referred to as the anomalous x - ray pulsars ( axp ; van paradijs et al .
1995 ) . nearly half
are located at the centers of snrs , suggesting that they are relatively young ( @xmath10 yr - old ) . and
so far , no counterparts at other wavelengths have been identified for these x - ray bright objects .
the prototype for this class , the 7-s pulsar 1e 2259 + 586 in the @xmath11 yr old snr ctb109 , has been known for nearly two decades ( gregory & fahlman 1980 ) .
these are now about a dozen slow x - ray pulsars apparently associated with young snrs ( originally gregory & fahlman 1980 ; see also table i , gotthelf & vasisht 1998 for a recent summary ) .
these include the four known soft @xmath12-ray repeaters ( sgr ) , also likely to be associated with young snrs ( cline et al .
1982 ; kulkarni & frail 1993 ; vasisht et al . 1994 ) , which have recently been confirmed as slow rotators ( kouveliotou 1998 ) .
* in fact , there are currently more known slow , radio - quiet x - ray pulsars in the center of identified snr than confirmed crab - like radio pulsars ! * here , we present new results on three intriguing radio - quiet , x - ray bright neutron star candidates which we are studying closely .
these include the recently discovered 12-s x - ray pulsar in the snr kes73 , very likely to be an isolated `` magnetar '' , a pulsar with an enormous magnetic field ( b @xmath13 g ) ; the newly discovered ( march 1998 ) , bright 7-s pulsar near kes 75 , ax j1845 - 03 , which displays similar properties ; and a follow - up _ asca _ observation of rcw 103 , which helps resolve some long standing mysteries about this enigmatic object . the study of these and closely related objects are shedding new light on the evolution of young nss . the recent discovery of pulsed x - ray emission from the central compact source in snr kes73 ( vasisht & gotthelf 1997 ) was somewhat surprising , as this unresolved einstein source , 1e1841@xmath14045 , has been studied for some time ( kriss 1985 ; helfand 1994 ) .
the slow ( 12-s ) period is most unusual for a young pulsar ; if this is an `` isolated '' neutron star , then it is the one having the longest spin period ever observed .
this pulsar was initially detected during a 1993 _ asca _ x - ray observation of kes73 and confirmed with a weak archival _
rosat_detection , which indicated a unusually rapid spin - down rate ( vasisht & gotthelf 1997 ) .
we have recently obtained a new _ asca _
measurement of the pulsator which provides irrefutable confirmation of its remarkable spin - down . fig .
1 shows the periodigram for the two _ asca _ epochs ( from gotthelf & vasisht 1998 , in prep . ) . the pulsar is apparently spinning down rapidly at a rate of @xmath15 s / s , orders of magnitude faster than the crab - like pulsars .
this rate is consistent with that found using the _ rosat _ data and suggests a linear trend .
most importantly , the characteristic pulsar age is consistent with the age derived for the snr ( @xmath16 yrs ) .
the flux has remained steady between the two asca observations , consistent with the _ rosat _ flux measurement . and
the spectrum also remains unchanged , described by a steep power - law of photon index @xmath17 , unlike a typical crab - like or recycled pulsar .
these properties have important consequences for a young pulsar .
the rotational energy of 1e1841@xmath14045 is far too small to power its total x - ray emission of @xmath18 .
the maximum luminosity derivable just from spin - down is @xmath19 erg / s . on the other hand ,
the observed luminosity is appreciably low for an accretion powered binary system @xmath20 erg / s .
however , the derived luminosity , steep spectrum , and lack of stochastic variability makes an accretion scenario all but unlikely . in
3 we discuss the several models to explain the nature of this and other axps .
= 6.0 cm a recent automated search through the _ asca _ archive ( gotthelf & vasisht 1998 ; torii 1998 ) has revealed another axp - like slow pulsator . like the 12 s pulsar in kes73
, this 7 s rotator has a sinusoidal pulse shape and is modulated with a 35% amplitude .
the pulsar lies at the edge of the asca field containing the snr kes 75 , but their separation makes an association most unlikely .
= 6.0 cm = 5.0 cm on the basis of its properties , is most likely the latest example of an anomalous x - ray pulsar .
the period is nearly identical to that of the pulsar in .
the light curve is consistent with that of a steady source .
a preliminary search for variability fails to show doppler modulation which would signal the presence of a binary companion .
in addition , the low frequency spectrum shows no obvious pink noise typical of accretion powered sources .
the x - ray spectrum is remarkably steep , with a power law index @xmath21 .
the steep index is indicative of the tail of an intrinsically thermal spectrum .
an absorbed blackbody also gives a good fit with a temperature of @xmath22 kev , similar to that found for other axps .
a high foreground absorption suggests that the pulsar is distant and likely highly absorbed . uncertainty in the distance to the pulsar makes the luminosity difficult to estimate .
our best guess is 15 kpc , an estimate likely to be accurate to within a factor of two .
the isotropic x - ray luminosity is then @xmath23 erg s@xmath9 .
note that like for the kes73 pulsar , the spectrum of is unlike that of any accreting high - mass neutron star binary .
we await further multi - wavelength observations of necessary to fully describe this source and gauge its importance .
there is evidence in the most radio survey data for an underlaying shell - type snr ( b. m. gaensler , priv comm . ) .
if verified , this would be a spectacular endorsement of the axp hypothesis .
although only a future @xmath24 measurement can help determine its spin - down age .
the central x - ray source in , , has also been known for some time ( tuohy & garmire 1980 ) , but unlike kes73 , so far no pulsar has been identified . nevertheless
, several other properties of this source identify it as axp - like object . located in the center of a young snr
, the compact source has an unexpectedly steep spectral signature ( photon index @xmath25 ) with an x - ray luminosity of @xmath26 erg @xmath8 s@xmath9 ( gotthelf , petre & hwang 1997 ) .
but the lack of detected pulsations leave the nature of ambiguous and so far no clear interpretation of the origin of the x - ray emission has emerged .
progress on this front is suggested by a recent follow - up _ asca_observation of .
the flux from was seen to decrease by a factor of @xmath1 compared to the 1993 observation 3.5 yrs earlier ( see fig .
1 , petre & gotthelf 1998 , this proceedings ) .
this was quite unexpected as the count rates for the hri detections at two epochs spanning 16 yrs agreed to within @xmath27 .
the spectral fits to the original _ asca _ data only allowed a flux , at most , @xmath28 times higher that predicted by the hri measurements . our new result
, the detection of significant variability from , strongly rules out a simple cooling ns origin , as originally postulated by tuohy ( 1993 ) .
a fuller account of this result is presented elsewhere in this proceedings ( petre & gotthelf 1998 ) . in the next section ,
we present several models which may help to elucidate the aforementioned objects , and discuss their significance .
the spectral and temporal properties of and are similar to each other and to the other seemingly isolated , young axps . and these objects appear related to the sgrs . similarly ,
although no pulsation have been found from central source in , its spectral properties suggest a relation to these slow pulsars .
if these and other ns candidates like them were indeed born as fast rotators , then a mechanism must be found to slow them down to their currently observed rates .
the rapid but steady spin - down of the kes73 pulsar suggests a possibility .
the equivalent magnetic field for a rotating dipole is @xmath29 g , one of the highest magnetic fields observed in nature .
theory describing a ns with such an enormous field , a `` magnetar '' , has been worked out by duncan & thompson ( 1992 ) .
vasisht & gotthelf ( 1997 ) suggest that the kes73 pulsar was born as a magnetar @xmath30 yrs ago and has since spun down to a long period due to rapid dipole radiation losses .
this would then be the first direct evidence of such a magnetar ( see vasisht & gotthelf 1997 for a detail discussion ) . in the context of the above theory ,
may have been born a magnetar which has subsequently spun down . alternatively , some nss may be born as slow ( @xmath31 2 s ) rotators as considered by spruit & phinney ( 1998 ) , resulting from strong physical coupling between the progenitor s slowly rotating envelope and its pre - collapse core .
the enormous magnetic field postulated for these slow pulsars may provide a natural explanation for their radio - quiet nature .
it has been argued that for supercritical fields , magnetic pair creation @xmath32 , the source of electrons for radio emission , is suppressed ( baring & harding 1998 ) .
in now seems likely that the crab and vela pulsars are in fact rare , but highly visible , examples of young ns evolution .
they might have obtained their rapid initial spins through some impulsive mechanism that imparts both kicks and spins to young neutron stars ( spruit & phinney 1998 ) .
slow axp pulsars , exemplified by kes73 , are likely to be more common , but observationally less obvious manifestations of young nss .
the sgrs may represent an evolutionary stage during which young nss are likely to be produce bursts . under this scenario ,
the axps and sgr phenomena are closely related , linked by their strong magnetic field .
the discovery of new examples of slow pulsars is revolutionizing our understanding young ns evolution
it suggests that an alternative evolution scenario for young nss , is not only possible , but most probable .
we consider that many of the young nss `` missing '' in radio surveys can be accounted for by the above discussed radio - quiet nss . | the paradigm that young neutron stars ( nss ) evolve as rapidly rotating crab - like pulsars requires re - examination .
evidence is accumulating that , in fact , many young ns are slowly rotating ( @xmath0-s ) x - ray pulsars , lacking in detectable radio emission .
we present new results on three radio - quiet ns candidates associated with supernova remnants , which suggests that alternative evolutionary - paths exist for young pulsars .
these include the 12-s pulsator in kes 73 , the 7-s pulsar near kes 75 , and the enigmatic x - ray source in rcw 103 .
we postulate that such objects account for the apparent paucity of radio pulsars in supernova remnants .
# 1 # 2 _ mem . soc . astron . it . _
* # 1 * , # 2 # 1 # 2 _ the messenger _ * # 1 * , # 2 # 1 # 2 _ astron .
nach . _ * # 1 * , # 2 # 1 # 2 _ astron .
astrophys . _ * # 1 * , # 2 # 1 # 2 _ astron .
astrophys .
lett . _ * # 1 * , l#2 # 1 # 2 _ astron .
astrophys .
rev . _ * # 1 * , # 2 # 1 # 2 _ astron .
astrophys .
suppl .
ser . _ * # 1 * , # 2 # 1 # 2 _ astron .
j. _ * # 1 * , # 2 # 1 # 2 _ ann .
rev .
astron .
astrophys .
_ * # 1 * , # 2 # 1 # 2 _ astrophys .
j. _ * # 1 * , # 2 # 1 # 2 _ astrophys .
j. lett .
_ * # 1 * , l#2 # 1 # 2 _ astrophys .
j. suppl .
_ * # 1 * , # 2 # 1 # 2 _ astrophys .
space sci .
_ * # 1 * , # 2 # 1 # 2 _ adv .
space res .
_ * # 1 * , # 2 # 1 # 2 _ bull .
astron .
inst .
czechosl . _
* # 1 * , # 2 # 1 # 2 _ j. quant .
spectrosc .
. transfer _ * # 1 * , # 2 # 1 # 2 _ mon . not .
r. astr .
soc . _ * # 1 * , # 2 # 1 # 2 _ mem .
r. astr .
soc .
_ * # 1 * , # 2 # 1 # 2 _ phys .
lett .
rev . _ * # 1 * , # 2 # 1 # 2 _ publ .
astron .
. japan _ * # 1 * , # 2 # 1 # 2 _ publ .
astr .
soc .
pacific _ * # 1 * , # 2 # 1 # 2 _ nature _ * # 1 * , # 2 epsf.sty |
achieving rapid development in integrated photonic circuits depends on the all - optical elements , which are essential for high - speed processing of light signals .
nonreciprocal light propagation is an indispensable common trait for some optical elements , such as optical diodes , optical isolator , circulator , etc .
for example , the optical diode permits the light transport in only one direction but not the opposite direction .
the successful design of nonreciprocal light propagation devices relies on the breaking of the time - reversal symmetry .
thus , nonreciprocal light propagation is inherently difficult , even in theory because of the time - reversal symmetry of light - matter interaction @xcite .
motivated by the tremendous application of nonreciprocal electrical current propagation , an immense attention has been paid to the study of nonreciprocal light propagation . as a traditional method ,
a material with strong magneto - optical effects ( faraday rotation ) is often used to break the time - reversal symmetry for some optical devices @xcite . however
, unfortunately the requirement of the magneto - optical effect is the big size components and strong external magnetic fields , which are harmful for the on - chip optical nonreciprocal devices . beyond that
, one can also break the time - reversal symmetry and design the nonreciprocal optical devices by time - dependent effects @xcite , unbalanced quantum coupling @xcite or optical nonlinearity @xcite .
the ubiquitous optical nonlinearity in different optical systems has been extensively studied and further adopted in design of nonreciprocal light propagation devices .
for example , many schemes have been reported through the nonlinearity of the waveguides , such as the second order nonlinearity @xmath0 @xcite , dispersion - engineered chalcogenide @xcite , raman amplification @xcite and so on . on the other hand , duce to the high - quality factor @xmath1 and small mode volume @xmath2 of optical microcavities @xcite ,
it has attracted considerable interest for implementing nonreciprocal light propagation devices @xcite .
for instance , fan et al . achieved the experiment of nonreciprocal light propagation with the kerr and thermal nonlinearity in silicon microring resonators @xcite .
based on a nonlinearity of an optomechanical system , some schemes of nonreciprocal behavior have also been reported @xcite .
the above schemes , however , rely heavily on the strong nonlinearity , which is not easy to obtain , especially for few - photon situations .
recently , some works show that the nonlinearity in the coupled resonators can be greatly enhanced by the introducing optical gain in one resonator of the -symmetric system @xcite . and
an immense attention has been attracted to -symmetric system which has an interesting feature that non - hermitian hamiltonian can still have an entirely real spectrum with respect to the -symmetry @xcite .
in addition , two coupled resonators can be processed as a -systemic system @xcite .
more recently , a few of schemes of nonreciprocal light propagation have been proposed with -systemic coupled resonators system @xcite . for example , based on the inherent nonlinearity ( i.e. , gain - induced nonlinearity ) of the -systemic system , successful experiment has been carried out for nonreciprocal light propagation with two coupled whispering - gallery - mode ( wgm ) microresonators @xcite . note that through mechanical kerr nonlinearity , a theory scheme is also proposed for nonreciprocal phonon propagation with coupled mechanical resonators @xcite .
the weak mechanical kerr nonlinearity is greatly improved by the gain in one mechanical resonator of the -symmetry and results in the non - lossy and high isolation ratio nonreciprocal phonon propagation . in this paper
, we explore the optical nonlinearity induced by a single quantum emitter coupled to a microcavity beyond weak - excitation approximation . based on the optical nonlinearity ,
an alternative scheme is proposed for nonreciprocal light propagation in a system of two coupled cavities and a single quantum emitter coupled to one of the cavities .
the scheme reported here has some important features .
( i ) the optical nonlinearity of the hybrid system is greatly enhanced by the cavity gain .
and the giant nonlinearity leads to the non - lossy nonreciprocal light propagation with extremely high isolation ratio .
( ii ) nonreciprocal light propagation means allowing transport of light in one direction but not the opposite direction @xcite .
however , through adjusting proper parameters , to switch between the blocking and allowing directions is allowed for the present scheme . for different nonreciprocal light propagation cases ,
we can all obtain the non - lossy transmission with high isolation ratio in allowing directions .
in addition , it is achievable to switch between unidirectional and bidirectional transport regime .
( iii ) optical bistability or even optical multistability behavior is often induced by optical nonlinearity , and it will lead to unstable output . when the disturbance and perturbation of the system parameters are strong enough , the output field intensity will switch between the different metastable values of the optical bistability .
obviously , it is harmful for the nonreciprocal light propagation .
however , via choosing proper parameters range , we can avoid the interference of unstable output and obtain certain output intensity even for the strong disturbance of parameters .
this paper is organized as follows : in sec.ii , the physical model considered and the corresponding hamiltonian of the system is introduced . by applying the heisenberg - langevin formalism ,
we can obtain the evolution equations of the system . in sec.iii
, we investigate carefully the evolution equations and get the numerical steady solution .
then , we explore the spectra of the optical output fields and analyze the influence of parameters on nonreciprocal light propagation .
finally , we give a conclusion in sec.iv .
we consider the setup as shown in fig .
1 , where two single - mode optical microcavities of frequencies @xmath3 are directly coupled to each other with strength @xmath4 , which is very sensitive to the distance between two cavities .
the optical cavities are denoted by bosonic annihilation and creation operators @xmath5 and @xmath6 , respectively .
a two - level quantum emitter with transition frequency @xmath7 is embedded in the first ( j=1 ) cavity , and the cavity mode @xmath8 is coupled to the quantum emitter transition @xmath9 with the coupling strength @xmath10 .
we take the input probe field as @xmath11 , where @xmath12 and @xmath13 is the carrier frequency and the amplitude of the probe field propagating in the waveguide .
the amplitude of the input probe field @xmath13 is normalized to a photon flux at the input of the cavity and the directly related power is @xmath14 . under the rotating - wave and the electric - dipole approximation ,
the effective hamiltonian of the hybrid ( color online ) schematic diagram of the hybrid optical system consisting of two coupled cavities and a two - level quantum emitter coupled to one of them . ]
optical system is written in the rotating frame at the frequency of the probe field @xmath12 as @xcite @xmath15 where the symbols @xmath16 and @xmath17 mean the transition operator and the population operator of the two - level quantum emitter , respectively .
we assume the two cavities have the same resonant frequencies @xmath18 .
@xmath19 and @xmath20 are the corresponding frequencies detunings of the cavity field frequency from the probe field frequency and the quantum emitter transition frequency , respectively .
the coupling parameter @xmath21 describes the coupling loss rate between each cavity and the corresponding taper waveguide .
( color online ) for passive - passive cavity system , the output field @xmath22 as a function of the input field @xmath23 with ( a ) , ( b ) j=4@xmath24 and ( c ) , ( d ) g=2@xmath24 .
the other system parameters are chosen as @xmath25 , @xmath26 , @xmath27 , and @xmath28 , respectively.,title="fig:",scaledwidth=40.0% ] ( color online ) for passive - passive cavity system , the output field @xmath22 as a function of the input field @xmath23 with ( a ) , ( b ) j=4@xmath24 and ( c ) , ( d ) g=2@xmath24 .
the other system parameters are chosen as @xmath25 , @xmath26 , @xmath27 , and @xmath28 , respectively.,title="fig:",scaledwidth=40.0% ] ( color online ) for passive - passive cavity system , the output field @xmath22 as a function of the input field @xmath23 with ( a ) , ( b ) j=4@xmath24 and ( c ) , ( d ) g=2@xmath24 .
the other system parameters are chosen as @xmath25 , @xmath26 , @xmath27 , and @xmath28 , respectively.,title="fig:",scaledwidth=40.0% ] ( color online ) for passive - passive cavity system , the output field @xmath22 as a function of the input field @xmath23 with ( a ) , ( b ) j=4@xmath24 and ( c ) , ( d ) g=2@xmath24 .
the other system parameters are chosen as @xmath25 , @xmath26 , @xmath27 , and @xmath28 , respectively.,title="fig:",scaledwidth=40.0% ] based on the hamiltonian ( 1 ) , the heisenberg - langevin equations , which describe the evolution of the optical composite system , can be obtained .
we focus on the the mean response of the optical system , then the operators are reduced to their expectation values , and the heisenberg - langevin equations lead to two groups of nonlinear evolution equations . in this work ,
our goal is to consider the effect of the input field with different directions ( i.e. , forward and backward incidence ) on the output field . with the use of the mean - field assuption @xmath29 @xcite , two groups of nonlinear evolution equations of the hybrid system with the forward and backward incidence
are given as @xmath30 where the subscript @xmath31 and @xmath32 means the coupled microcavities is backward or forward incidence , respectively .
@xmath33 with @xmath34 are the mean values of the corresponding operators of the hybrid system .
@xmath35 , @xmath36 , @xmath37 where @xmath38 , @xmath39 , and @xmath40 with @xmath41 the spontaneous emission decay rate and @xmath42 the cavity intrinsic decay rate .
the evolution of the optical system varies with the direction of the input beam .
the phenomenon of direction - dependent evolution of the optical system can be displayed in the output field , which can be obtained through the following input - output relation @xcite @xmath43
in the present scheme , the strong nonlinearity of the system plays a significant role in enhancing the ability of nonreciprocity light propagation . and
it is induced by the interaction between the optical cavity and quantum emitter . in the above nonlinear evolution equations ( i.e. , eqs .
( 2 ) ) , which describe the evolution of the hybrid optical system , we pay attention to the nonlinear terms @xmath44 , @xmath45 and @xmath46
. with the weak - excitation approximation , these nonlinear terms are always discarded and the optical system will evolute in linear regime .
we value the nonlinear terms which lead to the optical bistable state , as shown in ref @xcite .
then with the parameters in the bistable region , the considered optical system may have two different metastable output values .
once the disturbance and perturbation of the system parameters become strong , the output field will switch between the different metastale values .
the uncertainty of the output field will be detrimental to the nonreciprocity light propagation .
( color online ) the output field ( a ) , ( c ) @xmath47 , ( b ) , ( d ) @xmath48 as a function of the input field @xmath23 with j=4@xmath24 and g=2@xmath24 .
the other system parameters are the same as in fig .
2.,title="fig:",scaledwidth=40.0% ] ( color online ) the output field ( a ) , ( c ) @xmath47 , ( b ) , ( d ) @xmath48 as a function of the input field @xmath23 with j=4@xmath24 and g=2@xmath24 .
the other system parameters are the same as in fig .
2.,title="fig:",scaledwidth=40.0% ] ( color online ) the output field ( a ) , ( c ) @xmath47 , ( b ) , ( d ) @xmath48 as a function of the input field @xmath23 with j=4@xmath24 and g=2@xmath24 . the other system parameters are the same as in fig .
2.,title="fig:",scaledwidth=40.0% ] ( color online ) the output field ( a ) , ( c ) @xmath47 , ( b ) , ( d ) @xmath48 as a function of the input field @xmath23 with j=4@xmath24 and g=2@xmath24 . the other system parameters are the same as in fig .
2.,title="fig:",scaledwidth=40.0% ] ( color online ) the output field @xmath47 ( red solid line ) and @xmath48 ( blue dashed line ) vary the input field @xmath23 with ( a ) @xmath25 , @xmath49 , @xmath26 , @xmath27 , @xmath28 , and @xmath50 for the passive - passive cavity system ; ( b ) @xmath25 , @xmath51 , @xmath26 , @xmath52 , @xmath53 , and @xmath50 for the unbroken @xmath54 phase ( i.e. , active - passive cavity system).,title="fig:",scaledwidth=45.0% ] ( color online ) the output field @xmath47 ( red solid line ) and @xmath48 ( blue dashed line ) vary the input field @xmath23 with ( a ) @xmath25 , @xmath49 , @xmath26 , @xmath27 , @xmath28 , and @xmath50 for the passive - passive cavity system ; ( b ) @xmath25 , @xmath51 , @xmath26 , @xmath52 , @xmath53 , and @xmath50 for the unbroken @xmath54 phase ( i.e. , active - passive cavity system).,title="fig:",scaledwidth=45.0% ] in order to avoid the negative effect of the inconclusive output field on the nonreciprocal light propagation , we begin to carefully consider the optical bistable behavior , i.e. , how does the output field depend on the system parameters . by seeking the numerical steady - state solution of eqs .(2 ) , we show the the steady - state output field intensity @xmath55 and @xmath56 versus the input filed intensity @xmath57 under various parametric conditions in fig .
the influence of the cavity - quantum emitter coupling strength @xmath10 on the behavior of the optical bistability is shown in fig .
2(a ) and fig .
2(b ) . the bistable threshold increases gradually as the cavity - quantum emitter coupling strength @xmath10 increases . more importantly , the area of the hysteresis loop becomes broader as the coupling strength @xmath10 increases from @xmath58 to @xmath59 .
conversely , the optical bistable regions will disappear when the cavity - quantum emitter coupling strength is small enough .
this is because that the optical bistability is caused by the nonlinear terms @xmath44 , @xmath45 and @xmath46 in eqs .
( 2 ) , and at a large extent , the increasing coupling strength @xmath10 between the cavity and the quantum emitter can enhance the nonlinearity of the optical system .
on the other hand , the cavity - cavity coupling strength @xmath4 has also an important influence on the optical bistability .
there is a competition between the cavity - quantum emitter coupling strength @xmath10 and cavity - cavity coupling strength @xmath4 in the input field of the optical system as shown in fig .
2(c ) and ( d ) .
when the cavity - quantum emitter coupling strength and cavity - cavity coupling strength are equal ( i.e. , @xmath60 ) , there is a remarkable optical bistable region , which quickly becomes narrow as the ratio of the cavity - cavity coupling strength to cavity - quantum emitter coupling strength increases .
once the cavity - cavity coupling has an overwhelming advantage against the cavity - quantum emitter coupling ( i.e. , @xmath61 ) , the nonlinearity of the system can be neglected and the optical bistable area will disappear completely .
besides the cavity - cavity coupling strength @xmath4 , the frequency detunings @xmath62 also affect the cavity - quantum emitter interaction , which induces the nonlinearity of the considered system .
the increasing frequency detuning will weaken the cavity - quantum emitter coupling and makes the optical bistable area small . as shown in fig .
3 , the optical bistable area gradually becomes narrow until disappears as the frequency detuning increases .
( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50 .
the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50 .
the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50 .
the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50 .
the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50 .
the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the resonant case , the transmission coefficient @xmath63 ( red solid line ) and @xmath64 ( blue dashed line ) , and the isolation ratio vary under the different values of ( a),(b ) the cavity - quantum emitter coupling strength @xmath65 with @xmath50 , and @xmath66 ; ( c ) , ( d ) the cavity - cavity coupling strength @xmath67 with @xmath68 , and @xmath51 ; ( e ) , ( f ) the input field @xmath23 with @xmath51 , and @xmath50
. the other system parameters are chosen as @xmath25 , @xmath52 , and @xmath53 , respectively.,title="fig:",scaledwidth=45.0% ] based on the above analysis , we begin to explore the relationship between the output and input intensity for the forward and backward propagation cases .
the transmission coefficients are defined as @xmath69 and the isolation ratio is @xmath70 which quantifies the isolation performance of the system . fig .
4(a ) and fig .
4(b ) correspond to the passive - passive cavities coupling and active - passive cavity coupling , respectively . in the color - code region ( b ) of fig .
4(a ) , the maximum isolation ratio is approximately @xmath71db when the output field intensity for the forward propagation stays at the lower launch . however , in the color - code region ( a ) of fig .
4(a ) , both the forward and backward propagation have only one stable output value .
thus , if we choose the system parameters in the color - code region ( a ) of fig .
4(a ) , the shortcoming of the color - code region ( b ) of fig .
4(a ) ( i.e. , the uncertain output field intensity ) caused by the optical bistability , can be overcomed , and we can also obtain the high isolation ratio .
the physical mechanism underlying the nonreciprocal light transport is rooted in the cavity - quantum emitter interaction inducing nonlinearity , which leads to localization - induced dynamical - intensity accumulation in the first cavity . the asymmetrical coupling breaks the time - reversal symmetry , which make the nonreciprocal light transport feasible , i.e. , the light transport from the cavity @xmath72 to the cavity @xmath73
is allowed and the light transport of the opposite direction is blocked . for the passive - passive cavity system , due to the decay rate of the cavities and emitter ,
the low transmissivity ( i.e.,weak output field intensity , about @xmath74 of the input field intensity ) is another obstacle to realize the nonreciprocal light transport .
( color online ) for the off - resonant case , the transmission coefficient @xmath64 ( blue dashed line ) and @xmath63 ( red solid line ) , and isolation ratio vary with ( a ) , ( b ) frequency detuning @xmath75 for @xmath76 ; with ( c ) , ( d ) frequency detuning @xmath77 for @xmath78 .
the other system parameters are chosen as @xmath25 , @xmath52 , @xmath53 , @xmath50 , @xmath66 , and @xmath51 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the off - resonant case , the transmission coefficient @xmath64 ( blue dashed line ) and @xmath63 ( red solid line ) , and isolation ratio vary with ( a ) , ( b ) frequency detuning @xmath75 for @xmath76 ; with ( c ) , ( d ) frequency detuning @xmath77 for @xmath78 .
the other system parameters are chosen as @xmath25 , @xmath52 , @xmath53 , @xmath50 , @xmath66 , and @xmath51 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the off - resonant case , the transmission coefficient @xmath64 ( blue dashed line ) and @xmath63 ( red solid line ) , and isolation ratio vary with ( a ) , ( b ) frequency detuning @xmath75 for @xmath76 ; with ( c ) , ( d ) frequency detuning @xmath77 for @xmath78 .
the other system parameters are chosen as @xmath25 , @xmath52 , @xmath53 , @xmath50 , @xmath66 , and @xmath51 , respectively.,title="fig:",scaledwidth=45.0% ] ( color online ) for the off - resonant case , the transmission coefficient @xmath64 ( blue dashed line ) and @xmath63 ( red solid line ) , and isolation ratio vary with ( a ) , ( b ) frequency detuning @xmath75 for @xmath76 ; with ( c ) , ( d ) frequency detuning @xmath77 for @xmath78 .
the other system parameters are chosen as @xmath25 , @xmath52 , @xmath53 , @xmath50 , @xmath66 , and @xmath51 , respectively.,title="fig:",scaledwidth=45.0% ] in order to enhance the transmissivity , we can choose the _ pt_-symmetric system ( i.e. , the active - passive cavity system ) as shown in fig .
the gain of the first cavity greatly improves the nonlinearity and promotes the field localization in the first cavity .
it allows the light propagation from the passive cavity to the active cavity and prevents the propagation in the opposite direction @xcite .
thus , the gain - loss balance of the _ pt_-symmetric system makes the non - lossy unidirectional light transport achievable . as shown in all the color - code areas of fig .
4(b ) , we can obtain the nonreciprocal light transport with over @xmath79 high transmissivity and about @xmath80 db isolation ratio .
among of the color - code areas of fig .
4(b ) , the area ( b ) is more suitable for the idea unidirectional light transport without the disturbance of the uncertain output field intensity caused by optical bistability because the output field intensity for the forward and backward propagations have only one stable value . with respect to the color - code area ( b ) of fig .
4(b ) , we begin to analyze the effect of the experimental parameters deviation on the unidirectional light transport in detail .
other parameters are fixed , and the non - lossy unidirectional light propagation from passive cavity to active cavity is not sensitive to the cavity - quantum emitter coupling .
when the cavity - quantum emitter coupling @xmath10 is close to zero , the linear system allows the non - lossy light propagation in both directions with the help of gain of the first cavity , but the isolation ratio declines sharply as shown in fig .
5(a ) and fig .
for the growing coupling strength @xmath10 , the nonlinearity of the system will also tend to increase and will be greatly enhanced by the gain of the first cavity .
the large nonlinearity is only in the first cavity ( i.e. , the active cavity ) and promotes the field localization in the active cavity and breaks the time - reversal symmetry of the considered optical system .
thus , the non - lossy light propagation in the backward direction is almost unaffected and the opposite direction propagation is blocked completely . when the cavity - quantum emitter coupling strength @xmath10 approaches @xmath81 , we can obtain the isolation ratio about @xmath80 db .
5(c ) and fig .
5(d ) show the influence of cavity - cavity coupling @xmath4 on the unidirectional light propagation .
the cavity - cavity coupling @xmath4 and the cavity - quantum emitter coupling @xmath10 compete for the input field of the system .
when the cavity - quantum emitter coupling @xmath4 is in the commanding position , the nonlinearity of system will decrease .
thus , the light transmissivity in the backward direction reduces with the decrease of cavity - cavity coupling , and the isolation ratio goes down as well . from fig .
5(c ) and fig .
5(d ) , we can see the effect of the change of the input field intensity on the unidirectional light propagation can be neglected .
in addition , the frequency detuning has different influence from the cavity - cavity coupling on the present scheme . as the frequency detuning increases , the nonlinearity of the system decreases .
6 shows the concrete effect of the frequency detuning on the unidirectional light propagation .
( color online ) the output field @xmath48 ( blue dashed line ) and @xmath47 ( red solid line ) vary with the input field @xmath23 for the resonant case .
the parameters are @xmath25 , @xmath51 , @xmath52 , @xmath53 , and @xmath82.,scaledwidth=45.0% ] we begin to describe an interesting feature of the optical system , i.e. , the allowed direction of the unidirectional light propagation can be reversed by adjusting the parameter @xmath4 . on the above analysis ,
the light propagation in backward direction is allowed and the opposite direction is blocked . when other parameters are fixed and the cavity - cavity coupling strength is weaker than the cavity - quantum emitter coupling strength ( for example , @xmath82 and @xmath51 ) , the nonlineraity of the system is strengthened . as shown in fig . 7 , the optical bistable threshold decreases rapidly and approaches to zero for both directions , and the optical bistable area is clearly expending especially for the backward direction . as a result , the optical bistable area overlaps for the different directions . in the color - code area of fig .
4 , we choose the lowest branch of the blue dash line which corresponds to the light propagation from the active cavity to the passive cavity , and then we get the high efficiency unidirectional light propagation with the isolation ratio about @xmath71 db . in addition , when the input field intensity increases and exceeds the corresponding maximum value in the optical bistable region ( for example , @xmath83 in fig .
7 ) , the present system will turn to the bidirectional transport regime .
in summary , we have proposed a theory scheme for nonreciprocal light propagation with the cavity - cavity coupling hybrid system .
we analyze the effect of the system parameters on the optical bistable behavior , and further study the nonreciprocal light propagation behavior in the passive - passive cavity system .
however , the interaction of the cavity - quantum emitter causes the weak nonlinearity , which is greatly enhanced by the cavity gain in the active - passive cavity system ( i.e. , -symmetric system ) .
thus , through balancing the gain and loss , we can obtain the non - lossy and high isolation ratio nonreciprocal light propagation in the unbroken phase . with appropriate parameters , we can eliminate the risk of the uncertain output field intensity from optical bistability .
in addition , the direction of the nonreciprocal light propagation can be switched by changing the cavity - cavity coupling strength @xmath4 , which is sensitive to the distance between the cavities .
the work is supported in part by national basic research program of china ( no 2012cb922103 ) and by the national science foundation ( nsf ) of china ( grant nos 11374116 ) and the fundamental research funds for the central universities .
the authors acknowledge professor ying wu for his enlightening suggestions and anonymous referees for their insightful comments .
99 h. a. haus , _ waves and fields in optoelectronics _
( prentice - hall , englewood cliffs , nj , 1984 ) . m. levy , j. opt .
b * 22 * , 254 ( 2005 ) .
h. dotsch , n. bahlmann , o. zhuromskyy , m. hammer , l. wilkens , r. gerhardt , p. hertel , and a. f. popkov , j. opt .
b * 22 * , 240 2005 .
d. jalas , a. petrov , m. eich , w. freude , s. fan , z. yu , r. baets , m. popovic , a. melloni , j. d. joannopoulos , et al .
photonics * 7 * , 579 ( 2013 )
. z. yu , and s. fan , nat . photonics . * 3 * , 91 ( 2009 ) . h. lira , z. yu , s. fan , and m. lipson , phys
* 109 * , 033901 ( 2012 ) .
k. xia , g. lu , g. lin , y. cheng , y. niu , s. gong , and j. twamley , phys .
a * 90 * , 043802 ( 2014 ) z. shen , y .- l .
zhang , y. chen , c .-
zou , y .- f .
xiao , x .- b .
zou , f .- w .
sun , g .- c .
guo , and c .- h .
dong , arxiv:1604.02297v1 x .- w .
xu and y. li , phys .
rev . a * 91 * , 053854 ( 2015 ) y. shen , m. bradford , and j .-
shen , phys .
* 107 * , 173902 ( 2011 ) .
k. gallo , and g. assanto , j. opt .
b. * 16 * , 267 ( 1999 ) .
k. gallo , g. assanto , k. r. parameswaran , and m. m. fejer , appl .
* 79 * , 314 ( 2001 ) .
h. z. shen , y. h. zhou , and x. x. yi , phys .
a * 90 * , 023849 ( 2014 ) .
x. guo , c .-
zou , h. jung , and h. x. tang , arxiv:1511.08112v1 c. g. poulton , r. pant , a. byrnes , s. fan , m. j. steel , and b. j. eggleton , opt . express .
* 20 * , 21235 ( 2012 ) .
m. krause , h. renner , and e. brinkmeyer , electron .
* 44 * , 691 ( 2008 ) .
k. j. vahala , nature ( london ) * 424 * , 839 ( 2003 ) .
h. j. kimble , _ cavity quantum electrodynamics , edited by p. berman _
( academic , san diego , 1994 ) . h. mabuchi , and a. c. doherty , science * 298 * , 1372 ( 2002 ) .
j. vuckovic , arxiv:1402.2541 .
l. fan , j. wang , l. t. varghese , h. shen , b. niu , y. xuan , andrew m. weiner , and m. qi , science * 335 * , 447 ( 2012 ) .
s. manipatruni , j. t. robinson , and m. lipson , phys . rev
* 102 * , 213903 ( 2009 ) .
m. hafezi , and p. rabl , opt .
express . * 20 * , 7672 ( 2012 ) .
h. xiong , l. g. si , x. yang , and y wu , appl .
lett . * 107 * , 091116 ( 2015 ) .
xu , y. li , a .- x .
chen , and y .- x .
liu , phys .
a * 93 * , 023827 ( 2016 ) j. li , x. zhan , c. ding , d. zhang , and y. wu , phys .
a * 92 * , 043830 ( 2015 ) j. zhang , b. peng , .
k. zdemir , y. x. liu , h. jing , x. y. l , y. l. liu , l. yang , and f. nori , phys .
b * 92 * , 115407 ( 2015 ) c. m. bender , and s. boettcher , phys . rev . lett . *
80 * , 5243 ( 1998 ) . c. m. bender , rep .
phys . * 70 * , 947 ( 2007 ) . b. peng , .
k. zdemir , f. lei , f. monifi , m. gianfreda , g. l. long , s. fan , f. nori , c. m. bender , and l. yang , nat . phys .
* 10 * , 394 ( 2014 ) ; l. chang , x. jiang , s. hua , c. yang , j.wen , l. jiang , g. li , g. wang , and m. xiao , nat . photonics . *
8 * , 524 ( 2014 ) .
b. peng , .
k. zdemir , s. rotter , h. yilmaz , m. liertzer , f. monifi , c. m. bender , f. nori , and l. yang , science * 346 * , 328 ( 2014 ) .
l. feng , z. j.wong , r .- m .
ma , y. wang , and x. zhang , science * 346 * , 972 ( 2014 ) . h. hodaei , m .- a .
miri , m. heinrich , d. n. christodoulides , and m. khajavikhan , science * 346 * , 975 ( 2014 ) .
l , h. jing , j .- y .
ma , and y. wu , phys . rev .
* 114 * , 253601 ( 2015 ) .
r. fleury , d. l. sounas , c. f. sieck , m. r. haberman , and a. al , science * 343 * , 516 ( 2014 ) .
y. wu , and x. yang , phys .
a * 76 * , 013832 ( 2007 ) .
x. yang , y. wu , and y. li , phys .
a * 55 * , 4545 ( 1997 ) .
s. huang and g. s. agarwal , phys .
a * 81 * , 033830 ( 2010 ) .
d. walls and g. milburm , _ quantum optics _
( springer , berlin , 1994 ) . c. w. gardiner and p. zoller , _ quantum noise _ ( springer , berlin , 2004 ) .
j. li , r. yu , and y. wu , phys . rev . a * 89 * , 015802 ( 2014 ) . | we propose an alternative scheme for nonreciprocal light propagation in two coupled cavities system , in which a two - level quantum emitter is coupled to one of the optical microcavities .
for the case of parity - time ( ) symmetric system ( i.e. , active - passive coupled cavities system ) , the cavity gain can significantly enhance the optical nonlinearity induced by the interaction between a quantum emitter and cavity field beyond weak - excitation approximation .
the giant optical nonlinearity results in the non - lossy nonreciprocal light propagation with high isolation ratio in proper parameters range . in addition , our calculations show that nonreciprocal light propagation will not be affected by the unstable output field intensity caused by optical bistability and we can even switch directions of nonreciprocal light propagation by appropriately adjusting the system parameters . |
many thanks to k. a. takeuchi for numerous fruitful exchanges regarding our work and for providing his png skewness splines & 1 + 1 kpz class experimental data .
we re very grateful , as well , to m. prhofer for making available the tw - gue & goe traces , and to t. imamura for kindly sharing his numerical rendering of the baik - rains f@xmath10 limit distribution . 10 for early , bedrock kpz developments ,
see : t. halpin - healy & y .- c .
zhang , phys .
rep . * 254 * , 215 ( 1995 ) ; j. krug , adv .
phys . * 46 * , 139 ( 1997 ) .
t. sasamoto & h. spohn , phys .
lett . * 104 * , 230602 ( 2010 ) ; g. amir , i. corwin , & j. quastel , commun .
pure appl .
math * 64 * , 466 ( 2011 ) ; p. calabrese , p. le doussal , and a. rosso , europhys .
lett . * 90 * , 20002 ( 2010 ) ; v. dotsenko , _ ibid , _ * 90 * , 20003 ( 2010 ) . c. a. tracy and h. widom , commun .
. phys . * 159 * , 151 ( 1994 ) ; _ ibid . _
* 177 * , 727 ( 1996 ) ; _ ibid . _
* 207 * , 665 ( 1999 ) .
m. l. mehta , _ random matrices _ ( elsevier press , 2004 ) ; also , of particular interest- c. nadal & s. majumdar , j. stat .
p04001 ( 2011 ) .
this most recent kpz installment is well summarized by : t. kriecherbauer & j. krug , j. phys .
a. * 43 * , 403001 ( 2010 ) ; i. corwin , random matrices : theory and applications * 1 * , 1130001 ( 2012 )
. s. m. ulam ,
monte carlo calculations in problems of mathematical physics , " in _ modern mathematics for the engineer , _ e. f. beckenbach , ed .
, ( mcgraw - hill , 1961 ) ; ann . rev .
* 1 * , 277 ( 1972 ) .
r. m. baer & p. brock , math . comp . *
22 * , 385 ( 1968 ) .
a. m. odlyzko & e. m. rains , att bell labs technical report ( 1999 ) ; j. h. kim , j. comb .
theory a*76 * , 148 ( 1996 ) .
a. m. vershik and s. v. kerov , soviet math . dokl . * 18 * , 527 ( 1977 ) ; func .
* 19 * , 21 ( 1985 ) ; also , b. f. logan and l. a. shepp , adv . in math .
* 26 * , 206 ( 1977 ) .
j. s. frame , g. de b. robinson , r. m. thrall , canad .
* 6 * , 316 ( 1954 ) ; c. schensted , _ ibid _ , * 13 * , 179 ( 1961 ) ; d. e. knuth , pac . j. math . * 34 * , 709 ( 1970 ) .
j. baik , p. deift & k. johansson , j. amer .
* 12 * 1119 ( 1999 ) ; d. aldous & p. diaconis , bull .
soc . * 36 * , 413 ( 1999 ) ; not all were surprised- esp . , a. okounkov , int . math .
res . not .
* 2000 * , 1043 , ( 2000 ) . t.
halpin - healy , phys .
* 109 * , 170602 ( 2012 ) ; t. halpin - healy , phys .
e * 88 * , 024118 ( 2013 ) .
m. kardar , g. parisi , and y .- c .
zhang , phys .
lett . * 56 * , 889 ( 1986 ) .
m. beccaria and g. curci , phys .
e * 50 * , 104 ( 1994 ) .
t. imamura and t. sasamoto , phys .
* 108 * , 190603 ( 2012 ) ; j. stat . phys . * 150 * , 908 ( 2013 ) .
k. a. takeuchi , phys .
lett . * 110 * , 210604 ( 2013 ) .
k. a. takeuchi & m. sano , phys .
lett . * 104 * , 230601 ( 2010 ) ; for 1 + 1 _ flat _ kpz class experiments , see k. a. takeuchi , m. sano , t. sasamoto , and h. spohn , sci . rep . * 1 * , 34 , ( 2011 ) ; k. a. takeuchi and m. sano , j. stat . phys . * 147 * , 853 ( 2012 ) ; k. a. takeuchi , arxiv:1310.0220 .
note , as well , related work on the kinetic roughening of flameless firelines : l. miettinen , m. myllys , j. merikoski , and j. timonen , eur .
j. b*46 * , 55 ( 2005 ) .
j. baik and e. m. rains , j. stat .
phys . * 100 * , 523 ( 2000 ) .
m. prhofer and h. spohn , phys .
lett . * 84 * , 4882 ( 2000 ) ; see , too , their earlier work- arxiv:9910.273 .
k. johansson , commun .
phys . * 209 * , 437 ( 2000 ) .
p. ferrari & r. frings , j. stat .
phys . * 144 * , 123 ( 2011 ) .
h. spohn , arxiv:1201.0645 .
j. krug and p. meakin , j. phys .
l987 ( 1990 ) ; for additional details , see- j. krug , p. meakin and t. halpin - healy , phys .
rev . a*45 * , 638 ( 1992 ) . here , we simply recall , for a dprm transfer matrix calculation done in a box of finite size @xmath96 there is a small positive shift , @xmath97 , upwards ( since @xmath98 for the dprm ) in the polymer free energy per unit length .
this is manifest as a condensed matter variant of the casimir effect , arising from a truncated sum over fourier modes & diminished entropy contribution ; see- m. e. fisher , j. stat . phys . * 34 * , 667 ( 1984 ) ; j. krug & l .- h .
tang , phys .
e * 50 * , 104 ( 1994 ) . in the case of 1 + 1 kpz stochastic growth models ,
the parameter @xmath24 can be determined by the _ steady - state _ width of the interface , which scales with the finite system size @xmath72 via the relation @xmath99=@xmath100 alternatively , the kpz nonlinearity @xmath26 is fixed by the tilt - dependent growth velocity : @xmath101=@xmath102 ; these matters are amply discussed by krug , meakin , & halpin - healy @xcite .
h. van beijeren , r. kutner , and h. spohn , phys .
* 54 * , 2026 ( 1985 ) ; d. a. huse , c. l. henley and d. s. fisher , _ ibid , _ * 55 * , 2924 ( 1985 ) ; l .- h .
gwa and h. spohn , _ ibid , _ * 68 * , 725 ( 1992 ) ; m. kardar , nucl .
b*290 * , 582 ( 1987 ) ; d. dhar , phase transitions * 9 * , 51 ( 1987 ) . f. bornemann , markov proc . relat .
fields * 16 * , 803 ( 2010 ) .
s. g. alves , t. j. oliveira and s. c. ferreira , europhys . lett . * 96 * , 48003 ( 2011 ) .
k. a. takeuchi , j. stat .
2012 * , p05007 ( 2012 ) .
s. g. alves , t. j. oliveira and s. c. ferreira , j. stat . mech .
( 2013 ) p05007 .
t. sasamoto & h. spohn , nucl .
b*834 * , 523 ( 2010 ) ; for their wasep - leveraged solution to the kpz equation w/ wedge ic , these authors find the mean of the gumbel distribution , @xmath103=0.577 , to be an essential ingredient .
regarding additive constant in kpz growth experiments , see- takeuchi & sano , sect .
3.4 of their jsp paper @xcite .
j. m. kim , m. a. moore , and a. j. bray , phys .
a*44 * , 2345 ( 1991 ) .
t. halpin - healy , phys .
rev . a*44 * , r3415 ( 1991 ) . t. j. oliveira , s. g. alves & s. ferreira , phys .
e * 85 * , 010601 ( 2012 ) .
the kpz history of the baik - rains constant @xmath73 has been documented , in a nutshell , in ref .
[ 12b ] , sect .
b. farnudi and d. vvedensky , phys .
e*83 * , 020103 ( 2011 ) ; evident , also , ref @xcite , table 1 & figs 3,4 therein .
we extract @xmath104=4.855 via the tw - goe variance ( not shown ) , using this value to make comparison of ballistic deposit _ stationary - state _ to br - f@xmath10 limit distribution .
we are imagining , here , an experimental system size @xmath72=10@xmath105 pixels , with roughly 10@xmath105 runs being made , yielding 10@xmath56 data for each time slice . this particular _ nonuniversal behavior _ of takeuchi s png model
is quite clearly seen in his skewness insert , figure 1b , ref .
also , for increasing @xmath76 , his png traces _ fall into the minimum from above , _ rather than rise up from below , as they do for the experimental data sets .
these png features are likewise shared by our gaussian dprm and ballistic deposits . | motivated by the recent exact solution of the _ stationary - state _ kardar - parisi - zhang ( kpz ) statistics by imamura & sasamoto ( phys .
rev .
lett . * 108 * , 190603 ( 2012 ) ) , as well as a precursor experimental signature unearthed by takeuchi ( phys
. rev .
lett . * 110 * , 210604 ( 2013 ) ) , we establish here the universality of these phenomena , examining scaling behaviors of directed polymers in a random medium , the stochastic heat equation with multiplicative noise , and kinetically roughened kpz growth models .
we emphasize the value of cross kpz - class universalities , revealing crossover effects of experimental relevance .
finally , we illustrate the great utility of kpz scaling theory by an optimized numerical analysis of the ulam problem of random permutations .
extremal paths through random energy landscapes ,
i.e. , directed polymers in random media " ( dprm ) , have long been a topic of great interest to statistical physicists , condensed matter theorists , and mathematicians alike @xcite . in two dimensions ,
the exact limit distributions of the dprm problem @xcite are of the celebrated tracy - widom ( tw ) type @xcite , best known perhaps from random matrix theory @xcite , as well as the famous ulam problem of random permutations @xcite .
nevertheless , in the bulk , the statistics of these extremal trajectories remains a challenging , rich , and quite difficult problem @xcite . in all dimensions , however , the constrained free - energy @xmath0 of these directed , extremal paths is dictated by the stochastic partial differential equation @xcite of kardar , parisi , & zhang ( kpz ) : @xmath1 where @xmath2 and @xmath3 are system - dependent parameters , the last setting the strength of the _ additive _ stochastic noise @xmath4 .
rigorous mathematical approaches , by contrast , have focussed on the related stochastic heat equation ( she ) with _ multiplicative _ noise , obtained from kpz via a hopf - cole transformation , @xmath5=@xmath6 ; here , with @xmath7 and it interpretation , this becomes an elemental , rescaled version @xcite of the she : @xmath8 governing the dprm partition function @xmath9 , a much better behaved quantity , well - regularized in the uv .
inspired by imamura & sasamoto s exact solution @xcite of the kpz equation _ stationary - state _ ( ss ) statistics , as well as takeuchi s subsequent re - examination @xcite of the tour - de - force kpz turbulent liquid crystal experiments @xcite , we focus here on the 1 + 1 dimensional dprm / she , making manifest its connection to the underlying baik - rains ( br ) f@xmath10 limit distribution @xcite relevant to this kpz class .
we examine , too , dprm scaling phenomena in the pt - line & pt - pt configurations , analog of kpz stochastic growth in _ flat _ &
curved _ geometries , governed by tracy - widom distributions appropriate to gaussian orthogonal & unitary matrix ensembles ; i.e. , tw - goe & gue , respectively .
our dprm / she results in this regard provide the final installment of the kpz triumvirate , complementing prior numerical confirmation of tracy - widom universality which had focused initially on kinetic roughening models such as polynuclear - growth ( png ) @xcite , as well as single - step , or totally asymmetric exclusion processes @xcite .
in fact , we go further here in the dprm / she context , making additional suggestions regarding experimental signatures of kpz class statistics , examining the interplay of transient & ss regimes and , finally , in the pt - pt tw - gue setting , illustrate the potency of kpz scaling theory @xcite to inform key aspects of the purely mathematical ulam problem .
we thus return the favor here , reiterating the very fruitful dialog between kpz physics and tw gue mathematics . in figure 1 , for starters
, we compare to tw - gue the shifted , rescaled distributions of our _ curved _ kpz class models , among them the : i ) pt - pt g@xmath11 dprm ; i.e. , extremal paths in the xy - plane , traveling in the [ 01]-direction , gaussian random energies @xmath12 with zero mean & variance @xmath13 on the sites of the square lattice ; side - steps are permitted , but incur a microscopic , elastic energy cost @xmath14=1 .
we consider extremal paths that commence at the origin , constrained to terminate t=300 steps later on the axis .
the g@xmath11 dprm trace in fig 1 represents an average over 10@xmath15 realizations of the random energy landscape .
similarly , ii ) pt - pt w / e dprm- here , the random site energies are exponentially distributed , p(@xmath16)=e@xmath17 , and the paths , on average , cut diagonally through the plane in the [ 11 ] direction ; at each step , there are two possibilities , either vertically ( i.e. , in y - direction ) or horizontally ( x - direction ) .
there is no elastic energy cost in this model , the total energy of the trajectory simply the sum , @xmath18=@xmath19 , of the random energies collected traveling from ( 0,0 ) to ( t , t ) . for our w / e dprm ,
we averaged over 4x10@xmath15 paths of length t=400 . next , the numerically demanding iii ) constrained she - it integrations , with kpz parameter @xmath20=10 ; we evolved the system through the forward & rear light cones of the initial & final points of the trajectory , averaging over @xmath21 runs , considering paths of 4000 steps , with time increment @xmath22t=0.05 .
universal limit distributions : dprm / she , ulam - lis , & tw - gue .
insets : extraction of ulam - lis asymptotic moments & baer - brock constant , via kpz scaling theory.,width=321 ] in the case of the pt - pt dprm models , we have extracted , from a first principles krug - meakin ( km ) finite - size scaling analysis @xcite , the nonuniversal , _ system - dependent _ kpz parameters @xmath23 , @xmath24=@xmath25 , and @xmath26 . note that the km toolbox @xcite yields the asymptotic free energy per step , @xmath23 , and the _ product _ @xmath27 , so that @xmath26 , itself , must be fixed by the parabolic dprm free - energy profile @xcite .
interestingly , in the following , only @xmath23 and the combination @xmath28=@xmath29 are needed , which we record in table i. an alternative approach , which highlights the utility of cross kpz - class studies emphasized in this paper , involves fixing the kpz scaling parameter @xmath28 directly from a fit , e.g. , to the tw - goe variance and then _ using it in the service of _ both the tw - gue limit distribution , our immediate focus , as well as stationary - state baik - rains pdf , which we consider later .
this we do , for illustrative purposes , in the case of she - it integration . with @xmath30=@xmath31
the known @xcite dprm free - energy fluctuation exponent , the central kpz scaling ansatz reads- @xmath32 with @xmath33 the underlying , order one tracy - widom gue statistical variable .
thus , the pt - pt dprm / she distributions of figure 1 have been cast in terms of @xmath33 , the canonical quantity intrinsic to the _ curved _ kpz class in this dimension .
while crafting figure 1 , we have relied upon the model - dependent _ asymptotic _ values for the distribution mean and variance , @xmath34=@xmath35 , our curve fits for the latter assuming the characteristic @xmath36 dprm finite - time correction .
our most accurate values for these limit distribution characteristics are obtained via the gaussian & exponential polymers , see table i ; compare , too , our model estimates for the distribution skewness @xmath37=@xmath38 and kurtosis @xmath39=@xmath40 the accepted values @xcite for these tw - gue quantities are : @xmath41 = @xmath42 our pt - pt dprm - she results in figure 1 , stacked up against the exact , known 1 + 1 curved kpz class tw - gue distribution , provide an interesting extremal path counterpoint to efforts on radial 2d eden clusters @xcite and kinetic roughening in droplet geometries @xcite , the latter studies involving png , single - step , ballistic deposition ( bd ) , as well as corner growth in a restricted - solid - on - solid " ( rsos ) model .
l*7c kpz system & @xmath23 & @xmath43 & @xmath44 & @xmath34 & @xmath45 & @xmath46 + g5@xmath47 dprm & -0.498021 & 0.50633 & -1.77097 & 0.8167 & 0.2304 & 0.0924 + w / e dprm & -2.0003 & 1.9882 & -1.77102 & 0.8116 & 0.2227 & 0.0908 + she - it & 2.48533 & 0.13345 & -1.7908 & 0.8121 & 0.2263 & 0.0893 + finally , in homage to the pioneering efforts of baer & brock @xcite , as well as the extraordinary numerical work of odlyzko & rains @xcite a generation later , we have performed , in a post - kpz / tw context , our own analysis of the ulam problem , studying the fluctuating statistics of the length , @xmath48 , of the longest increasing subsequence ( lis ) in a permutation of @xmath49 integers .
it was established early on , with much heavy - lifting and the securing of mathematical bounds , but then definitively by vershik & kirov @xcite , that asymptotically @xmath50 .
indeed , this was well - presented by baer & brock , see figure 1 and table 2 of their early paper , where they recorded exact enumerations of the lis pdf for @xmath51 using the rsk correspondence , hook formula , & young tableau mapping @xcite , complementing these results with approximate monte carlo data reaching up to @xmath49=10@xmath52 .
later , it was conjectured by odlyzko & rains , then firmly established by kim , that the standard deviation about this mean scaled as @xmath53 ; i.e. , the ulam - lis problem possessed a well - defined limit distribution in the variable @xmath54 .
baik , deift , & johansson then proved rigorously that the underlying , asymmetric non - gaussian distribution was , surprisingly , tw - gue from random matrix theory @xcite .
in fact , odlyzko & rains provided explicit monte carlo evidence in this regard , matching their semilog lis pdf , with 10@xmath55 permutations of length @xmath49=10@xmath56 , against a precise painlev ii rendering of tracy - widom gue .
interestingly , the odlyzko - rains data show a small , but clear , finite - time offset ( they quote @[email protected] ) in their lis pdf , which only stubbornly disappears as the authors heroically head toward asymptopia , pushing their simulations to @xmath49=10@xmath15 , where @[email protected] , & beyond ( @xmath49=10@xmath58 ) , nearly closing the gap between themselves and the known tw - gue universal mean : @xmath44=1.7711 .
since @xmath59=@xmath60 , random permutations of @xmath49=10@xmath58 digits are equivalent , roughly , to generating pt - pt dprm with t=10@xmath55 , quite a demanding task , indeed . here
, we take a different attitude , considering rather short permutation strings , with @xmath49=10@xmath61 - 10@xmath52 , essentially the regime where baer & brock concentrated their numerical enterprise , but invoke kpz scaling wisdom to , nevertheless , pin down very precise values for the moments ; see fig .
1 , where our ulam - lis pdf follows from 10@xmath15 random permutations of length @xmath49=10@xmath52 , as well as the figure insets , which document the kpz scaling of the truncated lis cumulants , allowing us to make fine asymptotic estimates : ( -1.7715,0.8135,0.2245,0.09102 ) , close to known tw - gue values . for the ulam - lis variance , skewness & kurtosis , lower inset
, we ve invoked an @xmath62 finite-time " correction , suggested by dprm lore .
the upper inset , which examines kpz scaling of the universal tw - gue mean , amends the odlyzko - rains ansatz to include an additive constant term , resulting in an @xmath63 finite - time correction to @xmath44 , plainly visible in the data . in our ulam - lis work ,
we ve come to refer to this quantity as baer - brock s constant- here estimated to be @[email protected] , given by the slope of the line in our plot of @xmath44 versus @xmath65 the existence of such a _ model - dependent additive constant _ is well - known to kpz practitioners @xcite , but has not , to our knowledge , been discussed for ulam - lis .
given the somewhat priviliged role of the random permutation problem amidst the menagerie of systems obeying tw - gue statistics- it is , by the hook formula , uniquely accessible to exact enumeration for finite @xmath49- we attach somewhat greater significance to this particular additive constant . 1 + 1 _ transient - regime _ & _ stationary - state _ kpz / dprm class : comparison to tw - goe & br - f@xmath10 limit distributions .
inset : universal _ variance - ratio _ as precursor signature of kpz ss statistics .
black traces document crossover scaling to the stationary - state , accessible via experiment .
, width=321 ] in figure 2 , for our g@xmath11 dprm , single - step growth , and she - it eulerian integration , we have paired up the distinct pdfs associated with the _ transient - regime _ & _ stationary - state _ statistics for the flat kpz geometry , making relevant comparisons to exact tracy - widom goe & baik - rains f@xmath10 distributions , respectively . for the _ transient _ dynamics , our dprm / she work harks back to early , pre - tw efforts @xcite , and dovetails with recent growth model studies @xcite , in which several kpz kinetic roughening pdfs , among them bd & rsos , are fit to tw - goe . as an indication of our findings here , we note specifically that the g@xmath11 dprm yields a variance , 0.63703 , quite close to the tw - goe result @xmath66=0.63805 , along with a skewness , 0.2976 , very much in line with the value , @xmath67=0.2935 , characteristic of that distribution .
more importantly , with regard to 1 + 1 kpz class _ stationary - state statistics , _ and comparison to the baik - rains limit distribution , our dprm / she results , well - matched by single - step , bd , & rsos findings , convincingly establish universality of the seminal png studies of prhofer & spohn @xcite- note the strong data collapse of our models upon the br - f@xmath10 trace , a highly nontrivial numerical matter . for the g@xmath11 dprm
, we have evolved the system to a late time t@xmath68=8x10@xmath52 , then study dynamic temporal correlations in the path _ free - energy , _
@xmath69=@xmath70 , during the subsequent time interval @xmath71=500 , averaging over 8x10@xmath52 runs in a system of size @xmath72=10@xmath52 . at @xmath71=500 , we note the approximate values : ( 1.1177,0.3432,0.252 ) which , extrapolated to infinite @xmath71 , yield ( 1.1466,0.3530,0.278 ) , quite close to the known @xcite br - f@xmath10 variance , skewness , & kurtosis : ( @xmath73,@xmath74,@xmath75)=(1.15039,0.35941,0.28916 ) . for rsos growth in the stationary - state , @xmath76=40k & @xmath71=240 , which gives an asymptotic estimate @xmath771.1331 for the baik - rains constant @xcite , while ballistic deposition , with @xmath76=50k & @xmath71=500 , produces 1.1382 for this quantity .
we emphasize that the plotted distributions of figure 2 are not the raw data , but have been rescaled , each in turn , to represent the _ system - dependent _ asymptotic variances , carefully extracted via kpz scaling theory .
included as fig 2 inset , we consider the time dependence of the _ ratio of variances _ in kpz transient & stationary - state regimes .
we are motivated here by takeuchi s recent study @xcite of the png model , as well his attempts to tease from the 1 + 1 kpz class turbulent liquid - crystal data @xcite , some sign of the ss statistics .
although the br - f@xmath10 limit distribution itself remained well - beyond the reach of these experiments , takeuchi discovered an impressive precursor signature of the kpz stationary - state , evident as a _
skewness minimum , _ which we confirm shortly with our own models , establishing its universal kpz aspect .
given the intrinsic importance of the stationary - state dynamics @xcite , we propose here an alternative signature , relying upon a more statistically tame quantity which , asymptotically , is fixed universally- @xmath78 by the ratio of the tw - goe variance & baik - rains constant . since the _ variance - ratio , _ like the skewness @xmath37 , requires no knowledge of @xmath28 , a quick examination of this quantity in experiments may provide a glimpse of a key universal property of the kpz stationary - state . within figure 2 inset ,
we document ( solid symbols ) the scaling of this interesting ratio for g@xmath11 dprm , she - it , and three distinct kpz growth models .
all 5 systems converge convincingly- even the contrarian bd , well known for its curmudgeonly behavior @xcite . for the she - it case , @xmath76=10k & @xmath71=100 as above , per our rendering of the br distribution .
however , we also include within the inset , data sets associated with lesser values of @xmath76 & hundredfold fewer runs , to study _ crossover effects , _ indicating how this scaling phenomena might appear in a more constrained experimental context .
the black traces , top down , correspond to @xmath76=3 , 20 , 100 , 500 , though all possess @xmath71=100 , with statistical averaging now done @xcite over a 10@xmath56 , rather than 10@xmath15 points .
the goal here is to provide an extra tack upon the 1 + 1 kpz class _ stationary - state _
statistics , and to further complement the spectacular experimental results already in hand for the transient _ flat _ tw - goe and _ curved _ tw - gue kpz classes .
takeuchi _ skewness minimum _ , evident in our dprm , she - it , bd , rsos & single - step kinetic roughening models .
upper & lower dot - dashed lines correspond to asymptotic skewness , @xmath74=0.35941 & @xmath67=0.2935 , of the baik - rains f@xmath10 & tracy - widom goe limit distributions , respectively .
inset : data sets , kpz turbulent liquid crystal experiments- ref .
@xcite .
the emergence of the minimum for our she - it integration , revealed as @xmath76 grows from 25 , 50 , to 100 , is manifest within the experiment proper , where the symbols indicate distinct times , smallest to largest , @xmath76=2,6,10,18,30,54,60 , right to left.,width=321 ] lastly , with figure 3 , we establish _ universality _ of takeuchi s skewness minimum @xcite , heightening its utility as an experimental precursor signature of the 1 + 1 kpz class stationary - state statistics . for our gaussian dprm , she - it integration , and kpz growth models ,
we have plotted up the fluctuation pdf skewness @xmath37 as a function of the dimensionless scaling parameter @xmath79=@xmath80 where , in each case , the numerics are evolved to a time @xmath76 , the skewness then tracked through a subsequent time interval @xmath81 . to consider the consequences of an _ insufficiently mature _ kinetically - roughened state , we include she - it results for @xmath76=25 , 50 , & 100 respectively , shown in the ascending , dashed curves , solid in the last instance .
it is interesting to observe how , with increasing @xmath76 , the traces develop to reveal a nearly full - fledged takeuchi minimum ; i.e. , for @xmath76=25 , the skewness min is entirely absent , though the curve possesses a suggestive inflection point . at @xmath76=50
, the minimum has appeared , but only gains a fuller expression at @xmath76=100 .
note , however , that once the skewness minimum has emerged , there is very little movement in its precise location- this is especially true of the @xmath82 ordinate , slightly less so for its abscissa .
the evolution , with increasing @xmath76 , of the universal kpz class skewness minimum is tied primarily to its _ curvature _ and , in the 1 + 1 dimensional case , to the steepening slope for @xmath83 , as @xmath84 .
we mention , in passing , that our she - it , single - step , and rsos simulations share a common _ nonuniversal behavior _ at very early @xmath81 , which gives rise to a _ model - dependent _ maximum and subsequent ( @xmath85 ) monotonic decrease in @xmath86 , evident in the figure .
most interestingly , this peculiar feature is actually manifest within takeuchi s @xcite liquid - crystal kpz experimental data set proper ( ! ) , shown as an inset within figure 3 and contrasts , intriguingly , with his own png simulation results , which indicate a weakly divergent , _ increasing _
@xmath37 for the smallest @xmath79 values @xcite .
these system - dependent details are small , but curious matters , slightly off - stage , but nevertheless quite apparent to those involved in the nuts & bolts of simulation . from an experimental point of view
, however , the takeuchi minimum itself is the essential focus and sits center - stage , providing a _ crucial , tell - tale precursor signature of 1 + 1 kpz class stationary - state statistics .
_ hence , see figure 3 , our g@xmath11 dprm results , for which we have devoted the greatest numerical investment , making a 1/4 million runs in a system size l=10@xmath52 , resulting in a data set of 2.5 billion samples . included , too , in the figure itself , as well as the inset , is takeuchi s skewness spline ( black traces ) , carefully crafted via his png work .
the agreement between takeuchi s png spline and our mix of kpz data sets , and that of the g@xmath11 dprm in particular , is quite suggestive , indeed . from our simulations ,
we locate the takeuchi minimum at @[email protected] , for @xmath79=@[email protected] .
we note , that the nonuniversal behavior at small @xmath85 evident in takeuchi s png simulation , giving rise to a climbing , positive @xmath37 is , in fact , shared by our g@xmath11 dprm and bd models .
of course , it is only in the double limit , @xmath84 , then @xmath88 , with @xmath89 that an interpolating skewness spline , originating at the tw - goe value , @xmath67=0.2935 , for large @xmath90 , would descend through the takeuchi minimum at the sweetspot near @xmath91 , then exhibit the correct asymptotic approach to the stationary - state baik - rains value , @xmath74=0.35941 , for vanishing small @xmath92 .
to manifest the latter would be quite difficult , even numerically .
finally , we mention that the _ kurtosis minimum _ we find for the g@xmath11 dprm , at @xmath93=0.117 , is quite near takeuchi s png value for this more demanding , higher cumulant ratio . given the tougher statistics , as well as the smaller difference here between the tw - goe kurtosis , @xmath94=0.1652 , and our measured k@xmath95 , eliciting this delicate feature experimentally will remain , no doubt , a most challenging task . in the interim , we await a clever experimental implementation of the kpz _ stationary - state _ initial condition , which might allow _ direct _ access to the baik - rains f@xmath10 .
given the universality established via figure 2 , with our dprm / she , single - step , bd , and rsos results , this limit distribution represents , most assuredly , the relevant fixed point pdf . |
a spinfoam sum over a given two - complex @xmath0 , formed by faces @xmath1 joining along edges @xmath2 in turn meeting at vertices @xmath3 , is defined by the expression z_= _ j_f , i_e _ f d_j_f _ v a_v(j_f , i_e ) , [ z ] where @xmath4 is the vertex amplitude " and @xmath5 is the face amplitude " .
the sum is over an assignment @xmath6 of an irreducible representation of a compact group @xmath7 to each face @xmath1 and of an intertwiner @xmath8 to each edge @xmath2 of the two - complex .
the expression is often viewed as a possible foundation for a background independent quantum theory of gravity @xcite . in particular ,
a vertex amplitude @xmath4 that might define a quantum theory of gravity has been developed in @xcite and is today under intense investigation ( see @xcite ) .
but what about the measure factor " given by the face amplitude @xmath5 ? what determines it ?
the uncertainty in determining the face amplitude has been repeatedly remarked @xcite .
one way of fixing the face amplitude which can be found in the literature , for example , is to derive the sum for general relativity ( gr ) starting from the analogous sum for a topological @xmath9 theory , and then implementing the constraints that reduce @xmath9 to gr as suitable constraints on the states summed over .
for instance , in the euclidean context gr is a constrained @xmath10 @xmath9 theory .
the state sum is well understood for @xmath10 @xmath9 theory : its face amplitude is the dimension of the @xmath10 representation @xmath11 .
the simplicity constraint fixes this to be of the form @xmath12 where @xmath13 and @xmath14 is the barbero - immirzi parameter , and therefore d_j_f = ( 2j_++1)(2j_-+1)=(2_+j_f+1)(2_-j_f+1 ) .
[ bffa ] however , doubts can be raised against this argument .
for instance , alexandrov @xcite has stressed the fact that the implementation of second class constraints into a feynman path integral in general requires a modification of the measure , and here the face amplitude plays precisely the role of such measure , since @xmath15 .
do we have an independent way of fixing the face amplitude ? here
we argue that the face amplitude is uniquely determined for any spinfoam sum of the form by three inputs : ( a ) the choice of the boundary hilbert space , ( b ) the requirement that the composition law holds when gluing two - complexes ; and ( c ) a particular locality " requirement , or , more precisely , a requirement on the local composition of group elements .
we argue below that these requirements are implemented if @xmath16 is given by the expression z_=du_f^v _ v a_v(u_f^v ) _ f ( u_f^v_1 ...
u_f^v_k ) , [ main ] where @xmath17 , @xmath18 are the vertices surrounding the face @xmath1 , and @xmath19 is the vertex amplitude @xmath4 expressed in the group element basis @xcite .
then we show that this expression leads directly to , with arbitrary vertex amplitude , but a fixed choice of face amplitude , which turns out to be the dimension of the representation @xmath20 of the group @xmath7 , d_j= dim(j ) .
[ d ] in particular , for quantum gravity this implies that the @xmath9 face amplitude is ruled out , and should be replaced ( both in the euclidean and in the lorentzian case ) by the @xmath21 dimension d_j= 2j+1 .
equation is the key expression of this paper ; we begin by showing that @xmath10 bf theory ( the prototypical spinfoam model ) can be expressed in this form ( section [ bfsec ] ) .
then we discuss the three requirements above and we show that implements these requirements . ( section [ inputs ] ) .
finally we show that gives with the face amplitude ( section iv ) .
the problem of fixing the face amplitude has been discussed also by bojowald and perez in @xcite .
bojowald and perez demand that the amplitude be invariant under suitable refinements of the two - complex .
this request is strictly related to the composition law that we consider here , and the results we obtain are consistent with those of @xcite .
, @xmath22 and @xmath23 associated to a portion of a face @xmath1 of the two - complex.,scaledwidth=30.0% ] it is well known that the partition function for @xmath9 theory can be rewritten in the form ( see @xcite ) z_=du_e _ f ( u_e_1 ...
u_e_k ) , [ zbf ] where @xmath23 are group elements associated to the oriented _ edges _ of @xmath0 , and @xmath24 are the edges that surround the face @xmath1 . let us introduce group elements @xmath25 , labelled by a vertex @xmath3 and an adjacent edge @xmath2 , such that u_e = h_veh_ve^-1 where @xmath3 and @xmath26 are the source and the target of the edge @xmath2 ( see figure [ fig1 ] ) .
then we can trivially rewrite as z_=dh_ve _ f ( ( h_v_1e_1 h^-1_v_2 e_1 ) ... ( h_v_k e_k h^-1_v_1 e_k ) ) .
[ zbf3 ] now define the group elements u_f^v = h^-1_ve h_ve associated to a single vertex @xmath3 and two edges @xmath2 and @xmath27 that emerge from @xmath3 and bound the face @xmath1 ( see figure [ fig1 ] ) . using these ,
we can rewrite as z_= dh_ve du_f^v _ v , f^v ( u_f^v , h^-1_ve h_ve ) _ f ( u_f^v_1 ...
u_f^v_k ) , where the first product is over faces @xmath28 that belong to the vertx @xmath3 , and then a product over all the vertices of the two - complex .
notice that this expression has precisely the form , where the vertex amplitude is a_v(u_f^v)=dh_ve _ f^v ( u_f^v , h_ve h_ve^-1 ) , which is the well - known expression of the 15j wigner symbol ( the vertex amplitude of @xmath9 in the spin network basis ) in the basis of the group elements .
we have shown that the @xmath9 theory spinfoam amplitude can be put in the form .
we shall now argue that is the _ general _ form of a local spinfoam model that obeys the composition law .
\(a ) _ hilbert space structure_. equation is a coded expression to define the amplitudes w_(j_l , i_n ) = _ j_f , i_e _ f d_j_f _ v a_v(j_f , i_e;j_l , i_n ) , [ w ] defined for a two - complex @xmath0 _ with boundary _ , where the boundary graph @xmath29 if formed by links @xmath30 and nodes @xmath31 .
the spins @xmath32 are associated to the links @xmath30 , as well as to the faces @xmath1 that are bounded by @xmath30 ; the intertwiners @xmath33 are associated to the nodes @xmath31 , as well as to the edges @xmath2 that are bounded by @xmath31 .
the amplitude of the vertices that are adjacent to these boundary faces and edges depend also on the external variables @xmath34 . in a quantum theory ,
the amplitude @xmath35 must be interpreted as a ( covariant ) vector in a space @xmath36 of quantum states .
has two disconnected components interpreted as in `` and an out '' spaces , then @xmath36 can be identified as the tensor product of the in `` and an out '' spaces of non - relativistic quantum mechanics . in the general case
, @xmath36 is the boundary quantum state in the sense of the boundary formulation of quantum theory @xcite .
] we assume that this space has a hilbert space structure , which we know .
in particular , we assume that _ = l_2[g^l , du_l ] [ n ] where @xmath37 is the number of links in @xmath38 and @xmath39 is the haar measure .
thus we can interpret as w_(j_l , i_n ) = _ l , i_n|w [ wbraket ] where @xmath40 is the spin network function _
l|j_l , i_n=_j_l , i_n(u_l)= _
l r^j_l(u_l)_n i_n . here
@xmath41 are the representation matrices in the representation @xmath20 and the @xmath42 form an orthonormal basis in the intertwiner space .
see for instance @xcite for details
. using the scalar product defined by , we have _
l , i_n|j_l , i_n&= & du_l _ j_l , i_n(u_l ) + & = & _ ldim(j_l ) _
j_lj_l _ n _ i_ni_n . where @xmath43 is the dimension of the representation @xmath20 . therefore the spin - network functions
@xmath44 are not normalized .
( these @xmath43 normalization factors are due to the convention chosen : they have nothing to do with the dimension of the representation that appears in . )
the resolution of the identity in this basis is 11=_j_l , i_n ( dim(j_l ) ) @xmath45@xmath46 .
+ ( b ) _ composition law .
_ in non relativistic quantum mechanics , if @xmath47 is the evolution operator from time @xmath48 to time @xmath49 , the composition law reads u(t_2,t_0)=u(t_2,t_1)u(t_1,t_0 ) .
that is , if @xmath50 is an orthonormal basis , condition of the spinfoam sum .
consider for simplicity a two - complex @xmath51 without boundary , obtained by gluing two two - complexes @xmath52 and @xmath53 along their common boundary @xmath38 .
then we require that @xmath54 satisfies the composition law z__1_2=__2|w__1 , as discussed by atiyah in @xcite .
notice that to formulate this condition we need the hilbert space structure in the space of the boundary states .
+ is attached to a link of the boundary spin - network and satisfies equation .,scaledwidth=30.0% ] \(c ) _ locality .
_ as a vector in @xmath36 , the amplitude @xmath35 can be expressed on the group element basis w(u_l)&=&_l|w + & = & _ j_l , i_n ( dim(j_l ) ) _ j_l , i_n(u_l ) w(j_l , i_n ) .
similarly , the vertex amplitude can be expanded in the group element basis a_v(u_f^v)&=&_f^v|a_v [ va ] + & = & _ j_f^v , i_n^v ( dim(j_f^v ) ) _ j_f^v , i_n^v(u_f^v ) a_v(j_f^v , i_n^v).notice that here the group element @xmath55 and the spin @xmath56 are associated to a vertex @xmath3 and a face @xmath1 adjacent to @xmath3 .
similarly , the intertwiner @xmath57 is associated to a vertex @xmath3 and a node @xmath31 adjacent to @xmath3 .
consider a boundary link @xmath30 that bounds a face @xmath1 ( see figure [ fig2 ] ) .
let @xmath18 be the vertices that are adjacent to this face .
we say that the model is local if the relation between the boundary group element @xmath58 and the vertices group elements @xmath55 is given by u_l = u_f^v_1 ... u_f^v_k .
[ locality ] in other words : if the boundary group element is simply the product of the group elements around the face .
notice that a spinfoam model defined by is local and satisfies composition law in the sense above .
in fact , generalizes immediately to w_(u_l)&=&du_f^v _ v a_v(u_f^v ) _ internal f ( u_f^v_1 ...
u_f^v_k ) + & & _ external f ( u_f^v_1 ... u_f^v_ku^-1_l ) .
[ main2 ] here the first product over @xmath1 is over the ( internal " ) faces that do not have an external boundary ; while the second is over the ( external " ) faces @xmath1 that are also bounded by the vertices @xmath59 and by the the link @xmath30 .
it is immediate to see that locality is implemented , since the second delta enforces the locality condition .
furthermore , when gluing two amplitudes along a common boundary we have immediately that du_l w__2(u_l)=z__1_2 because the two delta functions containing @xmath58 collapse into a single delta function associated to the face @xmath30 , which becomes internal .
thus , is a general form of the amplitude where these conditions hold . in @xcite , bojowald and perez
have considered the possibility of fixing the face amplitude by requiring the amplitude of a given spin / intertwiner configuration to be equal to the amplitude of the same spin / intertwiner configuration on a finer two - simplex where additional faces carry the trivial representation .
this requirement imply essentially that the amplitude does not change by splitting a face into two faces .
it is easy to see that satisfies this condition .
therefore satisfies also the bojowald - perez condition .
finally , let us show that implies _ and _ . to this purpose , it is sufficient to insert into .
this gives z_&=&du_f^v _ v _ j_f^v , i_n^v ( dim(j_f^v ) ) _ j_f^v , i_n^v(u_f^v ) a_v(j_f^v , i_n^v ) + & & _ f ( u_f^v_1 ... u_f^v_k ) . [ main22 ] expand then the delta function in a sum over characters z_&=&du_f^v _ v _ j_f^v , i_n^v ( dim(j_f^v ) ) _ j_f^v , i_n^v(u_f^v ) a_v(j_f^v , i_n^v ) + & & _ f _ j_fdim(j_f ) tr(r^j_f(u_f^v_1)r^j_f(u_f^v_k ) ) .
we can now perform the group integrals .
each @xmath55 appears precisely twice in the integral : once in the sum over @xmath56 and the other in the sum over @xmath6 . each integration gives a delta function @xmath60 , which can be used to kill the sum over @xmath56 dropping the @xmath3 subscript .
following the contraction path of the indices , it is easy to see that these contract the two intertwiners at the opposite side of each edge .
since intertwiners are orthonormal , this gives a delta function @xmath61 which reduces the sums over intertwiners to a single sum over @xmath62 .
bringing everything together , and noticing that the @xmath43 factor from the group integrations cancels the one in the integral , we have z_=_j_f i_n _ f dim(j_f ) _ v a_v(j_f^v , i_n^v ) .
this is precisely equation , with the face amplitude given by .
notice that the face amplitude is well defined , in the sense that it can not be absorbed into the vertex amplitude ( as any _ edge _ amplitude can ) .
the reason is that any factor in the vertex amplitude depending on the spin of the face contributes to the total amplitude at a power @xmath63 , where @xmath63 is the number of sides of the face .
the face amplitude , instead , is a contribution to the total amplitude that does not depend on @xmath63 .
this is also the reason why the normalization chosen for the spinfoam basis does not affect the present discussion : it affects the expression for the vertex amplitude , not that for the face amplitude . by an analogous calculation
one can show that the same result holds for the amplitudes @xmath54 : equation follows from expanded on a spin network basis .
in conclusion , we have shown that the general form of the partition function , which implements locality and the composition law , implies that the face amplitude of the spinfoam model is given by the dimension of the representation of the group @xmath7 which appears in the boundary scalar product ( [ n ] ) . in general relativity , in both the euclidean and the lorentzian cases ,
the boundary space is _ & = & l_2[su(2)^l , du_l ] , therefore the face amplitude is @xmath64 , and not the @xmath10 dimension , as previously supposed .
notice that such @xmath65 amplitude defines a theory that is far less divergent than the theory defined by .
in fact , the potential divergence of a bubble is suppressed by a power of @xmath20 with respect to . in @xcite
, it has been shown that the @xmath65 face amplitude yields a _ finite _ main radiative correction to a five - valent vertex if all external legs set to zero .
we thank elena magliaro , antonino marcian , claudio perini , and simone speziale for useful discussions .
eb gratefully acknowledges funding for this work through a marie curie fellowship of the european union .
a. perez , `` spin foam models for quantum gravity , '' _ class .
grav . _ * 20 * ( 2003 ) r43 , http://arxiv.org/abs/gr-qc/0301113[gr-qc/0301113 ] . j. w. barrett and l. crane , `` relativistic spin networks and quantum gravity , '' _ j. math .
* 39 * ( 1998 ) 32963302 , http://arxiv.org/abs/gr-qc/9709028[gr-qc/9709028 ] .
j. engle , r. pereira , and c. rovelli , `` the loop - quantum - gravity vertex - amplitude , '' _ phys .
* 99 * ( 2007 ) 161301 , http://arxiv.org/abs/0705.2388[0705.2388 [ gr - qc ] ] .
j. engle , r. pereira , and c. rovelli , `` flipped spinfoam vertex and loop gravity , '' _ nucl .
* b798 * ( 2008 ) 251290 , http://arxiv.org/abs/0708.1236[0708.1236 [ gr - qc ] ] .
r. pereira , `` lorentzian lqg vertex amplitude , '' _ class .
* 25 * ( 2008 ) 085013 , http://arxiv.org/abs/0710.5043[0710.5043 [ gr - qc ] ] .
j. engle , e. livine , r. pereira , and c. rovelli , `` lqg vertex with finite immirzi parameter , '' _ nucl .
phys . _ * b799 * ( 2008 ) 136149 , http://arxiv.org/abs/0711.0146[0711.0146 [ gr - qc ] ] .
e. r. livine and s. speziale , `` a new spinfoam vertex for quantum gravity , '' _ phys .
* d76 * ( 2007 ) 084028 , http://arxiv.org/abs/0705.0674[0705.0674 [ gr - qc ] ] . l. freidel and k. krasnov , `` a new spin foam model for 4d gravity , '' _ class . quant .
* 25 * ( 2008 ) 125018 , http://arxiv.org/abs/0708.1595[0708.1595 [ gr - qc ] ] .
w. kaminski , m. kisielowski , and j. lewandowski , `` spin - foams for all loop quantum gravity , '' http://arxiv.org/abs/0909.0939[0909.0939 [ gr - qc ] ] .
c. rovelli , `` a new look at loop quantum gravity , '' ( 2010 ) http://arxiv.org/abs/1004.1780[1004.1780 [ gr - qc ] ] .
a. perez and c. rovelli , `` a spin foam model without bubble divergences , '' _ nucl .
phys . _ * b599 * ( 2001 ) 255282 , http://arxiv.org/abs/gr-qc/0006107[gr-qc/0006107 ] .
a. perez , `` finiteness of a spinfoam model for euclidean quantum general relativity , '' _ nucl .
* b599 * ( 2001 ) 427434 , http://arxiv.org/abs/gr-qc/0011058[gr-qc/0011058 ] .
l. crane , a. perez , and c. rovelli , `` perturbative finiteness in spin - foam quantum gravity , '' _ phys .
* 87 * ( 2001 ) 181301 . c. perini , c. rovelli , and s. speziale ,
`` self - energy and vertex radiative corrections in lqg , '' _ phys .
lett . _ * b682 * ( 2009 ) 7884 , http://arxiv.org/abs/0810.1714[0810.1714 [ gr - qc ] ] .
s. alexandrov , `` the new vertices and canonical quantization , '' http://arxiv.org/abs/1004.2260[1004.2260 [ gr - qc ] ] . e. bianchi , e. magliaro and c. perini , `` spinfoams in the holomorphic representation , '' ( 2010 ) http://arxiv.org/abs/1004.4550[1004.4550 [ gr - qc ] ] .
r. oeckl , `` a ` general boundary ' formulation for quantum mechanics and quantum gravity , '' _ phys . lett . _ * b575 * ( 2003 ) 318324 , http://arxiv.org/abs/hep-th/0306025[hep-th/0306025 ] . c. rovelli , _ quantum gravity _ , cambridge university press , cambridge , uk ( 2004 ) . | the structure of the boundary hilbert - space and the condition that amplitudes behave appropriately under compositions determine the face amplitude of a spinfoam theory . in quantum gravity
the face amplitude turns out to be simpler than originally thought . |
we are studying the semileptonic decays @xmath0 , @xmath1 , @xmath2 , @xmath3 , and @xmath4 and the corresponding decays with a strange spectator quark . for a companion study of purely leptonic decays , see @xcite . the ckm matrix element @xmath5 , for example
, is obtained from the differential semileptonic decay rate for @xmath6 at total leptonic four - momentum @xmath7 @xcite : @xmath8 the unknown hadronic form factor @xmath9 is to be determined in lattice gauge theory from the matrix element of the weak vector current @xmath10 , @xmath11
since the heavy - light meson decays involve light quarks , it is important to choose an @xmath12 lattice fermion implementation with good chiral properties . to this end
we have been experimenting with an action proposed by degrand , hasenfratz , and kovcs @xcite , which introduces , in effect , a cutoff - dependent form factor at the quark - gluon vertex to suppress lattice artifacts at the level of the cutoff .
the action is the usual clover action but with a gauge background constructed by replacing the usual gauge links by ape - smoothed links @xcite with coefficient @xmath13 for the forward link and @xmath14 for the sum of staples .
the smoothed link is projected back to su(3 ) . this smoothing process is repeated @xmath15 times . for the present experiment
we take @xmath16 and @xmath17 .
these values are to be kept constant in the continuum limit , thus giving the local continuum fermion action .
this `` fattening '' process reduces problems with `` exceptional '' configurations that obstruct extrapolations to light quark mass @xcite .
calculations were done on an archive of 200 @xmath18 gauge configurations , generated with two flavors of dynamical staggered quarks of mass @xmath19 at the one - plaquette coupling @xmath20 , corresponding to a lattice spacing ( from the rho mass ) of about 0.11 fm .
the fat clover propagator was generated for three `` light '' ( spectator and recoiling ) quarks and five `` heavy '' ( decaying and recoiling ) quarks over a mass range @xmath21 .
the coefficient of the clover term @xmath22 was set to 1 .
the mass of the lightest fat clover quark was adjusted to give the same pion mass as the staggered fermion goldstone boson .
we use the fermilab program through @xmath23 for the quark wave function normalization , including the three - dimensional rotation@xcite with coefficient @xmath24 .
the light meson source is placed at @xmath25 and the heavy - light meson at @xmath26 , with antiperiodic boundary conditions in @xmath27 .
we treat three values of the heavy - light - meson momentum and 21 values of the three - momentum transfer at the current vertex .
computations are in progress .
results are presented for a subset of about half of the 200 configurations including only the two lightest spectator quark masses .
an example of the meson dispersion relation is shown in fig . [
fig : disp_rel ] .
it is quite satisfactory .
the form factor is extracted by amputating the external meson legs at present , by dividing by @xmath28 $ ] , where the @xmath29 meson energy @xmath30 and recoil meson energy @xmath31 are taken from central values of a fit to the corresponding two - point dispersion relations .
the diagonal vector form factor at zero three - momentum transfer gives the vector current renormalization factor @xmath32 .
it is shown as a function of the inverse meson mass in fig .
[ fig : z_v ] for the two currently available choices of the spectator quark mass .
we see that this nonperturbative renormalization constant is within @xmath33% of unity .
we test the soft pion theorem @xcite which states that in the chiral limit @xmath34 .
the same action and configurations are used to get @xmath35 @xcite .
both spectator and recoil quark masses ( @xmath36 and @xmath37 ) are extrapolated to zero .
if we use @xmath38 = a + bm + cm^\prime$ ] we obtain fig .
[ fig : soft_pion2 ] , a disagreement similar to that found by jlqcd @xcite .
if we include an extra term @xmath39 as advocated by maynard @xcite the theorem is satisfied , but with large extrapolated errors .
we hope our eventual full data sample will help resolve these complexities @xcite .
sample form factors for the process @xmath40 are shown in fig .
[ fig : bs_to_k ] .
fattening has allowed us to obtain results for an ostensibly @xmath12 action on unquenched lattices for quark masses at least as low as @xmath41 with no noticeable trouble from exceptional configurations .
our experiment raises a number of important questions : will a one - loop - perturbative determination of current renormalization factors be adequate ? how much fattening is good ?
does fattening push us farther from the continuum limit for some quantities ?
work is in progress .
this work is supported by the us national science foundation and department of energy and used computer resources at the san diego supercomputer center ( npaci ) , university of utah ( chpc ) , oak ridge national laboratory ( ccs ) , and the pittsburgh supercomputer center .
99 presentation by s. gottlieb ( this conference ) .
j.d . richman and p.r .
burchat , rev .
* 67 * ( 1995 ) 893 .
t. degrand , a. hasenfratz , and t. kovcs , nucl .
* b547 * ( 1999 ) 259 . m. falcioni , m. paciello , g. parisi , b. taglienti , nucl .
b * 251 * [ fs13 ] ( 1985 ) 624 ; m. albanese _ et al .
_ , phys .
b * 192 * ( 1987 ) 163 . for a discussion of scaling and chiral zero modes with this action , see m. stephenson , c. detar , t. degrand , and a. hasenfratz , in progress ( 1999 ) . for a discussion of perturbative renormalization with this action , see the presentation by c. bernard ( this conference ) .
a. el - khadra , a. kronfeld , and p. mackenzie , phys . rev .
* d 55 * ( 1997 ) 3933 .
g. burdman and j.f .
donoghue , phys .
* b280 * ( 1992 ) 287 ; m.b .
wise , phys .
* d45 * ( 1992 ) 2188 .
h. matsufuru _ et al .
_ , nucl .
b ( proc . suppl . )
* 63a - c * ( 1998 ) 368 . c. maynard ( ukqcd ) , nucl .
b ( proc . suppl . )
* 73 * ( 1999 ) 396 .
presentation by v. lesk ( this conference ) .
plenary talk by s. hashimoto ( this conference ) . | we are studying a variety of semileptonic decays of heavy - light mesons in an effort to improve the determination of the heavy - quark standard - model ckm matrix elements .
our fermion action is a novel , improved `` fat '' clover action that promises to reduce problems with exceptional configurations .
dynamical sea quarks are included in a mixed approach , _
i.e. _ we use staggered sea quarks and fat - clover valence quarks . here
we report preliminary results . |
the recently released data of the wilkinson microwave anisotropy probe ( wmap ) confirmed the earlier cobe - dmr s observation about the deficiency in fluctuation power at the largest angular scales @xcite .
the amount of quadrupole and octopole modes of the cmb temperature fluctuations is anomalously low if compared to the prediction of the @xmath8cdm model .
it implies that the initial density perturbations are significantly suppressed on scales equal to or larger than the hubble radius .
models of structure formation with a cut - off power spectrum of perturbation on large scales provide a better fit to the cmb temperature fluctuations .
the most likely cut - off wavelength derived from the wmap data @xcite actually is the same as that determined by the cobe - dmr @xcite .
the super - horizon suppression is difficult to make compatible with models which produce pure adiabatic ( isentropic ) perturbations .
however , it might be explained if the perturbations are hybrid .
the different behavior of adiabatic and isocurvature ( entropic ) perturbations around the horizon scale can be used to construct power spectra with a super - horizon suppression .
the wmap data show not only a possible non - zero fraction of isocurvature fluctuations in the primordial density perturbations , also the correlation between the adiabatic and the isocurvature components @xcite . these results then turn into the constraints on the multi - component inflationary models , as the initial perturbations generated from these models are principally hybrid @xcite .
the double and multi - field models have been extensively studied in this context @xcite . in this paper
we will investigate the hybrid perturbations created by an inflation with thermal dissipation , the warm inflation scenario @xcite . in the scheme of the thermal dissipative inflation
the universe contains a scalar field and a thermal bath during the inflation era .
the two components are coupled via the thermal dissipation .
in addition to fitting the amplitude and the power law index of the power spectrum given by the cobe data @xcite , the thermal dissipative inflation leads to a super - horizon suppression of the perturbations by a factor @xmath9 @xcite .
recently , it has been found that the warm inflation of a spontaneous symmetry breaking potential with strong dissipation is capable of accommodating a running spectral index @xmath10 of the primordial perturbations , and generally yields @xmath11 on large scales and @xmath12 on small scales @xcite .
our purpose here is to study the fractional power of the isocurvature perturbations , as well as the cross correlation between the adiabatic and the isocurvature fluctuations in the thermal dissipative inflationary model .
in contrast to a single or a double field inflations , the evolution of the universe in the thermal dissipative inflation does not need a stage of non - thermal post - inflationary reheating .
as long as the damping coefficent @xmath3 satisfies the criterion given in @xcite , @xmath13 , where @xmath14 , and @xmath4 stand for the planck energy , the energy scale , and the hubble expansion of the inflaton respectively , the dissipation is effective enough to make the temperature of the radiation component increase continuously during the inflationary epoch .
the universe would eventually enter the radiation - dominated phase when the temperature is high enough so that the radiation component prevails .
since the evolution of entropy only depends upon the thermal dissipative process during inflation , the entropic perturbations are not contaminated by the entropy production in the reheating stage .
therefore , the primordial hybrid perturbations induced by the thermal dissipation can be calculated unambiguously .
the dynamical background of the thermal dissipative inflation model has been investigated within the framework of quantum field theory .
it has been shown that the dissipation may amount to the coupling of the inflaton to a large number of particle species @xcite . in this sense , the two - field model and the thermal dissipation model can be considered as two extremes among multi - component inflations .
the former adds one more field to the single inflaton , while the later has a large number of additional fields .
the adiabatic and the isocurvature perturbations in the thermal dissipative model have been estimated in @xcite . yet
, these calculations are not immune from the problems induced by gauge issues which are crucial for thermal dissipative perturbations @xcite . in particular when interactions between the inflaton and the thermal bath are substantial
, the commonly used adiabatic / isocurvature decomposition is not gauge - independent on the ground of super - horizon .
therefore , we must take a full relativistic treatment to analyze the evolution of the hybrid perturbations generated in the thermal dissipative inflation
. moreover , the fluctuations of the radiation component have not been carefully considered in previous works .
although the energy fluctuations of the radiation component are always less than that of the inflaton field , they are not negligible in examining the relative phase between the adiabatic and the isocurvature perturbations .
this paper is organized as follows . in
ii we introduce the thermal dissipative inflationary model in relativistic covariant form .
the initial adiabatic - to - isocurvature ratio is given in iii .
iv presents a full relativistic calculation on the super - horizon evolution of adiabatic and isocurvature perturbations .
the numerical result of the spectrum of the adiabatic - to - isocurvature ratio is also given in
we then summarize our findings in v. the appendices provide the necessary details of the relativistic theory of linear perturbations .
we consider a universe consisting of a scalar inflaton field @xmath15 , and a radiation component which mimics the thermal bath . the total energy - momentum tensor is @xmath16 where subscripts @xmath17 and @xmath18 are respectively for the scalar field and the radiation , and the latin indices run from 1 to 4 .
the energy - momentum tensor can be decomposed into fluid quantities as @xcite @xmath19 in which @xmath20 , @xmath21 , and @xmath22 can be any timelike vector field , i.e. @xmath23 , which is generally taken to be the average velocity vector .
the total energy density of matter measured by an observer @xmath24 is represented by @xmath25 , while @xmath26 and @xmath27 denote the isotropic and anisotropic pressures respectively .
the quantity @xmath28 prescribes the energy flux relative to @xmath24 . with @xmath22 , the line element of the spacetime
can be written as @xmath29 where @xmath30 is a projection tensor which maps points into the rest space of the observer @xmath24 .
hence at the space - time point @xmath31 , the observer @xmath24 assigns to the event @xmath32 a spatial separation @xmath33 , and a time separation @xmath34 from him . for a minimally coupled scalar field @xmath15 ,
the lagrangian is @xmath35 where @xmath36 is a self - interaction potential .
the fluid quantities of the energy - momentum tensor for the @xmath15 field are then @xmath37 assuming the radiation component is a relativistic ideal fluid , we have @xmath38 where @xmath39 is temperature , @xmath40 and @xmath41 is the effective number of degrees of freedom at temperature @xmath39 . as mentioned previously , the thermal dissipative inflation is a multi - component model with very high multiplication , the parameter @xmath41 amounts to the actual number of components . for simplicity , we will absorb the factor @xmath42 into @xmath39 , and use @xmath43 in what follows .
the total energy - momentum conservation of the system is governed by @xmath44 the interactions between the @xmath15 field and the thermal bath are characterized by the force vectors defined as @xmath45 obviously , we have @xmath46 the interaction term @xmath47
( @xmath48 is for @xmath15 or @xmath49 ) can be further decomposed into @xmath50 the quantities @xmath51 and @xmath52 are , respectively , the temporal and the spatial components of @xmath47 .
they describe the energy and the momentum exchange between the scalar field and the thermal bath . substituting ( 10 ) into ( 8)
, one has @xmath53 the dissipation of the scalar field can be modeled by @xmath54 where @xmath3 can be a function of @xmath15 , and is always positive . with eq .
( 10 ) we have @xmath55 and @xmath56
to find the background solutions , we consider all quantities being uniform and isotropic . accordingly , the space - time of the universe assumes the flat friedman - robertson - walker metric , @xmath57.\ ] ] the einstein equations of the expanding universe yield @xmath58 and @xmath59,\ ] ] where @xmath60 represents the adiabatic index of the thermal radiation , @xmath61 denotes the cosmic scale factor , and @xmath62 is the hubble parameter .
the equation of motion for the scalar field @xmath15 is @xmath63 for the uniform background field @xmath15 , the @xmath64 term in ( 19 ) can be ignored , and we have @xmath65 where the @xmath66 denotes @xmath67 .
the equation of motion for the radiation component ( the thermal bath ) is derived from the first law of thermodynamics as @xmath68 thus , the quantity @xmath3 in ( 12 ) represents the dissipation coefficient " , which describes the production of radiation from the @xmath15 field .
furthermore , for the background solution , eq .
( 14 ) gives @xmath69 this is expected as the uniform and isotropic scalar field comoves with the radiation field , the net momentum exchange between them vanishes .
when the potential energy @xmath70 is dominant , i.e. @xmath71 , the hubble parameter @xmath72 depends largely on @xmath70 , and the universe undergoes an inflation . in the slow - roll regime @xmath73 ,
( 19 ) yields @xmath74 the scalar field approximates the trajectory @xmath75 with @xmath76 .
thus , with eqs .
( 19 ) and ( 20 ) , the behavior of the radiation component at the inflationary phase can be characterized by @xmath77 this solution actually can be obtained directly from ( 20 ) by considering the slow rolling condition @xmath78 @xcite . hence , eq . ( 24 ) is still true even when @xmath3 is @xmath15-dependent .
numerical solutions to eqs .
( 19 ) and ( 20 ) show that ( 24 ) is indeed available when @xmath3 is a power law function of @xmath15 @xcite .
thus , the temperature of the thermal bath increases mostly during the inflation era . in the standard inflation scenario which corresponds to the @xmath79 case ,
the radiation is blown off as @xmath80 .
therefore , the solution [ eq .
( 24 ) ] is independent of the initial value of @xmath81 when @xmath82 is greater than @xmath83 .
equations ( 19 ) and ( 20 ) imply that the @xmath15 field dissipation would eventually heat up the universe , giving rise to the co - existence of the scalar field and a radiation component with temperature larger than the hawking temperature , i.e. @xmath84 under this condition , the large scale reheating is unnecessary for inflation with dissipation .
that is , the inflation regime will smoothly transfer to the radiation - dominated regime when @xmath39 is high enough , and the radiation component becomes dominant @xcite .
numerical solutions to ( 19 ) and ( 20 ) have shown the smooth transition @xcite .
this feature is critical for calculating the primordial entropic fluctuations because the initial perturbations will be unaffected by the large scale post - inflationary entropic process , such as the reheating .
the initial perturbations of the @xmath15 field is calculated by the linearly perturbed field equation upon the space - time background ( 15 ) . because the primordial fluctuations are produced well within the hubble radius
, one can use the calculation without considering the gravitational gauge @xcite .
for instance , the fourier mode of the @xmath15 field perturbations with comoving wavenumber @xmath85 is characterized by a langevin - like equation @xmath86 the noise term @xmath87 given by thermal fluctuations is gaussian .
the statistical property of the noise @xmath87 can be determined by means of the fluctuation - dissipation relation @xcite @xmath88 @xmath89 with @xmath90 , and @xmath91 denotes averaging over an ensemble .
the relation that @xmath92 for the quantum case without dissipation is easily recovered . taking the slow - roll condition @xmath93 into account ,
the @xmath94 term in eq .
( 26 ) can be ignored at the horizon - crossing where @xmath95 .
accordingly , the correlation function of the fluctuations is given by @xmath96 and the amplitude for the horizon sized perturbation @xmath97 is about @xmath98 which implies the thermal fluctuations dominate over the quantum ones in the era of @xmath99 . taking the ensemble average
, we obtain the perturbation @xmath100 just outside the horizon as @xmath101 the perturbations in the @xmath15 field energy density is then @xmath102\cdot ht^2\delta\phi,\ ] ] where we have used the slow - roll condition ( 22 ) , and @xmath103 .
we consider the radiation component in thermal equilibrium with a temperature @xmath39 . at sub - horizon scales , fluctuations in the thermal radiation
can be estimated by @xmath104 @xcite , where @xmath105 denotes the total number of the relativistic particles within the horizon @xmath106 .
since the photon number density is proportional to @xmath107 , and the volume within the hubble radius is about @xmath108 , we find that @xmath109 accordingly , the energy fluctuations caused by @xmath110 are characterized by @xmath111 where eq.(30 ) has been used .
it should be emphasized that ( 34 ) qualifies only the relation between the variances of @xmath110 and @xmath97 , but not their phases .
principally thermal fluctuations of radiation component is random - phased with respect to the @xmath15 field fluctuations , i.e. @xmath112 . with the help of the background solutions ( 22 ) and ( 24 ) , the relation between the fluctuations in the thermal energy and the @xmath15 field energy can be established , @xmath113 therefore , @xmath110 is always less than @xmath114 in situations either @xmath115 or @xmath116 .
apparently , the perturbed cosmic matter is just about isothermal .
the energy perturbations of the @xmath15 field and the radiation component can be decomposed into adiabatic ( ad ) and isocurvature ( en ) modes as @xmath117 by definition , the adiabatic perturbations @xmath118 are given by @xmath119 while the isocurvature mode satisfies @xmath120 therefore , the adiabatic perturbations can be rewritten as @xmath121 and the entropic perturbations are characterized by @xmath122 where we have used @xmath123 from eq.(5 ) , and @xmath124 from ( 6 ) and ( 24 ) .
the radiation part @xmath125 in the above definition ( 41 ) is noteworthy as it may be compatible to the inflaton part @xmath126 , even though the condition @xmath127 is always fulfilled .
that is , @xmath110 must not be ignored when treating the isocurvature perturbation no matter how small it may be . considering the case when @xmath128
, one can obtain the variances for the adiabatic and the entropic perturbations by virtue of eqs .
( 30 ) , ( 32 ) and ( 35 ) : @xmath129^{1/2 } = \left(\frac{\rho_{(\phi)}}{h_{(\phi)}}\right)\cdot \left[1+\frac{\gamma_h}{4(\gamma_h+3)^2}\right]^{1/2 } \left(1+\frac{\gamma_h}{3}\right)^{-1},\ ] ] @xmath130^{1/2 } = \left(\frac{\rho_{(\phi)}}{h_{(\phi)}}\right)\cdot \left[1+\frac{9\gamma_h^{-1}}{4\left(\gamma_h+3\right)^2}\right]^{1/2}.\ ] ] the correlation between the two perturbation components is given by @xmath131\left(1+\frac{\gamma_h}{3}\right)^{-1}.\ ] ] equation ( 44 ) shows that even when the fluctuations of the @xmath15 field and the radiation are uncorrelated , the correlation between the adiabatic and the entropic perturbations can be significant .
in these initial conditions all @xmath15 and @xmath132-dependent quantities as well as @xmath133 , @xmath134 , and @xmath135 are taken to be their values at the @xmath85-mode horizon - crossing times .
in super - horizon regions , physical quantities should be expressed in a gauge - independent fashion .
however , the adiabatic / isocurvature decomposition of perturbations shown in the last section are not gauge - independent , i.e. either eqs.(40 ) or ( 41 ) are not gauge invariant variables . as usual @xcite
, we choose the condition to fix the coordinates and to make ( 40 ) and ( 41 ) physically meaningful .
as shown in appendix c , a gauge invariant ( gi ) variable @xmath136 can be defined as @xmath137 where @xmath138 is the perturbed variable of 3-space curvature [ see eq .
thus , if we fix the coordinates by @xmath139 , the so - called uniform - curvature gauge ( ucg ) condition , then @xmath136 becomes @xmath118 [ eq.(40 ) ] .
thus , the initial condition and the evolution of the super - horizon - sized adiabatic perturbations can be described properly by the equations of @xmath136 in the ucg .
on the other hand , we can define another gi variable @xmath140 using eqs.(c11 ) and ( c12 ) as @xmath141 where @xmath142 and @xmath143 is the linear perturbed variables of energy exchange between the @xmath15 field and the thermal bath [ eqs.(14 ) and ( a17 ) ] .
evidently , @xmath140 is nothing but @xmath144 if replacing @xmath114 and @xmath110 in ( 41 ) respectively by @xmath145 and @xmath146 .
we may call @xmath140 a modified isocurvature ( entropic ) perturbation . as a consequence
, there exists no suitable coordinate fixing to make both @xmath118 and @xmath144 simultaneously a gi variable as there is an energy exchange between the background components . in this regard , the decomposition of perturbations by ( 40 ) and ( 41 ) at super - horizon scales does not have clear physical meanings .
instead , one should decompose the super hubble perturbations into the adiabatic and the modified isocurvature modes .
the initial condition and the evolution of the modified entropic perturbation outside the horizon can then be described unambiguously by the equations of @xmath140 in the ucg .
moreover , at the commencing of radiation regime when the inflation ends , @xmath147 , the energy ex change in the background ceases and ( 46 ) reduces to ( 41 ) .
therefore , perturbations defined by eq . ( 46 ) can be explained as the isocurvature perturbations at the onset of the radiation - dominated epoch .
the initial condition of @xmath142 can be estimated via ( a17 ) as @xmath148 comparing to ( 30 ) and ( 32 ) , @xmath142 is always less than @xmath114 , but it can be larger than @xmath149 if @xmath150 . since @xmath151 , we have @xmath152 .
hence , ( 40 ) remain valid even when @xmath114 and @xmath110 are replaced by @xmath145 and @xmath146 . more importantly ,
since @xmath143 involves the effect of @xmath132 , it is correlated with the @xmath15 field fluctuations .
the perturbations @xmath143 and @xmath153 actually are in phase .
therefore , it is interesting to consider the case of imposing the phase correlation into the initial conditions . connecting @xmath114 and @xmath110 by the relation ( 35 ) , eqs .
( 40 ) and ( 41 ) yield @xmath154 \left(1+\frac{\gamma_h}{3}\right)^{-1},\ ] ] @xmath155,\ ] ] and @xmath156 \left[1+\frac{\gamma_h^{1/2}}{2(\gamma_h+3)}\right ] \left(1+\frac{\gamma_h}{3}\right)^{-1}.\ ] ] consequently , @xmath144 and @xmath118 are in phase when @xmath157 , but are anti - correlated when @xmath158 .
the evolution of the linear perturbations outside the hubble radius is described by a set of equations of all perturbed matter variables , such as @xmath153 and @xmath110 , as well as perturbed variables of the space - time metric .
the sophisticated formalism of deriving these equations are given in appendices .
for the ucg , these equations are shown as ( d1 ) - ( d8 ) in appendix d. with the solutions of eqs .
( d1 ) - ( d8 ) , it is straightforward to trace the evolution of relevant variables of the perturbations .
the evolutionary features of the super - horizon - sized perturbations can be seen from eqs .
( d7 ) and ( d8 ) , which are @xmath159 \delta\dot{\phi } + \left[\frac{k^2}{a^2}+v''+\frac{16\pi g}{3 } \frac { \dot{\phi}v'}{h}\right]\delta\phi \nonumber \\ & = & -\left[\frac{32\pi g}{3 } v\frac{\dot{\phi}}{h } + \left(\gamma\dot{\phi}+2v'\right)\right][\dot{\chi}+h\chi ] + \frac{2}{3}\frac{k^2}{a^2}\dot{\phi}\chi,\end{aligned}\ ] ] @xmath160\dot{\chi } + \left[4\pi g\left(\frac{8}{3}v-\dot{\phi}^2-\frac{4}{3}\rho_{(r)}\right)- \frac{2}{3}\frac { k^2}{a^2}\right]\chi \nonumber \\ & = & -4\pi g\left[-\frac{2}{3}\frac{\dot{\phi}}{h}\delta\dot{\phi}+ \frac{4}{3}\frac{v'}{h}\delta\phi\right].\end{aligned}\ ] ] these equations portray the linear evolution of the @xmath85-mode perturbation of the scalar field @xmath161 , and that of the shear @xmath162 of the space - time metric @xcite .
the friedmann eq .
( 17 ) implies @xmath163 . for modes at sub - horizon
scales @xmath164 , we have @xmath165 $ ] .
accordingly , if taking the right hand side to be zero , eq .
( 51 ) within the horizon is exactly the same as the linearized eq . (
18 ) , or ( 26 ) with the slow - rolling condition .
since the initial perturbation of @xmath15 field is sub - horizon - scaled , and is governed by ( 18 ) or ( 26 ) , it is consistently to assume the initial conditions for @xmath162 and @xmath166 to be zero .
equations ( 51 ) and ( 52 ) can be regarded as two coupled oscillators with a time dependent mass , damping coefficients and coupling coefficients . whenever the mass " becoming negative , the perturbations undergo a decaying or growing process .
this gives rise to the gravitational clustering .
the `` mass '' of the @xmath162-oscillator generally is positive for modes beyond the hubble radius @xmath167 .
therefore , the perturbations @xmath161 will not be magnified by gravity outside the horizon , and will retain approximately their initial values .
consequently , variations in both @xmath168 and @xmath169 are insignificant during their super - horizon journey .
we calculate the power fraction @xmath0 of isocurvature perturbations , and the cross - correlation @xmath1 between the adiabatic and the isocurvature perturbations at the end of the inflation epoch when the universe enters the radiation regime .
these two quantities are defined by , respectively , @xmath170 @xmath171 two sets of initial conditions , ( 42 ) - ( 44 ) and ( 48 ) - ( 50 ) , are imposed respectively to calculate @xmath0 and @xmath2 .
both quantities depend mainly upon the dissipation parameter @xmath135 .
the results are plotted in figs . 1 and 2 .
figure 1 shows the prominent mixture and correlation of two distinguished perturbation components for the case where the fluctuations of the @xmath15 field and of the thermal bath are totally uncorrelated .
the @xmath135-dependence of the curves for @xmath0 and @xmath1 in fig .
2 reveals the similar behavior as in fig . 1 , but with different amounts .
in particular , the isocurvature perturbations and the adiabatic ones are in phase when @xmath172 , but are anti - correlated when @xmath158 .
we show that the thermal dissipative inflation produces the initial hybrid perturbations with significant correlation .
this is largely due to the coexistence of two components , the radiation and the @xmath15 field , during the inflationary epoch .
the evolution of the density perturbation of radiation and that of the @xmath15 field on scales larger than horizon are governed by different equations ( d3 ) and ( d7 ) .
consequently after reentering the horizon , the density perturbation of radiation and that originated from @xmath15 field generally are different .
the super - horizon analysis is essential to reveral the formation of the hybrid perturbations .
we have calculated the power fraction of the isocurvature perturbations ( @xmath0 ) , and the correlation between the adiabatic and the isocurvature perturbations ( @xmath1 ) at the end of the inflationary epoch .
since the transition from the inflationary era to the radiation - dominated period is smooth without an intervening reheating process , the values of @xmath0 and @xmath1 can be directly used as the initial conditions for the radiation regime .
we found that , for each @xmath85-mode , @xmath0 and @xmath1 are mainly determined by the parameter @xmath135 at the instant as the perturbation mode crosses outside the hubble horizon .
that is , @xmath0 and @xmath2 depends entirely upon the dissipation of the inflaton .
when taking into account the effect of energy exchange , @xmath2 is more sensitive to the dissipation parameter @xmath135 . for strong dissipation cases where @xmath173 the adiabatic and the isocurvature perturbations are in phase . under weak dissipation @xmath174 , however ,
the two perturbation components are anti - correlated . given that the current observational constraints on @xmath0 and @xmath2 are rather diverse@xcite , we do not make the detail parameter fitting in this paper , but disscuss some properties of the hybrid perturbations which are useful for further model testing .
apparently , in this thermal dissipative model @xmath0 and @xmath2 are not independent of each other .
the predicted @xmath0-@xmath2 relations can directly be seen from figs . 1 and 2 .
these relations hold for all @xmath85-modes , but they are not associated with the parameter @xmath3 . on large scales these predictions can be confronted with the cmb observation without considering the effects of evolution of perturbation during the radiation era .
the relation @xmath175 shown in fig .
2 can then be tested by the observed adiabatic / isocurvature ratio and correlation .
for instance , to realize the thermal dissipative inflation , the dissipation parameter @xmath135 can be taken in the range @xmath176 @xcite .
consequently , @xmath0 amounts to about 10% and @xmath177 to + 0.3 .
secondly , the @xmath85-dependence of @xmath2 in this model is governed by the @xmath85-dependence of @xmath135 .
on the other hand , eqs .
( 42)-(43 ) or ( 48)-(49 ) shows that the two differences @xmath178 are also specified by the @xmath85-dependence of @xmath135 .
therefore , the difference between the spectral indices , or the running spectral index , of the adiabatic and the isocurvature perturbations should also be determined by the @xmath85-dependence of the correlation @xmath2 .
for instance , if @xmath2 are @xmath85-independent for some range of @xmath85 , the spectral indices should be fixed without any changes in that range . on the contrary , if there exists difference between the spectral indices of the two perturbation components , we should see a @xmath85-dependent @xmath1 .
this property is a robust prediction regardless of the initial conditions [ eqs .
( 42)-(44 ) or ( 48)-(50 ) ] in use , and is effective to testing the thermal dissipative inflation model .
wl is supported by the national science council , taiwan , roc under the grant nsc91 - 2112-m-001 - 026 .
appendixes a - d will give all formulas needed for a full relativistic theory of linear perturbations for a system consisting of a @xmath15 field and a fluid , such as radiation .
most material is taken from @xcite .
the perturbed metric with respect to @xmath179 [ eq.(3 ) ] is @xmath180 . in this paper
, we are interested only in the scalar - type perturbations . for this purpose ,
the perturbations can be described by four variables@xcite @xmath181 where the comoving spatial metric tensor @xmath182 is defined by @xmath183 .
the vertical bar indicates a covariant derivative based on @xmath184 . from eq.(5 )
, the linearly perturbed energy density is @xmath185 the first term on the r.h.s . of ( a5 ) can be calculated by @xmath186 by means of ( a1 ) , we have @xmath187 .
therefore , ( a6 ) yields @xmath188 thus , ( a5 ) becomes @xmath189 similarly , the perturbation in the pressure of the @xmath15 field is @xmath190 for the background solution , @xmath191 . since @xmath192 , the linearly
perturbed @xmath193 can be described by @xmath194 where @xmath195 since @xmath196 [ eq . ( 5 ) ] , the perturbed anisotropic pressure @xmath197 . from eq .
( 6 ) , the perturbed variables of the thermal bath are given by @xmath198 where @xmath199 is the energy density flux of radiation .
using eqs .
( 13 ) and ( a6 ) the perturbed temporal component of the interaction term is @xmath200 similarly , the perturbed spatial component can be described by @xmath201
based on the perturbed variables given in a , the evolution of these variables is governed by the following equations : _ definition of @xmath42 _ @xmath202 _ adm energy constraint _
@xmath203 _ momentum constraint _
@xmath204 _
adm propagation _
@xmath205 _ raychaudhuri equation _
@xmath206,\ ] ] _ energy conservation of radiation _
@xmath207 _ momentum conservation of radiation _
@xmath208 _ energy conservation of scalar field _
@xmath209\alpha,\ ] ] the momentum conservation of scalar field yields an identity .
considering a gauge transformation @xmath210 , in which @xmath211 is an infinitesimal quantity , any tensor quantity @xmath212 transforms as @xmath213 , where @xmath214 is the lie derivative of @xmath212 in the 4-vector field @xmath211 .
we may split the tensor @xmath212 into its background and perturbed values ; i.e. , @xmath215 .
the transformation of the perturbed parameter becomes @xmath216 . applying the transformation to the metric tensor
, we have @xmath217 since the background 3-space is homogeneous and isotropic , all the perturbation variables are gauge independent under purely spatial gauge transformations . the perturbation equations presented above are also independent of the spatial gauge transformation . under the temporal transformation factor @xmath218 ,
the perturbed variables are given by @xcite : @xmath219 equations ( c6 ) to ( c9 ) are applicable to components @xmath17 , @xmath18 as well as the total fluids . based on the time - gauge transformations [ eqs .
( c2)-(c9 ) ] and background equations of motion [ eqs .
( 19 ) and ( 20 ) ] , one can construct the following gauge invariant variables : @xmath220 @xmath221 @xmath222 @xmath223
using the uniform - curvature gauge ( ucg ) , the equations of the perturbed variables can be obtained from eqs .
( b1 ) to ( b8 ) by setting up @xmath224 .
we have @xmath225 @xmath226 @xmath227h\chi -\dot{\phi}\delta\dot{\phi}-v'\delta\phi,\ ] ] @xmath228 @xmath229 @xmath230 @xmath231 \delta\dot{\phi } - \left[\frac{k^2}{a^2}+v''+\frac{16\pi g}{3 } \frac{\dot{\phi}v'}{h}\right]\delta\phi \nonumber \\ & & -\left[\frac{32\pi g}{3 } v\frac{\dot{\phi}}{h}+ \left(\gamma\dot{\phi}+2v'\right)\right]\dot{\chi } \nonumber \\ & & -\left[\left(\frac{32\pi g}{3 } v- \frac{2}{3}\frac{k^2}{a^2}\right)\dot{\phi}+ \left(\gamma\dot{\phi}+2v'\right)h\right]\chi~,\end{aligned}\ ] ] @xmath232\dot{\chi}- \left[4\pi g\left(\frac{8}{3}v-\dot{\phi}^2-\frac{4}{3}\rho_{(r)}\right)- \frac{2}{3}\frac{k^2}{a^2}\right]\chi \nonumber \\ & & -4\pi g\left[-\frac{2}{3}\frac{\dot{\phi}}{h}\delta\dot{\phi}+ \frac{4}{3}\frac{v'}{h}\delta\phi\right],\end{aligned}\ ] ] l. a. kofman , phys . lett . * b173 * , 400(1986 ) ; d. polarski and a. a. starobinsky , nucl . phys .
* b 385 * , 623 ( 1992 ) ; d. polarski and a. a. starobinsky , phys . rev .
* d 50 * , 6123 ( 1994 ) . f. di marco , f. finelli , r. brandenberger , phys . rev .
* d67 * , 063512 ( 2003 ) d. langlois , phys .
d59 , 123512 , ( 1999 ) ; d. wands , n. bartolo , s. matarrese & a. riotto , phys .
d66 , 043520 , ( 2002 ) ; l. amendola , c. gordon , d. wands & m. sasaki , phys .
88 , 211302 , ( 2002 ) . | we investigate the primordial scalar perturbations in the thermal dissipative inflation where the radiation component ( thermal bath ) persists and the density fluctuations are thermally originated .
the perturbation generated in this model is hybrid , i.e. it consists of both adiabatic and isocurvature components .
we calculate the fractional power ratio ( @xmath0 ) and the correlation coefficient ( @xmath1 ) between the adiabatic and the isocurvature perturbations at the commencing of the radiation regime .
since the adiabatic / isocurvature decomposition of hybrid perturbations generally is gauge - dependent at super - horizon scales when there is substantial energy exchange between the inflaton and the thermal bath , we carefully perform a proper decomposition of the perturbations .
we find that the adiabatic and the isocurvature perturbations are correlated , even though the fluctuations of the radiation component is considered uncorrelated with that of the inflaton .
we also show that both @xmath0 and @xmath2 depend mainly on the ratio between the dissipation coefficient @xmath3 and the hubble parameter @xmath4 during inflation .
the correlation is positive ( @xmath5 ) for strong dissipation cases where @xmath6 , and is negative for weak dissipation instances where @xmath7 .
moreover , @xmath0 and @xmath2 in this model are not independent of each other .
the predicted relation between @xmath0 and @xmath1 is consistent with the wmap observation .
other testable predictions are also discussed . |
this paper has two goals : to study the relationships between the hierarchies of convexity conditions and tame combing conditions on a cayley complex corresponding to a given group , and to significantly expand the understanding of the cayley complex of thompson s group @xmath0 with respect to the standard finite presentation with two generators and two relators .
several notions of almost convexity for groups have been developed in geometric group theory , from the most restrictive property defined by cannon to the weakest notion of minimal almost convexity introduced by kapovich . for a group @xmath3 with finite generating set @xmath4 , almost convexity conditions for different classes of functions measure , in terms of the given function , how close balls in the cayley graph for @xmath5 are to being convex sets ( see section [ sub : ac ] for the formal definition ) .
results of thiel and elder and hermiller , respectively , show that cannon s almost convexity and minimal almost convexity , respectively , are not quasi - isometry invariants .
mihalik and tschantz introduced the notion of a tame 1-combing of a 2-complex , and in particular of the cayley complex of a group presentation , in the context of studying properties of 3-manifolds .
hermiller and meier refined the definition of tame combing to differentiate between types of tameness functions , analogously to almost convexity conditions . for a group @xmath3 with finite presentation @xmath6 ,
intuitively the radial tameness function measures the relationship , for any loop , between the size of the ball in the cayley complex containing the loop and the size of the ball needed to contain a disk filling in that loop ( see section [ sub : tc ] for the formal definitions ) .
hermiller and meier showed that the advantage of studying balls in a cayley complex from the viewpoint of tame combings and radial tameness functions is that the classes of tame combable groups are , up to lipschitz equivalence of radial tameness functions ( for example , linear functions or exponential functions ) , invariant under quasi - isometry , and hence under change of presentation . in the same paper
they also showed that groups which are almost convex with respect to several classes of functions are contained in the quasi - isometry invariant class of groups admitting a 1-combing with a linear radial tameness function . in section [ sub : hier ] , we give a more complete discussion of the hierarchies of almost convexity conditions and tame combing functions , and their interconnections . in seeking to further understand the correspondence between these two hierarchies
, we use geometric information from the cayley complex to construct tame 1-combings with linear tameness functions for two groups : thompson s group @xmath0 and the solvable baumslag - solitar group @xmath1 for @xmath7 .
cleary and taback showed that thompson s group @xmath0 is not almost convex with respect to the standard generating set , and belk and bux showed that thompson s group @xmath0 is not minimally almost convex with respect to the standard finite generating set .
elder and hermiller showed that the groups @xmath1 for @xmath8 are also not minimally almost convex with respect to their usual generating set ; moreover , miller and shapiro showed shown that the group @xmath1 does not satisfy cannon s almost convexity condition for any generating set . combining these
then provides the first examples of groups which admit a combing with a linear radial tameness function ( with respect to any choice of finite presentation ) , but which are not minimally almost convex on a particular finite generating set .
these also provide the first examples of groups which admit a combing with a linear radial tameness function but which do not satisfy cannon s almost convexity condition with respect to every finite presentation . in the case of thompson s group @xmath0
, our proof also gives significant new insight into the cayley complex of this group . despite the prevalence of @xmath0 in geometric group theory ,
a detailed understanding of the cayley complex for the standard finite presentation @xmath9,[x_0x_1^{-1},x_0^{-2}x_1x_0 ^ 2 ] \rangle\ ] ] has been elusive . in an intricate analysis
, guba showed that thompson s group @xmath0 has a quadratic isoperimetric function , but it is as yet unknown if thompson s group is automatic , nor even if it is asynchronously combable .
the tame 1-combing we construct for @xmath0 utilizes the nested traversal paths defined by cleary and taback in .
we show that these paths yield a set of quasigeodesic normal forms for the group . extending these paths to a tame 1-combing of the cayley 2-complex @xmath10 for the presentation above
then requires a careful , detailed classification of the edges and 2-cells of @xmath10 .
in particular we analyze which edges do not lie on these normal form paths and for each such edge , we characterize which 2-cells of @xmath10 adjacent to that edge have the property that their other boundary edges lie `` closer '' to the identity vertex @xmath11 . our analysis and measure of closeness to the identity use combinatorial properties computed from the group elements labeling the vertices adjacent to that edge .
the paper is organized as follows . in section 2
, we provide an overview of the development of the notions of convexity and combings for groups , and the relations between them . in section 3
, we provide a brief introduction to thompson s group @xmath0 , and define the set of normal forms which will be used in the construction of the tame 1-combing .
we also show that this set of normal forms satisfies a quasi - geodesic property . in section 4
, we construct the 1-combing of @xmath0 , and in section 5 we show that this combing satisfies a linear radial tameness function , as stated in theorem [ thm : flintame ] . in section 6
we show that @xmath12 with @xmath2 has a 1-combing which satisfies a linear radial tameness function , proving theorem [ thm : bsp ] .
finally , section 7 is devoted to the proof of theorem [ thm : bscoeff1 ] , verifying that although @xmath12 with @xmath13 and the standard presentation admits a 1-combing with a linear radial tameness function , this tameness function must have a multiplicative constant greater than 1 .
for a group @xmath3 with a finite inverse - closed generating set @xmath4 , we let @xmath14 denote the cayley graph of @xmath3 with respect to @xmath4 , and let @xmath15 denote the word metric with respect to this generating set . the pair @xmath5 satisfies the almost convexity condition @xmath16 for a function @xmath17 if there is an @xmath18 such that for every two points @xmath19 in the sphere @xmath20 ( centered at the identity ) with @xmath21 and every natural number @xmath22 , there is a path inside the ball @xmath23 from @xmath24 to @xmath25 of length no more than @xmath26 .
every group satisfies the almost convexity condition @xmath27 for the function @xmath28 , as two points in the ball of radius @xmath29 can always be connected by a path of length @xmath30 which remains inside @xmath23 , simply by going to the identity and returning outward . thus the weakest nontrivial almost convexity condition for a pair @xmath5 is @xmath27 for the function @xmath31 .
kapovich and riley have shown that this minimal almost convexity condition ( mac ) implies finite presentation of the group and the existence of an algorithm for constructing the cayley graph . at the other end of the spectrum , @xmath5 is almost convex ( ac ) in the sense of cannon if it satisfies @xmath27 for a constant function @xmath32 . between the constant function and @xmath31
, there are a number of other possible functions which give rise to a range of almost convexity conditions .
for example , ponaru studied the property @xmath27 for sublinear functions @xmath32 .
let @xmath33 be a finitely presented group , with @xmath4 an inverse - closed generating set .
let @xmath10 denote the cayley complex corresponding to this presentation , with 0- and 1-skeletons @xmath34 and @xmath35 ; that is , @xmath36 is the cayley graph with respect to this presentation . in order to have a notion of a ball centered at the identity @xmath11 in the 2-complex @xmath10 , the notion of distance between the vertices of a cayley graph
is extended to a notion of level on the entire complex .
the following definition is equivalent to that in .
[ def : level ] 1 . if @xmath37 is a vertex in @xmath38 , the _ level _ @xmath39 is defined to be the word length @xmath40 with respect to the generating set @xmath4 .
2 . if @xmath41 , then @xmath42 is in the interior of some edge with vertices @xmath43
. then let @xmath44 3 . if @xmath45 , then @xmath42 is in the interior of some 2-cell with vertices @xmath46 along the boundary , and @xmath47 where if @xmath48 is the set of relators , and for each @xmath49 , @xmath50 is the number of letters in the relator @xmath51 , then @xmath52 . intuitively , a @xmath53-combing of a group @xmath3 with generating set @xmath4 is a choice of path in the cayley graph @xmath14 from the identity to each group element . to obtain a @xmath54-combing for @xmath55 ,
a @xmath53-combing is extended continuously through the @xmath54-skeleton of the cayley complex .
viewing the ball of radius @xmath56 in @xmath10 as the set of points of level at most @xmath56 , a radial tameness function @xmath57 for a @xmath54-combing ensures that once a combing path leaves the ball of radius @xmath58 , it never returns to the ball of radius @xmath56 .
the pair @xmath55 satisfies the tame combing condition @xmath59 for a function @xmath60 if there is a continuous function @xmath61 { \rightarrow}x$ ] satisfying : 1 . for all @xmath62 , @xmath63 and @xmath64 , 2 .
@xmath65 ) \subseteq x^1 $ ] , and 3 . for all @xmath66 , @xmath67 , and @xmath68 ,
if @xmath69 , then @xmath70 . the function @xmath71 is a _
1-combing _ of @xmath10 , and @xmath72 is a _ radial tameness function _ for @xmath71 .
a continuous function @xmath73 { \rightarrow}x^1 $ ] with @xmath63 and @xmath64 for all @xmath74 is called a _ 0-combing _ for the pair @xmath5 .
the restriction of a 1-combing to the vertices of @xmath10 is a 0-combing . in , hermiller and meier
show that the condition @xmath59 is quasi - isometry invariant , and thus is independent of the choice of presentation for the group , up to a lipschitz equivalence on the radial tameness functions .
hence it makes sense to define the class of groups admitting a tame 1-combing with a linear radial tameness function , and also classes with polynomial and exponential radial tameness functions .
a pair @xmath55 may satisfy a variety of almost convexity and tame combing conditions . in figure
[ fig : bigpicture ] below , we illustrate what is known about the relevant relationships between the different classes of almost convexity functions , and the different classes of possible radial tameness functions which arise from tame 1-combings . in particular ,
all descending vertical arrows in figure [ fig : bigpicture ] follow immediately from the definitions . node at ( 0,0 ) [ shape = rectangle , draw ] ( a2 ) minimally almost convex with @xmath75 node at ( 0,4 ) [ shape = rectangle , draw ] ( a4 ) poenaru almost convex with @xmath76 sublinear node at ( 0,8 ) [ shape = rectangle , draw ] ( a5 ) almost convex with @xmath77 node at ( 0,10 ) [ shape = rectangle ] ( a6 ) * hierarchy of almost * node at ( 0,9.5 ) [ shape = rectangle ] ( a6 ) * convexity conditions * node at ( 8,-2 ) [ shape = rectangle , draw , double ] ( b6 ) @xmath58 recursive node at ( 8,0 ) [ shape = rectangle , draw , double ] ( b5 ) @xmath58 exponential node at ( 8,2 ) [ shape = rectangle , draw , double ] ( b3 ) @xmath58 linear node at ( 8,5 ) [ shape = rectangle , draw ] ( b4 ) @xmath78 node at ( 8,8 ) [ shape = rectangle , draw ] ( b1 ) @xmath79 node at ( 8,10 ) [ shape = rectangle ] ( b8 ) * hierarchy of radial * node at ( 8,9.5 ) [ shape = rectangle ] ( b9 ) * tameness functions * node at ( 5,8.5 ) [ shape = rectangle ] ( x1 ) node at ( .9,2 ) [ shape = rectangle ] ( qq ) node at ( 5.25,3.6 ) [ shape = rectangle ] ( qqq ) , node at ( 8.75,3.5 ) [ shape = rectangle ] ( xxx ) * node at ( 4.2,1.4 ) [ shape = rectangle ] ( ww ) , , * ; ( 8.25,2.45)(8.25,4.55 ) ; ( 8,7.55)(8,5.45 ) ; ( 7.75,4.55)(7.75,2.45 ) ; ( 8,1.55)(8,0.45 ) ; ( 8,-0.55)(8,-1.45 ) ; ( 2.8,8)(7,8 ) ; ( -.25,.45)(-.25,3.55 ) ; ( .25,3.55)(.25,.45 ) ; ( 0,7.55)(0,4.45 ) ; ( 0,1.75)(.5,2.25 ) ; ( 8,3.25)(8.5,3.75 ) ; ( 4,3.5)(6.75,2.5 ) ; ( 6.75,1.75)(3.75,0.5 ) ; ( 5,1.25)(5.5,0.95 ) ; these two chains of conditions are tied together at the base by the results of hermiller and meier which show that a pair @xmath5 is almost convex if and only if there is a set of defining relations @xmath80 such that the pair @xmath81 admits a 1-combing satisfying the radial tameness function @xmath82 . in theorem d of
, hermiller and meier showed that the property @xmath27 with @xmath32 sublinear , together with a linear isodiametric function ( a combination of properties motivated by work of ponaru in ) , imply the existence of a 1-combing with a linear radial tameness function . for a pair @xmath5
satisfying @xmath27 for any function @xmath17 such that @xmath83 for all @xmath84 , it follows from riley and induction that for all @xmath85 , we have @xmath86 , where @xmath87 is a constant .
hence the property @xmath27 with @xmath32 sublinear implies a linear isodiametric function , and so this extra assumption was redundant . as a consequence , it follows that @xmath27 with @xmath32 sublinear implies the condition @xmath59 with @xmath72 linear .
tantalizing questions to consider , given these results , involve the potential connections between weaker notions of almost convexity and radial tameness functions .
as yet , there are few examples known , other than for groups satisfying the condition @xmath27 with @xmath32 sublinear , of groups with 1-combings admitting restricted tameness functions . in this paper the results of theorems [ thm : flintame ] and [ thm : bsp ]
show that the quasi - isometry independent class @xmath88 of groups with a 1-combing satisfying a linear radial tameness function contains groups which are not even minimally almost convex for some particular generating set , giving the diagonal non - implication in figure [ fig : bigpicture ] .
however , this leaves open the question of whether every group in @xmath88 , and in particular whether @xmath0 and @xmath1 with @xmath2 , might have some generating set with respect to which it is minimally almost convex .
other intriguing questions involve the possibility of upward implications in either of the two hierarchies . in theorem
[ thm : bscoeff1 ] we show the vertical non - implication for tameness functions drawn in figure [ fig : bigpicture ] . for almost convexity ,
elder and hermiller have exhibited a pair @xmath5 which is minimally almost convex but does not satisfy the condition @xmath27 with @xmath32 sublinear .
it is still an open question whether there can be a pair @xmath5 satisfying the ponaru @xmath27 condition with @xmath32 sublinear that does not also satisfy cannon s ac property .
we present a brief introduction to thompson s group @xmath0 and refer the reader to for a more detailed discussion , with historical background .
in addition , in section [ ntnf s ] we define a the set of normal forms for @xmath0 which will be used in our construction of a 1-combing .
thompson s group @xmath0 has a standard infinite presentation : @xmath89 the elements @xmath90 and @xmath91 are sufficient to generate the entire group , since powers of @xmath90 conjugate @xmath91 to @xmath92 for @xmath93 .
only two relators are required for a presentation with the generating set @xmath94 , resulting in the finite presentation for @xmath0 : @xmath9,[x_0x_1^{-1},x_0^{-2}x_1x_0 ^ 2 ] \rangle.\ ] ] this is the most commonly used finite generating set and presentation for thompson s group @xmath0 , and in this paper we will build the 1-combing for @xmath0 using the cayley complex for this presentation . with respect to the infinite presentation given above , each element @xmath95 can be written in normal form as @xmath96 with @xmath97 , @xmath98 and @xmath99 .
furthermore , we require that when both @xmath92 and @xmath100 occur , so does @xmath101 or @xmath102 , as discussed by brown and geoghegan .
we will use the term _ infinite normal form _ to mean this normal form , and write @xmath103 where @xmath104 is the maximal subword of this normal form with positive exponents , and @xmath105 is the maximal subword with negative exponents .
elements of @xmath0 can be viewed combinatorially as pairs of finite binary rooted trees , each with the same number of edges and vertices , called _
tree pair diagrams_. let @xmath106 be a finite rooted binary tree .
we define a _ caret _ of @xmath106 to be a vertex of the tree together with two downward oriented edges , which we refer to as the left and right edges of the caret .
right ( respectively left ) child _ of a caret @xmath107 is defined to be a caret which is attached to the right ( resp
. left ) edge of @xmath107 . if a caret @xmath107 does not have a right ( resp
. left ) child , we call the right ( resp
. left ) leaf of @xmath107 _ exposed_. the caret itself is _ exposed _ if both of its leaves are also leaves of the tree ; that is , the caret has no children . for
a given tree @xmath106 , let @xmath108 denote the number of carets in @xmath106 .
we number the carets from @xmath54 through @xmath108 in infix order .
the infix ordering is carried out by numbering the left descendants ( the left child and all descendants of the left child ) of a caret @xmath107 before numbering @xmath107 , and the right descendants of @xmath107 afterward .
we use the infix numbers as names for the carets , and the statement @xmath109 for two carets @xmath110 and @xmath56 simply expresses the relationship between their infix numbers . in a tree pair diagram @xmath111
, we refer to the pair of carets with infix number @xmath110 , one in each tree , as the _ caret pair @xmath110_. the left ( resp .
right ) side of a binary rooted tree @xmath106 consists of the left ( resp .
right ) edge of the root caret , together with the left ( resp .
right ) side of the subtree consisting of all left ( resp .
right ) descendants of the root caret .
a caret in a tree @xmath106 is said to be a _ right _
_ left _ ) caret if one of its edges lies on the right ( resp .
left ) side of @xmath106 .
the root caret can be considered either left or right .
all other carets are called _ interior _ carets .
we also number the leaves of the tree @xmath106 from left to right , from @xmath53 through @xmath108 .
an element @xmath95 is represented by an equivalence class of tree pair diagrams , among which there is a unique reduced tree pair diagram .
we say that a pair of trees is _ unreduced _ if , when the leaves are numbered from @xmath53 through @xmath108 , there is a caret in both trees with two exposed leaves bearing the same leaf numbers .
if so , we remove that pair of carets , and renumber the carets in both trees . repeating this process until there are no such pairs
produces the unique _ reduced _ tree pair diagram representing @xmath112 .
the equivalence of these two interpretations of thompson s group is given using the infinite normal form for elements with respect to the standard infinite presentation , and the concept of leaf exponent . in a single tree @xmath106
whose leaves are numbered from left to right beginning with @xmath53 , the _ leaf exponent _ @xmath113 of leaf number @xmath114 is defined to be the integral length of the longest path of left edges from leaf @xmath114 which does not reach the right side of the tree . given the reduced tree pair diagram @xmath111 representing @xmath115 , compute the leaf exponents @xmath113 for all leaves @xmath114 in @xmath106 , numbered @xmath53 through @xmath116 .
the negative part of the infinite normal form for @xmath112 is then @xmath117 .
we compute the exponents @xmath118 for the leaves of the tree @xmath119 and thus obtain the positive part of the infinite normal form as @xmath120 .
many of these exponents will be @xmath53 , and after deleting these , we can index the remaining terms to correspond to the infinite normal form given above , following . as a result of this process
, we often denote the unique reduced tree pair diagram for @xmath112 by @xmath121 , since the first tree in the pair determines the terms in the infinite normal form with negative exponents , and the second tree determines those terms with positive exponents .
we refer to @xmath122 as the negative tree in the pair , and @xmath123 as the positive tree . child child child child ; child childchild child ; child childchild child childchild ; child childchild childchild child ; group multiplication is defined as follows when multiplying two elements represented by tree pair diagrams .
let @xmath124 and @xmath125 . to form the product @xmath126 , we take unreduced representatives of both elements , @xmath127 and @xmath128 , respectively , in which @xmath129 . the product
is then represented by the ( possibly unreduced ) pair of trees @xmath130 .
if the fewest possible carets are added to the tree pairs for @xmath37 and @xmath131 in order to make @xmath129 , and yet the pair @xmath130 is unreduced , we say that a caret must be _ removed _ to reduce the tree pair diagram for @xmath126 . given any @xmath121 in @xmath0 , let @xmath132 denote the number of carets in either tree of a reduced tree pair diagram representing @xmath112 . for any natural number @xmath114 , let @xmath133 ( respectively @xmath134 ) denote the tree with @xmath114
right ( respectively left ) carets , and no other carets ; if @xmath135 , @xmath136 ( or @xmath137 ) denotes the empty tree . for @xmath103 , where as above @xmath104 and @xmath105 are the positive and negative subwords of the infinite normal form , the tree pair diagram @xmath138 represents @xmath139 , and @xmath140 represents @xmath105 .
however , one of these tree pair diagrams may not be reduced .
if the last @xmath114 carets of @xmath122 ( respectively @xmath123 ) are all right carets , then at least @xmath141 of them must be removed in order to produce the reduced tree pair diagram for @xmath105 ( respectively @xmath104 ) .
the inverse of @xmath112 is represented by the reduced tree pair diagram @xmath142 . for a word @xmath143 ,
let @xmath144 denote the number of letters in the word @xmath145 , and for an element @xmath95 , let @xmath146 be the length of the shortest word over the generating set @xmath147 that represents @xmath112 . following the notation of horak , stein and taback , the length @xmath146 can be described in terms of the reduced tree pair diagram @xmath148 for @xmath112 , with carets numbered in infix order .
first , we say that caret number @xmath110 in a tree @xmath106 has _
type n _ if caret @xmath149 is an interior caret which lies in the right subtree of @xmath110 .
caret pair @xmath110 in the reduced tree pair diagram @xmath150 is a _ penalty caret pair _ if either 1 .
caret @xmath110 has type @xmath151 in either @xmath122 or @xmath123 , and is not a left caret in either tree , or 2 .
caret @xmath110 is a right caret in both @xmath122 and @xmath123 and caret @xmath110 is neither the final caret in the tree pair diagram , nor a left caret in either tree . using this notation ,
the following lemma is proved in .
[ lem : penalty ] for @xmath152 , the length @xmath153 , where @xmath154 is total number of carets in both trees of the reduced tree pair diagram which are not right carets , and @xmath155 is the number of penalty caret pairs .
it then follows that @xmath156 is a good estimate for the @xmath146 .
lemma [ lengthisn ] makes this relationship precise and is used in the proof that the tameness function of the combing we construct below is linear .
[ lengthisn ] for @xmath95 , @xmath157 .
lemma [ lem : penalty ] shows that each caret pair in the reduced tree pair diagram for @xmath112 contributes 0 , 1 , 2 , 3 or 4 to @xmath146 , and the upper bound on @xmath146 follows immediately .
the caret pair can contribute either 0 , 1 , or 2 to @xmath154 , and can contribute another 2 if it is a penalty pair . in order for a caret pair to contribute zero to the word length of the element
, both carets must be on the right side of the tree in order to not contribute to @xmath154 , and either one is the root ( in which case the pair is not a penalty pair because the root is also a left caret ) , or the pair is the last caret pair .
so at most two caret pairs do not contribute anything to @xmath146 , which yields the lower bound on @xmath146 .
finally , we include here a lemma which will be used in section [ sec : combing ] and describes a family of words in @xmath0 which are always nontrivial . [ consecutiveindices ]
let @xmath158 , and suppose @xmath159 where for each @xmath160 , either @xmath161 or @xmath162 .
in addition , suppose that for each @xmath160 , if @xmath163 , then @xmath164 or @xmath165 .
then @xmath166 in @xmath0 .
we prove the lemma by induction on @xmath114 .
the base case @xmath167 is trivial .
suppose @xmath168 , and the indices of the generators satisfy the hypothesis of the lemma , and let @xmath160 be the smallest index appearing in @xmath112 .
we will show that if @xmath169 , then we can obtain a shorter word satisfying the conditions on indices which is also @xmath54 in @xmath0 , contradicting the inductive hypothesis . utilizing the representation of elements of @xmath0 as piecewise linear homeomorphisms of the unit interval
( see for details ) , @xmath170 has a breakpoint at @xmath171 , and the right derivative is @xmath172 at that breakpoint , but @xmath173 has support in @xmath174\subset [ 1-(1/2)^i , 1]$ ] for @xmath175 .
it follows that the net exponent of all generators @xmath170 occurring in @xmath112 must be zero .
we can write @xmath176 , where for each @xmath177 , @xmath178 is a nontrivial word in generators of the form @xmath179 for @xmath180 and @xmath181 , except possibly @xmath182 and @xmath183 which may be zero . note that if either @xmath182 or @xmath183 are zero , then necessarily @xmath184 , and if both are zero , then @xmath185 . in any case ,
@xmath186 , and for some pair of indices @xmath29 and @xmath187 , @xmath188 and @xmath189 . case 1 : if for some @xmath177 , @xmath190 and @xmath191 , then let @xmath192 be the word obtained from @xmath178 by increasing the index of each generator by 1
. then @xmath193 in @xmath0 .
furthermore , as @xmath160 is the minimal index in the word @xmath112 , we know that @xmath192 begins and ends with @xmath194 , @xmath195 ends in @xmath196 ( or does not exist if @xmath197 ) , and @xmath198 begins in @xmath196 ( or does not exist if @xmath199 ) , and so replacing @xmath200 by @xmath192 produces a word of length @xmath201 satisfying the hypotheses of the lemma .
case 2 : if no such index @xmath177 exists , and neither @xmath182 nor @xmath183 is zero , then @xmath112 begins with @xmath92 and ends with @xmath100 . therefore , since @xmath169 in @xmath0 , @xmath202 as well , and if @xmath203 is the word of length @xmath201 obtained from @xmath112 by deleting the first and last letters , then @xmath204 in @xmath0 , and @xmath203 satisfies the hypotheses of the lemma .
case 3 : if no such index @xmath177 exists , and @xmath205 , then @xmath206 and @xmath207 . let @xmath208 be the word @xmath209 with the index of each generator increased by one .
then since @xmath208 ends in @xmath194 and @xmath210 begins in @xmath196 , then replacing @xmath211 by @xmath212 results in a word of the same length which still satisfies the hypotheses .
either this new word satisfies the conditions of case 2 , or else it does not end in @xmath100 .
but if not , then @xmath213 , and one can do a similar substitution at that end to obtain a new word ending in @xmath100 and beginning in @xmath92 which satisfies the conditions of case 2 . applying the argument in case 2 to this new word yields a word of length @xmath201 satisfying the hypotheses of the lemma . in general , there are many minimal length representatives of elements of f with respect to the standard finite generating set , and fordham described effective methods for finding all such minimal length representatives .
cleary and taback described a straightforward procedure which canonically produces a minimal length element ( with respect to the generating set @xmath214 ) for a purely positive or purely negative element in @xmath0 ; that is , an element @xmath112 whose infinite normal form @xmath215 satisfies @xmath216 ( hence contains only terms with positive exponents ) or @xmath217 ( hence contains only terms with negative exponents ) .
they call these paths _ nested traversal paths _ due to their construction .
the combing paths used below will be built from concatenating and then freely reducing these nested traversal paths .
let @xmath95 be a strictly negative element ; that is , @xmath217 , and @xmath112 is represented by a reduced tree pair diagram of the form @xmath218 , where @xmath219 is a tree consisting only of @xmath220 right carets . to construct the nested traversal path corresponding to @xmath112 , we proceed as follows .
we number the carets of the tree @xmath122 in infix order , beginning with @xmath54 .
we proceed through the carets in infix order , adding generators @xmath90 , @xmath221 , and @xmath222 to the right end of the nested traversal path at each step according to the following rules . 1 .
if the infix number of the caret is @xmath54 , add nothing to the nested traversal path .
2 . if the caret is a left caret with infix number greater than @xmath54 , add @xmath221 to the nested traversal path .
3 . if the caret is an interior caret , let @xmath106 be the right subtree of the caret .
if @xmath106 is nonempty , add @xmath223 to the nested traversal path , where @xmath224 is the nested traversal path obtained by following these rules for the carets of @xmath106 .
if the caret is an interior caret and the right subtree of @xmath106 is empty , then add @xmath225 to the nested traversal path .
5 . if the caret is a right non - root caret , and its right subtree @xmath106 contains an interior caret , add @xmath226 to the nested traversal path , where @xmath224 is as above .
6 . if the caret is a right non - root caret , and its right subtree @xmath106 contains no interior carets , then add nothing to the nested traversal path .
it is proved in that this method produces a minimal length word representing a negative element @xmath105 of @xmath0 , with respect to the generating set @xmath227 .
we denote this nested traversal path for @xmath105 by @xmath228 .
for a reduced tree pair diagram , the situation in rule ( 6 ) above never arises . including it allows one to extend the algorithm to tree pair diagrams obtained by appending only right carets to both the last leaf in the tree @xmath122 and to the last leaf of @xmath219 without changing the word produced by the algorithm .
we define the _ nested traversal normal form _
@xmath229 of an element @xmath230 as follows .
let @xmath231 be the infinite normal form for @xmath112 .
then the element @xmath232 , represented by the ( not necessarily reduced ) tree pair diagram @xmath233 , is strictly negative , and so has a nested traversal path formed according to the above rules , which is not affected by the possible reduction of the diagram , according to rule ( 6 ) of the procedure above .
hence we can define the nested traversal normal form for @xmath104 to be @xmath234 .
it follows from the nested traversal construction that the words @xmath235 and @xmath228 are freely reduced , considered separately . however , their concatenation @xmath236 may not be , so we define @xmath229 , the nested traversal normal form for @xmath112 , to be the result of freely reducing the word @xmath237 .
note that @xmath229 is not necessarily a minimal length word representing the element @xmath112 .
cleary and taback show in the proof of theorem 6.1 of that along a strictly negative nested traversal normal form @xmath238 , the number of carets in the tree pair diagrams corresponding to the prefixes @xmath239 for @xmath240 never decreases , that is , @xmath241 .
this follows from the construction of the paths : the multiplication @xmath242 never causes a reduction of carets .
we prove below that the same holds for the general nested traversal normal form @xmath229 .
our proof of the tameness of the 0-combing given by the nested traversal normal forms uses this property combined with the relationship between @xmath146 and @xmath156 described in lemma [ lengthisn ] .
[ nincreases ] for @xmath95 , if @xmath243 , then @xmath244 for all @xmath245 .
we first claim that for any @xmath246 and generator @xmath247 , carets can not be both added and removed in the process of multiplying @xmath248 .
we check one case of this for the reader , in which @xmath249 and the right child of the root caret in @xmath250 exists and has an exposed left leaf labeled @xmath84 . in this case
it is necessary to add a single caret to leaf @xmath84 of both @xmath250 and @xmath251 , which has exposed leaves numbered @xmath84 and @xmath252 . before reduction of carets , we obtain a possibly unreduced tree pair diagram @xmath253 in which leaves @xmath84 and @xmath252 of @xmath254 no longer form a caret .
any exposed caret with leaves numbered greater than @xmath252 in @xmath255 has its leaf labels increased by @xmath54 in @xmath253 .
thus if a caret pair is exposed in @xmath253 , it would have been exposed in @xmath255 .
however , @xmath255 was reduced , and hence @xmath253 is reduced as well , and so equals @xmath256 .
other cases are checked similarly .
note that @xmath257 and @xmath258 can be related in one of the following three ways . either , * @xmath259 if carets must be added in order to perform the multiplication , or * @xmath260 if no carets must be added in order to perform the multiplication and no carets must be removed in order to reduce the resulting tree pair diagram , or *
@xmath261 if no carets must be added in order to perform the multiplication , but carets must be removed to reduce the resulting tree pair diagram .
we remark that in any case , @xmath257 and @xmath258 differ by at most 1 when @xmath262 and by at most 2 when @xmath263 .
in addition to the possible change in the number of carets , the resulting trees are rearranged slightly , in very constrained ways .
we will first prove the theorem in the case where @xmath264 .
observe that for any @xmath246 , either * @xmath265 has one more interior caret than @xmath251 , both @xmath266 and @xmath250 have the same number of interior carets , and @xmath267 , or * @xmath266 has one fewer interior caret than @xmath250 , both @xmath265 and @xmath251 have the same number of interior carets , and @xmath268 . to see this , note that if carets must be added in order to perform the multiplication @xmath269 ; that is , if either the root of @xmath250 does not have a right child , or this root does have a right child but this child does not have a left child , then @xmath265 has one more interior caret than does @xmath251 . in that case , @xmath266 is the tree @xmath250 with one or two right carets added to the last leaf , and the number of interior carets in the negative tree is preserved .
also as noted above , no carets can be removed , and hence @xmath270 . on the other hand , if no carets are added in performing the multiplication @xmath269 , then @xmath268 .
moreover , during the multiplication process , the left child of the right child of the root of @xmath250 is an interior caret , but in the tree @xmath266 , either this caret has been removed in the multiplication process , or the caret with this same number is a right caret , and hence either way , the number of interior carets in @xmath266 is strictly less than that of @xmath250 . if carets are removed , they must be the final one or two carets of the tree pair , and these must be right carets in the tree @xmath251 , so the number of interior carets in the positive tree is left unchanged .
similarly , one can verify that either * @xmath271 has one more interior caret than @xmath250 , both @xmath272 and @xmath251 have the same number of interior carets , and @xmath273 , or * @xmath272 has one fewer interior caret than @xmath251 , both @xmath271 and @xmath250 have the same number of interior carets , and @xmath274 .
furthermore , if @xmath275 , then @xmath276 and @xmath250 have the same number of interior carets , as do @xmath277 and @xmath251 . now in the (
possibly not freely reduced ) word @xmath236 , all occurrences of @xmath91 precede all occurrences of @xmath225 .
if @xmath122 has @xmath29 interior carets and @xmath123 has @xmath187 interior carets , then by construction , @xmath236 has precisely @xmath187 occurrences of @xmath91 followed by @xmath29 occurrences of @xmath225 .
thus in any prefix of @xmath229 , if @xmath278 it is always true that the number of interior carets in the tree pair diagram corresponding to @xmath279 is one more than the number of interior carets in the tree pair diagram corresponding to @xmath280 . hence , it follows from the previous observations that if @xmath269 is a prefix of @xmath235 , then @xmath265 has one more interior caret than @xmath251 , and @xmath281 .
similarly , if @xmath282 is a prefix of @xmath228 , then @xmath283 has one more interior caret than @xmath284 , and @xmath285 , which proves the theorem in the case @xmath264 .
in addition , we can conclude from this analysis a few more facts about the relationship between the reduced word @xmath229 and the potentially longer word @xmath236 , which we note here for use again later .
in particular , @xmath286 , where @xmath287 and @xmath288 for some @xmath289 , and if @xmath209 and @xmath210 are both nonempty words , then either @xmath209 ends in @xmath91 or @xmath210 begins with @xmath225 , but not both . moreover , for any prefix @xmath290 of @xmath209 , @xmath291 contains no interior carets . to prove the theorem for the cases where @xmath292
, we repeatedly refer to the following six facts , each of which can be deduced by carefully following the process of multiplying by a generator .
1 . @xmath293 if and only if the first caret of @xmath251 is exposed , and the first two right carets of @xmath250 have no left children . 2 .
@xmath294 if and only if the last caret of @xmath251 is exposed , and the last two left carets of @xmath250 have no right children . 3 .
if @xmath295 , then the root caret of @xmath296 has a left child . 4 .
if @xmath297 , then the root caret of @xmath298 has a right child .
if @xmath299 , then the second right caret of @xmath271 has a left child , so in particular the root caret has a right child . 6 .
if @xmath300 , then the root caret of @xmath266 has a right child .
we proceed by induction to prove the theorem . clearly @xmath301
, so now assume that @xmath302 for all @xmath303 .
case 1 : @xmath304 . either @xmath305 , @xmath306 , or @xmath307 .
but then , since by the inductive hypothesis @xmath308 , facts 4 , 5 , and 6 show that the root caret of @xmath309 always has a right child .
hence , fact 2 above implies that @xmath310 .
case 2 : @xmath311 .
if @xmath312 or @xmath306 , then since by the inductive hypothesis @xmath308 , facts 3 and 5 above show that either the root caret or the second right caret of @xmath309 has a left child .
hence , in these cases fact 1 above implies that @xmath310 .
so we must check the one remaining possibility , that @xmath307 .
we claim that for @xmath307 , either the root caret of @xmath309 has a left child , or the first caret of @xmath313 is not exposed , which will again imply by fact 1 above that @xmath310 .
to verify the claim , first note that since the letter @xmath91 can not occur in @xmath228 , the word @xmath290 is a prefix of @xmath209 , and so neither @xmath314 nor @xmath309 contain any interior carets .
now if the root caret of @xmath314 has a left child , then so does the root caret of @xmath309 . on the other hand ,
if @xmath314 consists only of right carets , then @xmath313 is obtained by hanging a caret from the second leaf of @xmath315 , so that in this case , the first caret of @xmath313 is not exposed .
in addition , we remark that the nested traversal forms , which are certainly not in general geodesics , are at least quasigeodesics . to see this , it is helpful to first make some preliminary observations about nested traversal paths for strictly negative words , each of which can be deduced from the algorithm for the construction of nested traversal paths .
[ nestedtraversalproperties ] 1 . a word @xmath316 , where @xmath317 for each @xmath160 , is a nested traversal path if and only if the word satisfies the following three conditions : 1 .
the word is freely reduced .
the exponent sum of @xmath90 in any prefix @xmath318 for @xmath319 is not positive .
3 . if @xmath320 for some @xmath114 , then @xmath321 for @xmath322 .
+ hence , any prefix of a nested traversal path is again a nested traversal path .
2 . let @xmath316 be a nested traversal path with reduced tree pair diagram @xmath323 for some @xmath324 .
then the first caret of @xmath325 is exposed if and only if @xmath326 and the exponent sum of @xmath90 in every nonempty prefix @xmath318 is strictly negative .
note that the final caret of @xmath325 , always a right caret , is never exposed .
3 . if @xmath316 is a nested traversal path with reduced tree pair diagram @xmath323 for some @xmath324 , the numbers labeling the exposed carets of @xmath325 can be algorithmically determined as follows .
caret 2 is exposed if @xmath327 , and not if @xmath326 .
so by reading through the prefix @xmath328 , it can be determined whether or not caret 2 is exposed .
inductively , suppose that by reading through a prefix @xmath318 , you have decided whether or not carets @xmath329 through @xmath177 are exposed .
then for caret @xmath330 , if + @xmath331 in the latter three cases , then move on to @xmath332 to make a decision about whether caret @xmath333 is exposed . with this observation in hand
, we are ready to prove that nested traversal normal forms are quasigeodesics .
recall that for a group @xmath3 with finite generating set @xmath4 , a word @xmath145 is a @xmath334-quasigeodesic for constants @xmath335 and @xmath336 if the unit speed path @xmath337 \rightarrow \gamma(g , a)$ ] labeled by @xmath145 satisfies @xmath338 for all @xmath339 $ ] . for every @xmath95 the nested traversal normal form
@xmath229 is a @xmath334-quasigeodesic with @xmath340 and @xmath341 .
let @xmath95 . from facts noted in the proof of theorem [ nincreases ] , we can write @xmath342 and @xmath343 for some @xmath344 and each @xmath345 .
it follows from the formula for length in lemma [ lem : penalty ] that each pair of carets in @xmath122 and @xmath123 , other than the first pair and the last pair , contribute some nonzero number between one and 4 to @xmath146
. however , each such caret pair contributes at most 6 to @xmath346 , so the length contribution to the normal form is at most 6 times the contribution to length in the cayley graph .
the first and last caret pairs are a slightly special case , since together they may contribute 0 , 1 , or 2 to both @xmath146 and @xmath346 .
however , one checks that if there is no contribution to @xmath146 from these carets , then the first caret in both trees is the root caret , and so the contribution to @xmath346 from these carets will be zero as well .
thus we obtain @xmath347 for any nested traversal normal form @xmath229 .
we now show that these inequalities hold for an arbitrary subword of @xmath229 .
so suppose @xmath348 , with @xmath349 .
if @xmath350 is a letter of @xmath209 , then since @xmath351 is a geodesic , then @xmath352 . the same argument , using the fact that @xmath228 is a geodesic , holds in the case where @xmath161 is a letter in @xmath210 .
now assume @xmath290 is a prefix of @xmath209 and @xmath353 is a suffix of @xmath210 .
let @xmath354 be the suffix of @xmath209 starting with @xmath161 , and let @xmath355 be the prefix of @xmath210 ending with @xmath350 ; then @xmath356 .
we prove below that for the subword @xmath357 , @xmath358 and @xmath359 for some @xmath360 .
this then implies that @xmath361 , and therefore @xmath357 is a nested traversal normal form , and @xmath362 . since @xmath363 and @xmath364 are prefixes of @xmath228 and @xmath365 , they are also nested traversal paths according to the first part of observation [ nestedtraversalproperties ] .
let @xmath366 denote the reduced tree pair diagram for @xmath363 and let @xmath367 be the reduced tree pair for @xmath364 .
there is a two step process to transform the pair of trees @xmath368 into the reduced tree pair diagram for @xmath356 . in the first step ,
if the numbers @xmath114 and @xmath324 of carets in each of these trees are not equal , we add a string of @xmath369 right carets to the final leaf of the smaller tree .
the second step consists of reducing the resulting tree pair .
note that by construction , the final caret of both @xmath254 and @xmath370 can not be exposed , and so any carets added in the first step will not be removed in the second step .
each time a caret is removed , there is a corresponding change to the pair of nested traversal paths , which we track below ; at the end of this process , we obtain the nested traversal paths for @xmath371 and @xmath372 .
if caret 1 is exposed in both trees , it must be that @xmath373 , so removing this caret pair corresponds algebraically to canceling the central @xmath374 pair to obtain the word @xmath375 .
note that both @xmath376 and @xmath377 are both again nested traversal paths . now repeat , and eventually reach a point where caret 1 is not exposed in one of @xmath378 and @xmath379 , for some @xmath360 .
note that if @xmath380 , caret 1 can not be exposed in both trees .
hence this process must successfully terminate .
next we check for possible reduction of caret pairs numbered greater than 1 .
part 3 of observation [ nestedtraversalproperties ] shows that caret 2 can only be exposed in both trees @xmath378 and @xmath379 if @xmath380 and both @xmath381 and @xmath355 start with the letter @xmath225 ; however , @xmath355 is a prefix of the word @xmath210 and @xmath354 is a suffix of @xmath209 , and the word @xmath382 is freely reduced . for caret pairs numbered between 3 and @xmath383 , the algebraic criteria from part ( 3 ) of observation [ nestedtraversalproperties ] by which we check for exposure of these carets only depends upon letters in the words @xmath355 and @xmath381 which , as noted above , are prefixes of the words @xmath210 and @xmath384 , respectively . since the tree pair diagram for @xmath382 is reduced , then none of these carets can be removed .
carets numbered above @xmath383 were added in the first step , and hence also can not be removed .
hence , @xmath385 , so @xmath361 , and the inequalities follow . thus nested traversal normal forms are ( 6,0)-quasigeodesics .
in this section , we construct a @xmath54-combing of the group @xmath0 with respect to the presentation @xmath9,[x_0x_1^{-1},x_0^{-2}x_1x_0 ^ 2 ] \rangle~;\ ] ] in section [ sec : tameness ] we will show that this combing satisfies a linear radial tameness function . let @xmath10 be the cayley complex for this presentation .
we first construct a @xmath53-combing of @xmath0 with respect to @xmath386 by defining a continuous function @xmath73 \rightarrow x^1 $ ] where , for any @xmath95 , the restriction @xmath387 \rightarrow x^1 $ ] is labeled by the nested traversal normal form @xmath229 for @xmath112 .
we call this 0-combing @xmath71 the _ nested traversal 0-combing_. now we must extend this 0-combing to a 1-combing .
all edges in the cayley graph fall into one of two categories , good " and bad " .
the good edges consist of those edges where the combing path to one endpoint contains the other endpoint , and thus points along that edge are combed through the @xmath54-skeleton .
the bad edges include all of those edges where this is not the case , and thus the points along the edge in question must be combed through the @xmath329-skeleton . to make this more formal , we introduce some notation . for each @xmath95 ,
let @xmath388 be the 0-combing path in @xmath36 from the identity to @xmath112 ( labeled by @xmath229 ) , and let @xmath389 be the inverse path from @xmath112 to the identity .
recall that each directed edge in the cayley graph @xmath390 is labeled either by the generator @xmath90 or the generator @xmath91 .
we formalize the notion of good and bad edges in the following definition .
if the set of endpoints of an edge @xmath391 is of the form @xmath392 , we denote the edge as @xmath393 , and if the set of endpoints of an edge @xmath391 is of the form @xmath394 , we denote the edge as @xmath395
. moreover , for the edge @xmath396 , where @xmath397 , if the loop @xmath398 is homotopic to the trivial loop in the cayley graph , we call the edge @xmath396 a _ good edge _ , and if not , we call the edge @xmath396 a _ bad edge_. theorem [ combingextends ] describes conditions on the cayley complex which allow us to extend the nested traversal 0-combing to a 1-combing .
[ combingextends ] let @xmath399 be the set of bad edges in the cayley complex @xmath10 with respect to the nested traversal 0-combing .
suppose that 1 .
there is a partial ordering of @xmath399 with the property that for any edge @xmath400 , the set of edges @xmath401 is finite , and 2 .
there is a function @xmath107 from @xmath399 to the set of 2-cells of @xmath10 , so that for every @xmath402 the edge @xmath391 is on the boundary of @xmath403 , and whenever @xmath404 is another edge on the boundary of @xmath403 , then @xmath405 .
then the nested traversal 0-combing @xmath73 \rightarrow x^1 $ ] can be extended to a 1-combing @xmath406 \rightarrow x^2 $ ] .
we remark that the hypotheses of the theorem imply that the mapping from bad edges to 2-cells is injective .
let @xmath407 be the set of good edges .
we extend the 0-combing in two stages .
first , extend @xmath73 \rightarrow x^1 $ ] to @xmath408 \rightarrow x^1 $ ] using the homotopies for the good edges .
next , note that the partial ordering on @xmath399 is well - founded , and so we may apply noetherian induction to define @xmath71 on bad edges as follows .
suppose we have already extended the combing to @xmath409 \rightarrow x^2 $ ] , where @xmath410 for a particular edge @xmath391 . since @xmath403 is a 2-cell , there is a homotopy from @xmath391 , through @xmath403 , to the remainder of the boundary @xmath411 excluding the interior @xmath412 of @xmath391 , which fixes the endpoints of @xmath391 .
more specifically , let @xmath413 { \rightarrow}c(e)$ ] satisfy : for each point @xmath110 in the edge @xmath391 , @xmath414 and @xmath415 ; the image @xmath416 ; and for the endpoints @xmath37 and @xmath131 of @xmath391 and for all @xmath417 $ ] , @xmath418 and @xmath419 .
since all edges in @xmath420 are in @xmath421 , the combing @xmath422 \rightarrow x^2 $ ] provides combing paths from the identity to each of the points of @xmath420 .
reparametrize these paths and concatenate them with the paths from the homotopy @xmath423 to define the homotopy @xmath424 { \rightarrow}x$ ] .
this yields a homotopy @xmath425 \rightarrow x^2 $ ] .
then , by induction , the 0-combing extends to a 1-combing @xmath406 \rightarrow x^2 $ ] .
the remainder of this section is devoted to establishing the hypotheses of theorem@xmath426 .
the goal of this section is to identify the good edges in @xmath427 .
this is accomplished in the following theorem .
[ goodedges ] let @xmath158 have reduced tree pair diagram @xmath148 . if any one of the following four conditions holds , then the edge @xmath396 , with @xmath397 , is a good edge : 1 .
the index @xmath428 .
the tree @xmath122 has at most two right carets .
3 . the tree @xmath122 has at least three right carets , no carets need be removed to reduce the tree pair diagram for @xmath429 , and all carets following the third right caret in @xmath122 , if any , are right carets .
the tree @xmath122 has at least three right carets but no interior carets , caret @xmath84 must be removed to reduce the tree pair diagram for @xmath429 , and caret @xmath84 is the first exposed caret in @xmath123 .
we prove this theorem in two lemmas , considering separately the cases @xmath393 and @xmath395 .
to prove each lemma , we simply compare the nested traversal forms for the two endpoints of the edge in each situation in the hypotheses of theorem [ goodedges ] .
[ x0generator ] let @xmath95 .
then @xmath393 is a good edge .
let @xmath430 .
we compare @xmath229 and @xmath431 , and show that either @xmath432 or @xmath433 , and so it follows immediately that @xmath393 is a good edge . as usual
, let @xmath121 and @xmath434 denote reduced tree pair diagrams .
the tree pair diagram @xmath435 for @xmath221 is given in figure [ fig : generators ] .
suppose first that @xmath122 has only one right caret , the root caret .
the left subtree of the root caret must then be nonempty ; let @xmath436 be this subtree , and let @xmath437 be the substring of @xmath228 consisting of all generators corresponding to carets in @xmath436 .
then @xmath438 . in multiplying @xmath430 ,
a caret is appended to the rightmost leaf of each of the trees for @xmath112 , and the tree @xmath436 is appended to the leftmost leaf of the trees @xmath439 and @xmath440 for @xmath221 .
then @xmath434 where @xmath441 consists of a root caret with a left child whose left subtree is @xmath436 , and @xmath442 is @xmath123 with a single caret appended to its rightmost leaf .
this pair is reduced , so no caret is removed in performing this product .
the tree @xmath441 has a left caret between the subtree @xmath436 and the root , so @xmath443 . since @xmath442 is just @xmath123 with a single caret appended to its rightmost leaf , @xmath444 .
hence in this case , @xmath445 .
therefore , if @xmath229 does not end in @xmath90 , then @xmath432 . however , if @xmath229 does end in @xmath90 , then @xmath446 .
for the remainder of this proof suppose that @xmath122 has at least two right carets .
let @xmath436 be the left subtree of the root caret , let @xmath447 denote the left subtree of the right child of the root caret , and let @xmath448 denote the right subtree of the right child of the root .
let @xmath437 , @xmath449 , and @xmath450 denote the subwords of @xmath228 corresponding to the carets of these subtrees ; note that any of these trees can be the empty tree , and if so , the corresponding subword will be empty
. define @xmath451 if the tree @xmath436 is the empty tree @xmath452 with no carets , and @xmath453 if @xmath454 , so that @xmath455 is the contribution of the root caret of @xmath122 to the nested traversal normal form @xmath228 .
the nested traversal normal form for @xmath112 is then @xmath456 in this case no carets need to be added to the tree pair for @xmath112 in order to perform the multiplication @xmath457 ; the trees @xmath436 , @xmath447 , and @xmath448 must be appended to leaves 0 , 1 , and 2 , respectively of the trees @xmath439 and @xmath440 .
a caret must be removed from the product @xmath457 to obtain the reduced tree pair diagram if and only if the trees @xmath436 and @xmath447 are both empty , and caret 1 in the tree @xmath123 is exposed . in this case , @xmath442 is the tree @xmath123 with the first caret removed .
note that caret 1 of @xmath123 contributed nothing to the nested traversal normal form @xmath365 , and that caret 2 of @xmath123 must also be a left caret , and so contributed @xmath221 .
the latter caret is caret 1 of @xmath442 .
hence @xmath458 . analyzing the negative trees
, we note that the tree @xmath441 is the tree with a single left caret , namely the root caret , having a right subtree given by @xmath448 , and so @xmath459 . if @xmath460 for some @xmath461 , then @xmath433 . on the other hand ,
if @xmath462 for any @xmath114 , then since @xmath448 is the right subtree of the root of @xmath441 , this subtree gives a nonempty contribution to the nested traversal path , and hence @xmath463 can not end with the letter @xmath221 .
thus when freely reducing the word @xmath464 , only one @xmath374 is removed , and @xmath433 . finally , suppose that no carets need to be removed in the multiplication @xmath457
. then @xmath465 and @xmath466 .
therefore @xmath467 . if @xmath460 , then @xmath468 , and so either @xmath432 , when @xmath229 does not end in @xmath90 , or else @xmath446 .
on the other hand if @xmath462 , then @xmath469 .
since @xmath463 can not end in @xmath221 , @xmath433 .
next , we turn to edges of the form @xmath395 .
for such an edge , the case where @xmath122 has at least three right carets is by far the most complicated , so before stating the desired lemma , we establish some useful notation for that case .
[ not:4trees ] for any @xmath470 such that @xmath122 has at least 3 right carets :
* let @xmath436 denote the left subtree of the root caret , @xmath447 denote the left subtree of the right child of the root , and @xmath471 and @xmath472 denote the left and right subtrees , respectively , of the third right caret of @xmath122 .
* define @xmath132 ( as above ) , and @xmath473 and @xmath474 . *
define @xmath475 to be the number of the first exposed caret of @xmath123 . to understand the good edges , and later the definition of the partial order on the edges
, one must first understand explicitly how the tree pair diagram for @xmath112 may change when @xmath112 is multiplied by @xmath225 . to form the product @xmath429 ,
if @xmath122 contains at least 3 right carets , then no carets must be added to the trees of the reduced pair diagram for @xmath112 , but the subtrees @xmath436 , @xmath447 , @xmath471 , and @xmath472 are appended to the leaves numbered 0 , 1 , 2 , and 3 , respectively , of the trees in the reduced tree pair diagram for @xmath225 , which is given in figure [ fig : generators ] . continuing the case that @xmath122 contains at least 3 right carets ,
let @xmath254 be the negative tree of the intermediate step in the multiplication , @xmath429 before any carets are removed to reduce the tree pair diagram .
the carets of @xmath122 and @xmath254 with the same number have the same type ( left , right or interior ) in both trees , with the exception of the caret numbered @xmath476 , which is a right caret in @xmath122 and an interior caret in @xmath254 .
this is the only caret that can be exposed in @xmath254 but not in @xmath122 , and hence the only caret that might be removed if @xmath254 is not reduced . as a consequence
, a caret must be removed in the multiplication @xmath429 if and only if the following property holds : @xmath477 : : @xmath478 , @xmath479 , and @xmath123 has an exposed caret at caret number @xmath480 .
we are now ready to describe edges of the form @xmath395 that are good edges .
[ x1generator ] let @xmath158 .
then @xmath395 is a good edge if any of the following are satisfied : 1 .
@xmath122 has at most two right carets .
@xmath122 has 3 or more right carets , property @xmath477 is not satisfied and @xmath481 for some @xmath482 .
@xmath122 has 3 or more right carets but no interior carets , property @xmath477 holds and the number @xmath84 of the caret that cancels satisfies @xmath483 , so caret @xmath84 is the first exposed caret in @xmath123 . let @xmath484 .
again we compare @xmath229 with @xmath485 , and we claim that when @xmath112 satisfies either of the first two conditions of the hypothesis , then @xmath486 . if @xmath112 satisfies the third condition , then @xmath487 . so in all cases it follows immediately that the edge @xmath395 is good .
we proceed by cases according to which hypothesis is satisfied .
the first two are straightforward , but for the third we separate into subcases . _ case 1 .
@xmath122 has fewer than three right carets .
_ suppose first that @xmath122 has only one right caret , and @xmath436 is the left subtree of this root caret . in multiplying @xmath484 ,
two right carets are appended to the rightmost leaf of each of the trees for @xmath112 , and @xmath436 is appended to the leftmost leaf of the trees for @xmath225 , but no carets are removed .
the two appended right carets in @xmath251 contribute nothing to @xmath488 , so @xmath489 .
the root caret of @xmath250 has left subtree @xmath436 , and the right child of the root has left subtree consisting of a single interior caret which contributes @xmath225 to the nested traversal normal form @xmath490 .
then @xmath491 , so it follows that @xmath486 .
the proof in the case that @xmath122 has two right carets is similar . for the remainder of the proof , assume that @xmath122 has at least three right carets .
let @xmath437 , @xmath455 , @xmath449 , @xmath492 , and @xmath493 be the subwords of the nested traversal normal form @xmath228 corresponding to the carets of @xmath436 , the root , @xmath447 , @xmath471 , and @xmath472 , respectively . in this case the nested traversal normal form for @xmath112 is then @xmath494 _ case 2 .
no carets must be removed to create the reduced tree pair diagram for @xmath484 , and @xmath481 for some @xmath495 . _
it follows immediately that @xmath496 and hence @xmath497 . from the discussion of @xmath122 and @xmath498 above ,
only caret number @xmath499 makes a different contribution to the respective nested traversal normal forms , yielding : @xmath500 comparing these words with the corresponding words @xmath236 given above yields @xmath486 .
caret @xmath84 is removed when we form the product @xmath429 ( equivalently , property @xmath477 holds ) , @xmath501 , and @xmath122 has no interior carets .
_ from @xmath477 , this caret necessarily has caret number @xmath502 . as @xmath122 has no interior carets in case 3 , we must have @xmath503 , the tree with @xmath504 left carets , where @xmath505 and @xmath481 for some @xmath461
. then @xmath506 .
when caret @xmath84 is removed to form the tree pair diagram for @xmath357 , we see that @xmath250 then has @xmath507 left carets including the root and @xmath508 right non - root carets , and so @xmath509 as well .
note that @xmath510 , and so caret @xmath84 of @xmath123 is neither the first nor the last caret of this tree .
then this is an interior caret of @xmath123 which is an exposed caret , in particular it has an empty right subtree .
this caret will contribute @xmath225 to the nested traversal normal form @xmath365 .
the tree @xmath251 is the tree @xmath123 with caret @xmath84 removed . for @xmath511 ,
let @xmath512 denote caret @xmath177 of the tree @xmath123 .
caret @xmath513 contributes nothing to the nested traversal normal form @xmath365 .
whenever @xmath514 , the unexposed caret @xmath512 is either an interior caret or a right caret , and in both cases @xmath512 has a nonempty right subtree containing the interior caret @xmath515 .
hence each of these carets @xmath512 adds @xmath221 to @xmath365 before the subword @xmath225 corresponding to caret @xmath515 , and also adds either @xmath516 or @xmath90 to @xmath365 after this subword .
then @xmath517 for some word @xmath518 , and hence @xmath519 , so @xmath520 . to analyze the nested traversal normal forms @xmath488 and @xmath485 further
, we now divide into four subcases , as follows . _
suppose @xmath521 .
_ then it follows that @xmath522 , and the tree @xmath251 is @xmath123 with caret @xmath522 removed .
hence @xmath523 , and so @xmath487 .
_ case 3b .
suppose @xmath524 , caret @xmath84 is the left child of its parent caret in @xmath123 , and @xmath525 .
_ then it follows that all other carets of @xmath123 are right carets , or else a caret with infix number less than @xmath84 would be the first exposed caret in @xmath123 .
thus @xmath526 .
the tree @xmath251 contains only right carets , and so @xmath527 .
therefore the nested traversal normal form for @xmath112 is @xmath528 and so @xmath487 . _
suppose that @xmath524 , caret @xmath84 is the left child of its parent caret in @xmath123 , and @xmath529 .
_ caret @xmath530 is the parent of caret @xmath531 in this case .
if @xmath530 is an interior caret of @xmath123 , then @xmath530 is an interior caret contained in the right subtree of carets @xmath532 for all @xmath533 , and so in the tree @xmath251 , these carets @xmath512 also contain an interior caret in their right subtrees . if instead @xmath530 is a right caret , then @xmath512 is a right caret for all @xmath534 .
note that the final caret @xmath220 of @xmath122 is exposed , and the tree pair @xmath150 is reduced , so caret number @xmath535 of @xmath123 is not exposed .
then the left subtree of the latter caret contains an interior caret @xmath536 of @xmath123 with @xmath537 , and hence this interior caret is contained in the right subtrees of all of the carets @xmath512 with @xmath538 .
then for both types of parent caret the nested traversal path for @xmath539 is the same as that for @xmath232 except that the @xmath225 subword corresponding to caret @xmath84 is removed .
then @xmath540 , and so once again @xmath487 . _
suppose that @xmath524 , and caret @xmath84 is the right child of its parent in @xmath123 .
_ since @xmath541 , @xmath542 , and caret @xmath531 is the first exposed caret in @xmath123 , then caret @xmath543 must be an interior caret in @xmath123 , which is contained in the right subtree of each @xmath512 with @xmath544 .
then the nested traversal path for @xmath539 is the same as that for @xmath232 except that the @xmath545 subword of @xmath365 corresponding to carets @xmath546 and @xmath515 is replaced with the word @xmath225 corresponding to the caret @xmath543 of @xmath251 . in this case
@xmath547 , ( where @xmath548 ) , and @xmath549 . therefore , @xmath550 .
lemmas [ x0generator ] and [ x1generator ] complete the proof of theorem [ goodedges ] . in the next sections , we will most frequently apply the contrapositive of theorem [ goodedges ] , rewritten below following notation [ not:4trees ] .
[ newbad ] let @xmath95 .
if @xmath396 is a bad edge , then @xmath551 , the tree @xmath122 has at least 3 right carets , and either 1 .
@xmath553 , @xmath554 , property @xmath477 holds , and @xmath555 .
3 . @xmath553 , @xmath556 , and property @xmath477 holds . from theorem [ goodedges ] parts ( 1 ) and ( 2 ) we know that @xmath551 and @xmath122 contains at least 3 right carets .
if no caret is removed in the multiplication @xmath429 ( that is , if property @xmath477 fails ) , then part ( 3 ) of theorem [ goodedges ] shows that we must have @xmath552 .
if a caret is removed in the multiplication @xmath429 , then property @xmath477 holds , additionally , if we are not in either of the cases ( 1 ) or ( 3 ) of this corollary , then we have @xmath557 and @xmath554 . in this case , @xmath122 has no interior carets , so @xmath558 and the caret that is canceled in the multiplication is caret number @xmath480 , which must be exposed in @xmath123 .
it follows from part ( 4 ) of theorem [ goodedges ] that this caret is not the first exposed caret in @xmath123 , and so @xmath559 .
however , two consecutive carets can not be exposed , and we conclude that @xmath560 . furthermore ,
if @xmath561 , then caret 1 would be exposed in both @xmath122 and @xmath123 and the tree pair diagram would not be reduced .
hence , @xmath562 , and case ( 2 ) of the corollary holds .
we now define a partial order on the set of all bad edges @xmath395 as required for theorem@xmath426 .
this partial order is based on numerical measures related to the tree pair diagram for @xmath112 .
these include @xmath156 , as well as @xmath563 and @xmath564 , the number of carets in the subtrees @xmath436 and @xmath472 defined in notation [ not:4trees ] above . to order the edges @xmath395 and @xmath565 where the values @xmath566 and @xmath567 are the same for both elements , we first need to construct , for each fixed number @xmath114 , several different partial orderings of the set of all rooted binary trees with @xmath114 carets . before explaining these posets , we first need some additional combinatorial information associated to a rooted binary tree . [ generaltinfo ] let @xmath106 be a rooted , binary tree . * we order the right carets of @xmath106 in infix order , and call them @xmath568 , where @xmath569 is the root caret of @xmath106 .
let @xmath570 be the ( possibly empty ) left subtree of caret @xmath51 .
let @xmath571 , where @xmath160 is the smallest index with @xmath572 , with the property that for every @xmath573 , @xmath574 is empty .
* similarly , we call the left carets of @xmath106 , in infix order , @xmath575 , where @xmath576 is the root caret of @xmath106 , and let @xmath577 be the ( possibly empty ) right subtree of caret @xmath578 .
then let @xmath579 , where @xmath160 is the smallest index , @xmath580 , such that @xmath581 is empty for every @xmath582 .
* let @xmath583 where @xmath114 is the number of right carets in @xmath106 ; that is , @xmath584 is the number of carets in @xmath106 up to and including caret @xmath585 .
* let @xmath586 where @xmath587 is the number of left carets in @xmath106 ; that is , @xmath588 is the number of carets in @xmath106 after , and including , caret @xmath589 .
we remark that the simple condition of whether a tree consists either only of right carets or only of left carets , which was critical in recognizing bad edges in corollary [ newbad ] , simply translates into whether @xmath590 or @xmath591 equals zero .
more precisely , the condition @xmath592 ( respectively @xmath593 ) is equivalent to @xmath594 ( respectively @xmath595 ) .
similarly , the condition @xmath596 ( respectively @xmath597 ) is equivalent to @xmath598 ( respectively @xmath599 ) . in order to sort , rather than simply recognize , the bad edges , however , we need to keep track of the numerical values @xmath590 and @xmath591 .
consider the set of rooted binary trees with @xmath114 carets .
we define the * right poset of rooted binary trees with @xmath114 carets * which will be used to order edges @xmath395 where @xmath600 . for each tree
@xmath601 with @xmath114 carets with @xmath602 , we define the tree @xmath603 as follows : * if @xmath604 is odd , and @xmath605 , the left subtree of the root caret of @xmath601 , is empty , @xmath603 is the tree formed by rotating @xmath601 to the left at caret @xmath569 .
that is , if @xmath37 is the element of @xmath0 with tree pair diagram @xmath606 where @xmath133 is the tree consisting of @xmath114 right carets , then @xmath607 has ( possibly unreduced ) tree pair diagram @xmath608 . *
if @xmath604 is odd , and @xmath605 is not empty , @xmath603 is the tree formed by rotating @xmath601 to the right at caret @xmath569 .
that is , if @xmath37 is the element of @xmath0 with tree pair diagram @xmath606 , then @xmath609 has tree pair diagram @xmath608 . *
if @xmath604 is even , and @xmath610 , the left subtree of the right child of the root caret of @xmath601 , is empty , @xmath603 is the tree formed by rotating @xmath601 to the left at caret @xmath611 .
if @xmath37 is the element of @xmath0 with tree pair diagram @xmath606 , then @xmath612 has tree pair diagram @xmath608 . *
if @xmath604 is even , and @xmath610 is not empty , @xmath603 is the tree formed by rotating @xmath601 to the right at caret @xmath611 .
if @xmath37 is the element of @xmath0 with tree pair diagram @xmath606 , then @xmath613 has tree pair diagram @xmath608 .
now declare @xmath614 for every @xmath601 .
we claim that the transitive closure of this order is a well - founded partial order , with unique minimal element @xmath133 , the tree with @xmath114 right carets . to see this ,
notice that @xmath615 if and only if @xmath616 . now @xmath617 , and if @xmath618 , then @xmath604 and @xmath619 have different parities .
so if @xmath620 for some positive integer @xmath84 , this implies that there is a word @xmath621 ( where possibly the first and/or last generators are absent ) which is trivial in @xmath0 , contradicting lemma [ consecutiveindices ] .
since there are only a finite number of trees with @xmath114 carets , @xmath622 for some @xmath587 , and hence @xmath623 for some @xmath84 .
hence , we see that this is a partial order with a unique minimal tree @xmath133 , which is less than all other trees in the poset .
we denote the order in this poset by @xmath624 .
we now define the * left posets of rooted binary trees with @xmath114 carets * , which will be used to sort bad edges @xmath395 for which @xmath625 . using the method given above
, we could have constructed a poset using @xmath591 , @xmath577 , and @xmath626 instead of @xmath590 , @xmath570 , and @xmath627 , replacing the words rotate left `` by rotate right '' and vice - versa .
this yields a dual poset , where the minimal element is the tree @xmath134 consisting of only left carets .
we denote relationships in this order by @xmath628 . however ,
in some cases we will need a modification of this left poset in order to sort our edges , depending on an index @xmath629 . for any natural numbers @xmath630 and @xmath631 , let @xmath632 be a tree consisting of @xmath114 carets , none of which are interior , so that the root caret has infix number @xmath330 .
note that @xmath633 , the tree consisting of @xmath114 left carets . in the left poset with order relation @xmath634
, there is a unique path from each tree to the minimal element @xmath633 , and hence there also is a unique ( undirected ) path from each tree to @xmath632 . for each @xmath635
, we form a new poset , reordering the trees by declaring @xmath636 if @xmath637 is on the unique path from @xmath638 to @xmath632 . for each such @xmath177 ,
the new poset now has least element @xmath632 , and whereas @xmath639 , exactly the reverse holds in @xmath640 , namely @xmath641 .
if @xmath642 or @xmath114 , we use the original poset , and declare @xmath643 .
thus we have constructed only @xmath201 distinct posets in all , for each @xmath644 . in the trivial cases @xmath167 and @xmath645 , simply declare @xmath643 for any @xmath646 . to summarize : for each natural number @xmath644 , we have defined @xmath141 distinct partial orderings of the set of rooted binary trees with @xmath114 carets .
there is a unique right poset which has as minimal element @xmath133 , which will be used to sort bad edges @xmath395 with @xmath647 ; there is a family of @xmath201 distinct left posets which have , respectively , the trees @xmath632 for @xmath648 as unique minimal elements , which will be used to sort edges with @xmath625 and @xmath649
. the following notation , based on the quantities introduced in notation [ not:4trees ] and definition [ generaltinfo ] , will simplify the description of the ordering .
[ not : ss ] let @xmath395 be a bad edge , for an element @xmath650 .
* let @xmath651 .
* let @xmath652 .
* let @xmath653 .
* let @xmath654 .
* let @xmath655 be the infix number of the right caret of @xmath122 whose left subtree is not empty , but whose right subtree is either empty or consists only of right carets .
if no such caret exists , @xmath122 consists only of right carets , and we set @xmath656 . given above , the subtree @xmath472 ( resp .
@xmath436 ) has four right ( resp
. left ) carets .
we compute the following quantities : @xmath657 and @xmath658.,width=192 ] in the following definition , we define a set of comparisons between certain pairs of bad edges .
we then prove that the transitive closure of this set of order relationships is a partial order .
some details of this partial order ( particularly the fourth set of comparisons ) may seem mysterious at this point , but they are exactly the relationships needed for the cell map from the set of bad edges into the 2-cells which is defined in the next section to satisfy the hypotheses of theorem [ combingextends ] .
[ def : order ] let @xmath395 and @xmath659 be bad edges .
we say @xmath660 in the following situations : 1 .
if @xmath661 .
2 . if @xmath662 , @xmath465 , both @xmath663 and @xmath664 , and either : 1 . @xmath665 and @xmath666 , or 2 .
@xmath667 , @xmath668 , and @xmath669 .
3 . if @xmath662 , @xmath465 , @xmath670 , and either : 1 . @xmath671 and @xmath668 or 2 .
@xmath672 , @xmath666 , and @xmath673 for @xmath674 .
if @xmath675 , @xmath465 , exactly one of the pair @xmath676 is zero , and either : 1 .
@xmath677 , @xmath678 or @xmath329 , and @xmath679 , or 2 .
@xmath680 , @xmath681 , @xmath666 , and @xmath682 .
the transitive closure of the set of order relationships defined above is a partial order satisfying the property that for all bad edges @xmath391 , the set of bad edges less than @xmath391 with respect to this partial order is finite . in order to show this is a partial order , we must show that for every set of bad edges satisfying @xmath683 , @xmath684 .
suppose @xmath683 .
if @xmath685 is not constant for all @xmath160 , then @xmath686 , and so @xmath687 . so we may assume @xmath688 for @xmath689 .
next we observe that if @xmath690 for every @xmath689 , then for each @xmath160 either @xmath691 , or else @xmath692 and @xmath693 , so @xmath684 .
similarly , if @xmath694 for every @xmath160 , either @xmath695 , or else @xmath696 and @xmath697 for @xmath698 , so @xmath684 .
therefore , if @xmath699 the value of the variable @xmath590 must change twice in the sequence of edges between a strictly positive value and 0
. thus there must be indices for which the value of @xmath590 increases from 0 to a ( strictly ) positive number and for which the value decreases from positive to 0 .
in particular , there must be some index @xmath160 for which @xmath700 or 2 , @xmath701 , and @xmath702 .
but since for every index @xmath177 we have @xmath703 , then @xmath704 , and hence @xmath684 . finally , since the subset of all edges @xmath395 with a fixed value of @xmath156 is finite , the finiteness condition is satisfied and this partial order is well - founded . in this section
we define a mapping @xmath107 from the set of bad edges to the set of 2-cells in the cayley complex .
we will set up the map @xmath107 so that the bad edge @xmath395 is on the boundary of the cell @xmath705 . in order to specify this mapping
, we will first define notation for 2-cells in the cayley complex with a specified basepoint and orientation . for each vertex @xmath112 and edge @xmath395 in the cayley complex
, there are eight 2-cells containing this edge in their boundaries . for four of these 2-cells , there are 10 edges on the boundary ; these are the 2-cells labeled @xmath706 and @xmath707 in figure [ fig : relators1 ] . for the other four 2-cells
whose boundaries contain @xmath395 , there are 14 boundary edges ; these are the 2-cells labeled @xmath708 and @xmath709 in figure [ fig : relators2 ] . in each of these 2-cells , in addition
to @xmath395 the boundary contains three other edges of the form @xmath710 for some @xmath711 , and none of the @xmath712 edges in the boundary of a particular @xmath329-cell are adjacent . the edge @xmath395 will be referred to as the top @xmath712 edge in these eight 2-cells .
the @xmath712 edges closest to @xmath112 and @xmath429 are the left and right side edges @xmath713 and @xmath714 , respectively , and the last @xmath712 edge is the bottom edge @xmath715 . for a bad edge @xmath395 , the 2-cell @xmath705 must be chosen from among these eight cells . the map will be defined so that @xmath716 can be represented by a ( not necessarily reduced ) tree pair diagram @xmath717 , where the negative trees @xmath122 and @xmath718 differ by a single rotation at a particular caret , and the positive trees satisfy @xmath719 .
the notation @xmath720 ( where @xmath80 stands for relator , ) has been motivated by this .
the letter @xmath721 or @xmath722 depends on whether the rotation needed to transform @xmath122 to @xmath718 takes place at a left or right caret of @xmath122 .
the superscript @xmath723 takes into account the direction of this rotation , and the subscript @xmath84 specifies at which caret the rotation takes place .
more specifically , in the case of a rotation at a left caret , @xmath724 means this caret is the left child of the root of @xmath122 , while @xmath522 means rotation is at the left child of the left child of the root . in the case of a rotation at a right caret ,
if caret @xmath587 is the right child of the right child of the root of @xmath122 , then @xmath724 means rotating at the right child of caret @xmath587 , and @xmath522 means rotating at the right child of the right child of caret @xmath587 .
( 0,0)(1,0 ) ; ( 1,0)(2,0 ) ; ( 2,0)(3,0 ) ; ( 3,0)(4,0 ) ; ( 4,0)(5,0 ) ; ( 5,0)(6,0 ) ; ( 6,0)(7,0 ) ; ( 7,0)(8,0 ) ; ( 0,0 ) circle ( 2pt ) ; ( 2,0 ) circle ( 2pt ) ; ( 4,0 ) circle ( 2pt ) ; ( 6,0 ) circle ( 2pt ) ; ( 8,0 ) circle ( 2pt ) ; ( 0,2)(1,2 ) ; ( 1,2)(2,2 ) ; ( 2,2)(3,2 ) ; ( 3,2)(4,2 ) ; ( 4,2)(5,2 ) ; ( 5,2)(6,2 ) ; ( 6,2)(7,2 ) ; ( 7,2)(8,2 ) ; ( 0,2 ) circle ( 2pt ) ; ( 2,2 ) circle ( 2pt ) ; ( 4,2 ) circle ( 2pt ) ; ( 6,2 ) circle ( 2pt ) ; ( 8,2 ) circle ( 2pt ) ; ( 0,0)(0,1 ) ; ( 0,1)(0,2 ) ; ( 8,0)(8,1 ) ; ( 8,1)(8,2 ) ; ( 1,0 ) node [ anchor = north]@xmath90 ; ( 3,0 ) node [ anchor = north]@xmath90 ; ( 5,0 ) node [ anchor = north]@xmath91 ; ( 7,0 ) node [ anchor = north]@xmath90 ; ( 1,2 ) node [ anchor = south]@xmath90 ; ( 3,2 ) node [ anchor = south]@xmath90 ; ( 5,2 ) node [ anchor = south]@xmath91 ; ( 7,2 ) node [ anchor = south]@xmath90 ; ( 0,1 ) node [ anchor = east]@xmath91 ; ( 8,1 ) node [ anchor = west]@xmath91 ; ( 4,1.5 ) node [ anchor = north ] * @xmath725 * ; ( 4,0 ) node [ anchor = north ] @xmath716 ; ( 4,2.2)(4,3 ) ; ( 4,3 ) node[anchor = south]@xmath112 ; ( 0,2 ) node [ anchor = east]@xmath726 ; ( 8,2 ) node [ anchor = west]@xmath727 ; ( 10,0)(11,0 ) ; ( 11,0)(12,0 ) ; ( 12,0)(13,0 ) ; ( 13,0)(14,0 ) ; ( 14,0)(15,0 ) ; ( 15,0)(16,0 ) ; ( 16,0)(17,0 ) ; ( 17,0)(18,0 ) ; ( 10,0 ) circle ( 2pt ) ; ( 12,0 ) circle ( 2pt ) ; ( 14,0 ) circle ( 2pt ) ; ( 16,0 ) circle ( 2pt ) ; ( 18,0 ) circle ( 2pt ) ; ( 10,2)(11,2 ) ; ( 11,2)(12,2 ) ; ( 12,2)(13,2 ) ; ( 13,2)(14,2 ) ; ( 14,2)(15,2 ) ; ( 15,2)(16,2 ) ; ( 16,2)(17,2 ) ; ( 17,2)(18,2 ) ; ( 10,2 ) circle ( 2pt ) ; ( 12,2 ) circle ( 2pt ) ; ( 14,2 ) circle ( 2pt ) ; ( 16,2 ) circle ( 2pt ) ; ( 18,2 ) circle ( 2pt ) ; ( 10,0)(10,1 ) ; ( 10,1)(10,2 ) ; ( 18,0)(18,1 ) ; ( 18,1)(18,2 ) ; ( 11,0 ) node [ anchor = north]@xmath90 ; ( 13,0 ) node [ anchor = north]@xmath90 ; ( 15,0 ) node [ anchor = north]@xmath91 ; ( 17,0 ) node [ anchor = north]@xmath90 ; ( 11,2 ) node [ anchor = south]@xmath90 ; ( 13,2 ) node [ anchor = south]@xmath90 ; ( 15,2 ) node [ anchor = south]@xmath91 ; ( 17,2 ) node [ anchor = south]@xmath90 ; ( 10,1 ) node [ anchor = east]@xmath91 ; ( 18,1 ) node [ anchor = west]@xmath91 ; ( 14,1.5 ) node [ anchor = north ] * @xmath728 * ; ( 14,0 ) node [ anchor = north ] @xmath716 ; ( 14,2.2)(14,3 ) ; ( 14,3 ) node[anchor = south]@xmath112 ; ( 10,0 ) node [ anchor = east]@xmath726 ; ( 18,0 ) node [ anchor = west]@xmath727 ; ( 0,0)(1.33,0 ) ; ( 1.33,0)(2.66,0 ) ; ( 2.66,0)(4,0 ) ; ( 4,0)(5.33,0 ) ; ( 5.33,0)(6.66,0 ) ; ( 6.66,0)(8,0 ) ; ( 0,0 ) circle ( 2pt ) ; ( 2.66,0 ) circle ( 2pt ) ; ( 5.33,0 ) circle ( 2pt ) ; ( 8,0 ) circle ( 2pt ) ; ( 0,2)(.75,2 ) ; ( .75,2)(1.5,2 ) ; ( 1.5,2)(2.25,2 ) ; ( 2.25,2)(3,2 ) ; ( 3,2)(4,2 ) ; ( 3.75,2)(5,2 ) ; ( 5,2)(5.75,2 ) ; ( 5.75,2)(6.5,2 ) ; ( 6.5,2)(7.25,2 ) ; ( 7.25,2)(8,2 ) ; ( 0,2 ) circle ( 2pt ) ; ( 1.5,2 ) circle ( 2pt ) ; ( 3,2 ) circle ( 2pt ) ; ( 5,2 ) circle ( 2pt ) ; ( 6.5,2 ) circle ( 2pt ) ; ( 8,2 ) circle ( 2pt ) ; ( 0,0)(0,1 ) ; ( 0,1)(0,2 ) ; ( 8,0)(8,1 ) ; ( 8,1)(8,2 ) ; ( 1.33,0 ) node [ anchor = north]@xmath90 ; ( 4,0 ) node [ anchor = north]@xmath91 ; ( 6.66,0 ) node [ anchor = north]@xmath90 ; ( .75,2 ) node [ anchor = south]@xmath90 ; ( 2.25,2 ) node [ anchor = south]@xmath90 ; ( 4,2 ) node [ anchor = south]@xmath91 ; ( 5.75,2 ) node [ anchor = south]@xmath90 ; ( 7.25,2 ) node [ anchor = south]@xmath90 ; ( 0,1 ) node [ anchor = east]@xmath91 ; ( 8,1 ) node [ anchor = west]@xmath91 ; ( 4,1.5 ) node [ anchor = north ] * @xmath729 * ; ( 2.66,0 ) node [ anchor = north ] @xmath716 ; ( 3,2.2)(3,3 ) ; ( 3,3 ) node[anchor = south]@xmath112 ; ( 0,2 ) node [ anchor = east]@xmath726 ; ( 8,2 ) node [ anchor = west]@xmath727 ; ( 10,2)(11.33,2 ) ; ( 11.33,2)(12.66,2 ) ; ( 12.66,2)(14,2 ) ; ( 14,2)(15.33,2 ) ; ( 15.33,2)(16.66,2 ) ; ( 16.66,2)(18,2 ) ; ( 10,2 ) circle ( 2pt ) ; ( 12.66,2 ) circle ( 2pt ) ; ( 15.33,2 ) circle ( 2pt ) ; ( 18,2 ) circle ( 2pt ) ; ( 10,0)(10.75,0 ) ; ( 10.75,0)(11.5,0 ) ; ( 11.5,0)(12.25,0 ) ; ( 12.25,0)(13,0 ) ; ( 13,0)(14,0 ) ; ( 14,0)(15,0 ) ; ( 15,0)(15.75,0 ) ; ( 15.75,0)(16.5,0 ) ; ( 16.5,0)(17.25,0 ) ; ( 17.25,0)(18,0 ) ; ( 10,0 ) circle ( 2pt ) ; ( 11.5,0 ) circle ( 2pt ) ; ( 13,0 ) circle ( 2pt ) ; ( 15,0 ) circle ( 2pt ) ; ( 16.5,0 ) circle ( 2pt ) ; ( 18,0 ) circle ( 2pt ) ; ( 10,0)(10,1 ) ; ( 10,1)(10,2 ) ; ( 18,0)(18,1 ) ; ( 18,1)(18,2 ) ; ( 11.33,2 ) node [ anchor = south]@xmath90 ; ( 14,2 ) node [ anchor = south]@xmath91 ; ( 16.66,2 ) node [ anchor = south]@xmath90 ; ( 10.75,0 ) node [ anchor = north]@xmath90 ; ( 12.25,0 ) node [ anchor = north]@xmath90 ; ( 14,0 ) node [ anchor = north]@xmath91 ; ( 15.75,0 ) node [ anchor = north]@xmath90 ; ( 17.25,0 ) node [ anchor = north]@xmath90 ; ( 10,1 ) node [ anchor = east]@xmath91 ; ( 18,1 ) node [ anchor = west]@xmath91 ; ( 14,1.5 ) node [ anchor = north ] * @xmath730 * ; ( 13,0 ) node [ anchor = north ] @xmath716 ; ( 12.66,2.2)(12.66,3 ) ; ( 12.66,3 ) node[anchor = south]@xmath112 ; ( 10,0 ) node [ anchor = east]@xmath726 ; ( 18,0 ) node [ anchor = west]@xmath727 ; ( 0,0 ) circle ( 2pt ) ; ( 1,0 ) circle ( 2pt ) ; ( 2,0 ) circle ( 2pt ) ; ( 3,0 ) circle ( 2pt ) ; ( 4,0 ) circle ( 2pt ) ; ( 5,0 ) circle ( 2pt ) ; ( 6,0 ) circle ( 2pt ) ; ( 0,2 ) circle ( 2pt ) ; ( 1,2 ) circle ( 2pt ) ; ( 2,2 ) circle ( 2pt ) ; ( 3,2 ) circle ( 2pt ) ; ( 4,2 ) circle ( 2pt ) ; ( 5,2 ) circle ( 2pt ) ; ( 6,2 ) circle ( 2pt ) ; ( .5,0 ) node [ anchor = north]@xmath90 ; ( 1.5,0 ) node [ anchor = north]@xmath90 ; ( 2.5,0 ) node [ anchor = north]@xmath90 ; ( 3.5,0 ) node [ anchor = north]@xmath91 ; ( 4.5,0 ) node [ anchor = north]@xmath90 ; ( 5.5,0 ) node [ anchor = north]@xmath90 ; ( .5,2 ) node [ anchor = south]@xmath90 ; ( 1.5,2 ) node [ anchor = south]@xmath90 ; ( 2.5,2 ) node [ anchor = south]@xmath90 ; ( 3.5,2 ) node [ anchor = south]@xmath91 ; ( 4.5,2 ) node [ anchor = south]@xmath90 ; ( 5.5,2 ) node [ anchor = south]@xmath90 ; ( 0,1 ) node [ anchor = east]@xmath91 ; ( 6,1 ) node [ anchor = west]@xmath91 ; ( 3,1.5 ) node [ anchor = north ] * @xmath731 * ; ( 0,0)(.5,0 ) ; ( .5,0)(1,0 ) ; ( 1,0)(1.5,0 ) ; ( 1.5,0)(2,0 ) ; ( 2,0)(2.5,0 ) ; ( 2.5,0)(3,0 ) ; ( 3,0)(3.5,0 ) ; ( 3.5,0)(4,0 ) ; ( 4,0)(4.5,0 ) ; ( 4.5,0)(5,0 ) ; ( 5,0)(5.5,0 ) ; ( 5.5,0)(6,0 ) ; ( 0,2)(.5,2 ) ; ( .5,2)(1,2 ) ; ( 1,2)(1.5,2 ) ; ( 1.5,2)(2,2 ) ; ( 2,2)(2.5,2 ) ; ( 2.5,2)(3,2 ) ; ( 3,2)(3.5,2 ) ; ( 3.5,2)(4,2 ) ; ( 4,2)(4.5,2 ) ; ( 4.5,2)(5,2 ) ; ( 5,2)(5.5,2 ) ; ( 5.5,2)(6,2 ) ; ( 0,0)(0,1 ) ; ( 0,1)(0,2 ) ; ( 6,0)(6,1 ) ; ( 6,1)(6,2 ) ; ( 3,0 ) node [ anchor = north ] @xmath716 ; ( 3,2.2)(3,3 ) ; ( 3,3 ) node[anchor = south]@xmath112 ; ( 0,2 ) node [ anchor = east]@xmath726 ; ( 6,2 ) node [ anchor = west]@xmath727 ; ( 9,0 ) circle ( 2pt ) ; ( 10,0 ) circle ( 2pt ) ; ( 11,0 ) circle ( 2pt ) ; ( 12,0 ) circle ( 2pt ) ; ( 13,0 ) circle ( 2pt ) ; ( 14,0 ) circle ( 2pt ) ; ( 15,0 ) circle ( 2pt ) ; ( 9,2 ) circle ( 2pt ) ; ( 10,2 ) circle ( 2pt ) ; ( 11,2 ) circle ( 2pt ) ; ( 12,2 ) circle ( 2pt ) ; ( 13,2 ) circle ( 2pt ) ; ( 14,2 ) circle ( 2pt ) ; ( 15,2 ) circle ( 2pt ) ; ( 9.5,0 ) node [ anchor = north]@xmath90 ; ( 10.5,0 ) node [ anchor = north]@xmath90 ; ( 11.5,0 ) node [ anchor = north]@xmath90 ; ( 12.5,0 ) node [ anchor = north]@xmath91 ; ( 13.5,0 ) node [ anchor = north]@xmath90 ; ( 14.5,0 ) node [ anchor = north]@xmath90 ; ( 9.5,2 ) node [ anchor = south]@xmath90 ; ( 10.5,2 ) node [ anchor = south]@xmath90 ; ( 11.5,2 ) node [ anchor = south]@xmath90 ; ( 12.5,2 ) node [ anchor = south]@xmath91 ; ( 13.5,2 ) node [ anchor = south]@xmath90 ; ( 14.5,2 ) node [ anchor = south]@xmath90 ; ( 9,1 ) node [ anchor = east]@xmath91 ; ( 15,1 ) node [ anchor = west]@xmath91 ; ( 12,1.5 ) node [ anchor = north ] * @xmath732 * ; ( 9,0)(9.5,0 ) ; ( 9.5,0)(10,0 ) ; ( 10,0)(10.5,0 ) ; ( 10.5,0)(11,0 ) ; ( 11,0)(11.5,0 ) ; ( 11.5,0)(12,0 ) ; ( 12,0)(12.5,0 ) ; ( 12.5,0)(13,0 ) ; ( 13,0)(13.5,0 ) ; ( 13.5,0)(14,0 ) ; ( 14,0)(14.5,0 ) ; ( 14.5,0)(15,0 ) ; ( 9,2)(9.5,2 ) ; ( 9.5,2)(10,2 ) ; ( 10,2)(10.5,2 ) ; ( 10.5,2)(11,2 ) ; ( 11,2)(11.5,2 ) ; ( 11.5,2)(12,2 ) ; ( 12,2)(12.5,2 ) ; ( 12.5,2)(13,2 ) ; ( 13,2)(13.5,2 ) ; ( 13.5,2)(14,2 ) ; ( 14,2)(14.5,2 ) ; ( 14.5,2)(15,2 ) ; ( 9,0)(9,1 ) ; ( 9,1)(9,2 ) ; ( 15,0)(15,1 ) ; ( 15,1)(15,2 ) ; ( 12,0 ) node [ anchor = north ] @xmath716 ; ( 12,2.2)(12,3 ) ; ( 12,3 ) node[anchor = south]@xmath112 ; ( 9,0 ) node [ anchor = east]@xmath726 ; ( 15,0 ) node [ anchor = west]@xmath727 ; ( 0,4 ) circle ( 2pt ) ; ( 1.5,4 ) circle ( 2pt ) ; ( 3,4 ) circle ( 2pt ) ; ( 5,4 ) circle ( 2pt ) ; ( 6.5,4 ) circle ( 2pt ) ; ( 8,4 ) circle ( 2pt ) ; ( 0,6 ) circle ( 2pt ) ; ( 1,6 ) circle ( 2pt ) ; ( 2,6 ) circle ( 2pt ) ; ( 3,6 ) circle ( 2pt ) ; ( 5,6 ) circle ( 2pt ) ; ( 6,6 ) circle ( 2pt ) ; ( 7,6 ) circle ( 2pt ) ; ( 8,6 ) circle ( 2pt ) ; ( .75,4 ) node [ anchor = north]@xmath90 ; ( 2.25,4 ) node [ anchor = north]@xmath90 ; ( 4,4 ) node [ anchor = north]@xmath91 ; ( 5.75,4 ) node [ anchor = north]@xmath90 ; ( 7.25,4 ) node [ anchor = north]@xmath90 ; ( .5,6 ) node [ anchor = south]@xmath90 ; ( 1.5,6 ) node [ anchor = south]@xmath90 ; ( 2.5,6 ) node [ anchor = south]@xmath90 ; ( 4,6 ) node [ anchor = south]@xmath91 ; ( 5.5,6 ) node [ anchor = south]@xmath90 ; ( 6.5,6 ) node [ anchor = south]@xmath90 ; ( 7.5,6 ) node [ anchor = south]@xmath90 ; ( 0,5 ) node [ anchor = east]@xmath91 ; ( 8,5 ) node [ anchor = west]@xmath91 ; ( 4,5.5 ) node [ anchor = north ] * @xmath733 * ; ( 0,4)(.75,4 ) ; ( .75,4)(1.5,4 ) ; ( 1.5,4)(2.25,4 ) ; ( 2.25,4)(3,4 ) ; ( 3,4)(4,4 ) ; ( 4,4)(5,4 ) ; ( 5,4)(5.75,4 ) ; ( 5.75,4)(6.5,4 ) ; ( 6.5,4)(7.25,4 ) ; ( 7.25,4)(8,4 ) ; ( 0,6)(.5,6 ) ; ( .5,6)(1,6 ) ; ( 1,6)(1.5,6 ) ; ( 1.5,6)(2,6 ) ; ( 2,6)(2.5,6 ) ; ( 2.5,6)(3,6 ) ; ( 3,6)(4,6 ) ; ( 4,6)(5,6 ) ; ( 5,6)(5.5,6 ) ; ( 5.5,6)(6,6 ) ; ( 6,6)(6.5,6 ) ; ( 6.5,6)(7,6 ) ; ( 7,6)(7.5,6 ) ; ( 7.5,6)(8,6 ) ; ( 0,4)(0,5 ) ; ( 0,5)(0,6 ) ; ( 8,4)(8,5 ) ; ( 8,5)(8,6 ) ; ( 3,4 ) node [ anchor = north ] @xmath716 ; ( 3,6.2)(3,7 ) ; ( 3,7 ) node[anchor = south]@xmath112 ; ( 0,6 ) node [ anchor = east]@xmath726 ; ( 8,6 ) node [ anchor = west]@xmath727 ; ( 0,2 ) circle ( 2pt ) ; ( 1.5,2 ) circle ( 2pt ) ; ( 3,2 ) circle ( 2pt ) ; ( 5,2 ) circle ( 2pt ) ; ( 6.5,2 ) circle ( 2pt ) ; ( 8,2 ) circle ( 2pt ) ; ( 0,0 ) circle ( 2pt ) ; ( 1,0 ) circle ( 2pt ) ; ( 2,0 ) circle ( 2pt ) ; ( 3,0 ) circle ( 2pt ) ; ( 5,0 ) circle ( 2pt ) ; ( 6,0 ) circle ( 2pt ) ; ( 7,0 ) circle ( 2pt ) ; ( 8,0 ) circle ( 2pt ) ; ( .75,2 ) node [ anchor = south]@xmath90 ; ( 2.25,2 ) node [ anchor = south]@xmath90 ; ( 4,2 ) node [ anchor = south]@xmath91 ; ( 5.75,2 ) node [ anchor = south]@xmath90 ; ( 7.25,2 ) node [ anchor = south]@xmath90 ; ( .5,0 ) node [ anchor = north]@xmath90 ; ( 1.5,0 ) node [ anchor = north]@xmath90 ; ( 2.5,0 ) node [ anchor = north]@xmath90 ; ( 4,0 ) node [ anchor = north]@xmath91 ; ( 5.5,0 ) node [ anchor = north]@xmath90 ; ( 6.5,0 ) node [ anchor = north]@xmath90 ; ( 7.5,0 ) node [ anchor = north]@xmath90 ; ( 0,1 ) node [ anchor = east]@xmath91 ; ( 8,1 ) node [ anchor = west]@xmath91 ; ( 4,1.5 ) node [ anchor = north ] * @xmath734 * ; ( 0,2)(.75,2 ) ; ( .75,2)(1.5,2 ) ; ( 1.5,2)(2.25,2 ) ; ( 2.25,2)(3,2 ) ; ( 3,2)(4,2 ) ; ( 4,2)(5,2 ) ; ( 5,2)(5.75,2 ) ; ( 5.75,2)(6.5,2 ) ; ( 6.5,2)(7.25,2 ) ; ( 7.25,2)(8,2 ) ; ( 0,0)(.5,0 ) ; ( .5,0)(1,0 ) ; ( 1,0)(1.5,0 ) ; ( 1.5,0)(2,0 ) ; ( 2,0)(2.5,0 ) ; ( 2.5,0)(3,0 ) ; ( 3,0)(4,0 ) ; ( 4,0)(5,0 ) ; ( 5,0)(5.5,0 ) ; ( 5.5,0)(6,0 ) ; ( 6,0)(6.5,0 ) ; ( 6.5,0)(7,0 ) ; ( 7,0)(7.5,0 ) ; ( 7.5,0)(8,0 ) ; ( 0,0)(0,1 ) ; ( 0,1)(0,2 ) ; ( 8,0)(8,1 ) ; ( 8,1)(8,2 ) ; ( 3,0 ) node [ anchor = north ] @xmath716 ; ( 3,2.2)(3,3 ) ; ( 3,3 ) node[anchor = south]@xmath112 ; ( 0,0 ) node [ anchor = east]@xmath726 ; ( 8,0 ) node [ anchor = west]@xmath727 ; rewriting the result of corollary [ newbad ] using the quantities in notation [ not : ss ] , we have that the bad edge @xmath395 satisfies either @xmath663 or else property @xmath477 holds and either @xmath735 or @xmath555 .
it will be useful to re - organize these cases for the definition of the map @xmath107 , as follows .
let @xmath95 .
if @xmath396 is a bad edge , then @xmath551 , the tree @xmath122 has at least 3 right carets , and either 1 .
@xmath663 , 2 .
@xmath681 , @xmath736 , property @xmath477 holds , @xmath737 , and either 1 .
@xmath738 and @xmath739 , 2 .
@xmath740 and @xmath741 , or 3 .
@xmath742 and @xmath743 with @xmath744 , + or 3 .
@xmath681 , @xmath735 , property @xmath477 holds , and the conditions of case ( 2 ) are not satisfied .
the proof of this corollary follows directly from corollary [ newbad ] , using the fact that when @xmath745 then @xmath746 , and is left to the reader . using these cases
, we will choose @xmath705 to accomplish the following : * if @xmath663 , then @xmath472 is not the minimal element @xmath747 relative to @xmath624 ; in this case @xmath705 is chosen so that either @xmath748 , or @xmath749 , @xmath750 and @xmath751 ( see part ( 1 ) of the definition below ) . *
if @xmath681 , but @xmath436 is not the minimal tree relative to @xmath752 , @xmath705 is chosen ( in parts ( 2c ) and ( 3 ) ) so that either @xmath748 , or @xmath749 , @xmath753 and @xmath754 . * finally ,
if both @xmath436 and @xmath472 are minimal , then @xmath705 is chosen ( in parts ( 2a ) and ( 2b ) ) so that caret @xmath475 is removed in moving around the 2-cell from @xmath112 to @xmath716 , so @xmath755 .
[ def : collapse ] we define a map @xmath107 from the set of bad edges to the set of @xmath329-cells in several cases .
consider a bad edge @xmath395 , and let @xmath756 .
let @xmath605 be the left subtree of the root of @xmath472 , and let @xmath610 be the left subtree of the right child of the root of @xmath472 .
similarly , let @xmath757 be the right subtree of the root caret of @xmath436 , and let @xmath758 be the right subtree of the left child of the root caret of @xmath436 . 1 . if @xmath663 and : * if @xmath759 is odd , and @xmath605 is empty , then define @xmath760 . *
if @xmath759 is odd , and @xmath605 is not empty , let @xmath761 . * if @xmath759 is even , and @xmath610 is empty , let @xmath762 . * if @xmath759 is even , and @xmath610 is not empty , let @xmath763 .
if @xmath681 , @xmath736 , property @xmath477 holds , @xmath764 , and : 1 .
if @xmath765 and @xmath766 , then let @xmath767 .
if @xmath768 and @xmath769 , then let @xmath770 .
3 . if @xmath771 and @xmath772 for @xmath773 , let @xmath770 .
if @xmath681 , @xmath774 , property @xmath477 holds , and the conditions of case ( 2 ) are not satisfied , and : * if @xmath775 is odd , and @xmath757 is empty , then let @xmath770 . *
if @xmath775 is odd , and @xmath757 is not empty , let @xmath776 . * if @xmath775 is even , and @xmath758 is empty , let @xmath767 . * if @xmath775 is even , and @xmath758 is not empty , let @xmath777 .
see figures [ fig : cell1 ] and [ fig : cell2 ] for examples of bad edges and their corresponding two cells .
figures [ fig : trees1 ] and [ fig : trees2 ] show the tree pair diagrams corresponding to the elements @xmath112 and @xmath716 , where @xmath715 is the edge across the two - cell from the bad edge @xmath395 .
( 0,0)(1,0 ) ; ( 1,0)(2,0 ) ; ( 2,0)(3,0 ) ; ( 3,0)(4,0 ) ; ( 4,0)(5,0 ) ; ( 5,0)(6,0 ) ; ( 6,0)(7,0 ) ; ( 7,0)(8,0 ) ; ( 0,0 ) circle ( 2pt ) ; ( 2,0 ) circle ( 2pt ) ; ( 4,0 ) circle ( 2pt ) ; ( 6,0 ) circle ( 2pt ) ; ( 8,0 ) circle ( 2pt ) ; ( 0,2)(1,2 ) ; ( 1,2)(2,2 ) ; ( 2,2)(3,2 ) ; ( 3,2)(4,2 ) ; ( 4,2)(5,2 ) ; ( 5,2)(6,2 ) ; ( 6,2)(7,2 ) ; ( 7,2)(8,2 ) ; ( 0,2 ) circle ( 2pt ) ; ( 2,2 ) circle ( 2pt ) ; ( 4,2 ) circle ( 2pt ) ; ( 6,2 ) circle ( 2pt ) ; ( 8,2 ) circle ( 2pt ) ; ( 0,0)(0,1 ) ; ( 0,1)(0,2 ) ; ( 8,0)(8,1 ) ; ( 8,1)(8,2 ) ; ( 1,0 ) node [ anchor = north]@xmath90 ; ( 3,0 ) node [ anchor = north]@xmath90 ; ( 5,0 ) node [ anchor = north]@xmath91 ; ( 7,0 ) node [ anchor = north]@xmath90 ; ( 1,2 ) node [ anchor = south]@xmath90 ; ( 3,2 ) node [ anchor = south]@xmath90 ; ( 5,2 ) node [ anchor = south]@xmath91 ; ( 7,2 ) node [ anchor = south]@xmath90 ; ( 0,1 ) node [ anchor = east]@xmath91 ; ( 8,1 ) node [ anchor = west]@xmath91 ; ( 4,1.5 ) node [ anchor = north ] * @xmath778 * ; ( 4,0 ) node [ anchor = north ] @xmath716 ; ( 4,2.2)(4,3 ) ; ( 4,3 ) node[anchor = south]@xmath779 ; ( 0,0 ) node [ anchor = east ] @xmath726 ; ( 8,0 ) node [ anchor = west ] @xmath727 ; = [ sibling distance=10 mm ] child child child child child child child child child child childchild childchild ; = [ sibling distance=15 mm ] = [ sibling distance=15 mm ] = [ sibling distance=15 mm ] = [ sibling distance=15 mm ] = [ sibling distance=10 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child child child child child ; ( 0,0)(1,0 ) ; ( 1,0)(2,0 ) ; ( 2,0)(3,0 ) ; ( 3,0)(4,0 ) ; ( 4,0)(5,0 ) ; ( 5,0)(6,0 ) ; ( 6,0)(7,0 ) ; ( 7,0)(8,0 ) ; ( 8,0)(9,0 ) ; ( 9,0)(10,0 ) ; ( 10,0)(11,0 ) ; ( 11,0)(12,0 ) ; ( 12,0)(13,0 ) ; ( 13,0)(14,0 ) ; ( 0,0 ) circle ( 2pt ) ; ( 2,0 ) circle ( 2pt ) ; ( 4,0 ) circle ( 2pt ) ; ( 6,0 ) circle ( 2pt ) ; ( 8,0 ) circle ( 2pt ) ; ( 10,0 ) circle ( 2pt ) ; ( 12,0 ) circle ( 2pt ) ; ( 14,0 ) circle ( 2pt ) ; ( 1,0 ) node [ anchor = north]@xmath90 ; ( 3,0 ) node [ anchor = north]@xmath90 ; ( 5,0 ) node [ anchor = north]@xmath90 ; ( 7,0 ) node [ anchor = north]@xmath91 ; ( 9,0 ) node [ anchor = north]@xmath90 ; ( 11,0 ) node [ anchor = north]@xmath90 ; ( 13,0 ) node [ anchor = north]@xmath90 ; ( 0,4)(1.4,4 ) ; ( 1.4,4)(2.8,4 ) ; ( 2.8,4)(4.2,4 ) ; ( 4.2,4)(5.6,4 ) ; ( 5.6,4)(7,4 ) ; ( 7,4)(8.4,4 ) ; ( 8.4,4)(9.8,4 ) ; ( 9.8,4)(11.2,4 ) ; ( 11.2,4)(12.6,4 ) ; ( 12.6,4)(14,4 ) ; ( 0,4 ) circle ( 2pt ) ; ( 2.8,4 ) circle ( 2pt ) ; ( 5.6,4 ) circle ( 2pt ) ; ( 8.4,4 ) circle ( 2pt ) ; ( 11.2,4 ) circle ( 2pt ) ; ( 14,4 ) circle ( 2pt ) ; ( 1.4,4 ) node [ anchor = south]@xmath90 ; ( 4.2,4 ) node [ anchor = south]@xmath90 ; ( 7,4 ) node [ anchor = south]@xmath91 ; ( 9.8,4 ) node [ anchor = south]@xmath90 ; ( 12.6,4 ) node [ anchor = south]@xmath90 ; ( 0,0)(0,2 ) ; ( 0,2)(0,4 ) ; ( 14,0)(14,2 ) ; ( 14,2)(14,4 ) ; ( 0,2 ) node [ anchor = east]@xmath91 ; ( 14,2 ) node [ anchor = west]@xmath91 ; ( 7,2 ) node * @xmath780 * ; ( 5.6,4.15)(5.6,5 ) ; ( 5.6,5 ) node[anchor = south]@xmath781 ; ( 7,3.6 ) node @xmath395 ; ( 2,0)(2,-1.5 ) ; ( 2,-1.5)(2,-3 ) ; ( 12,0)(12,-1.5 ) ; ( 12,-1.5)(12,-3 ) ; ( 2,-1.5 ) node [ anchor = east ] @xmath91 ; ( 12,-1.5 ) node [ anchor = west ] @xmath91 ; ( 2,-3)(4,-3 ) ; ( 4,-3)(6,-3 ) ; ( 6,-3)(7,-3 ) ; ( 7,-3)(8,-3 ) ; ( 8,-3)(10,-3 ) ; ( 10,-3)(12,-3 ) ; ( 2,-3 ) circle ( 2pt ) ; ( 6,-3 ) circle ( 2pt ) ; ( 8,-3 ) circle ( 2pt ) ; ( 12,-3 ) circle ( 2pt ) ; ( 6,-3 ) node [ anchor = north]@xmath782 ; ( 4,-3 ) node [ anchor = south]@xmath90 ; ( 10,-3 ) node [ anchor = south]@xmath90 ; ( 7,-3 ) node [ anchor = south]@xmath91 ; ( 7,-2 ) node * @xmath783 * ; ( 6,-.15)(6,-.9 ) ; ( 6,-1 ) node @xmath784 ; ( 7,.1 ) node [ anchor = south]@xmath565 ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child child child ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child child child ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child child child ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child child child ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child ; = [ sibling distance=25 mm ] = [ sibling distance=10 mm ] = [ sibling distance=10 mm ] child child child child child child child child child child ; in the following theorem , we verify that the map defined above and the partial order on the set of bad edges satisfy the hypothesis of theorem [ combingextends ] .
in addition , we prove another fact which will be used later in showing that the combing satisfies a linear radial tameness function .
[ thm : boundaryorder ] if @xmath395 is a bad edge , then all other vertices @xmath785 on the boundary of @xmath705 have @xmath786 . furthermore , every edge of the form @xmath659 along the boundary is either a good edge , or precedes @xmath395 in the ordering of the bad edges .
let @xmath715 be the bottom @xmath712 edge in the @xmath329-cell @xmath705 , and @xmath713 ( respectively @xmath714 ) be the left ( respectively right ) side @xmath712 edges .
the first statement in the theorem is a consequence of the following observation .
the tree @xmath122 has enough carets in the left subtree of the root caret , and in both subtrees of the right child of the root caret to ensure that as we read around @xmath705 to the left , starting from @xmath112 , terminating at @xmath716 , and form the successive products , no carets ever need to be added to the tree pair diagrams in order to perform these multiplications .
the same holds for the path from @xmath429 , around to the right ending at @xmath787 . since @xmath788
, it follows that for each vertex @xmath785 of @xmath705 , @xmath786 .
in addition , if @xmath675 , then @xmath465 . to prove the second statement of the theorem , we proceed by cases according to the size of @xmath759 . in each case
we show that @xmath789 , or else @xmath659 is a good edge .
we consider separately the three subcases of @xmath659 for @xmath790 .
1 . case 1 : @xmath791 . in this case
@xmath792 for @xmath793 .
also , note that @xmath794 .
1 . @xmath795 . in this case
either : * @xmath796 ( and @xmath797 by ( 1 ) of definition [ def : order ] if @xmath713 is a bad edge ) , or * @xmath798 , @xmath799 , @xmath800 and @xmath801 ( and @xmath797 by ( 2a ) of definition [ def : order ] if @xmath713 is a bad edge ) , or * @xmath798 and @xmath677 .
but one checks that if @xmath677 , then @xmath802 , @xmath803 , and @xmath804 .
but since no carets are ever added in moving from @xmath805 to @xmath726 , @xmath713 is a good edge .
2 . @xmath806 . if it is not the case that @xmath807 , then it is easily checked through the definition of @xmath808 that @xmath809 and @xmath810 differ only in the configuration of the carets in the left subtree of the root . therefore , the argument for @xmath713 goes through exactly , replacing @xmath726 by @xmath727 .
3 . @xmath811 . in this case
either : * @xmath748 ( and @xmath812 by ( 1 ) of definition [ def : order ] if @xmath715 is a bad edge ) , or * @xmath749 and @xmath813 , in which case @xmath750 , and @xmath814 .
then @xmath815 , which implies that @xmath816 , ( and @xmath812 by ( 2b ) of definition [ def : order ] if @xmath715 is a bad edge ) , or * @xmath749 and @xmath817
. however , this can happen only when @xmath804 for @xmath803 , @xmath818 , and @xmath819 ( and @xmath812 by ( 4a ) of definition [ def : order ] if @xmath715 is a bad edge ) .
2 . case 2 : @xmath681 . in this case ,
@xmath820 for @xmath821 .
also , note that @xmath822 .
1 . @xmath795 . in this case , @xmath823 .
now either : * @xmath796 ( and @xmath797 by ( 1 ) of definition [ def : order ] if @xmath713 is a bad edge ) , or * @xmath798 and @xmath824 , and hence @xmath668 ( and @xmath797 by ( 3a ) of definition [ def : order ] if @xmath713 is a bad edge ) , or * @xmath798 and @xmath799 .
however , this only occurs if @xmath825 , and then @xmath826 and @xmath801 ( and @xmath797 by ( 4b ) of definition [ def : order ] if @xmath713 is a bad edge ) .
2 . @xmath806 .
if @xmath714 is a bad edge , then @xmath827 implies that property @xmath477 holds . in this case , @xmath828 because a caret is removed when moving from @xmath112 to @xmath429 .
3 . @xmath811 .
then either : * in cases ( 2a ) and ( 2b ) of definition [ def : collapse ] , @xmath748 , since caret @xmath475 is removed in moving from @xmath726 to @xmath805 ( and @xmath812 by ( 1 ) of definition [ def : order ] if @xmath715 is a bad edge ) . * in cases ( 2c ) and ( 3 ) of definition [ def : collapse ] , either @xmath748 ( and @xmath812 by ( 1 ) of definition [ def : order ] if @xmath715 is a bad edge ) , or @xmath749 , @xmath817 , @xmath753 , @xmath829 , and @xmath830 ( and @xmath812 by ( 3b ) of definition [ def : order ] if @xmath715 is a bad edge ) .
since all hypotheses of theorem [ combingextends ] have now been verified , theorem [ combingextends ] shows that the nested traversal 0-combing @xmath71 extends to a 1-combing @xmath831 { \rightarrow}x$ ] .
the fact that our combing @xmath71 satisfies a linear radial tameness function will follow from the fact that the number of carets in the tree pair diagrams representing the vertices along a nested traversal normal form path never decreases , and from the close relationship between word length over the alphabet @xmath214 and the number of carets .
first , we extend the concept of the number of carets in a tree pair diagram from @xmath832 to all of @xmath10 . for any @xmath833 , we define @xmath834 and @xmath835 by cases . 1 . if @xmath836 , then @xmath837 , and we let @xmath838 , the number of carets in either tree of a reduced tree pair diagram for @xmath37 .
2 . if @xmath41 , then @xmath42 is on the interior of some edge , with vertices @xmath43
. then define @xmath839 , and @xmath840 .
3 . if @xmath45 , then @xmath42 is in the interior of some 2-cell , with vertices @xmath841 along the boundary .
then we define @xmath842 , and @xmath843 .
the following lemma proves that using this expanded notion of the number of carets of @xmath833 , the number of carets does not decrease along the combing paths defined by @xmath71 .
[ lem : maxmax ] for any @xmath62 and @xmath844 , we have @xmath845 , where @xmath71 is the @xmath54-combing defined in section [ sec : combing ] . in the case where @xmath74 , from theorem [ nincreases ] we know that along the nested traversal normal form @xmath846 , we have @xmath847 . for @xmath848 ,
if @xmath42 is in the interior of a good edge the conclusion of this lemma follows from the previous sentence .
if @xmath42 is in the interior of a bad edge @xmath391 , then the inequality follows from noetherian induction and the fact that for @xmath145 on any bad edge @xmath391 and @xmath785 on the complement of the edge @xmath391 in the closure of the 2-cell @xmath403 , we have @xmath849 as shown in theorem [ thm : boundaryorder ] .
the next lemma relates the level of @xmath833 to the quantities @xmath835 and @xmath834 .
recall that when @xmath74 , the level of @xmath42 and @xmath850 , the word length of @xmath42 with respect to @xmath4 , are identical .
the lengths of the two relators in this presentation are 10 and 14 , so the constant @xmath107 used in defining the level of a point in the interior of a 2-cell of the cayley complex for this presentation of @xmath0 will be @xmath851 .
[ lem : nlev ] for any @xmath833 we have @xmath852 and additionally @xmath853 .
when @xmath74 , lemma [ lengthisn ] gives @xmath854 .
therefore @xmath855 for @xmath74 .
if , on the other hand , @xmath848 , then @xmath42 is in the interior of an edge , whose endpoints are @xmath856 .
now @xmath857 @xmath858 but , on the other hand , @xmath859 in summary , in this case , we have @xmath860 and finally , if @xmath45 , @xmath42 is in the interior of some 2-cell , with @xmath861 on the boundary , then @xmath862 @xmath863 on the other hand , @xmath864 @xmath865 and so , in this case , we have @xmath866 this establishes the first statement of the lemma . now if @xmath867 , @xmath868 . for @xmath848 , @xmath869 , since one either needs to add at most 1 caret ( or can cancel at most one caret ) when multiplying by @xmath870 , and one needs to add at most two carets ( or can cancel at most two carets ) when multiplying by @xmath871 . now
the relators in our presentation of @xmath0 have length either 10 or 14 , and two vertices @xmath872 and @xmath112 on the boundary of a relator can be at most seven edges apart .
furthermore , examining the relators , we see that at most two of these seven edges correspond to multiplication by @xmath871 . therefore , for @xmath873 , @xmath874 .
we are now able to prove that the combing @xmath71 defined in section [ sec : combing ] satisfies a linear radial tameness function .
[ thm : flintame ] thompson s group @xmath0 is in @xmath59 with @xmath72 linear . more specifically , the cayley complex of the presentation @xmath875,[x_0x_1^{-1},x_0^{-2}x_1x_0 ^ 2 ] \rangle$ ] has a 1-combing admitting a radial tameness function of @xmath876 .
let @xmath61 \rightarrow x$ ] be the 1-combing of @xmath0 constructed in the previous section .
suppose that @xmath877 , @xmath878 , and @xmath879 .
in lemma [ lem : nlev ] we have shown that @xmath880 , which implies that @xmath881 , or @xmath882 . from lemma [ lem : maxmax ] we have @xmath883 , and so @xmath884 .
the last statement in lemma [ lem : nlev ] also shows that @xmath885 , and so @xmath886 . using lemma [ lem : nlev ]
once more , we obtain @xmath887 .
in this section we prove the following . [ thm : bsp ] for every natural number @xmath2 , the group @xmath888 is in the class @xmath59 with @xmath72 linear . moreover , the cayley complex for the presentation @xmath889 has a 1-combing admitting a radial tameness function of @xmath890({h}+ 4)$ ] where @xmath891 . throughout this section ,
let @xmath2 , @xmath892 , @xmath891 , and @xmath893 .
let @xmath10 denote the cayley 2-complex associated with the presentation @xmath894 of @xmath3 . give the cayley graph @xmath895 the path metric .
for a word @xmath872 over the alphabet @xmath4 , let @xmath896 denote the length of the word @xmath872 , and let @xmath897 denote the length of a geodesic ( with respect to the path metric on @xmath898 ) representative of the element of @xmath3 represented by @xmath872 .
each element of @xmath3 has a particularly simple , not necessarily geodesic , normal form @xmath899 with @xmath900 and @xmath901 .
the combing we construct will be based on this set of normal forms .
the following lemma , which characterizes a set of geodesics for elements of @xmath3 and relates them to the normal forms above , is a direct consequence of elder and hermiller .
[ lem : bsgeod ] let @xmath37 be an arbitrary element of @xmath3
. then @xmath37 is represented by a geodesic word @xmath112 satisfying one of the following .
1 . @xmath902 with @xmath903 , @xmath904 , @xmath905 , @xmath906 for @xmath907 , @xmath908 , and either @xmath909 or ( @xmath910 and @xmath380 ) .
2 . @xmath911 with @xmath912 , @xmath905 , @xmath906 for @xmath907 , @xmath913 , and either @xmath909 or @xmath914 .
moreover , in each case , @xmath37 also has a ( not necessarily geodesic ) representative of the form @xmath899 with @xmath915 , @xmath916 , and @xmath917 .
the next two lemmas show that lower and upper bounds on the length of a geodesic representative of @xmath899 in the cayley graph imply lower and upper bounds , respectively , on the value of @xmath918 .
[ lem : levimpliesh ] if @xmath919 , @xmath920 , @xmath921 , and @xmath922 , then @xmath923 .
we will prove the contrapositive ; suppose that @xmath919 , @xmath920 , @xmath921 , and @xmath924 .
let @xmath112 be a word in one of the forms ( 1)-(2 ) from lemma [ lem : bsgeod ] that is a geodesic representative of the element of @xmath3 that is also represented by @xmath925 . as @xmath112 is a geodesic
, it follows that @xmath926 is simply the length @xmath927 of the word @xmath112 . first note that if @xmath928 , then @xmath929 and either @xmath380 or @xmath930 . in both of these instances
, we have @xmath931 .
for the rest of the proof we suppose that @xmath932 . in both cases ( 1)-(2 ) , we have @xmath933 , and hence @xmath934 . since each @xmath906 for @xmath935 , then @xmath936 , where the last inequality uses the hypothesis that @xmath2 . plugging this into the previous inequality , and using the fact that @xmath932 , gives @xmath937 .
then @xmath938 , and so @xmath939 .
since @xmath940 , this gives @xmath941 .
if @xmath112 is of the form in ( 1 ) with @xmath942 , then @xmath943 if @xmath112 is of the form in ( 1 ) with @xmath944 , then @xmath945 for @xmath112 of the form ( 2 ) with @xmath946 , we have @xmath947 and finally , for @xmath112 in form ( 2 ) with @xmath948 , we have @xmath949 hence , in all possible cases , @xmath950 , and so this nonnegative integer satisfies @xmath951 .
[ lem : himplieslev ] if @xmath919 , @xmath920 , @xmath952 , and @xmath953 , then @xmath954 .
suppose that @xmath919 , @xmath920 , @xmath952 , and @xmath955 .
let @xmath112 be a word in one of the forms ( 1)-(2 ) from lemma [ lem : bsgeod ] that is a geodesic representative of the element of @xmath3 also represented by @xmath925 . in both cases
, we have @xmath956 , and in particular we must have @xmath932 . using the fact that @xmath957 ( see the proof of lemma [ lem : levimpliesh ] ) and the inequality @xmath908 yields @xmath958 , and so @xmath959
. then @xmath960 .
note that the inequality @xmath961 implies that @xmath962 .
we again consider the length of @xmath112 in each case . if @xmath112 is of the form in ( 1 ) , then @xmath963 for @xmath112 of the form ( 2 ) with @xmath946 , we have @xmath964 and finally , for @xmath112 in form ( 2 ) with @xmath948 ,
we have @xmath965 and hence @xmath966 thus in all possible cases we have @xmath967 . the cayley complex @xmath10 can be constructed using rectangles homeomorphic to @xmath968 \times [ 0,1]$ ] , with the top labeled @xmath24 and oriented to the right , the bottom labeled @xmath969 and also oriented to the right , and the left and right sides labeled @xmath970 and oriented upward .
gluing these rectangles along commonly labeled and oriented sides , the cayley complex @xmath10 is homeomorphic to the product @xmath971 of the real line with a tree @xmath106 .
the projection maps @xmath972 and @xmath973 are continuous , and we can write a point @xmath833 uniquely as @xmath974 $ ] . the vertices of @xmath106 are the projections via @xmath975 of the vertices of @xmath10 .
two vertices of @xmath10 project to the same vertex of @xmath106 if and only if there is a path in @xmath36 labeled by a power of @xmath24 between the two vertices .
each edge of @xmath106 can be considered as oriented upward with a label @xmath970 , the projection under @xmath975 of edges labeled by @xmath970 in the cayley complex .
each vertex of @xmath106 is the initial vertex for @xmath976 edges and the terminal vertex for one edge .
the projection @xmath977 maps a vertex @xmath899 to the real number @xmath978 .
the points on a vertical edge between vertices @xmath899 and @xmath979 also all map under @xmath977 to @xmath978 , and the projection @xmath977 maps the horizontal edge from @xmath899 to @xmath980 homeomorphically to the interval from @xmath978 to @xmath981 . on a rectangular ( @xmath968 \times [ 0,1]$ ] ) 2-cell ,
the top left and top right vertices have the form @xmath982 $ ] and @xmath983 $ ] , respectively .
two points @xmath984 $ ] and @xmath985 $ ] on the left and right sides of this 2-cell , respectively , determine a horizontal line segment if @xmath986 is a point on the unique edge in the tree @xmath106 oriented toward the vertex @xmath785 .
the projection @xmath987 maps this horizontal line segment homeomorphically to the interval from @xmath988 to @xmath989 in @xmath990 , and the projection @xmath975 is constant on this segment .
let @xmath991 be the initial vertex of the edge in @xmath106 whose terminus is @xmath785 , and let @xmath29 be any real number in the interval from @xmath988 to @xmath989 .
the two points @xmath992 $ ] and @xmath993 $ ] are on the bottom and top sides of this 2-cell , respectively , and they determine a vertical line segment in the 2-cell which maps via @xmath977 constantly to @xmath29 , and which maps via @xmath975 homeomorphically to the edge from @xmath991 to @xmath785 .
it will frequently be useful to move from points in the interiors of 1-cells or 2-cells to vertices in the cayley complex .
if @xmath145 is a vertex in @xmath10 , let @xmath994 .
if @xmath145 is in the interior of a 1-cell in @xmath10 labeled @xmath970 directed upward , then let @xmath995 be the initial vertex of that edge . if @xmath145 is in the interior of a 1-cell in @xmath10 labeled @xmath24 directed right , then let @xmath996 be the endpoint of that edge whose image under @xmath977 has the maximum absolute value .
finally if @xmath145 is in the interior of a 2-cell , let @xmath996 be the bottom ( left or right ) corner of that rectangular 2-cell whose image under @xmath977 has the maximum absolute value .
in a 2-cell there are @xmath997 vertices , so the difference in levels of vertices in that cell is at most h+ 2 , resulting in a bound on the difference between the level of the 2-cell and the level of any vertex in that cell , as well .
then in all cases , we have @xmath998 and @xmath999 .
we will call @xmath996 the _ vertex associated to _ @xmath145 . more information on cayley complexes for baumslag - solitar groups can be found in or .
next , we apply the lemmas above to prove theorem [ thm : bsp ] .
we first define a 1-combing @xmath61 { \rightarrow}x$ ] as follows .
let @xmath42 be an arbitrary point in @xmath36 .
since @xmath106 is a tree , there is a unique geodesic in @xmath106 from @xmath1000 to @xmath1001 .
this geodesic first follows a ( possibly empty ) edge path in the direction opposite to each edge orientation from @xmath1000 down to a point @xmath1002 ( which we will call the _ nadir _ of @xmath42 ) , and then follows a ( possibly empty ) edge path in the same direction as each edge orientation up to @xmath1001 . if @xmath1003 so that the upward portion is nonconstant , then the nadir @xmath1002 must be a vertex of @xmath106 . let the path @xmath1004 { \rightarrow}t$ ] follow the geodesic from @xmath1000 to @xmath1002 with constant speed , and let the path @xmath1005 { \rightarrow}t$ ] follow the geodesic from @xmath1002 to @xmath1001 with constant speed ( with respect to the path metric on @xmath106 ) .
define the path
@xmath1006 { \rightarrow}x$ ] by @xmath1007 $ ] for @xmath1008 $ ] , @xmath1009 $ ] for @xmath1010 $ ] , and @xmath1011 $ ] for @xmath1012 $ ] . in the first third of the interval
this path goes directly downward , in the second third it travels horizontally , and in the last third it goes directly upward in the cayley complex @xmath10 .
note that some of these three component paths may be constant .
we will refer to this path as the _ dhu - path _ for @xmath42 .
continuity of the function @xmath61 { \rightarrow}x$ ] defined by these dhu - paths follows from the continuity of the two projection functions @xmath977 and @xmath975 . for a vertex @xmath867 regarded as an element of @xmath3 , the representative @xmath899 of @xmath42 from lemma [ lem : bsgeod ] satisfies @xmath1013 , and so the path @xmath1006 { \rightarrow}x^1 $ ] follows the edge path labeled by the word @xmath899 and remains in the 1-skeleton of @xmath10 .
hence @xmath71 is a 1-combing , which we will refer to as the _ dhu - combing_. in order to show that the dhu - combing satisfies a linear radial tameness function , we will show that for the constants @xmath1014 and @xmath1015 , whenever @xmath62 , @xmath1016 , @xmath1017 , @xmath1018 , and @xmath1019 , we have a contradiction . to that end ,
fix a point @xmath42 in @xmath36 , @xmath1016 , and @xmath1017 .
let @xmath1020 , @xmath1021 , @xmath1022 , and @xmath1023 , and assume that @xmath1024 and @xmath1025 .
case i. _ suppose that @xmath1026)$ ] .
_ then both @xmath872 and @xmath112 are points on the downward portion of the dhu - path for @xmath42 , on the infinite ray labeled @xmath1027 going down from @xmath11 in @xmath10 .
now @xmath1028 is a geodesic in the cayley graph for all @xmath915 , and so traveling along a downward path , the level is a nondecreasing function .
then we have @xmath1029 .
hence we obtain the required contradiction in this case .
_ suppose that @xmath1030 ) \setminus \{\sigma\}$ ] .
_ in this case the dhu - combing path for @xmath42 has a nontrivial horizontal component and @xmath112 is in its image .
if @xmath1031 , then let @xmath1032 and @xmath1033 ; otherwise @xmath1034 and we let @xmath1035 and @xmath1036 . again applying the fact that the level is a nondecreasing function on a downward path from the identity @xmath11 , we also have that @xmath1037 with @xmath1038 and @xmath1039 .
now the points @xmath1040 and @xmath112 are both on the horizontal portion of the dhu - path for @xmath42 , satisfying @xmath1041 and @xmath1042 .
the geodesic in @xmath106 from @xmath1000 to @xmath1001 may not have an upward component in case ii , and so the nadir @xmath1002 may not be a vertex of @xmath106 .
this implies that the points @xmath1040 and @xmath112 may not be in @xmath36 .
let @xmath1043 and @xmath1044 be the vertices associated to @xmath1040 and @xmath112 , respectively . since @xmath1045 is on the @xmath1027 ray in @xmath106 , the associated vertices project to a vertex @xmath1046 for some integer @xmath915 on this ray .
the construction of the associated vertices implies that @xmath1047 is represented by a word @xmath1048 and @xmath1044 is represented by a word @xmath1049 with @xmath940 and @xmath1050 , so @xmath1051 .
define @xmath1052 . then @xmath1053
and as a consequence @xmath1054 as well .
similarly since @xmath1055 , we have @xmath1056 now @xmath1057 , and so we may apply lemma [ lem : levimpliesh ] to @xmath1048 , yielding the inequality @xmath1058 .
combining the inequalities at the ends of the last two paragraphs together with the value @xmath1059 gives @xmath1060 .
lemma [ lem : himplieslev ] applied to the word @xmath1049 with @xmath1061 says that @xmath1062 . however , from the previous paragraph we have @xmath1063 , a contradiction .
_ suppose that @xmath1064 ) \setminus \{\tau\}$ ] .
_ in this case the dhu - combing path for @xmath42 has a nontrivial upward component , and @xmath112 is in its image . as in case ii , define @xmath1032 and @xmath1033 if @xmath1065 , and define @xmath1066 and @xmath1067 if @xmath1068 .
then @xmath1037 with @xmath1038 , @xmath1039 , and @xmath1040 is either on the horizontal or the upward portion of the dhu - path for @xmath42 .
the geodesic in @xmath106 from @xmath1069 to @xmath1001 must have an upward component in case iii , and hence the nadir @xmath1002 is a vertex of @xmath106 .
then @xmath1013 for some integer @xmath1070 . the dhu - path for @xmath42 travels from @xmath1037 to @xmath1071 either via a nontrivial upward path , or else through a horizontal and then nonconstant upward path .
the dhu - paths for @xmath1040 and @xmath112 are reparameterizations of the portion of the dhu - path for @xmath42 traveling from @xmath11 to each endpoint , and so they have the same nadir @xmath1072 .
moreover , we have @xmath1073 , and there is an upward path in @xmath106 from @xmath1074 to @xmath1075 .
although the horizontal portion of the dhu - path for @xmath42 must stay in the 1-skeleton of @xmath10 ( since it projects to @xmath1076 ) , the upward portion of the dhu - path for @xmath42 may leave @xmath36 , and so @xmath1040 and @xmath112 may not be in @xmath36 .
let @xmath1047 and @xmath1044 be the vertices associated to @xmath1040 and @xmath112 , respectively .
it follows from the definition of associated vertices that these vertices satisfy @xmath1077 , @xmath1078 , and there is there is a ( possibly empty ) upward path in @xmath106 from @xmath1079 to @xmath1080 .
using lemma [ lem : bsgeod ] , the vertex @xmath1044 is represented by a word @xmath899 and the vertex @xmath1047 is represented by a word @xmath1081 .
the relations between these associated vertices above imply that @xmath1082 and @xmath1083 .
the definition of associated vertices implies that @xmath1084 , and hence @xmath1085 as in case ii . as a consequence we have both @xmath1054 and @xmath1086 as well .
also , as in case ii , the inequality @xmath1055 implies that @xmath1087 . combining inequalities from above
, we also have @xmath1088 .
the rest of the proof in this case is similar to that in case ii . in particular ,
lemma [ lem : levimpliesh ] applied to @xmath1089 yields the inequality @xmath1090 . combining this with the inequality @xmath1091 from above yields @xmath1060 . in turn
, using lemma [ lem : himplieslev ] with the word @xmath899 and @xmath1061 shows that @xmath1092 , contradicting the inequality @xmath1093 found above .
having achieved a contradiction in each case , this shows that the dhu - combing for the group @xmath888 and generating set @xmath1094 satisfies a radial tameness function @xmath1095 for the linear function @xmath1096 with the constants @xmath1097 and @xmath1098 .
in this section we show that the linear coefficient for a linear tame combing can be bounded away from 1 for a specific generating set .
[ thm : bscoeff1 ] for every natural number @xmath13 , the group @xmath1099 with the generating set @xmath1100 does not admit a 1-combing with radial tameness function of the form @xmath1101 for any constant @xmath1102 .
let @xmath13 and let @xmath10 be the cayley complex of the presentation @xmath894 , described in section [ sec : bslinear ] .
suppose to the contrary that @xmath61 { \rightarrow}x$ ] is a 1-combing with radial tameness function @xmath1101 . replacing @xmath1102 by any larger constant results in another radial tameness function satisfied by the 1-combing @xmath71 ,
so we may assume that @xmath1102 is a natural number larger than four .
consider the word @xmath1103 . since @xmath1104 in the group ,
this word labels a loop in the cayley graph @xmath36 based at the vertex corresponding to the identity @xmath11 of @xmath3 .
let @xmath1105 be the subcomplex of points on the vertices and edges along this loop .
the restriction @xmath1106 \rightarrow x$ ] of the 1-combing then defines a homotopy from the identity vertex to the loop @xmath1107 , and so the image @xmath1108)$ ] is the image of a disk filling in the loop @xmath1107 .
since the cayley complex @xmath10 is the product of the real line @xmath990 with the tree @xmath106 ( described in section [ sec : bslinear ] ) , this complex is aspherical .
then the image set @xmath1108)$ ] must include all of the points in the rectangle of points @xmath1109 with projections @xmath1110 and @xmath1111 on the geodesic in @xmath106 from @xmath1069 to @xmath1112 ; that is , the rectangle in @xmath10 bounded by the loop labeled @xmath1113 based at @xmath54 , including this boundary loop .
( the image @xmath1108)$ ] must also contain all of the points in the rectangle of @xmath10 bounded by the loop @xmath1113 based at @xmath24 . ) in particular , the vertex corresponding to the element @xmath1114 represented by the word @xmath1115 , where @xmath891 as before , is in @xmath1108)$ ] .
we obtain two estimates for @xmath1116 which , taken together , contradict our assumption that @xmath1117 .
first , we observe that the points in the set @xmath1107 all lie on the ( geodesic ) paths @xmath1118 or @xmath1119 starting at @xmath54 , and so the levels of all of the points in @xmath1107 are at most @xmath1120
. then the level of every point in the image set @xmath1108)$ ] must be at most @xmath1121 . hence @xmath1122 . on the other hand , note that @xmath1123 .
thus the element @xmath1124 of @xmath892 is also represented by the word @xmath1125 .
we claim that the word @xmath872 is a geodesic .
first note that since @xmath37 is a nontrivial power of the generator @xmath24 , we have @xmath1126 in lemma [ lem : bsgeod ] , and the geodesic word @xmath112 representing @xmath37 provided by the lemma is in the form ( 2 ) , @xmath1127 with @xmath1128 , @xmath906 for @xmath1129 , and @xmath1130 .
we will show that in fact the words @xmath872 and @xmath112 are the same .
so far we have @xmath1131 ; that is , @xmath1132 , and hence @xmath1133 .
if @xmath1134 , then @xmath1135 since @xmath909 , this shows that we must have @xmath1136 .
if @xmath1137 , then @xmath1138 again resulting in a contradiction . hence @xmath1139 .
subtracting once more , we get @xmath1140 and so @xmath1141 . using induction , then @xmath1142 for all @xmath1143 .
hence @xmath112 and @xmath872 are the same word , and the word @xmath872 is a geodesic .
this gives us another way to compute the word length over @xmath4 of @xmath37 , since @xmath872 is a geodesic representative of @xmath37 , and so @xmath1144 .
earlier in this proof we had @xmath1145 , which gives @xmath1146 .
the hypothesis that @xmath13 gives @xmath1147 , and so we have @xmath1148 , contradicting our choice of @xmath1102 . james belk and kai - uwe bux .
thompson s group @xmath0 is maximally nonconvex . in _ geometric methods in group theory _ ,
volume 372 of _ contemp .
_ , pages 131146 .
soc . , providence , ri , 2005 .
carsten thiel . .
bonner mathematische schriften [ bonn mathematical publications ] , 234 . universitt bonn mathematisches
institut , bonn , 1992 .
dissertation , rheinische friedrich - wilhelms - universitt bonn , bonn , 1991 . | we give the first examples of groups which admit a tame combing with linear radial tameness function with respect to any choice of finite presentation , but which are not minimally almost convex on a standard generating set .
namely , we explicitly construct such combings for thompson s group @xmath0 and the baumslag - solitar groups @xmath1 with @xmath2 . in order to make this construction for thompson s group @xmath0
, we significantly expand the understanding of the cayley complex of this group with respect to the standard finite presentation .
in particular we describe a quasigeodesic set of normal forms and combinatorially classify the arrangements of 2-cells adjacent to edges that do not lie on normal form paths . |
[ sec : intro ] we show that a hamiltonian diffeomorphism of a closed symplectically aspherical manifold has infinitely many periodic points .
more precisely , we prove that such a diffeomorphism with finitely many fixed points has simple periodic points of arbitrarily large period .
for tori , this fact , recently established by hingston , @xcite , was conjectured by conley , @xcite and is frequently referred to as the _ conley conjecture_. ( see also @xcite and references therein for similar results for hamiltonian diffeomorphisms and homeomorphisms of surfaces . ) the proof given here uses some crucial ideas from @xcite , but is completely self - contained .
the main result of the paper is [ thm : main ] let @xmath0 be a hamiltonian diffeomorphism of a closed symplectically aspherical manifold @xmath1 .
assume that the fixed points of @xmath2 are isolated .
then @xmath2 has simple periodic points of arbitrarily large period .
we refer the reader to section [ sec : general ] for the definitions . here
we only point out that a hamiltonian diffeomorphism is the time - one map of a time - dependent hamiltonian flow and that the manifolds @xmath1 with @xmath3 ( e.g. , tori and surfaces of genus greater than zero ) are among symplectically aspherical manifolds .
thus , theorem [ thm : main ] implies in particular the conley conjecture for tori , @xcite , and the results of @xcite on hamiltonian diffeomorphisms of such surfaces .
[ cor : main ] a hamiltonian diffeomorphism @xmath2 of a closed symplectically aspherical manifold has infinitely many simple periodic points .
the example of an irrational rotation of @xmath4 shows that in general the requirement that @xmath1 is symplectically aspherical can not be completely eliminated ; see , however , @xcite .
let @xmath5 be a periodic in time hamiltonian giving rise to @xmath2 .
since periodic points of @xmath2 are in one - to - one correspondence with periodic orbits of the time - dependent hamiltonian flow @xmath6 , theorem [ thm : main ] and corollary [ cor : main ] can be viewed as results about periodic orbits of @xmath5 .
then , in both of the statements , the periodic orbits can be assumed to be contractible .
( it is not hard to see that contractibility is a property of a fixed point rather than of an orbit , independent of the choice of @xmath5 . )
finally note that , as simple examples show , the assumption of theorem [ thm : main ] that the fixed points of @xmath2 are isolated can not be dropped as long as the periodic orbits are required to be contractible .
there are numerous parallels between the hamiltonian conley conjecture considered here and its lagrangian counterpart ; see , e.g. , @xcite and references therein .
the similarity between the two problems goes beyond the obvious analogy of the statements and can also easily be seen on the level of the proofs , although the methods utilized in @xcite are quite different from the floer homological techniques used in the present paper .
thus , for instance , our proposition [ prop2 ] plays the same role as bangert s homological vanishing method originating from @xcite in , e.g. , @xcite . in the framework of symplectic topology
, there are two essentially different approaches to proving results along the lines of the conley conjecture . the first approach , due to conley and salamon and zehnder , @xcite , is based on an iteration formula for the conley
zehnder index , asserting that the index of an isolated weakly non - degenerate orbit either grows linearly under iterations or its absolute value does not exceed @xmath7 , where @xmath8 .
this , in particular , implies that the local floer homology of such an orbit eventually becomes zero in degree @xmath9 as the order of iteration grows , provided that the orbit remains isolated .
( we refer the reader to sections [ sec : prelim ] and [ sec : lfh ] for the definitions .
the argument of salamon and zehnder , @xcite , does not rely on the notion of local floer homology , but this notion becomes indispensable in the proof of theorem [ thm : main ] . )
since the floer homology of @xmath1 in degree @xmath9 is non - zero , it follows that when all one - periodic orbits are weakly non - degenerate , new simple orbits must be created by large prime iterations to generate the floer homology in degree @xmath9 ; see @xcite for details .
the second approach comprises a broad class of methods and is based on the idea that a hamiltonian @xmath5 with sufficiently large variation must have one - periodic orbits with non - vanishing action . since iterating a hamiltonian diffeomorphism @xmath2 has the same effect as , roughly speaking , increasing the variation of @xmath5 , one can expect @xmath2 to have infinitely many periodic points .
when a sufficiently accurate upper or lower bound on the action is available , the orbits can be shown to be simple .
the results obtained along these lines are numerous and use a variety of symplectic topological techniques and assumptions on @xmath1 and @xmath5 . for instance , if the support of @xmath5 is displaceable and the variation of @xmath5 is greater than the displacement energy @xmath10 of the support , one - periodic orbits with action in the range @xmath11 $ ] have been shown to exist for many classes of symplectic manifolds and hamiltonians ; see , e.g. , @xcite .
then , the _ a priori _ bound on action implies the existence of simple periodic orbits with non - zero action and arbitrarily large period .
these methods do not rely on particular requirements on the fixed points of @xmath2 , but the assumption that the support is displaceable appears at this moment to be crucial . within
this broad class is also a group of methods applicable to hamiltonians @xmath5 with sufficiently degenerate large or `` flat '' maximum and detecting orbits with action slightly greater than the maximum of @xmath5 ; see , e.g. , @xcite .
iterating @xmath2 can be viewed as stretching @xmath5 near its maximum , and thus increasing its variation .
hence , methods from this group can also be used in some instances to prove the existence of simple periodic orbits of large period . here the condition that the maximum is in a certain sense flat is crucial , but the assumption that the support of the hamiltonian is displaceable is less important and , in some cases , not required at all .
in fact , what appears to matter is that the set where the maximum is attained is relatively small ( e.g. , symplectic as in @xcite or displaceable as in @xcite or just isolated as in @xcite ) .
it is one of these methods , combined with the conley
zehnder approach , that we use in the proof of the conley conjecture .
a work of hingston @xcite clearly suggests the idea , which is central to our proof , that the two approaches outlined above can be extended to cover the case of an arbitrary hamiltonian .
namely , the method of @xcite detects infinitely many simple periodic points of arbitrarily large periods , unless there exists a strongly degenerate @xmath12-periodic point @xmath13 such that the local floer homology groups of the @xmath12-th iteration @xmath14 and of a large iteration of @xmath14 at @xmath13 are non - zero in degree @xmath9 ; see section [ sec : proof ]
. then we show ( proposition [ prop1 ] ) that the @xmath15 is the time-@xmath12 flow of a @xmath12-periodic hamiltonian , say @xmath16 , such that @xmath13 is a ( constant ) local maximum of @xmath16 for all @xmath17 and this maximum is in a certain sense very degenerate ; cf .
( however , the hessian @xmath18 need not be identically zero . )
finally , we prove ( proposition [ prop2 ] ) that large iterations of @xmath5 have periodic orbits with actions arbitrarily close to the action of the iterated hamiltonian at @xmath13 .
these orbits are necessarily simple due to the lower and upper bounds on the action .
proposition [ prop2 ] is established by using a simple squeezing argument akin to the ones from @xcite .
this concludes the proof of the theorem .
this argument is extremely flexible and readily extends to manifolds convex at infinity or geometrically bounded and wide ; see @xcite and @xcite for the definitions
. we will give a detailed proof of the conley conjecture for such manifolds elsewhere .
in section [ sec : prelim ] , we set notation and conventions , briefly review elements of floer theory , and also discuss the properties of loops of hamiltonian diffeomorphisms relevant to the proof .
local floer homology is the subject of section [ sec : lfh ] . in section [ sec :
proof ] , we state propositions [ prop1 ] and [ prop2 ] mentioned above and derive theorem [ thm : main ] from these propositions .
proposition [ prop1 ] reduces the problem to the case of a hamiltonian with strict , but `` flat '' , local maximum .
this proposition is proved in sections [ sec : pr - prop1 ] and [ sec : gf ] by adapting an argument from @xcite .
proposition [ prop2 ] asserting the existence of simple periodic orbits of large period for such a hamiltonian and completing the proof of theorem [ thm : main ] is established in section [ sec : prop2-pf ] .
the author is deeply grateful to baak grel , doris hein , ely kerman , and felix schlenk for their numerous valuable remarks and suggestions .
[ sec : prelim ] [ sec : notation ] [ sec : general ] throughout the paper , a smooth @xmath19-dimensional manifold is denoted by @xmath20 and @xmath1 stands for a symplectic manifold of dimension @xmath21 , equipped with a symplectic form @xmath22 .
the manifold @xmath23 is always assumed to be closed and _ symplectically aspherical _ , i.e. , @xmath24 , where @xmath25 is the first chern class of @xmath1 ; see , e.g. , @xcite .
an almost complex structure @xmath26 on @xmath1 is said to be _ compatible with @xmath22 _ if @xmath27 is a riemannian metric on @xmath1 .
when @xmath28 depends on an extra parameter @xmath17 ( time ) , this condition is required to hold for every @xmath17 .
a ( time - dependent ) metric of the form @xmath27 is said to be compatible with @xmath22 .
the group of linear symplectic transformations of a finite - dimensional linear symplectic space @xmath29 is denoted by @xmath30 .
we will also need the fact that @xmath31 ( see , e.g. , @xcite ) , and hence @xmath32 . to be more specific , fixing a linear complex structure @xmath26 on @xmath33 , compatible with @xmath22 ,
gives rise to an inclusion @xmath34 of the unitary group into the symplectic group .
this inclusion is a homotopy equivalence .
the isomorphism @xmath31 is the composition of the isomorphism @xmath35 , the isomorphism of the fundamental groups induced by @xmath36 ( the unit circle in @xmath37 ) , and the identification @xmath38 arising from fixing the counter clock - wise orientation of @xmath39 .
note that the resulting isomorphism is independent of the choice of @xmath26 .
the _ maslov index _ of a loop in @xmath30 is the class of this loop in @xmath40 .
we use the notation @xmath39 for the circle @xmath41 and the circle @xmath42 of circumference @xmath43 is denoted by @xmath44 .
all hamiltonians @xmath5 on @xmath1 considered in this paper are assumed to be @xmath45-periodic ( in time ) , i.e. , @xmath46 .
we set @xmath47 for @xmath48 .
the hamiltonian vector field @xmath49 of @xmath5 is defined by @xmath50 .
let @xmath51 be a contractible loop .
the action of @xmath5 on @xmath52 is defined by @xmath53 here @xmath54 is the negative symplectic area bounded by @xmath52 , i.e. , @xmath55 where @xmath56 is such that @xmath57 .
the least action principle asserts that the critical points of @xmath58 on the space of all contractible maps @xmath51 are exactly the contractible @xmath45-periodic orbits of the time - dependent hamiltonian flow @xmath59 of @xmath5 .
when the period @xmath45 is clear from the context and , in particular , if @xmath60 , we denote the time-@xmath45 map @xmath61 by @xmath62 .
the action spectrum @xmath63 of @xmath5 is the set of critical values of @xmath58 .
this is a zero measure , closed set ; see , e.g. , @xcite . in this paper
we are only concerned with contractible periodic orbits . _
a periodic orbit is always assumed to be contractible , even if this is not explicitly stated_. a @xmath45-periodic orbit @xmath52 of @xmath5 is _ non - degenerate _ if the linearized return map @xmath64 has no eigenvalues equal to one .
following @xcite , we call @xmath52 _ weakly non - degenerate _ if at least one of the eigenvalues is different from one .
when all eigenvalues are equal to one , the orbit is said to be _
strongly degenerate_. when @xmath52 is non - degenerate or even weakly non - degenerate , the so - called _ conley zehnder index _ @xmath65 is defined , up to a sign , as in @xcite .
more specifically , in this paper , the conley
zehnder index is the negative of that of @xcite . in other words , we normalize @xmath66 so that @xmath67 when @xmath52 is a non - degenerate maximum of an autonomous hamiltonian with small hessian . more generally , when @xmath5 is autonomous and @xmath52 is a non - degenerate critical point of @xmath5 such that the eigenvalues of the hessian ( with respect to a metric compatible with @xmath22 ) are less than @xmath68 , the conley
zehnder index of @xmath52 is equal to one half of the negative signature of the hessian .
sometimes , the same hamiltonian can be treated as @xmath45-periodic for different values of @xmath43 .
for instance , an autonomous hamiltonian is @xmath45-periodic for every @xmath43 and a @xmath45-periodic hamiltonian can also be viewed as @xmath69-periodic for any integer @xmath70 . in this paper
, it will be essential to keep track of the period . unless specified otherwise , every hamiltonian @xmath5 considered here is originally one - periodic and @xmath45 is always an integer .
when we wish to view @xmath5 as a @xmath45-periodic hamiltonian , we denote it by @xmath71 and refer to it as the _
@xmath45-th iteration _ of @xmath5 .
( the parentheses here are used to distinguish iterated hamiltonians from families of hamiltonians , say @xmath72 , parametrized by @xmath73 . ) since @xmath71 is regarded as a @xmath45-periodic hamiltonian , it makes sense to speak only about @xmath45-periodic ( or @xmath69-periodic ) orbits of @xmath71 .
clearly , @xmath45-periodic orbits of @xmath71 are simply @xmath45-periodic orbits of @xmath5 . when @xmath74 is a one - periodic orbit of @xmath5 , its _
@xmath45-th iteration _ is the obviously defined map @xmath75 obtained by composing the @xmath45-fold covering map @xmath76 with @xmath52 .
thus , @xmath77 is a @xmath45-periodic orbit of @xmath5 and @xmath71 .
we call a @xmath45-periodic orbit _ simple _ if it is not an iteration of an orbit of a smaller period .
as is well - known , the fixed points of @xmath78 are in one - to - one correspondence with ( not - necessarily contractible ) one - periodic orbits of @xmath5 .
likewise , the @xmath45-periodic points of @xmath62 , i.e. , the fixed points of @xmath79 , are in one - to - one correspondence with ( not - necessarily contractible ) @xmath45-periodic orbits . in the proof of theorem [ thm : main ] , we will work with ( contractible ! )
periodic orbits of a hamiltonian @xmath5 whose time - one flow is @xmath2 . in fact , as is easy to see from section [ sec : loops ] , the free homotopy type of the one - periodic orbit of @xmath5 through a fixed point @xmath13 of @xmath2 is completely determined by @xmath2 and @xmath13 and is independent of the choice of @xmath5 .
the same holds for @xmath45-periodic points and orbits .
hence , `` contractible fixed points or periodic points '' of @xmath2 , i.e. , those with contractible orbits , are well defined and we will establish theorem [ thm : main ] for points in this class . when @xmath80 and @xmath5 are two ( say , one - periodic ) hamiltonians , the composition @xmath81 is defined by the formula @xmath82 the flow of @xmath81 is @xmath83 . in general , @xmath81 is not one - periodic .
however , this will be the case if , for example , @xmath84 .
another instance when the composition @xmath81 of two one - periodic hamiltonians is automatically one - periodic is when the flow @xmath85 is a loop of hamiltonian diffeomorphisms , i.e. , @xmath86 .
[ sec : norms ] in what follows , it will be convenient to use a somewhat unconventional terminology and work with @xmath87-norms of functions , vector fields , etc .
taken with respect to a coordinate system .
let @xmath88 be a coordinate system on a neighborhood @xmath89 of a point @xmath90 , i.e. , @xmath88 is a diffeomorphism @xmath91 sending @xmath13 to the origin .
( thus , @xmath89 is a part of the data @xmath88 . )
let @xmath92 be a function on @xmath89 or on the entire manifold @xmath20 .
the @xmath87-norm @xmath93 of @xmath92 with respect to @xmath88 is , by definition , the @xmath87-norm of @xmath92 on @xmath89 with respect to the flat metric associated with @xmath88 , i.e. , the pull - back by @xmath88 of the standard metric on @xmath94 .
the @xmath87-norm with respect to @xmath88 of a vector field or a field of operators on @xmath89 is defined in a similar fashion . likewise , the norm @xmath95 of a vector @xmath96 in a finite - dimensional vector space @xmath33 with respect to a basis @xmath97 is the norm of @xmath96 with respect to the euclidean inner product for which @xmath97 is an orthonormal basis .
the norm of an operator @xmath98 with respect to @xmath97 is defined in a similar way .
when @xmath88 is a coordinate system near @xmath13 , we denote by @xmath99 the natural coordinate basis in @xmath100 arising from @xmath88 .
[ exam : norm1 ] let @xmath101 be a linear map with all eigenvalues equal to zero . then @xmath102 can be made arbitrarily small by choosing a suitable basis @xmath97 . in other words , for any @xmath103 there exists @xmath97 such that @xmath104 .
indeed , in some basis @xmath105 is given by an upper triangular matrix with zeros on the diagonal ; @xmath97 is then obtained by appropriately scaling the elements of this basis .
[ exam : norm2 ] restricting @xmath88 to a smaller neighborhood of @xmath13 reduces the norm of @xmath92 .
however , one can not make , say , @xmath106 arbitrarily small by shrinking @xmath89 unless @xmath107 and @xmath108 .
indeed , @xmath109 .
it is clear that for a fixed basis @xmath97 in @xmath100 and a function @xmath92 near @xmath13 there exists a coordinate system @xmath88 with @xmath110 , such that @xmath106 is arbitrarily close to @xmath111 . in this section ,
we briefly recall the notion of filtered floer homology for closed symplectically aspherical manifolds .
all definitions and results mentioned here are quite standard and well known and we refer the reader to floer s papers @xcite , to @xcite , or to @xcite for introductory accounts of the construction of floer homology in this ( or more general ) setting .
let us first focus on one - periodic hamiltonians .
consider a hamiltonian @xmath5 such that all one - periodic orbits of @xmath5 are non - degenerate .
this is a generic condition and we will call such hamiltonians _ non - degenerate_. let @xmath28 be a ( time - dependent : @xmath112 ) almost complex structure on @xmath1 compatible with @xmath22 . for
two one - periodic orbits @xmath113 of @xmath5 denote by @xmath114 the space of solutions @xmath115 of the floer equation @xmath116 which are asymptotic to @xmath113 at @xmath117 , i.e. , @xmath118 point - wise as @xmath119 .
the energy @xmath120 of a solution @xmath121 of the floer equation , , is @xmath122 where we set @xmath123 . every finite energy solution of is asymptotic to some @xmath113 and @xmath124 when @xmath26 meets certain standard regularity requirements that hold generically ( see , e.g. , @xcite ) , the space @xmath114 is a smooth manifold of dimension @xmath125 .
this space carries a natural @xmath126-action @xmath127 and we denote by @xmath128 the quotient @xmath129 . when @xmath130 , the set @xmath128 is finite and we denote the number , @xmath131 , of points in this set by @xmath132 .
let @xmath133 be two points outside @xmath63 .
denote by @xmath134 the vector space over @xmath135 generated by one - periodic orbits of @xmath5 with @xmath136 and @xmath137 .
the floer differential @xmath138 is defined by @xmath139 where the summation is over all @xmath140 such that @xmath141 and @xmath142 .
as is well - known , @xmath143 .
the homology @xmath144 of the resulting complex is called the _ filtered floer homology _ of @xmath5 for the interval @xmath145 .
thus , @xmath146 is the ordinary floer homology .
it is a standard fact that @xmath147 . in general
, @xmath148 depends on the hamiltonian @xmath5 , but not on @xmath26 .
the subcomplexes @xmath149 , where @xmath150 , form a filtration of the total floer complex @xmath151 , called the _ action filtration _ , and @xmath152 can be identified with the complex @xmath153 ; see , e.g. , @xcite .
let now @xmath154 be three points outside @xmath63 . then , similarly , @xmath155 is a subcomplex of @xmath156 , and @xmath157 is naturally isomorphic to @xmath158 . as a result
, we have the long exact sequence @xmath159 the filtered floer homology of @xmath5 is defined even when the periodic orbits of @xmath5 are not necessarily non - degenerate , provided that @xmath133 are outside @xmath63 . namely , let @xmath160 be a @xmath161-small perturbation , rather than the difference @xmath162 , as a perturbation of @xmath5 .
however , it is the difference @xmath162 that is required to be @xmath161-small .
] of @xmath5 with non - degenerate one - periodic orbits .
the filtered floer homology @xmath148 of @xmath5 is by definition @xmath163 .
( clearly , @xmath133 are still outside @xmath164 . )
these groups are canonically isomorphic for different choices of @xmath160 ( close to @xmath5 ) and the results discussed here hold for @xmath148 , @xcite . in fact , when an assertion concerns individual hamiltonians ( as opposed to families of hamiltonians ) , it is usually sufficient to prove the assertion in the non - degenerate case , for then it extends `` by continuity '' to all hamiltonians .
these constructions carry over to @xmath45-periodic hamiltonians word - for - word by replacing one - periodic orbits with @xmath45-periodic ones . when @xmath5 is one - periodic , but we treat it as @xmath45-periodic for some integer @xmath43 , we denote the resulting floer homology groups @xmath165 .
[ sec : homotopy ] consider two non - degenerate hamiltonians @xmath166 and @xmath167 .
let @xmath72 be a homotopy from @xmath166 to @xmath167 . by definition
, this is a family of hamiltonians parametrized by @xmath168 such that @xmath169 when @xmath73 is large negative and @xmath170 when @xmath73 is large positive .
( strictly speaking , the notion of homotopy includes also a family of almost complex structures @xmath171 ; see , e.g. , @xcite .
we suppress this part of the homotopy structure in the notation . ) assume , in addition , that the homotopy is _ monotone decreasing _ , i.e. , @xmath172 is a decreasing function of @xmath73 for all @xmath173 and @xmath112 .
( thus , in particular , @xmath174 . )
then , whenever @xmath133 are outside @xmath175 and @xmath176 , the homotopy @xmath72 induces a homomorphism of complexes @xmath177 by the standard continuation construction ; see , e.g. , @xcite .
namely , for a one - periodic orbit @xmath178 of @xmath166 and a one - periodic orbit @xmath140 of @xmath167 , let @xmath179 be the space of solutions of with @xmath72 on the right hand side , asymptotic to @xmath178 and , respectively , @xmath140 at @xmath117 . under the well - known regularity requirements on @xmath26 and @xmath72 , the space @xmath180 is a smooth manifold of dimension @xmath181 ; see , e.g. , @xcite .
moreover , @xmath180 is a finite collection of points when @xmath182 .
the map @xmath183 is defined by @xmath184 where the summation is over all @xmath140 such that @xmath182 and @xmath142 .
the induced _ homotopy map _ in the filtered floer homology , also denoted by @xmath183 , is independent of the decreasing homotopy @xmath72 and commutes with the maps from the long exact sequence , see , e.g. , @xcite . by `` continuity '' in the hamiltonians and the homotopy , this construction extends to all ( not necessarily non - degenerate ) hamiltonians and all decreasing ( but not necessarily regular ) homotopies as long as @xmath133 are not in @xmath175 and @xmath176 . a ( non - monotone )
homotopy @xmath72 from @xmath166 to @xmath167 with @xmath185 and @xmath186 outside @xmath187 for all @xmath73 gives rise to an isomorphism between the groups @xmath188 , and hence , in particular , @xmath189 see @xcite .
this isomorphism is defined by breaking the homotopy @xmath72 into a composition of homotopies @xmath190 close to the identity .
each of the homotopies @xmath190 and its inverse homotopy increase action by no more than some small @xmath191 .
then , it is shown that the map in @xmath192 induced by @xmath190 is an isomorphism .
although this construction requires additional choices , it is not hard to see that the isomorphism is uniquely determined by the homotopy @xmath72 .
furthermore , commutes with the maps from the long exact sequence , provided that all three points @xmath154 are outside @xmath187 for all @xmath73 ; @xcite .
note also that when @xmath72 is a decreasing homotopy , the isomorphism coincides with @xmath183 .
[ ex : isospec ] a homotopy @xmath72 is said to be _ isospectral _ if @xmath187 is independent of @xmath73 . in this case , the isomorphism is defined for any @xmath133 outside @xmath187 .
for instance , let @xmath193 , where @xmath112 and @xmath194 $ ] , be a family of loops of hamiltonian diffeomorphisms based at @xmath195 , i.e. , @xmath196 for all @xmath73 . in other words , @xmath193 is a based homotopy from the loop @xmath197 to the loop @xmath198 .
let @xmath199 be a family of one - periodic hamiltonians generating these loops and let @xmath5 be a fixed one - periodic hamiltonian . then @xmath200 is an isospectral homotopy , provided that @xmath201 are suitably normalized .
( namely , @xmath202 for all @xmath73 ; see section [ sec : loops ] . )
it is easy to see that if @xmath203 for all @xmath73 , the isomorphism intertwines monotone homotopy homomorphisms from @xmath80 to @xmath166 and to @xmath167 , i.e. , the diagram @xmath204_{\psi_{k , h^0}}\ar[rd]^{\psi_{k , h^1 } } & \\ { \hf^{(a,\,b)}_*(h^0 ) } \ar[r]^{\cong } & { \hf^{(a,\,b)}_*(h^1 ) } } \ ] ] is commutative .
note that it is not at all clear whether the same is true if we only require that @xmath205 and @xmath206 .
we conclude this section by establishing a technical result , used later on in the proof , giving a criterion for a monotone homotopy map to be non - zero .
[ lemma : non - zero ] let @xmath72 be a monotone decreasing homotopy such that a point @xmath13 is a non - degenerate constant one - periodic orbit of @xmath72 and @xmath207 for all @xmath73 and @xmath17 .
then the monotone homotopy map @xmath208 is non - trivial , provided that @xmath209 .
[ rmk : non - zero ] in fact , we will prove a stronger result .
let us perturb @xmath166 and @xmath167 away from @xmath13 , making these hamiltonians non - degenerate .
then @xmath13 is a cycle in @xmath210 and @xmath211 and , moreover , this cycle is not homologous to any cycle that does not include @xmath13 .
( this is easy to see from the energy estimates ; see , e.g. , @xcite . ) in particular , @xmath212\neq 0 $ ] in @xmath213 and @xmath214 and we will show that @xmath183 sends @xmath212\in\hf^{(a,\,b)}_*(h^0)$ ] to @xmath212\in \hf^{(a,\,b)}_*(h^1)$ ] .
moreover , a simple modification of our argument proves the following : there exist @xmath161-small non - degenerate perturbations @xmath215 of @xmath166 and @xmath216 of @xmath167 for which @xmath13 is still a non - degenerate constant one - periodic orbit , and a regular monotone decreasing homotopy @xmath217 from @xmath215 to @xmath216 such that the cycle @xmath13 for @xmath215 is connected to @xmath13 for @xmath216 by an odd number of homotopy trajectories and all such trajectories are contained in a small neighborhood of @xmath13 .
( note that we do not assume that @xmath13 is a non - degenerate constant one - periodic orbit of @xmath217 for all @xmath73 or that @xmath218 .
of course , the lemma can be further generalized .
for instance , the constant orbit @xmath13 can be replaced by a fixed one - periodic orbit .
let us perturb the homotopy on the complement of a neighborhood @xmath89 of @xmath13 , keeping the homotopy monotone decreasing , to ensure that all but a finite number of the hamiltonians @xmath72 are non - degenerate . in particular , we will assume that @xmath166 and @xmath167 are such .
this can be achieved by an arbitrarily small perturbation of @xmath72 .
we keep the notation @xmath72 for the perturbed homotopy .
if the homotopy @xmath72 were regular , we would simply argue that the constant connecting trajectory @xmath219 is the only connecting trajectory from @xmath13 for @xmath166 to @xmath13 for @xmath167 .
indeed , @xmath220 for any such connecting trajectory @xmath96 , and thus @xmath96 must be constant . however , while it is easy to guarantee that @xmath72 is regular away from @xmath13 by reasoning as in , e.g. , @xcite , it is not _ a priori _ obvious that the transversality requirements can be satisfied for @xmath219 because of the constraint @xmath207 . rather than checking regularity of @xmath121 by a direct calculation , we chose to circumvent this difficulty .
as in the proof of ( see , e.g. , @xcite ) , we can break the homotopy @xmath72 by reparametrization of @xmath73 into a composition of homotopies @xmath221 from @xmath222 to @xmath223 with @xmath224 and @xmath225 .
these homotopies are monotone , since @xmath72 is monotone , and close to the identity homotopy . for every @xmath191 ,
this can be done so that the inverse homotopy @xmath226 from @xmath227 to @xmath228 increases the actions by no more than @xmath191 .
without loss of generality , we may assume that the hamiltonians @xmath228 are non - degenerate . since all `` direct '' homotopies are monotone decreasing , we have @xmath229 observe that it suffices to establish the lemma for @xmath185 and @xmath186 arbitrarily close to @xmath230 .
let @xmath89 be so small that @xmath13 is the only one - periodic trajectory of @xmath72 entering @xmath89 for all @xmath73 .
( since @xmath13 is isolated for all @xmath72 , such a neighborhood @xmath89 exists . ) there exists a constant @xmath231 such that every floer anti - gradient trajectory @xmath96 connecting @xmath13 with any other one - periodic orbit with action in the range @xmath145 has energy @xmath232 for any regular hamiltonian in the family @xmath72 ( cf .
in particular , this holds for @xmath228 and @xmath227 and , moreover , for every regular hamiltonian in the homotopy @xmath221
. we will pick @xmath185 and @xmath186 so that @xmath233 and @xmath234
. then , for every @xmath228 the point @xmath13 is a cycle in @xmath235 and @xmath13 is not homologous to any cycle that does not include @xmath13 .
now it is sufficient to prove that @xmath236 sends @xmath212\in\hf^{(a,\,b)}_*(k^i)$ ] to @xmath212\in\hf^{(a,\,b)}_*(k^{i+1})$ ] . to this end
, let us first prove that @xmath237)\neq 0 $ ] .
we may assume that none of the points @xmath185 , @xmath238 , @xmath186 , @xmath239 is in @xmath240 or in @xmath241 .
it is easy to see ( see , e.g. , @xcite ) that @xmath242 is the natural `` quotient - inclusion '' map , i.e. , the composition of the `` quotient '' and `` inclusion '' maps @xmath243 note that @xmath244 is completely determined by @xmath72 and @xmath89 and is independent of how @xmath72 is broken into the homotopies @xmath221 , and thus of @xmath191 .
pick @xmath191 so that @xmath245 and @xmath246
. then @xmath13 is a cycle in @xmath247 , which is not homologous to any cycle that does not include @xmath13 . as a consequence , @xmath212\neq 0 $ ] in both of the floer homology groups in and @xmath248)=[p]$ ]
therefore , @xmath237)\neq 0 $ ] in @xmath249 . to show that @xmath237)=[p]$ ]
, we need to refine our choice of @xmath244 .
note that there exists @xmath231 such that , in addition to the above requirements , every @xmath221-homotopy trajectory starting at @xmath13 and leaving @xmath89 has energy greater than @xmath244 .
then , clearly , the same is true for every sufficiently @xmath161-small perturbation @xmath250 of @xmath221 .
again , @xmath244 depends only on @xmath72 and @xmath89 , but not on breaking @xmath72 into the homotopies @xmath221 .
( the existence of @xmath231 with these properties readily follows from energy estimates for connecting trajectories , cf .
pick a @xmath161-small regular perturbation @xmath250 of @xmath221 .
we may still assume that @xmath250 is monotone decreasing and @xmath13 is a non - degenerate constant orbit of @xmath251 and @xmath252 .
however , @xmath13 is not required to be a constant one - periodic orbit of @xmath250 for all @xmath73 ; nor is @xmath253 constant as a function of @xmath73 and @xmath17 .
clearly , the inverse homotopy to @xmath250 does not increase action by more than @xmath191 .
hence , in the homological analysis of the previous paragraph we can replace @xmath228 and @xmath227 by @xmath254 and @xmath252 .
in fact , the original and perturbed hamiltonians have equal filtered floer homology for relevant action intervals and the maps @xmath236 and @xmath226 are induced by the maps for @xmath250 acting on the level of complexes .
therefore , @xmath255)=\psi_{k^i , k^{i+1}}([p])\neq 0 $ ] in @xmath256 , and thus @xmath257 in @xmath258 .
since @xmath233 and @xmath259 and every connecting orbit leaving @xmath89 must have energy greater than @xmath244 , we conclude that @xmath260 in the floer complexes , and hence in the floer homology .
[ sec : loops ] in this section , we recall a few well - known facts about loops of hamiltonian diffeomorphisms of @xmath1 .
we will focus on loops parametrized by @xmath39 , but obviously all results discussed here hold for loops of any period .
furthermore , throughout the paper all loops @xmath261 are assumed to be based at @xmath195 , i.e. , @xmath262 ; contractible loops are thus required to be contractible in this class .
recall that , as is proved in @xcite , the filtered floer homology of the hamiltonian @xmath5 is determined , up to a shift of filtration , entirely by the time - one map @xmath62 and is independent of the hamiltonian @xmath5 .
this fact translates to geometric properties of loops of hamiltonian diffeomorphisms , which are briefly reviewed below , and is actually proved by first establishing these properties .
let @xmath263 , @xmath112 , be a loop generated by a periodic hamiltonian @xmath264 . then
all orbits @xmath265 of @xmath266 with @xmath112 and @xmath173 are one - periodic and lie in the same homotopy class .
hence , every orbit of @xmath264 is contractible by the arnold conjecture .
the action @xmath267 is independent of @xmath173 ( see , e.g. , @xcite ) and @xmath268 , where @xmath269 is the symplectic volume of @xmath1 .
the latter identity is easy to prove when the loop @xmath266 is contractible ( see , e.g. , @xcite ) ; in the general case , this is a non - trivial result , @xcite . for @xmath52 as above pick a trivialization of @xmath270 that extends to a trivialization of @xmath271 along a disk bounded by @xmath52 . using this trivialization
, we can view the maps @xmath272 as a loop in @xmath273 .
hence , the linearization @xmath274 along @xmath52 gives rise to an element in @xmath275 , which could be called the maslov index , @xmath276 , of the loop @xmath266 if it were non - trivial ; cf.@xcite .
the maslov index is well defined : it is independent of @xmath52 , the trivialization and the disc .
( the latter follows from the fact that @xmath277 . ) however , as is well known and as we will soon reprove , @xmath278 ; see also , e.g. , @xcite .
let @xmath5 be a periodic hamiltonian on @xmath1 . recall that @xmath279 is the hamiltonian generating the flow @xmath280 .
this hamiltonian is automatically one - periodic and its time - one map is @xmath62 .
we claim that there exists an isomorphism of filtered floer homology @xmath281 indeed , composition with @xmath263 sends one - periodic orbits of @xmath5 to one - periodic orbits of @xmath282 with shift of action by @xmath283 and shift of conley
zehnder indices by @xmath284 .
( see , e.g. , @xcite ; the negative sign is a result of the difference in conventions . ) furthermore , let @xmath121 be a floer anti - gradient trajectory for @xmath5 and a time - dependent almost complex structure @xmath26 . then , as a straightforward calculation shows , @xmath285 is a floer anti - gradient trajectory for @xmath282 and the almost complex structure @xmath286 .
furthermore , it is clear that the transversality requirements are satisfied for @xmath287 if and only if they are satisfied for @xmath288 .
therefore , the composition with @xmath289 commutes with the floer differential and thus induces an isomorphism of floer complexes ( and hence homology groups ) shifting action by @xmath283 and grading by @xmath284 . applying this construction to the full floer homology @xmath290
, we see that the grading shift must be zero , i.e. , @xmath278 .
when the loop @xmath289 is contractible , the existence of an isomorphism readily follows from ; see example [ ex : isospec ] . however , it is not clear whether this is the same isomorphism as constructed above .
let now @xmath266 be a loop of ( the germs of ) hamiltonian diffeomorphisms at @xmath173 generated by @xmath264 .
in other words , the maps @xmath266 and the hamiltonian @xmath264 are defined on a small neighborhood of @xmath13 and @xmath291 for all @xmath112 .
then the action @xmath283 and the maslov index @xmath276 are introduced exactly as above with the orbit @xmath52 taken sufficiently close to @xmath13 .
in fact , we can set @xmath292 and hence @xmath293 and @xmath276 is just the maslov index of the loop @xmath294 in @xmath295 .
note that in this case @xmath276 need not be zero . we conclude this section by giving a necessary and sufficient condition , to be used later , for @xmath289 to extend to a loop of global hamiltonian diffeomorphisms of @xmath1 .
[ lemma : loop - ext ] let @xmath266 , @xmath112 , be a loop of germs of hamiltonian diffeomorphisms at @xmath173 .
the following conditions are equivalent : * the loop @xmath289 extends to a loop of global hamiltonian diffeomorphisms of @xmath1 , * the loop @xmath289 extends to a loop of global hamiltonian diffeomorphisms of @xmath1 , contractible in the class of loops fixing @xmath13 , * the loop @xmath289 is contractible in the group of germs of hamiltonian diffeomorphisms at @xmath13 , * @xmath278 .
the implications ( ii)@xmath296(i ) and ( iii)@xmath296(iv ) are clear and ( i)@xmath296(iv ) is established above . to prove that ( iv)@xmath296(iii ) , we identify a neighborhood of @xmath13 in @xmath1 with a neighborhood of the origin in @xmath297 .
then , as is easy to see , the loop @xmath266 is homotopy equivalent to its linearization @xmath274 , a loop of ( germs of ) linear maps . by the definition of the maslov index , @xmath274 is contractible in @xmath298 if and only if @xmath299 .
to complete the proof of the lemma , it remains to show that ( iii)@xmath296(ii ) . to this end , let us first analyze the case where @xmath266 is @xmath300-close to the identity .
fixing a small neighborhood @xmath89 of @xmath13 , we identify a neighborhood of the diagonal in @xmath301 with a neighborhood of the zero section in @xmath302 .
then the graphs of @xmath266 in @xmath301 turn into lagrangian sections of @xmath302 .
these sections are the graphs of exact forms @xmath303 on @xmath89 , where all @xmath304 are @xmath161-small and @xmath305
. then we extend ( the germs of ) the functions @xmath304 to @xmath161-small functions @xmath306 on @xmath1 such that @xmath307 .
the graphs of @xmath308 in @xmath309 form a loop of exact lagrangian submanifolds which are @xmath300-close to the zero section . thus
, this loop can be viewed as a loop of hamiltonian diffeomorphisms of @xmath1 .
it is clear that the resulting loop is contractible in the class of loops fixing @xmath13 . to deal with the general case ,
consider a family @xmath310 , @xmath311 $ ] , of local loops with @xmath312 and @xmath313 .
let @xmath314 be a partition of the interval @xmath315 $ ] such that the loops @xmath316 and @xmath317 are @xmath300-close for all @xmath318 .
in particular , the loop @xmath319 is @xmath300-close to @xmath320 , and thus extends to a contractible loop @xmath321 on @xmath1 . arguing inductively ,
assume that a contractible extension @xmath322 of @xmath316 has been constructed .
consider the loop @xmath323 defined near @xmath13 .
this loop is @xmath300-close to the identity , for @xmath317 and @xmath316 are @xmath300-close .
hence , @xmath324 extends to a contractible loop @xmath325 on @xmath1 .
then @xmath326 is the required extension of @xmath317 , contractible in the class of loops fixing @xmath13 .
[ rmk : loop - ext ] it is clear from the proof of lemma [ lemma : loop - ext ] that the extension of the germ of a loop near @xmath13 to a global loop fixing @xmath13 can be carried out with some degree of control of the @xmath87-norm and the support of the loop
. we will need the following simple fact , which can be easily verified by adapting the proof of the implication ( iii)@xmath296(ii ) .
assume that @xmath289 is the germ of a loop near @xmath13 and the linearization of @xmath289 at @xmath13 is equal to the identity : @xmath327 for all @xmath17 .
then @xmath289 extends to a loop @xmath328 of global hamiltonian diffeomorphisms of @xmath1 such that @xmath328 is contractible in the class of loops fixing @xmath13 and having identity linearization at @xmath13 .
[ sec : lfh ] [ sec : lmh ] let @xmath329 be a smooth function on a manifold @xmath20 and let @xmath330 be an isolated critical point of @xmath92 . fix a small neighborhood @xmath89 of @xmath13 containing no other critical points of @xmath92 and consider a small generic perturbation @xmath331 of @xmath92 in @xmath89 . to be more precise ,
@xmath331 is morse inside @xmath89 and @xmath300-close to @xmath92 .
then , as is easy to see , for any two critical points of @xmath331 in @xmath89 , all anti - gradient trajectories connecting these two points are contained in @xmath89 .
moreover , the same is true for broken trajectories connecting these two points . as a consequence ,
the vector space ( over @xmath135 ) generated by the critical points of @xmath331 in @xmath89 is a complex with ( morse ) differential defined in the standard way .
( see , e.g. , @xcite . ) furthermore , the continuation argument shows that the homology of this complex , denoted here by @xmath332 and referred to as the _ local morse homology _ of @xmath92 at @xmath13 , is independent of the choice of @xmath331 .
this construction is a particular case of the one from @xcite .
assume that @xmath13 is a non - degenerate critical point of @xmath92 of index @xmath70 .
then @xmath333 when @xmath334 and @xmath335 otherwise .
[ exam : morse - max ] when @xmath13 is a strict local maximum of @xmath92 , we have @xmath336 . indeed , in this case , as is easy to see from standard morse theory , @xmath337 where @xmath191 is assumed to be small and such that @xmath338 is a regular value of @xmath92 .
we will need the following two properties of local morse homology : 1 .
let @xmath339 , @xmath194 $ ] , be a family of smooth functions with _ uniformly isolated _
critical point @xmath13 , i.e. , @xmath13 is the only critical point of @xmath339 , for all @xmath73 , in some neighborhood of @xmath13 independent of @xmath73 .
then @xmath340 is constant throughout the family , and hence @xmath341 ; cf .
* lemma 4 ) .
the function @xmath92 has a ( strict ) local maximum at @xmath13 if and only if @xmath342 , where @xmath343 . the first assertion , ( lm1 ) , is again established by the continuation argument ; cf . @xcite .
we emphasize that here the assumption that @xmath13 is uniformly isolated is essential and can not be replaced by the weaker condition that @xmath13 is just an isolated critical point of @xmath339 for all @xmath73 .
( example : @xmath344 on @xmath126 with @xmath345 .
the author is grateful to doris hein for this remark . ) regarding ( lm2 ) first note that , by example [ exam : morse - max ] , @xmath342 when @xmath92 has a strict local maximum at @xmath13 .
the converse requires a proof although the argument is quite standard .
denote by @xmath266 the anti - gradient flow of @xmath92 .
let @xmath346 be a closed connected neighborhood of @xmath13 with piecewise smooth boundary @xmath347 such that whenever @xmath348 and @xmath349 the entire trajectory segment @xmath350 with @xmath351 $ ] is also in @xmath346 , and @xmath13 is the only critical point of @xmath92 contained in @xmath346 .
we call @xmath346 a gromoll meyer neighborhood of @xmath13 .
it is not hard to show that @xmath13 has an ( arbitrarily small ) gromoll meyer neighborhood ; see @xcite or @xcite .
( strictly speaking , the above definition is slightly different from the one used in @xcite .
however , the existence proof given in @xcite goes through with no modifications . )
when @xmath331 is a @xmath161-small generic perturbation of @xmath92 supported in @xmath346 , the morse complex of @xmath352 is defined and its homology is equal to @xmath332 .
( the fact that @xmath143 follows from the requirements on @xmath346 . ) assume that @xmath342 .
then there exists a non - zero cycle @xmath353 of degree @xmath19 in the morse complex of @xmath352 .
let @xmath33 be the closure of the union of the unstable manifolds of @xmath352 for all local maxima entering @xmath353 .
the set @xmath33 is the closure of a domain with piecewise smooth boundary .
the condition that @xmath353 is a cycle implies that for every critical point @xmath178 of @xmath331 in @xmath33 , the intersection of the unstable manifold of @xmath178 with @xmath346 is contained entirely in the interior of @xmath33 .
hence , @xmath354 , and thus @xmath355 .
it follows that at every smooth point @xmath356 , the gradient @xmath357 either points inward or is tangent to @xmath347 . consider a gromoll meyer neighborhood @xmath358 of @xmath13 .
note that for a small generic @xmath191 the connected component @xmath346 of @xmath359 containing @xmath13 is also a gromoll meyer neighborhood . clearly , when @xmath13 is not a local maximum of @xmath92 , there are smooth points on @xmath347 where @xmath360 points inward , provided that @xmath191 is small . as a consequence of the above analysis ,
@xmath361 if @xmath13 is not a local maximum .
this completes the proof of the implication @xmath362 .
generalizing example [ exam : morse - max ] and the proof of ( lm2 ) , it is not hard to relate local morse homology to local homology of a function , introduced in @xcite ; see also @xcite . however , we do not touch upon this question , for such a generalization is not necessary for the proof of theorem [ thm : main ] . in the setting of local homology ,
the analogues of ( lm1 ) and ( lm2 ) are established in @xcite and , respectively , in @xcite .
[ sec : lfh2 ] let @xmath52 be an isolated one - periodic orbit of a hamiltonian @xmath363 . pick a sufficiently small tubular neighborhood @xmath89 of @xmath52 and consider a non - degenerate @xmath161-small perturbation @xmath160 of @xmath5 supported in @xmath89 . more specifically ,
let @xmath89 be a neighborhood of @xmath364 , where @xmath52 is viewed as a curve in the extended phase space @xmath365 , and let @xmath160 be a hamiltonian @xmath161-close to @xmath5 , equal to @xmath5 outside of @xmath89 , and such that all one - periodic orbits of @xmath160 that enter @xmath89 are non - degenerate .
( such perturbations @xmath160 do exist ; see ( * ? ? ?
* theorem 9.1 ) . ) abusing notation , we will treat @xmath89 simultaneously as an open set in @xmath1 and in @xmath366 .
consider one - periodic orbits of @xmath160 contained in @xmath89 .
every anti - gradient trajectory @xmath121 connecting two such orbits is also contained in @xmath89 , provided that @xmath367 and @xmath368 are small enough .
indeed , the energy @xmath120 is equal to the difference of action values on the periodic orbits , and thus is bounded from above by @xmath369 .
the @xmath161-norm of @xmath160 is bounded from above by a constant independent of @xmath160 , say @xmath370 .
therefore , @xmath371 is pointwise uniformly bounded by @xmath369 , and it follows that @xmath121 takes values in @xmath89 ; see @xcite .
note also that for a suitable small perturbation of a fixed almost complex structure on @xmath1 the transversality requirements are satisfied for moduli spaces of floer anti - gradient trajectories connecting one - periodic orbits @xmath160 contained in @xmath89 , @xcite . by the compactness theorem ,
every broken anti - gradient trajectory @xmath121 connecting two one - periodic orbits in @xmath89 lies entirely in @xmath89 . hence , the vector space ( over @xmath135 ) generated by one - periodic orbits of @xmath160 in @xmath89 is a complex with ( floer ) differential defined in the standard way .
the continuation argument ( see , e.g. , @xcite ) shows that the homology of this complex is independent of the choice of @xmath160 and of the almost complex structure .
we refer to the resulting homology group @xmath372 as the _ local floer homology _ of @xmath5 at @xmath52 .
homology groups of this type were first considered ( in a more general setting ) by floer in @xcite ; see also ( * ? ? ?
* section 3.3.4 ) .
in fact , an orbit @xmath52 can be replaced by a connected isolated set @xmath373 of one - periodic orbits of @xmath5 ; @xcite .
( note that @xmath374 is constant , for @xmath58 is continuous and @xmath63 is nowhere dense . )
assume that @xmath52 is non - degenerate and @xmath375 .
then @xmath376 when @xmath334 and @xmath377 otherwise
. we will need the following properties of local floer homology : 1 .
let @xmath72 , @xmath194 $ ] , be a family of hamiltonians such that @xmath52 is a _ uniformly isolated _ one - periodic orbit for @xmath72 , i.e. , @xmath52 is the only periodic orbit of @xmath72 , for all @xmath73 , in some open set independent of @xmath73 .
then @xmath378 is constant throughout the family , and hence @xmath379 .
this is again an immediate consequence of the continuation argument .
however , it is worth pointing out that unless @xmath72 is monotone decreasing , the isomorphism @xmath379 is not induced by the homotopy @xmath72 in the same sense as the homomorphism @xmath183 is induced by a monotone homotopy ; see .
the isomorphism in question is constructed similarly to by breaking @xmath72 into a composition of homotopies close to the identity .
local floer homology spaces are building blocks for filtered floer homology .
namely , essentially by definition , we have the following 1 .
let @xmath380 be such that all one - periodic orbits @xmath381 of @xmath5 with action @xmath230 are isolated .
( as a consequence , there are only finitely many such orbits . )
then , if @xmath191 is small enough , @xmath382 in particular , if all one - periodic orbits @xmath52 of @xmath5 are isolated and @xmath383 for some @xmath70 and all @xmath52 , we have @xmath384 by the long exact sequence of filtered floer homology .
the effect on local floer homology of the composition of @xmath5 with a loop of hamiltonian diffeomorphisms is the same as in the global setting and is established in a similar fashion ; see section [ sec : loops ] . 1 .
let @xmath263 be a loop of hamiltonian diffeomorphisms of @xmath1 .
then @xmath385 for every isolated one - periodic orbit @xmath52 of @xmath5 , where @xmath386 stands for the one - periodic orbit @xmath387 of @xmath282 corresponding to @xmath52 ; see section [ sec : loops ] .
as is clear from the definition of local floer homology , @xmath5 need not be a function on the entire manifold @xmath1 it is sufficient to consider hamiltonians defined only on a neighborhood of @xmath52 . for the sake of simplicity , we focus on the particular case , relevant here , where @xmath388 is a constant orbit , and hence @xmath389 for all @xmath112 . then ( lf1 ) still holds and ( lf3 ) takes the following form : 1 . let @xmath263 be a loop of hamiltonian diffeomorphisms defined on a neighborhood of @xmath13 and fixing @xmath13 ( i.e. , @xmath291 for all @xmath112 ) .
then @xmath390 where @xmath391 is the maslov index of the loop @xmath392 .
note that in ( lf3 ) , in contrast with ( lf4 ) , we _ a priori _ know that @xmath393 as is pointed out in section [ sec : loops ] . hence , the shift of degrees does not occur when @xmath266 is a global loop . in other words , comparing ( lf3 ) and ( lf4 ) , we can say that the group @xmath372 is completely determined by the hamiltonian diffeomorphism @xmath394 and its fixed point @xmath395 , while the germ of @xmath62 at @xmath13 determines @xmath396 only up to a shift in degree .
the degree depends on the class of @xmath59 in the universal covering of the group of germs of hamiltonian diffeomorphisms .
finally note that in the construction of local floer homology the hamiltonian @xmath5 need not have period one .
the definitions and results above extend word - for - word to @xmath45-periodic hamiltonians and , in particular , to the @xmath45-th iteration @xmath71 of a one - periodic hamiltonian @xmath5 as long as the @xmath45-periodic orbit in question is isolated .
[ sec : lfh - lmh ] a fundamental property of floer homology is that @xmath397 is equal to the morse homology of a smooth function on @xmath1 ( and thus to the homology of @xmath1 ) .
the key to establishing this fact is identifying @xmath397 with @xmath398 , when the hamiltonian @xmath5 is autonomous and @xmath161-small ; see @xcite .
a similar identification holds for local floer homology .
we consider here the case of @xmath45-periodic hamiltonians , for this is the ( superficially more general ) situation where the results will be applied in the subsequent sections .
[ ex : lfh - lmh ] assume that @xmath13 is an isolated critical point of an autonomous hamiltonian @xmath399 and @xmath400 then @xmath401 . indeed ,
when the condition is satisfied , the hamiltonians @xmath402 , @xmath403 $ ] , have no non - trivial @xmath45-periodic orbits ( uniformly ) near @xmath13 .
( see @xcite or the proof of lemma [ lemma : lfh - lmh ] below . )
thus , @xmath13 is a uniformly isolated @xmath45-periodic orbit of @xmath402 for @xmath404 $ ] when @xmath191 is small , and @xmath405 is constant throughout this family by ( lf1 ) .
the argument of @xcite shows that the floer complex of @xmath406 is equal to the local morse complex of @xmath399 when @xmath73 is close to zero . in what follows
, we will need a slightly more general version of this fact , where the hamiltonian is `` close '' to a function independent of time .
[ lemma : lfh - lmh ] let @xmath399 be a smooth function and let @xmath80 be a @xmath45-periodic hamiltonian , both defined on a neighborhood of a point @xmath13 .
assume that @xmath13 is an isolated critical point of @xmath399 , and the following conditions are satisfied : * the inequalities @xmath407 and @xmath408 hold pointwise near @xmath13 for all @xmath48 .
( the dot stands for the derivative with respect to time . ) * the hessians @xmath409 and @xmath410 and the constant @xmath191 are sufficiently small .
namely , @xmath411 and @xmath412 then @xmath13 is an isolated @xmath45-periodic orbit of @xmath80 .
furthermore , 1 .
@xmath413 ; 2 .
if @xmath414 , the functions @xmath415 for all @xmath17 and @xmath399 have a strict local maximum at @xmath13 . the requirement of this lemma , asserting that @xmath80 is in a certain sense close to @xmath399 , plays a crucial role in our proof of theorem [ thm : main ] ( cf .
lemmas [ lemma : local ] and [ lemma : local2 ] ) and in the argument of @xcite . to the best of the author
s knowledge , this requirement is originally introduced in ( * ? ? ?
* lemma 4 ) as that @xmath80 is relatively autonomous . in
what follows , we will sometimes call @xmath399 a _ reference function _ and say that the pair @xmath416 meets the requirements of lemma [ lemma : lfh - lmh ] .
since the statement is local , we may assume that @xmath417 .
consider the family of hamiltonians @xmath418 starting with @xmath419 and ending with @xmath420 .
we claim that @xmath292 is a uniformly isolated @xmath45-periodic orbit of @xmath421 for @xmath194 $ ] .
we show this by adapting the proof of ( * ? ? ?
* proposition 17 , p.184 ) .
fix @xmath422 and let @xmath423 be the ball of radius @xmath424 centered at @xmath13 . since @xmath13 is a constant @xmath45-periodic orbit of @xmath421 , every @xmath45-periodic orbit @xmath52 of @xmath421 with @xmath395 sufficiently close to @xmath13 is contained in @xmath423 .
recall also that @xmath425 for any map @xmath426 with zero mean . applying this inequality to @xmath427
, we obtain @xmath428 furthermore , from the first requirement on @xmath429 and @xmath430 , it is easy to see that @xmath431 pointwise . hence , @xmath432 once holds and @xmath422 is small , we have @xmath433 therefore , @xmath434 . in other words , @xmath52 is a constant loop , and thus a critical point of @xmath435 for @xmath48 .
then , by , @xmath436 . as a consequence , @xmath292 since @xmath13 is an isolated critical point of @xmath399 .
this shows that @xmath13 is a uniformly isolated @xmath45-periodic orbit of @xmath421 . by ( lf1 ) ,
the local floer homology @xmath437 is constant throughout the family @xmath421 , and @xmath438 . as a consequence of
, the condition of example [ ex : lfh - lmh ] is satisfied . applying this example , we conclude that @xmath439 .
this proves ( a ) . by ( lm2 )
, @xmath13 is an isolated local maximum of @xmath440 , and hence , as is easy to see from the first condition of the lemma , @xmath13 is a uniformly isolated critical point of @xmath435 for @xmath441 $ ] and every fixed @xmath48 .
now , by ( lm1 ) and ( lm2 ) applied to @xmath442 , all functions @xmath435 , and , in particular , @xmath443 , have a ( strict ) local maximum at @xmath13 .
this proves ( b ) and concludes the proof of the lemma .
[ sec : proof ] as has been pointed out above , it is sufficient to prove the theorem for ( contractible ! ) periodic orbits of a hamiltonian @xmath5 generating @xmath2 rather than for all periodic points of @xmath2 .
let @xmath363 be a one - periodic hamiltonian with finitely many one - periodic orbits @xmath444 .
then , these orbits are isolated and the action spectrum of @xmath5 is comprised of finitely many points . for every one - periodic orbit @xmath444 of @xmath5 denote by @xmath445 the degrees of roots of unity , different from 1 , among the floquet multipliers of @xmath444 . arguing by contradiction , assume that for every sufficiently large integer @xmath12 , all @xmath12-periodic orbits of @xmath5 are iterated or , in other words , @xmath62 has only finitely many _ simple periods _ ,
i.e. , periods of simple , non - iterated , orbits .
in particular , every periodic orbit of @xmath5 with sufficiently large period is iterated .
let @xmath446 be the finite collection of integers comprised of all simple periods ( greater than 1 ) and the degrees @xmath447 for all one - periodic orbits @xmath444 .
then , in particular , every @xmath12-periodic orbit is an iterated one - periodic orbit when @xmath12 which is not divisible by any of the integers @xmath448 .
moreover , all @xmath12-periodic orbits are isolated and @xmath449 .
recall also that , when @xmath444 is a weakly non - degenerate one - periodic orbit of @xmath5 and @xmath12 is a sufficiently large integer , not divisible by @xmath445 , we have @xmath450 indeed , as is shown in @xcite , for a generic perturbation of @xmath5 supported near @xmath444 , the orbit @xmath451 splits into non - degenerate orbits with conley
zehnder index different from @xmath9 .
next observe that there exists a strongly degenerate one - periodic orbit @xmath52 of @xmath5 such that @xmath452 is an isolated @xmath453-periodic orbit for some sequence @xmath454 and @xmath455 where all @xmath453 are divisible by @xmath456 and none of @xmath453 is divisible by @xmath446 . to prove this , first note that by for any sufficiently large integer @xmath12 , not divisible by @xmath446 , there exists a totally degenerate one - periodic orbit @xmath324 such that @xmath457 .
( otherwise , held for all @xmath12-periodic orbits , and we would have @xmath458 by ( lf2 ) . ) pick an infinite sequence @xmath459 of such integers satisfying the additional requirement that @xmath460 is divisible by @xmath461 for all @xmath462 .
( for instance , we can take @xmath463 , where @xmath464 is a sufficiently large prime . ) as we have observed , for every @xmath461 there exists a totally degenerate one - periodic orbit @xmath465 such that @xmath466 . since there are only finitely many distinct one - periodic orbits , one of the orbits @xmath52 among the orbits @xmath465 and some infinite subsequence @xmath453 in @xmath461 satisfy .
( we also re - index the subsequence @xmath453 to make it begin with @xmath456 . )
let @xmath467 .
we will use the orbit @xmath52 and the sequence @xmath453 to prove * _ for every @xmath191 there exists @xmath468 such that for any @xmath469 and some @xmath470 in the range @xmath471 , depending on @xmath45 , we have @xmath472 _ the theorem readily follows from the claim .
indeed , set @xmath473 .
then , if @xmath469 is such that @xmath474 is not divisible by @xmath446 , we have @xmath475 .
thus , for any fixed @xmath191 and @xmath476 , the interval @xmath477 contains no action values of @xmath478 when @xmath45 is sufficiently large .
this contradicts the claim .
( note that we have used the assumption that @xmath2 has finitely many simple periods twice : the first time to find the orbit @xmath52 and the sequence @xmath453 and the second time to arrive at the contradiction with the claim . ) to establish the claim , it is convenient to adopt the following [ def : sympl - deg ] a one - periodic orbit @xmath52 of a one - periodic hamiltonian @xmath5 is said to be a _ symplectically degenerate _
maximum if there exists a sequence of loops @xmath465 of hamiltonian diffeomorphisms
such that @xmath479 , i.e. , @xmath465 sends @xmath13 to @xmath52 , for some point @xmath173 and all @xmath480 and @xmath17 , and such that the hamiltonians @xmath228 given by @xmath481 and the loops @xmath465 have the following properties : * the point @xmath13 is a strict local maximum of @xmath482 for all @xmath112 and all @xmath480 , * there exist symplectic bases @xmath483 in @xmath484 such that @xmath485 * the linearization of the loop @xmath486 at @xmath13 is the identity map for all @xmath480 and @xmath487 ( i.e. , @xmath488 for all @xmath112 ) and , moreover , the loop @xmath489 is contractible to @xmath195 in the class of loops fixing @xmath13 and having the identity linearization at @xmath13 .
[ rmk : sympl - deg ] regarding ( k1 ) and ( k3 ) note that since @xmath479 , the point @xmath13 is a fixed point of the flow @xmath490 of @xmath228 , and thus a critical point of @xmath491 for all @xmath17 .
furthermore , @xmath13 is also a fixed point of the loop @xmath486 for all @xmath480 and @xmath487 , for @xmath492 , and hence @xmath493 .
we refer the reader to section [ sec : norms ] for the definition and discussion of the norm with respect to a basis , used in ( k2 ) .
the hamiltonians @xmath228 and @xmath5 have the same time - one flow and there is a natural one - to - one correspondence between ( contractible ) one - periodic orbits of the hamiltonians . the hamiltonians @xmath228 can be chosen so that @xmath494 is constant and equal to @xmath495 . in what follows , we will always assume that @xmath228 is normalized in this way .
then the corresponding orbits of @xmath228 and @xmath5 have equal actions and , in particular , all hamiltonians @xmath228 have the same action spectrum and action filtration ; see section [ sec : loops ] .
symplectically degenerate maxima are further investigated in @xcite . in particular , it is shown there that condition ( k3 ) is superfluous ; see ( * ? ? ?
* remark 5.5 ) .
this fact is not used in the present paper .
assume that @xmath16 has a strict local maximum at @xmath13 ( and @xmath496 ) and @xmath497 for all @xmath17 .
then @xmath13 is a symplectically degenerate maximum of @xmath5 .
indeed , we can take @xmath498 and @xmath499 and any fixed symplectic basis as @xmath483 .
more generally , vanishing of the hessian may be replaced by the condition that @xmath500 can be made arbitrarily small by a suitable choice of @xmath97 , cf .
this condition is satisfied , for instance , when @xmath5 is autonomous and all eigenvalues of the linearization of @xmath49 at @xmath13 are equal to zero ; see lemma [ lemma : la ] .
[ ex : sympl - deg ] assume that @xmath52 is a symplectically degenerate maximum of @xmath5 .
let @xmath160 be a hamiltonian generating the flow @xmath501 , where @xmath289 is a loop of hamiltonian diffeomorphisms .
then , the periodic orbit @xmath502 of @xmath160 is a symplectically degenerate maximum of @xmath160 as is easy to verify .
( in other words , symplectic degeneracy is a property of the fixed point @xmath395 of the time - one map @xmath62 . ) for instance , in the notation of definition [ def : sympl - deg ] , the constant orbit @xmath13 is a symplectically degenerate maximum of each hamiltonian @xmath228 .
now we are in a position to state the two results that we need to complete the proof of theorem [ thm : main ] .
the first result gives a floer homological criterion for an isolated , strongly degenerate orbit @xmath52 to be a symplectically degenerate maximum , and thus translates local floer homological properties of @xmath52 to geometrical features of a constant orbit @xmath13 of hamiltonians @xmath228 .
the second one asserts non - vanishing of the filtered floer homology of an iterated hamiltonian @xmath71 for an interval of actions just above the action @xmath503 , provided that @xmath52 is a symplectically non - degenerate maximum of @xmath5 . when applied to the hamiltonian @xmath504 in place of @xmath5 , where @xmath456 is as in the claim
, these results will yield the claim .
[ prop1 ] let @xmath52 be a strongly degenerate isolated one - periodic orbit of @xmath5 such that its @xmath505-th iteration @xmath506 is also isolated and @xmath507 then @xmath52 is a symplectically degenerate maximum of @xmath5 .
[ rmk : comp ] note that , similarly to definition [ def : sympl - deg ] , requirement is a condition on the fixed point @xmath395 of @xmath62 , independent of a particular choice of @xmath5 .
[ prop2 ] let @xmath52 be a symplectically degenerate maximum of @xmath5 and let @xmath495 .
then for every @xmath191 there exists @xmath468 such that @xmath508 combining the propositions , we conclude that whenever a strongly degenerate one - periodic orbit @xmath52 of @xmath5 satisfies the hypotheses of proposition [ prop1 ] , for every @xmath191 there exists @xmath468 such that @xmath509 where @xmath495 and @xmath476 .
to prove the claim , first note that although propositions [ prop1 ] and [ prop2 ] are stated for one - periodic hamiltonians , similar results hold , of course , for hamiltonians and orbits of any period .
thus , consider the hamiltonian @xmath504 in place of @xmath5 and the isolated orbit @xmath510 in place of @xmath52 in and .
then the requirement is met due to : @xmath511 if @xmath480 is large enough , since @xmath454 .
furthermore , @xmath512 and follows immediate from .
it remains to establish propositions [ prop1 ] and [ prop2 ] to complete the proof of the theorem .
it is illuminating to compare the above proof with the argument due to salamon and zehnder from @xcite asserting that ever large prime is a simple period whenever all one - periodic orbits of @xmath5 are weakly non - degenerate .
( in particular , the number of simple periods less than or equal to @xmath70 is of order at least @xmath513 . ) in the context of the present paper relying , of course , on @xcite , this is an immediate consequence of . to be more specific , if @xmath12 is a large prime and all @xmath12-periodic orbits are iterated ,
holds for all weakly non - degenerate one - periodic orbits and @xmath458 by ( lf2 ) , if there are no totally degenerate one - periodic orbits .
when such one - periodic orbits exist , we can no longer use the salamon
zehnder argument to conclude that every large prime is a simple period or even to establish the existence of infinitely many simple periods .
the reason is that in this case the argument implies that for every large prime @xmath12 there is a one - periodic orbit @xmath52 such that @xmath514 .
it is unclear , however , if @xmath515 , and hence whether or not @xmath52 is a symplectically degenerate maximum .
[ sec : pr - prop1 ] our goal in this section is to construct the hamiltonians @xmath228 and the loops @xmath465 meeting requirements ( k1k3 ) .
this construction relies on two technical lemmas , proved in section [ sec : gf ] , and is carried out in several steps . in section [ sec : near - p ]
, we construct the hamiltonians @xmath228 and the loops @xmath465 near @xmath13 .
we begin by proving in section [ sec : step2 ] that the time one map @xmath516 can be made @xmath300-close to @xmath195 by an appropriate choice of a canonical coordinate system @xmath88 near @xmath13 .
this is essentially an elementary linear algebra fact ( lemma [ lemma : la ] , proved in section [ sec : la ] ) , asserting that a strongly degenerate linear symplectomorphism can be made arbitrarily close to the identity by conjugation within the linear symplectic group . as a consequence , near @xmath13 , the map @xmath2 is given by a generating function @xmath399 in the coordinate system @xmath88 . in section [ sec : step3 ] ,
we show that on a neighborhood of @xmath13 there exists a hamiltonian @xmath80 with time - one flow @xmath2 , which is in a certain sense close to @xmath399 . here
, the key result is lemma [ lemma : local ] spelling out the relation between @xmath399 and @xmath80 and established in section [ sec : gf ] . choosing a sequence of coordinate systems @xmath517 so that @xmath518 , we obtain a sequence of hamiltonians @xmath228 defined near @xmath13 and meeting requirement ( k2 ) .
then , again near @xmath13 , the loop @xmath465 is defined by @xmath492 .
utilizing condition , we show in section [ sec : step4 ] that the maslov index of @xmath465 is zero .
this enables us to relate homological properties of @xmath292 to the geometrical properties of @xmath228 near @xmath13 and prove ( k1 ) as a consequence of lemma [ lemma : lfh - lmh ] .
( assertion ( k2 ) easily follows from the construction of @xmath228 . )
property ( k3 ) is proved in section [ sec : k3 ] . at this stage , we further specialize our choice of canonical coordinate systems @xmath517 to ensure that all flows @xmath519 have the same linearization at @xmath13 .
then , the first part of assertion ( k3 ) is obvious . by lemma [ lemma : loop - ext ]
, the loops @xmath465 extend to @xmath1 , for @xmath520 .
this , in turn , gives an extension of @xmath228 to @xmath1 .
carrying out these extensions with some care , we can guarantee that ( k3 ) holds in its entirety .
first recall that for any contractible , closed curve @xmath523 there exists a contractible loop of hamiltonian symplectomorphisms @xmath266 for which @xmath52 is an integral curve , i.e. , @xmath524 ; cf .
* section 9 ) . for the sake of completeness ,
let us outline a proof of this fact .
consider a smooth family of closed curves @xmath525 , @xmath526 $ ] , connecting the constant loop @xmath527 to @xmath528 .
it is easy to show that there exists a smooth family of hamiltonians @xmath199 such that for every @xmath17 , the curve @xmath529 is an integral curve of @xmath199 with respect to @xmath73 , i.e. , @xmath530 . let @xmath531 be the time - one flow ( in @xmath73 ) of this family , parametrized by @xmath112 . then @xmath524 .
the family of hamiltonians @xmath199 can be chosen so that @xmath532
. then @xmath266 is a loop of hamiltonian diffeomorphisms with @xmath533 .
as readily follows from the construction , the loop @xmath289 is contractible . composing @xmath6 with the loop @xmath534 and adding , if necessary , a time - dependent constant function to the resulting hamiltonian @xmath535 , we may assume without loss of generality that @xmath388 is a fixed point of the flow @xmath536 for all @xmath112 and @xmath537
. then @xmath535 has the same time - one map and the same filtered floer homology as @xmath5 .
by example [ ex : sympl - deg ] and remark [ rmk : comp ] , it is sufficient to prove the proposition for @xmath535 .
thus , we will assume from now on that @xmath292 and keep the notation @xmath5 for the modified hamiltonian @xmath535 .
[ [ the - construction - of - the - hamiltonians - ki - and - the - loops - eta_i - near - p ] ] the construction of the hamiltonians @xmath228 and the loops @xmath465 near @xmath13 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ our main objective in this section is to show that for every @xmath103 , there exists a symplectic basis @xmath97 in @xmath484 and a hamiltonian @xmath80 on a neighborhood of @xmath13 such that the time - one flow of @xmath80 is @xmath2 , condition ( k1 ) is satisfied , and @xmath538 then , clearly , there exists a sequence of symplectic bases @xmath483 and a sequence of hamiltonians @xmath228 meeting requirements ( k1 ) and ( k2 ) .
the loop @xmath465 is defined near @xmath13 by @xmath539 . *
_ for any @xmath103 there exists a symplectic basis @xmath97 in @xmath484 such that @xmath540 . as a consequence ( cf .
example [ exam : norm1 ] ) , for any @xmath103 there exists a canonical system of coordinates @xmath88 on a neighborhood @xmath89 of @xmath13 such that the @xmath541-distance from @xmath2 to the identity is less than @xmath542 .
_ [ lemma : la ] let @xmath543 be a linear symplectic map of a finite dimensional symplectic vector space @xmath29 such that all eigenvalues of @xmath544 are equal to one .
then @xmath544 is conjugate in @xmath545 to a linear map which is arbitrarily close to the identity
. indeed , since @xmath13 is a strongly degenerate fixed point of @xmath5 , all eigenvalues of @xmath546 are equal to one .
thus , the desired statement follows from this lemma applied to @xmath547 .
the proof of the lemma is elementary and provided for the sake of completeness in section [ sec : la ] .
here we only mention that @xmath544 is given by an upper triangular matrix in some basis @xmath97 and , by scaling the elements of @xmath97 appropriately , one can make @xmath544 arbitrarily close to the identity , cf .
example [ exam : norm1 ] .
hence , we only need to show that @xmath97 and the scaling can be made symplectic . [
sec : step3 ] pick a system @xmath88 of canonical coordinates near @xmath13 such that @xmath2 is @xmath541-close to the identity .
in particular , @xmath548 is small .
furthermore , the map @xmath2 is given , near @xmath13 , by a generating function @xmath399 . the precise definition of @xmath399 and the relation between @xmath399 and @xmath2 and @xmath88 are immaterial at the moment and these issues will be discussed in section [ sec : gf ] . at this stage , we only need to know that @xmath399 is defined on a neighborhood of @xmath13 and uniquely determined by @xmath88 and @xmath2 .
( to make this statement accurate , let us agree that a canonical coordinate system is comprised of _ ordered _ pairs of functions @xmath549 such that @xmath550 .
thus , each coordinate function is assigned to either @xmath551- or @xmath552-group . )
moreover , @xmath399 has the following properties : the second item ( gf2 ) requires , perhaps , a clarification .
first note that @xmath555 stands here for the @xmath541-distance from @xmath2 to @xmath195 ; see section [ sec : norms ] .
furthermore , @xmath399 and @xmath555 depend on @xmath88 .
thus , in ( gf2 ) , we view both @xmath556 and @xmath555 as functions of @xmath88 with @xmath2 fixed and the second item asserts that @xmath557 , where @xmath558 is independent of @xmath88 , provided that @xmath555 is small enough
. more generally , let @xmath92 and @xmath559 be non - negative functions of @xmath88 and some ( numerical ) variables .
we write @xmath560 , when @xmath561 pointwise , where @xmath558 is independent of @xmath88 .
the notation @xmath562 will be used when @xmath563 pointwise as functions of other variables , with @xmath564 depending on @xmath88 and possibly becoming arbitrarily large .
furthermore , we denote by @xmath565 the ball of radius @xmath424 with respect to @xmath88 centered at @xmath13 .
[ lemma : local ] let @xmath88 be a coordinate system near @xmath13 such that @xmath555 is small . then for every sufficiently small @xmath422 ( depending on @xmath88 ) , there exists a one - periodic hamiltonian @xmath415 on @xmath565 such that 1 .
the time one - flow @xmath566 of @xmath80 is @xmath2 , 2 .
@xmath13 is an isolated critical point of @xmath415 and @xmath567 , 3 .
@xmath568 , 4 .
the following estimates hold pointwise near @xmath13 : @xmath569 and @xmath570 where the dot denotes the time derivative of a vector field
. note that in ( iv ) we could have written @xmath548 in place of @xmath571 by ( gf2 ) .
the important point here is that @xmath548 and @xmath572 can be made arbitrarily small by choosing an appropriate coordinate system @xmath88 .
then , shrinking the domain of @xmath80 , we can also make the right hand sides in the estimates ( iii ) and ( iv ) arbitrarily small .
a proof of lemma [ lemma : local ] can be extracted from @xcite .
however , to make our proof of theorem [ thm : main ] self - contained , we provide a detailed argument . deferring this to section [ sec : gf ]
, we proceed with the proof of proposition [ prop1 ] .
[ sec : step4 ] let @xmath80 be a hamiltonian on a neighborhood of @xmath13 , such that ( i)(iv ) of lemma [ lemma : local ] are satisfied and @xmath555 is small .
our first goal is to prove that @xmath80 meets requirements ( k1 ) and ( k2 ) . to establish ( k1 ) , consider the loop @xmath575 , where @xmath576 .
thus , @xmath577 note that @xmath578 , i.e. , @xmath579 with @xmath112 is a loop of hamiltonian symplectomorphisms near @xmath13 .
we denote this loop by @xmath580 .
the @xmath45-th iteration @xmath581 of @xmath580 is simply @xmath579 with @xmath48 .
let @xmath583 be the time - average of @xmath415 , i.e. , @xmath584 a straightforward calculation utilizing lemma [ lemma : local ] and and detailed in section [ sec : bark - k ] shows that the requirements of lemma [ lemma : lfh - lmh ] are met , for any fixed @xmath45 , by the pair @xmath585 , provided that @xmath555 is small enough . in other words , these requirements are satisfied when @xmath2 is @xmath541-close to the identity and @xmath583 is taken as the reference function in lemma [ lemma : lfh - lmh ] ( denoted there by @xmath399 ) . in particular , @xmath13 is an isolated @xmath505-periodic orbit .
we set , @xmath586 , where @xmath505 is as in .
next note that @xmath589 for any @xmath590 . by ( lf4 ) , @xmath591 as long as @xmath13 is an isolated one - periodic orbit of @xmath71 . applying this identity to @xmath586
, we conclude from that @xmath592 in particular , since @xmath593 , @xmath594 combining and , we conclude that @xmath393 .
a different proof of this fact , relying on the properties of the mean conley zehnder index ( see @xcite ) , can be found in ( * ? ? ?
* section 5.2 ) .
furthermore , when @xmath2 is sufficiently @xmath541-close to the identity , the requirements of lemma [ lemma : lfh - lmh ] with @xmath60 and the hamiltonians @xmath80 and @xmath399 as in lemma [ lemma : local ] are obviously met due to ( gf2 ) , , and ( iv ) .
thus , by lemma [ lemma : lfh - lmh](b ) , the function @xmath415 has strict local maximum at @xmath13 .
this proves ( k1 ) . applying this construction to a sequence of symplectic bases @xmath483 in @xmath484 such that @xmath598
, we obtain a sequence of hamiltonians @xmath228 , meeting requirements ( k1 ) and ( k2 ) , and also the loops @xmath465 .
we emphasize that @xmath228 and @xmath599 have so far been defined only on a neighborhood of @xmath13 .
the goal of this auxiliary section , which is included for the sake of completeness , is to show that , as stated above , the pair @xmath585 satisfies the hypotheses of lemma [ lemma : lfh - lmh ] with @xmath45 fixed . to this end ,
note first that by lemma [ lemma : local](iv ) , we have @xmath600 pointwise near @xmath13 .
( here the coordinate system @xmath88 is suppressed in the notation . )
let us integrate the first of these inequalities with respect to @xmath17 over @xmath44 . then , since @xmath399 is independent of time , we have , again pointwise near @xmath13 , @xmath601 with @xmath602 .
thus , @xmath603 and , as a consequence , @xmath604 then @xmath605 likewise , @xmath606 therefore , @xmath607 where @xmath608 and all inequalities are pointwise .
recall now that we can make @xmath609 arbitrarily small ( with @xmath45 fixed ) by making a suitable choice of @xmath88 and then requiring @xmath422 to be sufficiently small .
it follows that we can also make @xmath191 arbitrarily small . in the same vein
, the left hand side of can be made arbitrarily small .
furthermore , since @xmath13 is an isolated critical point of @xmath399 , it is also an isolated critical point of @xmath583 . therefore , the pair @xmath585 satisfies the hypotheses of lemma [ lemma : lfh - lmh ] . as has been pointed out in section [ sec : step4 ] , a pair of function satisfying the hypotheses of lemma [ lemma : local ] also satisfies the hypotheses of lemma [ lemma : lfh - lmh ] .
we have shown that @xmath585 satisfies the conditions of lemma [ lemma : lfh - lmh ] whenever @xmath416 meets the requirements of lemma [ lemma : local ] .
moreover , by arguing as in this section , it is not hard to show that @xmath585 satisfies the conditions of lemma [ lemma : local ] ( and hence of lemma [ lemma : lfh - lmh ] ) once @xmath416 does .
we omit this ( straightforward ) calculation , for it is never used in the proof . here
@xmath228 is the sequence of hamiltonians constructed in section [ sec : step4 ] using lemma [ lemma : local ] .
we prove lemma [ lemma : special ] in section [ sec : gf ] along with lemma [ lemma : local ] . at this point
, we only note that , as will become clear in section [ sec : gf ] , the linearized flow @xmath610 is completely determined by @xmath2 and the basis @xmath483 . in particular , the linearization is independent of the extension of @xmath483 to a canonical coordinate system @xmath517 near @xmath13 .
( here , we use a convention similar to that of section [ sec : step3 ] for canonical coordinate systems : a symplectic basis is divided into two groups of @xmath9 vectors spanning lagrangian subspaces and this division is a part of the structure of a symplectic basis . ) let us now extend the loops @xmath465 and the hamiltonians @xmath228 to @xmath1 so that the remaining part of requirement ( k3 ) is met : the loop @xmath486 is contractible to @xmath195 in the class of loops with identity linearization at @xmath13 .
recall that the maslov index of the loop @xmath465 is zero , as is shown in section [ sec : step4 ] .
hence , by lemma [ lemma : loop - ext ] , each of these loops extends to a loop of hamiltonian diffeomorphisms of @xmath1 , contractible in the class of loops fixing @xmath13 .
let us fix such an extension for @xmath613 .
for the sake of simplicity we denote this extension by @xmath613 again .
consider now the loop @xmath614 .
then @xmath615 . hence , by lemma
[ lemma : loop - ext ] and remark [ rmk : loop - ext ] , @xmath616 extends to a loop of hamiltonian diffeomorphisms of @xmath1 , contractible in the class of loops with identity linearization at @xmath13 . keeping the notation @xmath616 for this extension , we set @xmath617 .
it is clear that @xmath465 is contractible in the class of loops with identity linearization at @xmath13 .
[ lemma : la2 ] let @xmath543 be a linear symplectic map of a finite dimensional symplectic vector space @xmath29 .
assume that all eigenvalues of @xmath544 are equal to one .
then @xmath33 can be decomposed as a direct sum of two lagrangian subspaces @xmath618 and @xmath619 with @xmath620 .
moreover , by a suitable choice of @xmath621 preserving the subspaces @xmath618 and @xmath619 , the map @xmath622 can be made arbitrarily close to the identity . in the former case
, we decompose @xmath33 as @xmath627 , where the superscript @xmath22 denotes the symplectic orthogonal .
it is easy to see that this decomposition is preserved by @xmath544 and @xmath628 .
now the assertion follows from the induction hypothesis applied to @xmath629 . in the latter case ,
pick a symplectic subspace @xmath626 complementary to @xmath625 in @xmath630 and an isotropic subspace @xmath358 complementary to @xmath630 in @xmath33 .
( we are assuming at the moment that @xmath631 , i.e. , @xmath618 is not lagrangian . )
thus , @xmath632 and @xmath633 .
furthermore , @xmath80 and @xmath630 are preserved by @xmath544 ; the spaces @xmath626 and @xmath358 can be canonically identified with @xmath634 and @xmath635 , respectively ; and @xmath636 .
note that @xmath544 induces a symplectic linear map @xmath637 with all eigenvalues equal to one and the identity map @xmath638 on @xmath639 .
hence , using the decomposition @xmath640 we can write @xmath544 in the block upper - triangular form @xmath641 where @xmath642 and @xmath643 and @xmath644 .
( there are relations between these operators , resulting from the fact that @xmath544 is symplectic . )
consider a block - diagonal symplectic linear transformation of the form @xmath645 where @xmath646 is symplectic , @xmath647 is invertible , and we have identified @xmath358 with @xmath648
. then @xmath649 by the induction assumption , there exists a decomposition @xmath650 , where @xmath651 , and transformations @xmath652 preserving this decompositions and making @xmath653 arbitrarily close to @xmath654 .
set @xmath655 and @xmath656 .
then @xmath620 and the decomposition @xmath657 is preserved by @xmath658 .
furthermore , noticing that @xmath659 is close to zero when @xmath660 is close to zero , we can pick @xmath660 to make the off - diagonal entries in @xmath544 arbitrarily small . with this choice of @xmath658 ,
the map @xmath622 is close to @xmath661 if @xmath662 is close to @xmath654 . when @xmath663 is lagrangian ( i.e. , @xmath664 ) , no induction reasoning is needed .
we simply set @xmath665 and let @xmath666 be an arbitrary complementary lagrangian subspace .
then the map @xmath544 is decomposed as a two - by - two block upper - triangular matrix , and , similarly to the argument above , @xmath660 is chosen to make the off - diagonal block arbitrarily small . indeed , @xmath13 is a symplectically degenerate maximum of @xmath825 as is pointed out in example [ ex : sympl - deg ] .
the hamiltonians @xmath825 and @xmath5 have the same time-@xmath45 flow and there is a natural one - to - one correspondence between ( contractible ) @xmath45-periodic orbits of the hamiltonians , for @xmath826 with @xmath827 .
due to our normalization of @xmath825 , the corresponding @xmath45-periodic orbits of @xmath825 and @xmath5 have equal actions and , in particular , @xmath828 has the same action spectrum and action filtration in the floer complex as @xmath71 ; see section [ sec : loops ] . as a consequence , @xmath829 thus , the proposition holds for @xmath5 if ( and only if ) it holds for @xmath825 .
furthermore , when @xmath5 is replaced by @xmath825 , the loops @xmath830 get replaced by the loops @xmath831 which have the identity linearization at @xmath13 by ( k3 ) . with these observations in mind , we establish the proposition by using the squeezing method of @xcite .
namely , closely following @xcite , we construct functions @xmath833 such that @xmath834 ( see fig . 1 ) and such that the map @xmath835 in the filtered floer homology for the interval @xmath836 induced by a monotone homotopy from @xmath837 to @xmath838 is non - zero .
this map factors as @xmath839 and , therefore , @xmath840 as required .
the hamiltonian @xmath837 depends only on @xmath5 and @xmath841 . outside a ball @xmath739 of radius @xmath736 , centered at @xmath13 ,
the function @xmath837 is constant and equal to @xmath842 .
( here the distance is taken with respect to some fixed metric compatible with @xmath22 . ) within @xmath739 , the hamiltonian @xmath837 is a function of the distance to @xmath13 , equal to @xmath843 when the distance is small , dropping to some constant @xmath844 , and then increasing to @xmath842 near the boundary of @xmath739 .
the period @xmath45 is required to be large enough , i.e. , @xmath845 , where @xmath468 is determined by @xmath837 .
the function @xmath838 and the constant @xmath846 depend on @xmath845 .
the condition that @xmath13 is a symplectically degenerate maximum of @xmath5 is used in the construction of @xmath838 and also in proving that @xmath847 .
the function @xmath838 is constructed as follows .
pick @xmath480 so that @xmath848 is small .
( here , as above , @xmath849 is normalized by @xmath850 . )
there exists a bump function @xmath851 , supported near @xmath13 , with non - degenerate maximum at @xmath13 and @xmath852 and such that @xmath853 is also small . then @xmath854 . setting @xmath838 to be the hamiltonian generating the flow @xmath855 , normalized by @xmath856
, we note that @xmath857 .
hence , @xmath834 .
the hamiltonian @xmath838 has the same filtered floer homology as @xmath399 , and we show that @xmath858 using ( k3 )
. in the standard canonical coordinates @xmath741 on @xmath297 , set @xmath859 .
all orbits of @xmath860 are closed and have period @xmath861 .
fix a ball @xmath862 of radius @xmath422 , centered at the origin @xmath13 .
consider a rotationally symmetric function @xmath399 on @xmath297 supported in @xmath863 .
the function @xmath399 depends only on the distance to @xmath13 and it will be convenient in our analysis to also view @xmath399 as a function of @xmath864 .
assume , in addition , that @xmath399 has the following properties ( see fig .
2 ) : * @xmath399 is decreasing as a function of @xmath864 ; * @xmath865 and @xmath866 is increasing on some closed ball @xmath867 ; and * on the shell @xmath868 , * * @xmath399 is concave , i.e. , @xmath869 , on @xmath870 $ ] , where @xmath871 , * * @xmath399 is convex ( @xmath872 ) on @xmath873 $ ] , where @xmath874 , * * @xmath399 has constant slope ( @xmath875 ) on the interval @xmath876 $ ] , where @xmath877 is irrational .
we will refer to @xmath399 as a _
standard bump function _ on @xmath297 and we will call @xmath878 and @xmath879 and @xmath424 and other constants from the construction of @xmath399 the parameters of @xmath399 . the trivial one - periodic orbits of @xmath399 ( i.e. , its critical points ) are either contained in @xmath880 or in the complement to @xmath881 .
the orbits from the first group form a closed ball ( possibly of zero radius ) centered at @xmath13 and have action @xmath353 ; the orbits from the second group are exactly the points where @xmath882 .
let @xmath885 be the standard @xmath161-small ( periodic in time ) perturbation of @xmath399 as the ones considered in , e.g. , @xcite , and still supported in @xmath863 .
for such a perturbation each sphere filled in by one - periodic orbits of @xmath399 breaks down into @xmath21 non - degenerate orbits . within @xmath886
, we may assume that @xmath885 is still autonomous and rotationally symmetric and @xmath887 .
( hence the only one - periodic orbit of @xmath885 in @xmath886 is the trivial orbit @xmath13 of conley - zehnder index @xmath9 . )
as is well known , the filtered floer complex of @xmath885 and the filtered floer homology of @xmath885 ( and @xmath399 ) are still defined , say , for any positive interval of actions @xmath888 even though @xmath297 is not compact ; see , e.g. , @xcite .
here we adopt the conventions of @xcite .
it is not hard to see that @xmath885 has no one - periodic orbits of index @xmath7 .
furthermore , it has exactly two one - periodic orbits of index @xmath9 .
one of these is the constant orbit @xmath13 .
the second orbit @xmath140 arises from the sphere of periodic orbits farthest from the origin .
this sphere has radius @xmath894 and @xmath895 , up to an error of order @xmath896 . by
, @xmath140 is outside the range of action .
finally , @xmath885 has two one - periodic orbits of index @xmath892 , but only one of them , @xmath178 , has action in @xmath893 .
( the action of the second orbit is approximately equal to @xmath897 . )
the orbit @xmath178 arises from the sphere of periodic orbits closest to the origin .
this is the sphere of radius @xmath898 and @xmath899 , up to an error of order @xmath896 .
( see @xcite or section [ sec : k+ ] , where we analyze in detail periodic orbits of a function similar to @xmath399 . )
* @xmath885 has no one - periodic orbits of index @xmath7 ; * @xmath13 is the only one - periodic orbit of @xmath885 with index @xmath9 and action in @xmath893 ; * @xmath178 is the only one - periodic orbit of @xmath885 with index @xmath892 and action in @xmath893 ; * the connecting map from the long exact sequence @xmath902 is an isomorphism sending @xmath903 $ ] to @xmath212 $ ] , and hence @xmath904 . let @xmath89 be a small neighborhood of @xmath173 .
fixing a canonical coordinate system on @xmath89 , denote the open ball of radius @xmath422 in @xmath89 , centered at @xmath13 , by @xmath423 and let @xmath905 be the boundary of @xmath423 .
we define a bump function @xmath399 on @xmath1 exactly as for @xmath297 by using the coordinate system in @xmath89 .
furthermore , since @xmath1 is compact , now we need not assume that @xmath399 is supported in @xmath863 . instead we just require @xmath399 to be constant outside @xmath863 . in other words , we allow @xmath399 to be shifted up and down .
the description of periodic orbits of @xmath885 and the floer homology of @xmath399 given in section [ sec : bump - r2n ] extends word - for - word to this case , provided that @xmath863 is sufficiently small ( e.g. , displaceable ) and the variation @xmath906 is sufficiently large .
the requirement @xmath891 is replaced by that @xmath907 , where @xmath908 depends only on @xmath863 and goes to zero as @xmath909 ; see , e.g. , @xcite .
( hypothetically , @xmath908 is equal to the displacement energy of @xmath863 , although the estimate we have been able to prove is somewhat weaker . )
the requirement carries over to this case unchanged when @xmath399 is supported in @xmath863 , and is , in general , replaced by @xmath910 [ sec : traj ] let us show that by making @xmath424 sufficiently small , we can ensure that the floer gradient trajectories of @xmath399 from @xmath178 to @xmath13 are close to @xmath13 .
( when @xmath399 is a bump function on @xmath297 , all such trajectories are contained in @xmath911 by the maximum principle . ) to this end , pick a ball @xmath912 contained in @xmath89 and fix once and forever a compatible with @xmath22 almost complex structure @xmath913 on @xmath1 coinciding with the standard complex structure on a neighborhood of @xmath914 .
consider holomorphic curves @xmath96 in @xmath915 with boundary in @xmath916 and such that the part of the boundary of @xmath96 lying in @xmath917 is non - empty .
( then the part of the boundary of @xmath96 in @xmath918 is also non - empty due to the maximum principle . )
denote by @xmath919 the infimum of the areas of such curves @xmath96 .
it is easy to see that @xmath920 remains separated from zero as @xmath909 .
( otherwise we would have @xmath921 for some fixed @xmath422 as is clear from considering the intersections with @xmath915 of holomorphic curves whose areas approach zero . ) in other words , @xmath922 .
replacing @xmath793 by @xmath923 , we see that there exists @xmath924 such that @xmath925 let @xmath399 be an arbitrary bump function @xmath399 such that holds and @xmath926 .
assume that @xmath191 and @xmath927 satisfy .
then for a perturbation @xmath885 of @xmath399 as above and any regular perturbation of @xmath26 of @xmath913 all floer anti - gradient trajectories from @xmath178 to @xmath13 are contained in @xmath928 .
assume the contrary .
then for some @xmath885 close to @xmath399 and for some sequence of regular perturbations @xmath929 , there exists a sequence of connecting trajectories @xmath930 from @xmath178 to @xmath13 , leaving a neighborhood of @xmath931 . observe that the part of @xmath930 contained in @xmath932 is a @xmath933-holomorphic curve . by the compactness theorem , in the limit we have a @xmath913-holomorphic curve @xmath96 in @xmath932 with non - empty boundary in @xmath917 . by the definition of @xmath934 , the area of @xmath96 is greater than @xmath934 .
therefore , the same is true for the part of @xmath930 contained in @xmath932 when @xmath933 is close to @xmath913 .
thus , @xmath935 by .
this is impossible , for @xmath936 [ sec : k+ ] without loss of generality , we may assume that @xmath937 . furthermore , throughout this section we will keep the notation and convention of section [ sec : bump ] . in particular
, we fix a system of canonical coordinates on a neighborhood @xmath89 of @xmath13 and let , as in section [ sec : bump ] , the function @xmath864 on @xmath89 be one half of the square of the distance to @xmath13 with respect to this coordinate system .
* @xmath939 ; * @xmath940 ; * on the shell @xmath868 the function @xmath837 is monotone decreasing , as a function of @xmath864 , and * * @xmath837 is concave ( @xmath941 ) on @xmath942 $ ] , where @xmath871 , * * @xmath837 is convex ( @xmath943 ) on @xmath873 $ ] , where @xmath874 , * * @xmath837 has constant slope ( @xmath944 ) on the interval @xmath945 $ ] , where @xmath877 is irrational ; * @xmath946 on the shell @xmath947 , where the constant @xmath185 is to be specified latter ; * @xmath837 is monotone increasing on the shell @xmath948 , and * * @xmath837 is convex ( @xmath943 ) on @xmath949 $ ] , where @xmath950 , * * @xmath837 is concave ( @xmath941 ) on @xmath951 $ ] , where @xmath952 , * * @xmath837 , as a function of @xmath864 , has constant slope ( @xmath944 ) on the interval @xmath953 $ ] , where @xmath877 is irrational ; * @xmath954 on @xmath955 , with the constant @xmath956 to be specified .
furthermore , we extend @xmath837 to @xmath1 by setting it to be constant and equal to @xmath842 on the complement of @xmath89 . the constant @xmath842 is chosen so that @xmath957 on @xmath1 and @xmath956 .
note that within @xmath739 , the function @xmath837 is a standard bump function of section [ sec : bump ] .
this bump function has variation @xmath958 which , due to the requirement @xmath959 , may be very small . the neighborhood @xmath89 is chosen so that @xmath843 is a strict global maximum of @xmath5 on @xmath89 and @xmath89 is displaceable in @xmath1 by a hamiltonian diffeomorphism .
the values @xmath960 are chosen arbitrarily , with the only restriction that @xmath961 . to pick @xmath424 ,
we fix @xmath191 and also fix a compatible with @xmath22 almost complex structure @xmath913 on @xmath1 coinciding with the standard complex structure on a neighborhood of the closed ball @xmath914 .
the radius @xmath424 is chosen so that @xmath962 where the upper bound @xmath963 is as in section [ sec : traj ] .
the radius @xmath964 is chosen arbitrarily with the only restriction that @xmath965 .
the constants @xmath185 and @xmath842 are picked so that @xmath844 and @xmath939 on @xmath966 $ ] and on @xmath967 ( this may require @xmath185 to be very close to @xmath230 . ) likewise , on the intervals @xmath968 $ ] and @xmath969 $ ] , the behavior of @xmath837 is specified to guarantee that @xmath939 .
finally , we will also impose the condition that @xmath970 [ sec : orbits ] in this section , we analyze the relevant @xmath45-periodic orbits of @xmath837 when @xmath45 is sufficiently large . since @xmath837 is autonomous ,
its @xmath45-periodic orbits can simply be treated as one - periodic orbits of @xmath972 .
furthermore , it is clear that all @xmath45-periodic orbits of @xmath837 outside @xmath89 are trivial .
those in @xmath89 are either trivial or fill in spheres of certain radii . replacing @xmath837 by its standard time - dependent @xmath161-small perturbation @xmath973 as in @xcite and section [ sec : bump ] results in each of these spheres splitting into @xmath21 non - degenerate orbits . here , as in section [ sec : bump ] , we are primarily interested in the orbits with index @xmath7 or @xmath9 or @xmath892 and action in the interval @xmath974 for some small @xmath927
. we will show that these orbits are essentially the same as for a bump function with ( large ) variation @xmath975 .
the perturbation @xmath973 is similar to @xmath885 from section [ sec : bump ] .
we emphasize that @xmath837 is perturbed not only within the shells @xmath976 and @xmath977 where non - trivial periodic orbits are , but also within the ball @xmath978 where @xmath979 . on this ball
, @xmath973 is a monotone decreasing function of @xmath864 with a non - degenerate maximum at @xmath13 equal to @xmath230 .
this function is @xmath161-close to the constant function @xmath837 so that ( for a fixed @xmath45 ) , the function @xmath980 is @xmath161-close to @xmath981 on @xmath978 . in particular ,
the eigenvalues of the hessian @xmath982 are close to zero and the conley
zehnder index of the constant @xmath45-periodic orbit @xmath13 of @xmath973 is @xmath9 . in what follows
, we will always assume that @xmath973 is as close to @xmath837 as necessary .
in the shell @xmath947 and in the complement to @xmath983 we keep @xmath973 constant and equal to @xmath837 .
the trivial @xmath45-periodic orbits of @xmath837 are the points of @xmath886 ( with action @xmath981 ) , the points of @xmath985 ( with action @xmath986 ) , and the points of @xmath987 ( with action @xmath988 ) . here , only the points of @xmath886 have action within the range in question . indeed , @xmath989 by and @xmath990 , for @xmath991 by .
thus , @xmath13 is the only trivial @xmath45-periodic orbit of @xmath973 with action in @xmath974 ; it has index @xmath9 . the first group is formed by the @xmath45-periodic orbits in the shell @xmath992 .
these orbits fill in a finite number of spheres @xmath993 of radii @xmath994 with @xmath995 the orbits on @xmath993 have action @xmath996 .
once the hamiltonian @xmath837 is replaced by @xmath973 , a sphere @xmath993 breaks down into @xmath21 non - degenerate orbits . the conley
zehnder indices of these orbits are @xmath997 as is proved in @xcite .
only one of the orbits in this group has index from @xmath7 to @xmath892 and action in the range @xmath974 .
this is a periodic orbit , denoted by @xmath178 , of index @xmath892 and action @xmath998 ( up to an error of order @xmath999 ) , arising from the sphere @xmath1000 .
the conley
zehnder indices of the remaining orbits are greater than @xmath892 although some of these orbits may have action within the range @xmath974 . the second group is comprised of the @xmath45-periodic orbits in the shell @xmath1001 .
these orbits fill in the spheres @xmath1002 of radii @xmath1003 with @xmath1004 the orbits on @xmath1002 have action @xmath1005 . again , once the hamiltonian @xmath837 is replaced by @xmath973 , each sphere @xmath1002 breaks down into @xmath21 non - degenerate orbits .
the conley
zehnder indices of these orbits are @xmath1006 ; see @xcite . only the orbits arising from @xmath1007 and @xmath1008 can have index @xmath7 or @xmath9 or @xmath892 .
( other spheres give rise to orbits of index greater than @xmath1009 . )
however , the orbits coming from the spheres @xmath1007 and @xmath1008 have action not exceeding @xmath1010 if @xmath973 is close to @xmath837 . by
, these orbits are outside the action range @xmath974 .
the @xmath45-periodic orbits in the shell @xmath1011 are in the third ( possibly empty ) group .
these orbits fill in the spheres @xmath1012 of radii @xmath1013 with @xmath1014 and the orbits on @xmath1012 have action @xmath1015 .
hence , all of these orbits are outside of the range of action @xmath974 .
the fourth group , which may also be empty , is formed by the @xmath45-periodic orbits in the shell @xmath1016 .
these orbits fill in a finite collection of spheres @xmath1017 of radii @xmath1018 such that @xmath1019 and the orbits on @xmath1017 have action @xmath1020 , which can be in the interval @xmath974 .
however , calculating the conley
zehnder indices of the resulting orbits of @xmath973 as in @xcite , it is easy to see that the sphere @xmath1017 brakes down into non - degenerate orbits of @xmath973 with indices @xmath1021 . in particular , all resulting orbits have indices not exceeding @xmath1022 , and none of the orbits has index @xmath7 or @xmath9 or @xmath892 . to summarize , the perturbation @xmath973 has only one @xmath45-periodic orbit of index @xmath9 with action in @xmath974 this is the trivial orbit @xmath13 and only one orbit , namely @xmath178 , of index @xmath892 with action within this range .
the action of @xmath13 is @xmath981 and the action of @xmath178 is @xmath998 up to an error of order @xmath999 .
there are no orbits with index @xmath7 and action in the range @xmath974 .
[ sec : fh - k+ ] as in the previous section , assume that @xmath45 is sufficiently large and @xmath927 is small ( independently of @xmath45 ) .
explicitly , now we require in addition to that @xmath1023 where @xmath1024 is defined in section [ sec : bump - w ] . in this section
we prove [ lemma : fh1 ] under the above assumptions on the function @xmath837 , the period @xmath45 , and @xmath841 and @xmath1025 , we have @xmath1026 with generators @xmath212 $ ] and , respectively , @xmath903 $ ] .
moreover , the connecting map @xmath1027 is an isomorphism . since is satisfied , the results of the previous section apply , and @xmath178 and @xmath13 are the only @xmath45-periodic orbits of @xmath973 of index @xmath7 or @xmath9 on @xmath892 with action in the range @xmath974 .
it is clear that @xmath212 $ ] is the generator of @xmath1028 .
furthermore , @xmath178 is the only @xmath45-periodic orbit of index @xmath892 with action in @xmath1029 and there are no @xmath45-periodic orbits of index @xmath9 with action in this interval .
hence , the homology @xmath1030 , generated by @xmath903 $ ] , is either zero or @xmath135 .
( the former is _ a priori _ possible , for in fact there exists a @xmath45-periodic orbit of index @xmath1031 with action in @xmath1029 . )
to finish the proof of the lemma , it is sufficient now to show that the connecting map is onto or , equivalently , @xmath1032 i.e. , the number of floer anti - gradient trajectories for @xmath1033 from @xmath178 to @xmath13 is odd . within @xmath739 ,
the hamiltonian @xmath972 coincides with a standard bump function @xmath399 whose variation @xmath1034 is greater than @xmath1035 by .
thus , the assumptions of section [ sec : bump - w ] are satisfied , and @xmath1036 .
furthermore , @xmath1033 agrees with @xmath885 on @xmath739 . due to our choice of @xmath424 ,
the condition holds and lemma [ lemma : traj ] is applicable .
therefore , @xmath885 , and hence @xmath1033 , have an odd number of floer anti - gradient trajectories from @xmath178 to @xmath13 contained in @xmath739 . moreover , every floer anti - gradient trajectory for @xmath1033 from @xmath178 to @xmath13 is automatically in @xmath739 .
this is established by arguing exactly as in the proof of lemma [ lemma : traj ] with @xmath399 replaced by @xmath1033 and using again . as a consequence , @xmath1033 and
@xmath399 have the same floer anti - gradient trajectories from @xmath178 to @xmath13 , and the total number of such trajectories is odd .
this concludes the proof of and of the lemma .
recall that the function @xmath837 and the parameter @xmath191 were fixed above , while @xmath45 and @xmath1025 have been variable . at this point
, we also fix a large period @xmath45 meeting the requirement and such that @xmath1037 . then , condition is satisfied if @xmath927 is small , and hence lemma [ lemma : fh1 ] applies . in this section ,
we construct a hamiltonian @xmath1038 , depending on @xmath45 , such that @xmath1039 and @xmath1040 , and the connecting map @xmath1041 is an isomorphism if @xmath846 is sufficiently small .
such that the hamiltonian @xmath1044 generating the flow @xmath1045 has a strict local maximum at @xmath13 and @xmath1046 moreover , the loop @xmath324 has identity linearization at @xmath13 , i.e. , @xmath1047 for all @xmath112 , and is contractible to @xmath195 in the class of loops with identity linearization at @xmath13 .
( see ( k3 ) and section [ sec : prop2-outline ] . )
let @xmath1048 be a homotopy from @xmath324 to the identity such that @xmath1049 and let @xmath1050 be the one - periodic hamiltonian generating @xmath1051 and normalized by @xmath1052 .
the condition @xmath1053 is equivalent to that @xmath1054 . as usual
, we normalize @xmath80 by requiring that @xmath567 or , equivalently , by @xmath1055 . without loss of generality
, we may also assume that @xmath1056 , where @xmath33 is the domain of the coordinate system @xmath88 and @xmath978 is the ball from the construction of @xmath1057 ; see section [ sec : k+ ] .
note that this ball is taken with respect to the original metric and is not related to @xmath88 .
let @xmath399 be a bump function , `` centered '' at @xmath13 , with respect to the coordinate system @xmath88 .
as in section [ sec : bump - w ] , we do not require @xmath399 to be supported in @xmath33 , but only constant outside @xmath33 . thus , @xmath1058 on @xmath1059 .
we may assume that @xmath1060 .
it is also clear that @xmath399 can be chosen so that the variation of @xmath1071 , equal to @xmath1072 , is much larger than @xmath1073 .
hence , as shown in section [ sec : bump ] , and we have the isomorphism @xmath1074 provided that @xmath846 is sufficiently small . finally note that @xmath202 for all @xmath73 , for @xmath1075 .
therefore , the functions @xmath1066 have equal filtered floer homology for any period @xmath45 . in particular , the filtered floer homology of @xmath1076 and of @xmath1077 are the same and the latter is identical to the ( one - periodic ) filtered floer homology of @xmath1071 .
thus , we obtain the desired isomorphism . by construction , @xmath1078 . a monotone decreasing homotopy from @xmath837 to @xmath838 induces maps of filtered floer homology , which , as is well known , commute with the maps from the long exact sequence . in particular , combining the monotone homotopy maps with the connecting maps for @xmath837 and for @xmath838 , we obtain the following commutative diagram : @xmath1079 to prove the theorem , it is sufficient to show that the right vertical arrow @xmath658 in the diagram , i.e. , the homotopy map @xmath1080 is an isomorphism .
indeed , the rows of are isomorphisms , and hence the left vertical arrow is an isomorphism whenever @xmath658 is an isomorphism . since @xmath834 by , the left vertical arrow factors as @xmath1081 and , as a consequence , the middle group is non - zero as desired . to show that @xmath658 is an isomorphism ,
first observe that since @xmath1070 for all @xmath73 and @xmath1082 and @xmath1068 , the diagram @xmath1083_\psi\ar[rd ] & \\ { \hf_n^{(tc-\delta_t,\,tc+\delta_t)}\big(h_-^{(t)}\big ) } \ar[r]^{\cong } & { \hf_n^{(tc-\delta_t,\,tc+\delta_t)}\big(f^{(t)}\big ) } } \ ] ] is commutative , where the horizontal isomorphism is induced by the isospectral homotopy @xmath1066 and the remaining two arrows are monotone homotopy maps .
( see section [ sec : homotopy ] and , in particular , . )
recall also that @xmath837 and @xmath399 are autonomous .
it remains to prove that the diagonal arrow , which can be identified with @xmath1084 is an isomorphism .
consider a @xmath161-small autonomous perturbation @xmath1085 of @xmath837 such that @xmath1086 is negative definite and @xmath1087 .
( it is straightforward to construct @xmath1088 by modifying @xmath837 on a neighborhood of @xmath886 . )
then @xmath1089 and @xmath1090 can be made arbitrarily small , for @xmath1091 .
we take @xmath1088 such that @xmath1092 is @xmath161-close to @xmath972 , and , in particular , @xmath1093 is small . essentially by definition ,
the filtered floer homology of @xmath1092 is isomorphic to the filtered floer homology of @xmath972 , and it suffices to show that @xmath1094 is an isomorphism for some small @xmath846 independent of the choice of @xmath1088 . recall that @xmath981 is an isolated action value of @xmath1071 and @xmath1095 , and hence of @xmath1096 , for a generic choice of parameters of these functions .
fix @xmath846 meeting the requirement and such that @xmath981 is the only action value of these hamiltonians in @xmath1097 . consider the linear decreasing homotopy @xmath1098 .
since both of the hessians @xmath1086 and @xmath410 are negative definite , @xmath1099 is also negative definite .
thus , @xmath13 is a non - degenerate critical point of @xmath1100 for all @xmath194 $ ] with @xmath1101 .
furthermore , by and since @xmath1102 is small , @xmath1103 . as a consequence , @xmath13 is a uniformly isolated one - periodic orbit of @xmath1104 ; see , e.g. , @xcite or section [ sec : lfh - lmh ] . by lemma [ lemma : non - zero ] ,
the homotopy map is non - zero , and hence an isomorphism .
this concludes the proof of proposition [ prop2 ] and of theorem [ thm : main ] . a slightly different proof of the proposition , although based on the same ideas as the present argument and following the same line of reasoning , can be found in ( * ? ? ?
* section 5 ) .
m. poniak , floer homology , novikov rings and clean intersections , in _ northern california symplectic geometry seminar _ , 119181 , amer .
ser . 2 , * 196 * , amer . math .
soc . , providence , ri , 1999 .
a. weinstein , _ lectures on symplectic manifolds .
_ expository lectures from the cbms regional conference held at the university of north carolina , march 812 , 1976 .
regional conference series in mathematics , no .
american mathematical society , providence , r.i . , 1977 .
in this section , we recall the definition of a generating function on @xmath297 and set the stage for proving lemma [ lemma : local ] . the material reviewed here is absolutely standard
it goes back to poincar and we refer the reader to ( * ? ? ?
* appendix 9 ) and @xcite for a more detailed discussion of generating functions .
let us identify @xmath297 with the lagrangian diagonal @xmath667 via the projection to the first factor , where @xmath668 is equipped with the symplectic structure @xmath669 , and fix a lagrangian complement @xmath358 to @xmath670 .
thus , @xmath668 can now be treated as @xmath671 .
let @xmath2 be a hamiltonian diffeomorphism defined on a neighborhood of the origin @xmath13 in @xmath297 and such that @xmath672 is sufficiently small .
then the graph @xmath373 of @xmath2 is close to @xmath670 , and hence @xmath373 can be viewed as the graph in @xmath671 of an exact form @xmath673 near @xmath674 .
( we normalize @xmath399 by @xmath675 . ) the function @xmath399 , called the generating function of @xmath2 , has the following properties : for instance , it is clear that the critical points of @xmath399 are in one - to - one correspondence with the fixed points of @xmath2 .
if @xmath13 ( the origin ) is an isolated fixed point of @xmath2 , the origin is also an isolated critical point of @xmath399 .
hence , ( gf1@xmath678 ) holds .
the second property of @xmath399 , ( gf2@xmath678 ) , is also easy to check ; see the references above .
the function @xmath399 depends on the choice of the lagrangian complement @xmath358 to @xmath670 . to be specific ,
we take as @xmath358 the linear subspace of vectors of the form @xmath679 in @xmath668 , where @xmath680 and @xmath681 are the standard canonical coordinates on @xmath297 , i.e. , @xmath682 . in the setting of section [ sec : step3 ] ,
let @xmath88 be a coordinate system near @xmath173 .
using @xmath88 , we identify a neighborhood of @xmath13 in @xmath1 with a neighborhood of the origin in @xmath297 , keeping the notation @xmath13 for the origin . with this identification , @xmath2 defined near @xmath173 turns into a hamiltonian diffeomorphism @xmath683 defined near the origin @xmath684 . by definition , @xmath685 .
abusing notation , we denote the resulting hamiltonian diffeomorphism @xmath683 near @xmath684 by @xmath2 again . by our background assumptions
, @xmath13 is an isolated fixed point of @xmath2 , and thus ( gf1 ) and ( gf2 ) follow immediately from ( gf1@xmath678 ) and ( gf2@xmath678 ) , respectively .
[ lemma : local2 ] let @xmath2 be a hamiltonian diffeomorphism of a neighborhood of the origin @xmath686 .
assume that @xmath13 is an isolated fixed point of @xmath2 and @xmath672 is so small that the generating function @xmath399 is defined .
then for every sufficiently small @xmath422 ( depending on @xmath2 ) , there exists a one - periodic hamiltonian @xmath415 on the ball @xmath423 of radius @xmath424 centered at @xmath13 such that 1 .
the time one - flow @xmath566 of @xmath80 is @xmath2 , 2 .
@xmath13 is an isolated critical point of @xmath415 and @xmath687 for all @xmath112 , 3 .
@xmath688 , 4 . the upper bounds @xmath689 and @xmath690 hold pointwise near @xmath13 .
the notation used here is similar to that of section [ sec : step3 ] .
for instance , should be read as that its left hand side is pointwise bounded from above by @xmath691 , where @xmath692 is independent of @xmath2 and @xmath693 depends on @xmath2 and can be arbitrarily large .
the proof of the lemma is organized as follows .
first we consider the time - dependent hamiltonian @xmath694 generating the flow @xmath695 given by the family of generating functions @xmath696 , @xmath596 $ ] , and verify ( i)(iv ) for @xmath694 .
the time - one flow of @xmath694 is @xmath2 .
however , in general , the hamiltonian @xmath694 is _ not _ periodic in time .
hence , as the next step , we modify @xmath694 to obtain the required periodic hamiltonian @xmath80 and then again check that the new hamiltonian @xmath80 satisfies ( i)(iv )
. consider the family of generating functions @xmath696 with @xmath697 $ ] .
the family of graphs @xmath698 of @xmath699 in @xmath671 beginning with @xmath700 and ending with @xmath701 can be viewed as a family of graphs of hamiltonian diffeomorphisms @xmath695 near @xmath13 with @xmath702 and @xmath703 .
thus , @xmath695 is a time - dependent hamiltonian flow with the time - one map @xmath2 , defined near @xmath13 .
let @xmath704 be the hamiltonian generating this flow , normalized by @xmath705 .
let @xmath709 . then , as is well known and can be checked by a simple calculation , we have @xmath710 differentiating with respect to time , we obtain the following expression for the hamiltonian @xmath706 ( cf .
@xcite ) : @xmath711 where @xmath712 here @xmath713 stands for the matrix of partial derivatives @xmath714 and , similarly , @xmath715 is the matrix @xmath716 . clearly , @xmath722 uniformly in @xmath17 . in particular , @xmath717 is indeed a diffeomorphism , fixes @xmath13 , and is , moreover , @xmath300-close to the identity when @xmath2 is close to @xmath195 .
furthermore , @xmath723 pointwise near @xmath13 and uniformly in @xmath17 . hence , @xmath105 is invertible near @xmath13 .
since @xmath13 is an isolated critical point of @xmath399 by ( gf1@xmath678 ) , @xmath13 is also an isolated zero of @xmath429 , and thus an isolated zero of @xmath724 .
this gives a direct proof of ( ii ) for @xmath694 .
since @xmath725 , the linearization at @xmath13 yields @xmath726 here , @xmath727 and @xmath728 are the linear hamiltonian vector fields on @xmath729 with quadratic hamiltonians @xmath730 and , respectively , @xmath410 .
furthermore , it is easy to see from and that @xmath731 and @xmath732 are both close to @xmath733 , with error @xmath734 , and hence are small . combining these observations ,
we conclude that @xmath735 proving ( iii ) for @xmath694 .
[ sec : proofkf1 ] turning to the proof of ( iv ) for @xmath694 , observe that for every small @xmath736 there exists @xmath422 such that @xmath737 and @xmath738 are in @xmath739 for all @xmath740 $ ] and all @xmath741 in @xmath423 .
furthermore , it is clear that @xmath742 to establish the upper bound of ( iv ) , let us first show that @xmath743 for every @xmath721 in @xmath423 and all @xmath740 $ ] .
we have @xmath744 where in the last inequality we used the fact that , by , @xmath745 thus , we only need to show that @xmath746 by , we have @xmath747 since @xmath748 and @xmath725 , the first summand is obviously @xmath749 .
( when @xmath750 , both @xmath751 and @xmath752 are , by , in the ball of radius @xmath749 . )
the second summand is bounded as @xmath753 here , the first and the last factors are @xmath754 , and hence bounded from above by a constant independent of @xmath2 , when @xmath2 is sufficiently close to @xmath195 .
the middle factor is @xmath755 , for @xmath756 and , as a consequence , @xmath757 thus , the second summand is @xmath755 , which completes the proof of .
then @xmath758 by , the second term is bounded as @xmath759 and the first term is @xmath760 by .
this proves the pointwise estimate at @xmath761 in place of @xmath721 .
since @xmath717 is a diffeomorphism fixing @xmath13 , the upper bound in its original form ( at @xmath721 ) follows . arguing exactly as in the proof of
, it is easy to show that @xmath762 and , as a consequence , @xmath763\cdot\| x_{f}(z)\|\leq \|x_f(\kappa^t(z))\|.\ ] ] therefore , to establish for @xmath694 , it is sufficient to prove the upper bound @xmath764 for @xmath750 . differentiating with respect to @xmath17 and setting @xmath751
, we obtain @xmath765 to prove , we will estimate all three terms in this identity . as a straightforward calculation shows , @xmath766 thus , @xmath767 and @xmath768 furthermore , @xmath769 as follows from the definition of @xmath717 and . using again the inequality @xmath770 , we see that @xmath771 where @xmath558 is independent of @xmath2 , when @xmath2 is sufficiently close to @xmath195 . then ,
since @xmath772 , we have @xmath773 and @xmath774 for @xmath775 . fix a monotone increasing function @xmath776 \to [ 0,\ , 1]$ ] such that @xmath777 when @xmath17 is near @xmath778 and @xmath779 when @xmath17 is near @xmath780 .
this function is independent of @xmath2 , and hence @xmath781 and @xmath782 are bounded from above by constants independent of @xmath2 .
the hamiltonian @xmath80 is the one generating the flow @xmath783 explicitly , since @xmath399 is autonomous , @xmath784 it is clear that @xmath785 when @xmath17 is close to @xmath778 and @xmath780 and hence @xmath80 can be viewed as a hamiltonian one - periodic in @xmath17 . also
, @xmath786 , i.e. , requirement ( i ) is satisfied .
as has been pointed out , the second condition , ( ii ) , follows from ( gf1@xmath678 ) and ( iv ) which is proved below . the hamiltonian vector field of @xmath80 is @xmath789
where @xmath790 .
since @xmath399 is autonomous , @xmath791 . in particular , @xmath792 when @xmath721 is close to @xmath13 . also note that when @xmath750 , the point @xmath790 is in @xmath739 , where the radius @xmath793 satisfies .
differentiating with respect to @xmath17 , we obtain @xmath795(w ) .
\end{split}\ ] ] arguing as in section [ eq : kf1k ] , we see that the norm of the first term in this sum is @xmath796 .
similarly , the same holds for the second term by for @xmath694 .
( in both cases we use to relate @xmath797 and @xmath798 and also the fact that @xmath799 when @xmath750 . )
to estimate the third term , it is sufficient to show that @xmath800(w)\|=\big ( o(\|d^2 f_p\|)+o_\varphi(r)\big)\|x_f(w)\|,\ ] ] for then , by , this term is @xmath796 when @xmath750 . to prove , observe that @xmath801(w)\big\|&=\big\|[x_f , x_{{\tilde{k}}_t}-x_f](w)\big\|\\ & \leq \alpha(w)\|x_f(w)\|+\beta(w)\|x_{{\tilde{k}}_t}(w)-x_f(w)\| , \end{split}\ ] ] where the functions @xmath802 and @xmath803 are bounded from above by the partial derivatives of @xmath804 and , respectively , @xmath429 at @xmath805 .
hence , both of these functions are @xmath806 and follows from for @xmath694 .
let @xmath547 and let @xmath657 be the decomposition of @xmath807 from lemma [ lemma : la2 ] . pick a linear canonical coordinate system @xmath741 on @xmath484 , which is compatible with the decomposition , i.e. , such that the @xmath178-coordinates span @xmath618 and the @xmath140-coordinates span @xmath619 .
by lemma [ lemma : la2 ] , we can do this so that @xmath808 is small in this coordinate system , and thus @xmath2 is given by the generating function @xmath399 .
denote by @xmath809 the hessian of @xmath399 at @xmath13 and let @xmath810 be the linear hamiltonian vector field of @xmath809 on @xmath33 .
linearizing at @xmath13 , we see that @xmath544 and @xmath809 are related via the equation @xmath811 here @xmath812 is obtained from @xmath544 by replacing its @xmath140-component by the identity map , i.e. , @xmath813 in the decomposition @xmath657 , where @xmath814 .
note that uniquely determines @xmath810 .
it is clear from and that @xmath810 and @xmath815 depend only on the decomposition @xmath657 .
hence , any other coordinate system compatible with this decomposition will give rise to the same quadratic form @xmath809 and the same maps @xmath815 .
due to , the linearization @xmath818 is equal to @xmath819 hence , @xmath820 also depends only on the decomposition , but not on the coordinate system as long as the latter is compatible with the decomposition . in other words ,
all such coordinate systems result in the flows @xmath85 with linearization @xmath820 .
lemma [ lemma : special ] follows now from lemma [ lemma : la2 ] , which guarantees that there exist symplectic bases @xmath97 ( or , equivalently , linear canonical coordinate systems ) compatible with @xmath657 and making @xmath821 arbitrarily small .
recall that @xmath822 due to ( k1 ) and that all eigenvalues of @xmath544 are equal to one . combining these facts with the normal forms of quadratic hamiltonians ( see ( * ? ? ?
* appendix 7 ) and @xcite ) , it is not hard to show that @xmath823 in some symplectic basis compatible with the decomposition @xmath824 . then it is straightforward to write down an explicit expression for @xmath815 and @xmath820 . this
, however , does not lead to any simplification in the line of reasoning used here , for the required result readily follows from and . | we prove the conley conjecture for a closed symplectically aspherical symplectic manifold : a hamiltonian diffeomorphism of a such a manifold has infinitely many periodic points .
more precisely , we show that a hamiltonian diffeomorphism with finitely many fixed points has simple periodic points of arbitrarily large period .
this theorem generalizes , for instance , a recent result of hingston establishing the conley conjecture for tori . |
the experimental technique of trapped one - dimensional atomic gases @xcite has now become a fundamental tool for revealing nontrivial effects in quantum many - body systems @xcite . for the interacting bose gas in one dimension ( 1d ) , the first set of exact results goes back to the pioneering work of girardeau @xcite on the impenetrable bose gas where the strong interacting limit is considered .
the 1d bose gas interacting with the delta - function potentials , i.e. the lieb - liniger ( ll ) model , gives a solvable model for interacting bosons in 1d @xcite , where it is integrable even when the interaction parameter is generic .
for the impenetrable bose gas which corresponds to the tonks - girardeau ( tg ) limit , i.e. the strong interacting limit of the ll model , the one - body reduced density matrix is derived and successfully expressed in terms of the determinant of a fredholm operator @xcite .
the exact result is followed by several important developments in mathematical physics @xcite .
furthermore , dynamical correlation functions of the ll model are now systematically derived @xcite .
soliton - like localized excitations in a periodic 1d bose gas have attracted much interest recently and have been studied theoretically @xcite . here
we remark that dark solitons are created experimentally in cold atoms @xcite , for instance , by the phase - imprinting method @xcite ( see also @xcite ) .
localized quantum states are important for investigating dynamical responses of interacting quantum systems .
quantum dark solitons in confining potentials are studied by semiclassical quantization @xcite , and those in thermal equilibrium of a quasi-1d bose gas by generating classical field ensembles @xcite .
however , it is not clear even at zero temperature how we can construct quantum states associated with dark solitons in the many - body system of the ll model .
let us consider the gross - pitaevskii ( gp ) equation , which describes bose - einstein condensation ( bec ) in the mean - field approximation @xcite .
we also call it the nonlinear schrdinger equation .
the gp equation has dark soliton solutions for the repulsive interactions , while it has bright soliton solutions for the attractive interactions @xcite .
it was conjectured that dark solitons are closely related to lieb s type - ii excitations , i.e. one - hole excitations , by carefully studying the dispersion relations @xcite .
the dispersion relations of the ll model are briefly compared with those of the classical nonlinear schrdinger equation in the weak coupling limit @xcite .
however , it has not been shown how one can construct such a quantum state that leads to a dark soliton in the classical limit or what kind of physical quantity can show a property of a dark soliton for some quantum state . here
we remark that each of the type - ii eigenstates has a flat density profile since the bethe ansatz eigenstates are translationally invariant .
moreover , we consider dark solitons under the periodic boundary conditions , which are expressed in terms of elliptic functions @xcite . in this paper we demonstrate that a quantum state derived from the bethe ansatz eigenvectors of the ll model by taking the fourier transform of the type - ii excitations over a branch @xcite has many aspects closely related to classical dark solitons of the gp equation under the periodic boundary conditions .
we call the state the _ quantum dark soliton state _ and a classical dark soliton under the periodic boundary conditions simply a _ classical dark soliton_. through the bethe ansatz we show several corresponding aspects in the weak coupling regime .
firstly , the density profile of the quantum dark soliton state at an initial time is consistent with that of a classical dark soliton .
here we call the graph of the expectation value of the local density operator for a given state versus the position coordinate the density profile of the state , and for a quantum dark soliton state
we simply call it the density profile of quantum dark soliton ; we call the graphs of the square amplitude and phase in the complex scalar field of a classical dark soliton versus the position coordinate the density profile and phase profile of classical dark soliton , respectively . secondly , in time evolution the density profile of quantum dark soliton coincides with that of the corresponding classical dark soliton over the whole graph and they move together with the same velocity for a long period of time . thirdly , for the matrix element of the bosonic field operator between two quantum dark soliton states where one state has @xmath0 particles and another @xmath1 particles , the profiles of the square amplitude and phase at the initial time graphically agree with those of classical dark soliton , respectively . in time
evolution the profiles of square amplitude and phase overlap with those of classical dark soliton , respectively , over the whole region and the corresponding profiles move together in time for a long period of time . here
we remark that a classical dark soliton parametrized by elliptic functions becomes a standard dark soliton with open boundaries by simultaneously sending the system size to infinity and the elliptic modulus to unity .
furthermore , in order to illustrate the method for constructing quantum dark solitons , in the 1d free fermions we show from the anti - commutation relations that a notch appears in the density profile of some superposition of one - hole excitations .
interestingly , the density profile of the fermionic state coincides with that of quantum dark soliton for the 1d bose gas in the large coupling case , i.e. near the tg limit , not only at an initial time but also during the quantum dynamics for some period of time
. the time evolution of the expectation value of the local density operator in the 1d bose gas should be important also from the renewed interest on the fundamentals of quantum statistical mechanics @xcite .
the density profile of a quantum dark soliton state has initially a localized notch but collapses slowly to a flat profile in time evolution @xcite .
the relaxation behavior is consistent with the viewpoints of equilibration of an isolated quantum system @xcite and thermalization due to the typicality of quantum states @xcite .
we now argue that the density profile of quantum dark soliton has a finite lifetime but it is much longer than the quantum speed limit time of the quantum state in the weak coupling case . here
we remark that the lifetime of a generic state is given by the quantum speed limit time for it @xcite . by observing the exact time evolution of the density profile of quantum dark soliton
we estimate its lifetime .
we shall show in section 5 that the observed life time is inversely proportional to the coupling constant .
we thus suggest that the localized density profile of a quantum dark soliton is much more stable than the density profile of a generic quantum state in the weak coupling case . here
we remark that girardeau and wright discussed a permanent quantum soliton for impenetrable bosons in 1d @xcite , which corresponds to the infinite coupling case of the ll model .
we also argue that the behavior of the wavefunctions is nontrivial in the weak coupling limit for the ll model .
the wavefunctions do not simply become close to such wavefunctions of the 1d free bosons that could be consistent with the mean - field picture , when the coupling constant approaches zero but it takes a nonzero value .
the exact bethe ansatz wavefunctions consist of a large number of terms such as @xmath2 .
furthermore , we can show that many - body correlations increase if we increase the particle number @xmath1 while we keep the coupling constant small but fixed and the particle density constant .
for instance , the zero mode fraction in the ground state of the 1d bose gas , which we regard as the condensate fraction of bec , becomes small and decreases to zero if the number of particles increases while the coupling constant and the particle density are fixed @xcite .
we suggest that it is also the case for quantum dark solitons i.e. , that some properties of quantum dark solitons corresponding to classical dark solitons do not necessarily hold if we increase the particle number @xmath1 while we keep the coupling constant small but fixed and the density of particles fixed .
in fact , it is not _ a priori _ clear how valid the mean - field approximation is for quantum dark solitons even in the weak coupling case . here
we remark that the breakdown of mean - field theory is addressed for impenetrable bosons @xcite , and a long - wavelength theory beyond the mean - field approximation is discussed for low - dimensional bose liquids @xcite .
let us consider the hamiltonian of the ll model for @xmath1 bosons with coupling constant @xmath3 @xmath4 here we impose periodic boundary conditions with length @xmath5 .
we employ a system of units with @xmath6 , where @xmath7 is the particle mass .
we introduce the canonical bose field operator @xmath8 with the commutation relations @xcite @xmath9=\delta(x - x ' ) , \\ & [ \hat { \psi}(x , t ) , \hat { \psi}(x',t)]=[\hat { \psi}^{\dagger}(x , t ) , \hat { \psi}^{\dagger}(x',t)]=0 . \end{aligned}\ ] ] the second - quantized hamiltonian of the ll model is written in terms of the field operator @xmath8 as @xmath10 , \label{h2}\end{aligned}\ ] ] where @xmath11 is the chemical potential .
the heisenberg equation of motion of this system has the form @xmath12 in the classical limit where the quantum field operator @xmath8 is replaced by a complex @xmath3-number field @xmath13 , equation ( [ eq : qnse ] ) becomes the following partial differential equation @xmath14 we call it the nonlinear schrdinger equation .
the equation ( [ eq : nse ] ) can be solved by the inverse scattering method and has soliton solutions @xcite ( see also @xcite ) .
we recall that it has dark soliton solutions for the repulsive interactions @xmath15 , while it has bright soliton solutions for the attractive interactions @xmath16 . for the attractive case ,
bright solitons are analytically derived from the quantum wave packets constructed through the fourier transform of the bound states @xcite .
the construction was extended for optical fibers @xcite .
time evolution of quantum states of bright solitons was investigated theoretically @xcite and experimentally @xcite .
the effect of quantum noises on the quantum states of bright solitons was also studied @xcite .
the contents of the paper consists of the following . in section 2 we briefly introduce the notation for the bethe ansatz , characterize the type ii excitation branch , and then construct quantum dark soliton states .
we argue that the structure of quantum dark soliton states is analogous to the dirichlet kernel , and show for free fermions that the density profile of a superposition of one - hole excited states has a notch .
it illustrates the construction of quantum dark soliton states . in section 3
we derive one - soliton solutions of the nonlinear schrdinger equation ( [ eq : nse ] ) under the periodic boundary conditions .
we express the classical dark solitons in terms of elliptic functions similarly as in ref .
@xcite . by sending the elliptic modulus @xmath17 to 1 ,
the classical dark solitons under the periodic boundary conditions become the standard classical dark solitons of the nonlinear schrdinger equation with open boundaries . furthermore , we calculate the logarithmic corrections to them , in particular , to the chemical potential and the soliton velocity , by which we can confirm the numerical estimates of the soliton parameters shown in section 4 at least approximately . in section 4 we show several corresponding aspects between the quantum dark soliton states and the corresponding classical dark solitons , which hold in the weak coupling case and should be rigorously valid when the coupling constant approaches zero with the density and the particle number @xmath1 fixed .
firstly , we show that the density profile of quantum dark soliton at an initial time is consistent with that of classical dark soliton .
secondly , we show that in exact time evolution the density profile of quantum dark soliton as a whole coincides with that of classical dark soliton for a long period of time .
thirdly , we show that both the square amplitude and phase profiles for the matrix element of the quantum field operator @xmath18 in the ll model between the quantum dark soliton states with @xmath0 and @xmath1 particles overlap with those of classical dark soliton , respectively , all over the profiles . the square amplitude and phase profiles move in time evolution with exactly the same speed as those of classical dark soliton , respectively .
furthermore , we remark that the density profile of a sum of one - hole excitations in the 1d free fermions overlap with that of quantum dark soliton in the 1d bose gas with the large coupling constant , i.e. in a regime close to the tg limit . for the matrix element of the fermionic field operator between two such fermionic states the square amplitude and phase profiles have several features in common with those of quantum dark soliton in the 1d bose gas for the large coupling case , respectively , although they do not overlap each other .
we thus obtain useful tools to describe approximately the profiles of quantum dark soliton in the large coupling case . however , it is still quite nontrivial that the profiles of quantum dark soliton in the weak coupling case are consistent with those of classical dark soliton . in section 5
we argue that the lifetime of a notch in the density profile of quantum dark soliton is much longer than that of a generic state when the coupling constant @xmath3 is very small .
we compare the lifetime of a notch in the density profile of quantum dark soliton with the quantum speed limit time , and show that the ratio of the former to the latter increases as the coupling constant decreases .
it becomes infinitely large as the coupling constant approaches zero .
thus , although the localized density profile collapses in time evolution , we call the quantum states quantum dark soliton states .
in the framework of the bethe ansatz method , the bethe ansatz wavefunction which is expressed in terms of quasi - momenta @xmath19 @xmath20 , \nn\db & a_\sg=(-1)^\sg \prod_{j>\ell}^n\[k_{\sg_j}-k_{\sg_\ell}-\i c \ , \sign(x_j - x_\ell)\ ] \end{aligned}\ ] ] gives an eigenstate of the ll model , if the quasi - momenta satisfy the bethe ansatz equations @xmath21 in the logarithmic form , we have @xmath22 here @xmath23 s are integers for odd @xmath1 and half - odd integers for even @xmath1 .
we call them the bethe quantum numbers .
the total momentum @xmath24 and energy eigenvalue @xmath25 are written in terms of the quasi - momenta as @xmath26 in the case of repulsive interaction @xmath15 , the bethe equations have a unique real solution @xmath27 if we specify a set of bethe quantum numbers @xmath28 @xcite .
the configuration of the bethe quantum numbers for the ground state is given by the regular array around the center : @xmath29 it is depicted in the top panel of figure [ ij1h1p ] ( a ) in the case of @xmath30 .
we can construct low - lying excited states systematically by putting holes and placing particles in the regular array of the bethe quantum numbers for the case of the ground state .
the arrays of the bethe quantum numbers for one - hole excitations are shown in the second panel of figure [ ij1h1p ] ( a ) . in each sequence there
exists an unoccupied number in the middle of the sequence of the five red circles , which we call the one - hole .
we remark that the one - hole excitations are also called the type ii excitations @xcite .
similarly as in the second panel , the arrays of the bethe quantum numbers for one - particle excitations are shown in the third panel of figure [ ij1h1p ] ( a ) .
the right most red particle corresponds to the one - particle in each array .
the one - particle excitations are also called the type i excitations @xcite .
the dispersion relation of the one - particle excitations and that of the one - hole excitations are plotted with filled circles and filled squares , respectively , in figure [ ij1h1p ] ( b ) .
the former branch corresponds to the type i excitations and the latter branch the type ii excitations . .
( b ) dispersion relations of the 1-hole excitations and 1-particle excitations in the case of @xmath30 .
a triangle represents the ground state , big circles represent 1-particle excitations , and squares represent 1-hole excitations .
small dots are the 1-particle 1-hole excitations . ]
we now explain how we construct a series of quantum states @xcite through the type ii excitations .
furthermore , we call them quantum dark soliton states . in the type ii branch , for each integer @xmath31 in the set @xmath32 , we consider momentum @xmath33 .
we denote the normalized bethe eigenstate of @xmath1 particles with total momentum @xmath24 by @xmath34 , which we call a one - hole excited state .
the bethe quantum numbers @xmath23 s of the one - hole excitation @xmath35 are given by @xmath36 for any value of @xmath37 satisfying @xmath38 we define the coordinate state @xmath39 by the discrete fourier transformation : @xmath40 shows a density notch as will be shown in section 4 .
the formal structure of quantum dark soliton states is analogous to a truncated series of the delta function . for periodic functions with period @xmath5 the delta function
is expressed in the form of an infnite series @xmath41 we now truncate it by keeping only such terms with integers @xmath42 satisfying @xmath43 @xmath44 it is called the dirichlet kernel in the fourier series analysis .
it has the peak around at @xmath45 with peak height @xmath46 and oscillating behavior in the region for @xmath47 with period @xmath48 .
here we assume that @xmath49 is independent of @xmath5 . for @xmath50 ,
the oscillations in the graph of the dirichlet kernel become of very short wavelength , and we may call them ripples .
the integral of the squared amplitude of the dirichlet kernel over the period @xmath5 is given by @xmath51 we therefore normalize the function @xmath52 by dividing it by the square root @xmath53 @xmath54 in terms of the normalized momentum eigenfunctions @xmath55 where @xmath33 the normalized function @xmath56 is expressed as @xmath57 it is analogous to the structure of the states of quantum dark solitons , where the number @xmath58 in ( [ eq : xp ] ) corresponds to the number of states @xmath1 in , and the single - particle momentum eigenstates @xmath59 to the one - hole excitations @xmath60 consisting of @xmath1 particles . by considering the phase factors @xmath61 in the graphs
are shifted by @xmath37 in the @xmath62-direction .
we have @xmath63 for the truncated series of the delta function modified with the phase factors . through the anti - commutation relation of the field operators
we now show that a notch appears in the density profile of a linear combination of one - hole excited states in the 1d free fermions . under the periodic conditions of length @xmath5
the momenta are given by @xmath64 with some integers @xmath65 in the 1d free fermions .
we denote by @xmath66 and @xmath67 for some integers @xmath68 the creation and annihilation operators of the one - dimensional free fermions with momenta @xmath69 , respectively .
we assume that an infinite series of integers @xmath65 for @xmath70 cover all integers .
the field operator and its conjugate are given by @xmath71 let us take a set of arbitrary integers @xmath72 and consider the state of @xmath73 fermions with momenta @xmath74 s @xmath75 for simplicity , we may assume that @xmath76 is the ground state . by applying the field operator @xmath77 to it
, we have @xmath78 here we have assumed the ordering of fermionic operators as @xmath79 .
we regard states @xmath80 as ` one - hole excitations ' of @xmath81 .
we define the state @xmath82 by the sum over all one - hole excitations derived from @xmath83 @xmath84 let us define the local density @xmath85 for the state @xmath86 of the 1d free fermions by @xmath87 we now show that the local density vanishes at the origin : @xmath88 for the state .
we first recall that the square of the field operator at the origin vanishes , @xmath89 , due to the anti - commutation relations of the field operators .
thus , we have @xmath90 we can show that the integral of the local density over the entire interval @xmath5 is given by a large positive value such as @xmath91 @xmath92 it follows that the local density can not be always equal to zero , although it vanishes at the origin .
therefore , for the state @xmath82 which is given by the sum over all the one - hole excited states of @xmath83 , the local density @xmath93 has a notch at least at the origin . in section 4
we shall construct the free - fermionic analog of quantum dark soliton states for the 1d free fermions and show that the density profile coincides with those of the ll model if the coupling constant is large enough such as @xmath94 .
let us assume the traveling wave solution with velocity @xmath95 : @xmath96 to the nonlinear schrdinger equation .
hereafter in section 3 we denote @xmath97 simply by @xmath98 .
we denote the time derivative by @xmath99 and the second spatial derivative by @xmath100 , where @xmath101 denotes @xmath102 with @xmath103 .
we therefore have @xmath104 we denote the solution of eq .
( [ eq2 ] ) by @xmath105 as a function of @xmath62 , for simplicity .
we express the complex scalar field @xmath105 in terms of the amplitude @xmath106 and the phase @xmath107 , both of which are real , as @xmath108 by substituting ( [ eq : ap ] ) into equation we derive a pair of coupled equations from the real and imaginary parts , respectively .
@xmath109 it follows from equation that we have @xmath110 . by integrating it once with respect to @xmath62
, we have @xmath111 here @xmath112 denotes a constant of integration . substituting ( [ vpd ] ) into and multiplying it by @xmath113 ,
we have @xmath114 integrating this equation with respect to @xmath62 , we have @xmath115 where @xmath116 is another constant of integration .
thus , we have @xmath117 where the potential function @xmath118 is given by @xmath119 let us assume that the cubic equation : @xmath120 has three distinct real roots , @xmath121 , @xmath122 , and @xmath123 .
we put them in increasing order : @xmath124 .
the coefficients of the cubic equation are expressed in terms of the roots as follows .
@xmath125 we now derive a solution @xmath126 to equation ( [ eq : u ] ) such that it satisfies @xmath127 in the interval of @xmath62 with @xmath128 .
we set the initial condition : @xmath129 , for simplicity .
it thus follows from equation ( [ eq : u ] ) that we have @xmath130 let us define the modulus of jacobi s elliptic functions , @xmath17 , by @xmath131 it is clear that @xmath132 . through the transformation of variables from @xmath133 to @xmath134 @xmath135 we express the integral ( [ eq : integral - x ] ) in terms of the elliptic integral of the first kind : @xmath136 we therefore have the solution to eq .
( [ eq : u ] ) in terms of jacobi s elliptic function @xmath137 @xmath138 we now integrate equation . here
we remark that the constant @xmath112 is given by @xmath139 . in terms of the elliptic integral of the third kind
we have @xmath140 where @xmath141 corresponds to the constant of integration .
we now express @xmath121 , @xmath122 and @xmath123 in terms of the complete elliptic integrals . here
we consider the periodic boundary conditions ( pbc ) for the square amplitude @xmath126 and the phase @xmath142 of the classical complex scalar field @xmath143 , and the normalization condition of the square amplitude @xmath126 .
then , the elliptic modulus @xmath17 is related to the depth of the classical dark soliton , i.e. the minimum value @xmath121 of the square amplitude @xmath126 .
we express the soliton velocity @xmath95 in terms of the elliptic integral of the second kind , and determine the chemical potential @xmath11 .
suppose that the square amplitude @xmath126 of a classical dark soliton increases from @xmath121 to @xmath122 and then returns back to @xmath121with period @xmath5 .
we therefore assign the conditions at the initial and the middle points as @xmath144 it follows from that in terms of the complete elliptic integral of the first kind , @xmath145 , we have @xmath146 we have @xmath147 . here and
hereafter we often denote @xmath145 by @xmath148 , suppressing the modulus @xmath17 .
we assume the periodic boundary condition for the phase : @xmath149 .
suppose that the soliton velocity is positive : @xmath150 .
it follows that we have @xmath151 and the velocity @xmath95 is expressed in terms of the complete elliptic integral of the third kind as @xmath152 we impose the normalization condition to the square amplitude @xmath126 @xmath153 where @xmath154 is the density of the number of particles @xmath1 , i.e. we have @xmath155 . by the formula we transform equation ( [ eq : rho ] ) in terms of jacobi s @xmath156 function into the following .
@xmath157 thus , by the formula in terms of the complete elliptic integral of the second kind @xmath158 we have @xmath159 by solving equations , and with respect to the roots @xmath121 , @xmath122 and @xmath123 , we express them in terms of the complete elliptic integrals as follows : @xmath160 since the minimum value of the amplitude must be non - negative , i.e. @xmath161 , we assign the following condition @xmath162 thus , if the coupling constant @xmath3 , the system size @xmath5 , the density @xmath42 and the elliptic modulus @xmath17 satisfy condition ( [ eq : nonnegative ] ) , the classical dark soliton exists as a solution of the nonlinear schrdinger equation ( i.e. the gp equation ) . the soliton velocity @xmath95 and the chemical potential @xmath11
are given by @xmath163 let us explain how we evaluate the parameters of a dark soliton such as @xmath42 and @xmath17 in order to compare the profiles of the dark soliton with those of a given quantum dark soliton . here
we remark that when a quantum state is given , parameters @xmath3 , @xmath5 and the particle number @xmath1 have been specified .
we determine the elliptic modulus @xmath17 from the depth of the dark soliton , i.e. @xmath121 : we solve equation as an equation for modulus @xmath17 .
we assign the value @xmath164 on the parameter of density @xmath42 for the density profile of a quantum state . however , we determine the density @xmath42 numerically for the squared amplitude profile of the matrix element of the bosonic field operator between different quantum states , since it may take a value smaller than @xmath164 . for the density profile of a quantum state @xmath165 ,
the integral of the expectation value @xmath166 of the local density operator over the whole space is equal to @xmath164 .
however , for the matrix element @xmath167 of the quantum field operator @xmath168 between two states @xmath165 and @xmath169 , the integral of the squared amplitude @xmath170 over the whole space may be smaller than the value of @xmath164 . in this case
we determine the parameter @xmath42 numerically by taking the integral of the squared amplitude @xmath171 over the whole space . here
we remark that the symbols will be defined in section 4.1.1 . in summary ,
after we fix the value of density @xmath42 , we solve equation as a function of modulus @xmath17 and estimate its numerical value very precisely . in section 4
we shall make the profiles of classical dark soliton for given density profiles of quantum dark soliton by the method explained in the above .
we define the function @xmath172 by @xmath173 $ ] .
it appears in the third term of the expression for the chemical potential @xmath11 .
we can show that it is a monotonically decreasing function of modulus @xmath17 , which yields the inequality @xmath174 .
this gives an upper bound for the absolute value of soliton velocity @xmath95 : @xmath175 , where @xmath176 here we remark that the critical velocity is defined for the system of a finite size .
we shall denote by @xmath177 the critical velocity for an infinite system .
we now express a classical dark soliton explicitly in terms of the complete elliptic integrals .
the square amplitude is given by @xmath178 we shall show that the second term in the right - hand side of equation leads to the logarithmic correction associated with the conservation of the number of particles .
let us introduce a parameter @xmath179 by @xmath180 we denote the argument @xmath181 in by @xmath182 .
then , the square amplitude is expressed as follows : @xmath183 we now show that the phase @xmath142 is expressed in terms of jacobi s theta function .
let us define a pure imaginary number @xmath184 with @xmath185 by @xmath186 by formula ( [ eq : ei3 ] ) we evaluate the elliptic integral of the third kind with jacobi s zeta and theta functions @xcite , and we express the phase @xmath142 as the logarithm of a ratio of theta functions @xmath187 let us consider the limit of sending the system size @xmath5 to infinity ( @xmath188 ) and the modulus @xmath17 to 1 ( @xmath189 ) simultaneously so that the ratio @xmath190 is kept constant .
we denote the ratio by @xmath191 : @xmath192 . here
we remark that we keep the density @xmath154 constant .
we define @xmath193 by the simultaneous limit of sending @xmath5 to infinity and @xmath17 to 1 with @xmath194 .
@xmath195 let us introduce the complementary modulus @xmath196 by @xmath197 .
we define @xmath198 and @xmath199 by @xmath200 and @xmath201 , respectively .
we remark that @xmath148 and @xmath202 give quarter periods of jacobi s elliptic functions .
we introduce a parameter @xmath31 by @xmath203 , which is small when modulus @xmath17 is close to 1 . making use of formulae , , and legendre s relation
we expand @xmath148 s and @xmath25 s with respect to @xmath31 as follows : @xmath204 we have @xmath205 , since we have fixed the ratio @xmath194 .
we recall that the soliton velocity @xmath95 is expressed in terms of the complete elliptic integral of the third kind . through
we express it in terms of jacobi s zeta function as follows : @xmath206 here we recall that parameter @xmath68 has been defined by .
let us now evaluate @xmath207 .
we first express the left - hand side of in terms of @xmath179 as @xmath208 when the modulus @xmath17 is close to 1 , by making use of jacobi s imaginary transformation of @xmath209 function : @xmath210 and then by expressing it in terms of jacobi s eta and theta functions , we show @xmath211 through we express the value of zeta function @xmath212 in terms of @xmath184 as follows .
@xmath213 applying the expansions and to , we evaluate the velocity @xmath95 upto the order of logarithmic terms @xmath214 and through we evaluate the chemical potential up to the order of logarithmic terms @xmath215 it follows that chemical potential @xmath11 approaches @xmath216 when we send system size @xmath5 to infinity .
we denote the limiting value by @xmath217 .
here we remark that the logarithmic corrections correspond to the finite size corrections since we have @xmath218 . moreover , the chemical potential @xmath11 is larger than @xmath219 when the system size is finite since we have @xmath220 for @xmath221
. we can confirm such behavior by checking the numerical estimates for the parameters of classical dark solitons listed in section 4 . through jacobi s imaginary transformation for @xmath156 function
we have @xmath222 we therefore have @xmath223 it is easy to show that the asymptotic expansion of the square amplitude up to the order of @xmath224 , i.e. the order of @xmath225 , satisfies the normalization condition . at @xmath226 , the square amplitude @xmath126
is expressed in terms of @xmath227 by @xmath228 thus , the critical velocity @xmath229 with the infinite system size ( @xmath230 ) is given by @xmath231 .
here we recall @xmath232 .
it follows that we have @xmath233 \right\ } .
\end{aligned}\ ] ] here @xmath193 is expressed as @xmath234 .
we have thus derived the square amplitude of one - soliton solution due to tsuzuki @xcite . applying
to we derive the following expansion in terms of @xmath235 for any system size @xmath5 : @xmath236 making use of we derive the expansion of @xmath142 up to the order of the inverse of @xmath237 as follows : @xmath238 therefore , in the limit of sending @xmath17 to 1 , we obtain the phase of the dark soliton solution by tsuzuki @xcite : @xmath239 thus , we have shown that the classical dark soliton under the pbcs approaches the dark soliton solution by tsuzuki @xcite through the simultaneous limit of sending the system size to infinity and the modulus @xmath17 to 1 .
we define the local density operator @xmath240 by @xmath241 , in the second - quantized system of the 1d bose gas interacting through the delta - function potentials . here
we denote by @xmath242 the field operator at the initial time @xmath243 : @xmath244 .
we now consider the graph of the expectation value of the local density operator for a quantum state @xmath245 ( i.e. @xmath246 ) versus the position coordinate @xmath62 . here
we recall that we call it the density profile of the state @xmath245 .
we evaluate the expectation value of the local density operator for the state @xmath247 by the form factor expansion .
we express the expectation value as the sum over the form factors between the bethe eigenstates @xmath248 \bra p ' , n|\hat{\rho}(0)|p , n\ket \ , .
\label{eq : dp - formfactor}\end{aligned}\ ] ] here @xmath34 and @xmath249 are the normalized bethe eigenstates of @xmath1 particles in the type ii branch ( i.e. they are one - hole excitations ) and have total momentum @xmath33 and @xmath250 , respectively . here
the form factors @xmath251 are effectively calculated @xcite by the determinant formula for the norms of bethe eigenstates @xcite and that of the form factors of the density operator @xcite @xmath252 where the quasimomenta @xmath253 and @xmath254 lead to the eigenstates @xmath255 and @xmath256 , respectively .
we use the abbreviations @xmath257 and @xmath258 .
the kernel @xmath259 is defined by @xmath260 .
the matrix @xmath261 is called the gaudin matrix , whose @xmath262 th element is given by @xmath263-{\hat k}(k_{j,\ell } ) \quad \mbox{for } \quad j , \ell=1 , 2 , \cdots , n .
\label{eq : gaudin}\ ] ] the matrix elements of the @xmath264 by @xmath264 matrix @xmath265 are given by @xmath266 + \frac{\prod^n_{a=1}(k'_a - k_j)}{\prod^n_{a\neq j}(k_a - k_j ) } \({\hat k}(k_{j,\ell})-{\hat
k}(k_{n,\ell})\ ) .
\label{eq : matrixu_density}\end{aligned}\ ] ] , i.e. @xmath267 versus @xmath62 , is shown by a red solid line to each value of @xmath3 with @xmath268 and 100 .
the density profile of classical dark soliton @xmath269 with parameters shown in table [ density_table20-t1 ] is plotted with a blue broken line to each value of @xmath3 with @xmath268 and 100 . ] , i.e. @xmath270 versus @xmath62 , is shown by a red solid line to each value of @xmath3 with @xmath268 and 100 .
the density profile of classical dark soliton @xmath271 with parameters shown in table [ density_table500-t2 ] is plotted with a blue broken line to each of @xmath3 with @xmath268 and 100 . ]
.parameters for the density profiles of the classical dark solitons with @xmath272 plotted by blue broken lines in fig .
[ density20 ] and in the first and the second columns of fig .
[ density_dynamics ] . [ cols="^,^,^,^,^",options="header " , ] the square amplitude profile @xmath273 of quantum dark soliton is consistent with the density profile of classical dark soliton in the weak coupling case of @xmath274 , as shown in the upper left panel of fig .
[ field_amp20 ] . moreover , the phase profiles @xmath275 $ ] of quantum dark soliton perfectly agree with those of classical dark soliton for @xmath274 and @xmath276 , as graphs . here
, no deviation is found between the quantum and classical profiles .
we thus suggest that both the profiles of square amplitude @xmath273 and phase @xmath275 $ ] for the matrix element @xmath277 between the quantum dark soliton states should be consistent with those of classical dark soliton , respectively , in the weak coupling limit : @xmath278 . for @xmath279 the profiles of square amplitude @xmath273 and phase @xmath275/\pi$ ] are plotted in figs [ field_amp500 ] and [ field_phase500 ] , respectively .
the density profile @xmath271 and phase profile @xmath280/\pi = \varphi(x - l/2)/\pi$ ] of classical dark soliton are shown by broken blue lines .
the numerical estimates of the soliton parameters are given in table [ density_table500 ] .
they are obtained by the numerical method of section 3.2.4 .
the phase profiles @xmath275/\pi$ ] of the matrix element @xmath277 perfectly agree with those of the corresponding classical dark solitons for the cases of @xmath281 and 1 , respectively . on the other hand ,
the square amplitude profile @xmath273 of the matrix element @xmath277 is not necessarily in very good agreement with the density profile of classical dark soliton even for the case of @xmath274 .
we suggest that it is due to the effect of many - body correlations in the 1d bose gas such as observed in the study of the bec fraction @xcite as mentioned in introduction .
we also suggest that if the coupling constant @xmath3 becomes much smaller , the square amplitude profiles @xmath273 should be in much better agreement with the density profiles of classical dark soliton @xmath282 .
we now show the time evolution of the square amplitude profile and the phase profile of the matrix element of the bosonic quantum field operator with respect to the quantum dark soliton state with @xmath1 particles and that of @xmath0 particles .
we denote by @xmath283 the matrix element of the bosonic quantum field operator @xmath284 between the quantum dark soliton states @xmath39 and @xmath285 as follows .
@xmath286 here we recall that the two quantum states have different particle numbers but have the same system size @xmath5 . the snapshots at times @xmath287 for the profiles of the square amplitude and the phase of the matrix element @xmath283 of the bosonic quantum field operator @xmath284 between the quantum dark soliton states @xmath39 and @xmath285 are plotted by red solid lines for the cases of @xmath288 and @xmath289 in figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] , respectively . here
, we consider two values of the coupling constant such as @xmath274 and @xmath290 to each of the particle number @xmath1 in the cases of @xmath291 and 500 . for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of quantum dark soliton with @xmath288 for @xmath274 are shown by red solid lines in the first and second columns from the left , respectively .
the snapshots at times @xmath287 of the density profile @xmath294 and phase profile @xmath295/\pi$ ] of classical dark soliton are plotted by blue broken lines in the first and second columns from the left , respectively . here , the soliton parameters are given in table [ density_table20 ] .
( ii ) snapshots at times @xmath287 for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of quantum dark soliton with @xmath288 for @xmath290 are shown by red solid lines in the third and fourth columns , respectively .
the snapshots of the density profile @xmath294 and phase profile @xmath295/\pi$ ] of classical dark soliton are shown by blue broken lines in the third and fourth columns from the left , respectively . here , the soliton parameters are shown in table [ density_table20 ] . ]
the quantum and classical profiles are in good agreement in many aspects of the time evolution as shown in figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] .
we now describe three aspects as follows .
firstly , in the weak coupling case of @xmath274 , both the time evolution of the square amplitude profile @xmath296 and that of the phase profile @xmath293/\pi$ ] are consistent with those of the corresponding classical dark soliton , @xmath294 and @xmath295/\pi$ ] , respectively . the consistency of the quantum profiles of the square amplitude and the phase with the classical profiles of the corresponding classical dark soliton holds over a long period of time such as at least up to @xmath297 as shown in the first and second columns of panels in fig . [ field_n20_dynamics ] from the left .
secondly , the quantum profiles move with the same speed as the corresponding classical profiles , as shown in all the eight columns of panels in figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] . in the first and third columns of figs .
[ field_n20_dynamics ] the positions of the notches in the square amplitude profiles @xmath292 are in very good agreement with those of the corresponding classical dark solitons . in the second and fourth columns of fig .
[ field_n20_dynamics ] , the step positions in the phase profiles @xmath293/\pi$ ] are consistent with those of the phase profiles @xmath295/\pi$ ] of the corresponding classical dark solitons .
thirdly , as far as the phase profiles are concerned , the quantum and classical profiles coincide with each other very well for almost all cases as shown in figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] .
in fact , not only the step position in the profile of @xmath293 $ ] moves together with that of the phase @xmath295 $ ] of the corresponding classical dark soliton , but also the whole profile itself coincides with the corresponding classical one .
\(i ) case of @xmath291 for @xmath274 with @xmath274 and @xmath288 , i.e. the weak coupling case , for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of the matrix element @xmath283 , the snapshots at times @xmath287 are plotted by red solid lines in the first and second columns of panels of fig . [ field_n20_dynamics ] from the left , respectively . here , five snapshots are shown in each column of panels for @xmath298 and 40 . the snapshots at times @xmath287 for the density profile @xmath299 and phase profile @xmath295/\pi = \varphi(x - vt - l/2)/\pi$ ] of classical dark soliton are shown by blue broken lines in the first and second columns , respectively . the numerical estimates of the soliton parameters are given in table [ density_table20 ] .
\(ii ) case of @xmath291 for @xmath290 with @xmath290 and @xmath288 , for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of the matrix element @xmath283 , the snapshots at times @xmath287 are plotted by red solid lines in the third and fourth columns of panels of fig . [ field_n20_dynamics ] from the left , respectively . here
, five snapshots are shown in each column of panels for @xmath300 and 8 . the density profile @xmath299 and phase profile @xmath295/\pi = \varphi(x - vt - l/2)/\pi$ ] of classical dark soliton are shown by blue broken lines in the third and fourth columns , respectively .
the numerical estimates of the soliton parameters are given in table [ density_table20 ] .
we observe that by comparing the snapshots of @xmath274 with those of @xmath290 , the density notch in the square amplitude profile @xmath292 at @xmath243 decays in a shorter period of time as the coupling constant becomes larger , as shown in the first and third columns of fig .
[ field_n20_dynamics ] for @xmath274 and @xmath290 , respectively .
however , the position of the density notch moves together with that of classical dark soliton , although the notch itself becomes wider rather rapidly in time . in the phase profiles ,
the abrut step becomes softer and milder gradually in time , but the change in the phase profile is much slower than in the square amplitude profile .
\(iii ) case of @xmath279 for @xmath274 with @xmath274 and @xmath289 , for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of the matrix element @xmath283 , the snapshots at times @xmath287 are plotted by red solid lines in the first and second columns of panels of fig . [ field_n500_dynamics ] from the left , respectively . here , five snapshots are taken with much longer intervals of time than in the case of @xmath288 in fig . [ field_n20_dynamics ] such as for @xmath301 and 4000 in the case of the square amplitude @xmath292 and for @xmath302 and 16000 in the case of the phase @xmath293 $ ] .
the profiles of square amplitude @xmath303 and phase @xmath295/\pi = \varphi(x - vt - l/2)/\pi$ ] are shown by blue broken lines in the first and second columns of fig .
[ field_n500_dynamics ] from the left , respectively .
the numerical estimates of the soliton parameters are given in table [ density_table500 ] .
we observe that the square amplitude @xmath292 for the matrix element of the quantum field operator @xmath284 approaches zero in time evolution .
the decaying behavior is clear for @xmath290 as shown in the third column from the left of figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] . in the first column of fig .
[ field_n500_dynamics ] from the left , we also observe it for @xmath274 and @xmath289 over a very long period of time up to @xmath304 .
however , for the density profiles of quantum dark soliton , the density approaches a nonzero constant value in time evolution , as shown in fig .
[ density_dynamics ] .
we suggest that the quantum dark soliton states with different particle numbers such as @xmath1 and @xmath0 do not have the same set of energy eigenvalues in common , so that the off - diagonal matrix element @xmath305 vanishes in time , finally .
\(iv ) case of @xmath279 for @xmath290 with @xmath290 and @xmath289 , for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] in the matrix element @xmath283 , the snapshots at times @xmath287 are plotted by red solid lines in the third and fourth columns of panels of fig . [ field_n500_dynamics ] from the left , respectively . here
, five snapshots are taken for almost the same intervals of time as in the case of fig .
[ field_n20_dynamics ] such as for @xmath306 and 12 in both cases of the profiles of square amplitude @xmath292 and phase @xmath293 $ ] . the density profile @xmath299 and phase profile @xmath295/\pi = \varphi(x - vt - l/2)/\pi$ ] of classical dark soliton are shown by blue broken lines in the third and the fourth columns of fig .
[ field_n500_dynamics ] from the left , respectively .
the numerical estimates of the soliton parameters are given in table [ density_table500 ] .
here we remark that when we evaluate the matrix element @xmath283 straightforwardly by making use of the form factor formula , the phase of the matrix element @xmath283 increases with a constant velocity in time , and the phase profile moves downwards with a constant velocity in the animation .
we suggest that it is related to the difference between the zero point energy of the eigenstates with @xmath0 particles and that of @xmath1 particles . in figs .
[ field_n20_dynamics ] and [ field_n500_dynamics ] we plotted the phase profiles which are subtracted by the constant shift in time . for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of quantum dark soliton with @xmath289 for @xmath274 are shown by red solid lines in the first and second columns from the left , respectively .
the snapshots at times @xmath287 for the density profile @xmath294 and phase profile @xmath295/\pi$ ] of classical dark soliton are plotted by bule broken lines in the first and second columns from the left , respectively . here , the soliton parameters are given in table [ density_table500 ] .
( ii ) snapshots at times @xmath287 for the profiles of square amplitude @xmath292 and phase @xmath293/\pi$ ] of quantum dark soliton with @xmath289 for @xmath290 are shown by red solid lines in the third and fourth columns , respectively . the snapshots at times @xmath287 for the density profile @xmath294 and phase profile @xmath295/\pi$ ] of classical dark soliton are shown in the third and fourth columns from the left , respectively . here , the soliton parameters are shown in table [ density_table500 ] . ] as the quantum dark soliton states for the 1d free fermions with @xmath1 and @xmath0 particles we consider the following @xmath307 we calculate the matrix element of the field operator between the two states as @xmath308 here , the exponential sum is given by the dirichlet kernel .
for instance , when we take @xmath309 and momenta @xmath310 for @xmath311 , -1 , and @xmath312 where @xmath313 we have @xmath314 the square - amplitude and phase profiles for the matrix element of the fermionic field operator have several features in common with those of the 1d bose gas in the large coupling case shown in figs . [ field_amp20 ] , [ field_phase20 ] , [ field_amp500 ] , and [ field_phase500 ] , although they do not overlap with those of the 1d bose gas even in the large coupling case .
for instance , the square - amplitude profile vanishes at the origin , and it has a density notch around at the origin and small ripples at the shoulders of the notch ; the phase profiles change signs at the origin , etc ..
for a given quantum state we denote the expectation value of the hamiltonian measured from the ground state energy by @xmath315 , and the square root of the variance of the expectation value of the hamiltonian by @xmath316 .
we define the quantum speed limit time @xmath317 @xcite by @xmath318 we remark that the quantum speed limit time gives a typical lifetime of a generic quantum state .
let us evaluate the quantum speed limit time @xmath317 for the quantum dark soliton state @xmath319 in the weak coupling limit where we send the coupling constant @xmath3 to zero .
we solve the bethe ansatz equations ( [ bae ] ) in the weak coupling limit .
here we recall that the bethe quantum numbers for one - hole excitations are defined by eq .
( [ eq : ij - one - hole ] ) .
let us consider an expansion of function @xmath320 which is valid if the absolute value of @xmath62 is larger than 1 .
@xmath321 here @xmath322 for @xmath323 while @xmath324 for @xmath325 .
applying this expansion we show in the weak coupling case ( @xmath326 ) @xmath327 we have the solution of one - hole excited state @xmath328 approximately as follows .
@xmath329 therefore , the one - hole excitation @xmath328 has the energy eigenvalue @xmath330 .
let us denote by @xmath331 the expectation value of the energy eigenvalues for the quantum dark soliton state @xmath319 .
we have @xmath332 and hence @xmath333 since the ground - state energy is given by 0 in the weak coupling limit , we denote it also by @xmath315 .
we calculate the variance of the energy eigenvalues @xmath319 as @xmath334 .
we thus have the square root of the variance , @xmath316 , as @xmath335 it follows that the quantum speed limit time for the quantum dark soliton state in the weak coupling limit is given by that for the inverse of the square root of variance @xmath336 @xmath337 here we remark that the quantum speed limit time @xmath317 is constant in the weak coupling limit where only the coupling constant @xmath3 approaches zero while the density @xmath42 and the particle number @xmath1 are kept constant .
the estimates of the relaxation time ( or collapse time ) for the density notch in the density profile of quantum dark soliton for the case of @xmath338 are plotted against coupling constant @xmath3 in the double logarithmic scale in fig .
[ rt_n20 ] . by observing the time evolution of the density profile of quantum dark soliton for a given value of the coupling constant @xmath3 we estimated the relaxation time by the time when the smallest value of the density notch in the density profile reaches the value of @xmath339 .
here we recall that at the initial time @xmath243 the density profile of quantum dark soliton is consistent with that of classical dark soliton , and then the bottom level of the density profile of quantum dark soliton increases gradually in time evolution . in fig .
[ rt_n20 ] we observe that the collapse time @xmath340 increases as the coupling constant @xmath3 approaches zero .
it is almost inversely proportional to the coupling constant @xmath3 in the weak coupling regime : @xmath341 hence we suggest that it becomes infinitely large in the weak coupling limit .
the quantum speed limit time @xmath317 for quantum dark soliton states approaches a non - zero constant value when we send the coupling constant @xmath3 to zero , if we keep the system size @xmath5 and the number of particles @xmath1 fixed .
in fact , in the weak coupling limit , all the solutions of the bethe ansatz equations are given by some integral multiples of @xmath342 .
hence , the quantum speed limit time of any quantum state is given by some finite value .
on the other hand , the collapse time of the density notch in the density profiles of quantum dark soliton is approximately given by the inverse of the coupling constant , @xmath343 , in the very weak coupling case .
therefore , the collapse time of the density notch in the density profile of quantum dark soliton is much longer than the quantum speed limit time when the coupling constant @xmath3 is very small . for the quantum dark soliton state with @xmath1 particles in the system size
@xmath5 we have approximately @xmath344 in the case of @xmath288 we have a rough estimate : @xmath345 . here
we recall that we have set @xmath346 . in fig .
[ rt_n20 ] the collapse time is given by approximately 10.0 for @xmath276 .
thus , for @xmath347 the collapse time of the density profile in the quantum dark soliton state is much longer than the quantum speed limit time .
we conclude that the density notch in the density profile of quantum dark soliton , i.e. that of state @xmath319 , has a much longer lifetime than a generic quantum state when the coupling constant @xmath3 is very small but nonzero in the 1d bose gas interacting through the repulsive delta - function potentials .
the authors would like to thank dr .
a. del campo for useful comments , in particular , on the quantum speed limit .
the present research is partially supported by jsps kakenhi ( grants - in - aid for scientific research ) grant no . 15k05204 .
we define the elliptic integrals of the first , second and third kind , denoted by @xmath348 , @xmath349 and @xmath350 , respectively , as follows . @xmath351
the complete elliptic integrals of the first , second and third kind , denoted by @xmath145 , @xmath158 and @xmath352 , respectively , are derived by substituting @xmath353 in ( [ ei1 ] ) , ( [ ei2 ] ) and ( [ ei3 ] ) , i.e. @xmath354 , @xmath355 and @xmath356 , respectively .
we define the hypergeometric function @xmath357 in terms of the gamma function @xmath358 by @xmath359 the complete elliptic integrals of the first and second kind , @xmath145 and @xmath158 , are expressed as @xcite @xmath360 jacobi s @xmath209 function is defined by the inverse of the elliptic integral of the first kind @xmath361 we therefore have @xmath362 .
jacobi s @xmath363 and @xmath156 functions are related to @xmath364 as @xmath365 hereafter we shall often suppress the modulus @xmath17 .
jacobi s amplitude function @xmath366 is defined by the inverse of @xmath367 as @xmath368 we express @xmath369 and @xmath370 as @xmath371 and @xmath372 . making the transformation of variables @xmath373 in the integral expression of @xmath158 such
as we have @xmath374 due to the periodicity and the parity invariance of @xmath156 function , we have @xmath375 we introduce the complementary modulus @xmath196 by @xmath376 .
we define @xmath198 and @xmath199 by @xmath377 and @xmath378 , respectively . here
we remark that jacobi s elliptic functions have quarter periods @xmath148 and @xmath379 .
we can show legendre s relation @xcite @xmath380 let us define @xmath381 and @xmath382 by @xmath383 and @xmath384 , respectively @xcite .
we have @xmath385 we can show @xmath386 by making transformation of variable @xmath387 in eq .
, we have @xmath388 we define jacobi s zeta function @xmath389 by @xmath390 we introduce jacobi s theta function by @xmath391 in terms of @xmath392 we have @xmath393 by taking the logarithmic derivative of jacobi s theta function we have @xmath394 let us take a complex number @xmath395 satisfying the following relation : @xmath396 we can express the elliptic integral of the third kind in terms of jacobi s zeta function as follows .
@xmath397 we express the complete elliptic integral of the third kind as follows .
@xmath398 we define jacobi s eta function by @xmath399 in terms of eta and theta functions jacobi s @xmath209 , @xmath363 and @xmath156 functions are given by @xmath400 jacobi s imaginary transformation for @xmath209 , @xmath363 and @xmath156 functions is given by @xmath401 jacobi s imaginary transformation of the elliptic integral of the second kind is given by @xmath402 we can derive jacobi s imaginary transformation of zeta function from @xmath403 it follows that we have @xmath404 therefore , in terms of @xmath405 , we have @xmath406 by taking the logarithmic derivative of we expand jacobi s zeta function through as follows .
in order to reduce the short - wavelength oscillations or ripples such as shown in the graph of the dirichlet kernel @xmath52 we may modify the truncated sum with some weights .
let us denote by @xmath17 the wavenumbers @xmath408 for integers @xmath42 .
we take an integer @xmath409 and set @xmath410 as a reference wavenumber .
we define @xmath411 by @xmath412 and denote the difference of wavenumbers by @xmath413 . with the gaussian weights @xmath414 we modify the truncated sum as @xmath415 in terms of @xmath411 we express it as @xmath416 we can show that the sum over @xmath411 in does not become very large in the region of @xmath62 satisfying @xmath417 . if @xmath418 , each term in the sum has its absolute value larger than 1 , and hence the sum can be much larger than @xmath58 which is the number of the terms in the sum . here
we remark that the bound @xmath419 is given by @xmath420 if we set @xmath421 .
we may therefore assume that the sum over @xmath411 in is approximately constant in some region near the origin such as @xmath422 .
evaluating the sum by an integral we approximate the gaussian weighted sum @xmath423 as @xmath424 here the error function @xmath425 is given by @xmath426 typically , by putting @xmath427 , i.e. half the period of the short - distance oscillations or ripples in the graph of the function @xmath52 , the ripples disappear , and the gaussian weighted sum @xmath428 approximately becomes a smooth gaussian wave packet with the standard deviation @xmath429 .
d. c. wadkin - snaith and d. m. gangardt , phys .
* 108 * , 085301 ( 2012 ) .
t. karpiuk , p. deuar , p. bienias , e. witkowska , k. pawlowski , m. gajda , k. rzazewski and m. brewczyk , phys .
* 109 * , 205302 ( 2012 ) .
a. j. leggett , rev .
* 73 * ( 2001 ) 307356 .
m. rigol , v. dunjko and m. olshanii , nature * 452 * , 854 ( 2008 ) .
s. goldstein , t. hara and h. tasaki , new j. phys . * 17 * , 045002 ( 2015 ) l. mandelstam and i. g. tamm , j. phys .
( ussr ) * 9 * , 249 ( 1945 ) .
n. margolus and l. b. levitin , physica d * 120 * , 188 ( 1998 ) .
v. giovannetti , s. lloyd , and l. maccone , phys .
a * 67 * , 052109 ( 2003 ) .
m. d. girardeau and e. m. wright , phys .
. lett . * 84 * , 5691 ( 2000 ) .
j. sato , e. kaminishi , and t. deguchi , arxiv:1303.2775 [ cond-mat.quantum-gas ] .
girardeau and e.m .
wright , phys .
84 * , 5239 ( 2000 ) .
e. b. kolomeisky , t. j. newman , j. p. straley , and x. qi , phys .
lett . * 85 * , 1146 ( 2000 ) .
t. kawata and h. inoue , j. phys .
44 * , 1722 ( 1978 ) . c. r. nohl , ann
. phys . * 96 * , 234 ( 1976 ) ; m. wadati , m. sakagami , j. phys .
jpn . * 53 * , 1933 ( 1984 ) ; m. wadati , a. kuniba , t. konishi , j. phys .
. jpn . * 54 * , 1710 ( 1985 ) ; m. wadati , a. kuniba , j. phys .
soc . jpn . *
55 * , 76 ( 1986 ) .
y. lai et al . , phys .
a * 40 * , 854 ( 1989 ) .
s. j. carter et al . , phys .
58 * , 1841 ( 1987 ) .
m. rosenbluh and r.m .
shelby , phys .
lett . * 66 * , 153 ( 1991 ) ; p. d. drummond et al . , nature * 365 * , 307 ( 1993 ) .
k " artner et al . , phys .
a * 53 * , 454 ( 1996 ) .
j. f. corney et al . , j. opt .
b * 18 * , 153 ( 2001 ) .
a. d. martin and j. ruostekoski , phys .
lett . * 104*,194102 ( 2010 ) .
new j. phys .
* 12 * , 055018 ( 2010 ) .
e. t. copson , _ an introduction to the theory of functions of a complex variable _ ( oxford university press , oxford , 1978 ) .
m. gaudin , `` _ la fonction donde de bethe _ '' , masson ( paris ) ( 1983 ) ; v. e. korepin , commun . math
. phys . * 86 * , 391 ( 1982 ) .
n. a. slavnov , teor .
* 79 * , 232 ( 1989 ) ; * 82 * , 389 ( 1990 ) .
p. calabrese and j .- s .
caux , j. stat .
( 2007 ) p08032 .
t. kojima , v.e .
korepin , n.a .
slavnov , commun .
* 188 * , 657 ( 1997 ) j .- s .
caux , p. calabrese and n. a. slavnov , j. stat . mech .
p01008 ( 2007 ) . | we present a series of quantum states that are characterized by dark solitons of the nonlinear schrdinger equation ( i.e. the gross - pitaevskii equation ) for the one - dimensional ( 1d ) bose gas interacting through the repulsive delta - function potentials .
the classical solutions satisfy the periodic boundary conditions and we simply call them classical dark solitons . through exact solutions
we show corresponding aspects between the states and the solitons in the weak coupling case : the quantum and classical density profiles completely overlap with each other not only at an initial time but also at later times over a long period of time , and they move together with the same speed in time ; the matrix element of the bosonic field operator between the quantum states has exactly the same profiles of the square amplitude and the phase as the classical complex scalar field of a classical dark soliton not only at the initial time but also at later times , and the corresponding profiles move together for a long period of time .
we suggest that the corresponding properties hold rigorously in the weak coupling limit . furthermore , we argue that the lifetime of the dark soliton - like density profile in the quantum state becomes infinitely long as the coupling constant approaches zero , by comparing it with the quantum speed limit time .
thus , we call the quantum states _ quantum dark soliton states_. |
after the recent discovery of neutrino oscillations by the superkamiokande experiment @xcite , supersymmetric ( susy ) models with left - right symmetric gauge groups have attracted a great deal of attention .
these models provide a natural framework for implementing the seesaw mechanism @xcite which explains the existence of the small neutrino masses .
the implications of these models have been considered in ref.@xcite , in the case of the gauge group @xmath15 , and in ref.@xcite for the susy pati - salam ( ps ) @xcite model based on the gauge group @xmath16 .
it was shown that they lead to a constraint version of the minimal supersymmetric standard model ( mssm ) .
0.25 cm recently , it was demonstrated @xcite that the n=1 susy ps model can emerge as the effective four dimensional limit of brane models from type i string theory .
this result provides further motivation for considering the phenomenological and cosmological implications of this model . also , it is know @xcite that the gauge symmetry @xmath1 can arise from the weakly coupled heterotic string as well .
0.25 cm hybrid inflation @xcite has been extensively studied @xcite in the case of the susy model based on the gauge group @xmath17 .
however , in trying to extend this scheme to @xmath1 , we encounter the following difficulty .
the spontaneous symmetry breaking of @xmath1 to the standard model gauge group @xmath18 leads to the existence of topologically stable magnetic monopole solutions .
this is due to the fact that the second homotopy group of the vacuum manifold @xmath19 is non - trivial and equal to the set of integers @xmath20 .
these monopoles carry two units of dirac magnetic charge @xcite .
inflation is terminated abruptly when the system reaches a critical point ( instability ) on the inflationary trajectory and is followed by a ` waterfall ' regime during which the spontaneous breaking of @xmath1 occurs .
the appropriate higgs fields develop their non - zero vacuum expectation values ( vevs ) starting from zero and they can end up at any point of the vacuum manifold with equal probability . as a consequence ,
magnetic monopoles are copiously produced @xcite by the kibble mechanism @xcite leading to a cosmological disaster .
0.25 cm in this paper , we propose a specific susy model based on @xmath1 which avoids this cosmological catastrophe .
this is achieved by including a non - renormalizable term in the part of the superpotential involving the inflaton system and causing the breaking of @xmath1 .
it is worth mentioning that an analogous non - renormalizable term was also used in ref.@xcite for the same purpose . in that case , however , the leading renormalizable term was eliminated by imposing a discrete symmetry .
here , we keep this leading term along with the non - renormalizable contribution .
the picture that emerges turns out to be considerably different .
in particular , there exists a non - trivial ( classically ) flat direction along which @xmath1 is spontaneously broken with the appropriate higgs fields acquiring constant values .
this direction can be used as inflationary trajectory with the necessary inclination obtained from one - loop radiative corrections @xcite in contrast to the model of ref.@xcite , where a classical inclination was present .
another difference is that here the termination of inflation is abrupt ( as in the original hybrid inflationary scenario ) and not smooth as in ref.@xcite .
nevertheless , no magnetic monopoles are formed in this transition since @xmath1 is already broken during inflation .
0.25 cm we show that , for a certain range of parameters , the system always passes from the above mentioned inflationary trajectory before falling into the susy vacuum .
thus , the magnetic monopole problem is solved for all initial conditions .
it is interesting to note that the idea of breaking the gauge symmetry before ( or during ) inflation in order to avoid monopoles was also employed in ref.@xcite .
however , the monopole problem was solved only for a certain ( wide ) class of initial values of the fields .
0.25 cm the constraints on the quadrupole anisotropy of the cosmic microwave background radiation from the cosmic background explorer ( cobe ) @xcite measurements can be easily met with natural values ( of order @xmath2 ) of the relevant coupling constant and a grand unification theory ( gut ) scale @xmath21 close to ( or somewhat smaller than ) the susy gut scale . note that the mass scale in the model of ref.@xcite , which uses only renormalizable couplings in the inflationary superpotential , is considerably smaller .
our model possesses a number of other interesting features too .
the @xmath5-problem of mssm is solved @xcite via a peccei - quinn ( pq ) symmetry which also solves the strong cp problem .
although the baryon ( @xmath22 ) and lepton ( @xmath23 ) numbers are explicitly violated , the proton life time is considerably higher than the present experimental limits .
light neutrinos acquire masses by the seesaw mechanism and the baryon asymmetry of the universe can be generated through a primordial leptogenesis @xcite .
the gravitino constraint @xcite on the reheat temperature , the low deuterium abundance limits @xcite on the baryon asymmetry of the universe and the requirement of almost maximal @xmath6 mixing from superkamiokande @xcite can be met for @xmath5- and @xmath24-neutrino masses restricted by superkamiokande and the large angle msw solution of the solar neutrino puzzle respectively .
the required values of the relevant coupling constants are more or less natural .
0.25 cm the plan of the paper is as follows . in sec.2
, we introduce our susy model which is based on the gauge group @xmath1 and motivate the inclusion of a non - renormalizable coupling in the inflaton sector of the theory .
the full superpotential and its global symmetries are then discussed together with the solution of the @xmath5-problem via the pq symmetry of the model . in sec.3 ,
the hybrid inflationary scenario is studied in detail in this model .
the structure of the potential is carefully analyzed .
we calculate the one - loop radiative corrections along the inflationary trajectory by first deriving ( see appendix ) the mass spectrum of the theory during inflation .
the parameters of the model are then restricted by employing cobe results . in sec.4
, we discuss the reheating process following inflation , neutrino masses and mixing and baryogenesis via leptogenesis .
we show that all the relevant constraints can be satisfied with natural values of the coupling constants .
we summarize our conclusions in sec.5 .
in the susy ps model , the left - handed quark and lepton superfields are accommodated in the following representations : @xmath26 where the subscript @xmath27 denotes the family index @xcite .
the @xmath1 gauge symmetry can be spontaneously broken to @xmath18 by a pair of higgs superfields @xmath28 acquiring non - vanishing vevs in the right - handed neutrino direction , @xmath29 , @xmath30 . 0.25 cm the two low energy higgs doublets of the mssm are contained in the following representation : @xmath31 after @xmath1 breaking , the bidoublet higgs field @xmath32 splits into two higgs doublets @xmath33 , @xmath34 , whose neutral components subsequently develop weak vevs @xmath35 and @xmath36 with @xmath37
. 0.25 cm the breaking of @xmath1 can be achieved by introducing a gauge singlet superfield @xmath38 , which has a trilinear ( renormalizable ) coupling to @xmath39 , @xmath40 .
the resulting scalar potential automatically possesses an in - built ( classically ) flat direction along which inflation can take place @xcite with the system driven by an inclination from one - loop radiative corrections @xcite .
the @xmath1 gauge symmetry is restored along this trajectory and breaks spontaneously only at the end of inflation when the system falls towards the susy vacua .
this transition leads to a cosmologically unacceptable copious production of doubly charged magnetic monopoles @xcite .
one way to resolve this problem is to use , for inflation , another flat direction in which @xmath1 is already broken .
such a trajectory naturally appears if we include the next order non - renormalizable superpotential coupling of @xmath38 to @xmath39 , @xmath40 too .
we find that , together with the usual flat direction with unbroken @xmath1 , an extra flat trajectory along which @xmath1 is spontaneously broken to @xmath18 emerges .
the termination of inflation can then take place with @xmath1 already broken and no monopoles being produced .
0.25 cm an important issue is the generation of the @xmath5-term of mssm .
this could be easily achieved @xcite by coupling @xmath38 to the electroweak higgs superfields and using the fact that @xmath38 , after gravity - mediated susy breaking , develops a vev . however , this is not totally satisfactory since the inflaton decays into electroweak higgs superfields via an unsuppressed ( renormalizable ) coupling . as a consequence , the gravitino constraint @xcite on the reheat temperature implies @xcite unnaturally small values for the relevant coupling constants ( of order @xmath41 or so ) .
we thus prefer to follow here ref.@xcite and impose a pq symmetry on the superpotential by introducing a pair of gauge singlet superfields @xmath42 , @xmath43 .
the pq breaking occurs at an intermediate scale by the vevs of @xmath42 , @xmath43 and the @xmath5-term is generated via a non - renormalizable coupling of @xmath42 and @xmath32 .
the inflaton can be made to decay into right - handed neutrinos by introducing into the scheme quartic ( non - renormalizable ) superpotential couplings of @xmath40 to @xmath44 and the gravitino constraint can be satisfied with more natural values of the parameters .
it should be noted that the presence of these quartic terms is anyway necessary for generating intermediate scale masses for the right - handed neutrinos and , thus , masses for the light neutrinos via the seesaw mechanism .
finally , in order to give superheavy masses to @xmath45 and @xmath46 , we introduce @xcite an @xmath9 6-plet superfield @xmath47 which , under @xmath18 , splits into @xmath48 and @xmath49 .
0.25 cm the superpotential of the model , which incorporates all the above couplings , is @xmath50 where @xmath51 gev is a superheavy string scale .
also , @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 and @xmath57 can be made positive by field redefinitions , while @xmath58 is chosen positive for simplicity ( it could be genuinely complex ) . here , we are in the basis where @xmath59 s are diagonal .
0.25 cm note that the existence of the third ( non - renormalizable ) coupling in the right hand side of eq.([eq : superpot ] ) is an automatic consequence of the first two couplings which constitute the standard superpotential for hybrid inflation .
indeed , the operator @xmath60 is neutral under all the r and non - r symmetries of this standard superpotential and , thus , the above coupling , which is crucial for our inflationary scheme , can not be forbidden .
our only assumption is then that the dimensionless coefficient @xmath58 of this non - renormalizable coupling is of order unity so that it can be comparable to the trilinear coupling in the standard superpotential whose coefficient @xmath53 is typically relatively small ( @xmath61 ) .
this non - renormalizable coupling can then play an important role in the inflationary scenario .
the non - renormalizable couplings @xmath62 with @xmath63 are also allowed .
they are , however , subdominant to the leading non - renormalizable coupling ( with @xmath64 ) in the relevant region of the field space even if their coefficients are of order one . the r symmetry ( see below ) of the standard superpotential , extended to higher orders , is needed for avoiding terms non - linear in @xmath38 which can destroy inflation .
the couplings , then , permitted between @xmath38 , @xmath39 , @xmath40 are the ones already mentioned modulo arbitrary multiplications by @xmath65 , @xmath66 .
the new couplings do not affect the potential for inflation .
0.25 cm in addition to @xmath1 , the superpotential in eq.([eq : superpot ] ) possesses two global anomalous symmetries , a r symmetry @xmath67 and a pq symmetry @xmath68 .
the r and pq charges of the superfields are assigned as follows : @xmath69 to avoid undesirable mixing of @xmath70 and @xmath32 or @xmath71 and @xmath39 , we also impose a discrete @xmath72 symmetry ( know as ` matter parity ' ) , under which @xmath70 and @xmath71 change sign
. 0.25 cm additional superpotential terms allowed by the symmetries of the model are @xmath73 modulo arbitrary multiplications by non - negative powers of the combinations @xmath74 , @xmath65 , @xmath66 ( this applies to the terms in eq.([eq : superpot ] ) too ) .
note that the @xmath9 indices in all couplings except the last three in eq.([terms ] ) and the combinations @xmath65 , @xmath66 are contracted between @xmath75 s and @xmath76 s , while in these terms and combinations the four @xmath9 indices of the @xmath75 s or @xmath76 s are contracted with an @xmath77 .
the soft susy breaking and instanton effects explicitly break @xmath67 to @xmath78 , under which @xmath79 , and @xmath68 to @xmath80 .
these two discrete symmetries are spontaneously broken by the vevs of @xmath81 and would create a domain wall problem if the pq transition took place after inflation . when @xmath39 , @xmath40 , @xmath42 and @xmath43 acquire non - vanishing vevs , the symmetry which is left unbroken is @xmath82 .
0.25 cm we can assign baryon number @xmath83 to all color triplets ( antitriplets ) . recall that there are ( anti)triplets not only in @xmath84 but also in @xmath85 .
lepton number is then defined via @xmath86 .
@xmath22 ( and @xmath23 ) violation comes from the last three terms in eq.([terms ] ) ( and the combinations @xmath65 , @xmath66 ) which give couplings like @xmath87 , @xmath88 with appropriate coefficients . also , the terms @xmath89 and @xmath90 give rise to the @xmath22 ( and @xmath23 ) violating couplings @xmath91 , @xmath92 .
all other combinations are @xmath22 ( and @xmath23 ) conserving since @xmath75 s are contracted with @xmath76 s .
0.25 cm the dominant contribution to proton decay comes from effective dimension five operators generated by one - loop diagrams with two of the @xmath93 , @xmath45 or one of the @xmath93 , @xmath45 and one of the @xmath94 , @xmath95 circulating in the loop .
the amplitudes corresponding to these operators are estimated to be at most of order @xmath96 ( @xmath97 is the gravitino mass ) .
this makes the proton practically stable .
furthermore , the dominant contribution to the majorana mass term of light neutrinos comes from @xmath98 and is utterly small .
so the seesaw mechanism is the only source of light neutrino masses .
0.25 cm the @xmath5-term is generated , as mentioned , by a non - renormalizable superpotential coupling which contains the electroweak higgs and the @xmath42 superfield after the breaking of @xmath68 by @xmath99 , @xmath100 .
the relevant part of the scalar potential for the pq breaking is given by @xcite @xmath101 where @xmath102 is the dimensionless coefficient of the soft susy breaking term corresponding to the @xmath103 term in eq.([eq : superpot ] ) . here , the phases @xmath104 , @xmath105 and @xmath106 of @xmath102 , @xmath42 and @xmath43 are taken to satisfy the relation @xmath107 and @xmath108 , @xmath109 are assumed equal which minimizes the potential . for @xmath110 ,
the absolute minimum of this potential is given by @xcite @xmath111 hence the pq symmetry breaking scale is of order @xmath112 gev and the @xmath5-term of the mssm is @xmath113 as desired .
0.25 cm note that the zero temperature pq potential ( in eq.([potn0 ] ) ) , shown in fig.1 , possesses two local minima , the trivial one at @xmath114 and the pq minimum which , for @xmath115 , is the absolute minimum .
these minima are separated by a sizable potential barrier which prevents a successful transition from the trivial to the pq vacuum .
taking the one - loop temperature corrections @xcite to the potential into account , one can show that the pq vacuum remains the absolute minimum at least for temperatures below the reheat temperature @xmath116 gev .
the trivial vacuum is still protected by a potential barrier .
we , thus , conclude that if , after inflation , the system emerges in the trivial vacuum the completion of the pq transition will be practically impossible .
we are obliged to assume that the pq symmetry is already broken before or during inflation .
the pq vacuum then remains stable after inflation and reheating .
there is yet another reason which disfavors a pq transition after inflation .
the vevs of @xmath42 , @xmath117 break spontaneously the @xmath78 symmetry ( @xmath79 ) and the @xmath80 subgroup of @xmath68 which is left unbroken by instantons .
this would lead to disastrous domain walls in the universe .
0.25 cm in concluding this section , we emphasize that the extra couplings which we incorporated into our basic inflationary superpotential ( see eq.([eq : superpot ] ) ) are simple , quite general , and well - motivated . in particular , the couplings @xmath118 , @xmath119 were included in order to solve the @xmath5-problem .
however , they also provide a solution to the strong cp problem .
moreover , the presence of @xmath120 is necessary for the generation of the right - handed neutrino masses .
it is an extra benefit , though , that , via these same couplings , the inflaton decays to right - handed neutrinos , thereby leading to a successful and ` natural ' ( no tiny couplings ) reheating of the universe with the observed baryon asymmetry generated via leptogenesis ( see below ) .
finally , @xmath89 , @xmath121 were added for merely giving superheavy masses to @xmath45 , @xmath46 .
it is worth mentioning that the emerging theory is ` natural ' in the sense that all couplings which are consistent with a simple and general set of global symmetries ( @xmath67 , @xmath68 and @xmath72 ) can be allowed without jeopardizing our inflationary scenario or leading into trouble with other requirements such as proton stability .
the part of the superpotential in eq.([eq : superpot ] ) which is relevant for inflation is given by @xmath122 where @xmath52 is a superheavy mass scale close to the gut scale .
note that the rest of the superpotential in eq.([eq : superpot ] ) does not affect the inflationary dynamics . the scalar potential obtained from @xmath123 is given by @xmath124 where the scalar components of the superfields are denoted by the same symbols as the corresponding superfields .
vanishing of the d - terms is achieved with @xmath125 ( @xmath40 ( @xmath39 ) lies in the @xmath126 ( @xmath94 ) direction ) . restricting ourselves to this direction and performing an appropriate r transformation
, we can bring the complex field @xmath38 to the real axis , @xmath127 , where @xmath128 is a normalized real scalar field .
an @xmath129-independent flat trajectory suitable for hybrid inflation driven by radiative corrections is obtained in the direction @xmath130 along which the potential takes the form @xmath131 ^ 2 + \kappa^2 \sigma^2 \vert h^c \vert^2 \left[1 - \frac{2 \beta}{\kappa m_s^2 } \vert h^c \vert^2 \right]^2 .\label{eq : inflpot}\ ] ] we can now rewrite this potential in terms of the dimensionless variables @xmath132 we get @xmath133 where @xmath134 .
this potential is a simple extension of the standard potential for susy hybrid inflation ( obtained from eq.([vtilde ] ) by putting @xmath135 ) and appears generically in a wide class of models which incorporate the leading non - renormalizable correction to the standard hybrid inflationary superpotential .
( recall that this correction is naturally present as we explained . )
it is , thus , interesting to study in some detail the structure of this potential and the new inflationary possibilities which typically emerge .
0.25 cm for constant @xmath136 , the potential in eq.([vtilde ] ) has the following extrema @xmath137 . \label{y3}\end{aligned}\ ] ] note that the first two extrema are @xmath128-independent .
as it turns out , @xmath138 is a local minimum ( maximum ) if @xmath139 , while @xmath140 is a local minimum ( maximum ) if @xmath141 @xmath142 .
inflation will take place when the system is trapped along the @xmath140 minimum .
we restrict ourselves to @xmath143 since , in this case , the inflationary trajectory ( at @xmath140 ) is destabilized in the direction of the real part of @xmath144 before @xmath145 reaches zero .
inflation can then be terminated with the system falling towards the susy vacua ( see below ) following the direction @xmath146 , where the potential is given by eq.([vtilde ] ) .
the potential at @xmath140 is @xmath147 , while at @xmath138 is @xmath148 .
so that , for @xmath149 , the extremum at @xmath138 lies higher ( lower ) than the one at @xmath140 .
0.25 cm for @xmath150 , the discriminant @xmath151 appearing under the square root in eq.([y3 ] ) is negative if @xmath136 lies between the positive numbers @xmath152 and non - negative otherwise .
so , for @xmath153 , the extrema at @xmath154 do not exist .
note , however , that @xmath155 does not necessarily guarantee the presence of these extrema .
this requires that , in addition , the right hand side of eq.([y3 ] ) is non - negative .
an important ingredient on which this requirement depends is the sign of the expression @xmath156 which is positive ( negative ) for @xmath157 .
one can show that , for @xmath158 , @xmath159 ( equality holds for @xmath160 ) and @xmath161 .
also , @xmath162 for @xmath163 .
0.25 cm for @xmath164 , we always have @xmath155 .
in addition , note that @xmath165 for @xmath166 , @xmath167 for @xmath168 and @xmath169 always .
it is interesting to point out that , at @xmath170 , above which the extremum at @xmath140 turns into a local minimum , @xmath171 coincides with @xmath140 .
the minima at @xmath154 , for @xmath172 , become supersymmetric , i.e. , @xmath173 . for reasons to become obvious later , we consider the @xmath174 minimum at @xmath172 as the relevant susy vacuum of the theory
. 0.25 cm taking into account these facts , we can distinguish five cases with qualitatively different structure of the potential : @xmath175 for @xmath176 , we have @xmath177 .
one can then show that , for fixed @xmath178 , there exist two local minima at @xmath138 and @xmath140 ( the interesting inflationary trajectory ) and a local maximum at @xmath171 between them ( see fig.2 ) .
for @xmath136 between @xmath179 and @xmath180 , the trivial extremum at @xmath138 becomes a local maximum and the extremum at @xmath171 disappears ( see fig.3 ) . in this range of @xmath136 , the system can freely fall into the desirable ( inflationary ) minimum at @xmath140 even if it was initially along the trivial trajectory at @xmath138 ( remember that the extremum at @xmath140 lies lower than the one at @xmath138 in this case ) .
as we further decrease @xmath136 to become smaller than @xmath180 , a pair of two new extrema , a local minimum at @xmath174 and a local maximum at @xmath171 , are created between @xmath138 and @xmath140 .
as @xmath136 crosses @xmath181 , the local maximum at @xmath171 crosses @xmath140 becoming a local minimum ( see fig.4 ) . at the same time , the local minimum at @xmath140 turns into a local maximum and inflation is terminated with the system falling into the local minimum at @xmath174 which at @xmath172 develops into a susy vacuum ( see below ) .
0.2 cm @xmath182 for @xmath183 , we have @xmath184 .
the situation for @xmath178 and @xmath185 is similar to the previous case .
for @xmath136 between @xmath179 and @xmath186 , the @xmath138 extremum becomes a local maximum and a local minimum at @xmath174 appears between @xmath138 and @xmath171 . as @xmath136 decreases below @xmath186 , the extrema at @xmath154 disappear and there exists no obstacle for the system to fall to @xmath140 even if it was initially at @xmath138 .
the extrema at @xmath154 reappear as @xmath136 becomes smaller than @xmath180 .
0.2 cm @xmath187 for @xmath188 , @xmath189 .
the behavior of the potential for @xmath190 and @xmath191 is similar to the previous cases . for @xmath192
, however , the extremum at @xmath138 becomes a local maximum and a local minimum at @xmath174 appears between @xmath138 and @xmath171 .
notice that , in this case , although the extremum at @xmath140 lies lower than the one at @xmath138 , there is no range of @xmath136 where the system can fall into @xmath140 if it was initially at @xmath138 .
instead , it ends up directly in @xmath174 from @xmath138 and monopoles can be copiously produced .
of course , if the system happens to be at @xmath140 from the beginning , there is no production of monopoles .
0.2 cm @xmath193 for @xmath194 , the situation is exactly as in case @xmath187 with the only difference that the extremum at @xmath140 now lies higher than the one at @xmath138 .
0.2 cm @xmath195 for @xmath196 , we have @xmath197 .
it turns out that , for @xmath198 , the local minima at @xmath138 and @xmath140 ( which lies higher ) are again separated by a local maximum at @xmath171 .
as @xmath136 crosses @xmath181 , the @xmath171 local maximum turns into minimum and crosses @xmath140 which becomes a local maximum .
there is then no obstacle to keep the system from falling into @xmath138 even if it was at @xmath140 .
subsequently , when @xmath136 becomes smaller than @xmath179 , @xmath138 turns into a local maximum and the system falls into @xmath174 with a copious production of magnetic monopoles .
0.25 cm we will restrict ourselves here to the first two cases above @xmath199 .
we saw that , in these cases , even if the system starts along the trivial valley at @xmath138 , it always falls into the ( classically ) flat direction at @xmath140 . the relevant part of inflation can then take place along this trajectory with the inflaton being driven by radiative corrections @xcite .
so , @xmath1 is already broken during inflation and there is no production of magnetic monopoles at the end of inflation where the system falls into the @xmath174 minimum .
case @xmath187 could also solve the monopole problem provided the system starts at @xmath140 .
case @xmath193 , although quite similar to case @xmath187 , is more tricky requiring further study .
the reason is that , since @xmath140 lies higher than @xmath138 , the system oscillates over the local maximum at @xmath138 after falling from @xmath140 . finally , case @xmath195 is always unacceptable since the system , for all initial conditions , falls to @xmath174 from @xmath138 and the copious production of monopoles is unavoidable .
0.25 cm as we already mentioned , after inflation ends , the system falls into the minimum at @xmath174 which , at @xmath172 , develops into the final susy vacuum of the theory
. however , the system could , in principle , fall into the minimum at @xmath171 which appears only after the instability of the inflationary trajectory at @xmath200 is reached .
( the minimum at @xmath171 also develops into a susy vacuum at @xmath172 . )
we will argue that this does not happen .
for the values of the parameters used here , the potential barrier separating the inflationary path at @xmath140 and the minimum at @xmath174 is considerably reduced in the last e - folding or so .
( the peak of this barrier coincides with the maximum at @xmath171 . ) as a consequence , the rate per unit volume and time of forming bubbles of the @xmath174 minimum ceases to be exponentially suppressed .
an order of magnitude estimate then shows that the decay of the false vacuum at @xmath140 to the minimum at @xmath174 is completed within a fraction of one e - folding .
this happens before the appearance of the minimum at @xmath171 , i.e. , before the system reaches the critical point at @xmath200 ( but very close to this point ) .
moreover , in the last stages of inflation , the above barrier is small enough to be overcome by the inflationary density perturbations .
this can also accelerate the completion of this transition .
0.25 cm to avoid confusion we should mention here that @xmath201 is not an extra free parameter .
it depends on the coupling @xmath53 and the superheavy mass scale @xmath52 ( we put @xmath202 ) .
the values of @xmath53 and @xmath52 will be related by calculating the quadrupole anisotropy of the cosmic microwave background radiation @xmath203 as a function of the number of e - foldings of our present horizon @xmath204 and compare it with the measurements of cobe @xcite .
the parameter @xmath201 then becomes a function of the basic coupling constant of the scheme @xmath53 .
so , searching for solutions with @xmath201 in the desirable range and also satisfying all the other requirements which we will discuss below is a highly non - trivial task .
0.25 cm as already mentioned , the interesting part of inflation takes place when the system is trapped along the trajectory at @xmath140 .
inflation is driven by the constant classical energy density on this trajectory which also breaks susy .
this breaking gives rise to non - trivial radiative corrections @xcite which lift the ( classical ) flatness of this trajectory producing a necessary inclination for driving the inflaton towards the susy vacua .
as will be seen later , the slow - roll conditions @xcite are satisfied and inflation continues essentially till @xmath136 reaches @xmath181 , where the inflationary trajectory is destabilized . to calculate the one - loop radiative corrections at @xmath140 we need to construct the mass spectrum of the theory on this path where both @xmath1 and susy are broken .
details of the calculation can be found in the appendix .
we summarize the results in table 1 .
[ ht ] [ cols="^,^",options="header " , ] table 1 : the mass spectrum of the model as the system moves along the inflationary trajectory at @xmath140 .
the parameter @xmath205 is the vev of @xmath94 , @xmath126 on this trajectory and @xmath206 is the @xmath1 gauge coupling constant .
0.3 cm using this spectrum , we can now calculate the one - loop radiative correction to the potential along the inflationary trajectory from the coleman - weinberg formula @xcite @xmath207 where the sum extends over all helicity states @xmath208 , @xmath209 and @xmath210 are the fermion number and squared mass of the @xmath208th state and @xmath211 is a renormalization mass scale .
we find that the inflationary effective potential is given by @xmath212\right ) , \label{vinf}\end{aligned}\ ] ] where @xmath213 and @xmath214 .
we see that the only non - zero contributions to the effective potential come from the @xmath215-dependent part of the spectrum ( in the first two lines of table 1 ) .
this is a consequence of the fact that , along the inflationary trajectory , susy breaking , due to the presence of non - zero vacuum energy density , occurs only in the inflaton sector . in particular
, there is mass splitting only in the supermultiplet which contains the complex scalar field @xmath216 ( see appendix ) .
0.25 cm note that radiative corrections lift the ( classical ) flatness of the inflationary trajectory providing the necessary inclination for driving the inflaton field @xmath38 towards zero .
it is important to observe that although the effective potential in eq.([vinf ] ) does depend on the unknown scale @xmath211 , its inclination ( derivative with respect to @xmath128 ) is @xmath211-independent .
this is due to the fact that the supertrace of @xmath217 ( @xmath218 being the mass squared matrix ) appearing in eq.([deltav ] ) is , as one can readily deduce using the spectrum in table 1 , @xmath128-independent .
this is an important property since otherwise @xmath203 and @xmath204 would depend on the unknown mass parameter @xmath211 .
0.25 cm inflation is terminated only very close to the critical point @xmath219 ( @xmath220 or @xmath170 ) after which the inflationary path is destabilized and the system falls into @xmath174 .
this can be checked , for all relevant values of the coupling constants , by employing the slow - roll parameters @xmath104 and @xmath221 @xcite .
it turns out that @xmath222 for @xmath223 , while @xmath224 exceeds unity only for @xmath225 s extremely close to 1 .
0.25 cm the quadrupole anisotropy of the cosmic microwave background radiation can be calculated to be @xcite : @xmath226 where @xmath227 gev is the planck scale , @xmath228 and @xmath229 with @xmath230 , @xmath231 being the value of @xmath128 when our present horizon scale crossed outside the inflationary horizon .
the coupling constant @xmath53 can be evaluated from @xcite @xmath232 now , using the cobe constraint , @xmath233 @xcite , taking @xmath234 and eliminating @xmath235 between eqs.([eq : dtt ] ) and ( [ eq : kappa ] ) , we can obtain @xmath236 and , consequently , @xmath201 and @xmath52 as functions of @xmath53 ( we put @xmath202 and @xmath237 ) .
the inflationary scale @xmath238 and the spectral index of density perturbations @xmath239 are then also found as functions of @xmath53 and are depicted in figs.5 and 6 respectively .
0.25 cm the susy minimum can be obtained from @xmath174 in eq.([y3 ] ) by putting @xmath240 .
the common vev @xmath241 of @xmath39 and @xmath40 at this minimum is then given by @xmath242 and is shown in fig.7 as a function of @xmath53 .
a complete inflationary scenario should be followed by a successful reheating which can generate the observed baryon asymmetry of the universe .
we will now turn to the discussion of this reheating process in our model .
the inflaton consists of the two complex scalar fields @xmath38 and @xmath105 , which have equal masses given by @xmath243 at the end of inflation , the two fields @xmath38 and @xmath105 oscillate about the susy minimum and decay into a pair of right - handed sneutrinos ( @xmath244 ) and neutrinos ( @xmath245 ) respectively .
the masses of these ( s)neutrinos are generated , after the breaking of @xmath1 , by the superpotential coupling @xmath246 in eq.([eq : superpot ] ) and turn out to be @xmath247 this same coupling together with the terms in eq.([eq : susyinfl ] ) constitute the part of the superpotential which is relevant for the decay of the inflaton .
we obtain the lagrangian terms @xmath248 for the decay of @xmath38 and @xmath249 for the decay of @xmath105 respectively . from eqs.([eq : ls ] ) and ( [ eq : ltheta ] ) , we deduce that @xmath38 and @xmath105 have equal decay widths given by @xmath250 provided that @xmath251 . to minimize the number of small couplings we assume that @xmath252 so that the coupling @xmath253 can be of order one .
the inflaton then decays into the second heaviest right - handed neutrino superfield with mass @xmath254 .
( there always exist @xmath253 s smaller than one which satisfy the second inequality in this equation for all relevant values of the other parameters . )
thus the reheat temperature @xmath255 after inflation , for the mssm spectrum , is given by @xcite @xmath256 the gravitino constraint @xcite gives an upper bound on @xmath255 of about @xmath257 gev for gravity - mediated susy breaking with universal boundary conditions .
to maximize the naturalness of the model , we take the maximal value of @xmath254 ( and thus @xmath258 ) allowed by the gravitino constraint . note that these values of @xmath254 turn out to be about two orders of magnitude lower than the corresponding values of @xmath259 and , thus , the first inequality in eq.([eq : ineq ] ) is well satisfied .
0.25 cm analysis @xcite of the chooz experiment @xcite shows that the solar and atmospheric neutrino oscillations decouple , allowing us to concentrate on the two heaviest families .
the light neutrino mass matrix is then given by @xmath260 where @xmath261 is the dirac neutrino mass matrix with positive eigenvalues @xmath262
( @xmath263 ) , and @xmath264 the majorana mass matrix of the right - handed neutrinos with positive eigenvalues @xmath265 ( @xmath266 ) given in eq.([m0 ] ) .
the two positive eigenvalues of @xmath267 will be denoted by @xmath268 ( or @xmath7 ) and @xmath269 ( or @xmath8 ) , with @xmath270 .
the determinant and trace invariance of @xmath271 provide us with two constraints @xcite on the mass parameters @xmath272 , @xmath262 , @xmath265 and the angle @xmath105 and phase @xmath273 of the rotation matrix which diagonalizes the right - handed neutrino mass matrix @xmath264 in the basis where @xmath261 is diagonal .
0.25 cm the bounds on @xmath7 from the small or large angle msw solution of the solar neutrino puzzle are respectively @xmath274 ev or @xmath275 ev @xcite . as we will see below , the latter solution is favored in our model .
the @xmath24-neutrino mass is restricted in the range
@xmath276 ev @xmath277 ev from the results of superkamiokande @xcite which also imply almost maximal @xmath6 mixing , i.e. , @xmath278 .
assuming that the dirac mixing angle @xmath279 ( i.e. , the mixing angle in the absence of right - handed neutrino majorana masses ) is negligible , we find @xcite @xmath280 , where @xmath281 is the rotation angle which diagonalizes @xmath267 .
0.25 cm an important constraint comes from the baryon asymmetry of the universe . in this model , a primordial lepton asymmetry
is generated @xcite by the decay of the superfield @xmath282 which emerges as the decay product of the inflaton .
( this lepton asymmetry is subsequently partially converted into baryon asymmetry by electroweak sphalerons . )
the superfield @xmath282 decays into electroweak higgs and ( anti)lepton superfields .
the resulting lepton asymmetry is @xcite @xmath283 where @xmath284 and @xmath285 .
this is related to the baryon asymmetry @xmath286 by @xmath287 for the spectrum of the mssm @xcite .
thus , the low deuterium abundance constraint @xcite on the baryon asymmetry of the universe @xmath288 gives @xmath289 .
0.25 cm due to the presence of @xmath9 in @xmath1 , the dirac mass parameter @xmath290 coincides with the asymptotic value of the top quark mass .
taking renormalization effects into account , in the context of the mssm with large @xmath291 , we find @xcite @xmath292 .
0.25 cm for each value of @xmath53 , the majorana masses @xmath265 are fixed .
taking @xmath272 and @xmath290 also fixed in their allowed ranges , we are left with only three undetermined parameters @xmath273 , @xmath105 and @xmath293 which are further restricted by four constraints : almost maximal @xmath6 mixing ( @xmath294 ) , the leptogenesis restriction ( @xmath295 ) and the constraints from the trace and determinant invariance of @xmath271 .
it is highly non - trivial that solutions satisfying all the above requirements can be found with natural values of @xmath53 ( of order @xmath2 ) and @xmath293 of order @xmath179 gev .
typical solutions can be constructed , for instance , for @xmath296 , which corresponds to @xmath297 , @xmath298 gev , @xmath299 gev , @xmath300 gev and @xmath301 gev ( @xmath302 ) .
( remember @xmath202 , @xmath303 gev and @xmath304 gev . ) taking , for example , @xmath305 ev , @xmath306 ev and @xmath307 gev , we find @xmath308 gev , @xmath309 , @xmath310 and @xmath311 for @xmath312 .
0.25 cm it is interesting to note that the mass scale @xmath313 is of order @xmath3 gev which is consistent with the unification of the gauge couplings of the mssm .
also , the values of the @xmath5-neutrino mass , for which solutions are found , turn out to be consistent with the large rather than the small angle msw mechanism .
0.25 cm in summary , we found that , within the framework of our inflationary scheme , a number of cosmological and phenomenological constraints could be easily satisfied .
these constraints include the gravitino bound on the reheat temperature , successful baryogenesis in the universe , and consistency of neutrino masses and mixing with the solar and atmospheric neutrino oscillations .
it should be pointed out that , for this success of our model , no extra structure was needed to be added to it .
actually , the possibility of accounting for these requirements is more or less automatic in this inflationary scheme and can , in principle , remain in gauge groups other than @xmath1 too .
we have constructed a susy gut model based on the @xmath1 gauge symmetry group .
this model is consistent with all particle physics and cosmological requirements .
the @xmath5-problem is solved by introducing a global anomalous pq symmetry @xmath68 , which also solves the strong cp problem .
although baryon and lepton numbers are violated in the superpotential , the proton turns out to be practically stable .
susy hybrid inflation is ` naturally ' and successfully incorporated in this model but in an unconventional way . in the standard realizations of susy hybrid inflation ,
the superpotential involves only renormalizable couplings of the gut higgs superfields and a gauge singlet .
we have modified this picture by including the next order non - renormalizable coupling too .
in contrast to the usual case , inflation now takes place along a classically flat direction where the gauge symmetry ( @xmath1 ) is spontaneously broken to @xmath18 . as a consequence , after inflation ends , there is absolutely no production of doubly charged magnetic monopoles , which are associated with the breaking of @xmath1 .
thus , the cosmological catastrophe one would encounter by employing the usual inflationary scheme in the susy ps theory is avoided .
our mechanism is crucial for the viability of any model containing cosmologically disastrous topological defects such as magnetic monopoles or domain walls and leads to complete absence of such objects .
it is interesting to point out that , although the usual trajectory with unbroken @xmath1 also exists , there is a range of parameters for which the system finally inflates along the non - trivial path before falling into the susy vacua .
thus , the monopole problem can be solved for all possible initial conditions .
0.25 cm the classical flatness of the inflationary valley is lifted by one - loop radiative corrections which produce an inclination for driving the inflaton towards the susy vacua .
the measurements of cobe can be easily reproduced with natural values ( of order @xmath2 ) of the relevant coupling constant .
the gut mass scale comes out equal to ( or somewhat smaller than ) the susy gut scale and can certainly be much closer to it than in the standard susy inflationary scheme .
the spectral index of density perturbations ranges between about 1 and 0.9 .
0.25 cm after inflation ends , the inflaton oscillates about the susy vacuum and decays into the second heaviest right - handed neutrino superfield thereby reheating the universe
. the subsequent decay of these right - handed neutrinos to lepton and electroweak higgs superfields generates a lepton asymmetry which is then partially converted to baryon asymmetry by the electroweak instantons .
we require that the so obtained baryon asymmetry of the universe is consistent with the low deuterium abundance constraint .
we also take almost maximal @xmath6 mixing as indicated by superkamiokande . the @xmath5- and @xmath24-neutrino masses
are restricted by the msw resolution of the solar neutrino puzzle and the heaviest dirac neutrino mass by @xmath9 symmetry .
we find that all these requirements can be met with natural values ( of order @xmath2 ) of the relevant coupling constant .
note that the second heaviest dirac neutrino mass turns out to be of order @xmath179 gev and masses of @xmath314 consistent with the large rather than the small angle msw mechanism are favored . 0.25 cm finally , we would like to point out that , although we restricted our discussion to the ps gauge group , this new susy hybrid inflationary scenario is of much wider applicability .
the reason for choosing this particular framework for our presentation is that @xmath1 is one of the simplest gauge groups with magnetic monopoles .
the scenario can be readily extended to other semi - simple gauge groups such as the trinification group @xmath315 which emerges from string theory .
of course , the superfields @xmath39 and @xmath40 should be replaced by the appropriate pair of superfields @xmath316 and @xmath317 from the @xmath318 @xmath319-plet and @xmath320-plet which cause the breaking of the trinification group to @xmath17 . further breaking of @xmath17 to @xmath18
is achieved by a similar pair of superfields with vevs , though , in the right - handed neutrino direction .
our scheme could also be extended to simple gauge groups such as @xmath321 .
the breaking of @xmath321 can be achieved by including , among other representations , a pair of @xmath322 , @xmath323 higgs superfields acquiring vevs in the right - handed neutrino direction .
inflation and reheating are expected to be quite similar to the ones discussed here with the only complication that more higgs superfield representations such as @xmath324 and @xmath325 will be involved for gauge symmetry breaking to @xmath18 .
the magnetic monopole problem can then be solved only if some of these fields are non - zero too on the inflationary trajectory .
this work was supported by european union under the tmr contract no .
erbfmrx - ct96 - 0090
. s. k. is supported by a ministerio de educacion y cultura research grant .
also , q. s. would like to acknowledge the doe support under grant number de - fg02 - 91er40626 . finally , r.
j. and s. k. would like to thank g. senjanovi for discussions .
in this appendix , we sketch the derivation of the mass spectrum of the model when the system is trapped along the inflationary trajectory at @xmath140 . during inflation ,
the fields @xmath39 , @xmath40 acquire vevs in the @xmath94 , @xmath126 direction which break the gauge symmetry @xmath1 down to @xmath18 .
these vevs are given by @xmath326 and we can write @xmath327 and @xmath328 . one can show that the scalar potential in eq.([eq : infpot ] ) does not generate masses for the scalar components of the superfield @xmath39 @xmath329 in the directions @xmath93 , @xmath45 , @xmath95 ( @xmath330 , @xmath46 , @xmath331 ) . on the contrary , a simple calculation yields that the normalized real scalar fields @xmath332 and @xmath333 acquire non - zero masses given by @xmath334 respectively .
the superpotential in eq.([eq : susyinfl ] ) gives rise to just one massive majorana fermion with @xmath335 corresponding to the direction @xmath336 .
we see that the susy breaking along the inflationary trajectory , which is due to the non - zero vacuum energy density @xmath337 , produces a mass splitting in the @xmath94 , @xmath126 supermultiplets .
actually , as we will show , this is the only place where such a mass splitting appears .
0.25 cm the d - term contribution to the scalar masses can be found from @xmath338 where @xmath206 is the @xmath1 gauge coupling constant and the sum extends over all the generators @xmath339 of @xmath1 .
the part of this sum over the generators @xmath340 of @xmath9 and @xmath341 of @xmath342 gives rise to a mass term for the normalized real scalar field @xmath343 with @xmath344 as one can show by using the above expansion of @xmath94 , @xmath126 . 0.25 cm the gauge bosons @xmath345 can acquire masses from the lagrangian terms @xmath346 taking the contribution of @xmath347 and @xmath348 again , we obtain a mass term for the normalized gauge field @xmath349 with @xmath344 ( the real field @xmath350 , which is so far left massless , is absorbed by this gauge boson ) .
0.25 cm fermion masses get also contributions from the lagrangian terms @xmath351 where @xmath352 is the gaugino corresponding to @xmath339 and @xmath353 , @xmath354 represent the chiral fermions belonging to the superfields @xmath40 , @xmath39 respectively .
concentrating again on @xmath347 , @xmath348 , we obtain a dirac mass term between the chiral fermion in the superfield @xmath355 and the gaugino @xmath356 with @xmath344 .
this completes the analysis of the @xmath94 , @xmath126 sector together with the gauge supermultiplet in the @xmath357 direction .
0.25 cm the eight normalized real scalar fields @xmath358 , @xmath359 ( three colors ) , @xmath360 , @xmath361 acquire mass terms from the d - term contribution in eq.([a2 ] ) with @xmath362 . indeed , the part of the sum in this equation over the generators @xmath363 of @xmath342 gives @xmath364 similarly , the sum over the @xmath9 generators @xmath365 and @xmath366 @xmath367 with @xmath368 @xmath369 and @xmath370 @xmath371 in the @xmath372 @xmath373 entries respectively and zero everywhere else generates the masses of @xmath374 , @xmath375 ( three colors ) . using eq.([a3 ] ) , one can show that the eight gauge bosons @xmath376 , @xmath377 , @xmath378 , @xmath379 @xmath367 become massive with @xmath362 ( they absorb the real fields @xmath380 , @xmath381 , and @xmath382 , @xmath383 ( three colors ) ) . 0.25 cm the chiral fermions @xmath384 and @xmath385 combine with the gauginos @xmath386 and @xmath387 respectively to form two dirac fermion states with @xmath362 as one deduces from eq.([a4 ] ) . similarly , @xmath388 and @xmath389 ( three colors ) together with @xmath390 and @xmath391 give six more dirac fermions with the same mass squared
0.25 cm the only superfields left are the @xmath45 , @xmath46 , @xmath392 , @xmath393
. they do not mix with the rest of the spectrum , as one can easily show , and acquire masses from the last two superpotential terms in eq.([eq : superpot ] ) .
these terms can be explicitly written as @xmath394 the scalar potential then contains the terms @xmath395 and we obtain six complex scalars @xmath396 with @xmath397 and six complex scalars @xmath398 with @xmath399 . also , the chiral fermions @xmath400 and @xmath401 combine to give three dirac fermions with @xmath397 , while @xmath402 and @xmath403 give three dirac fermions with @xmath399 .
0.25 cm we see that all the fields acquire non - zero masses but susy is broken only in the sector of the two real scalar fields with masses given in eq.([a1 ] ) and the majorana fermion with @xmath404 . in all other supermultiplets ,
the fermionic and bosonic components have equal masses . as a consequence ,
these supermultiplets give zero contribution to the supertraces @xmath405 for any integer @xmath406 ( and in general to the supertrace of any function of the mass squared matrix @xmath218 ) .
thus , in calculating any such supertrace , we only have to consider the two real scalars and the majorana fermion mentioned above .
note that , in particular , their contribution to @xmath407 is zero .
99 t. kajita , talk given at xviiith international conference on neutrino physics and astrophysics ( neutrino98 ) , takayama , japan , 4 - 9 june , 1998 . 0.2 cm m. gell - mann , p. ramond and r. slansky , in _
supergravity _ ,
p. van nieuwenhuizen and d. freeman ( north holland , amsterdam , 1979 ) p. 315 ; + t. yanagida , prog .
64 ( 1980 ) 1103 ; + r. mohapatra and g. senjanovi , phys .
44 ( 1980 ) 912 . 0.2 cm k. babu , b. dutta and r. mohapatra , phys . rev
d 60 ( 1999 ) 095004 ; + phys . lett .
b 458 ( 1999 ) 93 ; phys . rev .
d 61 ( 2000 ) 091701
. 0.2 cm s. khalil and q. shafi , phys .
d 61 ( 2000 ) 035003 ; + s. king and q. shafi , phys . lett .
b 442 ( 1998 ) 135 . 0.2 cm j. c. pati and a. salam , phys .
d 10 ( 1974 ) 275 . 0.2 cm g. shiu and s .- h .
h. tye , phys .
d 58 ( 1998 ) 106007 . 0.2 cm i. antoniadis and g. k. leontaris , phys .
b 216 ( 1989 ) 333 .
0.2 cm i. antoniadis , g. k. leontaris and j. rizos , phys .
b 245 ( 1990 ) 161 . 0.2 cm a. d. linde , phys . rev .
d 49 ( 1994 ) 748 . 0.2 cm g. lazarides , r. schaefer and q. shafi , phys .
d 56 ( 1997 ) 1324 . 0.2 cm g. lazarides and n. d. vlachos , phys . lett .
b 459 ( 1999 ) 482 . 0.25 cm g. lazarides , prhep - trieste99/008 [ hep - ph/9905450 ]
. 0.2 cm g. lazarides , m. magg and q. shafi , phys .
b 97 ( 1980 ) 87 . 0.2 cm
g. lazarides and c. panagiotakopoulos , phys .
d 52 ( 1995 ) 559 .
0.2 cm t. w. b. kibble , j. phys . a 9 ( 1976 ) 387 .
0.2 cm g. dvali , r. schaefer and q. shafi , phys .
73 ( 1994 ) 1886 .
0.2 cm l. covi , g. mangano , a. masiero and g. miele , phys .
b 424 ( 1998 ) 253 . 0.2 cm g. f. smoot _
et al . _ ,
396 ( 1992 ) l1 ; c. l. bennett
_ et al . _ ,
j. lett . 464
( 1996 ) 1 . 0.2 cm g. lazarides and q. shafi , phys . rev .
d 58 ( 1998 ) 071702 . 0.2 cm m. fukugita and t. yanagita , phys .
b 174 ( 1986 ) 45 ; + w. buchmller and m. plmacher , phys . lett .
b 389 ( 1996 ) 73 ; + g. lazarides and q. shafi , phys
b 258 ( 1991 ) 305 . + for a recent review in the context of susy hybrid inflation see g. lazarides , sprin .
trac . in mod .
163 ( 2000 ) 227 [ hep - ph/9904428 ] .
0.2 cm m. yu . khlopov and a. d. linde , phys .
b 138 ( 1984 ) 265 ; + j. ellis , j. e. kim and d. nanopoulos , phys . lett .
b 145 ( 1984 ) 181 . 0.2 cm s. burles and d. tytler , astrophys . journal 499 ( 1998 ) 699 ; ibid . 507
( 1998 ) 732 .
0.2 cm e. j. copeland , a. r. liddle , d. h. lyth , e. d. stewart and d. wands , phys . rev .
d 49 ( 1994 ) 6410 . 0.2 cm g. dvali , g. lazarides and q. shafi , phys .
b 424 ( 1998 ) 259 . 0.2 cm g. lazarides and n. d. vlachos , phys .
b 441 ( 1998 ) 46 . 0.2
cm l. dolan and r. jackiw , phys .
d 9 ( 1974 ) 3320 ; + s. weinberg , phys .
d 9 ( 1974 ) 3357 .
0.2 cm r. jeannerot and j. lesgourgues , phys .
d 62 ( 2000 ) 023514 . 0.25 cm a. r. liddle and d. h. lyth , phys .
231 ( 1993 ) 1 . 0.2 cm s. coleman and e. weinberg , phys . rev .
d 7 ( 1973 ) 1888 .
0.2 cm c. giunti , hep - ph/9802201 .
0.2 cm m. apollonio _
b 420 ( 1998 ) 397 . 0.2 cm g. lazarides , q. shafi and n. d. vlachos , phys .
b 427 ( 1998 ) 53 . 0.2 cm j. n. bahcall , p. i. krastev and a. yu .
smirnov , phys .
d 58 ( 1998 ) 096016 ; ibid .
d 60 ( 1999 ) 093001 .
0.2 cm l. e. ibez and f. quevedo , phys .
b 283 ( 1992 ) 261 . 0.2 cm | we show how hybrid inflation can be successfully realized in a supersymmetric model with gauge group @xmath0 . by including a non - renormalizable superpotential term , we generate an inflationary valley along which @xmath1 is broken to the standard model gauge group .
thus , catastrophic production of the doubly charged magnetic monopoles , which are predicted by the model , can not occur at the end of inflation .
the results of the cosmic background explorer can be reproduced with natural values ( of order @xmath2 ) of the relevant coupling constant , and symmetry breaking scale of @xmath1 of the order of @xmath3 gev .
the spectral index of density perturbations lies between unity and @xmath4 .
moreover , the @xmath5-term is generated via a peccei - quinn symmetry and proton is practically stable .
baryogenesis in the universe takes place via leptogenesis . the low deuterium abundance constraint on the baryon asymmetry ,
the gravitino limit on the reheat temperature and the requirement of almost maximal @xmath6 mixing from superkamiokande can be simultaneously met with @xmath7 , @xmath8 and heaviest dirac neutrino mass determined from the large angle msw resolution of the solar neutrino problem , the superkamiokande results and @xmath9 symmetry respectively . ba-99 - 82 , ftuam-99 - 42 , +
ic/99/187 , ut - stpd-1/00 . 0.3 cm * inflation and monopoles in supersymmetric + 0.15 cm @xmath10 * 0.35 cm r. jeannerot@xmath11 , s. khalil@xmath12 , g. lazarides@xmath13 and q. shafi@xmath14 + + + + + + |
photonic crystals are artificial low - loss dielectric structures with periodic modulation of refractive index , which have attracted considerable attention in the last two decades . due to bragg reflection
, electromagnetic ( optical ) waves can not propagate through such structures in certain directions , at certain frequencies .
hence , photonic crystals can control the propagation of electromagnetic waves in novel ways , with obvious application to dielectric mirrors , dielectric waveguides , and dielectric laser cavities . as a way to efficiently inject light into a photonic crystal ( pc ) waveguide
, it has recently been proposed to use surface electromagnetic waves ( sew)@xcite . in those papers , the photonic crystal was a two dimensional array of rods , of infinite length normal to the plane of incidence .
instead , we have studied sew on a semi - infinite one - dimensional ( 1d ) photonic crystal sketched in fig .
while retaining all useful properties of 2d and 3d photonic crystals , a 1d dielectric periodic structure with high refractive index contrast is more attractive from a technological point of view
. the usual theoretical methods for wave propagation in 1d photonic crystals , including sew , are the floquet - bloch modal formalism , coupled wave theory , and the transfer matrix method . among these three ,
the coupled wave approach@xcite offers superior physical insight and gives simple analytical results in limiting cases .
unfortunately , the conventional coupled wave theory of kogelnik fails in the case of high refractive index contrast , which is essential for a functional 1d photonic crystal . in this paper
, we apply our recently developed semiclassical version of coupled wave theory@xcite to sew on 1d photonic crystals .
the method is analytically almost as simple as conventional coupled wave theory , and is practically exact for the achievable ratios ( e.g. 1.5:4.6 ) of the indices of refraction of the materials available to build devices .
we present here a unified description of te and tm sew . a detailed account of the properties of the te surface modes has recently been given by us in ref .
@xcite ; here we complement these findings with those for tm modes , which are slightly more complex due to the presence of brewster points in the bandgaps . as a result , we thoroughly clarify the systematics of solutions for surface em waves in semi - infinite 1d photonic crystals .
our method is formally quite different from that recently presented in ref .
@xcite , or those in ref .
@xcite , so in section ii we provide a short summary of the transfer matrix approach , in the notation of our previous work@xcite . in section
iii we rederive the exact equations for sew of tm modes and obtain from them various analytic approximations for a semi - infinite crystal .
the analogous equations for te modes were given in ref .
@xcite . with these in hand , we discuss systematics of sew . in section
iv we apply the semiclassical approximations of refs .
@xcite and @xcite to surface waves , and show that the second approximation is very accurate both for the dispersion relation and the bandgap boundaries .
we wish to describe surface states that form at the interface between a medium of low refractive index , @xmath0 , and a semi - infinite 1-d photonic crystal with layers of refractive indices @xmath1 and @xmath2 and thicknesses @xmath3 and @xmath4 .
we choose a coordinate system in which the layers have normal vector along oz . as shown in fig .
[ dsfig01 ] , the crystal is capped by a layer of the same material but different width , @xmath5 . for convenience of presentation , we split this termination layer of index of refraction @xmath1 and width @xmath5 into two sublayers , of lengths @xmath6 .
the first sublayer extends from @xmath7 to @xmath8 .
then the periodic array that forms the 1d photonic crystal consists of `` cells '' each made of three uniform layers of widths @xmath9 , @xmath4 and @xmath10 whose respective indices of refraction are @xmath1 , @xmath2 and @xmath1 .
( if @xmath11 , the unit cell will have reflection symmetry , which simplifies some relations , but does not change any physical results . ) the first cell , given index @xmath12 , ranges from @xmath8 to @xmath13 ; the second is given index @xmath14 , and ranges from @xmath15 to @xmath16 , etc .
the p - th cell runs from @xmath17 to @xmath18 and has @xmath19 when @xmath20 or @xmath21 and @xmath22 when @xmath23 .
we choose @xmath24 . for monochromatic te waves the electric field is parallel to the oy axis .
as in ref .
@xcite , we write @xmath25 where @xmath26 is the angular frequency , @xmath27 . is the vacuum wavenumber and @xmath28 is the ( constant ) @xmath29-component of the wavevector of modulus @xmath30 . for an electromagnetic wave entering the 1d photonic crystal from a uniform medium
, one has @xmath31 where @xmath32 is the angle of incidence measured from the normal . for monochromatic tm waves it is the magnetic field which lies parallel to the oy axis .
following ref .
@xcite , we write @xmath33 p. kramper , m. agio , c. m. soukoulis , a. bimer , f. mller , r. wehrspohn , u. gsele , and v. sandoghdar , _
`` highly directional emission from photonic crystal waveguides of subwavelength width '' _ , phys .
lett . * 92 * , 113903 ( 2004 ) . | we study surface states of 1d photonic crystals using a semiclassical coupled wave theory . both te and tm modes are treated .
we derive analytic approximations that clarify the systematics of the dispersion relations , and the roles of the various parameters defining the crystal . |
approaching grand unification and supersymmetry in higher dimensions has been an intriguing possibility .
one of the problems to be solved in grand unification is the doublet - triplet splitting . in the four - dimensional ( 4d ) minimal su(5 )
grand unified model , an adjoint higgs field is responsible for breaking the unified gauge group to the standard model gauge group , whereas a fundamental higgs field leads to breaking electroweak symmetry .
this fundamental higgs field includes a color - triplet higgs field as well as the weak - doublet higgs field under the standard model gauge group . at tree level ,
the triplet and doublet higgs fields acquire their masses depending on the parameters in the potential of the original adjoint and fundamental higgs fields .
it is natural that these masses are of the same order . because the doublet higgs field whose vacuum expectation value is developed at the electroweak scale should be lighter than the triplet higgs field , a tuning
would be needed unless the hierarchy is generated by any mechanism .
in addition , even if the masses are tuned at tree level , radiative corrections can break it . in theory with extra dimensions ,
higher - dimensional gauge invariance can consist of a unified gauge group while the standard model gauge group only survives on 4d .
the unified gauge group is reduced to the standard model gauge group via boundary conditions in the direction of the extra dimensions . if the fundamental su(5 ) higgs field in the role of the electroweak symmetry breaking propagates in higher dimensions , the mass splitting of the triplet and doublet higgs fields can be obtained as a result of the boundary conditions @xcite .
such higher - dimensional grand unified models have been widely studied @xcite-@xcite .
employing boundary conditions provides various interesting application .
if in constructing 4d grand unified models one requires that the unified gauge group is a simple group or a direct product of simple groups , that it contains the standard model gauge group as a subgroup , that its rank is four and that it has complex representation , the candidates of gauge group could be not only su(5 ) but also su(3)@xmath0@xmath1su(3)@xmath2 .
an unfavorable reason of adopting su(3)@xmath0@xmath1su(3)@xmath2 would be that the inclusion of matter is not minimal .
if quarks are transformed as @xmath3 under su(3)@xmath0@xmath1su(3)@xmath2 , the electric charge matrix can be assigned as @xmath4 . then integer electric charges
are also made from @xmath5 because the adjoint representation includes components with the integer charge @xmath6 and the third symmetric representation also includes components with the integer charge @xmath7 .
if leptons are assigned in these representations , extra fields in addition to standard model fields would be involved .
such additional fields should be decoupled at low energies as they are heavy . in the case where the theory is based on higher - dimensionsional gauge invariance ,
the mass splitting of extra fields and matter fields can be obtained as a result of boundary conditions similar to the mass splitting for the higgs field .
this type of decoupling is also used for avoiding another disputable feature in the 4d minimal su(5 ) unified model : the fermion mass relations . at a unification scale ,
down - type quarks and charged leptons have the identical yukawa coupling ( matrix in flavor space , while up - type yukawa matrix is symmetric ) .
their mass eigenvalues are equal . for one - loop mass correction arising from fermion self - energy with helicity flip ,
the ratio of the down - type quark masses to the charged lepton masses is described in powers of gauge coupling constants . for the third generation
, it provides a successful prediction and for the first two generations , the prediction seems unfavorable . in higher - dimensional gauge theory ,
down - type quarks and charged leptons for the first two generations may be taken to arise from distinct origin of multiplets as extra components are decoupled via boundary conditions
. then the unfavorable fermion mass relation disappears .
supersymmetry breaking transmitted via extra dimensions can be a solution to no experimentally incompatible flavor changing neutral current . in gaugino mediation @xcite@xcite ,
supersymmetry is broken in a sector spatially separated from supersymmetric standard model sector and gauginos acquire masses at high energy . for squarks and sleptons , the positive masses squared are generated at low energy by renormalizaiton group flow .
the regularities required to avoid flavor changing neutral currents are automatically obtained since the gauge interactions do not distinguish generations . for the renormalization group equations ,
a simple possiblity of the intitial condition is that the gaugino masses have unified values as an input at high energy .
if this is taken seriously , it would be natural that gaugino - mediated supersymmetry breaking is incorporated into grand unified models .
if the doublet - triplet splitting by boundary conditions , no fermion mass relations for the first two generations and gaugino - mediated supersymmetry breaking are taken into account , the simplest setup would be to consider two extra dimensions . although the doublet - triplet splitting and no fermion mass relations can be simultaneously treated for one extra dimension , the source of supersymmetry breaking in gaugino mediation
should not be directly coupled to the matter superfields propagating in the extra dimension . for such unified models , to contain the weak - doublet higgs fields in an adjoint representation ( a possiblity of gauge - higgs unification ) and to introduce right - handed neutrino motivate that the original higher dimensional gauge group is larger than the standard model gauge group .
we consider 6d su(6 ) supersymmetric models on an orbifold , where the sizes of two extra dimensions are different . to solve the doublet - triplet splitting , higgs fields
propagate in 5d or 6d .
we choose the smaller compactification radius as the unification scale where the unified gauge group is broken by boundary conditions .
the other compactification radius is limited so that field - theoretical description is valid .
our first model has the weak - doublet higgs fields in 5d chiral superfields , which is a version with gaugino - mediated supersymmetry breaking of ref.@xcite where the doublet - triplet splitting , no proton decay from operators of dimension four or five , no mass relations for the first two generations and gauge coupling unification were achieved .
our other model has the higgs doublets in the 6d gauge multipet , which is a 6d version with gaugino mediated supersymmetry breaking of ref.@xcite where yukawa couplings were given without conflicting with higher - dimensional gauge invariance and the sizes of the yukawa couplings arised from wave - function profiles of the matter zero modes determined by bulk mass parameters . in both models ,
we examine gauge coupling constants beyond the energy scale of the smaller compactification radius which is the unification scale .
we find very nontrivial high energy behavior of gauge coupling constants . within the region where the larger compactification radius has such a size that the gauge coupling constants are not blow - up nor zero , we find realisic patterns of yukawa coupling constants .
proton stability is achieved by r invariance .
although higgs fields propagate in 5d or 6d , gaugino - mediated supersymmetry breaking is quite similar to the minimal supersymmetric standard model ( mssm ) case with only gauge superfields propagating in bulk .
the paper is organized as follows . in section 2
we present an explicit 6d component action starting with the action already known in the 4d superfield formalism .
general properties of orbifold parity , mode expansion , couplings on fixed lines and fixed points , localization and dceoupling of fields on fixed lines are shown also with explicit equations . in section 3 our approach to examine high energy behavior of gauge coupling constants
gauge coupling corrections , yukawa couplings , proton stability and supersymmetry breaking are examined for the model with higgs fields in 5d multiplets in section 4 and for the model with higgs fields in the 6d gauge multiplet in section 5 . in section 6 conclusion
we work with 6d @xmath8 theory which corresponds to @xmath9 in the language of 4d .
the theory is vector - like and it is manifest to be free from 6d anomaly .
we consider a vector multiplet in 6d bulk .
the field contents are the vector superfield @xmath10 and the three chiral superfields @xmath11 , @xmath12 and @xmath13 .
these superfields are written in the wess - zumino gauge for @xmath10 in terms of 4d superfields as @xmath14 where @xmath15 .
the two - component weyl fermions @xmath16 are left - handed . for the three left - handed chiral superfields ,
the 4d coordinates are indicated in the @xmath17-basis @xmath18 .
the coordinates @xmath19 and @xmath20 are labeled as a kind of internal spin in 4d . in the left - hand sides of the component expressions for @xmath11 , @xmath12 and @xmath13 , ( [ sig5y])-([phiy ] ) , the factor @xmath21 is included so that kinetic terms such as @xmath22 will be canonically normalized . for the lowest components of @xmath11 and @xmath12 ,
the complex scalar fields @xmath23 and @xmath24 are written as @xmath25 the 4d real scalars @xmath26 and @xmath27 are identified with the fifth- and sixth - components of the 6d gauge field , respectively .
the superfields @xmath10 , @xmath11 , @xmath12 and @xmath13 have dimensionality @xmath28=\left[\textrm{mass}\right]^0 $ ] and @xmath29=\left[\sigma_6\right]= \left[\phi\right]=\left[\textrm{mass}\right]$ ] which are the same as in the 4d case .
the difference of dimensionality from the 4d case is included as the 6d gauge coupling has dimensionality @xmath30=\left[\textrm{mass}\right]^{-1}$ ] . in the wess - zumino gauge with these superfields ,
the 6d supersymmetric action is written as @xcite@xcite @xmath31)\right)\right]_f \nonumber \\ & \!\!\!+\!\!\!&{1\over 8kg^2 } \textrm{tr}\bigg[\sum_{i=5}^6 \left(2(-\partial_i+\sigma_i^\dag)e^{2v } ( \partial_i+\sigma_i)e^{-2v } + ( \partial_i e^{2v } ) ( \partial_i e^{-2v})\right ) + 2\phi^\dag e^{2v } \phi e^{-2v}\bigg]_d\bigg\rbrace , \nonumber \\
\label{actions6}\end{aligned}\ ] ] where extracting @xmath32-term is represented as the square brackets with the subscript @xmath32 such as @xmath33_f$ ] and extracting @xmath34-term is represented as @xmath33_d$ ] with the notation of a vector superfield @xmath35 .
the action is invariant under the transformation with @xmath36 @xmath37 in the action ( [ actions6 ] ) , the numerical coefficient in front of @xmath38 is determined by the normalization of the gauge field kinetic term @xmath39 . the numerical coefficient of the khler term including @xmath11 is determined by the normalization of @xmath40 .
the khler term with @xmath11 also includes @xmath41 .
the khler term with @xmath13 includes @xmath42 .
the numerical coefficient of the superpotential is determined by the normalization of @xmath43 after the auxiliary field @xmath32 is eliminated .
since the theory contains 4d @xmath8 supersymmetry ( @xmath44 4d @xmath9 ) , two of the chiral superfields may be paired via complex conjugate in which representation matrices of gauge group can be @xmath45 and @xmath46 . in the above notation , this property is implicit and all the representation matrices of gauge group are taken as @xmath45 .
the superfields are written as @xmath47 , @xmath48 , @xmath49 and @xmath50 , where @xmath51 and @xmath52 .
the generators @xmath53 with @xmath54 and the structure constants are shown in appendix [ lam ] . from the action ( [ actions6 ] )
, the component action is written as @xmath55 ) -\sqrt{2}\phi\left[\lambda_5,\lambda_6\right ] + \textrm{h.c.}\bigg ) \nonumber \\ & \!\!\!+\!\!\ ! & { 1\over 2 } \sum_{j=5}^6 \bigg\lbrace -\partial_j a^n \partial_j a_n -\sqrt{2}i(\partial_j a^n)({\cal d}_n \sigma_j ) + \sqrt{2}i({\cal d}^n \sigma_j)^*(\partial_j a_n ) -2({\cal d}^n \sigma_j)^ * ( { \cal d}_n \sigma_j ) \nonumber \\ & & \qquad + \bigg ( -2i(\partial_j\lambda)\lambda_j -\sqrt{2}(\partial_j d)\sigma_j -2\sqrt{2}i\bar{\lambda}_j\bar{\lambda}\sigma_j + 2\sqrt{2}i(\bar{\lambda}_j\sigma_j\bar{\lambda } ) + \textrm{h.c.}\bigg ) \nonumber \\ & & \qquad -2i\bar{\lambda}_j \bar{\sigma}^n { \cal d}_n\lambda_j + 2f_j^ * f_j+2\sigma_j^ * d\sigma_j -2\sigma_j^ * \sigma_j d\bigg\rbrace \nonumber \\ & \!\!\!-\!\!\ ! & ( { \cal d}^n\phi)^*({\cal d}_n\phi ) -i\bar{\chi}\bar{\sigma}^n{\cal d}_n\chi + f^ * f + \phi^*d\phi -\phi^*\phi d \nonumber \\ & & \qquad + \bigg ( -\sqrt{2}i\bar{\chi}\bar{\lambda}\phi + \sqrt{2}i(\bar{\chi}\phi\bar{\lambda } ) + \textrm{h.c . } \bigg)\bigg ] . \label{componentl6}\end{aligned}\ ] ] where @xmath56 = if^{abc}\sigma_j^{*b}\sigma_j^c t^a$ ] . here
the covariant derivatives @xmath57 acts on fields as @xmath58 , \end{aligned}\ ] ] where @xmath59 , @xmath60 and @xmath61 are similar and the derivative @xmath62 are given by @xmath63 , \qquad~ \nabla_6 f_5=\partial_6f_5 + \sqrt{2}\left[\sigma_6,f_5\right ] , \\
\nabla_5 \lambda_6 & \!\!=\!\!&\partial_5\lambda_6 + \sqrt{2}\left[\sigma_5,\lambda_6\right ] , \qquad~ \nabla_6 \lambda_5 = \partial_6\lambda_5 + \sqrt{2}\left[\sigma_6,\lambda_5\right ] .\end{aligned}\ ] ] from the action ( [ componentl6 ] ) , the equations of motion for auxiliary fields are @xmath64 the equations of motion would change if couplings of fixed lines and fixed points are included .
the corresponding equations and some formula in using superfields are given in appendix [ ap : sf ] . in obtaining @xmath65 in eq.([f5eq ] ) and @xmath66 in eq.([f6eq ] ) , partial integrals have been used . in eq.([deq ] ) , @xmath67 leads to derivative terms over extra components such as @xmath68 whereas it prevents terms such as @xmath69 .
this correctly produces the kinetic term of the 6d gauge field .
after equations of motion for auxiliary fields are employed , the action is written as @xmath70 we obtain the bosonic part as @xmath71\right)^2 -\left[\varphi,\phi\right ] \left[\phi^*,\varphi\right ] -\left[\varphi^*,\phi\right ] \left[\phi^*,\varphi^*\right ] .
\label{bosop}\end{aligned}\ ] ] here the field strength of 6d gauge field @xmath72 and the covariant derivative of @xmath73 are given by @xmath74 with the indices @xmath75 and the complex scalar field @xmath76 is made out of @xmath77 and @xmath78 as @xmath79 whose the covariant derivative is given by @xmath80 . for the fermionic part @xmath81 , we obtain @xmath82 here the 6d fermions are defined in terms of two - component weyl fermions as @xmath83 these 6d fermions satisfy the symplectic majorana condition @xmath84 with @xmath85 .
the notation of spinors , gamma matrices and charge conjugation matrix is shown in appendix [ spinor ] .
we choose radii of two extra dimensions as @xmath86 and impose the orbifold parity @xmath87 in each of @xmath19 and @xmath20 directions .
the theory has four fixed lines at @xmath88 and @xmath89 and four fixed points at @xmath90 , @xmath91 , @xmath92 , @xmath93 .
the boundary conditions of fields are specified by the orbifold parity . the boundary condition for the vector multiplet
are given as follows : for the direction of @xmath19 , @xmath94 for the direction of @xmath20 , @xmath95 these have the periodicity @xmath96 where @xmath97 .
the existence of fixed lines and fixed points explicitly breaks gauge invariance in 6d . at these lines and points , at most the restricted gauge invariance
is preserved @xcite .
the gauge transformation parameters obey the same boundary conditions as the corresponding gauge fields .
the consistency condition of these restricted gauge invariance gives a constraint to the selection of boundary conditions .
for example , as seen from properties of the structure constants shown in appendix [ lam ] , breaking of gauge group such as su(6)@xmath98so(6 ) or independently so(6)@xmath98su(3)@xmath1u(1 ) can be achieved by boundary conditions . on the other hand ,
breaking of gauge group such as su(6)@xmath98so(6)@xmath98su(3)@xmath1u(1 ) ( for example , su(6)@xmath98so(6 ) in the direction of @xmath20 and so(6)@xmath98su(3)@xmath1u(1 ) in the direction of @xmath19 ) by neumann and dirichlet boundary conditions is seen to be incompatible with restricted gauge invariance .
it was explicitly shown that an example of boundary conditions to break restricted gauge symmetry gives rise to breakdown of ward - takahashi identity and unitarity violation @xcite . imposing boundary conditions with the orbifold parity ensures the restricted gauge invariance .
we introduce mode functions in which the property of orbifold parity is manifest . for @xmath99 where four signs indicate the orbifold parity at @xmath100 , @xmath101 in order , a generic 6d massless field @xmath102 is expanded in terms of 4d mode as @xmath103 the mode expansions for all possible parities are shown in appendix [ xpan ] .
as mentioned above , @xmath73 has dimensionality @xmath104=\left[\textrm{mass}\right]$ ] .
each 4d mode has the mass squared tabulated in table [ tabmass ] .
.4d modes and masses squared.[tabmass ] [ cols= " < , < , < , < , < , < " , ] in addition , the dimension - six proton decay by colored higgs exchange is absent unlike the model with higgs fields as 5d chiral superfields , while the dimension - six proton decay by @xmath105 and @xmath106 boson exchange would exist .
the two mssm higgs fields which are part of the 6d gauge multiplet propagate in 6d .
one may expect that the higgs superfields would be directly coupled to supersymmetry breaking source as @xmath107_f \nonumber \\ & & + { 1\over 2}\left [ \left(\left({1\over m}s^\dag + { 1\over m^2 } s^\dag s\right ) h_1 h_2 + \textrm{h.c.}\right ) + { 1\over m^2 } s^\dag s \left ( h_1^\dag h_1 + h_2^\dag h_2 \right ) \right]_d , \label{higgsino}\end{aligned}\ ] ] at @xmath108 .
such couplings appeared in gaugino mediation with the standard model gauge fields and higgs fields and their superpartners in bulk @xcite .
if the coupling @xmath109 is allowed , the coupling seems to lead to higgs mass terms . at the fixed point @xmath91
, restricted gauge symmetry has su(4)@xmath1 su(2)@xmath1u(1 ) .
the field @xmath11 is written as @xmath110 if the coupling @xmath111 is allowed , this seems @xmath112 term .
however , the operators @xmath113 and @xmath114 are gauge non - invariant due to @xmath115-factor of the transformation ( [ sig5 ] ) @xmath116 thus in this model , higgs masses and @xmath112 and @xmath117 terms are not generated by contact interactions . at @xmath118
the model has nonzero gaugino mass and negligible soft scalar mass . for electroweak symmetry breaking , the quartic terms of higgs doublet appear from @xmath34-term . in the operator ( [ bosop ] ) , they are included in the term @xmath119 . since there is no term such as eq.([higgsino ] )
, @xmath112 term must be generated via other mechanism . in case with chern - simons terms ,
it was proposed that combining @xmath32-components of compensator and radion as well as constant superpotentials can generate @xmath112 term @xcite . in this mechanism ,
nontrivial radion potential is crucial .
it has also been shown that a potential of radion with radius stabilization is generated in a model with compensator , radion and constant superpotentials @xcite@xcite and that this radius stabilization is sensitive to field content @xcite .
whether a radion potential relevant to generation of @xmath112 term arises in the present models is an open problem .
we have studied gauge coupling corrections and yukawa coupling constants in 6d unified models based on the gauge group su(6 ) . for a possibility of gauge - higgs unification and inclusion of right - handed neutrino , gauge group larger than the standard model gauge group has been motivated .
we have employed two extra dimensions to approach the doublet - triplet splitting , no fermion mass relations for the first two generations and gaugino - mediated supersymmetry breaking ( no supersymmetry flavor problem ) .
the source of supersymmetry breaking is spatially separated from fixed lines and fixed points where matter fields reside . by orbifold ,
theory on fixed lines and fixed points has restricted gauge symmetry .
heavy fields decouple by boundary conditions .
the models are 6d and 4d anomaly free and proton stability is achieved by r invariance as a subgroup of su(2)@xmath120 . for energy dependence of gauge coupling constants , we have adopted 4d kk effective theory by giving a 6d component action and the fixed - line and fixed - point actions .
the number of kk mode is regarded as energy - dependent . at each scale @xmath121 ,
@xmath122 , @xmath123 , @xmath124 , 4d kk states with the corresponding masses become dynamical . for the effective 4d theory , restricted gauge invariance
has the gauge group su(3)@xmath0@xmath1 su(2)@xmath2@xmath1u(1)@xmath125@xmath1u(1)@xmath126 .
the gauge group u(1)@xmath127 is spontaneously broken by fixed - point couplings .
up to the unification scale , zero mode contributes to renormalization group evolution . since zero mode is right - handed neutrino and its superpartner as well as the field content of mssm , successful su(5 ) relations about the gauge coupling unification , weak mixing angle and charge quantization
have been achieved . beyond the unification scale ,
gauge couplings are sensitive to the field content . in the model with 5d higgs ,
the sum of the coefficients of @xmath128 function relevant to kk mode is @xmath129 independently of gauge group .
this leads to strong coupling at high energies . in the model with the higgs
superfields in the 6d gauge multiplet , @xmath130 .
asymptotic freedom is more strong compared to the 4d unified model .
these rapid energy dependence is likely to constrain the scale @xmath131 . in the two models
, it may be chosen as @xmath132 . for @xmath133 , @xmath132 and @xmath134 ,
realistic patterns of yukawa couplings have been obtained from fixed - point coupling in the model with 5d higgs fields and from the gauge interaction in the model with higgs fields in the 6d gauge multiplet . for the values of paramters ,
small numbers required for the hierarchy of yukawa coupling constants have been generated via volume suppression in the model with 5d higgs fields and exponential suppression dependent on 5d masses in the model with higgs fields in the gauge multiplet .
we have fixed the smaller compactification radius @xmath133 . from a viewpoint of lsp
, it may be important to examine the case @xmath135 in a 6d setup where no fermion mass relations for the first two generations and gaugino - mediated supersymmetry breaking still hold .
gaugino masses as an input at high energies influence the whole values of supersymmetry breaking masses . for a large @xmath136 , spectrum avoiding the lightest stau as lsp
was obtained for a moderate @xmath137 and for @xmath135 @xcite . in models with the two extra dimensions radiative correction may also affect spectrum @xcite .
there is still room to be examined about this point .
i thank masud chaichian , nobuhito maru and norisuke sakai for useful discussions .
this work is supported by bilateral exchange program between japan society for the promotion of science and the academy of finland .
the generators of su(6 ) transformations , @xmath138 ( @xmath139 ) are @xmath140 where dots indicate 0 .
the @xmath138 obey the following commutation relationship : @xmath141 \equiv \lambda_a \lambda_b -\lambda_b \lambda_a = 2i f_{abc}\lambda_c .
\label{commu}\end{aligned}\ ] ] the @xmath142 are odd under the permutation of any pair of indices . the nonzero values are tabulated in table [ fabc ] .
@xmath143 @xmath144 @xmath145 @xmath146 @xmath147 among the generators of su(6 ) transformations , the 15 generators @xmath148 form a subalgebra .
the 15 generators listed above are written also as generators with antisymmetric two indices , @xmath149 @xmath150 .
the generators are given in a matrix form as @xmath151 the @xmath149 obey the following commutation relationship of so(6 ) transformations : @xmath152 = -i(\delta_{bc}{\cal t}_{ad}-\delta_{ac}{\cal t}_{bd } -\delta_{bd}{\cal t}_{ac}+\delta_{ad}{\cal t}_{bc}).\end{aligned}\ ] ] we define @xmath153 ( @xmath154 ) @xmath155 then eq.([commu ] ) becomes @xmath156 = 2i f_{abc}\lambda_c + 2i f_{ab\widehat{c}}\lambda_{\widehat{c } } , \\ & & \left[\lambda_a , \lambda_{\widehat{b}}\right ] = 2i f_{a\widehat{b}c}\lambda_c + 2i f_{a\widehat{b}\widehat{c}}\lambda_{\widehat{c } } , \\ & & \left[\lambda_{\widehat{a } } , \lambda_{\widehat{b}}\right ] = 2i f_{\widehat{a}\widehat{b}c}\lambda_c + 2i f_{\widehat{a}\widehat{b}\widehat{c}}\lambda_{\widehat{c } } , \end{aligned}\ ] ] where @xmath157 run except for @xmath158 . from table
[ fabc ] , it is immediately seen that @xmath159 the nonzero @xmath160 are tabulated in table [ fwabc ] .
@xmath161 @xmath162 @xmath163 @xmath164 @xmath165 @xmath166 the nonzero @xmath167 are tabulated in table [ fabcwww ] .
@xmath168 @xmath169 @xmath170 @xmath171 @xmath172 from tables [ fwabc ] and [ fabcwww ] , it is seen that @xmath173 form the algebra of su(3 ) transformations and that @xmath174 which commutes @xmath175 form the algebra of u(1 ) transformations . in 5d and 6d
, the spinor representation is pseudo real @xcite @xmath176 where @xmath177 denotes 5d and 6d charge conjugation matrices collectively and @xmath178 denotes 0-component of 5d and 6d gamma matrices collectively .
the gamma matrices and charge conjugation matrix to satisfy the pseudo - real condition ( [ pser ] ) are listed in the following .
the 4d gamma matrices are @xmath179 where @xmath180 and @xmath181 .
the 4d chirality matrix is @xmath182 the 4d charge conjugation matrix is @xmath183 where @xmath184 .
4d dirac fermions and the conjugate are @xmath185 the 5d gamma matrices are @xmath186 and @xmath187 .
the 5d charge conjugation matrix is @xmath188 the matrix has the property , @xmath189 @xmath190 the 5d minimal spinor is written as two symplectic majorana spinors or a dirac spinor .
5d symplectic majorana spinors are @xmath191 where @xmath192 and @xmath193 are 4d two - component weyl spinors .
these satisfy the symplectic majorana condition @xmath194 a 5d dirac spinor is written in terms of 4d weyl spinor as @xmath195 the 6d gamma matrices are @xmath196 @xmath197 where @xmath198 and @xmath199 .
the 6d chirality matrix is @xmath200 the 6d charge conjugation matrix is @xmath201 @xmath202 @xmath203 the 6d symplectic majorana spinors are written as @xmath204 where @xmath205 are four 4d two - component left - handed weyl spinors .
these satisfy the symplectic majorana condition given in ( [ 6dsymp ] ) .
the spinor superfield is given by @xmath206 , \nonumber\end{aligned}\ ] ] where @xmath207 .
on the fixed line @xmath208 , the left - handed chiral superfield @xmath209 and its complex conjugate @xmath210 are expanded with component fields as @xmath211 the action @xmath212 ( [ actions5 ] ) is written in terms of component fields as @xmath213 , \label{s5after}\end{aligned}\ ] ] where @xmath214 for @xmath215 , the operator @xmath62 is defined as @xmath216 and @xmath217 fields confined on @xmath218 , @xmath100 and their actions are obtained in a parallel way . for a generic 4d field
@xmath219 the action @xmath220 ( [ actions4 ] ) becomes @xmath221 .\end{aligned}\ ] ] with @xmath222 here superpotential and supersymmetry - breaking terms are omitted . from @xmath223 , @xmath212 and @xmath220 , the auxiliary field part of the action
is @xmath224 ) + \textrm{h.c . }
\bigg ) \nonumber \\ & & + { 1\over 2}\sum_{j=5}^6 ( -\sqrt{2 } ( \partial_j d)\sigma_j -\sqrt{2 } \sigma_j^ * \partial_j d + 2f_j^*f_j + 2\sigma_j^ * d\sigma_j -2\sigma_j^*\sigma_j
d ) + f^*f+\phi^*d\phi -\phi^*\phi d \bigg ] \nonumber \\ & & + \delta(x^6 ) \bigg [ f_l^ * f_l + \phi_l^ * d\phi_l+f_l^c f_l^{c * } -\phi_l^c d\phi_l^{c * } + \bigg ( f_l^c\nabla_5 \phi_l + \sqrt{2}\phi_l^c f_5 \phi_l + \phi_l^c \nabla_5 f_l + \textrm{h.c.}\bigg ) \bigg ] \nonumber \\ & & + \delta(x^5)\delta(x^6 ) \left[f_p^ * f_p+ \phi_p^ * d\phi_p\right ] \bigg\rbrace .\end{aligned}\ ] ] the equations of motion for auxiliary fields are @xmath225 here partial integrals are used in obtaining @xmath65 in eq.([ap : f5eq ] ) , @xmath66 in eq.([ap : f6eq ] ) and @xmath226 in eq.([ap : fheq ] ) . the operator of bosonic part of 6d bulk action ( [ componentl6 ] ) is @xmath227\right ) + \left[\phi^*,\phi\right ] \bigg)^2 \nonumber \\ & & - \left(\partial_5\sigma_6-\partial_6\sigma_5 + \sqrt{2}\left[\sigma_5,\sigma_6\right]\right ) \left(\partial_5\sigma_6^*-\partial_6\sigma_5^ * -\sqrt{2}\left[\sigma_5^*,\sigma_6^*\right]\right ) \nonumber \\ & & - \left(\partial_j\phi+\sqrt{2}\left[\sigma_j,\phi\right]\right ) \left(\partial_j\phi^*-\sqrt{2}\left[\sigma_j^*,\phi^*\right]\right ) .
\nonumber \\ & \!\!\!\cong\!\!\ ! & -{1\over 4}f_{nm}f^{nm } + \sum_{j=5}^{6}\bigg ( -{1\over 2}\partial_ja^n\partial_j a_n + ( \partial_j a^n)(\partial_n a_j + i\left[a_n , a_j\right ] )
\nonumber \\ & & -{1\over 2}(\partial^n a_j -i\left[a_j , a^n\right ] ) ( \partial_n a_j + i\left[a_n , a_j\right ] ) \bigg ) -{1\over 2 } \left(\partial_5 a_6-\partial_6 a_5 + i\left[a_5,a_6\right]\right)^2 \nonumber \\ & & -\sum_{s=1}^2 ( \partial^n\pi_s+i\left[a^n,\pi_s\right ] ) ( \partial_n\pi_s+i\left[a_n,\pi_s\right ] ) - ( { \cal d}^n \phi)^*({\cal d}_n\phi ) \nonumber \\ & & -\sum_{j=5}^{6}\sum_{s=1}^2 ( \partial_j\pi_s+i\left[a_j,\pi_s\right])^2 -\sum_{j=5}^{6}(\partial_j\phi + i\left[a_j,\phi\right ] ) ( \partial_j\phi^*-i\left[\phi^*,a_j\right ] ) \nonumber \\ & & -{1\over 2 } ( \left[\phi^*,\phi\right])^2 -\sum_{s=1}^2\left[\pi_s,\phi\right ] \left[\phi^*,\pi_s\right ] .
\label{lagbos}\end{aligned}\ ] ] here @xmath228 indicates that both hand sides are equal to each other when the trace @xmath229 is taken into account .
the bosonic operator ( [ lagbos ] ) is written in terms of bulk bosons as in eq.([bosop ] ) .
the operator of fermionic part of 6d bulk action ( [ componentl6 ] ) is @xmath230 + \textrm{h.c.}\right ) \nonumber \\ & & + \sum_{j=5}^6 \left(-i\bar{\lambda}_j\bar{\sigma}^n { \cal d}_n\lambda_j -i(\mathfrak{d}_j\lambda)\lambda_j + i\bar{\lambda}_j ( \mathfrak{d}_j\bar{\lambda})\right ) \nonumber \\ & & -i\bar{\chi}\bar{\sigma^n}{\cal d}_n\chi + \left(-\sqrt{2}i\bar{\chi}\left[\bar{\lambda},\phi\right ] + \textrm{h.c.}\right ) .
\label{lagfer}\end{aligned}\ ] ] here @xmath231 the operator ( [ lagfer ] ) becomes eq.([ferop ] ) . | we study six - dimensional ( 6d ) su(6 ) supersymmetric models where the doublet - triplet splitting , quark - lepton mass relations and gaugino - mediated supersymmetry breaking are taken into account .
we find that effective 4d gauge coupling constants have highly nontrivial behavior between two compactification scales .
it is shown that realistic patterns of yukawa coupling constants are obtained for valid values of parameters and that hierarchical numbers are generated via suppression by extra - dimensional effects .
hip-2008 - 09/th + * gauge and yukawa couplings in 6d supersymmetric su(6 ) models * nobuhiro uekusa _ department of physics , university of helsinki + and helsinki institute of physics , + p.o .
box 64 , fin-00014 helsinki , finland _
+ _ e - mail _ : [email protected] |
in this paper we study the root distribution of a sequence of polynomials satisfying one of the following three - term recurrences : @xmath10 with certain initial conditions and @xmath11 polynomials in @xmath8 with complex coefficients . for the study of the root distribution of other sequences of polynomials that satisfy three - term recurrences , see @xcite and @xcite .
in particular , we choose the initial conditions so that the generating function is @xmath12 where @xmath3 , @xmath4 , or @xmath5 .
we notice that the root distribution of @xmath0 will be the same if we replace @xmath13 in the numerator by any monomial @xmath14 .
if @xmath14 is not a monomial , the root distribution will be different .
the quadratic case @xmath3 is not difficult and it is also mentioned in @xcite .
we present this case in section 2 because it gives some directions to our main cases , the cubic and quartic denominators @xmath2 , in sections 3 and 4 .
our approach uses the concept of @xmath9-analogue of the discriminant ( @xmath9-discriminant ) introduced by ismail @xcite .
the @xmath9-discriminant of a polynomial @xmath15 of degree @xmath16 and leading coefficient @xmath17 is @xmath18 where @xmath19 , @xmath20 are the roots of @xmath15 .
this @xmath9-discriminant is @xmath21 if and only if a quotient of roots @xmath22 equals @xmath9 . as @xmath23 , this @xmath9-discriminant becomes the ordinary discriminant which is denoted by @xmath24 . for the study of resultants and ordinary discriminants and their various formulas , see @xcite , @xcite , @xcite , and @xcite .
we will see that the concept of @xmath9-discriminant is useful in proving connections between the root distribution of a sequence of polynomials @xmath0 and the discriminant of the denominator of its generating function @xmath25 .
we will show in the three cases mentioned above that the roots of @xmath0 lie on a portion of a real algebraic curve ( see theorem [ quadratic ] , theorem [ cubic ] , and theorem [ quartic ] ) .
for the study of sequences of polynomials whose roots approach fixed curves , see @xcite .
other studies of the limits of zeros of polynomials satisfying a linear homogeneous recursion whose coefficients are polynomials in @xmath8 are given in @xcite .
the @xmath9-discriminant will appear as the quotient @xmath9 of roots in @xmath26 of @xmath2 .
one advantage of looking at the quotients of roots is that , at least in the three cases above , although the roots of @xmath0 lie on a curve depending on @xmath6 and @xmath7 , the quotients of roots @xmath27 of @xmath2 lie on a fixed curve independent of these two polynomials .
we will show that this independent curve is the unit circle in the quadratic case and two peculiar curves ( see figures 1 and 2 in sections 3 and 4 ) in the cubic and quartic cases . from computer experiments ,
this curve looks more complicated in the quintic case @xmath28 ( see figure 3 in section 4 ) . as an application of these theorems
, we will consider an example where @xmath29 and @xmath30 .
we will see that the roots of @xmath0 lie either on portions of the circle of radius @xmath31 or real intervals depending on the value @xmath32 compared to the critical values @xmath21 and @xmath33 .
also , the endpoints of the curves where the roots of @xmath0 lie are roots of @xmath25 .
interestingly , the critical values @xmath21 and @xmath33 are roots of the double discriminant @xmath34
in this section , we will consider the root distribution of @xmath0 when the denominator of the generating function is @xmath3 .
[ quadratic ] let @xmath0 be a sequence of polynomials whose generating function is @xmath35 where @xmath6 and @xmath7 are polynomials in @xmath8 with complex coefficients .
the roots of @xmath0 which satisfy @xmath36 lie on the curve @xmath37 defined by @xmath38 and are dense there as @xmath39 .
suppose @xmath40 is a root of @xmath0 which satisfies @xmath41 .
let @xmath42 and @xmath43 be the roots of @xmath44 . if @xmath45 then @xmath46 .
in this case @xmath40 belongs to @xmath37 , and we only need to consider the case @xmath47 . by partial fractions , we have @xmath48 thus if we let @xmath49 then @xmath9 is an @xmath50-st root of unity and @xmath51 . by the definition of @xmath9-discriminant in , @xmath9 is a root of @xmath52 which equals @xmath53 this implies that @xmath54 thus @xmath55 since @xmath9 is an @xmath50-th root of unity .
the map @xmath56 maps an open neighborhood @xmath57 of a point on @xmath37 onto an open set which contains a point @xmath58 , where @xmath9 is an @xmath50-th root of unity , when @xmath59 is large . from
, there is a solution of @xmath0 in @xmath57 .
the density of the roots of @xmath0 follows .
we consider an example in which the generating function of @xmath0 is given by @xmath60 where @xmath30 .
let @xmath61 .
we exhibit the three possible cases for the root distribution of @xmath0 depending on @xmath32 : 1 . if @xmath62 , the roots of @xmath0 lie on the two real intervals defined by @xmath63 2 .
if @xmath64 , the roots of @xmath0 can lie either on the half circle @xmath65 , @xmath66 , or on the real interval defined by @xmath67 .
3 . if @xmath68 , the roots of @xmath0 lie on two parts of the circle @xmath65 restricted by @xmath69 . indeed , by complex expansion , we have @xmath70 where @xmath71 theorem [ quadratic ] yields three cases : @xmath72 , @xmath73 or @xmath74 . since @xmath75 , all these cases give @xmath76 .
we note that if @xmath73 then the condition @xmath77 reduces to @xmath78 suppose @xmath62 .
then the condition @xmath76 implies that the roots of @xmath0 are real .
the condition @xmath77 becomes @xmath79 suppose @xmath64 .
the roots of @xmath0 lie either on the half circle @xmath73 , @xmath66 ( from the inequality ) , or on the real interval given by @xmath67 ( from the inequality ) . if @xmath68 then the roots of @xmath0 lie on the two parts of the circle @xmath73 restricted by @xmath69 ( from the inequality ) .
we notice that in this example , the inequality @xmath77 gives the endpoints of the curves where the roots of @xmath0 lie .
thus , these endpoints are roots of @xmath80 . moreover the critical values of @xmath32 , which are @xmath21 and @xmath33 , are roots of the double discriminant of the denominator @xmath81 this comes from the fact that the endpoints of the fixed curves containing the roots of @xmath0 are the roots of @xmath82 .
when this discriminant has a double root as a polynomial in @xmath8 , some two endpoints of the fixed curves coincide .
that explains the change in the shape of the root distribution .
in this section we show that in the cubic case @xmath4 , the roots of @xmath0 lie on a portion of a real algebraic curve . as we see in the proof of theorem [ quadratic ]
, we can first consider the distribution of the quotients of roots @xmath83 of @xmath2 , and then we can relate to the root distribution of @xmath0 using the @xmath9-discriminant .
while in the previous section this quotient lies on the unit circle , in this section we show that this quotient lie on the curve in figure 1 .
[ cubicqlemma ] suppose @xmath84 are complex numbers such that @xmath85 and @xmath86 then @xmath87 and @xmath88 lie on the union @xmath89 where the cartesian equations of @xmath90 , @xmath91 and @xmath92 are given by @xmath93 and are dense there as @xmath39 .
we can rewrite as @xmath94 where we can replace @xmath88 by @xmath95 . by multiplying both sides by @xmath96
, we note that there are at most @xmath97 solutions @xmath98 counting multiplicity .
let @xmath99 where @xmath100 . from implicit differentiation , we can check that the equation has roots at @xmath101 with multiplicity @xmath102 . after subtracting this number of roots from @xmath97 ,
we conclude that there are at most @xmath103 roots @xmath104 .
we first show that if @xmath105 is a root , then so is @xmath106 . from the two equations in the hypothesis
, we note that @xmath107 and @xmath108 subtracting @xmath13 , then dividing by @xmath109 and multiplying both sides by @xmath110 , we obtain @xmath111 by interchanging the summation and reversing the index of summation we obtain @xmath112 hence we have symmetry between @xmath109 and @xmath113 in the two double summations .
our goal is to show that the number of roots @xmath104 on @xmath89 is at least @xmath103 , counting multiplicities .
then all roots will lie on @xmath89 since we have at most @xmath103 roots @xmath104 . by the symmetry of roots mentioned above ,
if @xmath105 is a solution in @xmath90 then @xmath114 is a solution in @xmath115 .
hence there is a bijection between roots in @xmath116 , @xmath91 and @xmath92 .
thus if @xmath117 contains at least @xmath118 roots then all of the roots lie on @xmath89 .
let @xmath119 be a root on @xmath117 .
then the equation @xmath85 gives @xmath120 .
thus gives @xmath121 write @xmath122 where @xmath123 , @xmath124
. then complex expansion yields @xmath125 divide @xmath126 , replace @xmath126 by @xmath127 and combine the first two terms to obtain @xmath128 we note that the right side has different signs if @xmath129 and @xmath130 .
thus we can apply the intermediate value theorem on several intervals whose boundaries are the solutions of @xmath131 .
the equations @xmath131 give @xmath132 the condition @xmath133 and the fact that @xmath99 , @xmath100 , yield @xmath134 if @xmath135 , we have at least @xmath118 roots coming from @xmath118 intervals formed by the @xmath136 points @xmath137 where @xmath138 . if @xmath139 , we have at least @xmath118 roots coming from @xmath118 intervals formed by the @xmath136 points @xmath140 if @xmath141 , we have at least @xmath118 roots coming from @xmath118 intervals formed by the @xmath136 points @xmath137 where @xmath142 .
the density follows from the distribution of @xmath118 roots mentioned above .
the lemma follows .
[ cubic ] let @xmath0 be a sequence of polynomials whose generating function is @xmath143 where @xmath6 and @xmath7 are polynomials in @xmath8 with complex coefficients .
the roots of @xmath0 which satisfy @xmath36 lie on the curve @xmath144 defined by @xmath145 and are dense there as @xmath39 . for a little simplification ,
we consider the roots of @xmath146 .
let @xmath40 be a root of @xmath146 which satisfies @xmath41 .
let @xmath42 , @xmath43 and @xmath147 be the roots of @xmath148 .
it suffices to consider @xmath149 . by partial fractions ,
the function @xmath150 is @xmath151 we expand @xmath152 using geometric series and write the expression above as @xmath153 since @xmath40 is a root of @xmath146 , we have @xmath154 we divide this equation by @xmath155 and let @xmath156 , @xmath157 to obtain @xmath158 where @xmath159 since @xmath160 .
the equation can be written as @xmath161 since @xmath162 and @xmath163 , this equation becomes @xmath164 let @xmath165 and @xmath166 and add 1 to both sides .
then @xmath167 thus @xmath87 and @xmath88 ( and also @xmath168 and @xmath169 ) lie on the curve given in lemma [ cubicqlemma ] .
since @xmath168 and @xmath169 are given by quotients of two roots , they are roots of the @xmath9-discriminant given by @xmath170 this gives @xmath171 it remains to show that the map @xmath172 maps the curve in figure [ quotientcubicfig ] to the real interval @xmath173 $ ] .
let @xmath9 be a point on the this curve .
we note that @xmath174 since @xmath9 lies on the curve in figure [ quotientcubicfig ] , we have the three possible cases @xmath175 , @xmath176 or @xmath177 . in the first case , @xmath178 since @xmath179 . in the second and third cases , @xmath178 since @xmath180 and @xmath181 .
furthermore , @xmath182 attains its minimum and maximum when @xmath183 and @xmath184 respectively .
the density of the roots of @xmath0 follows from similar arguments as in the proof of theorem [ quadratic ] .
in this section , we will show that in the case @xmath185 the roots of @xmath0 lie on a portion of a real algebraic curve .
similar to the approach in the previous sections , we first consider the distribution of the quotients of roots of @xmath2 . before looking at these quotients , let us recall that the chebyshev polynomial of the second kind @xmath186 is @xmath187 where @xmath188 suppose @xmath189 such that @xmath190 .
let @xmath191 and @xmath192 . if @xmath193 is a positive integer then @xmath194 by analytic continuation , we can extend this identity to any pair of complex numbers @xmath195 and @xmath196 with @xmath197 suppose @xmath40 is a root of @xmath0 and @xmath198 is a quotient of two roots in @xmath26 of @xmath199 .
then the set of all such quotients belongs to the curve depicted in figure 2 , where the cartesian equation of the quartic curve on the left is @xmath200 and the curve on the right is the unit circle with real part at least @xmath201 .
all such quotients are dense on this curve as @xmath39 . for each @xmath202 ,
let @xmath42 , @xmath43 , @xmath147 , and @xmath203 be the roots of the denominator @xmath199 . by partial fractions , we have @xmath204 where @xmath205
let @xmath206 , @xmath207 , @xmath208 .
for a little reduction in the powers of @xmath209 , @xmath210 , we will consider the roots of the polynomial @xmath211 .
we put all terms of @xmath212 over a common denominator and then divide the numerator by @xmath213 .
the condition @xmath214 implies @xmath215 the fact @xmath216 gives @xmath217 from the symmetric reductions , the right side of , after being divided by @xmath218 , is a polynomial in @xmath168 of degree @xmath219 .
we used a computer algebra system to check for the root distribution of this polynomial in the case @xmath220 .
we now assume that @xmath221 .
we will show that the number of roots @xmath168 lying on the two curves in figure [ quotientquarticfig ] is at least @xmath219 .
the first step is to show that if the set of @xmath168 belongs to the unit circle with @xmath222 and is dense there as @xmath39 then the set of @xmath169 and @xmath223 belongs to the quartic curve given in the lemma and is dense on this quartic curve as @xmath39 .
then we will find the number of roots @xmath168 on the unit circle with @xmath222 .
suppose @xmath224 lies on the unit circle and @xmath225 .
we note that @xmath169 and @xmath223 are the two roots of the equation @xmath226 thus the quadratic formula gives @xmath227 splitting the real and imaginary parts of the function on the left side , we leave it to the reader to check that this function maps the interval @xmath225 to the quartic curve @xmath228 we now compute the number of roots @xmath224 with @xmath229 .
we first consider @xmath230 .
let @xmath231 equation gives @xmath232 where @xmath233 we divide by @xmath218 and rewrite it in terms of chebyshev polynomials : we divide this equation by @xmath235 and write @xmath236 in terms of chebyshev polynomials .
we obtain @xmath237 where @xmath238 finally , from we can replace @xmath239 by @xmath240 and use the recurrence definition of the chebyshev polynomials to rewrite this equation in the symmetric form below : @xmath241 from this symmetric form , the right expression remains the same if we interchange @xmath109 and @xmath242 or if we interchange @xmath243 and @xmath244 ( from and ) .
thus the numbers of roots @xmath168 are the same in the two cases @xmath245 and @xmath246 .
it is sufficient to count the number of roots @xmath245 or @xmath247 .
let @xmath248 and @xmath249 where @xmath250 , @xmath251 .
the idea is to show that in this case the summand @xmath252 dominates the right expression of . since @xmath253 and @xmath254 in are real numbers and the chebyshev polynomials in this equation are either even or odd , we can apply the intermediate value theorem .
we note that has different signs when @xmath255 and when @xmath256 .
suppose @xmath257 and @xmath250 . since @xmath258 it suffices to show @xmath259 let @xmath260 .
using the fact that @xmath261 and @xmath262 , we obtain the following upper bound for the right hand side of : @xmath263 since @xmath264 where @xmath265 and @xmath250 , we have @xmath266 thus @xmath267 this inequality and give @xmath268 from the definition of the chebyshev polynomial , we have @xmath269 with similar computations for @xmath270 , we obtain @xmath271 since @xmath257 and @xmath272 , the right side is at least @xmath273 .
we combine this with to have when @xmath275 the inequality follows . by the intermediate value theorem
, we have at least one root when @xmath276 changes between @xmath277 and @xmath13 with @xmath250 . from the formula , the number of roots @xmath168 when @xmath245 is at least @xmath278 . by symmetry , the number of roots @xmath230 with @xmath279 on the unit circle is at least @xmath280 . note that each of these roots gives two more roots @xmath169 and @xmath223 on the quartic curve .
it remains to check the multiplicities of @xmath281 in the equation .
we note that this equation has a root @xmath282 with multiplicity at least 1 . in the case
@xmath282 we obtain four more roots @xmath169 , @xmath283 , @xmath223 , and @xmath284 .
we now consider the case @xmath285 .
the equation @xmath286 where @xmath287 gives @xmath288 or @xmath289 when @xmath290 and @xmath291 or @xmath292 when @xmath293 .
hence each of the roots @xmath285 gives us another root at @xmath277 with the same multiplicity . to check the multiplicities at @xmath285
, we need to differentiate the equation with respect to @xmath168 .
we obtain its derivatives by applying implicit differentiation to the equation @xmath286 . after substituting @xmath285 in and its derivatives
, we see that the multiplicity of @xmath294 is @xmath295 all the roots counted on the table lie on the curves given in the lemma .
the number of roots counted equals the number of possible roots which is @xmath219 . also , as a consequence of the intermediate value theorem applied to the intervals formed by @xmath257 , the roots @xmath168 are dense on the portion of the unit circle with real part at least @xmath201 .
the lemma follows .
let @xmath0 be a sequence of polynomials whose generating function is @xmath296 where @xmath6 and @xmath7 are polynomials in @xmath8 with complex coefficients
. the roots of @xmath0 which satisfy @xmath36 lie on the curve @xmath297 defined by @xmath298 and are dense there as @xmath39 . from the definition of @xmath9-discriminant in
, we have @xmath299 if @xmath9 is a quotient of two roots of @xmath300 , then @xmath301 let @xmath182 be the function on the right side .
we note that @xmath182 maps @xmath302 with @xmath222 to the real interval @xmath303 $ ] since @xmath304 if @xmath9 is a point on the quartic curve in lemma [ quarticqlemma ] then @xmath9 and @xmath168 are related by @xmath305 multiplying this equation by @xmath306 , we obtain @xmath307 thus by the definition of @xmath182 , we have @xmath308 . since @xmath309 the roots of @xmath0 lie on the curve @xmath297 .
the density of these roots follows from arguments similar to those in the proof of theorem [ quadratic ] .
* remark : * one may try to find the root distribution of @xmath0 in the case @xmath310 . from computer experiments ,
the distribution of the quotients of roots of @xmath2 in the case @xmath311 is given in the figure below .
let @xmath0 be a sequence of polynomials whose generating function is @xmath312 where @xmath6 and @xmath7 are polynomials in @xmath8 with complex coefficients .
the roots of @xmath0 which satisfy @xmath36 lie on the curve @xmath313 defined by @xmath314 and are dense there as @xmath39 .
s. beraha , j. kahane , n. j. weiss , limits of zeros of recursively defined families of polynomials , studies in foundations and combinatorics , pp .
213232 , adv . in math .
suppl . stud .
, 1 , academic press , new york - london , 1978 . | let @xmath0 be a sequence of polynomials whose generating function @xmath1 is the reciprocal of a bivariate polynomial @xmath2 .
we show that in the three cases @xmath3 , @xmath4 and @xmath5 , where @xmath6 and @xmath7 are any polynomials in @xmath8 with complex coefficients , the roots of @xmath0 lie on a portion of a real algebraic curve whose equation is explicitly given .
the proofs involve the @xmath9-analogue of the discriminant , a concept introduced by mourad ismail . |
the notion of an @xmath0-algebra ( _ i.e. _ algebra over the sphere spectrum ) is well - known in homotopy theory ( _ cf . _ _ e.g. _ @xcite ) : in the categorical form used in this paper and implemented by the concept of a discrete @xmath1-ring , it was formalized in the late 90 s .
@xmath0-algebras are also intimately related to the theory of brave new rings ( first introduced in the 80 s ) and of functors with smash products ( fsp ) for spectra in algebraic topology ( _ cf .
_ @xcite ) .
the goal of this article is to explain how the implementation of the notion of an @xmath0-algebra in arithmetic , in terms of segal s @xmath1-rings , succeeds to unify several constructions pursued in recent times aimed to define the notion of absolute algebra " . in particular
, we refer to the development , for applications in number theory and algebraic geometry , of a suitable framework apt to provide a rigorous meaning to the process of taking the limit of geometry over finite fields @xmath2 as @xmath3 . in our previous work
we have met and used at least three possible categories suitable to handle this unification : namely the category @xmath4 of monods as in @xcite , the category @xmath5 of hyperrings of @xcite and finally the category @xmath6 of semirings as in @xcite.in @xcite , n. durov developed a geometry over @xmath7 intended for arakelov theory applications by implementing monads as generalizations of classical rings . in his work
certain combinatorial structures replace , at the archimedean place(s ) , the geometric constructions classically performed , at a finite prime ideal , in the process of reduction modulo that ideal . in the present article
we argue that all the relevant constructions in _ op.cit . _ can be naturally subsumed by the well - known theory of @xmath0-algebras and segal s @xmath1-sets - space is needed in homotopy theory and
is simply that of a simplicial @xmath1-set . ] in homotopy theory .
this theory is taken up as the natural groundwork in the recent book @xcite . in proposition
[ assembly ] we prove that the assembly map of @xcite provides a functorial way to associate an @xmath0-algebra to a monad on pointed sets . while in the context of @xcite the tensor product @xmath8 produces an uninteresting output isomorphic to @xmath9 , in proposition [ tensq ] we show that the same tensor square , re - understood in the theory of @xmath0-algebras , provides a highly non - trivial object . as explained in @xcite , the category of pointed @xmath1-sets is a symmetric closed monoidal category and the theory of generalized schemes of b. ten and m. vaqui in @xcite applies directly to this category while it could not be implemented in the category of endofunctors under composition of @xcite which is not symmetric .
we endow the arakelov compactification @xmath10 of @xmath11 with a natural structure of a sheaf @xmath12 of @xmath0-algebras and each arakelov divisor provides a natural sheaf of modules over that structure sheaf .
moreover , this new structure of @xmath10 over @xmath0 endorses a one parameter group of weakly invertible sheaves whose tensor product rule is the same as the composition rule of the frobenius correspondences over the arithmetic site @xcite .
the fundamental advantage of having unified the various attempts done in recent times to provide a suitable definition of absolute algebra " by means of the well established concept of @xmath0-algebra is that this latter notion is at the root of the theory of topological cyclic homology which can be understood as cyclic homology over the absolute base @xmath0 .
segal s @xmath1-spaces form a model for stable homotopy theory and they may be equivalently viewed as simplicial objects in the category of @xmath1-sets ( _ cf._@xcite ) .
note that general @xmath1-spaces in _ op.cit .
_ are not assumed to be special or very special , and it is only by dropping the very special condition that one can use the day product .
moreover , the prolongation of a @xmath1-space to a spectrum is a purely formal construction and does not need the special / very special hypotheses .
the very special condition only appears for fibrant " objects in the appropriate quillen model category .
thus , the stable homotopy category , understood using general @xmath1-spaces , is properly viewed as the derived category of the algebraic category of @xmath1-sets , _ i.e. _ of @xmath0-modules .
topological cyclic homology then appears as the direct analogue of ordinary cyclic homology of rings when one replaces the category of abelian groups by the category of @xmath1-sets . in particular
it is now available to understand the new structure of @xmath10 using its structure sheaf @xmath12 and the modules .
our original motivation for using cyclic homology in the arithmetic context arose from the following two results:@xmath13 in @xcite we showed that cyclic homology ( a fundamental tool in noncommutative geometry ) determines the correct infinite dimensional ( co)homological theory for arithmetic varieties to recast the archimedean local factors of serre as regularized determinants .
the key operator in this context is the generator of the @xmath14-operations in cyclic theory.@xmath15 l. hesselholt and i. madsen have proven that the de rham - witt complex , an essential ingredient of crystalline cohomology , arises naturally when one studies the topological cyclic homology of smooth algebras over a perfect field of finite characteristic ( _ cf . _ _ e.g. _ @xcite ) .
our long term objective is to use cyclic homology over the absolute base @xmath0 and the arithmetic site defined in @xcite to obtain a global interpretation of @xmath16-functions of arithmetic varieties .
before recalling the definition of a @xmath1-set we comment on its conceptual yet simple meaning .
a @xmath1-set is the most embracing generalization of the datum provided on a set by a commutative addition with a zero element .
the commutativity of the addition is encoded by the fact that the output of a ( finite ) sum depends only upon the corresponding finite set ( with multiplicity ) .
a preliminary definition of a @xmath1-set is given by a covariant functor @xmath17 from the category @xmath18 of finite sets to the category @xmath19 of sets , mapping a finite set @xmath20 to the set @xmath21 of all possible strings of elements indexed by @xmath20 which can be added .
the action of @xmath17 on morphisms encodes the addition , while its covariance embodies the associativity of the operation .
one introduces a base point @xmath22 in the labelling of the sums and a base point @xmath23 to implement the presence of the @xmath24-element ( neutral for the sum ) .
thus @xmath17 is best described by a pointed covariant functor from the category @xmath25 of finite pointed sets to the category @xmath26 of pointed sets .
note that while the category of covariant functors @xmath27 is a topos the category of pointed covariant functors @xmath28 is no longer such and admits an initial object which is also final , in direct analogy with the category of abelian groups .
the theory reviewed in chapter ii of @xcite is that of segal s @xmath1-spaces , _ i.e. _ of simplicial @xmath1-sets . the category of @xmath1-spaces admits a natural structure of model category ( _ cf.__op.cit . _
definition 2.2.1.5 ) allowing one to do homotopical algebra .
the transition from @xmath1-sets to @xmath1-spaces parallels the transition from abelian groups to chain complexes of abelian groups in positive degrees and the constructions of topological hochschild and cyclic homology then become parallel to the construction of their algebraic ancestors .
for each integer @xmath29 , one introduces the pointed finite set @xmath30 , where @xmath24 is the base point .
let @xmath31 be the small , full subcategory of @xmath25 whose objects are the sets @xmath32 s , for @xmath29 .
the morphisms in the category are maps between pointed finite sets , preserving the base points .
the notion of a discrete @xmath1-space , _ i.e. _ of a @xmath1-set suffices for our applications and at the same time it is intimately related to the topos @xmath33 of covariant functors @xmath34 .
the small category @xmath35 is a pointed category _
i.e. _ it admits a ( unique ) initial and final object , namely the object @xmath36 formed by the base point alone . in general , if @xmath37 is a pointed category one defines a @xmath1-object of @xmath37 to be a covariant functor @xmath38 preserving the base point .
[ defngamset ] a @xmath1-set @xmath17 is a functor @xmath39 between pointed categories .
the morphisms @xmath40 between two @xmath1-sets are natural transformations of functors .
the category @xmath41 of @xmath1-sets is a symmetric , closed , monoidal category ( _ cf . _
@xcite , chapter ii ) .
the monoidal structure is given by the smash product of @xmath1-sets which is an example of a day product : we shall review this product in [ salgsect ] .
the closed structure property is shown in @xcite ( _ cf . _
also @xcite theorem 2.1.2.4 : replace the category @xmath42 with @xmath43 ) .
the internal @xmath44 structure is given by @xmath45 where @xmath46 .
for two pointed sets @xmath47 their smash product is defined by collapsing the set @xmath48 to @xmath49 in the product @xmath50 .
our next task is to explain how the closed monoidal category @xmath41 encompasses several attempts to model a notion of an absolute algebra " .
let @xmath51 be a commutative monod denoted additively and with a @xmath24 element .
one defines a functor @xmath52 by setting @xmath53 in the last formula @xmath54 and the sum over the empty set is the @xmath24 element .
moreover the base point of @xmath55 is @xmath56 , @xmath57 .
the formula is meaningful in view of the commutativity and associativity of the monod @xmath51 .
one easily checks that the maps @xmath58 with @xmath59 do not create any problem for the functoriality .
the functor @xmath60 is in fact a particular case of the covariant functor @xmath61 , @xmath62 defined by the formula @xmath63 .
more precisely , @xmath60 is obtained by restricting to @xmath25 and by subsequently dividing using the equivalence relation @xmath64 by assuming that @xmath65 preserves the base point and if @xmath66 , one gets @xmath67 next , we recall the notion of an @xmath0-algebra as given in @xcite ( definition 2.1.4.1 ) .
this requires to define first the smash product of two @xmath1-sets , _ i.e. _ of two pointed functors @xmath68 , @xmath69 .
the definition of the smash product @xmath70 is dictated by the internal @xmath44 structure of and the adjunction formula @xmath71 thus the smash product @xmath70 is such that for any @xmath1-set @xmath72 , a morphism @xmath73 is simply described by a natural transformation of bifunctors @xmath74 _ i.e. _ by maps of pointed sets , natural in both objects @xmath47 of @xmath35 @xmath75 the evaluation of the @xmath1-set @xmath70 on an object @xmath76 of @xmath25 is given by the following colimit @xmath77 where for any morphisms @xmath78 and @xmath79 in @xmath25 one uses the morphism @xmath80 provided that @xmath81 , with @xmath82 .
thus , with the exception of the base point , a point of @xmath83 is represented by the data @xmath84 given by a pair of objects @xmath47 of @xmath25 , a map @xmath85 and a pair of non - base points @xmath86 , @xmath87 . moreover ,
notice the following implication @xmath88 the colimit is in general not filtered .
the specialization of definition 2.1.4.1 . of @xcite to the case of @xmath1-sets yields the following [ defnsalg ] an @xmath0-algebra @xmath89 is a @xmath1-set @xmath90 endowed with an associative multiplication @xmath91 and a unit @xmath92 , where @xmath93 is the canonical inclusion functor .
since @xmath0 is the inclusion functor , the obvious identity map @xmath94 defines the product @xmath95 in @xmath0 which is clearly associative . by construction ,
any @xmath0-algebra is an algebra over @xmath0 .
this elementary categorical object can be naturally associated to @xmath7 _ i.e. _ to the most basic algebraic structure underlying the absolute ( geometric ) point .
a morphism @xmath96 of @xmath0-algebras is a morphism of the underlying @xmath1-sets which is compatible with the unit and with the product .
this latter condition is equivalent to the commutativity of the following diagram for any objects @xmath47 of @xmath35 @xmath97^{\mu_{\ca } } \ar[rr]^{\rho_x\wedge \rho_y } & & \ar[d]^{\mu_\cb } \cb(x)\wedge \cb(y ) \\
\ca(x\wedge y ) \ar[rr]_{\rho_{x\wedge y } } & & \cb(x\wedge y ) } \ ] ] we let @xmath98 be the category of @xmath0-algebras .
it follows from _ op.cit . _ ( _ cf . _
2.1.4.1.6 ) that , given an @xmath0-algebra @xmath99 and an integer @xmath100 , one obtains an @xmath0-algebra of @xmath101 matrices by endowing the @xmath1-set @xmath102 with the natural multiplication of matrices having only one non - zero entry in each column .
there is moreover a straightforward notion of module over an @xmath0-algebra ( _ cf . _
_ definition 2.1.5.1 ) : an @xmath0-module being just a @xmath1-set .
we begin this section by reviewing an easy construction of an @xmath0-algebra derived from a functor from the category of ( not necessarily commutative ) multiplicative monods with a unit and a zero element , to the category of @xmath0-algebras .
this construction is described in example 2.1.4.3 , 2 . of @xcite , where the monod @xmath51 is not assumed to have a @xmath24 element and the obtained @xmath0-algebra
is called spherical monod algebra .
let @xmath51 be a multiplicative monod with a multiplicative unit and a zero element @xmath24 .
we define the covariant functor @xmath103 with @xmath104 viewed as the base point and with maps @xmath105 , for @xmath65 .
[ mono2sss ] let @xmath51 be a multiplicative monod with a unit and a zero element .
then the product in @xmath51 endows @xmath106 with a structure of an @xmath0-algebra .
the product in @xmath51 , viewed as a map @xmath107 determines the following map , natural in both objects @xmath20 , @xmath108 of @xmath25 @xmath109 the multiplicative unit @xmath110 determines a natural transformation @xmath111 .
this construction endows @xmath106 with a structure of an @xmath0-algebra.the multiplicative monod @xmath112 ( frequently denoted by @xmath7 ) determines , in this framework , the canonical inclusion functor @xmath0 : _ i.e. _ @xmath113 .
conversely , given an @xmath0-algebra @xmath114 one obtains , using the product @xmath115 and the unit @xmath116 , a canonical structure of multiplicative monod ( with a base point @xmath24 and a multiplicative identity @xmath117 ) on the set @xmath118 .
[ monomono ] let @xmath99 be an @xmath0-algebra , @xmath51 a monod and @xmath119 a morphism of monods , then the following map defines a morphism of @xmath0-algebras @xmath120 the map @xmath121 gives the adjunction @xmath122 .
the unit @xmath116 for @xmath99 defines a natural transformation of functors compatible with the product .
since the product @xmath123 is natural in @xmath20 and @xmath108 , this shows that by taking @xmath124 , the morphism @xmath125 defines a natural transformation @xmath126 .
the associativity of @xmath127 shows that @xmath125 is multiplicative .
the required adjunction then follows using .note that the counit of the adjunction gives a canonical morphism @xmath128 .
a semiring @xmath129 is a set endowed with two binary operations : @xmath130 and @xmath131 .
the addition @xmath130 defines on @xmath129 the structure of an additive , commutative mono " i d with neutral element @xmath132 ; the multiplication @xmath133 is left and right distributive with respect to the addition and defines on @xmath129 the structure of a multiplicative mono " i d with neutral element @xmath117 and absorbing element @xmath24 .
we let @xmath134 be the category of semirings .
an important construction of an @xmath0-algebra is provided by the following result ( _ cf._@xcite ) [ sssalg ] let @xmath129 be a semiring , then the functor @xmath135 , @xmath136 is naturally endowed with a structure of an @xmath0-algebra . the additive structure on @xmath129 defines the @xmath1-set @xmath137 using .
we first describe the product and the unit @xmath138 . to define the product
we introduce the following map , natural in both objects @xmath47 of @xmath35 @xmath139 the naturality of the above operation follows from the bilinearity of the product in @xmath129 . the unit @xmath138 is defined as follows @xmath140 one obtains in this way a natural transformation since there is at most one non - zero value in the sum @xmath141 which computes @xmath142 .
moreover , a non - zero value occurs exactly when @xmath143 , which shows that @xmath144 .
notice that we have defined both the product and the transformation @xmath138 without using the additive group structure of a ring : in fact the semiring structure suffices .
finally , the axioms of @xmath0-algebras are checked in the same way as for rings ( _ cf._@xcite example 2.1.4.3 for details ) .
[ sssalg1 ] the functor @xmath145 from semirings to @xmath0-algebras is fully faithful .
one needs to check that , for two semirings @xmath146 the natural map @xmath147 is bijective . as a set ,
one obtains the bijection @xmath148 by using the map defined by @xmath149 , @xmath150 , which maps the base point to @xmath151 . the product in @xmath152
is recovered by the map @xmath153 , using the fact that @xmath154.notice that for a pointed functor @xmath155 of the form @xmath156 , with @xmath152 a semiring , one derives the special property that given two elements @xmath157 there exists a _ unique _
element @xmath158 whose images by the maps @xmath159 of the form @xmath160 are given by @xmath161 , @xmath162 .
one then obtains @xmath163 this shows that a morphism of functors @xmath164 determines , by restriction to @xmath165 , a homomorphism @xmath166 of semirings .
the uniqueness of this homomorphism is clear using the bijection @xmath167 .
the equality @xmath168 follows from the naturality of @xmath169 which implies that the projections @xmath170 , @xmath171 , fulfill @xmath172 the formula for the addition in @xmath173 retains a meaning also in the case where the above special property , for a pointed functor @xmath155 is relaxed by dropping the condition that the solution to @xmath174 , @xmath175 is unique . in this case , one can still define the following generalized addition _
i.e. _ the hyper - operation which associates to a pair @xmath176 a subset of @xmath177 @xmath178 this fact suggests that one can associate an @xmath0-algebra to a hyperring .
this construction will be described in more details in section [ secthyp ] .
the apparent simplicity of the smash product of two @xmath1-sets in the day product hides in fact a significant difficulty of the concrete computation of the colimit defined in . in this section
we consider the specific example of the @xmath0-algebra @xmath179 , where @xmath180 is the smallest semiring of characteristic one ( idempotent ) and compute explicitly the smash product @xmath181 .
we start by producing an explicit description of the @xmath0-algebra @xmath179 .
[ sssalg1 ] the @xmath0-algebra @xmath179 is canonically isomorphic to the functor @xmath182 which associates to an object @xmath20 of @xmath35 the set @xmath183 of subsets of @xmath20 containing the base point .
the functoriality is given by the direct image @xmath184 .
the smash product is provided by the map @xmath185 which associates to the pair @xmath186 the smash product @xmath187 .
the unit @xmath188 is given by the natural transformation @xmath189 defined by the map @xmath190 .
a map @xmath191 is specified by the subset @xmath192 .
one associates to @xmath193 the subset @xmath194 .
the formula @xmath195 which defines the functoriality shows that it corresponds to the direct image @xmath184 .
the last two statements are straightforward to check using the product in @xmath196 .
[ f2]it is interesting to compare the structures of the @xmath0-algebras @xmath179 and @xmath197 because , when evaluated on an object @xmath20 of @xmath35 , they both yield the same set . @xmath198 is indeed equal to the set @xmath183 of subsets of @xmath20 containing the base point : this equality is provided by associating to @xmath199 the subset @xmath194 .
the products and the unit maps are the same in both constructions since the product in @xmath200 is the same as that in @xmath196 .
the difference between the two constructions becomes finally visible by analyzing the functoriality property of the maps .
the simple rule @xmath184 holding in @xmath179 is replaced , in the case of @xmath197 by adding the further condition that on the set @xmath201 one only retains the points @xmath202 such that the cardinality of the preimage @xmath203 is an odd number .
it follows from the colimit definition that @xmath205 is the set @xmath206 of connected components of the following category @xmath207 .
the objects of @xmath207 are 4-tuples @xmath208 , where @xmath47 are objects of @xmath25 , @xmath209 is a morphism in that category and @xmath210 .
a morphism @xmath211 in @xmath207 is given by a pair of pointed maps @xmath78 , @xmath79 such that @xmath212 and @xmath213 .
notice that the set @xmath183 of subsets of @xmath20 containing the base point can be equivalently viewed as the set of all subsets of @xmath214 .
thus @xmath215 , @xmath216 , is determined by a pair @xmath186 of non - empty subsets , @xmath217 , @xmath218 .
one can thus encode an object @xmath208 of @xmath207 such that @xmath219 by a 5-tuple @xmath220 where @xmath217 , @xmath218 are non - empty sets .
a morphism @xmath211 of such 5-tuples is then given by a pair of pointed maps @xmath221 such that @xmath222 for @xmath223 , a morphism @xmath211 is given by a pair of pointed maps @xmath221 such that @xmath224 let @xmath208 be an object of @xmath207 then the _ support _ of @xmath225 is empty if @xmath226 and for @xmath216 , @xmath227 one lets @xmath228 an object @xmath225 of @xmath207 is _ degenerate _ iff @xmath229 .
in particular one sees that any object @xmath230 is degenerate .
[ assoc0 ] let @xmath231 be a morphism in @xmath207 then @xmath232 assume first that @xmath223 . then by the restriction of @xmath233 to @xmath234
is equal to @xmath24 .
thus @xmath229 .
assume now that @xmath235 .
for @xmath236 , one has @xmath237 , hence @xmath238 , @xmath239 and @xmath240 .
conversely , let @xmath241 , then since @xmath242 there exist @xmath243 , @xmath244 , @xmath245 , @xmath246 .
one then has @xmath247 since @xmath248 .
this shows that @xmath249.let @xmath250 be the object of @xmath207 where @xmath20 and @xmath108 are reduced to the base point .
one has [ assoc05 ] an object of @xmath207 is degenerate if and only if it belongs to the connected component of @xmath251 .
the condition @xmath229 defines a connected family of objects by lemma [ assoc0 ] .
thus it is enough to show that if @xmath229 then @xmath225 is in the same component as @xmath251 .
the inclusions @xmath252 give in general a morphism in @xmath207 @xmath253 if @xmath229 the restriction of @xmath233 to @xmath234 is equal to @xmath24 .
this shows that the unique pair of maps @xmath254 , @xmath255 gives a morphism in @xmath207 @xmath256 so that @xmath257 and @xmath251 are in the same component . to determine an explicit description of the ( pointed ) set @xmath204 we introduce the following terminology [ kmultigraph ] @xmath13
a @xmath258-_relation _ is a triple @xmath259 where @xmath17 and @xmath72 are non - empty finite sets and @xmath260 is a map of sets such that no line or column of the corresponding matrix is identically @xmath24 there exists @xmath261 such that @xmath262 and symmetrically @xmath263 , @xmath264 . ] .
@xmath15 a morphism @xmath265 between two k - relations @xmath259 and @xmath266 is a pair of surjective maps @xmath267 , @xmath268 such that @xmath269 .
@xmath270 a @xmath258-relation @xmath259 is said to be reduced when no line is repeated and no column is repeated .
we let @xmath271 be the category of @xmath258-relations .
the next lemma produces a retraction of the full subcategory of @xmath207 whose objects are non - degenerate , on @xmath271 .
[ assoc]@xmath13 let @xmath257 be a non - degenerate object of @xmath207 .
set @xmath272 , @xmath273 where @xmath274 and @xmath275 are the projections .
then the triple @xmath276 defines a k - relation .
@xmath15 a morphism @xmath277 in @xmath207 induces by restriction a morphism @xmath278 of k - relations .
@xmath270 for @xmath279 a k - relation , let @xmath280 , where @xmath281 is the unique extension of @xmath233 as a pointed map
. then @xmath282 extends to a functor from the category @xmath271 of @xmath258-relations to @xmath207 and one has @xmath283 .
@xmath284 the objects @xmath225 and @xmath285 belong to the same connected component of @xmath207 .
@xmath13 notice that @xmath286 is a subset of @xmath287 whose projections are @xmath288 and @xmath289 , thus the restriction of @xmath290 to @xmath291 defines a k - relation .
let @xmath231 be a morphism in @xmath207 between non - degenerate objects .
one has a commutative diagram @xmath292^{p_a } \ar[rr]^{f\times g } & & \ar[d]^{p_{a ' } } \supp(\alpha ' ) \\ f \ar[rr]_{f } & & f ' } \ ] ] where both @xmath293 and @xmath294 are surjective .
this shows that @xmath295 and similarly that @xmath296 .
thus the restrictions of @xmath297 and @xmath298 define surjections @xmath267 and @xmath268 , where @xmath272 , @xmath273 .
the restrictions @xmath299 , @xmath300 fulfill the equality @xmath301 and one thus obtains a morphism @xmath302 of k - relations.@xmath270 let @xmath303 be a morphism of k - relations , we extend both @xmath304 to maps of pointed sets @xmath305 , @xmath306 . then one easily checks that one obtains a morphism @xmath307 ( _ cf . _ ) .
one has by construction @xmath283 .
@xmath284 by the object @xmath257 is equivalent to @xmath308 .
we now prove that @xmath309 is equivalent to @xmath310 , where @xmath311 , @xmath312 , and @xmath313 .
let @xmath314 act as the identity on @xmath315 and map the elements of @xmath316 to the base point .
we define @xmath317 in a similar manner .
then by construction one has @xmath318 .
moreover for @xmath319 one has @xmath320 since both sides vanish unless @xmath321 and they agree on @xmath291 .
this shows that @xmath225 and @xmath285 belong to the same connected component .
every k - relation @xmath259 admits a canonical reduction @xmath322 .
this is obtained by dividing @xmath17 and @xmath72 by the following equivalence relations @xmath323 by construction , the value @xmath324 only depends upon the classes @xmath325 of @xmath326 and thus the canonical reduction map @xmath327 defines a morphism of k - relations .
more precisely [ assoc17]@xmath13 the reduction map defines an endofunctor @xmath328 of the category of @xmath258-relations which determines a retraction on the full subcategory of reduced @xmath258-relations.@xmath15 let @xmath329 be a morphism of @xmath258-relations , with @xmath330 reduced , then @xmath297 and @xmath298 are bijections and @xmath331 is an isomorphism . @xmath13 let @xmath332 be a morphism of @xmath258-relations .
we show that there exists a unique morphism @xmath333 making the following diagram commutative @xmath334^{r_c } \ar[rr]^{(f , g ) } & & \ar[d]^{r_{c ' } } c ' \\
r(c ) \ar[rr]_{(r(f),r(g ) ) } & & r(c ' ) } \ ] ] indeed , for @xmath335 one has @xmath336 , since using the equality @xmath212 and the surjectivity of @xmath298 one derives @xmath337 similarly for @xmath338 one has @xmath339 . thus @xmath297 and @xmath298 induce maps @xmath340 and @xmath341 on the reductions making the diagram commutative .
the surjectivity of @xmath297 and @xmath298 implies the surjectivity of the induced maps @xmath340 and @xmath341 .
@xmath15 let @xmath332 be a morphism of @xmath258-relations , with @xmath330 reduced .
then for @xmath342 let @xmath343 such that @xmath344 .
one has @xmath345 which shows that @xmath297 is injective .
thus both @xmath297 and @xmath298 are bijections and @xmath331 is an isomorphism . let @xmath346 be the set of isomorphism classes of reduced @xmath258-relations , and let @xmath347 be the set obtained by adjoining a base point .
we extend the association @xmath348 to a functor @xmath349 by letting a morphism @xmath350 act on elements of @xmath347 as follows @xmath351 where the right hand side reduces to the base point when @xmath352 is degenerate .
[ wedge2 ] @xmath13 for @xmath353 , the map @xmath354 induces a canonical bijection of sets @xmath355 @xmath15 the isomorphism @xmath356 induces an equivalence of functors @xmath357 .
@xmath13 by construction , @xmath205 is the union of its base point with the set @xmath206 of connected components of the category @xmath207 .
it follows from lemma [ assoc ] that the category @xmath271 of @xmath258-relations is a retraction of the category of non - degenerate objects of @xmath207 . thus using lemma [ assoc05 ]
one obtains @xmath358 .
moreover by lemma [ assoc17 ] the reduction @xmath359 gives a bijection @xmath360 .
@xmath15 this follows since by construction of @xmath181 one has @xmath361 which can be applied to any representative @xmath362 of a given class in @xmath206 .
[ pp1 ] @xmath13 for @xmath363 , let @xmath364 be the graph of the identity map on the set with @xmath365-elements .
then the @xmath117-relations @xmath364 belong to distinct connected components of the category @xmath366 and they define distinct elements of @xmath367 . @xmath15 the action of the cyclic group @xmath368 on @xmath367 is the transposition acting on isomorphism classes of reduced @xmath117-relations .
both its set of fixed points and its complement are infinite .
@xmath13 the statement follows from the fact that the @xmath117-relations @xmath364 are reduced and pairwise non - isomorphic.@xmath15 the action of the transposition @xmath369 replaces a reduced 1-relation @xmath225 by its transpose @xmath370 .
the fixed points are given by the reduced 1-relations such that @xmath225 is isomorphic to @xmath370 ; all the @xmath364 as in @xmath13 are thus fixed points .
the relations between finite sets of different cardinality determine infinitely many non fixed points .
@xmath13 in general , the natural map @xmath371 is _ not _ surjective .
this can be easily seen by choosing @xmath372 , since @xmath373 is finite unlike @xmath367 .
@xmath15 a square reduced relation is not necessarily symmetric .
for instance , among the @xmath374 reduced @xmath117-relations of figure [ krelations9 ] , only the following reduced 1-relation and its transposed @xmath375 which have equal number ( 3 ) of lines and columns are _ not symmetric _ , since the number of ( non - zero ) elements in the lines are respectively @xmath376 and @xmath377 which can not be matched by any pair of permutations .
-relations @xmath378[krelations9 ] ] -algebras[levels ] ]
the definition of [ hyperpart ] which gives back the classical addition on @xmath129 for the @xmath0-algebra @xmath137 associated to a semiring @xmath129 is multivalued for a general @xmath0-algebra @xmath379 .
the extension of ring theory when the additive group structure is replaced by a multivalued addition has been investigated by m. krasner in his theory of hyperrings and hyperfields . in our recent research
we have encountered these types of generalized algebraic structures in the following two important cases ( _ cf._@xcite ) :
@xmath380 the adle class space of a global field @xmath17 is naturally a hyperring and it contains the krasner hyperfield @xmath381 as a sub - hyperalgebra precisely because one divides the ring of the adles of @xmath17 by the non - zero elements @xmath382 of the field @xmath17 .
@xmath380 the dequantization process described by the `` universal perfection '' at the archimedean place of a number field yields a natural hyperfield structure @xmath383 on the set of real numbers and a natural hyperfield structure @xmath384 in the complex case .
the structure of a hyperfield is more rigid than that of a semifield since the operation @xmath385 is present in the hyper - context while it is missing for semirings .
this fact makes the requirement that non - zero elements are invertible more restrictive for hyperfields than for semifields .
a natural construction of hyperrings is achieved by simply dividing an ordinary ( commutative ) ring @xmath152 with a subgroup @xmath386 of the group of invertible elements of @xmath152 .
in fact , in corollary 3.10 of @xcite we proved that any ( commutative ) hyperring extension of the krasner hyperfield @xmath381 ( @xmath387 ) , without zero divisors and of dimension @xmath388 is of the form @xmath389 , where @xmath152 is a commutative ring and @xmath390 is the group of non - zero elements of a subfield @xmath315 .
moreover , theorem 3.13 of _ op.cit .
_ shows that a morphism of hyperrings of the above form lifts ( under mild non - degeneracy conditions ) to a morphism of the pairs @xmath391 .
in this section we provide the construction of the @xmath0-algebra associated to a hyperring of the form @xmath389 , where @xmath152 is a commutative ring and @xmath386 is a subgroup .
notice that the structure needed to define _ uniquely _ the @xmath0-algebra @xmath392 is provided by the pair @xmath393 and that this datum is more precise than simply assigning the hyperring @xmath389 .
[ sssalg3 ] let @xmath152 be a commutative ring and @xmath386 be a subgroup of the group of invertible elements of @xmath152 .
for each object @xmath20 of @xmath25 , let @xmath394 be the quotient of @xmath395 by the following equivalence relation @xmath396 then , the functor @xmath397 defines an @xmath0-algebra and the quotient map @xmath398 is a morphism of @xmath0-algebras .
we first check the functoriality of the construction , _
i.e. _ that for @xmath65 in @xmath25 , the map @xmath399 respects the equivalence relation @xmath400 . for @xmath401 ,
@xmath402 one has @xmath403 since these equations hold for the same @xmath298 and for all @xmath401 , one derives that @xmath404 .
the above equivalence relation is compatible with the product : in fact using the commutativity of @xmath152 , one has @xmath405 the unit @xmath406 defines by composition a natural transformation @xmath407 : @xmath408 it follows easily from the construction that the quotient map @xmath398 is a morphism of @xmath0-algebras . [ sssalgback ]
let @xmath152 be a commutative ring , let @xmath386 be a subgroup of the group of invertible elements of @xmath152 and @xmath392 the @xmath0-algebra defined in proposition [ sssalg3 ] .
then , the set @xmath409 endowed with the hyper - addition defined in and the product @xmath410 induced by the @xmath0-algebra structure , is canonically isomorphic to the hyperring @xmath389 . by using the bijection @xmath411 , @xmath412 which maps the base point to @xmath151 one derives the bijection of sets @xmath413 .
the product in @xmath409 is recovered by the map @xmath414 using the relation @xmath154 .
it is the same as the product in @xmath389 . by using and
, the hyper - addition is described , for @xmath415 as follows @xmath416 for @xmath417 and @xmath418 let @xmath419 with @xmath420 and @xmath421
. then the coset @xmath422 belongs to @xmath423 and all elements of @xmath423 are of this form . in this way one
recovers the hyper - addition on @xmath389 . given an @xmath0-algebra @xmath99 a necessary condition for @xmath99 being of the above type @xmath392 is that the equations @xmath174 , @xmath175 always admit solutions .
the lack of uniqueness of solutions is directly related to the multivalued nature of the addition .
the construction of the @xmath0-algebras @xmath392 given in proposition [ sssalg3 ] applies in particular when one uses the notion of @xmath424-model introduced in @xcite .
we recall that given a hyperfield @xmath425 , a @xmath424-_model _ of @xmath425 is by definition a triple @xmath426 where @xmath424 is a field@xmath427 is a homomorphism of hyperfields@xmath428 is a multiplicative section of @xmath169 . because @xmath428 is a multiplicative section of @xmath169 ,
the map @xmath169 is surjective and identifies the multiplicative monod @xmath425 with the quotient of @xmath424 by the multiplicative subgroup @xmath429 .
since @xmath427 is a homomorphism of hyperfields , one has @xmath430 for any @xmath431 .
thus @xmath169 defines a morphism of hyperfields @xmath432.a morphism @xmath433 of @xmath424-models is a field homomorphism @xmath434 such that @xmath435 and @xmath436 .
we recall that a @xmath424-model of @xmath425 is said to be _
universal _ if it is an initial object in the category of @xmath424-models of @xmath425 .
when such universal model exists one easily sees that it is unique up to canonical isomorphism : we denote it by @xmath437 . in this case
it is thus natural to associate to @xmath425 the @xmath0-algebra @xmath438 of its universal @xmath424-model @xmath439 , where @xmath440 .
[ shyper ] let @xmath441 , ( @xmath442 ) be the hyperfield of signs .
the associated @xmath424-model is @xmath443 where the morphism @xmath444 is given by the sign of rational numbers ( _ cf._@xcite ) .
the corresponding @xmath0-algebra is then @xmath445 .
one has @xmath446 and @xmath447 is the set of half lines @xmath16 from the origin in the rational plane @xmath448 ( including the degenerate case @xmath449 ) .
the maps @xmath450 and @xmath451 ( _ cf .
_ ) are the projections onto the two axes .
the map @xmath452 ( _ cf .
_ ) is the projection on the main diagonal .
it follows that the hyper - operation gives back the hyperfield structure on @xmath453 .
one may wonder how to relate the @xmath0-algebra @xmath445 with @xmath179 . to this end ,
one first considers the @xmath0-subalgebra @xmath454 which is defined using the sub - semiring @xmath455 .
the subset @xmath456 corresponds to the collection of half lines @xmath16 through the origin in @xmath457 which belong to the first quadrant @xmath458 .
one then defines a morphism of @xmath0-algebras @xmath459 by using the morphism of semirings @xmath460 and one also notes that @xmath461 induces a morphism of @xmath0-algebras @xmath462 .
one can thus describe the relation between @xmath99 and @xmath179 by the following map @xmath463 we expect that a similar diagram holds more generally when passing from a semifield of characteristic @xmath117 to a hyperfield and its @xmath424-model .
-algebras.[levels2 ] ] in this section we briefly discuss the concept of levels for @xmath0-algebras using which one may derive an explicit approximation of an @xmath0-algebra @xmath114 .
the first two levels of such approximation are obtained by restricting the functor @xmath99 to sets with @xmath464 elements to define the level @xmath117 , and to sets with @xmath465 elements to produce the level @xmath466 . in this way
, we obtain * level @xmath117 * the theory of multiplicative monods * level @xmath466 * the theory of hyperrings with partially defined addition . in [ sectlevel1 ]
we have seen that at level @xmath117 the @xmath0-algebra @xmath99 is well approximated by the @xmath0-subalgebra corresponding to the monod @xmath467 .
next we discuss the compatibility of the additive structure given in with the morphism of @xmath0-algebras @xmath468 . for a general mono " i d @xmath51 , implementing to @xmath106 with the same notations used there
, one gets @xmath469 one has : @xmath470 , while the elements of the set @xmath471 are the base point @xmath24 and the pairs @xmath472 for @xmath69 , @xmath473 , @xmath474 .
one also has @xmath475 and @xmath476 .
thus the only rule related to the addition that is retained at level @xmath117 simply states that the base point @xmath24 plays the role of the neutral element : @xmath477 this rule is of course preserved by the morphism of @xmath0-algebras @xmath468 . in [ fromrtohyp ] we have seen that the notion of an @xmath0-algebra is compatible with the operation of quotient by a subgroup of the multiplicative monod associated to level @xmath117 . in the following [ sectnorm ]
we shall describe how to associate a sub-@xmath0-algebra to a sub - multiplicative seminorm on a semiring . by combining these two constructions
one provides a good approximation to the description of level @xmath466 of an @xmath0-algebra .
this viewpoint shows that , in general , the operation defined by yields a partially defined hyper - addition so that the approximation of level @xmath466 is by hyperrings with partially defined addition .
the theory of @xmath0-algebras contains however all ( non - negative ) levels , even though only for the first level we have an easy explicit description .
this section is motivated by the theory developed in @xcite aiming to determine an `` absolute '' , algebraic foundation underlying arakelov geometry : in particular , we refer to the theory of monads which defines the groundwork of _ op.cit . _ first , in [ sectsemi ]
we explain how the basic construction of _ op.cit .
_ with the seminorms can be adapted in the framework of @xmath0-algebras .
then , in [ sectstructsh ] we describe a natural structure sheaf @xmath12 of @xmath0-algebras associated to the arakelov compactification @xmath10 of @xmath478 . as a topological space
, @xmath10 is obtained from the zariski topological space @xmath478 by adding a new point @xmath479 treated as an additional closed point .
this compactification is the same topological space as the space denoted by @xmath480 in _ op.cit .
_ ( _ cf .
_ [ durov1 ] ) .
we show that the arakelov divisors @xmath481 on @xmath10 give rise to sheaves of @xmath12-modules and that the smash product rule for these modules parallels the composition rule of frobenius correspondences introduced in @xcite .
let @xmath129 be a semiring .
by a sub - multiplicative seminorm on @xmath129 we mean a map @xmath482 , @xmath483 , such that @xmath484 , @xmath485 and the following inequalities hold @xmath486 the basic construction of @xcite adapts directly as follows [ sssalg2 ] @xmath13 let @xmath129 be a semiring and @xmath487 a sub - multiplicative seminorm on @xmath129 .
then the following set - up defines an @xmath0-subalgebra @xmath488 @xmath489 @xmath15 let @xmath490 be an @xmath129-module and @xmath491 a seminorm on @xmath490 such that @xmath492 , @xmath493 , @xmath494 .
then for any real @xmath495 the following structure defines a module @xmath496 over @xmath497 @xmath498 where @xmath499 is defined as in .
@xmath13 the sub - multiplicative seminorm on @xmath129 fulfills . in particular
, the triangle inequality shows that , using the formula @xmath500 , one derives @xmath501 thus , for any morphism @xmath502 , the map @xmath503 restricts to a map @xmath504 .
finally , the stability under product is derived from the sub - multiplicativity of the seminorm as follows @xmath505 @xmath15 the proof of @xmath13 shows that @xmath496 is a @xmath1-set .
it also shows that one obtains a map of sets , natural in both objects @xmath20 , @xmath108 of @xmath35 @xmath506 this gives the required @xmath497-module structure @xmath507 on @xmath496 .
[ ss1]@xmath13 the only multiplicative seminorm on the semifield @xmath196 is given by : @xmath484 , @xmath485 . in this case one obtains @xmath508 since the condition restricts the functor @xmath509 of lemma [ sssalg1 ] to the range of the unit map @xmath510.@xmath15 with the notations of , one also defines a submodule @xmath511 by setting @xmath512 we let @xmath10 be the set @xmath513 endowed with the topology whose non - empty open sets @xmath514 are the complements of finite sets @xmath17 of primes , with @xmath17 possibly containing @xmath479 .
in particular all non - empty open sets contain the generic point @xmath515 . for any open set @xmath516
the subset @xmath517 is open .
the structure sheaf @xmath518 is canonically a subsheaf of the constant sheaf @xmath519 : it associates to the complement of a finite set @xmath17 of primes the ring @xmath520 of fractions with denominators only involving primes in @xmath17 .
we view this ring as a subring of @xmath519 .
we shall construct the structure sheaf @xmath12 as a subsheaf of the constant sheaf of @xmath0-algebras @xmath521 .
first , we use the functor @xmath522 to transform @xmath518 into a sheaf of sub-@xmath0-algebras @xmath523 .
then , we extend this sheaf at the archimedean place @xmath524 by associating to any open set @xmath514 containing @xmath479 the @xmath0-subalgebra of @xmath521 given by @xmath525 where @xmath526 , @xmath527 is the usual absolute value on @xmath519 which induces a multiplicative seminorm on @xmath528 so that makes sense using .we recall that an arakelov divisor on @xmath10 is given by a pair @xmath529 where @xmath530 is a divisor on @xmath478 and @xmath531 is a real number @xmath532 .
one uses the notation @xmath533 .
the principal divisor @xmath534 associated to a rational number @xmath535 is @xmath536 . to a divisor @xmath530 on @xmath478
corresponds a sheaf @xmath537 of modules over @xmath518 and this sheaf is canonically a subsheaf of the constant sheaf @xmath519 .
[ propspecz ] @xmath13 the definition together with the inclusions @xmath538 and the restriction maps define a sheaf @xmath539 of @xmath0-algebras over @xmath540 .
@xmath15 let @xmath481 be an arakelov divisor on @xmath540 .
the following defines a sheaf @xmath541 of @xmath539-modules over @xmath540 extending the sheaf @xmath537 on @xmath11 @xmath542 the sheaf @xmath537 on @xmath11 is viewed as a subsheaf of the constant sheaf @xmath519 which gives a meaning to @xmath527 .
@xmath13 by construction the @xmath0-algebras @xmath543 are naturally @xmath0-subalgebras of @xmath521 .
moreover the restriction map @xmath544 for @xmath545 corresponds to the inclusion as @xmath0-subalgebras of @xmath521 .
these maps are injective and compatible with the product . to show that one has a sheaf of @xmath0-algebras it is enough to show that for each @xmath29
, the ( pointed ) sets @xmath546 form a sheaf of sets . given a covering @xmath547 , the elements @xmath548 of @xmath549 which agree on the intersections @xmath550 correspond to the same element of @xmath551 .
moreover this element belongs to @xmath546 since this means that @xmath552 for all @xmath553 , @xmath554 and , if @xmath555 , that @xmath556 which is true since @xmath557 for some @xmath558 .
@xmath15 let @xmath514 be a non - empty open set containing @xmath479 .
we apply proposition [ sssalg2 ] with @xmath559 and the module @xmath560 : these are naturally subsets of @xmath519 with the inherited ring and module structures .
thus the norms induced by the usual norm on @xmath519 satisfy the hypothesis of proposition [ sssalg2 ] and one obtains using the same argument as in @xmath13 , a sheaf @xmath541 of @xmath539-modules over @xmath10 .
note that at level @xmath117 , one has , for @xmath555 , @xmath561 where the positivity of a divisor is defined pointwise .
let @xmath99 be an @xmath0-algebra , @xmath51 a right @xmath99-module and @xmath562 a left @xmath99-module . by definition ( _ cf .
_ @xcite definition 2.1.5.3 ) the smash product @xmath563 is the coequalizer @xmath564 where the maps come from the two actions .
we shall use the following two properties of the smash product : the first states that @xmath565 and the second asserts that for filtered colimits of @xmath99-modules one has @xmath566 these properties can be shown by applying , for any @xmath0-module @xmath567 , the isomorphism @xmath568 , where @xmath569 is the internal one .
likewise for the tensor product of sheaves modules over sheaves of rings , the smash product @xmath570 is defined as the sheaf associated to the presheaf @xmath571 a similar argument as in proposition [ propspecz ] then applies showing that one has a natural morphism of @xmath539-modules @xmath572 next , we prove that the smash product rule for the sheaves @xmath541 , when taken over the structure sheaf @xmath539 is similar to the law of composition of frobenius correspondences over the arithmetic site of @xcite
. let @xmath533 and @xmath573 .
then the corresponding conditions at the level @xmath117 ( of the @xmath0-algebras ) determine the restrictions @xmath574 and @xmath575 .
the additive law , asserting that @xmath576 is an isomorphism , means here that the map @xmath577 is surjective , and fails to be so when both @xmath578 and @xmath579 are irrational while their product is rational .
the sheaf @xmath580 , which plays the role of the infinitesimal deformation of the identity correspondence in @xcite , is defined as follows ( using ) @xmath581 [ proppic ] @xmath13 the multiplication rule @xmath582 holds for the @xmath539-modules @xmath541 and @xmath583 except when @xmath578 and @xmath579 are irrationals while their product is rational . in that case one has @xmath584 @xmath15 two arakelov divisors @xmath585 are equivalent modulo principal divisors if and only if the sheaves @xmath541 and @xmath583 are isomorphic as abstract sheaves of @xmath539-modules .
@xmath13 one checks that @xmath576 is an isomorphism , when both @xmath578 and @xmath579 are rational : this is verified using and the fact that in that case the modules are locally trivial .
one then obtains in the case @xmath579 is rational using an increasing sequence @xmath586 such that @xmath587 $ ] for a fixed prime @xmath588 . on the open set @xmath589 one
gets that @xmath590 where @xmath591 is defined by replacing the component at @xmath479 of @xmath481 by @xmath592 .
one then applies to obtain .
finally , when @xmath579 is also irrational one uses an increasing sequence @xmath593 such that @xmath594 $ ] together with to obtain @xmath13 .
@xmath15 for any @xmath535 , one has a canonical isomorphism @xmath595 of sheaves of @xmath539-modules over @xmath10 which is defined by multiplication by @xmath596 .
more precisely , the module @xmath560 maps to @xmath597 by letting @xmath598 .
one also has @xmath599 this shows that if @xmath585 are equivalent modulo principal divisors , then the sheaves @xmath541 and @xmath583 are isomorphic as abstract sheaves of @xmath539-modules .
conversely , an isomorphism of abstract sheaves of @xmath539-modules induces an isomorphism on the set of global sections @xmath600 .
thus one obtains an invariant of the isomorphism class of abstract sheaf by setting @xmath601 where @xmath602 is the arakelov divisor with vanishing finite part .
one also has the identification @xmath603 thus @xmath604 using the multiplication rule @xmath13 one gets that @xmath605 . since every arakelov divisor is equivalent , modulo principal divisors , to a @xmath606 one gets @xmath15 .
[ picrem]to handle the non - invertibility of the sheaf @xmath541 for irrational values of @xmath607 in one can use @xmath580 ( _ cf . _ ) rather than @xmath539 as the structure sheaf .
the obtained @xmath0-algebras @xmath608 are no longer unital but one still retains the equality @xmath609 .
one defines in a similar manner the sheaves @xmath610 , using the strict inequality .
the multiplication rule is now @xmath611 this implies that the group of isomorphism classes of invertible sheaves of @xmath580-modules contains at least the quotient of the idele class group of @xmath519 by the maximal compact subgroup @xmath612 .
thus this new feature bypasses the defect of the construction of @xcite mentioned in the open question 4 .
6 of @xcite . [ cyclicrem]the above developments suggest that one should investigate the ( topological ) cyclic homology of @xmath540 , viewed as a topological space endowed with a sheaf of @xmath0-algebras and apply the topological hochschild theory to treat the cohomology of the modules @xmath541 .
[ ss12]proposition [ sssalg2 ] , combined with the general theory of schemes developed by b. ten and m. vaqui in @xcite for a symmetric closed monoidal category , provides the tools to perform the same constructions as in @xcite ( reviewed briefly below in [ durov1 ] ) , in the framework of @xmath0-algebras .
we shall not pursue this venue , rather we will now explain , using the assembly map of @xcite , a conceptual way to pass from the set - up of monads to the framework of @xmath0-algebras .
the main fact highlighted in this section is that the assembly map of @xcite determines a functorial way to associate an @xmath0-algebra to a monad on @xmath43 .
we derive as a consequence that several main objects introduced in @xcite can be naturally incorporated in the context of @xmath0-algebras .
we start off in [ durov1 ] by reviewing some basic structures defined in _ op.cit .
_ in [ sectassemb ] we recall the definition of the assembly map which plays a fundamental role , in proposition [ assembly ] , to define the functor that associates an @xmath0-algebra to a monad .
finally , proposition [ tensq ] of [ sectalgmon ] shows that the smash product @xmath0-algebra @xmath613 is ( non - trivial and ) not isomorphic to @xmath614 .
this result is in sharp contrast with the statement of @xcite that @xmath615 ( _ cf .
_ 5.1.22 p. 226 ) .
a monad on @xmath19 is defined by an endofunctor @xmath616 together with two natural transformations @xmath617 and @xmath618 which are required to fulfill certain coherence conditions ( @xcite ) .
the required associativity of the product @xmath127 is encoded by the commutativity of the following diagram , for any object @xmath20 of @xmath19 @xmath619^{\mu_{\sigma(x ) } } \ar[rr]^{\sigma(\mu(x ) ) } & & \ar[d]^{\mu_{x } } \sigma^2(x ) \\
\sigma^2(x ) \ar[rr]_{\mu_x}&&\sigma(x ) } \ ] ] there is a similar compatibility requirement for the unit transformation @xmath620 .
this set - up corresponds precisely to the notion of a monod in the monoidal category @xmath621 of endofunctors under composition , where @xmath622 is the identity endofunctor .
next , we review briefly some of the constructions of @xcite .
let @xmath129 be a semiring , then in _ op.cit .
_ one encodes @xmath129 as the endofunctor of @xmath129-linearization @xmath623 .
it associates to a set @xmath20 the set @xmath624 of finite , formal linear combinations of points of @xmath20 , namely of finite formal sums @xmath625 , where @xmath626 and @xmath627 .
one uses the addition in @xmath129 to simplify : @xmath628 . to a map of sets @xmath65
one associates the transformation @xmath629 the multiplication in @xmath129 is encoded by the elementary fact that a linear combination of finite linear combinations is still a finite linear combination , thus one derives @xmath630 this construction defines a natural transformation @xmath631 which fulfills the algebraic rules of a monad .
in fact , the rules describing a monad allow one to impose certain restrictions on the possible linear combinations which one wants to consider .
this fits well , in particular , with the structure of the convex sets so that one may for instance define the monad @xmath632 by considering as the ring @xmath129 the field @xmath633 of real numbers and by restricting to consider only the linear combinations @xmath634 satisfying the convexity condition @xmath635 .
one can eventually replace @xmath633 by @xmath519 and in this case one obtains the monad @xmath636 .
one may combine the monad @xmath636 with the ring @xmath9 as follows .
the first step takes an integer @xmath562 and considers the localized ring @xmath637 $ ] and the intersection @xmath638 .
this is simply the monad @xmath639 of finite linear combinations @xmath634 with @xmath640 and @xmath635 . by computing its prime
spectrum one finds that @xmath641 .
one then reinstalls the missing primes _ i.e. _ the set @xmath642 , by gluing @xmath643 with @xmath644 on the large open set that they have in common .
one obtains in this way the set @xmath645 finally , one eliminates the integer @xmath562 by taking a suitable projective limit under divisibility .
the resulting space is the projective limit space @xmath646 as a topological space it has the same topology ( zariski ) as that of @xmath644 with simply one further closed point ( _ i.e. _ @xmath479 ) added . at the remaining non - archimedean primes
this space fulfills the same topological properties as @xmath644 .
one nice feature of this construction is that the local algebraic structure at @xmath479 is modeled by @xmath636 .
as explained in 2.2.1.1 of @xcite , a @xmath1-set @xmath647 automatically extends , using filtered colimits over the finite subsets @xmath648 , to a pointed endofunctor @xmath649 let @xmath129 be a ( commutative ) semiring and @xmath137 be the associated @xmath0-algebra as in lemma [ sssalg ] .
then the extension @xmath650 is , except for the nuance given by the implementation of the base point in the construction , the same as the endofunctor @xmath651 of [ durov1 ] .
next , we review the definition of the assembly map in the simple case of @xmath1-sets : we refer to @xcite for more details .
let @xmath652 be a @xmath1-set , and @xmath653 its extension to a pointed endofunctor as in .
one derives a map natural in the two pointed sets @xmath47 ( _ cf . _ 2.2.1.2 of @xcite ) @xmath654 which is obtained as the family , indexed by the elements @xmath655 , of maps @xmath656 with @xmath657 similarly , we let @xmath647 be a second @xmath1-set , and @xmath658 its extension to a pointed endofunctor of @xmath43 .
one then constructs a map natural in @xmath20 and @xmath108 as follows @xmath659 where @xmath660 and @xmath661 . by restricting to finite pointed sets @xmath47 , _
i.e. _ to objects of @xmath662 , one obtains a natural transformation of bi - functors and hence a morphism of @xmath1-sets @xmath663 @xmath13 the functor @xmath670 coincides with the endofunctor @xmath671 . for a pointed set @xmath76
, we denote the elements of the set @xmath672 as formal sums @xmath673 , where only finitely many non - zero terms occur in the sum , and the coefficient of the base point @xmath251 is irrelevant , _ i.e. _ the term @xmath674 drops out .
let @xmath675 and @xmath676 , then one has @xmath677 the image of @xmath678 by @xmath679 is described by @xmath680 and one finally obtains @xmath681 by applying one has @xmath682 any element @xmath683 is represented by @xmath684 with @xmath85 and @xmath685 as above
. its image is described , using , by @xmath686 for @xmath32 an object of @xmath35 , an element of @xmath687 is of the form @xmath688 where @xmath689 is a finite set of indices .
let @xmath690 and @xmath691 .
let @xmath692 and @xmath693 .
let @xmath209 be given by @xmath694 if @xmath695 and @xmath696 , @xmath697 . with this choice for @xmath698 ,
one then obtains @xmath699 @xmath15 for @xmath667 , the element of @xmath204 associated to the @xmath258-relation @xmath259 corresponds to @xmath698 where @xmath700 , @xmath701 , @xmath702 @xmath703 , @xmath704 @xmath263 .
thus becomes @xmath705 .
notice that @xmath15 of proposition [ assembly0 ] implies that the assembly map is _ not _ injective for @xmath667 , since by corollary [ pp1 ] one knows that @xmath204 is infinite and countable for any @xmath706 , while @xmath707 remains finite for any choice of @xmath258 .
the constructions of @xcite for monads reviewed in [ durov1 ] can be easily reproduced in the context of @xmath1-sets and yield the @xmath0-algebras : @xmath708 and @xmath709 , where the multiplicative seminorms are _ resp . , _ the usual absolute value on @xmath633 and its restriction to @xmath519 .
this fact suggests the existence of a functor that associates to a monad on @xmath43 an @xmath0-algebra .
we first briefly comment on the use of the category of pointed sets in place of the category of sets as in @xcite .
one has a pair of adjoint functors @xmath710 where @xmath711 is the functor @xmath712 of implementation of a base point .
it is left adjoint to the forgetful functor @xmath713 .
the unit of the adjunction is the natural transformation @xmath714 which maps the set @xmath20 to its copy in the disjoint union @xmath715 .
the counit @xmath716 is the natural transformation which is the identity on @xmath717 while it maps the extra base point to the base point of @xmath20 .
let @xmath522 be a pointed endofunctor of @xmath43 , the composition @xmath718 is then an endofunctor of @xmath19 .
moreover given a natural transformation @xmath719 of endofunctors of @xmath43 , one obtains a natural transformation of endofunctors of @xmath19 , @xmath720 as follows @xmath721 this construction determines a natural correspondence between monads on @xmath43 and monads on @xmath19 .
note in particular that the monad on @xmath19 given by @xmath722 , where @xmath723 is the identity endofunctor is the monad called @xmath7 in @xcite .
moreover , for any semiring @xmath129 , with @xmath670 being the extension of @xmath137 to an endofunctor of @xmath43 as in , one checks that the natural monad structure on @xmath670 corresponds to the monad @xmath724 , _
i.e. _ that @xmath725 .
[ assembly ] let @xmath726 be a pointed algebraic monad on @xmath43 , then the restriction @xmath99 of @xmath726 to @xmath662 defines an @xmath0-algebra with the product @xmath727 defined by the composite of the assembly map and the product @xmath617 of the monad @xmath726 .
one uses the fact stated in remark 2.19 of @xcite that the assembly map makes the identity functor on @xmath1-sets a lax monoidal functor from the monoidal category @xmath728 to the monoidal category @xmath729 .
more explicitly , the composition @xmath727 becomes a morphism of @xmath1-sets @xmath730 .
the compatibility of the assembly map with the associativity shows that the commutative diagram gives the commutativity of the following diagram of maps of @xmath1-sets @xmath731^{m\wedge \id } \ar[rr]^{\id \wedge m } & & \ar[d]^{\mu\circ \alpha_{\sigma,\ca } } \ca\wedge \ca \\
\ca\wedge \ca \ar[rr]_{\mu\circ \alpha_{\sigma,\ca } } & & \ca } \ ] ] the compatibility with the unit is handled in a similar manner .
the assembly map makes the identity functor on @xmath1-sets a lax monoidal functor from the monoidal category @xmath728 to the monoidal category @xmath729 .
thus given an algebraic monad @xmath726 on @xmath43 one can associate to a ( left or right ) algebraic module @xmath490 on @xmath726 in the monoidal category @xmath728 a ( left or right ) module over the @xmath0-algebra @xmath99 of proposition [ assembly ] .
the underlying @xmath1-set is unchanged and the action of @xmath99 is obtained using the assembly map as in proposition [ assembly ] . besides the notion of ( left or right ) algebraic module @xmath490 on @xmath726 in the sense of the monoidal category @xmath728 as discussed in 4.7 of @xcite , a simpler category @xmath733 of modules over a monad is also introduced and used in 3.3.5 of _ op.cit . _ a module @xmath490 on a monad @xmath726 on @xmath43 in this sense is simply a pointed set @xmath490 together with a map of pointed sets @xmath734 fulfilling the two conditions : @xmath735 .
this notion of a module over a monad defined in @xcite implies that modules over @xmath724 correspond exactly to ordinary @xmath129-modules ( _ cf . _
_ or @xcite just before proposition 2 , p. 11 ) . in particular , by taking the monad @xmath736 ( which is denoted by @xmath737 in @xcite )
, one finds that any set @xmath51 is a module in the unique manner provided by the map @xmath738 .
this fact continues to hold if one considers the monad which corresponds to @xmath7 ( _ cf . _
@xcite 4.4 p. 24 ) .
then , one finds that the modules over @xmath7 are just pointed sets . in the framework of @xmath0-algebras ,
an @xmath129-module @xmath490 over a ring @xmath129 gives rise to an @xmath137 module @xmath499 but a general module over @xmath137 is not always of this form ( _ cf .
_ @xcite remark 2.1.5.2 , 4 ) .
this nuance between @xmath129-modules and @xmath137-modules luckily disappears at the homotopy level .
since this is a crucial fact , we shall comment on it here below : we refer to @xcite and @xcite for the full treatment .
let @xmath258 be a _ commutative _ @xmath0-algebra , then one defines the smash product @xmath739 of two @xmath258-modules as the coequalizer ( _ cf .
_ @xcite , definition 2.1.5.3 ) @xmath740 where the two maps are given by the actions of @xmath258 on @xmath51 and @xmath562 .
one obtains in this way the symmetric closed monoidal category of @xmath258-modules and a corresponding notion of @xmath258-algebra . in order to perform homotopy theory
, one passes to the associated categories of simplicial objects .
the fundamental fact ( _ cf .
@xcite , theorem 4.1 ) is that the homotopy category of @xmath614-algebras , obtained using the simplicial version of @xmath614-algebras , is equivalent to the homotopy category of ordinary simplicial rings .
this result continues to hold if one replaces @xmath9 by any commutative ring @xmath129 and simplicial rings by simplicial @xmath129-algebras .
this shows that , in the framework of @xmath0-algebras , nothing is lost by using the more relaxed notion of @xmath137-module in place of the more restricted notion based on monads . in @xcite ( _ cf .
_ 5.1.22 p. 226 ) one
gets the equality @xmath615 which is disappointing for the development of the analogy with the geometric theory over functions fields .
the following result shows the different behavior of the corresponding statement over @xmath0 .
assume that there exists an isomorphism of @xmath0-algebra @xmath741 .
it is then unique since , by proposition [ sssalg1 ] , the @xmath0-algebra @xmath614 has no non - trivial automorphism .
furthermore , by proposition [ assembly0 ] the assembly map @xmath742 is surjective and this fact implies that the composition @xmath743 is also surjective .
let @xmath744 be the natural transformation as in , then the composite @xmath745 is a morphism of @xmath1-sets : @xmath746 .
since @xmath744 is surjective , one then derives that the composite @xmath746 is also surjective , thus coming from a surjective endomorphism ( _ i.e. _ an automorphism ) of the abelian group @xmath9 ( here we refer to the proof of proposition [ sssalg1 ] ) .
these facts would imply that @xmath747 is also injective , hence an isomorphism @xmath748 .
the set @xmath749 is @xmath750 with @xmath24 as the base point .
the maps @xmath751 of act as follows @xmath752 thus @xmath753 is naturally isomorphic to the @xmath1-set which assigns to @xmath32 the laurent polynomials in @xmath258-variables modulo constants @xmath754^{\otimes k}/\z$ ] .
the three maps @xmath751 of determine the following maps @xmath755\to \z[t]$ ] @xmath756 in particular , we see that the map @xmath757 given by the pair of maps @xmath758 is not injective since one can add to @xmath759 any multiple of @xmath760 without affecting @xmath761 and @xmath762 . for the @xmath1-set @xmath614 the map @xmath763 given by the pair of maps @xmath764 is bijective ,
thus the @xmath1-set @xmath765 can not be isomorphic to @xmath614 and this determines a contradiction .
[ homtop]when working at the secondary level of topological spectra , it is a well known fact that the smash product of the eilenberg maclane spectra @xmath766 is not isomorphic to @xmath767 .
in fact , the corresponding homotopy groups are not the same . indeed , for any cofibrant spectrum @xmath20 one has the equality @xmath768 , where @xmath769 denotes the spectrum homology with integral coefficients .
applying this fact to @xmath770 , one interprets @xmath771 as the integral homology of the eilenberg - mac lane spectrum @xmath767 .
this homology is known to be finite in positive degrees but not trivial ( _ cf .
_ theorem 3.5 of @xcite ) .
this argument however , can not be applied directly to provide an alternative proof of proposition [ tensq ] since then one would need to compare the spectra @xmath772 and @xmath773 and this comparison is usually done by using a cofibrant replacement @xmath774 of the @xmath1-set @xmath614 .
a. connes , c. consani , _ characteristic one , entropy and the absolute point _ ,
`` noncommutative geometry , arithmetic , and related topics '' , the twenty - first meeting of the japan - u.s .
mathematics institute , baltimore 2009 , jhup ( 2012 ) , pp 75139 .
a. connes , c. consani , _ the universal thickening of the field of real numbers _ , to appear on advances in the theory of numbers thirteenth conference of the canadian number theory association ( 2014 ) ; arxiv:1202.4377 [ math.nt ] ( 2012 ) . | * abstract * we show that the basic categorical concept of an @xmath0-algebra as derived from the theory of segal s @xmath1-sets provides a unifying description of several constructions attempting to model an algebraic geometry over the absolute point .
it merges , in particular , the approaches using monods , semirings and hyperrings as well as the development by means of monads and generalized rings in arakelov geometry .
the assembly map determines a functorial way to associate an @xmath0-algebra to a monad on pointed sets .
the notion of an @xmath0-algebra is very familiar in algebraic topology where it also provides a suitable groundwork to the definition of topological cyclic homology .
the main contribution of this paper is to point out its relevance and unifying role in arithmetic , in relation with the development of an algebraic geometry over symmetric closed monoidal categories . |
a basic assumption underlying many evolutionary and economic game theoretical models is that individuals are the same except for possible differences in the strategy that they select .
in fact , a population of individuals involved in ecological or social interaction is considered to be heterogeneous .
for example , different individuals may have different fighting abilities or endowments @xcite , occupy different positions in contact networks specifying the peers with whom the game is played @xcite , or have different preferences over the objective outcome of the game .
the last situation is succinctly represented by the battle of the sexes game in which a wife and husband prefer to go to watch opera and football , respectively , whereas their stronger priority is on going out together @xcite ( the battle of the sexes game here is different from the one that models conflicts between males and females concerning parental investment as described in @xcite ) . in behavioral game experiments ,
the heterogeneity of subjects is rather a norm than exceptions ( e.g. , @xcite ) .
for example , some humans are cooperative in the public goods game and others are not ( e.g. , @xcite ) , and some punish non - cooperators more than others do @xcite
. evolution of strategies in such a heterogeneous population is the focus of the present paper .
this question has been examined along several lines . first , in theory of preference , it is assumed that individuals maximize their own idiosyncratic utilities that vary between individuals .
the utility generally deviates from the fitness on which evolutionary pressure operates ( e.g. , @xcite ) .
in fact , experimental evidence shows that individuals tend to imitate behavior of similar others in the context of diffusion of innovations @xcite and health behavior @xcite .
also in the context of economic behavior described as games , individuals may preferentially imitate similar others because similar individuals are expected to be interested in maximizing similar objective functions .
this type of behavior is not considered in previous preference models in which individuals can instantaneously maximize their own payoffs , and selection occurs on the basis of the fitness function common to the entire population .
the model proposed in this study deals with evolutionary dynamics in which individuals in a heterogeneous population mimic successful and similar others .
the similarity here refers to that in the idiosyncratic preference .
second , evolution in heterogeneous populations has been investigated with the use of the evolutionary bimatrix game @xcite .
a payoff bimatrix describes the payoff imparted to the two players in generally asymmetric roles . in its evolutionary dynamics
, a population is divided into two subpopulations , pairs of individuals selected from the different subpopulations play the game , and selection occurs within each subpopulation .
the population then has bipartite structure induced by the fixed role of individuals .
however , the most generic population structure for investigating interplay of evolution via social learning and idiosyncratic preferences would be a well - mixed population without fixed roles of individuals .
third , evolutionary game dynamics on heterogeneous social networks @xcite is related to evolution in heterogeneous populations . in most of the studies on this topic ,
the payoff to an individual per generation is defined as the obtained payoff summed over all the neighboring individuals .
then , cooperation in social dilemma games is enhanced on heterogeneous networks @xcite . in this framework , hubs ( i.e. , those with many neighbors ) and non - hubs are likely to gain different payoffs mainly because of their positions in the contact network . in particular ,
if the payoff of a single game is assumed to be nonnegative , hubs tend to earn more than non - hubs simply because hubs have more neighbors than non - hubs by definition @xcite .
however , as long as the contact network is fixed , a non - hub player will not gain a large payoff by imitating the strategy of a successful hub neighbor .
the number of neighbors serves as the resource of a player .
then , it may be more natural to assume that players imitate successful others with a similar number of neighbors . motivated by these examples ,
i examine evolutionary dynamics in which a player would imitate successful others having similar preferences or inhabiting similar environments .
i divide the players into two subpopulations depending on the subjective perception of the result of the game ; one may like a certain outcome of the game , and another may not like the same outcome .
imitation is assumed to occur within each subpopulation .
however , the interaction occurs as a well - mixed population .
i also assume that all the individuals have the same ability , i.e. , no player is more likely to `` win '' the game than others .
consider a population comprising two subpopulations of players such that the payoff matrix depends on the subpopulation .
the payoff is equivalent to the fitness in the present model .
i call the game the subjective payoff game .
each player , independent of the subpopulation , selects either of the two strategies denoted by @xmath0 and @xmath1 .
the case with a general number of strategies can be analogously formulated . the subjective payoff game and its replicator dynamics described in the following are a special case of the multipopulation game proposed before @xcite ( for slightly different variants , see @xcite ) . the population is infinite , well - mixed , and consists of a fraction @xmath2 ( @xmath3 ) of type @xmath4 players and a fraction @xmath5 of type @xmath6 players .
the subjective payoff matrices that an @xmath4 player and a @xmath6 player perceive as row player are defined by @xmath7 respectively .
it should be noted that the payoff that an @xmath4 player , for example , perceives depends on the opponent s strategy ( i.e. , @xmath0 or @xmath1 ) but not on the opponent s type ( i.e. , @xmath4 or @xmath6 ) .
the use of the two payoff matrices represents different idiosyncrasies in preferences in the two subpopulations .
alternatively , the payoff matrix differs by subpopulations because @xmath4 and @xmath6 players have different tendencies to transform the result of the one - shot game ( i.e. , one of the four consequences composed of a pair of @xmath0 and @xmath1 ) into the fitness .
for example , @xmath4 and @xmath6 players may benefit the most from mutual @xmath0 and mutual @xmath1 , respectively .
the fractions of @xmath4 and @xmath6 players that select strategy @xmath0 are denoted by @xmath8 and @xmath9 , respectively .
the fractions of @xmath4 and @xmath6 players that select strategy @xmath1 are equal to @xmath10 and @xmath11 , respectively .
the payoffs to an @xmath4 player with strategies @xmath0 and @xmath1 are given by @xmath12+b_x\left[p(1-x)+(1-p)(1-y)\right ] \label{eq : pi_xa}\ ] ] and @xmath13+d_x\left[p(1-x)+(1-p)(1-y)\right ] , \label{eq : pi_xb}\ ] ] respectively .
the payoff to a @xmath6 player is defined with @xmath4 replaced by @xmath6 in eqs . and .
i assume that in the evolutionary dynamics , the players can only copy the strategies of peers in the same subpopulation .
this assumption reflects the premise that the payoff in the present model is subjective such that the only comparison that makes sense is that between the players in the same subpopulation .
the replicator dynamics of the subjective payoff game is then defined by @xmath14\notag\\ = & x(1-x)\left\{(a_x - c_x)\left[px+(1-p)y\right ] + ( b_x - d_x)\left[p(1-x)+(1-p)(1-y)\right]\right\ } \label{eq : dx / dt m=2}\end{aligned}\ ] ] and @xmath15 + ( b_y - d_y)\left[p(1-x)+(1-p)(1-y)\right]\right\ } , \label{eq : dy / dt m=2}\end{aligned}\ ] ] where @xmath16 and @xmath17 represent the time derivatives .
if @xmath18 is an internal equilibrium ( i.e. , @xmath19 ) of the replicator dynamics given by eqs . and , @xmath20 + ( b_x - d_x)\left[p(1-x)+(1-p)(1-y)\right]=(a_y - c_y)\left[px+(1-p)y\right ] + ( b_y - d_y)\left[p(1-x)+(1-p)(1-y)\right]=0 $ ] must be satisfied .
however , this is impossible unless a degenerate condition @xmath21 is satisfied .
therefore , for a generic pair of payoff matrices , the replicator dynamics does not have an internal equilibrium .
three remarks are in order .
first , the absence of internal equilibrium implies that the present dynamics does not allow limit cycles @xcite .
second , the present result contrasts with that for a two - subpopulation dynamics in which the perceived payoff matrix depends on the opponent s subpopulation as well as on the focal player s subpopulation . in the latter case , an internal equilibrium or limit cycle can exist @xcite .
third , the present conclusion is different from that for the bimatrix game . in the bimatrix game , each player in subpopulation @xmath4 exclusively interacts with each player in subpopulation @xmath6 .
then , an internal equilibrium can exist , whereas , when it exists , it is either a saddle or a neutrally stable point surrounded by periodic orbits @xcite .
the replicator dynamics without population structure , which is referred to as the ordinary replicator dynamics in the following , is invariant under some transformations of the payoff matrix .
the dynamics given by eqs . and is also invariant under some payoff transformations .
first , trajectories of the ordinary replicator equation are invariant under the addition of a common constant to all the entries of the payoff matrix .
similarly , replacing the payoff matrices given by eq . by @xmath22 where @xmath23 and @xmath24 are arbitrary constants , does not alter the dynamics .
second , trajectories of the ordinary replicator equation are invariant under multiplication of all the entries of the payoff matrix by a common positive constant .
it only changes the time scale . in the present model , replacing eq . by @xmath25 where @xmath26 , does not alter the dynamics .
it should be noted that the multiplicative factor for the two payoff matrices has to be the same for the dynamics to be conserved .
third , in the ordinary replicator equation , adding a common constant to any column of the payoff matrix does not alter a trajectory . in the present model , replacing eq . by @xmath27
does not alter the trajectory for arbitrary @xmath28 , @xmath29 , @xmath30 , and @xmath31 .
this invariance is a generalization of the first invariance .
it is also equivalent to the invariance relationship found for a more general model @xcite .
let us calculate the conditions for the combination of pure strategies in each subpopulation to be evolutionarily stable strategies ( esss ) .
some definitions of ess for multipopulation games exist @xcite , and it seems that consensus on the definition of the ess in the case of multiple subpopulations has not been reached @xcite .
here i adhere to the definition given in @xcite ( also see @xcite ) , which was proposed for general two - subpopulation games in which intra - subpopulation and inter - subpopulation interactions yield a different payoff to a focal player .
i start with stating the definition of the ess by obeying @xcite .
consider a general two - subpopulation matrix game such that there are @xmath32 and @xmath33 pure strategies in subpopulations @xmath4 and @xmath6 , respectively .
a mixed strategy in @xmath4 and @xmath6 is parametrized as @xmath34 and @xmath35 , where @xmath36 ( @xmath37 ) and @xmath38 ( @xmath39 ) are the probabilities that the mixed strategy in @xmath4 and @xmath6 takes the @xmath40th strategy , respectively , @xmath41 , and @xmath42 denotes the transposition .
consider a population of resident players taking strategies @xmath43 and @xmath44 in subpopulations @xmath4 and @xmath6 , respectively .
i assume that the payoff to a player adopting mixed strategy @xmath45 in subpopulation @xmath4 embedded in this resident population is given by @xmath46 similarly , assume that the payoff to mixed strategy @xmath47 in subpopulation @xmath6 embedded in the same resident population is given by @xmath48 strategy @xmath49 is ess if for any @xmath50 , @xmath51 when the equality holds in eq .
, it is also required that an additional condition given by @xmath52 is satisfied .
in the case of the subjective payoff game , i obtain @xmath53 , @xmath54 @xmath55 , @xmath56 , @xmath57 , and @xmath58
. therefore , eqs . and
are reduced to @xmath59\notag\\ & \cdot \left[p\begin{pmatrix}x\\ 1-x\end{pmatrix } + ( 1-p)\begin{pmatrix}y\\ 1-y\end{pmatrix}\right]\ge 0 \label{eq : def ess 1 subjective}\end{aligned}\ ] ] and @xmath59\notag\\ & \cdot \left[p\begin{pmatrix}x^{\prime}\\ 1-x^{\prime}\end{pmatrix } + ( 1-p)\begin{pmatrix}y^{\prime}\\ 1-y^{\prime}\end{pmatrix}\right ] > 0 \label{eq : def ess 2 subjective}\end{aligned}\ ] ] respectively . in this section ,
let us identify the pure strategy esss of the subjective payoff game .
first , suppose that the population in which all players in both subpopulations adopt strategy @xmath0 is evolutionarily stable . by substituting @xmath60 in eq .
, i obtain @xmath61 because eq .
must hold true for @xmath62 and @xmath63 , a necessary condition reads @xmath64 . if @xmath64 is satisfied with equality , eq .
for the same @xmath62 and @xmath63 , i.e. , @xmath65 > 0 \label{eq : ess pure a 3}\ ] ] must be satisfied . by substituting @xmath66 in eq .
, i obtain @xmath67 .
the necessary conditions obtained so far are summarized as @xmath68 these conditions are the same as the ess conditions for the structureless population . by considering the mutant parametrized by @xmath69 and @xmath70 , i similarly obtain the necessary conditions for subpopulation @xmath6 as @xmath71 on the other hand , if eqs . and are satisfied , eqs . and are satisfied for any @xmath72 . therefore , eqs . and provide the necessary and sufficient conditions for strategy @xmath0 to be evolutionarily stable . in conclusion , @xmath0 is evolutionarily stable for the entire population if @xmath0 is evolutionarily stable in each subpopulation in the ordinary sense .
similarly , @xmath1 is an ess of the subjective payoff game if @xmath1 is evolutionarily stable in each subpopulation .
next , assume that the population in which all the players in subpopulations @xmath4 and @xmath6 adopt @xmath0 and @xmath1 , respectively , is evolutionarily stable . by substituting @xmath73 and @xmath74 in eq .
, i obtain @xmath75 + y^{\prime } \left[(a_y - c_y)p + ( b_y - d_y)(1-p)\right]\ge 0 .
\label{eq : ess pure lopsided 1}\ ] ] because eq . must hold true for @xmath62 and @xmath76 , a necessary condition reads @xmath77 if eq .
is satisfied with equality , eq .
for the same @xmath62 and @xmath76 , i.e. , @xmath78 > 0 , \label{eq : ess pure lopsided 4}\ ] ] must be satisfied . by substituting @xmath79 in eq .
, i obtain @xmath80 similarly , by considering the mutant parametrized by @xmath69 and @xmath70 , i obtain @xmath81 when eq . is satisfied with equality , eq . and @xmath82 lead to @xmath83 the population given by @xmath84 is an ess if eqs . and are satisfied , eq
. holds true when eq . is satisfied with equality , and eq
. holds true when eq .
is satisfied with equality .
it should be noted that the conditions given by eqs . and depend on @xmath2 . in this section ,
i show that the replicator equation of the subjective payoff game can not be mapped to the replicator equation of a bimatrix game .
it should be noted that the following arguments can be readily generalized to the case of an arbitrary number of strategies . in the bimatrix game in the well - mixed population
, we consider all possible pairs of a player in subpopulation @xmath4 and a player in subpopulation @xmath6 .
the two selected players are involved in a two person game , which is generally asymmetric .
the payoff bimatrix is given by @xmath85 where the first and second elements in each entry of the bimatrix represent the payoffs imparted to an @xmath4 player and @xmath6 player , respectively .
the row and column players correspond to subpopulations @xmath4 and @xmath6 , respectively .
the payoff to an @xmath4 player with strategies @xmath0 and @xmath1 is equal to @xmath86 and @xmath87 , respectively .
then , the replicator equation for subpopulation @xmath4 is given by @xmath88 .
\label{eq : dx / dt bimatrix game}\ ] ] similarly , @xmath89 .
\label{eq : dy / dt bimatrix game}\ ] ] if the dynamics given by eqs . and is equivalent to that for the subjective payoff game ( eqs . and ) , the comparison of eqs . and
yields @xmath90 because the right - hand side of eq . must be independent of @xmath8 except for the multiplication factor @xmath91 . because @xmath92 implies a structureless population
, @xmath93 holds true . under this condition , eq .
is reduced to @xmath94 .
then , @xmath95 and @xmath96 must hold true .
similarly , @xmath97 and @xmath98 must hold true . except for this degenerate case ,
the two dynamics are not mapped from one to the other .
consider the snowdrift game , also called the chicken game or the hawk - dove game , which represents a social dilemma situation @xcite .
the snowdrift game in two subpopulations in which the payoff matrix for a player depends on the opponent s subpopulation as well as the focal player s subpopulation is analyzed in @xcite . in the case without population structure ,
a standard payoff matrix for the snowdrift game is given by @xmath99 where @xmath100 .
strategies @xmath0 and @xmath1 correspond to cooperation and defection , respectively . if the opponent cooperates , it is better to defect . otherwise , it is better to cooperate .
the mixed population with a fraction of @xmath0 players given by @xmath101 is the unique ess .
consider the case in which the @xmath102 value depends on the subpopulation .
denote by @xmath103 and @xmath104 the subpopulation - dependent @xmath102 value such that @xmath105 and @xmath106 without loss of generality , i assume that @xmath107 .
equations and read , respectively , @xmath108 \right\},\\ \dot{y } = & y(1-y)\left\{\left(\beta_y-1\right)-\left(\beta_y-\frac{1}{2}\right)\left[px+(1-p)y\right ] \right\}.\end{aligned}\ ] ] therefore , i obtain @xmath109 equation implies that the replicator dynamics has the unique stable equilibrium whose location depends on the @xmath2 value . if @xmath110 , the stable equilibrium is located at @xmath111 an example of this case is shown in fig .
[ fig : snowdrift](a ) .
this equilibrium is an ess , which is shown in appendix with the use of the ess criterion established in @xcite . if @xmath112 , the stable equilibrium is located at @xmath113 an example of this case is shown in fig .
[ fig : snowdrift](b ) . in this case , the equilibrium is an ess because eqs .
, , , and are satisfied . finally ,
if @xmath114 , the stable equilibrium is located at @xmath115 an example of this case is shown in fig .
[ fig : snowdrift](c ) .
this equilibrium is also an ess ( see appendix for the proof ) . in all three cases , @xmath4 ,
i.e. , the subpopulation with the larger @xmath102 value , has a larger fraction of @xmath0 players than @xmath6 does .
in particular , at least one subpopulation is monomorphic for any @xmath2 ; @xmath0 monopolizes subpopulation @xmath4 , or @xmath1 monopolizes subpopulation @xmath6 .
even a slight difference in the payoff matrix ( i.e. , @xmath102 value ) in the two subpopulations yields polarization of the strategies .
similar polarization also occurs in the bimatrix snowdrift game .
however , the internal equilibrium exists but is a saddle in the case of the bimatrix game @xcite such that the mechanism is different from that for the subjective payoff snowdrift game .
consider a coordination game in which the players in different subpopulations have different preference of one strategy over the other . specifically , let us set the payoff matrix for subpopulations @xmath4 and @xmath6 to @xmath116 and @xmath117 respectively , where @xmath118 .
players in @xmath4 and @xmath6 prefer strategies @xmath0 and @xmath1 , respectively .
equations and can be regarded as a payoff bimatrix of the battle of the sexes game @xcite . however , in the bimatrix game , the population is composed of two subpopulations corresponding to the roles in the game . in contrast , the players in the present game do not have roles and interact in a well - mixed population
. given eqs . and ,
eqs . and
read @xmath119 , \label{eq : dx / dt coordination game}\\ \dot{y } = & y(1-y)\left[p(2x-1)+(1-p)(2y-1)+\alpha_y \right ] .
\label{eq : dy / dt coordination game}\end{aligned}\ ] ] near @xmath60 , @xmath120 is satisfied .
therefore , @xmath121 is a stable equilibrium of the replicator equations of the subjective payoff coordination game . because eqs . and are satisfied with eqs . and , @xmath121 is an ess .
similarly , because @xmath122 near @xmath123 , @xmath124 is a stable equilibrium . because eqs . and yield @xmath125 and @xmath126 , @xmath124 is another ess .
these results are consistent with those for the coordination game without population structure ; the two strategies are bistable . the subjective payoff version of the coordination game yields two phenomena that are absent in the same game without population structure .
first , the polarized configuration in which all the @xmath4 players adopt @xmath0 and all the @xmath6 players adopt @xmath1 is an stable equilibrium if @xmath127 when eq . is satisfied , eqs . and are satisfied with inequality such that this population is a pure - strategy ess .
equation is satisfied if @xmath2 is close to @xmath128 or if @xmath129 or @xmath130 , i.e. , the asymmetry in the liking of the two actions , is large . as an example , the attractive basins of the three equilibria for @xmath131 , @xmath132 , and @xmath133 are shown in fig .
[ fig : coordination game](a ) .
the final configuration of the population depends on the initial condition .
it should be noted that stable coexistence of the opposite pure strategies does not occur in the bimatrix coordination game ( i.e. , battle of the sexes game ) .
second , the fraction of players employing a strategy in a subpopulation can non - monotonically change in time . some non - monotonic trajectories starting from different initial conditions are shown in fig .
[ fig : coordination game](b ) with @xmath134 , @xmath135 , and @xmath133 .
@xmath136 holds true to the right of the thick solid line in fig .
[ fig : coordination game](b ) .
if the initial condition is located slightly right to this line , @xmath9 first decreases because @xmath137 holds true to the left of the thick dotted line .
if @xmath138 is not large , the trajectory eventually crosses the thick solid line ( @xmath139 ) such that @xmath8 starts to decrease .
if the initial @xmath16 value is large enough , the trajectory eventually crosses the thick dotted line ( @xmath140 ) such that @xmath9 starts to increase . in both cases ,
the trajectory shows non - monotonic behavior . such a non - monotonic behavior does not occur in the coordination game without population structure . in extensions of the ising model @xcite and the voter model @xcite , which are non - game population dynamics , idiosyncratic preferences of individuals lead to coexistence of different states , where states are equivalent to strategies in games .
the present results are consistent with these results in that idiosyncratic preferences let multiple states coexist when unanimity necessarily occurs in the absence of idiosyncrasy . in this section ,
i examine the possibility of cooperation in the iterated prisoner s dilemma ( ipd ) in which the unconditional cooperation ( c ) and unconditional defection ( d ) are strategies @xmath0 and @xmath1 , respectively .
i do not assume error in action implementation and do assume that a next round of the game between the same pair of players occurs with probability @xmath141 ( @xmath142 ) .
the following results also hold true if c is replaced by the tit - for - tat ( tft ) or the so - called grim strategy .
tft starts with cooperation and selects the previous action ( i.e. , cooperate or defect ) selected by the opponent .
grim strategy starts with cooperation and switches to permanent defection once the opponent ever defects .
the invariance of the following results holds true because the payoff matrix for the ipd , i.e. , eq .
below , does not change when c is replaced by tft or grim .
consider a standard payoff matrix for the single - shot prisoner s dilemma given by @xmath143 where @xmath144 and @xmath145 are the benefit and the cost of cooperation and satisfy @xmath146 . the expected payoff matrix for the ipd in the structureless population is given by @xmath147 if @xmath148 the prisoner s dilemma is effectively transformed into a coordination game such that mutual cooperation by c and mutual defection by d are bistable @xcite .
let us consider the situation in which two subpopulations possess different discount factors @xmath149 and @xmath150 .
in other words , assume @xmath151 and @xmath152 because the duration of ipd is the same for the two players , i interpret that @xmath4 players put more emphasis on long - term benefits than @xmath6 players .
specifically , i assume @xmath153 given eq . , eq .
implies @xmath137 for all @xmath154 .
therefore , d eventually occupies subpopulation @xmath6 .
i examine the possibility that cooperation occurs in subpopulation @xmath4 . on the line @xmath74 , eq .
is reduced to @xmath155 .
\label{eq : dx / dt ipd at y=0}\ ] ] therefore , the population in which d dominates in both subpopulations , i.e. , @xmath124 , is always a stable equilibrium of the replicator equations of the subjective payoff ipd game .
it is an ess because eqs . and imply @xmath125 and @xmath126 , respectively .
equation implies that the combination of c in subpopulation @xmath4 and d in subpopulation @xmath6 is a stable equilibrium if @xmath156 because eq .
and eq . are satisfied with inequality when eq
. holds true , this population is an ess . a large @xmath2 ( i.e. , large fraction of @xmath4 players ) and a large benefit - to - cost ratio @xmath157 lessen @xmath158 such that cooperation would occur .
the threshold discount factors @xmath159 and @xmath158 are compared in fig .
[ fig : ipd w_c ] for some @xmath2 values .
the figure indicates that for a wide range of @xmath2 , the condition for cooperation in the subjective payoff case is not very severe relative to the case without population structure .
in particular , both @xmath159 and @xmath158 tend to unity in the limit @xmath160 . as @xmath161
, it follows that @xmath162 and both @xmath159 and @xmath158 converge to 0 .
when @xmath163 , condition coincides with the condition for the risk dominance of c over d in the structureless population , i.e. , @xmath164 .
i proposed the so - called subjective payoff game and its replicator dynamics .
the model is mathematically a special case of the previously analyzed model with two subpopulations @xcite .
however , the present model is motivated by the possibility that different players may perceive the same result of the game to transform it to the fitness in different manners .
the model shows polarization in the snowdrift and coordination games , non - monotonic time courses in the coordination game , and a wide margin of cooperation in the ipd .
extension of the present model to the case of more than two strategies and more than two subpopulations is straightforward . generalizing the present results for such extended models warrants future work .
the replicator dynamics of the subjective payoff game is different from that of the bimatrix game ( section [ sub : not bimatrix ] ) .
in addition , the subjective payoff game can not be mapped to a model with strategy - dependent interaction rates , which does not have multiple subpopulations within each of which imitation occurs @xcite .
the subjective payoff game is also different from those in which interaction is confined in single subpopulations , such as group selection models @xcite , island model @xcite , and evolutionary set theory @xcite .
the present model is distinct from previous models of evolution of preference ( e.g. , @xcite ) and the so - called subjective game @xcite .
both in these and present models , the preference , or the subjective payoff , is assumed to be consistent within each individual @xcite . in these previous models , the utility that a player maximizes and the fitness on which the selection pressure operates are different .
a player in such a model is rational enough to be able to personally maximize the player s idiosyncratic utility . in the present model , as in standard evolutionary models , a player is subjected to bounded rationality and tends to imitate successful others ( i.e. , social learning ) .
the difference from standard evolutionary models is that , in the present model , each player limits the set of possible parents from whom the strategy is copied to those with the same idiosyncratic payoff . in this way
, the player can pursue both maximization of fitness via social learning and consistency with the player s idiosyncratic preference .
the subjective payoff game does not allow internal equilibria regardless of the stability .
this result has implications in games in which internal equilibria play an important role in structureless populations . in the snowdrift game , a mixture of the two strategies is stable under ordinary replicator dynamics .
in contrast , in the subjective payoff game , a slight difference in the payoff matrices perceived by the two subpopulations leads to polarization such that the two subpopulations tend to select the opposite strategies . the rock scissors paper game comprises three strategies that cyclically dominate one another .
it is straightforward to show that there is no internal equilibrium in the subjective payoff game with a general number of strategies .
therefore , the subjective payoff variant of the rock
scissors paper game lacks the internal equilibrium of any kind and limit cycles .
such a game behaves very differently from the same game played in the structureless population @xcite , bimatrix population @xcite , and two subpopulations with different social learning rates @xcite ; these models allow a unique internal equilibrium .
it may be interesting to examine the rock
scissors paper game under the current framework .
i assumed that players imitate others in the same subpopulation .
in fact , there may be competition of update rules between such players and those that imitate from the entire population .
nevertheless , at least near pure stable equilibria , the population is considered to be resistant against invasion by mutants that imitate from the entire population . to explain why ,
let us suppose that @xmath4 and @xmath6 players select @xmath0 and @xmath1 in the equilibrium .
a mutant that imitates from the entire population and attempts to invade subpopulation @xmath4 would sometimes select @xmath1 because @xmath6 players select @xmath1 . because @xmath0 , not @xmath1 , is the best response in this population ,
such a mutant is considered not able to invade the subpopulation of resident players .
therefore , the imitation rule considered in the present study is considered to have evolutionary stability , at least in this case .
in this section , i show that the equilibria given by eqs . and are esss of the subjective payoff snowdrift game . to this end
, i use a matrix criterion for the ess @xcite accommodated to the case of the two - strategy game .
assume that a population given by @xmath165 satisfies @xmath166\notag\\ \le & \left(x^*\ ; 1-x^*\right)\left[r_{xx}\begin{pmatrix } x^*\\ 1-x^*\end{pmatrix } + r_{xy}\begin{pmatrix } y^*\\ 1-y^*\end{pmatrix}\right ] \label{eq : taylor 1}\end{aligned}\ ] ] and @xmath167\notag\\ \le & \left(y^*\ ; 1-y^*\right)\left[r_{yx}\begin{pmatrix}x^*\\ 1-x^*\end{pmatrix } + r_{yy}\begin{pmatrix}y^*\\ 1-y^*\end{pmatrix}\right ] \label{eq : taylor 2}\end{aligned}\ ] ] for @xmath168 @xmath169 , @xmath170 , @xmath171 , and @xmath172 .
i also assume that eqs . and are satisfied with equality for @xmath173 if @xmath174 and @xmath175 , for @xmath176 if @xmath177 and @xmath175 , for @xmath178 if @xmath174 and @xmath179 , and for @xmath180 if @xmath177 and @xmath179 .
the criterion dictates that @xmath165 is an ess if and only if @xmath181 here , @xmath182 ( @xmath183 ) if and only if the payoff of a pure @xmath0 player in subpopulation @xmath4 ( @xmath6 ) and that of a pure @xmath1 player in subpopulation @xmath4 ( @xmath6 ) are the same in the equilibrium of interest . under the snowdrift game ,
the assumptions for @xmath165 are satisfied when @xmath165 is given by eq . or .
substitution of eqs .
, , , , , and in eq . yields @xmath184\cdot \left[x_1\left(\frac{1}{2}-\beta_x\right ) + y_1\left(\frac{1}{2}-\beta_y\right)\right ] < 0 .
\label{eq : matrix criterion sd}\ ] ] for the equilibrium given by eq . to be an ess , eq .
must be satisfied for @xmath185 and @xmath183 .
for the equilibrium given by eq . to be an ess , eq
. must be satisfied for @xmath182 and @xmath186 .
in fact , eq . is satisfied in both cases . therefore , the two equilibria are esss .
i thank julin garca and arne traulsen for valuable discussions and for reading of the manuscript .
i also acknowledge the support provided through grants - in - aid for scientific research ( no .
23681033 ) from mext , japan , the nakajima foundation , and crest , jst . and @xmath189 .
the filled circles indicate the esss .
( a ) @xmath190 .
( b ) @xmath191 .
( c ) @xmath131.,title="fig:",width=226 ] and @xmath189 .
the filled circles indicate the esss .
( a ) @xmath190 .
( b ) @xmath191 .
( c ) @xmath131.,title="fig:",width=226 ] and @xmath189 .
the filled circles indicate the esss .
( a ) @xmath190 .
( b ) @xmath191 .
( c ) @xmath131.,title="fig:",width=226 ] , @xmath192 , and @xmath193 , respectively .
the thick solid lines represent @xmath139 , i.e. , @xmath194 .
the thick dotted lines represent @xmath140 , i.e. , @xmath195 .
( a ) @xmath131 , @xmath132 , and @xmath133 .
( b ) @xmath134 , @xmath135 , and @xmath133 .
the thin solid curves in ( b ) represent trajectories converging to @xmath124 or @xmath192 . for calculating the attractive basins and individual trajectories , the euler scheme with @xmath196 was used.,title="fig:",width=302 ] , @xmath192 , and @xmath193 , respectively .
the thick solid lines represent @xmath139 , i.e. , @xmath194 .
the thick dotted lines represent @xmath140 , i.e. , @xmath195 .
( a ) @xmath131 , @xmath132 , and @xmath133 .
( b ) @xmath134 , @xmath135 , and @xmath133 .
the thin solid curves in ( b ) represent trajectories converging to @xmath124 or @xmath192 . for calculating the attractive basins and individual trajectories , the euler scheme with @xmath196 was used.,title="fig:",width=302 ] | in social situations with which evolutionary game is concerned , individuals are considered to be heterogeneous in various aspects . in particular , they may differently perceive the same outcome of the game owing to heterogeneity in idiosyncratic preferences , fighting abilities , and positions in a social network . in such a population , an individual may imitate successful and similar others , where similarity refers to that in the idiosyncratic fitness function .
i propose an evolutionary game model with two subpopulations on the basis of multipopulation replicator dynamics to describe such a situation . in the proposed model ,
pairs of players are involved in a two - person game as a well - mixed population , and imitation occurs within subpopulations in each of which players have the same payoff matrix .
it is shown that the model does not allow any internal equilibrium such that the dynamics differs from that of other related models such as the bimatrix game .
in particular , even a slight difference in the payoff matrix in the two subpopulations can make the opposite strategies to be stably selected in the two subpopulations in the snowdrift and coordination games .
keywords : evolutionary game ; replicator dynamics ; homophily ; cooperation |
we consider the harmonic map heat flow which is defined as the negative gradient flow of the dirichlet energy of maps between manifolds . indeed ,
if @xmath13 is a map from @xmath14 to a compact riemannian manifold @xmath15 , with second fundamental form @xmath16 , then @xmath13 solves @xmath17 we assume that the target manifold is the @xmath1-sphere @xmath18 .
then , becomes @xmath19 we will study the problem under an additional assumption of 1-corotational symmetry , namely that a solution of takes the form @xmath20 under this ansatz , the problem reduces to the one dimensional semilinear heat equation @xmath21 where @xmath22 $ ] .
the set of solutions to is invariant by the scaling symmetry @xmath23 the energy associated to is given by @xmath24(t ) = \int_0^{+\infty } \left(|\partial_r u|^2 + \frac{(d-1)}{r^2}\sin^2(u)\right)r^{d-1}dr,\ ] ] which satisfies @xmath25 = \lambda^{d-2}{\mathcal{e}}[u].\ ] ] the criticality of the problem is reflected by the fact that the energy is left invariant by the scaling property when @xmath26 , hence , the case @xmath27 corresponds to the energy supercritical case .
the problem is locally wellposed for data which are close in @xmath28 to a uniformly continuous map ( see koch and lamm @xcite ) or in @xmath29 by wang @xcite . actually , eells and sampson @xcite introduced the harmonic map heat flow as a process to deform any smooth map @xmath30 into a harmonic map via .
they also proved that the solution exists globally if the sectional curvature of the target manifold is negative .
there exist other assumptions for the global existence such as the image of the initial data @xmath31 is contained in a ball of radius @xmath32 , where @xmath33 is an upper bound on the sectional curvature of the target manifold @xmath34 ( see jost @xcite and lin - wang @xcite ) . without these assumptions ,
the solution @xmath35 may develop singularities in some finite time ( see for examples , coron and ghidaglia @xcite , chen and ding @xcite for @xmath27 , chang , ding and yei @xcite for @xmath26 ) . in this case , we say that @xmath35 blows up in a finite time @xmath36 in the sense that @xmath37 here we call @xmath38 the blowup time of @xmath39 . the blowup has been divided by struwe @xcite into two types : @xmath40 in @xcite , struwe shows that the type i singularities are asymptotically self - similar , that is their profile is given by a smooth shrinking function @xmath41 where @xmath42 solves the equation @xmath43 thus , the study of type i blowup reduces to the study of nonconstant solutions of equation .
+ when @xmath44 , by using a shooting method , fan @xcite proved that there exists an infinite sequence of globally regular solutions @xmath45 of which are called `` shrinkers ( corresponding to the existence of type i blowup solutions of ) , where the integer index @xmath46 denotes the number of intersections of the function @xmath45 with @xmath47 . more detailed quantitative properties of such solutions were studied by biernat and bizo @xcite , where the authors conjectured that @xmath48 is linear stable and provide numerical evidences supporting that @xmath48 corresponds to a generic profile of type i blow - up .
very recently , biernat , donninger and schrkhuber @xcite proved the existence of a stable self - similar blowup solution for @xmath49 . since is not time reversible , there exist another family of self - similar solutions that are called ' ' expanders `` which were introduced by germain and rupflin in @xcite .
these ' ' expanders " have been lately proved to be nonlinearly stable by germain , ghoul and miura @xcite . up to our knowledge ,
the question on the existence of type ii blowup solutions for remains open for @xmath44 .
+ when @xmath3 , bizo and wasserman @xcite proved that equation has no self - similar shrinking solutions . according to struwe @xcite
, this result implies that in dimensions @xmath3 , all singularities for equation must be of type ii ( see also biernat @xcite for a recent analysis of such singularities ) .
recently , biernat and seki @xcite , via the matched asymptotic method developed by herero and velzquez @xcite , construct for equation a countable family of type ii blowup solutions , each characterized by a different blowup rate @xmath50 where @xmath51 such that @xmath52 and @xmath53 is given by @xmath54 \quad \text{for}\;\ ; d \geq 7,\ ] ] where @xmath55 .
the blowup rate is in fact driven by the asymptotic behavior of a stationary solution of , say @xmath7 , which is the unique ( up to scaling ) solution of the equation @xmath56 and admits the behavior for @xmath57 large , @xmath58 ( see appendix in @xcite for a proof of the existence of @xmath7 ) . note that the case @xmath59 only happens in dimension @xmath60 . in this case , biernat @xcite used the method of @xcite and formally derived the blowup rate @xmath61 he also provided numerical evidences supporting that the case @xmath10 in or corresponds to a generic blowup solution . + in the energy critical case , i.e. @xmath26 , van de berg , hulshof and king @xcite , through a formal analysis based on the matched asymptotic technique of herrero and velzquez @xcite , predicted that there are type ii blowup solutions to of the form @xmath6 where @xmath62 is the unique ( up to scaling ) solution of , and the blowup speed governed by the quantized rates @xmath63 this result was later confirmed by raphal and schweyer @xcite .
note that the case @xmath10 was treated in @xcite and corresponds to a stable blowup .
in particular , the authors in @xcite , @xcite adapted the strategy developed by merle , raphal and rodnianski @xcite , @xcite for the study of wave and schrdinger maps to construct for equation type ii blowup solutions .
their method is based on two main steps : * construction of an adapted approximate blowup profile by solving elliptic equations . from the computation of the tails of the blowup profile , they are able to formally derive the blowup speed
. * control of the error by using an energy method involving `` lyapunov '' functionals adapted to the linearized flow around the ground state .
this energy method does nt involve neither spectral estimates nor maximum principles . in this work , by considering @xmath3 , we ask whether we can carry out the analysis of @xcite for the energy critical case @xmath64 , to the construction of blowup solutions for equation in the case @xmath3 .
it happens that the asymptotic behavior is perfectly suitable to replace the explicit profile for an implementation of the strategy of @xcite .
the following theorem is the main result of this paper .
[ theo:1 ] let @xmath3 and @xmath65 be defined as in , we fix an integer @xmath66 and an arbitrary sobolev exponent @xmath67 then there exists a smooth corotational radially symmetric initial data @xmath31 such that the corresponding solution to is of the form : @xmath68 where @xmath69 and @xmath70.\ ] ] moreover , the case @xmath10 corresponds to a stable blowup regime .
since @xmath71 for @xmath72 and @xmath73 for @xmath74 , the condition @xmath52 means that @xmath75 for @xmath60 and @xmath76 for @xmath74 . note that the condition @xmath52 allows to avoid the presence of logarithmic corrections in the construction of the approximate profile .
in other words , the case @xmath59 ( equivalent to @xmath10 and @xmath60 ) would involve an additional logarithmic gain related to the growth of the approximate profile at infinity , which turns out to be essential for the derivation of the speed .
although our analysis could be naturally extended to this case ( @xmath10 and @xmath60 ) with some complicated computation , we hope to treat this case in a separate work .
the quantization of the blowup rate is the same as the one obtained in @xcite .
note that the authors of @xcite only claim the existence result of a type ii blowup solution with the rate and say nothing about the dynamical description of the solution . on the contrary
, our result shows that the constructed solution blows up in finite time by concentration of a stationary state in the supercritical regime .
moreover , our constructed solution is in fact @xmath9 codimension stable in the sense that we will precise shortly .
fix @xmath51 such that @xmath52 , an integer @xmath77 and @xmath78 , our initial data that we are considering has the following form @xmath79 where @xmath80 is a deformation of the ground state @xmath7 .
in addition , @xmath81 corresponds to unstable directions of the flow in the topology @xmath82 in a certain neighborhood of the ground state @xmath7 .
we will prove that for all @xmath83 ( for some @xmath84 ) small enough , for all @xmath85 small enough , under a specific choice of the unstable directions @xmath86 the solution of with the data verifies the conclusion of theorem [ theo:1 ] .
this implies that our constructed solution is @xmath9 codimension stable .
in particular , the case @xmath10 corresponds to a stable type ii blowup regime , which is in agreement with numerical evidences given in @xcite .
it is worth to remark that the harmonic heat flow shares many features with the semilinear heat equation @xmath87 a remarkable fact that two important critical exponents appear when considering the dynamics of : @xmath88 correspond to the cases @xmath26 and @xmath60 in the study of equation respectively . when @xmath89 , giga and kohn @xcite , giga , matsui and sasayama @xcite showed that all blowup solutions are of type i. here the type i blowup means that @xmath90 otherwise we say the blowup solution is of type ii .
when @xmath91 , filippas , herrero and velzquez @xcite formally constructed for type ii blowup solutions in dimensions @xmath44 , however , they could not do the same in dimensions @xmath3 .
this formal result is partly confirmed by schweyer @xcite in dimension @xmath92 .
interestingly , collot , merle and raphal @xcite show that type ii blowup is ruled out in dimension @xmath3 near the solitary wave . when @xmath93 , matano and merle @xcite ( see also mizoguchi @xcite ) proved that only type i blowup occurs in the radial setting .
when @xmath94 , herrero and velzquez formally derived in @xcite the existence of type ii blowup solutions with the quantized rates : @xmath95 the formal result was clarified in @xcite , @xcite and @xcite .
the collection of these works yields a complete classification of the type ii blowup scenario for the radially symmetric energy supercritical case . in comparison to the case of the semilinear heat equation , it might be possibe to prove that all blowup solutions to equation are of type i in dimension @xmath96 .
however , due to the lack of monotonicity of the nonlinear term , the analysis of the harmonic heat flow is much more difficult than the case of the semilinear heat equation treated in @xcite .
let us briefly explain the main steps of the proof of theorem [ theo:1 ] , which follows the method of @xcite treated for the critical case @xmath26 .
we would like to mention that this kind of method has been successfully applied for various nonlinear evolution equations .
in particular in the dispersive setting for the nonlinear schrdinger equation both in the mass critical @xcite and mass supercritical @xcite cases ; the mass critical gkdv equation @xcite ; the energy critical @xcite , @xcite and supercritical @xcite wave equation ; the two dimensional critical geometric equations : the wave maps @xcite , the schrdinger maps @xcite and the harmonic heat flow @xcite ; the semilinear heat equation in the energy critical @xcite and supercritical @xcite cases ; and the two dimensional keller - segel model @xcite , @xcite . in all these works ,
the method is based on two arguments : * reduction of an infinite dimensional problem to a finite dimensional one , through the derivation of suitable lyapunov functionals and the robust energy method as mentioned in the two step procedure above . *
the control of the finite dimensional problem thanks to a topological argument based on index theory .
note that this kind of topological arguments has benn proved to be successful also for the construction of type i blowup solutions for the semilinear heat equation in @xcite , @xcite , @xcite ( see also @xcite for the case of logarithmic perturbations , @xcite , @xcite and @xcite for the exponential source , @xcite for the complex - valued case ) , the ginzburg - landau equation in @xcite ( see also @xcite for an earlier work ) , a non - variational parabolic system in @xcite and the semilinear wave equation in @xcite .
+ for the reader s convenience and for a better explanation , let s first introduce notations used throughout this paper .
+ - * notation .
* for each @xmath3 , we define @xmath97 where @xmath98 stands for the integer part of @xmath99 which is defined by @xmath100 . note that @xmath101 . indeed , if @xmath102 , then there is @xmath103 such that @xmath104 .
this only happens when @xmath71 or @xmath105 because @xmath106 $ ] .
the case @xmath71 gives @xmath60 and @xmath107 .
the case @xmath105 gives @xmath108 .
given a large integer @xmath109 , we set @xmath110 given @xmath111 and @xmath112 , we define @xmath113 and denote by @xmath114 let @xmath115 be a positive nonincreasing cutoff function with @xmath116 $ ] and @xmath117 on @xmath118 $ ] .
for all @xmath119 , we define @xmath120 we also introduce the differential operator @xmath121 and the schrdinger operator @xmath122 - * strategy of the proof . * we now resume the main ideas of the proof of theorem [ theo:1 ] , which follows the route map in @xcite and @xcite .
+ @xmath123 _ renormalized flow and iterated resonances .
_ following the scaling invariance of , let us make the change of variables @xmath124 which leads to the following renormalized flow : @xmath125 assuming that the leading part of the solution @xmath126 is given by the ground state profile @xmath7 admitting the asymptotic behavior , hence the remaining part is governed by the schrdinger operator @xmath127 defined by .
the linear operator @xmath127 admits the factorization ( see lemma [ lemm : factorl ] below ) @xmath128 which directly implies @xmath129 where from a direct computation , @xmath130 more generally , we can compute the kernel of the powers of @xmath127 through the iterative scheme @xmath131 which displays a non trivial tail at infinity ( see lemma [ lemm : genlk ] below ) , @xmath132 @xmath133 _ tail computation . _
following the approach in @xcite , we look for a slowly modulated approximate solution to of the form @xmath134 where @xmath135 with a priori bounds @xmath136 so that @xmath137 is in some sense homogeneous of degree @xmath138 in @xmath139 , and behaves better than @xmath140 at infinity .
the construction of @xmath137 with the above a priori bounds is possible for a specific choice of the universal dynamical system which drives the modes @xmath141 .
this procedure is called the _
tail computation_. let us illustrate the procedure of the _ tail computation_. we plug the decomposition into and choose the law for @xmath141 which cancels the leading order terms at infinity .
+ - at the order @xmath142 : it is not possible to adjust the law of @xmath139 for the first term , then @xmath143 , hence after an integration in time , @xmath144 and there is no blowup . ] and obtain from , @xmath145 - at the order @xmath146 : we obtain @xmath147 where @xmath148 corresponds to the interaction of the nonlinear terms . note from and
, we have @xmath149 and thus , @xmath150t_1.\ ] ] hence , we cancel the highest order growth for @xmath151 large through the choice @xmath152 we then solve for @xmath153 and check the improved decay @xmath154 - at the order @xmath155 : we get an elliptic equation of the following form @xmath156 from and , we have @xmath157t_{k},\ ] ] which leads to the choice @xmath158 to cancel at infinity the leading order growth .
we then solve for @xmath159 term and verify that @xmath160 for @xmath151 large .
we refer to proposition [ prop:1 ] for all details of the _ tail computation_. + @xmath161 _ the universal system of odes .
_ from the above procedure we deduce after @xmath162 iterations the following universal system of odes : @xmath163 unlike the critical case treated in @xcite , we do not need a logarithmic correction .
the set of solutions to ( see lemma [ lemm : solsysb ] below ) is explicitly given by @xmath164 in the original time variable @xmath165 , this implies that @xmath166 goes to zero in finite time @xmath38 with the asymptotic @xmath167 furthermore , the linearized flow of around the solution is explicit and manisfests @xmath9 unstable directions ( see lemma [ lemm : lisysb ] below ) .
this induces that the case @xmath10 corresponds to a stable type ii blowup regime .
+ @xmath168 _ the flow decomposition and the equations of modulation . _
consider @xmath80 as the approximate solution given by which from the construction will engender an approximate solution of the renormalized flow , @xmath169 where the modulation equation term is approximately given by : @xmath170t_i.\ ] ] we localize @xmath80 in the zone @xmath171 to prevent the appearance of the irrelevant growing tails for @xmath172 .
we then take initial data of the form @xmath173 where @xmath174 is small in some suitable sense and @xmath175 is chosen to be close to the exact solution .
the following decomposition of the flow is stanrdard consequence of the modulation theory @xmath176 where @xmath177 are the @xmath178 modulation parameters and they are fixed to insure the following orthogonality conditions : @xmath179 where @xmath180 ( see ) is a well adapted approximation of the kernel of @xmath127 which depends on a certain large constant @xmath181 . as a consequence , @xmath182 produces also a well adapted approximation of the powers of @xmath127 kernel .
thanks to the orthogonal decomposition we compute the modulation equations driving the parameters @xmath177 ( see lemmas [ lemm : mod1 ] and [ lemm : mod2 ] below ) , @xmath183 where @xmath184 is the error measured in certain norm well adapted to our estimations .
@xmath185 _ control of sobolev norms . _ from , we see that it is important to prove that the local norms of @xmath186 remain small enough and do not disturb the dynamical system .
this is accomplished via the control of the error @xmath186 through norms that are well adapted to our estimations and coercive to some weighted sobolev norms . in particular
, we have the following property which gives the control of some weighted sobolev norms under the orthogonality conditions ( see lemma [ lemm : coelk ] ) , @xmath187 where @xmath188 is given by . here
the factorization will simplify our proof . as in @xcite , @xcite and @xcite , the control of @xmath189
is done through the use of the linearized equation in the original variables @xmath190 , i.e. we work with @xmath191 in and not @xmath186 .
the energy estimate is of the form ( see proposition [ prop : e2k ] ) @xmath192 where the right hand side is the size of the error @xmath193 of the construction of our approximate profile @xmath80 above .
an integration of in time by using initial smallness assumptions , @xmath194 and @xmath195 yields the estimate @xmath196 which is good enough to control the local norms of @xmath186 and close the modulation equations .
note that we also need to control lower energies @xmath197 for @xmath198 because the control of the high energy @xmath189 alone is not sufficient to control the nonlinear term appearing after the linearization around @xmath80 .
in particular , we exhibit a lyapunov functional with the dynamical estimate @xmath199 then , an integration in time yields @xmath200 which is enough to control the nonlinear term .
let us remark that the condition @xmath201 ensures @xmath202 so that @xmath197 is always controlled .
by the coercivity of @xmath197 , this means that we are only able to control the sobolev norms @xmath203 for @xmath204 resulting in the asymptotic .
+ the above framework allows us to design an appreciated shrinking set ( see definition [ def : skset ] for a precise definition ) which traps the solution according to the asymptotic dynamic described in theorem [ theo:1 ]
. following lemma [ lemm : solsysb ] and [ lemm : lisysb ] , the control of the solution in this set reduces to the control of @xmath9 unstable modes @xmath205 , which is a finite - dimensional problem .
we then solve this finite - dimensional problem through a topological argument based on the brouwer fixed point theorem ( see the proof of proposition [ prop : exist ] ) , and the proof of theorem [ theo:1 ] follows .
+ the paper is organized as follows . in section [ sec:2 ] ,
we give the construction of the approximate solution @xmath80 of and derive estimates on the generated error term @xmath193 ( proposition [ prop:1 ] ) as well as its localization ( proposition [ prop : localprofile ] ) .
we also give in this section some elementary facts on the study of the system ( lemmas [ lemm : solsysb ] and [ lemm : lisysb ] ) .
section [ sec:3 ] is devoted to the proof of theorem [ theo:1 ] assuming a main technical result ( proposition [ prop : redu ] ) .
in particular , we give the proof of the existence of the solution trapped in some shrinking set to zero ( proposition [ prop : exist ] ) such that the constructed solution satisfies the conclusion of theorem [ theo:1 ] .
readers not interested in technical details may stop there . in section [ sec:4 ] ,
we give the proof of proposition [ prop : redu ] which gives the reduction of the problem to a finite - dimensional one ; and this is the heart of our analysis .
this section is devoted to the construction of a suitable approximate solution to by using the same approach developed in @xcite .
similar approachs can also be found in @xcite , @xcite , @xcite , @xcite , @xcite .
the construction is mainly based on the fact that the kernel of the linearized operator @xmath127 around @xmath7 is given explicitly in the radial setting . following the scaling invariance of
, we introduce the following change of variables : @xmath206 which leads to the following renormalized flow : @xmath207 where @xmath208 . noticing that in the setting , we have @xmath209 and since we deal with the finite time blowup of the problem , we would naturally impose the condition @xmath210 for some @xmath211 .
hence , @xmath212 blows up in finite time @xmath38 .
let us assume that the leading part of the solution of is given by the harmonic map @xmath7 , which is unique solution ( up to scaling ) of the equation @xmath213 we aim at constructing an approximate solution of close to @xmath7 .
the natural way is to linearize equation around @xmath7 , which generates the schrdinger operator defined by .
let us now recall the main properties of @xmath127 in the following subsection . in this subsection , we recall the main properties of the linearized operator around @xmath7 , which are fundamental for the construction of the approximate profile
. actually , the properties of the linearized operator are also important to derive the coercivity properties used for the high sobolev energy estimates .
let us start by recalling the following result from biernat @xcite , which gives the asymptotic behavior of the harmonic map @xmath7 : let @xmath3 , there exists a unique solution @xmath7 to equation , which admits the following asymptotic behavior : for any @xmath214 , + @xmath123 ( asymptotic behavior of @xmath7 ) @xmath215\quad & \text{as } \quad y \to + \infty , \end{array } \right.\ ] ] where @xmath65 is defined in , @xmath216 and the constant @xmath217 .
+ @xmath133 ( degeneracy ) @xmath218\quad & \text{as } \quad y \to + \infty .
\end{array } \right.\ ] ] the proof of is done through the introduction of the variables @xmath219 and @xmath220 and consists of the phase portrait analysis of the autonomous equation @xmath221 all details of the proof can be found at pages 184 - 185 in @xcite .
the proof of directly follows from the expansion .
the linearized operator @xmath127 displays a remarkable structure given by the following lemma : [ lemm : factorl ] let @xmath3 and define the first order operators @xmath222 where @xmath223 we have @xmath224 where @xmath225 stands for the conjugate hamiltonian .
the adjoint operator @xmath226 is defined with respect to the lebesgue measure @xmath227 we have @xmath228 since @xmath229 , one can express the definition of @xmath230 through the potential @xmath231 as follows : @xmath232 let @xmath233 be defined by @xmath234 then , a direct computation yields @xmath235 from and , we see that the kernel of @xmath236 and @xmath226 are explicit : @xmath237 hence , the elements of the kernel of @xmath127 are given by @xmath238 where @xmath239 can be found from the wronskian relation @xmath240 that is @xmath241 which admits the asymptotic behavior : @xmath242 from , we may invert @xmath127 as follows : @xmath243 the factorization of @xmath127 grants us to compute @xmath244 easily .
in particular , we have the following : [ lemm : inversionl ] let @xmath245 be a @xmath246 radially symmetric function and @xmath247 be given by , then @xmath248 from the relation , we compute @xmath249 applying @xmath236 to and using the cancellation @xmath250 , we obtain @xmath251 from the definition of @xmath236 , we write @xmath252 this concludes the proof of lemma [ lemm : inversionl ] .
we define a class of admissible functions which exhibit a relevant behavior both at the origin and infinity .
[ def : admitfunc ] fix @xmath253 , we say that a smooth function @xmath254 is admissible of degree @xmath255 if + @xmath256 @xmath245 admits a taylor expansion to all orders around the origin , @xmath257 @xmath258 @xmath245 and its derivatives admit the bounds , for @xmath259 , @xmath260 note from that @xmath261 is admissible of degree @xmath262 . one note that @xmath127 naturally acts on the class of admissible function in the following way : [ lemm : actionll ]
let @xmath245 be an admissible function of degree @xmath255 , then : + @xmath256 @xmath263 is admissible of degree @xmath264 .
+ @xmath265 @xmath266 is admissible of degree @xmath267 .
+ @xmath268 @xmath269 is admissible of degree @xmath270 .
@xmath271 this is simply a consequence of definition [ def : admitfunc ] .
+ @xmath161 we aim at proving that if @xmath245 is admissible of degree @xmath264 , then @xmath247 is admissible of degree @xmath270 .
to do so , we use lemma [ lemm : inversionl ] to estimate + - for @xmath272 , @xmath273 @xmath274 - for @xmath259 , @xmath275 @xmath276 from the last formula in and , we estimate @xmath277 using @xmath278 , we get @xmath279 by taking radial derivatives of @xmath278 , we obtain by induction @xmath280 this concludes the proof of lemma [ lemm : actionll ] .
the following lemma is a consequence of lemma [ lemm : actionll ] : [ lemm : genlk ] let the sequence of profiles @xmath281 then + @xmath256 @xmath282 is admissible of degree @xmath283 for @xmath284 .
+ @xmath258 @xmath285 is admissible of degree @xmath286 for @xmath287 .
@xmath123 we note from that @xmath261 is admissible of degree @xmath262 . by induction and part @xmath161 of lemma [ lemm :
actionll ] , the conclusion then follows .
+ @xmath133 we proceed by induction . for @xmath288 , we explicitly compute @xmath289 by using lemma [ lemm : inversionl ] and the expansion to get @xmath290 by induction , one can easily check that @xmath291 for @xmath292 .
hence , @xmath293 = \lambda \partial_y^m t_1 - ( 2 - \gamma - m)\partial_y^mt_1 = { \mathcal{o}}\left(y^{-\gamma - m}\right ) \quad \text{as } \quad y \to + \infty.\ ] ] since @xmath294 and @xmath295 are admissible of degree @xmath296 , we deduce that @xmath297 is admissible of degree @xmath298 .
we now assume the claim for @xmath299 , namely that @xmath285 is admissible of degree @xmath300 .
let us prove that @xmath301 is admissible of degree @xmath302 .
we use formula and definition to write @xmath303 from part @xmath123 , we know that @xmath304 is admissible of degree @xmath305 . from and , one can check that @xmath306 admits the asymptotic : @xmath307 and @xmath308 together with the induction hypothesis , we deduce that the right hand side of is admissible of degree @xmath286 .
the conclusion then follows by using part @xmath161 of lemma [ lemm : actionll ] .
this ends the proof of lemma [ lemm : genlk ] .
we close this subsection by defining a simple notion of homogeneous admissible function .
let @xmath109 be an integer and @xmath309 , we say that a function @xmath310 with @xmath311 is homogeneous of degree @xmath312 if it is a finite linear combination of monomials @xmath313 with @xmath314 admissible of degree @xmath264 in the sense of definition [ def : admitfunc ] and @xmath315 we set @xmath316 in this subsection , we use the explicit structure of the linearized operator @xmath127 to construct an approximate blow - up profile .
in particular , we claim the following : [ prop:1 ] let @xmath3 and @xmath109 be an integer .
let @xmath119 be a large enough universal constant , then there exist a small enough universal constant @xmath317 such that the following holds true .
let a @xmath318 map @xmath319 \mapsto ( -b^ * , b^*)^l,\ ] ] with a priori bounds in @xmath320 $ ] : @xmath321 then there exist homogeneous profiles @xmath322 such that @xmath323 produces an approximate solution to the remormalized flow : @xmath324 with the following property : + @xmath123 ( modulation equation ) @xmath325 \left[t_k + \sum_{j = k + 1}^{l+2}\frac{\partial s_j}{\partial b_k}\right],\ ] ] where we use the convention @xmath326 for @xmath327 . + @xmath133 ( estimate on the profiles ) the profiles @xmath328 are homogeneous with @xmath329 @xmath161 ( estimate on the error @xmath193 ) for all @xmath330 , there holds : + - ( global weight bound ) @xmath331 where @xmath332 , @xmath333 , @xmath334 are defined in and .
+ - ( improved local bound ) @xmath335 we aim at constructing the profiles @xmath328 such that @xmath336 defined from has the _ least possible growth _ as @xmath337 .
the structure of the linearized operator @xmath127 defined in is the main ingredient for the construction of the approximate profile .
this procedure will lead to the leading - order modulation equation @xmath338 which actually cancels the worst growth of @xmath339 as @xmath337 .
[ [ bullet - expansion - of - psi_b ] ] @xmath340 expansion of @xmath193 + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + . from and , we write @xmath341:= a_1 + a_2.\end{aligned}\ ] ] using the expression of @xmath342 and the definition of @xmath282 ( note that @xmath343 with the convention @xmath344 ) , we write @xmath345 + \sum_{k = 2}^{l+2}\big[\partial_s
s_k + { \mathscr{l}}s_k + b_1 \lambda s_k\big]\\ & = \sum_{k = 1}^l \big[(b_k)_st_k - b_{k+1}t_k + b_1b_k \lambda t_k \big ] + \sum_{k = 2}^{l+2}\big[\partial_s s_k + { \mathscr{l}}s_k + b_1 \lambda s_k\big]\\ & = \sum_{k = 1}^l \big[(b_k)_s - b_{k+1 } + ( 2k - \gamma)b_1b_k\big]t_k\\ & \qquad + \sum_{k = 1}^l\big[{\mathscr{l}}s_{k + 1 } + \partial_s s_k + b_1b_k
\big[\lambda t_k - ( 2k - \gamma)t_k\big ] + b_1 \lambda s_k\big]\\ & \qquad\qquad+ \big[{\mathscr{l}}s_{l+2 } + \partial_s s_{l+1 } + b_1\lambda s_{l+1}\big ] + \big[\partial_ss_{l+2 } + b_1\lambda s_{l+2}\big].\end{aligned}\ ] ] we now write @xmath346\frac{\partial s_k}{\partial b_j } \\ & \qquad - \sum_{j = 1}^l\big[(2j - \gamma)b_1b_j - b_{j + 1 } \big]\frac{\partial s_k}{\partial b_j}.\end{aligned}\ ] ] hence , @xmath347 + e_{l+2},\end{aligned}\ ] ] where for @xmath348 , @xmath349 + b_1 \lambda s_k - \sum_{j = 1}^{k-1}\big[(2j - \gamma)b_1b_j - b_{j + 1 } \big]\frac{\partial s_k}{\partial b_j},\ ] ] and for @xmath350 , @xmath351\frac{\partial s_k}{\partial b_j}.\ ] ] for the expansion of the nonlinear term @xmath352 , let us denote @xmath353 and use a taylor expansion to write ( see pages 1740 in @xcite for a similar computation ) @xmath354 = \frac{(d-1)}{2y^2}\left[\sum_{i = 2}^{l+2}p_i + r_1 + r_2\right],\end{aligned}\ ] ] where @xmath355 @xmath356 @xmath357 with @xmath358 and @xmath359 + e_{l+2 } + \frac{(d-1)}{2y^2}(r_1 + r_2)\ ] ] [ [ bullet - construction - of - s_k . ] ] @xmath340 construction of @xmath339 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + from the expression of @xmath193 given in , we construct iteratively the sequences of profiles @xmath360 through the scheme @xmath361 where @xmath362 we claim by induction on @xmath363 that @xmath364 is homogeneous with @xmath365 and @xmath366 from item @xmath161 of lemma [ lemm : actionll ] and , we deduce that @xmath339 is homogeneous of degree @xmath367 and from , we get @xmath368 which is the conclusion of item @xmath133 . +
let us now give the proof of and .
we proceed by induction .
+ - case @xmath369 : we compute explicitly from and , @xmath370,\ ] ] which directly follows . from lemma [ lemm :
genlk ] , we know that @xmath294 and @xmath297 are admissible of degree @xmath296 and @xmath298 respectively . using , one can check the bound @xmath371 since @xmath294 is admissible of degree @xmath296 , we have that @xmath372 by the leibniz rule and the fact that @xmath373 , we get that @xmath374 we also have the expansion near the origin , @xmath375 hence , @xmath376 is admissible of degree @xmath377 , which concludes the proof of for @xmath369 .
+ - case @xmath378 : estimate holds by direct inspection . let us now assume that @xmath339 is homogeneous of degree @xmath379 and prove that @xmath380 is homogeneous of degree @xmath381 . in particular , the claim immediately follows from part @xmath161 of lemma [ lemm : actionll ] once we show that @xmath382 is homogeneous with @xmath383 from part @xmath133 of lemma [ lemm : genlk ] and the a priori assumption , we see that @xmath384 is homogeneous of degree @xmath385 . from part @xmath123 of lemma [ lemm : actionll ] and
the induction hypothesis , @xmath386 is also homogeneous of degree @xmath385 . by definition , @xmath387 is homogeneous and has the same degree as @xmath339 .
thus , @xmath388 is homogeneous of degree @xmath385 . from definitions and ,
we derive @xmath389 it remains to control the term @xmath390 . from the definition ,
we see that @xmath390 is a linear combination of monomials of the form @xmath391 with @xmath392 recall from part @xmath123 of lemma [ lemm : genlk ] the bound @xmath393 and from the induction hypothesis and the a priori bound , @xmath394 together with the bound , we obtain the following bound at infinity , @xmath395 the control of @xmath396 follows by the leibniz rule and the above estimates . one can also check that @xmath397 is of order @xmath398 near the origin .
this concludes the proof of as well as part @xmath133 of proposition [ prop:1 ] .
+ [ [ bullet - estimate - on - psi_b . ] ] @xmath340 estimate on @xmath193 .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + from and , the expression of @xmath193 is now reduced to @xmath399 where @xmath400 , @xmath401 and @xmath402 are given by , and .
we start by estimating @xmath400 term defined by . since @xmath403 is homogeneous of degree @xmath404 and thus so are @xmath405 and @xmath406 .
this follows that @xmath400 is homogeneous of degree @xmath407 . using part @xmath133 of lemma [ lemm : actionll ] and the relation @xmath408 ( see )
, we estimate for all @xmath330 @xmath409 where @xmath410 , @xmath411 .
we now turn to the control of the term @xmath412 , which is a linear combination of terms of the form ( see ) @xmath413 with @xmath414 using the admissibility of @xmath415 and the homogeneity of @xmath416 , we get the bounds @xmath417 and @xmath418 where we used the fact that @xmath419 and @xmath420 , and similarly for higher derivatives by the leibniz rule .
thus , we obtain the round estimate for all @xmath330 , @xmath421 the term @xmath422 is estimated exactly as for the term @xmath412 using the definition .
similarly , the control of @xmath423 is obtained along the exact same lines as above .
this concludes the proof of .
the local estimate directly follows from the homogeneity of @xmath339 and the admissibility of @xmath282 .
this concludes the proof of proposition [ prop:1 ] .
we now proceed to a simple localization of the profile @xmath80 to avoid the growth of tails in the region @xmath424 .
more precisely , we claim the following : [ prop : localprofile ] under the assumptions of proposition [ prop:1 ] , we assume in addition the a priori bound @xmath425 where @xmath332 and @xmath426 are defined as in and .
then @xmath427 where @xmath428 satisfies the bounds : + @xmath256 ( large sobolev bound ) for all @xmath429 , @xmath430 and @xmath431 where @xmath333 and @xmath334 are defined by .
+ @xmath133 ( very local bound ) for all @xmath432 and @xmath330 , @xmath433 @xmath161 ( refined local bound near @xmath434 ) for all @xmath330 , @xmath435 by a direct computation , we have @xmath436 \\ & \quad + \theta_b \left
[ \partial_s \chi_{b_1 } - \left(\partial_{yy}\chi_{b_1 } + \frac{d-1}{y}\partial_y \chi_{b_1}\right ) + b_1 \lambda \chi_{b_1}\right ] - 2\partial_y \chi_{b_1}\partial_y\theta_b + b_1(1 - \chi_{b_1})\lambda q \\ & \qquad + \frac{(d-1)}{2y^2}\left[\sin(2\tilde q_b ) - \sin(2q ) - \chi_{b_1}(\sin(2q_b ) - \sin(2q ) ) \right].\end{aligned}\ ] ] according to and , we write @xmath437 where @xmath438}_{\hat \psi_b^{(2)}}\\ & \qquad + \underbrace { \theta_b \left [ \partial_s \chi_{b_1 } - \left(\partial_{yy}\chi_{b_1 } + \frac{d-1}{y}\partial_y \chi_{b_1}\right ) + b_1 \lambda \chi_{b_1}\right ] - 2\partial_y \chi_{b_1}\partial_y\theta_b}_{\hat \psi_b^{(3)}}.\end{aligned}\ ] ] the contribution of the term @xmath439 to the bounds , , and follows exactly the same as in the proof of and .
we therefore left to estimate the term @xmath440 .
since all the terms in the expression of @xmath440 are localized in @xmath441 except for the first one whose support is a subset of @xmath442 .
hence , the estimates and directly follow from and .
let us now find the contribution of @xmath443 to the bounds and .
we estimate @xmath444 hence , using the relation @xmath445 ( see ) and the definition of @xmath332 , we estimate for all @xmath330 , @xmath446 for the nonlinear term @xmath447 , we note from the admissibility of @xmath282 and the homogeneity of @xmath339 that the @xmath282 terms dominate for @xmath448 in @xmath342 .
thus , for @xmath448 , @xmath449 using and noting that @xmath447 is localized in @xmath441 , we obtain the round bound @xmath450 we then estimate for @xmath330 , @xmath451 to control @xmath452 , we first note from the definition and the assumption that @xmath453 using , we estimate for @xmath330 , @xmath454 gathering all the bounds yields @xmath455 the control of @xmath456 , @xmath457 and @xmath458 are obtained along the exact same lines as above .
this concludes the proof of and as well as proposition .
the construction of the @xmath80 profile formally leads to the finite dimensional dynamical system for @xmath311 by setting to zero the inhomogeneous @xmath459 term given in : @xmath460 unlike the critical case @xmath461 treated in @xcite , there is no more logarithmic correction to be taken into account in the system . in particular , the system admits explicit solutions and the linearized operator near these solutions is explicit .
[ lemm : solsysb ] let @xmath462 and the sequence @xmath463 then the explicit choice @xmath464 is a solution to . the proof of lemma [ lemm : solsysb ] can be left to the reader since it directly follows from an explicit computation .
we claim that the linearized flow of near the solution is explicit and exhibits @xmath9 unstable directions .
note that the stability is considered in the sense that @xmath465 in particular , we have the following result which was proved in @xcite : [ lemm : lisysb ]
let @xmath466 and note @xmath467 .
then , for @xmath468 , @xmath469,\ ] ] and @xmath470,\ ] ] where @xmath471 moreover , @xmath472 is diagonalizable : @xmath473 since we have an analogous system as the one in @xcite and the proof is essentially the same as written there , we kindly refer the reader to see lemma 3.7 in @xcite for all details of the proof .
this section is devoted to the proof of theorem [ theo:1 ] .
we hope that the explanation of the strategy we give in this section will be reader friendly .
we proceed in 3 subsections : + - in the first subsection , we give an equivalent formulation of the linearization of the problem in the setting .
+ - in the second subsection , we prepare the initial data and define the shrinking set @xmath474 ( see definition [ def : skset ] ) such that the solution trapped in this set satisfies the conclusion of theorem [ theo:1 ] .
+ - in the third subsection , we give all arguments of the proof of the existence of solutions trapped in @xmath474 ( proposition [ prop : exist ] ) assuming an important technical result ( proposition [ prop : redu ] ) whose proof is left to the next section .
then we conclude the proof of theorem [ theo:1 ] .
let @xmath109 be an integer and @xmath475 , we introduce the renormalized variables : @xmath476 and the decomposition @xmath477 where @xmath478 is constructed in proposition [ prop : localprofile ] and the modulation parameters @xmath479 are determined from the @xmath178 orthogonality conditions : @xmath480 where @xmath180 is a fixed direction depending on some large constant @xmath181 defined by @xmath481 with @xmath482 here , @xmath180 is build to ensure the nondegeneracy @xmath483 and the cancellation @xmath484 in particular , we have @xmath485 from , we see that @xmath186 satisfies the equation : @xmath486 where @xmath487 and @xmath488 is the linear part given by @xmath489q,\ ] ] and @xmath490 is the purely nonlinear term @xmath491.\ ] ] we also need to write the equation in the original variables .
to do so , let the rescaled linearized operator : @xmath492 and the renormalized function @xmath493 then from , @xmath191 satisfies the equation @xmath494 note that @xmath495 we describe in this subsection the set of initial data @xmath31 of the problem as well as the initial data for @xmath496 leading to the blowup scenario of theorem [ theo:1 ] . assume that @xmath497 satisfies @xmath498 from the continuity of the flow and a standard argument , we can propagate the smallness assumption on a small time interval @xmath499 .
hence , the decomposition , @xmath500 can be defined uniquely on the interval @xmath501 $ ] .
the existence of the decomposition is from the implicit function theorem and the explicit relations @xmath502 which induces the nondegeneracy of the jacobian @xmath503 in fact , the decomposition exists as long as @xmath504 and @xmath186 remains small in the energy topology .
we now set up the shrinking set for the control of the parameters @xmath505 and the error @xmath186 . to measure
the regularity of the map we use the following coercive norms of @xmath186 : + @xmath506 our construction is build on a careful choice of the initial data for the modulation parameter @xmath507 and the radiation @xmath186 at time @xmath508 .
in particular , we will choose it in the following way : [ def:1 ] given @xmath509 and @xmath334 as in and .
let consider the variable @xmath510 where @xmath467 is introduced in the linearization , namely that @xmath511 and @xmath512 refers to the diagonalization of @xmath472 .
let @xmath513 , we assume that * smallness of the initial perturbation for the @xmath514 unstable modes : @xmath515 * smallness for the @xmath514 stable modes of the initial perturbation : @xmath516 * normalization : we can assume up to a fixed rescaling : @xmath517 in particular , the initial data described in definition [ def:1 ] belongs to the following set which shrinks to zero as @xmath518 : [ def : skset ] given @xmath509 and @xmath334 as in and . for all @xmath519 and @xmath520 , we define @xmath521 as the set of all @xmath522 such that @xmath523 note from that the bounds given in definition [ def : skset ] imply that for @xmath509 small enough , @xmath524 hence , the choice of the initial data @xmath525 belongs in @xmath526 if @xmath527 is large enough .
the introduction of the high sobolev norm @xmath189 is reflected on the following relation : @xmath528 which is computed thanks to the @xmath529 orthogonality conditions ( see lemmas [ lemm : mod1 ] and [ lemm : mod2 ] below ) .
we claim the following proposition : [ prop : exist ] there exists @xmath530 such that for @xmath531 , there exists @xmath532 such that for all @xmath533 , there exists initial data for the unstable modes @xmath534^{\ell - 1},\ ] ] such that the corresponding solution @xmath535 for all @xmath536 .
let us briefly give the proof of proposition [ prop : exist ] .
let us consider @xmath519 and @xmath513 and @xmath525 as in definition [ def:1 ] .
we introduce the exit time @xmath537 and assume that for any choice of @xmath534^{\ell - 1},\ ] ] the exit time @xmath538 and look for a contradiction . by the definition of @xmath539 ,
at least one of the inequalities in that definition is an equality .
owing the following proposition , this can happens only for the components @xmath540 .
precisely , we have the following result which is the heart of our analysis : [ prop : redu ] there exists @xmath541 such that for each @xmath542 , there exists @xmath543 such that for all @xmath544 , the following holds : given the initial data at @xmath508 as in definition [ def:1 ] , if @xmath545 for all @xmath546 $ ] , with @xmath547 for some @xmath548 , then : + @xmath123 ( reduction to a finite dimensional problem ) @xmath549^{\ell - 1}.\ ] ] @xmath133 ( transverse crossing ) @xmath550 let us assume proposition [ prop : redu ] and continue the proof of proposition [ prop : exist ] . from part @xmath123 of proposition [ prop : redu ] , we see that @xmath551^{\ell - 1},\ ] ] and the following mapping @xmath552^{\ell - 1 } & \mapsto \partial \left([-1,1]^{\ell - 1}\right)\\ s_0^{\frac{\eta}{2}(1 - \delta)}\big({\mathcal{v}}_2(s_0 ) , \cdots , { \mathcal{v}}_\ell(s_0)\big ) & \to \frac{s_*^{\frac{\eta}{2}(1 - \delta)}}{k } \big({\mathcal{v}}_2(s _ * ) , \cdots , { \mathcal{v}}_\ell(s_*)\big)\end{aligned}\ ] ] is well defined . applying the transverse crossing property given in part @xmath133 of proposition [ prop : redu ] , we see that @xmath553 leaves @xmath521 at @xmath508 , hence , @xmath554 .
this is a contradiction since @xmath16 is the identity map on the boundary sphere and it can not be a continuous retraction of the unit ball .
this concludes the proof of proposition [ prop : exist ] , assuming that proposition [ prop : redu ] holds .
[ [ conclusion - of - theorem - theo1-assuming - proposition - propredu ] ] conclusion of theorem [ theo:1 ] assuming proposition [ prop : redu ] + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + . from proposition
[ prop : exist ] , we know that there exist initial data @xmath525 such that @xmath555 from , , we have @xmath556,\end{aligned}\ ] ] which yields @xmath557 we easily conclude that @xmath558 vanishes in finite time @xmath559 with the following behavior near the blowup time : @xmath560 which is the conclusion of item @xmath123 of theorem [ theo:1 ] .
for the control of the sobolev norms , we observe from and definition [ def : skset ] that @xmath561 from the relation @xmath562 , we deduce that @xmath563 , \quad \int |\nabla^\sigma q|^2 \to 0 \quad \text{as}\;\;s \to + \infty,\ ] ] which yields @xmath133 of theorem [ theo:1 ] .
in this section , we aim at proving proposition [ prop : redu ] which is the heart of our analysis .
we proceed in three separate subsections : - in the first subsection , we derive the laws for the parameters @xmath505 thanks to the orthogonality condition and the coercivity of the powers of @xmath127 . - in the second subsection , we prove the main monotonicity tools for the control of the infinite dimensional part of the solution .
in particular , we derive a suitable lyapunov functional for the @xmath189 energy as well as the monotonicity formula for the lower sobolev energy . - in the third subsection , we conclude the proof of proposition [ prop : redu ] thanks to the identities obtained in the first two parts .
we derive here the modulation equations for @xmath496 .
the derivation is mainly based on the orthogonality and the coercivity of the powers of @xmath127 .
let us start with elementary estimates relating to the fixed direction @xmath180 .
[ lemm : estphim ] given @xmath180 as defined in , we have the followings : @xmath564 @xmath565 arguing by induction , we assume that @xmath566 using the fact that @xmath567 is admissible of degree @xmath568 , we estimate from the definition , @xmath569 using the estimate for @xmath570 yields @xmath571 and @xmath572 this concludes the proof of lemma [ lemm : estphim ] . from the orthogonality conditions and equation , we claim the following : [ lemm : mod1 ] given @xmath333 , @xmath334 and @xmath509 as defined in and . for @xmath519 ,
we assume that there is @xmath573 such that @xmath545 for @xmath546 $ ] for some @xmath548 . then
, the followings hold for @xmath546 $ ] : @xmath574 and @xmath575 we start with the law for @xmath576 .
let @xmath577 where we recall that @xmath578 if @xmath579 .
now , we take the inner product of with @xmath580 and use the orthogonality to write @xmath581 from the definition , we see that @xmath180 is localized in @xmath582 . from and
, we compute by using the identity , @xmath583 + { \mathcal{o}}(m^c b_1d(t)).\ ] ] the error term is estimated by using with @xmath584 and lemma [ lemm : estphim ] , @xmath585 for the linear term , we apply lemma [ lemm : coelk ] with @xmath586 , @xmath587 hence , the cauchy - schwartz inequality yields , @xmath588 the remaining terms are easily estimated by using the following bound coming from lemma [ lemm : coelk ] and lemma [ lemm : coerl ] , @xmath589 which implies @xmath590 put all the above estimates into and use together with the relation , we arrive at @xmath591 for the modulation equations for @xmath514 with @xmath592 , we take the inner product of with @xmath593 and use the orthogonality to write for @xmath592 , @xmath594 proceed as for @xmath576 , we end up with @xmath595 similarly , we have by taking the inner product of with @xmath180 , @xmath596 from , and , we obtain the round bound @xmath597 the conclusion then follows by substituting this bound into , and .
this ends the proof of lemma [ lemm : mod1 ] . from the bound for @xmath598 given in definition [ def : skset ] and the modulation equation
, we only have the pointwise bound @xmath599 which is not good enough to close the expected one @xmath600 we claim the following improved modulation equation for @xmath576 : [ lemm : mod2 ] under the assumption of lemma [ lemm : mod1 ] , the following bound holds for all @xmath546 $ ] : @xmath601.\label{eq : odeblimproved}\end{aligned}\ ] ] we commute with @xmath602 and take the inner product with @xmath603 to get @xmath604 - \left<{\mathscr{l}}^l q , \chi_{b_0}\lambda q\right > \frac{d}{ds}\left[\frac{1 } { \left<\lambda q , \chi_{b_0}\lambda q\right > } \right ] \right\}\nonumber\\ & = \left<{\mathscr{l}}^l q , \lambda q \partial_s(\chi_{b_0})\right > - \left<{\mathscr{l}}^{l+1 } q , \chi_{b_0}\lambda q\right > + \frac{\lambda_s}{\lambda}\left<{\mathscr{l}}^l\lambda q , \chi_{b_0}\lambda q\right > \nonumber\\ & \quad - \left<{\mathscr{l}}^l \tilde{\psi}_b,\chi_{b_0}\lambda q\right > - \left < { \mathscr{l}}^l\widehat{\textup{mod}}(t ) , \chi_{b_0}\lambda q\right > + \left<{\mathscr{l}}^l\big({\mathcal{l}}(q ) - { \mathcal{n}}(q)\big ) , \chi_{b_0}\lambda q\right>.\label{eq : lqlamqt1}\end{aligned}\ ] ] we recall from that @xmath605 let us estimate the second term in the left hand side of .
we use cauchy - schwartz and lemma [ lemm : coelk ] to estimate @xmath606 we write @xmath607\right| & \lesssim \frac{|\left<{\mathscr{l}}^l q , \chi_{b_0}\lambda q\right>|}{\left<\lambda q , \chi_{b_0}\lambda q\right>^2 } \left|\frac{(b_1)_s}{b_1}\right|\int_{b_0 \leq y \leq 2b_0}|\lambda q|^2\\ & \quad \lesssim b_1 \frac{b_0^{\frac d2 - \gamma + 2\hbar + 2 } \sqrt{{\mathscr{e}}_{2\bbbk}}}{b_0^{2d - 4\gamma}}b_0^{d - 2\gamma } \lesssim \frac{\sqrt{{\mathscr{e}}_{2\bbbk}}}{b_0^{2\delta}},\end{aligned}\ ] ] where we used the relation . for the first three terms in the right hand side of
, we use the cauchy - schwartz , lemma [ lemm : coelk ] and the fact that @xmath229 to find that @xmath608 and @xmath609 and @xmath610 the error term is estimated by using , @xmath611 the last term in the right hand side of is estimated in the same way , @xmath612 for the remaining term , we recall that @xmath229 , @xmath613 for @xmath592 , and @xmath614 , from which @xmath615 from , and the fact that @xmath616 , we write @xmath617 \right|\\ & \quad \lesssim \sum_{k = 1}^{l } \left|(b_k)_s + ( 2k - \gamma)b_1b_l - b_{k+1 } \right| \left|\left<\sum_{j = k+1}^{l+2 } \frac{\partial \tilde{s}_j}{\partial b_k } , { \mathscr{l}}^l ( \chi_{b_0}\lambda q)\right>\right|\\ & \qquad + \left|\frac{\lambda_s}{\lambda } + b_1\right| \left|\left<\lambda \tilde{\theta}_b,{\mathscr{l}}^l ( \chi_{b_0}\lambda q)\right>\right|.\end{aligned}\ ] ] recall that @xmath282 is admissible of degree @xmath283 and @xmath339 is homogeneous of degree @xmath379 , we derive the round bounds for @xmath618 : @xmath619 thus , from lemma [ lemm : mod1 ] , we derive the bound @xmath620 the equation follows by gathering all the above estimates into , dividing both sides of by @xmath621 and using the relation .
this finishes the proof of lemma [ lemm : mod2 ] .
we derive in this subsection the main monotonocity formula for @xmath622 for @xmath623 .
we claim the following which is the heart of this paper : [ prop : e2k ] we have @xmath624\right\}\nonumber\\ & \qquad \qquad \leq \frac{b_1}{\lambda^{4\bbbk - d + 2}}\left[\frac{{\mathscr{e}}_{2\bbbk}}{m^{2\delta } } + b_1^{l+(1 - \delta)(1 + \eta)}\sqrt{{\mathscr{e}}_{2\bbbk } } + b_1^{2l + 2(1 - \delta)(1 + \eta)}\right],\label{eq : es2slya}\end{aligned}\ ] ] and for @xmath625 , @xmath626\right\ } \leq \frac{b_1}{\lambda^{4 m - d + 2}}\left[b_1^{m - \hbar -1 + ( 1 - \delta ) - c\eta}\sqrt{{\mathscr{e}}_{2 m } } + b_1^{2(m-\hbar -1 ) + 2(1 - \delta ) - c\eta}\right],\label{eq : es2slyam}\end{aligned}\ ] ] the proof uses some ideas developed in @xcite and @xcite . because the proof of follows exactly the same lines as for , we only deal with the proof of .
let us start the proof of
. * step 1 : suitable derivatives and energy identity . * for @xmath284
, we define the suitable derivatives of @xmath186 and @xmath191 as follows : @xmath627 where @xmath628 and @xmath629 satisfy and respectively , the linearized operator @xmath127 and @xmath630 are defined by and , @xmath236 and @xmath226 are the first order operators defined by and , and @xmath631 with @xmath632 admitting the asymptotic behaviors as in . with the notation , we note that @xmath633 recall from lemma [ lemm : factorl ] , we have the following factorization : @xmath634 where @xmath635 and @xmath636 with @xmath233 expressed in terms of @xmath231 as in .
+ we commute equation with @xmath637 and use the notation to derive @xmath638v + { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right).\ ] ] now commuting this equation with @xmath639 yields @xmath640v + { \mathscr{a}}_\lambda { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right),\ ] ] since @xmath641 , we then have @xmath642 using the definition of @xmath643 and an integration by parts , we write @xmath644 using the definition of @xmath226 and an integration by parts together with the definition of @xmath233 , we write @xmath645 from , we write @xmath646v + { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right)\right].\end{aligned}\ ] ] gathering all the above identities and using yields the energy identity @xmath647v + { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right)\right]\nonumber\\ & \quad + \int \left(\tilde{{\mathscr{l}}}_\lambda v_{2\bbbk -1 } + \frac{b_1(\lambda v)_\lambda}{\lambda^2r } v_{2\bbbk - 2}\right)\left[\frac{{\partial_t}v_\lambda}{r } v_{2\bbbk - 2 } + { \mathscr{a}}_\lambda [ { \partial_t } , { \mathscr{l}}^{\bbbk - 1}_\lambda]v + { \mathscr{a}}_\lambda { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right ) \right].\nonumber\end{aligned}\ ] ] we now estimate all terms in . the proof uses the coercivity estimate given in lemma [ lemm : coelk ] . in particular , we shall apply lemma [ lemm : coelk ] with @xmath586 to have the estimate @xmath648 * step 2 : control of the lower order quadratic terms . * let us start with the second term in the left hand side of . from and , we have the round bound @xmath649 using , and , we estimate @xmath650 for the third term in the right hand side of , we write @xmath651 a direct computation yields the round bound @xmath652 thus , we use , the cauchy - schwartz inequality and to estimate @xmath653 similarly , we have @xmath654 and @xmath655v \right| + \left|\int \left(\tilde{{\mathscr{l}}}_\lambda v_{2\bbbk -1 } + \frac{b_1(\lambda v)_\lambda}{\lambda^2r } v_{2\bbbk - 2}\right){\mathscr{a}}_\lambda [ { \partial_t } , { \mathscr{l}}^{\bbbk - 1}_\lambda]v \right|\\ & \leq \frac{1}{4}\int |\tilde{{\mathscr{l}}}_\lambda v_{2\bbbk -1}|^2 + c\left(\frac{b_1 ^ 2}{\lambda^{4\bbbk - d + 2}}{\mathscr{e}}_{2\bbbk } + \int \frac{\big|[{\partial_t } , { \mathscr{l}}_\lambda^{\bbbk - 1}]v\big|^2}{\lambda^2(1 + y^2 ) } + \int \big|{\mathscr{a}}_\lambda[{\partial_t } , { \mathscr{l}}_\lambda^{\bbbk - 1}]v\big|^2\right).\end{aligned}\ ] ] we claim the bound @xmath656v\big|^2}{\lambda^2(1 + y^2 ) } + \int \big|{\mathscr{a}}_\lambda[{\partial_t } , { \mathscr{l}}_\lambda^{\bbbk - 1}]v\big|^2 \lesssim \frac{b_1 ^ 2}{\lambda^{4\bbbk - d + 2}}{\mathscr{e}}_{2\bbbk},\ ] ] whose proof is left to appendix [ ap : estcomm ] .
the collection of all the above estimates to yields @xmath657 \right\ } & \leq -\frac{1}{4 } \int |\tilde{\mathscr{l}}_\lambda v_{2\bbbk - 1}|^2 + \frac{cb_1 ^ 2}{\lambda^{4\bbbk - d + 2}}{\mathscr{e}}_{2\bbbk}\nonumber\\ & \quad + \int \frac{b_1(\lambda v)_\lambda}{\lambda^2r } v_{2\bbbk - 1 } { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right)\nonumber\\ & \quad + \int \frac{b_1(\lambda v)_\lambda}{\lambda^2r } v_{2\bbbk - 2 } { \mathscr{a}}_\lambda{\mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right)\nonumber\\ & \quad + \int \tilde{{\mathscr{l}}}_\lambda v_{2\bbbk -1}{\mathscr{a}}_\lambda { \mathscr{l}}_\lambda^{\bbbk - 1}\left ( \frac{1}{\lambda^2 } { \mathcal{f}}_\lambda\right).\label{eq : enerid1}\end{aligned}\ ] ] * step 3 : further use of dissipation .
* we aim at estimating all terms in the right hand side of . from , and the cauchy - schwartz inequality , we write @xmath658 similarly , we have @xmath659 for the last term in , let us introduce the function @xmath660 and the decomposition @xmath661 where @xmath662 is referred to , @xmath663 , @xmath664 and @xmath665 are defined as in and respectively .
actually , we introduced the decomposition and @xmath666 to take advantage of the improved bound obtained in lemma [ lemm : mod2 ] .
we now write @xmath667 injecting all these bounds into yields @xmath668 \right\}\nonumber\\ & \qquad \leq -\frac{1}{8 } \int |\tilde{\mathscr{l}}_\lambda v_{2\bbbk - 1}|^2 + \frac{cb_1 ^ 2}{\lambda^{4\bbbk - d + 2}}{\mathscr{e}}_{2\bbbk } + \frac{1}{\lambda^{4\bbbk - d + 2}}\int { \mathscr{l}}^\bbbk q { \mathscr{l}}^\bbbk(\partial_s \xi_l)\nonumber\\ & \qquad \quad + \frac{b_1}{\lambda^{4\bbbk - d + 2 } } \sqrt{{\mathscr{e}}_{2\bbbk}}\left[\left(\int \frac{|{\mathscr{a}}{\mathscr{l}}^{\bbbk - 1}{\mathcal{f}}|^2}{1 + y^2 } \right)^\frac{1}{2 } + \left(\int \frac{|{\mathscr{l}}^{\bbbk - 1}{\mathcal{f}}|^2}{1 + y^4 } \right)^\frac{1}{2 } \right]\nonumber\\ & \qquad \quad + \frac{c}{\lambda^{4\bbbk - d + 2}}\left ( \sqrt{{\mathscr{e}}_{2\bbbk}}\left\|{\mathscr{l}}^\bbbk { \mathcal{f}}_0 \right\|_{l^2 } + \left\|{\mathscr{a}}{\mathscr{l}}^{\bbbk -1}{\mathcal{f}}_1 \right\|_{l^2}^2 \right ) .
\label{eq : enerid2}\end{aligned}\ ] ] * step 4 : estimates for @xmath662 term . *
recall from that we already have the following estimate for @xmath669 : @xmath670 * step 5 : estimates for @xmath663 term .
* we claim the following : @xmath671 where @xmath672 let us prove .
we only deal with the first term since the second term is estimated similarly .
we recall from the definition of @xmath663 , @xmath673\left(\tilde t_i + \sum_{j = i + 1}\frac{\partial s_j}{\partial b_i}\chi_{b_1}\right),\end{aligned}\ ] ] where @xmath674 is defined as in and we know from lemma [ lemm : genlk ] that @xmath140 is admissible of degree @xmath675 and from proposition [ prop:1 ] that @xmath676 is homogeneous of degree @xmath677 .
+ since @xmath678 and @xmath679 , we use lemma [ lemm : actionll ] to estimate @xmath680 where we used the algebra @xmath681 .
+ using the cancellation @xmath682 for @xmath683 and the admissibility of @xmath140 , we estimate @xmath684 using the homogeneity of @xmath676 , we estimate for @xmath683 , @xmath685 provided that @xmath686 .
the collection of the above bounds together with and yields @xmath687 the same estimate holds for @xmath688 by following the same lines as above .
this concludes the proof of .
+ we now prove .
let us write @xmath689\tilde t_i\\ & \quad + \sum_{i = 1}^{l}\big[(b_i)_s + ( 2i - \gamma)b_1 b_i - b_{i + 1}\big ] \chi_{b_1}\sum_{j = i+1}^{l+2}\frac{\partial s_j}{\partial b_i}\\ & \quad + \left[(b_l)_s + ( 2i - \gamma)b_1 b_l + \frac{d}{ds}\left\{\frac{\left<{\mathscr{l}}^lq , \chi_{b_0}\lambda q\right>}{\left<\lambda q , \chi_{b_0}\lambda q\right > } \right\ }
\right]\tilde t_l + \frac{\left<{\mathscr{l}}^lq , \chi_{b_0}\lambda q\right>}{\left<\lambda q , \chi_{b_0}\lambda q\right>}\partial_s\tilde{t}_l . \end{aligned}\ ] ] proceeding as for the proof of yields the estimate @xmath690 and @xmath691 from and , we have the bound @xmath692 we also have @xmath693 the collection of the above bounds together with lemmas [ lemm : mod1 ] and [ lemm : mod2 ] yields @xmath694 which is the conclusion of .
+ injecting the estimates , and into , we arrive at @xmath695 \right\}\nonumber\\ & \qquad \leq -\frac{1}{8 } \int |\tilde{\mathscr{l}}_\lambda v_{2\bbbk - 1}|^2 + \frac{b_1}{\lambda^{4\bbbk - d + 2}}\left(\frac{{\mathscr{e}}_{2\bbbk}}{m^{2\delta } } + b_1^{\eta(1 - \delta)}{\mathscr{e}}_{2\bbbk } + b_1^{l + ( 1 - \delta)(1 + \eta)}\sqrt{{\mathscr{e}}_{2\bbbk}}\right ) \nonumber\\ & \qquad \quad + \frac{b_1\sqrt{{\mathscr{e}}_{2\bbbk}}}{\lambda^{4\bbbk - d + 2}}\left[\left(\int \frac{|{\mathscr{a}}{\mathscr{l}}^{\bbbk - 1}{\mathcal{f}}_1|^2}{1 + y^2 } \right)^\frac{1}{2 } + \left(\int \frac{|{\mathscr{l}}^{\bbbk - 1}{\mathcal{f}}_1|^2}{1 + y^4 } \right)^\frac{1}{2 } \right]\nonumber\\ & \qquad \qquad + \frac{1}{\lambda^{4\bbbk - d + 2}}\left\|{\mathscr{a}}{\mathscr{l}}^{\bbbk -1}{\mathcal{f}}_1 \right\|_{l^2}^2 + \frac{1}{\lambda^{4\bbbk - d + 2}}\int { \mathscr{l}}^\bbbk q { \mathscr{l}}^\bbbk(\partial_s \xi_l )
. \label{eq : enerid3}\end{aligned}\ ] ] * step 6 : estimates for the linear small term @xmath664 .
* we claim the following @xmath696 we only deal with the estimate for the first term because the last two terms are estimated similarly .
let us rewrite from the definition of @xmath664 , @xmath697,\ ] ] where @xmath698 from the asymptotic behavior of @xmath7 given in , the admissibility of @xmath140 and the homogeneity of @xmath137 , we deduce that @xmath13 is a regular function both at the origin and at infinity .
we then apply the leibniz rule with @xmath586 and @xmath699 to write @xmath700,\ ] ] where @xmath701 with @xmath702 are defined by the recurrence relation given in lemma [ lemm : leibnizlk ] .
in particular , we have the following estimate @xmath703 hence , we estimate from , @xmath704\\ & \quad \lesssim b_1 ^ 2\sum_{m = 0}^{\bbbk - 1}\left[\int \frac{|q_{2 m + 1}|^2}{1 + y^{2 + 2(2\bbbk - 1 - 2 m - 1 ) } } + \int \frac{|q_{2m}|^2}{1 + y^{2 + 2(2\bbbk - 1 - 2 m ) } } \right]\\ & \qquad \lesssim b_1 ^
2\sum_{m = 0}^{\bbbk - 1}\left[\int \frac{|q_{2 m + 1}|^2}{1 + y^{2 + 4(\bbbk - 1 - m ) } } + \int \frac{|q_{2m}|^2}{1 + y^{4 + 4(\bbbk - 1 - m ) } } \right]\\ & \qquad \quad\lesssim b_1 ^ 2 { \mathscr{e}}_{2\bbbk}.\end{aligned}\ ] ] this concludes the proof of . + * step 7 : estimates for the nonlinear term @xmath665 .
* this is the most delicate point in the proof of .
we claim the following @xmath705 @xmath706 provided that @xmath509 and @xmath707 are small enough .
we only deal with the proof of since the same proof holds for .
+ - _ control for @xmath708 .
let us rewrite from the definition of @xmath665 , @xmath709.\ ] ] from and the admissibility of @xmath140 , we write @xmath710 where @xmath711
let @xmath712 $ ] and @xmath713 we obtain from proposition [ prop:1 ] and the expansion , @xmath714 with @xmath715 together with the taylor expansion of @xmath716 at @xmath717 , we write @xmath718 with @xmath719 from and , we have the expansion of @xmath490 near the origin , @xmath720 with @xmath721 from the definition of @xmath236 and @xmath226 ( see and ) , one can check that for @xmath708 , @xmath722 note from the asymptotic behavior of @xmath231 that @xmath723 for @xmath708 , which implies @xmath724 we then conclude @xmath725 - _ control for @xmath259 .
_ let us rewrite from the definition of @xmath665 , @xmath726 note from the definitions of @xmath236 and @xmath226 that @xmath727 from which and the leibniz rule , we write @xmath728 we aim at using the pointwise estimate to prove that for @xmath729 , @xmath730 and @xmath731 , @xmath732 which concludes the proof of . to prove , we distinguish in 3 cases : + - since @xmath733 , then @xmath734 .
although this is the simplest case , it gives us a basic idea to handle other cases . from , it is obvious to see that @xmath735 is uniformly bounded .
we write @xmath736 using , definition [ def : skset ] , @xmath737 and the fact that @xmath738 ( see ) , we estimate @xmath739 for the integral on the domain @xmath740 , let us write @xmath741 this concludes the proof of when @xmath734 . + - .
we first use the leibniz rule to write @xmath742 from which , @xmath743 we claim that for all @xmath744 and @xmath745 , @xmath746 which immediately follows for the case when @xmath747 . to prove , we proceed as for the case @xmath748 by splitting the integral in two parts as follows : @xmath749 where @xmath750 satisfy @xmath751 we now estimate @xmath752 .
+ - if @xmath753 is even , we take @xmath754 this gives @xmath755 using , we have the estimate @xmath756 - if @xmath753 is odd , we simply take @xmath757 which gives @xmath758 hence , @xmath759 recall from definition that for all even integer @xmath760 in the range @xmath761 , @xmath762 - if @xmath763 and @xmath764 , then @xmath765 - if @xmath766 , then @xmath767 because @xmath768 .
this follows @xmath769 hence , we obtain @xmath770 similarly , one prove that @xmath771 therefore , @xmath772 for @xmath773 .
this concludes the proof of as well as when @xmath747 .
+ - let us write from and , @xmath774 at this stage , we need to precise the decay of @xmath775 to archive the bound .
to do so , let us recall that @xmath140 is admissible of degree @xmath675 ( see lemma [ lemm : genlk ] ) and @xmath137 is homogeneous of degree @xmath776 ( see proposition [ prop:1 ] ) .
together with , we estimate @xmath777 let @xmath712 $ ] and @xmath778 .
we use the faa di bruno formula to write + @xmath779 for @xmath780 , we use to estimate @xmath781 from which , we have latexmath:[\[\label{est : psin1b0 } for @xmath740 , we use again to write for all @xmath783 , @xmath784 injecting and into , we arrive at @xmath785 where @xmath786 . arguing as for the proof of , we end up with @xmath787 for @xmath509 small enough .
this finishes the proof of as well as .
since the proof of follows exactly the same lines as for the proof of , we omit its proof here . inserting , and into and recalling from definition that @xmath788 , we arrive at @xmath789 \right\ } & \lesssim \frac{b_1}{\lambda^{4\bbbk - d + 2}}\left(\frac{{\mathscr{e}}_{2\bbbk}}{m^{2\delta } } + b_1^{l + ( 1 - \delta)(1 + \eta)}\sqrt{{\mathscr{e}}_{2\bbbk } } + b_1^{2l + 2(1 - \delta)(1 + \eta)}\right)\nonumber\\ & \qquad + \frac{1}{\lambda^{4\bbbk - d + 2}}\int { \mathscr{l}}^\bbbk q { \mathscr{l}}^\bbbk(\partial_s \xi_l ) .
\label{eq : enerid4}\end{aligned}\ ] ] * step 8 : time oscillations . * in this step , we want to find the contribution of the last term in to the estimate .
let us write @xmath790\right \}\nonumber\\ & \quad + \frac{4\bbbk - d + 2}{\lambda^{4\bbbk - d + 2 } } \frac{\lambda_s}{\lambda } \left[\int { \mathscr{l}}^\bbbk q { \mathscr{l}}^\bbbk \xi_l + \frac{1}{2}\int |{\mathscr{l}}^\bbbk \xi_l|^2\right]\nonumber\\ & \qquad - \frac{1}{\lambda^{4\bbbk - d + 2}}\int { \mathscr{l}}^\bbbk({\partial_s}q - { \partial_s}\xi_l ) { \mathscr{l}}^\bbbk \xi_l.\label{id : aextract}\end{aligned}\ ] ] from and , we have @xmath791 this follows @xmath792 since @xmath793 , we then write @xmath794\right\ } = \frac{d}{dt}\left(\frac{{\mathscr{e}}_{2\bbbk}}{\lambda^{4\bbbk - d}}\mathcal{o}(b_1^{\eta(1 - \delta ) } ) \right).\ ] ] noting from that @xmath795 , this gives @xmath796\right| \lesssim
b_1 b_1^{\eta(1 - \delta)}{\mathscr{e}}_{2\bbbk}.\ ] ] for the last term in , we use equation and the decomposition to write @xmath797 \nonumber\\ & + \int { \mathscr{l}}^{\bbbk}\big[-\tilde \psi_b - \widetilde{mod } + { \mathcal{h}}(q ) + { \mathcal{n}}(q)\big ] { \mathscr{l}}^{\bbbk}\xi_l.\label{eq : a3exp}\end{aligned}\ ] ] using , the admissibility of @xmath798 and the fact that @xmath799 if @xmath800 , we estimate @xmath801| ^2\\ & \lesssim b_1^{-2(1 - \delta)}{\mathscr{e}}_{2\bbbk}\int_{y \geq b_1}y^{2(2l - \gamma - 2(\bbbk + 1 ) ) } y^{d-1}dy\\ & \quad \lesssim b_1^{-2(1 - \delta)}{\mathscr{e}}_{2\bbbk}b_1^{(4 - 2\delta)(1 + \eta ) } \lesssim b_1 ^ 2b_1^{2\eta(1 - \delta)}{\mathscr{e}}_{2\bbbk}.\end{aligned}\ ] ] from which we obtain @xmath802 similarly , we have the estimate @xmath803 hence , using and , we get @xmath804 from , and , we have @xmath805 in the same manner , we have the estimate @xmath806 from which together and , we get the bound @xmath807 collecting these final bounds into yields @xmath808 substituting , , and into concludes the proof of as well as proposition [ prop : e2k ] .
we give the proof of proposition [ prop : redu ] in this subsection in order to complete the proof of theorem [ theo:1 ] .
note that this section corresponds to section 6.1 of @xcite . here
we follow exactly the same lines as in @xcite and no new ideas are needed .
we divide the proof into 2 parts : - _ part 1 : reduction to a finite dimensional problem . _
assume that for a given @xmath809 large and an initial time @xmath513 large , we have @xmath545 for all @xmath546 $ ] for some @xmath548 . by using , , and
, we derive new bounds on @xmath810 , @xmath811 for @xmath812 and @xmath813 for @xmath814 , which are better than those defining @xmath521 ( see definition [ def : skset ] ) .
it then remains to control @xmath815 .
this means that the problem is reduced to the control of a finite dimensional function @xmath815 and then get the conclusion @xmath123 of proposition [ prop : redu ] .
- _ part 2 : transverse crossing .
_ we aim at proving that if @xmath815 touches @xmath816 at @xmath817 , it actually leaves @xmath818 at @xmath819 for @xmath548 , provided that @xmath527 is large enough .
we then get the conclusion @xmath133 of proposition [ prop : redu ] . [
[ bullet - reduction - to - a - finite - dimensional - problem . ] ] @xmath820 reduction to a finite dimensional problem .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we give the proof of item @xmath123 of proposition [ prop : redu ] in this part .
given @xmath809 , @xmath513 and the initial data at @xmath508 as in definition [ def:1 ] , we assume for all @xmath546 $ ] , @xmath545 for some @xmath548 . we claim that for all @xmath546 $ ] , @xmath821 once these estimates are proved , it immediately follows from definition [ def : skset ] of @xmath474 that if @xmath822 , then @xmath823 must be in @xmath824 , which concludes the proof of item @xmath123 of proposition [ prop : redu ] . + before going to the proof of - , let us compute explicitly the scaling parameter @xmath558 .
to do so , let us note from and the a priori bound on @xmath825 given in definition [ def : skset ] that @xmath826 using yields @xmath827 from which we write @xmath828 we now integrate by using the initial data value @xmath829 to get @xmath830 \quad \text{for}\;\ ; s_0 \gg 1.\ ] ] this implies that @xmath831 - _ improved control of @xmath189 _ : we aim at using to derive the improved bound . to do so , we inject the bound of @xmath189 given in definition [ def : skset ] into the monotonicity formula and integrate in time by using @xmath829 : for all @xmath832 , @xmath833.\end{aligned}\ ] ] using , we estimate @xmath834 here we used the fact that the integral is divergent because @xmath835 = \frac{2\gamma l}{2\ell - \gamma } + { \mathcal{o}}_{l \to + \infty}(1 ) \gg -1.\ ] ] using again and the initial bound , we estimate @xmath836 for @xmath162 large enough . therefore , we obtain @xmath837 for @xmath838 large enough . this concludes the proof of .
+ - _ improved control of @xmath197 .
_ we can improve the control of @xmath197 by using the monotonicity formula .
we distinguish two cases : + - @xmath839 . from the bound of @xmath197 given in definition
[ def : skset ] and @xmath840 , we integrate in time @xmath841 by using @xmath829 to find that @xmath842\end{aligned}\ ] ] using the initial bound and , we estimate @xmath843 for @xmath527 large .
+ using and the identity @xmath844 we estimate @xmath845 similarly , thanks to the identity @xmath846 we obtain @xmath847 therefore , we deduce that @xmath848 for @xmath849 large , which yields the improved bound for @xmath850 .
+ - @xmath851 .
proceeding as the previous case , we arrive at @xmath852}}{\lambda(\tau)^{4 m - d}}d\tau\right].\end{aligned}\ ] ] from the identity @xmath853 together with , we estimate @xmath854}}{\lambda(\tau)^{4 m - d}}d\tau\\ & \quad \lesssim s^{-\frac{\ell(4 m - d)}{2\ell - \gamma}}\int_{s_0}^s\tau^{\frac{\ell(4 m - d)}{2\ell - \gamma } -\left[2 m - 2\hbar - 1 + 2(1 - \delta ) - \left(c + \frac k2\right)\eta\right ] } d\tau\\ & \qquad \lesssim s^{-\left[2(m - \hbar - 1 ) + 2(1 - \delta ) - \left(c + \frac k2\right)\eta\right]}\leq \frac 14 s^{-\left[2(m - \hbar - 1 ) + 2(1 - \delta ) - k\eta\right]}.\end{aligned}\ ] ] using , and the initial bound , we derive @xmath855 } \\ & \quad \leq \frac 14 s^{-\left[2(m - \hbar - 1 ) + 2(1 - \delta ) - k\eta\right]}.\end{aligned}\ ] ] this concludes the proof of .
+ - _ control of the stable modes @xmath514 s .
_ we now end the control of the stable modes @xmath856 , in particular , we prove .
we first treat the case when @xmath857 .
let @xmath858 then from and , @xmath859 it follows from the improved modulation equation , @xmath860 this follows @xmath861 integrating this identity in time from @xmath527 and recalling that @xmath829 yields @xmath862 using , @xmath840 , the initial bounds and together with , we estimate @xmath863 and @xmath864 therefore , @xmath865 which concludes the proof of for @xmath857 .
now we will propagate this improvement that we found for the bound of @xmath576 to all @xmath514 for all @xmath866 . to do so we do a descending induction where the initialization is for @xmath867 .
let assume the bound @xmath868 for @xmath869 and let s prove it for @xmath363 .
indeed , from and the induction bound , we have @xmath870 which follows @xmath871 integrating this identity in time as for the case @xmath857 , we end - up with @xmath872 where we used the initial bound , and @xmath873 .
this concludes the proof of .
- _ control of the stable mode @xmath874 .
_ we recall from and that @xmath875 where @xmath512 diagonalize the matrix @xmath472 with spectrum . from , and
, we estimate for @xmath876 , @xmath877 from , and the improved bound , we have @xmath878 using the diagonalization , we obtain @xmath879 s using again yields the control of the stable mode @xmath874 : @xmath880 thus from the initial bound , @xmath881 which yields for @xmath882 large enough .
+ [ [ bullet - transverse - crossing . ] ] @xmath883 transverse crossing .
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we give the proof of item @xmath133 of proposition [ prop : redu ] in this part .
we compute from and at the exit time @xmath817 : @xmath884\right)_{\big|_{s= s_1}}\\ & \qquad = \left(s^{\eta(1 - \delta ) - 1 } \bigg[\sum_{k = 2}^\ell \left[\frac{k\gamma}{2k - \gamma } + \frac{\eta}{2}(1 - \delta)\right]{\mathcal{v}}_k^2(s ) + { \mathcal{o}}\left(\frac{1}{s^{\frac{3}{2}\eta(1 - \delta)}}\right)\bigg ] \right)_{\big|_{s= s_1}}\\ & \quad \qquad \geq \frac{1}{s_1}\left[c(d,\ell ) \sum_{k = 2}^\ell |s_1^{\frac{\eta}{2}(1 - \delta)}{\mathcal{v}}_k(s_1)|^2 + { \mathcal{o}}\left(\frac{1}{s_1^{\frac{\eta}{2}(1 - \delta)}}\right)\right]\\ & \qquad \qquad \geq \frac{1}{s_1}\left[c(d,\ell ) + { \mathcal{o}}\left(\frac{1}{s_1^{\frac{\eta}{2}(1 - \delta)}}\right)\right ] > 0,\end{aligned}\ ] ] where we used item @xmath123 of proposition [ prop : redu ] in the last step .
this completes the proof of proposition [ prop : redu ] .
we give in this section the coercivity estimates for the operator @xmath127 as well as the iterates of @xmath127 under some suitable orthogonality condition .
we first recall the standard hardy type inequalities for the class of radially symmetric functions , @xmath885 for simplicity , we write @xmath886 and @xmath887 we have the following : [ lemm : hardy ] let @xmath3 and @xmath888 , then + @xmath123 ( hardy near the origin ) @xmath889 @xmath133 ( hardy away from zero for the non - critical exponent ) let @xmath890 , then @xmath891 @xmath161 ( hardy away from zero for the critical exponent ) let @xmath892 , then @xmath893 @xmath168 ( general weighted hardy ) for any @xmath894 , @xmath895 be an integer and @xmath896 , @xmath897 the proof can be found in @xcite , lemma b.1 . from the hardy type inequalities ,
we derive the following coercivity of @xmath226 : [ lemm : coerast ] let @xmath898 , there exists @xmath899 such that for all @xmath888 : @xmath900 we proceed into two steps : _ - step 1 : subcoercive estimate for @xmath226 : _ we first prove the following subcoercive bound for @xmath226 : for @xmath901 and @xmath898 , @xmath902 from the definition of @xmath226 and the asymptotic of @xmath231 given in , we use an integration by parts to estimate near the origin : @xmath903 away from the origin , we use to estimate : @xmath904 we make the change of variable @xmath905 and use the hardy inequality given in part @xmath133 of lemma [ lemm : hardy ] to write @xmath906 gathering the above bounds together with the trivial bound from , @xmath907 yields the subcoercivity .
+ _ - step 2 : coercivity of @xmath226 : _ we now argue by contradiction to show the coercivity of @xmath226 .
assume that does not hold .
up to a renormalization , we consider the sequence @xmath908 with @xmath909 this implies from , @xmath910 from , the sequence @xmath911 is bounded in @xmath912 .
hence , from a standard diagonal extraction argument , there exists @xmath913 such that up to a subsequence , @xmath914 and from the local compactness of one dimensional sobolev embeddings : @xmath915 this implies from and , @xmath916 which means that @xmath917 . on the other hand , from and the lower semi continuity of norms for the weak topology , we have @xmath918 hence , @xmath919 since @xmath920 near the origin
, we have @xmath921 which contradicts the a priori regularity of @xmath922 given in .
this concludes the proof of lemma [ lemm : coerast ] .
we also need the following subcoercivity of @xmath236 .
[ lemm : subcoera ] let @xmath923 and @xmath901 such that @xmath924 , where @xmath925 $ ] is defined by , there holds : @xmath926.\label{eq : subcoera}\end{aligned}\ ] ] assume in addition that @xmath927 where @xmath180 is defined in , we have @xmath928 the proof is very similar to the proof of lemma [ lemm : coerast ] .
we proceed into two steps .
the first step is to derive the subcoercive estimate . in the second step
, we use a compactness argument to show the coercivity of @xmath236 under a suitable condition . _ - step 1 : subcoercive estimate for @xmath236 .
_ from the definition of @xmath236 and the asymptotic of @xmath231 given in , we estimate near the origin @xmath929 away from the origin , we estimate from : @xmath930 we make the change of variable @xmath931 . from the assumption @xmath932
, we use the hardy inequality given in part @xmath133 of lemma [ lemm : hardy ] to write @xmath933 note also that we have the trivial bound from , @xmath934 the collection of the above bounds yields the subcoercivity .
+ _ - step 2 : coercivity of @xmath236 : _ arguing as the proof of , we end up with the existence of @xmath917 such that @xmath935 hence , from the definition of @xmath236 , we have @xmath936 if @xmath937 , we use the orthogonality condition to deduce that @xmath938 thus , @xmath939 .
if @xmath940 , we use the fact that @xmath941 as @xmath337 to estimate @xmath942 which contradicts with the regularity of @xmath922 .
this concludes the proof of lemma [ lemm : subcoera ] . from the coercivities of @xmath236 and @xmath226
, we claim the following coercivity for @xmath127 : [ lemm : coerl ] let @xmath284 , @xmath901 and @xmath943 large enough , then there exists @xmath944 such that for all @xmath888 satisfying the orthogonality @xmath945 where @xmath180 is defined by , @xmath333 is given in , there holds : @xmath946 and @xmath947 we proceed into two steps : + - _ step 1 : subcoercivity of @xmath127 .
_ we apply lemma [ lemm : coerast ] to @xmath948 with @xmath949 and note that @xmath950 to write @xmath951 applying lemma [ lemm : subcoera ] to @xmath245 with @xmath952 and noting that the condition @xmath953 is always satisfied ( if not , we have @xmath954 ) , we have @xmath955.\end{aligned}\ ] ] we apply again lemma [ lemm : subcoera ] to @xmath956 with @xmath957 to estimate @xmath958.\end{aligned}\ ] ] injecting these bounds into yields the subcoercive estimate for @xmath127 , @xmath959 .
\label{eq : subcoerl}\end{aligned}\ ] ] - _ step 2 : coercivity of @xmath127 .
_ we argue by contradiction .
assume that does not hold .
up to a renormalization , there exists a a sequence of functions @xmath908 such that @xmath960 this implies from , @xmath961 from , the sequence @xmath911 is bounded in @xmath962 .
hence , from a standard diagonal extraction argument , there exists @xmath963 such that up to a subsequence , @xmath964 and from the local compactness of one dimensional sobolev embeddings : @xmath965 and @xmath966 this implies from and , @xmath967 which means that @xmath917 . on the other hand , from and the lower semi continuity of norms for the weak topology , we deduce that @xmath922 is a non trivial function in the kernel of @xmath127 , namely that @xmath968 which follows @xmath969 where @xmath970 and @xmath971 two real numbers .
+ from and the lower semicontinuity , we have @xmath972 recall from that @xmath973 as @xmath974 .
this yields the estimate @xmath975 hence , @xmath976 . from , we have @xmath977 as @xmath337 .
if @xmath978 , we have @xmath979 hence , @xmath939 .
if @xmath980 , we use the orthogonality condition to deduce @xmath981 which yields @xmath939 , hence @xmath982 .
the contradiction then follows and the coercivity is proved .
the estimate simply follows from and .
this finished the proof of lemma [ lemm : coerl ] .
we are now in a position to prove the coercivity of @xmath983 under a suitable orthogonality condition .
we claim the following : [ lemm : coelk ] let @xmath284 and @xmath943 large enough , then there exists @xmath944 such that for all @xmath888 satisfying the orthogonal condition @xmath984 where @xmath333 is defined as in , there holds : @xmath985 we argue by induction on @xmath363 . for @xmath748 , we apply lemma [ lemm : coerast ] to @xmath948 with @xmath986 and @xmath987 , then lemma [ lemm : subcoera ] to @xmath245 with @xmath988 and @xmath989 to write @xmath990 note that we had to use the orthogonality condition @xmath991 when @xmath992 . in fact , the case @xmath992 only happens when @xmath60 . in this case , the condition @xmath993 is fulfilled when applying lemma [ lemm : coerast ] with @xmath988 and @xmath989 .
we now assume the claim for @xmath994 and prove it for @xmath995 .
we have the orthogonality condition @xmath996 let @xmath997 , then we have @xmath998 by induction hypothesis , we write @xmath999 note that we have the orthogonality condition @xmath1000 when @xmath1001 .
the case @xmath1002 implies that @xmath1003 hence , we use the coercivity bound to derive @xmath1004 which concludes the proof of lemma [ lemm : coelk ] .
we derive in this section interpolation bounds on @xmath186 which are the consequence of the coercivity property given in lemma [ lemm : coelk ] .
we have the following : [ lemm : interbounds ] @xmath1005 + @xmath123 weighted bounds for @xmath1006 : for @xmath1007 , @xmath1008 @xmath133 development near the origin : @xmath1009 with bounds @xmath1010
@xmath1011 @xmath161 bounds near the origin for @xmath1006 and @xmath1012 : for @xmath1013 , latexmath:[\ ] ] for @xmath1095 , we use and to estimate @xmath1096 for @xmath1097 , we apply with @xmath1098 and note from that @xmath1099 which yields @xmath1100 thus , @xmath1101v\right|^2 \lesssim \frac{b_1 ^ 2}{\lambda^{4\bbbk - d + 2 } } { \mathscr{e}}_{2\bbbk}.\ ] ] similarly , we use to get the estimate @xmath1102v\right|^2 \lesssim \frac{b_1 ^ 2}{\lambda^{4\bbbk - d + 2}}{\mathscr{e}}_{2\bbbk}.\ ] ] this concludes the proof of .
m. struwe . geometric evolution problems . in _ nonlinear partial differential equations in differential geometry ( park city , ut , 1992 ) _ , volume 2 of _ ias / park city math .
_ , pages 257339 .
soc . , providence , ri , 1996 . | we consider the energy - supercritical harmonic heat flow from @xmath0 into the @xmath1-sphere @xmath2 with @xmath3 . under an additional assumption of 1-corotational symmetry
, the problem reduces to the one dimensional semilinear heat equation @xmath4 we construct for this equation a family of @xmath5 solutions which blow up in finite time via concentration of the universal profile @xmath6 where @xmath7 is the stationary solution of the equation and the speed is given by the quantized rates @xmath8.\ ] ] the construction relies on two arguments : the reduction of the problem to a finite - dimensional one thanks to a robust universal energy method and modulation techniques developed by merle , raphal and rodnianski @xcite for the energy supercritical nonlinear schrdinger equation and by raphal and schweyer @xcite for the energy critical harmonic heat flow , then we proceed by contradiction to solve the finite - dimensional problem and conclude using the brouwer fixed point theorem
. moreover , our constructed solutions are in fact @xmath9 codimension stable under perturbations of the initial data . as a consequence
, the case @xmath10 corresponds to a stable type ii blowup regime .
t. ghoul@xmath11 , s. ibrahim@xmath12 and v. t. nguyen@xmath11 |
_ the level - set methods _ for the image segmentation have been studied and applied during the last two decades .
the level - set method applied in the image segmentation is typically an iterative method .
the segmentation starts with an initial curve @xmath0 representing an initial guess for the segmented object and it is evolved towards the segmented object by means of a suitable geometric law taking into account the direction to the segmented object and also the curvature of evolved curves . loosely speaking ,
the better the initial guess is , the better and faster the segmentation process is .
this is profitable for processing of time sequences where the final segmentation of one frame may serve as the initial guess for the next frame .
we refer the reader to a wide range of literature on this topic like caselles et al .
@xcite , handloviov et al .
@xcite , osher , paragios @xcite or sethian and references therein . in comparison to parametric models studied e.g. in bene et al .
@xcite and kass et al .
@xcite the level - set methods can handle topological changes and so one initial curve can split and segment more separate objects .
very different segmentation methods are _ the graph - cuts methods _
boykov et al .
@xcite ) which are based on graph theory and algorithms for finding _ the minimal cuts _ respectively _ the maximal flows_. it is also an optimization problem but algorithms for the minimal - cuts / maximal - flows find globally the best solution which is usually not the case of the level - set methods .
these algorithms are not iterative and they do not require initial curves . instead of it , they need initial seeds - one or more points or lines in the interior and exterior of the segmented object . in this article , we show how to incorporate the a - priori information in form of the initial seeds to the level - set method .
our main purpose is to propose a new constrained level - set method which can be applied to the image segmentation problems . in comparison to the classical level - set methods ( c.f .
@xcite ) , our method allows an expert to prescribe an a - priori information by marking parts which are surely inside or outside the segmented region .
such a possibility can be helpful for an expert conducting an interactive segmentation of medical data .
a numerical approximation scheme is based on the complementary - finite volumes method developed by handloviov et al . in @xcite
combined with the projected successive over - relaxation method for solving constrained problems proposed by mangasarian in @xcite and elliott , ockendon in @xcite .
the advantage of the constrained level - set method is demonstrated on several artificial images as well as on cardiac mri data .
the paper is organized as follows . in section [
sect : level - set - segmentation ] we show the common level - set method for the image segmentation together with the numerical scheme and successive over - relaxation ( sor ) method . section [ sect : constrained - level - set ] explains the constrained level - set method with appropriate numerical scheme . as a solver for the linear complementarity problem with range bounds we adopt the projected sor method .
comparison with the common level - set method and contributions of the constrained level - set method are demonstrated in section [ sect : results ] .
we assume an image is represented by the greyscale image function @xmath1 $ ] defined on a two dimensional rectangle @xmath2\times[0,l_2]$ ] . a common idea how to segment an object in the image is to start from a closed , embedded and smooth initial curve @xmath0 approximating the shape of the object and let it evolve towards the exact boundary of the object . to this end , we construct a family of evolving curves @xmath3 with the property that @xmath3 converges to the boundary of a segmented object as @xmath4 goes to infinity .
there are many ways how to construct such a flow of planar curves . among them
we will focus our attention to the flow of curves proposed in the active contour model ( c.f .
caselles et al . in @xcite ,
kichenassamy et al . in @xcite ) .
a problem of finding a boundary of an object in the image can be reformulated as a problem of construction of planar curves on which the gradient @xmath5 of the image intensity function @xmath6 is large .
assuming @xmath3 is a @xmath7 smooth curve we can evaluate the unit tangent vector @xmath8 and outer unit normal vector @xmath9 .
each point @xmath10 is evolved in the normal direction with the normal velocity @xmath11 .
although the velocity vector @xmath12 can be decomposed into its tangential and normal parts , it should be noted that only the motion in the normal direction has impact on the shape of the closed curve @xmath3 . following the active contour model ( c.f .
@xcite ) , mikula and evovi @xcite considered a generalized form of the normal velocity : @xmath13 where @xmath14 is the curvature of @xmath3 and @xmath15 . here
@xmath16 is a smooth edge detector function @xmath17 such that @xmath18 and @xmath19 , for some constant @xmath20 .
a typical example is the function @xmath21 where @xmath22 is the contrast parameter .
notice that , for a given smooth intensity function @xmath6 , the vector field @xmath23 has an important geometric property as it points towards edges in the image where the norm of the gradient @xmath5 is large ( c.f .
@xcite ) . notice that a possible lack of smoothness of @xmath6 ( e.g. due to a noise ) can be overcome by taking the convolution of @xmath6 with a smooth gaussian mollifier @xmath24 with the dispersion @xmath25 ( see @xcite ) .
the term @xmath26 pushes the evolved curve @xmath3 towards the edge of the image @xmath6 ( c.f .
the effect of the curvature term @xmath27 consists in smoothing the segmented curve by means of minimization of its total length .
this property makes the segmentation model robust for application even in the case of a noisy image .
notice that the term @xmath28 slows down the normal velocity in the vicinity of an edges of @xmath6 ( c.f .
@xcite ) . in the level - set method @xmath3
is given implicitly as @xmath29 where @xmath30 is a real valued smooth function defined on @xmath31 such that @xmath32 for all @xmath33 belonging to the interior of @xmath3 and @xmath34 for all @xmath33 belonging to the exterior of a jordan curve @xmath3 .
following derivation from @xcite , the level - set formulation of ( [ eqn : normal - velocity ] ) can be stated in terms of a solution @xmath30 to the initial - boundary value problem @xmath35 , \\ & { \ensuremath{\partial^{}_{\nu } { u } } } = 0\ & { \rm on}\ \partial \omega , \\ \label{eq : level - set - segmentation-3 } & u \mid_{t=0 } = u_{ini}\ & { \rm in}\ \omega,\end{aligned}\ ] ] where @xmath36 is the initial level - set function corresponding to the initial curve @xmath0 , @xmath37 and @xmath38 is the outer normal unit vector of the boundary @xmath39 of a computational domain @xmath31 . the quantity @xmath40 should be equal to @xmath41
. however , for practical purposes , it is regularized by means of the tichonov regularization , i.e. @xmath42 where @xmath43 is a small regularizing parameter .
we discretize the initial - boundary value problem ( [ eq : level - set - segmentation])([eq : level - set - segmentation-3 ] ) by means of the method of complementary finite volumes developed by handloviov et al .
@xcite in the context of a class of level - set equations arising in the image processing .
let @xmath44 be a time step for time discretization .
let @xmath45 be spatial discretization steps such that @xmath46 for some @xmath47 .
we define a numerical grid , its closure and its boundary as @xmath48 for a function @xmath49;{\ensuremath{\mathbbm{r}^{}}}\right)$ ] we define its piecewise constant approximation on @xmath50 at the time @xmath51 as a grid function defined by @xmath52 .
we furthermore introduce a dual mesh @xmath53 defined as @xmath54 \times \left [ \left(j-\frac{1}{2}\right ) h_2 , \left(j+\frac{1}{2}\right ) h_2 \right ] \mid \right .
\nonumber \\ &
& i = 1 \cdots n_1 - 1 , j = 1 \cdots n_2 - 1 \bigg\}. \label{dual - fvm - mesh}\end{aligned}\ ] ] for @xmath55 , @xmath56 , @xmath57 and @xmath58 fixed , we consider a finite volume @xmath59 of the dual mesh @xmath53 .
we denote its interior by @xmath60 , its boundary by @xmath61 and let @xmath62 be the volume of @xmath60 ( see fig . [
fig : fin - volume ] ) .
we also denote the set of all neighboring volumes of a finite volume @xmath59 by @xmath63 . for all inner finite volumes @xmath59 of the dual mesh @xmath53 ,
the boundary @xmath61 consists of four linear segments .
we denote them as @xmath64 for @xmath65 .
it means that @xmath64 is a boundary between the finite volumes @xmath59 and @xmath66 .
let @xmath67 be the length of this part of @xmath61 . dividing equation ( [ eq : level - set - segmentation ] ) by the term @xmath40 , integrating it over the interior @xmath60 of a finite volume @xmath59 we obtain the equation : @xmath68 applying the euler backward in time discretization of @xmath69 we end up with the following time semi - discretization of ( [ eq : level - set - segmentation ] ) : @xmath70 next , after long and tedious but straightforward calculations and evaluations described in appendix we are in a position to formulate full time - space discretization of the level set equation ( [ eq : level - set - segmentation ] ) .
it can be rewritten in a form of the following system of linear equations : the elements @xmath71 are defined in appendix ( see ( [ def : a-1])([def : a-6 ] ) and ( [ def : a-7])([def : a-8 ] ) ) . at each time level @xmath72
we can represent a solution @xmath73 by a long vector @xmath74 by mapping the node @xmath75 of the two dimensional spatial domain to the one - dimensional vector , i.e. @xmath76 for @xmath77 and @xmath78 and setting @xmath79 .
system of linear equations ( [ eq : algebraic - segmentation ] ) can be then rewritten as a linear equation @xmath80 for the solution vector @xmath74 .
the dimension of the square matrix @xmath81 as well as of the vectors @xmath82 and @xmath83 is @xmath84 .
we can solve the aforementioned linear problem ( [ eq : linear - problem ] ) by means of the _ successive over - relaxation method _
( sor ) in this linear equation solver we iteratively solve @xmath85 for @xmath86 .
we repeat the iterative process for @xmath87 until the prescribed tolerance level for the difference @xmath88 is achieved . the iterative sor scheme ( [ def : sor ] ) can be written in the operator form : @xmath89 where @xmath90 . here
@xmath91 stand for the diagonal , sub - diagonal and upper - diagonal parts of the matrix @xmath81 , respectively .
the parameter @xmath92 is the so - called relaxation parameter .
it can be used in order to speed up the convergence of @xmath93 by minimizing the spectral norm of the operator @xmath94 .
in this section we introduce our new constrained level - set method for image segmentation .
we suppose that there are two disjoint subdomains @xmath95 and @xmath96 of the domain @xmath31 such that @xmath95 is a subset of the interior of the segmented object and @xmath96 lies outside the segmented object .
furthermore , we suppose that there are two prescribed functions @xmath97 with the property such that @xmath98 in @xmath31 and @xmath99 in @xmath95 and @xmath100 in @xmath101 and @xmath102 in @xmath96 and @xmath103 in @xmath104 .
notice that any function @xmath30 fulfilling @xmath105 in @xmath31 must be negative in @xmath95 and positive in @xmath96 .
its zero level - set contains the set @xmath95 in its interior and @xmath96 in its exterior .
our purpose is to construct a solution @xmath106 such that it satisfies the level set equation ( [ eq : level - set - segmentation ] ) in the open region where @xmath107 .
moreover , we require that @xmath108 for all @xmath109 $ ] .
the reason why to prescribe range bounds on the level - set function @xmath30 is to keep the set @xmath95 inside and @xmath96 outside the zero level set @xmath110 approaching the boundary of a segmented object when @xmath111 . to this end
we consider the following partial differential inequality problem : @xmath112,\ s.t.\ w(x ) < u(x , t ) < v(x ) , \nonumber \\ u_t & \ge & q { \ensuremath { \nabla \cdot \left ( g^0 \frac{{\ensuremath { \nabla } } u}{q } \right)}}\ \hbox{for}\ ( x , t)\in\omega \times ( 0 , t],\
s.t.\ u(x , t)=w(x ) , \label{eq : constrained - level - set - segmentation } \\ u_t & \le & q { \ensuremath { \nabla \cdot \left ( g^0 \frac{{\ensuremath { \nabla } } u}{q } \right)}}\ \hbox{for}\ ( x , t)\in\omega \times ( 0 , t],\
s.t.\ u(x , t ) = v(x ) , \nonumber\end{aligned}\ ] ] @xmath113 it can be also viewed as the following variational inequality problem : find a solution @xmath114 where @xmath115 such that @xmath116 where @xmath117 is the cone : @xmath118 ( c.f . brezis ( * ? ? ?
* chapter 2 ) ) .
the operator @xmath119 is defined by @xmath120 and @xmath121 is the inner product in @xmath122 , i.e. @xmath123 the discretization of ( [ eq : constrained - level - set - segmentation])([eq : constrained - level - set - segmentation-3 ] ) follows exactly from the discretization of the level set equation ( [ eq : level - set - segmentation])([eq : level - set - segmentation-3 ] ) when taking into account the range bounds constraints @xmath124 for a.e . @xmath125 . at each time step
we have to construct a solution @xmath82 to the following linear complementarity problem : @xmath126 for @xmath86 . in order to solve the linear complementarity problem ( [ eq : linear - problem - inequality ] ) we make use of the so - called projected sor method ( psor ) @xcite adopted for our problem . for each index @xmath86 , we repeat the following up - dating of the vector @xmath127 : @xmath128 until the prescribed tolerance level for the difference @xmath88 is achieved .
suppose that the matrix @xmath81 has a positive diagonal , i.e. @xmath129 for each @xmath130 .
assume that @xmath131 as @xmath132 .
hence @xmath133 .
it means @xmath134_i , w_i\ } , v_i \}. \label{eqa}\ ] ] clearly , @xmath135 for each index @xmath130 .
if the strict inequality @xmath136 holds for some index @xmath130 , then @xmath137_i.\ ] ] it means that @xmath138 on the other hand , if @xmath139 then @xmath140 . according to ( [ eqa ] )
we conclude that @xmath141_i.\ ] ] since @xmath129 we obtain @xmath142 analogously , if @xmath143 for some index @xmath130 then it follows from ( [ eqa ] ) that @xmath144_i.\ ] ] thus @xmath145 hence the vector @xmath82 is a solution to the linear complementarity problem ( [ eq : linear - problem - inequality ] ) .
in this section , we present experimental results obtained by the constrained level - set method .
we first demonstrate the effect of the obstacle constraints applied to artificial images .
[ fig : no - obstacle ] a ) ( left ) shows an image we want to segment .
there are two rectangles with centers at points @xmath146 and @xmath147 . the width of the rectangles is @xmath148 and the height is @xmath149 .
the thickness of the rectangles edges is @xmath150 . on the inner edge of each rectangle
, there is a thin hole .
the image intensity function is defined on the domain @xmath151 .
the initial curve is a circle with the radius @xmath152 centered exactly between the rectangles .
its signed distance function ( taken as the initial level - set function @xmath36 ) is depicted on fig .
[ fig : no - obstacle ] a ) ( right ) .
the numerical mesh consisted of @xmath153 grid points .
if the thin holes in the rectangles should be taken into account we may want to segment only the rectangles edges .
it can be achieved by setting the regularizing parameter @xmath154 . in @xcite mikula and sarti interpreted the regularization parameter @xmath155 as a parameter by which we can control `` convexity '' of the final segmentation curve . the result taken at time @xmath156
is depicted at fig .
[ fig : no - obstacle ] b ) . on the other hand ,
if small holes are present by mistake , we may want the segmentation curve to fill them .
it may be achieved by setting @xmath157 . however , a convex hull of both rectangles is segmented as we can see on the fig .
[ fig : no - obstacle ] c ) .
the segmentation we aim to can be achieved by prescribing one constraint guaranteeing that the part of the image between the rectangles must be outside the segmentation domain .
we construct the constraint by putting a red bar placed on between the rectangles on the fig .
[ fig : one - obstacle ] a ) .
its width equals @xmath150 and the height is set to @xmath158 .
the constraint function @xmath159 is positive on the red bar region and negative everywhere else .
we again set @xmath157 . the result taken at time
@xmath160 is shown on fig .
[ fig : one - obstacle ] b ) . in our last synthetic example shown in fig . [
fig : two - obstacles ] we place one more constraint inside the left rectangle .
the segmentation poses features of both experiments on the figures [ fig : no - obstacle ] and [ fig : one - obstacle ] .
it shows that the constrained level - set method is capable of controlling which part of the image we want to segment . in the remaining two figures we present a real application of the constrained level - set method to segmentation of cardiac mri data .
first we want to segment both , the left and the right ventricle of a heart . the usual level - set method for the image segmentation ( c.f .
@xcite ) is not capable to separate them . in figure
[ fig : cardiac - mri-1 ] we present a ) the initial set - up of the constraint barriers ; b ) segmentation without obstacles involved ; c ) segmentation with constraints separating both ventricles well . figure [ fig : cardiac - mri-2 ] demonstrates more complex segmentation result in which we separated the left ventricle , septum and the right ventricle together with pericardial fat .
we proposed a new constrained level - set method for the image segmentation .
the method gives a possibility to an expert to prescribe an additional information concerning the expected shape of the segmented object . in this method
, we may preset fixed constraints telling us which parts of the segmented object should be inevitably inside the segmented region and some parts must remain outside .
the complementary finite - volume method was used for the numerical approximation of the level - set equation .
it was combined with the projected successive over relaxation method for solving the corresponding parabolic variational inequality problem .
we demonstrated the difference between the new and the usual level - set method on several artificial images as well as on data from the magnetic resonance .
in this section we show details of derivation of the full time - space discretization of ( [ eq : level - set - segmentation ] ) . as usual in the finite volume methods , we apply the stokes theorem to the right hand side of ( [ eq : level - set - segmentation ] ) .
we obtain @xmath161 where @xmath38 is the outer unit normal of @xmath61 .
we have to approximate the `` fluxes '' @xmath162 and the `` capacity '' @xmath163 numerically .
we assume that the `` fluxes '' are constant on each segment @xmath64 of the finite volume boundary @xmath61 .
it means that @xmath164 first we evaluate the norm of smoothed gradient of the image intensity function @xmath165 . here
@xmath166 is the gaussian mollifier .
we refer the reader to @xcite for details how to effectively calculate the convolution @xmath167 .
then @xmath168 on @xmath169 and we approximate @xmath170 on the finite volume edges as follows : @xmath171 denote @xmath172 .
the approximation of @xmath173 in the direction of the vector @xmath174 is obvious : @xmath175 in order to calculate remaining coordinates of @xmath173 which are perpendicular to @xmath174 we need to know the value @xmath176 at the ends of @xmath67 ( corners of the finite volume @xmath59 ) .
they can be only approximated using the value of @xmath176 at the neighboring volumes : @xmath177 where @xmath178 .
hence we obtain the approximation of @xmath173 in the direction perpendicular to @xmath174 in the form @xmath179 having calculated approximation of @xmath173 we can approximate the term @xmath180 as follows : @xmath181 where @xmath182 .
the `` capacity '' term @xmath183 and the function @xmath176 are approximated by a constant value on the finite volume @xmath59 .
taking the averaged value @xmath184 yields the approximation @xmath185 it follows from ( [ stokes ] ) that @xmath186 notice that @xmath187 ( see fig . [ fig : fin - volume ] ) .
in such a regular grid , one coordinate of @xmath38 is always vanishing .
it cancels one coordinate of @xmath188 in the inner product @xmath189 .
recall that @xmath67 attains only the values @xmath190 or @xmath191 .
we have @xmath192 finally , equation ( [ eq : constrained - level - set - segmentation ] ) is approximated as @xmath193 let us denote @xmath194 to approximate the neumann boundary condition @xmath195 on @xmath39 where @xmath38 is the outer unit normal of the domain @xmath31 we set @xmath196 which yields @xmath197
this work was partially supported by the grant lc06052 of ncmm research center and research direction project msm6840770010 of the ministry of education of the czech republic , the cvut student grant agency project no . 283 ohk4 - 009/10 p3913 and vega project 1/0747/12 .
99 , m. kimura , p. pau , d. evovi , t. tsujikawa , s. yazaki , application of a curvature adjusted method in image segmentation , bulletin of inst . of mathematics , academia sinica , new series * 3 * ( 2008 ) , 509524 . , a. sarti , parallel co - volume subjective surface method for 3d medical image segmentation , in : parametric and geometric deformable models : an application in biomaterials and medical imagery , volume - ii , springer publishers , ( eds . jasjit s. suri and aly farag ) , 2007 , pp .
123 - 160 . | we propose a new constrained level - set method for semi - automatic image segmentation .
the method allows to specify which parts of the image lie inside respectively outside the segmented objects .
such an a - priori information can be expressed in terms of upper and lower constraints prescribed for the level - set function .
constraints have the same meaning as the initial seeds of the graph - cuts based methods for image segmentation .
a numerical approximation scheme is based on the complementary - finite volumes method combined with the projected successive over - relaxation method adopted for solving range bounds constrained linear complementarity problems .
the advantage of the constrained level - set method is demonstrated on several artificial images as well as on cardiac mri data .
linear complementarity , image processing , segmentation , level - set method , projected successive over - relaxation method , graph cuts 90c33 , 35k55 , 35k52 , 53c44 , 74s10 , 74g15 |
the class of type iin supernovae ( sne iin hereafter ) , whose narrow h lines indicate strong interaction with dense circumstellar material ( csm ) , have challenged our understanding of stellar evolution and death .
their dense h - rich csm , progenitor instability , and high initial masses inferred from various clues suggest a link to the class of luminous blue variables ( lbvs ) , which are not supposed to be anywhere near core collapse in the standard scenario of massive single - star evolution ( see @xcite for a general review ) .
among well - studied examples of sne iin , the explosion of sn 2009ip in mid-2012 ( note that its discovery in 2009 was deemed to be a sn impostor ) is one of the most interesting , with by far the best observational characterization of a directly detected progenitor among any sn in history ( even sn 1987a ) .
it had a ( presumably ) quiescent progenitor star candidate detected in archival _ hubble space telescope _
( _ hst _ ) images , with a luminosity that implied a very high initial mass of at least 50 - 60 @xmath2 @xcite .
this source also showed a series of outbursts in the few years before the sn @xcite that were reminiscent of both s dor outbursts and giant eruptions of lbvs . unlike any progenitor source so far , high - quality spectra of these precursor outbursts were obtained , with a detailed analysis before the sn indicating a strong similarity to lbvs @xcite . with a quiescent and very luminous progenitor , an s dor outburst , several bright but brief sn impostor eruptions , and progenitor spectra resembling lbvs
, sn 2009ip provides a strong link between lbvs and sne iin .
the repeating variable source at the position of sn 2009ip began to brighten again in mid-2012 , but this time things were different .
spectra of the fainter 2012a peak showed very broad p cygni profiles with velocities of 13,000 km s@xmath3 , suggesting that the event was a core - collapse sn and not another lbv outburst @xcite .
the subsequent and brighter 2012b event showed a high peak luminosity and a spectrum typical of sne iin with strong csm interaction .
the 2012 sn - like event has already been discussed extensively in the literature @xcite . in these publications and in discourse at meetings
, there was some uncertainty and controversy about whether the 2012 event was a true core - collapse sn , since ( 1 ) csm interaction can provide bright transients even from relatively low - energy explosions , ( 2 ) the initial sn was somewhat fainter than standard sne ii - p , and ( 3 ) the rich observational dataset for the progenitor presented mysteries that were not easily explained by any existing model .
these are , however , expressions of the challenge in understanding sne iin and csm interaction , rather than arguments against a core - collapse event . while it is difficult to prove definitively that the event was a core collapse because of the masking of csm interaction , a sn is the most straightforward explanation of the data .
@xcite showed that all available evidence was consistent with the core collapse sn explosion of a blue supergiant that encountered strong csm interaction .
moreover , both line - profile evolution @xcite and spectrapolarimetry @xcite show that the csm interaction was highly aspherical and probably disk - like , forcing the kinetic energy budget of the event to be @xmath110@xmath4 ergs .
@xcite also argued for a disk - like csm based on narrow line ratios .
@xcite showed that the source at + 2 yr was consistent with steady ongoing csm interaction with no additional outbursts , adding further evidence in favor of a core - collapse event .
sn 2009ip provides our clearest example of pre - sn instability that leads to eruptive pre - sn mass loss in the few years before explosions , which may be associated with the final nuclear burning sequences in the last years of a massive star s life @xcite .
alternative non - terminal models involving binary mergers and accretion were also proposed for the 2012 event @xcite , but these can not supply the required 10@xmath4 ergs of kinetic energy . in this paper , we are mainly concerned with the host galaxy environment around sn 2009ip .
a fundamental interesting mystery was that while progenitor detections pointed to a very massive unstable star , the location of sn 2009ip was in the remote outskirts of its spiral host , far away from obvious signs of recent star formation and young stellar populations @xcite .
it was located about 5 kpc from the center of its relatively small host spiral galaxy ngc 7259 . by extrapolating the apparent metallicity gradient measured in the inner @xmath11.5
kpc out to the 5 kpc radius of sn 2009ip , @xcite infer a mildly subsolar metallicity at the sn site of @xmath5 .
sn 2009ip s progenitor can therefore be compared with populations of massive stars observed in the milky way and large magellanic cloud ( lmc ) .
an interesting result is that despite their high luminosities and high inferred initial masses , lbvs in the milky way and lmc appear to be relatively isolated compared to expectations for their presumed role in stellar evolution .
@xcite demonstrated that lbvs selectively avoid clusters of o - type stars , especially early o - types that are their presumed progenitors .
more importantly , lbvs are more dispersed on the sky than wr stars ; this rules out the standard picture wherein lbvs are a transitional phase between massive o - type stars and wr stars .
instead , @xcite suggested that most lbvs may be the result of interacting binary evolution , getting rejuvenated by either mass transfer or mergers .
this would make them stand out as anomolously young compared to their surrounding populations .
in other words , _ they are evolved massive blue stragglers_. they may become even more isolated upon receiving a kick from their companion star s sn , although it is not yet clear if a kick is required to explain their environments . also relevant to this story
is that sne iin in general appear to be less correlated with bright h@xmath0 in their host galaxies than other types of sne @xcite . while this has been interpreted as signifying lower initial masses by those authors , that interpretation has been a topic of debate @xcite .
as described below , sn 2009ip seems to follow a similar trend of not having bright h@xmath0 nearby , despite having a detection of a very massive and luminous progenitor star .
.new _ hst _ wfc - uvis images of sn 2009ip , including st magnitudes for sn 2009ip and background 3@xmath6 upper limits [ cols="<,^,^,^,^,^",options="header " , ] [ tab : s99 ] compared to a roughly contemporaneous spectrum of sn 2009ip ( blue ) taken with the mmt in 2015
november , as well as a late - time day 905 spectrum of sn 2005ip ( black ) from @xcite .
sn 2005ip had stronger csm interaction at almost 3 yr after explosion , and a spectrum with much higher signal to noise showing the fainter emission features that are expected to accompany strong h@xmath0 emission .
this sn 2005ip spectrum has been scaled to match the same h@xmath0 flux as sn 2009ip.,width=307 ] 10@xmath7 ergs s@xmath3 @xmath8 arcsec@xmath9 , but there are negative residuals at a similar level , so these could be subtraction artifacts . there is no extended emission beyond 0@xmath104 ( 40 pc ) from the sn site brighter than @xmath10.2@xmath1110@xmath7 ergs s@xmath3 @xmath8 arcsec@xmath9.,width=307 ]
the new _ hst _ images show a clearly detected point source at the location of sn 2009ip . based on the measured offset from a nearby bright star in the field
, this source is coincident with the position of the 1999 progenitor to within 0.043 arcsec ( roughly 10% of a wfc3 pixel ) .
the late - time source has a profile consistent with the point spread function in all the broad filters .
figure [ fig : psf ] shows the result of subtracting a model point spread function ( psf ) from the f657n image of sn 2009ip .
the model psf was generated using the tiny tim program .
there is residual emission after the psf subtraction at a typical brightness level of 0.5@xmath1110@xmath7 ergs s@xmath3 @xmath8 arcsec@xmath9 located within 0@xmath104 ( 40 pc ) of the sn site .
this could indicate some faint extended h ii region emission from a fading star forming region .
however , there are negative residuals at a similar level , so these are also consistent with being simple subtraction artifacts due to an imperfect psf match .
there is no detected emission brighter than @xmath12@xmath1110@xmath12 ergs s@xmath3 @xmath8 arcsec@xmath9 beyond 40 pc from the sn site .
we also tried using a psf made from a bright star on the same image , with similar results .
psf subtraction of the broadband images ( not shown ) yielded similarly inconclusive results . in the continuum images ,
the residuals allow for the presence of a faint sn light echo or extended and very faint ob association that is 2 - 5% as bright as the sn point source , but as with f657n , the residuals are also consistent with subtraction artifacts from an imperfect psf match .
the f555w source is slightly fainter than the f606w progenitor , but consistent with no change in the continuum level within the uncertainty . in a preliminary report of ground - based photometry
, @xcite also reported that sn 2009ip continued to fade , and by a few months after our _ hst_images were taken , it was fainter than the progenitor . the faint emission detected in the broad filters could be dominated by continuum from an underlying star cluster , late - time csm interaction emission from the fading sn itself , or a combination of these two . in fact
, there must be some contribution from csm interaction emission , due to the strong and broad h@xmath0 detected in the late - time spectrum ( fig .
[ fig : spec ] ) .
@xcite showed that there was still strong csm interaction at around 2 yr after explosion .
we can therefore rule out the possibility that there is no change in the brightness of the progenitor candidate ( as would be the case if it was a chance alignment with an unrelated star ) , and so the progenitor must have faded by whatever amount is now contributed by csm interaction luminosity .
the possibility that a progenitor survived but is now obscured by dust is hard to rule out because it requires some arbitrary amount of extinction to compensate for csm interaction .
however , @xcite noted that there was no clear indication of new dust formation in the late - time data for sn 2009ip , so this would argue that the progenitor candidate has in fact faded .
figure [ fig : sed ] compares the observed spectral energy distribution ( sed ) in the _ hst _ photometry to the hypothetical emission from underlying young star clusters using examples from starburst99 models @xcite .
we show models with ages of 5 , 10 , and 20 myr scaled to the f275w point only ( assuming the the optical / red emission is dominated by the sn ) , as well as these same cluster model spectra with reddening to match both the f275w and the f814w flux .
clusters much older than 20 myr would have turn off masses too low ( less than 12 @xmath2 ) .
the associated cluster terminal - age main sequence masses ( @xmath13 ) , absolute @xmath14 mags , and total cluster stellar masses are listed in table [ tab : s99 ] . the corresponding absolute @xmath14 mag and total cluster mass corrected for extinction
are also shown , along with the corresponding @xmath15 value . if we assume that the csm interaction dominates the optical / red emission , the unreddened clusters allowed by the f275w photometry imply total clutser masses that are unreasonably low ( lower than the progenitor mass of an individual star ; this is because they have been artificially scaled to match the photometry ) .
we discount these . adding reddening to the host cluster , we can account for the f275w and f814w fluxes reasonably well , but there is a f555w excess flux of about a factor of 2 above the model spectra , which could perhaps be accounted for by sn emission ( see below ) .
the stellar masses implied are more reasonable , but the problem with these reddened clusters is that the required reddening is larger than the value of the total @xmath15 inferred for the line of sight ( @xmath15=0.019 mag ; @xcite ) .
a dominant contribution from a reddened cluster therefore seems unlikely as well .
overall , we find that an underlying star cluster does not provide a compelling explanation for the observed photometry , but we can not rule out some contribution from a small cluster or association hidden by sn 2009ip .
another possibility is that the current photometry of sn 2009ip is dominated by late - time sn emission from ongoing csm interaction .
figure [ fig : spec ] shows the sn 2009ip photometry compared to a spectrum that we obtained at the mmt several months later in 2015 november .
this spectrum indicates that very strong h@xmath0 emission with a resolved width of about 2000 km s@xmath3 continued to be observed , requiring the presence of strong ongoing csm interaction at these late times . for comparison ,
figure [ fig : spec ] shows a late time ( day 905 ) spectrum of sn 2005ip from @xcite .
sn 2005ip was a type iin with very strong late - time csm interaction , and this has much higher signal to noise across a wide wavelength range that demonstrates the lower - level emission in many lines that may accompany bright h@xmath0 .
many of these fainter emission lines blend together to make a blue `` pseudo - continuum '' @xcite .
the sn 2005ip spectrum has been scaled to match the same h@xmath0 flux as measured in the spectrum of sn 2009ip , providing a rough indication of the pseudo - continuum emission we might expect from sn 2009ip .
we do not necessarily expect the late - time spectrum of sn 2009ip to be exactly the same as that of sn 2005ip .
the point of this comparison is to illustrate that the level of csm interaction we detect via the h@xmath0 emission in sn 2009ip should be accompanied by a host of other emission lines across a wide wavelength range that should contribute to the broad - band filters .
using sn 2005ip as a guide , it seems possible that csm interaction alone may account for the broadband flux and relatively blue color . a deeper spectrum of the late - time emission can confirm this hypothesis . because the late - time csm interaction emission must make some contribution to the observed f555w photometry in 2015 , we conclude that any underlying continuum source other than csm interaction should be fainter than the f606w progenitor seen in 1999 .
it is likely that we will need to wait several more years to confirm that this source has actually faded enough to place strong constraints on the progenitor .
the main goal of our proposed
observations was to investigate the environment around sn 2009ip , as discussed next .
the comparison between photometry of sn 2009ip and starburst99 model spectra in figure [ fig : sed ] shows that our broad - band images could easily detect any massive young clusters or ob associations in these images .
the sensitivity determined from 3@xmath6 upper limits in the background of these images indicates that we would have detected sources with @xmath16 of @xmath174.7 mag , almost 100 times fainter than the sn 2009ip point source .
this limit corresponds to a single mid to late o - type main sequence star , or a single red supergiant .
the f275w image rules out even relatively small clusters with ages younger than 10 myr or any early o - type main sequence stars ( initial masses above 40 @xmath2 ) ; these would be easily detected in the @xmath11 kpc environment around sn 2009ip .
no such bright clusters or luminous stars are seen .
there is only one source about 100 pc to the south east , which is , however , more than 10 times fainter than sn 2009ip at most wavelengths , and not detected in f275w .
we take this as indicating a remarkable lack of recent massive star formation in the vicinity over the past @xmath110 myr .
h@xmath0 also serves as a tracer of any recent or ongoing nearby star formation .
our f657n image was designed to be sensitive to even relatively modest h ii regions .
indeed , the f657n image reveals dozens of h ii regions with a wide range of surface brightness in the interior few kpc of the host galaxy nucleus , but there is a marked absence of any h ii emission in the outer regions of the galaxy beyond 34 kpc from the center .
the new f657n image of sn 2009ip and its surroundings is shown in figure [ fig : carina ] .
this image reveals no detectable h@xmath0 emission from any young h ii regions in the surrounding @xmath11 kpc around sn 2009ip , which is found about 5 kpc from the galaxy center .
there is a bright point source to the north east ( possibly a foreground star ) and two very faint point sources as well , but all these other sources have similar brightness in broadband continuum filters and so these are unlikely to be h ii regions .
sn 2009ip itself is much brighter in the f657n filter , consistent with its bright h@xmath0 emission detected in the spectrum ( fig .
[ fig : spec ] ) . to illustrate how the lack of h@xmath0 detection indicates a dearth of recent or ongoing massive star formation in the vicinity , figure [ fig : carina ] shows two panels that zoom - in on the surrounding @xmath11 kpc around sn 2009ip .
one of these is the observed f657n image , and the other is the same image with some fake comparison sources injected . as noted earlier in section 2 , we used h@xmath0 images of the orion nebula , the carina nebula , and the 30 dor complex , and scaled them to the appropriate pixel size as they would appear in these images if they resided at the distance of ngc 7259 .
these sources were then convolved to the fwhm of the f657n images and added to the fits files of the _ hst _ data at the appropriate surface brightness for the flux calibration in the _ hst _ images .
we also added these fake images to the left side of the larger scale image in the same figure .
all three nebulae would have been easily detected in the images , although the orion nebula is so small that it would have been an unresolved point source ( note however , that the orion nebula is in a larger complex of star formation that we did not include this in the fake injected image ) .
it is evident that the eastern side of ngc 7259 has no extreme mini - starbursts comparable to 30 dor , although there are a few detected h@xmath0 sources that have a scale in between 30 dor and carina , and several comparable to carina in the inner spiral arms of the galaxy .
there are a few dozen smaller h ii regions comparable to orion scattered around ngc 7259 , but none in the vicinity of sn 2009ip . with a presumed initial mass of 50 - 80 @xmath2 @xcite , and a lifetime of only @xmath14 myr
, one would expect sn 2009ip to reside in a massive young cluster with many other o - type stars , and to be surrounded by bright extended h ii region emission .
this is clearly not the case , nor are there any such h ii regons in the vicinity .
we dod not attempt to obtain deep radio data to search for free - free emission from h ii regions or sn remnants , although this may be interesting future observation to make .
we discuss implications corresponding to each type of region below .
_ 30 dor : _ big h ii region complexes like 30 dor can have ongoing massive star formation for 10 - 20 myr @xcite .
30 dor contains the very young ( 12 myr ) cluster r136 , as well as the @xmath110 myr old population that gave rise to sn 1987a . if sn 2009ip were in such a region , the new _ hst _ images would clearly detect bright extended h@xmath0 emission in the few 10@xmath18 pc surrounding sn 2009ip .
we can rule out the presence of such a large complex at the position of sn 2009ip or within several kpc .
_ carina : _ the carina nebula has birthed about 70 o - type stars @xcite , and is of sufficient scale to sample the upper part of the main sequence .
there are a handful of extremely massive stars in carina , including @xmath19 carinae and three very massive wnh stars .
moderately large h ii regions like carina fade in 5 myr ( carina is now @xmath13 myr old ) when the most massive o stars have died and the bulk of the ionizing photons shut off . with a presumed initial mass of 50 - 80 @xmath2 , one would expect sn 2009ip to come from a region that is no older than than this ( if it is a single star ) .
however , we can clearly rule out a bright extended h@xmath0 nebula like carina at the position of sn 2009ip .
after about 5 myr , there may be a larger and fainter h@xmath0 bubble .
this is harder to rule out , and depends on an adopted escape fraction of ionizing photons .
as noted earlier , our psf subtraction residuals allow for some faint emission within 40 pc of the sn site , but these could also be artifacts .
the lack of bright extended emission to a surface brightness about 10 times fainter than carina s outer shells @xcite would imply a region significantly older than 5 or 6 myr , and would also therefore cause us to expect that any stars with initial masses above 40 @xmath2 are long since dead .
the lack of any carina - like h ii region in the 1 kpc surrounding sn 2009ip also suggests that it is unlikely to be a runaway from such a region . _
orion : _ the orion nebula contains only one o - type star ( with an initial mass around 30 @xmath2 ) .
smaller h ii regions like this with only a couple o stars die after 5 - 6 myr as well , when any mid - o - type stars evolve off the main sequence and most of the ionizing photons shut off .
the surrounding gas shells become neutral and get much fainter and diffuse as they age . moreover , even when young and bright , such a region would be very small in terms of angular extent and would be unresolved by _ hst _ at 20 mpc
. such an h ii region could easily be the birth environment of sn 2009ip and may be hiding under the sn emission
. it would be unusual ( although perhaps not impossible with stochastic sampling of the initial mass function ; @xcite ) to find such a massive 50 - 80 @xmath2 progenitor being born in this type of region .
it seems more plausible that such an h ii region might give rise to a single very massive star like the progenitor of sn 2009ip through binary evolution ( see below ) .
although an orion - like region could be hiding underneath the current photometry of sn 2009ip , it is important to note that such regions are rarely found in complete isolation .
massive star formation tends to propagate from one region of a molecular cloud complex to another @xcite , leading to chains and complexes of h ii region emission in disk galaxies .
orion is part of a larger complex of star formation in the orion region , but no sign of any other nearby h ii regions are seen in the images of sn 2009ip . therefore , we have the additional unusual circumstance that sn 2009ip s birth environment would need to be a singular instance of star formation , or the last such episode of massive star formation in the vicinity .
all of these considerations point to ages that are longer than 5 myr , and more likely around 10 myr or more , suggesting initial masses around 20 @xmath2 .
this is in apparent contradiction with the high luminosity of the progenitor star in its early phase before the series of bright pre - sn eruptive events ( at least 50 - 60 @xmath2 ) .
one way to close the gap between the massive progenitor that was detected and the surprising lack of recent massive star formation in the vicinity would be to invoke binary evolution .
either sn 2009ip was a runaway star born at some other location , or it has been rejuvenated by binary interaction . if it is a runaway , the nearest indications of massive star formation in ngc 7259 are found at least 1.5 kpc away . to cover this distance in a time of @xmath13 myr would require a kick speed of 400 - 500 km s@xmath3 .
this would be a very fast kick for a massive companion of a sn .
moreover , such high runaway speeds seem hard to reconcile with the narrow emission lines in the spectrum of sn 2009ip ( arising in the pre - shock csm ) , whose centroid radial velocities are consistent to within @xmath20100 km s@xmath3 with the redshift of the host galaxy @xcite .
a more likely alternative may be that the progenitor of sn 2009ip increased its mass and luminosity late in life by accreting mass in a mass transferring interacting binary system , or by a stellar merger event ( see @xcite ) .
this would allow the stars to live a longer core h - burning main sequence lifetime of a lower - mass star , and to then get rejuvenated by mass accretion or merging to produce the luminous progenitor source that was detected in 1999 .
this is the sort of evolutionary scenario suggested for lbvs based on their remote environments @xcite , although it should be noted that sn 2009ip s environment seems unusually remote even among known lbvs .
the most isolated lbv known in nearby galaxies is r71 in the lmc , which is 300 pc from any o - type stars , but still on the outskirts of 30 dor @xcite .
@xcite pointed out , however , that rejuvenation by mass accetion or merging might produce an lbv - like star that appears very isolated even without a kick if it is a blue straggler in a relatively small , isolated , and aging star cluster .
this scenario requires the presence of a relatively modest @xmath110 myr - old star cluster or association at the position of sn 2009ip , which could be hidden by the current ongoing csm interaction .
observations after the csm interaction luminosity fades ( this may take several years ) can eventually test this hypothesis . a binary scenario for the progenitor of sn 2009ip
was invoked previously , independent of the remote environment .
the disk - like csm inferred to exist around sn 2009ip @xcite would be consistent with the notion that some strong binary interaction had occurred in the recent past .
the repeating brief luminosity spikes in 2009 - 2011 we also attributed to binary close encouters @xcite , either through direct stellar collisions at periaston in an eccentric system as in the @xmath19 car system @xcite or potentially due to periastron accretion events @xcite .
the remote environment discussed here strengthens the argument for a binary progenitor .
if the sort of binary rejuvenation discussed above is appropriate for sn 2009ip , then it is important to remember that the companion should still be there if the binary system did not merge .
this companion might be combined with any host cluster light in the late time photometry after the csm interaction luminosity fades .
if this sort of binary rejuvenation scenario is _ not _ the origin of sn 2009ip , then its very luminous progenitor source is difficult to understand . to produce such a luminous
, h - rich progenitor with a strong wind and eruptive variability may require a more exotic scenario invoving a binary system with a massive compact object accreting from a companion , for example , which might then lead to a core - collapse sn explosion . a pulsational pair instability ( ppi )
eruption @xcite is the only proposed single - star mechanism to produce a non - terminal eruption that could account for the total energetics of sn 2009ip s 2012 event , which has been shown to require a kinetic energy well above 10@xmath21 ergs because of the asymmetric csm and the large ejecta mass implied by the sustained broad - lined spectrum @xcite .
however , typically the ppi will have a large ejecta mass and expansion speeds @xmath205,000 km s@xmath3 @xcite , slower than the observed 13,000 km s@xmath3 broad lines in sn 2009ip s 2012a event @xcite .
the most energetic ppi events approaching 10@xmath4 ergs would require extremely high initial masses in excess of 100 @xmath2 , and it is uncertain if they occur at all at 0.5 @xmath22 due to mass loss .
it seems very unlikely that there would be no evidence whatsoever of recent massive star formation in the vicinity of sn 2009ip if it originated from a star born with an initial mass above 100 @xmath2 .
support was provided by the national aeronautics and space administration ( nasa ) through hst grant go-13787 from the space telescope science institute , which is operated by aura , inc .
, under nasa contract nas5 - 26555 .
support was also provided by the national science foundation ( nsf ) through grants ast-1210599 and ast-1312221 to the university of arizona .
this study used data products from the southern h - alpha sky survey atlas ( shassa ) , which is supported by the national science foundation . | we present late - time _ hubble space telescope _ ( _ hst _ ) images of the site of supernova ( sn ) 2009ip taken almost 3 yr after its bright 2012 luminosity peak .
sn 2009ip is now slightly fainter in broad filters than the progenitor candidate detected by _ hst _ in 1999 .
the current source continues to be dominated by ongoing late - time csm interaction that produces strong h@xmath0 emission and a weak pseudo - continuum , as found previously for 1 - 2 yr after explosion .
the intent of these observations was to search for evidence of recent star formation in the local ( @xmath11kpc ; 10 ) environment around sn 2009ip , in the remote outskirts of its host spiral galaxy ngc 7259 .
we can rule out the presence of any massive star - forming complexes like 30 dor or the carina nebula at the sn site or within a few kpc .
if the progenitor of sn 2009ip was really a 50 - 80 @xmath2 star as archival _ hst _ images suggested , then it is strange that there is no sign of this type of massive star formation anywhere in the vicinity .
a possible explanation is that the progenitor was the product of a merger or binary mass transfer , rejuvenated after a lifetime that was much longer than 4 - 5 myr , allowing its natal h ii region to have faded
. a smaller region like the orion nebula would be an unresolved but easily detected point source .
this is ruled out within @xmath11.5 kpc around sn 2009ip , but a small h ii region could be hiding in the glare of sn 2009ip itself . later
images after a few more years have passed are needed to confirm that the progenitor candidate is truly gone and to test for the presence of a small h ii region or cluster at the sn position .
[ firstpage ] circumstellar matter stars : evolution stars : winds , outflows supernovae : general supernovae : individual ( 2009ip ) |
in engineering mechanics , _ damage _ is understood as a load - induced evolution of microstructural defects , resulting in a reduced macroscopic material integrity .
the phenomenological constitutive models of damage incorporate the irreversible phenomena by reducing the secant modulus of elasticity depending on the _ internal _ damage variable .
since the seminal contribution of @xcite , it has been well - understood that such description within the framework of local ( i.e. scale - free ) continuum mechanics leads to an ill - posed problem , resulting in localization of damage growth into an arbitrarily small region . as a remedy to this pathology
, a plethora of non - local rate - independent continuum theories , based on integral , explicit and implicit gradient approaches , has been proposed to introduce an _ internal length scale _ into the description , see e.g. ( * ? ? ?
* chapter 26 ) for a representative overview . despite a significant increase in objectivity offered by the enhanced continuum theories , the non - local damage formulations often suffer from the fact that the non - local variables are introduced into the model in an ad - hoc fashion , thus violating basic constraints of thermodynamics .
in addition , due to violation of the principle of local action , such inconsistencies are rather difficult to detect , especially in the multi - dimensional setting , e.g. @xcite . fortunately , as first demonstrated by @xcite and later confirmed by a number of independent studies , e.g. @xcite , a simple one - dimensional study of the localization behavior can serve as a convenient `` filter '' test , allowing to pinpoint various inconsistencies in the constitutive model formulation .
motivation of the present work arose from an energetic non - local damage model proposed by @xcite , which combines the basic features of an engineering damage model due to @xcite with recent advances in the mathematical theory of rate - independent irreversible processes @xcite .
such a connection provides a powerful modeling and analysis framework , allowing at the same time a rigorous mathematical treatment of the complete damage @xcite , theoretically supported numerical implementation @xcite as well as a thermodynamically consistent variational formulation of the non - local damage evolution problem . in this contribution
, a localization study of the model will be performed to examine its qualitative behavior from the engineering viewpoint .
numerical as well as analytical results are provided to illustrate its basic features with emphasis on the proper representation of the damage process from its initiation until complete failure .
let us consider a prismatic bar of length @xmath0 , subjected to displacement - controlled uniaxial tensile loading . in the sequel
, the bar will be represented by the interval @xmath1 , with boundary @xmath2 ( consisting of two points ) subjected to the dirichlet loading @xmath3 ( see ) , where @xmath4 denotes the ( pseudo- ) time taken from interval @xmath5 . following the standard thermodynamic approach to constitutive modeling , cf .
* chapter 25 ) , a state of the system is described using an admissible displacement and damages fields @xmath6 and @xmath7 , respectively .
formally , we write @xmath8 where @xmath9 denotes the set of kinematically admissible displacements , @xmath10 stands for the set of admissible damage values , and @xmath11 is the sobolev space of functions with square - integrable weak derivatives , e.g. @xcite , often denoted as @xmath12 . within the adopted _ global _ energetic framework @xcite ,
the constitutive description of the damage model is provided by the _ globally stored energy _ functional @xmath13 and by the _ global dissipation distance _
@xmath14 physically , @xmath15 represents the energy reversibly stored in the system , and @xmath16 is the energy dissipated by changing the damage field from @xmath17 to @xmath18 . in eqs .
( [ eq : ge ] ) and ( [ eq : gdd ] ) , @xmath19 [ pa ] denotes the young modulus of the material , @xmath20 [ jm@xmath21 is the amount of energy needed to disintegrate a unit volume of the material , and @xmath22 [ m ] is a characteristic length of the model .
later it will become clear that the `` @xmath23 '' term appearing in ensures the irreversibility of the damage evolution ; i.e. at any point , the damage variable can not decrease in time .
now , given the dirichlet loading @xmath3 , energetic functionals @xmath15 and @xmath16 and initial data @xmath24 and @xmath25 , the _ energetic solution _ of the damage problem is provided by functions @xmath26 and @xmath27 satisfying @xcite : global stability : : for all @xmath28 $ ] , @xmath29 and @xmath30 @xmath31 energy equality : : for all @xmath28 $ ] @xmath32 where @xmath33 denotes the power of external loading .
although the previous two conditions present the formal definition of the energetic solution , the analysis itself will be performed using the time discretization technique , see e.g. @xcite for a nice exposition . to that end
, we introduce a uniform partitioning of the time interval @xmath34 and inductively solve the minimization problem @xmath35 \hskip 5 mm \mbox{for } k = 1 , 2 , \ldots , n\ ] ] note that the previous problem is independent from @xmath36 , which is consistent with the assumed rate - independent character of the damage process .
the theoretical results gathered in @xcite show that , under reasonable data qualification , the solution of the time - discretized problem converge to the energetic solution as @xmath37 .
having established the essentials of the global energetic framework , we will now proceed with the main goal of this contribution the analysis of the simple uniaxial tensile test .
intuitively , we expect that as long as the energy density at all points of the structure stays below @xmath20 , the response of the bar remains elastic and no damage evolution occurs within the structure .
after the energy density reaches the value of @xmath20 at a certain time @xmath38 , damage is initiated and localization occurs .
therefore , the explicit description of the localization phenomenon will build on the incremental variational principle ( [ eq : incr_min ] ) .
note that , for the sake of brevity , the time instant @xmath38 and the time increment @xmath39 will be omitted in the rest of this paper .
dropping the constraints on @xmath40 for a moment , the optimality conditions for the minimization problem ( [ eq : incr_min ] ) read @xmath41 \de x \geq 0\ ] ] for all admissible variations @xmath42 vanishing on @xmath2 and arbitrary @xmath43 . assuming the involved fields are smooth enough , we formally obtain @xmath44_{\bnd } & \geq & 0\end{aligned}\ ] ] using the arbitrariness of the variations and imposing the constraints @xmath45 finally leads to the system of the karush - kuhn - tucker conditions , cf .
* section 15.2.4 ) @xmath46 corresponds to the uniform stress condition @xmath47 while eqs .
( [ eq : el2 ] ) , ( [ eq : el3 ] ) and ( [ eq : el4 ] ) provide the link between strain and damage and describe the elastic , damaging and fully damaged part of the structure , respectively .
note that by virtue of , the stress @xmath48 remains to be well - defined even in the fully damaged region , where the physical meaning of displacements @xmath49 and strains @xmath50 is rather questionable .
this open the way to the rigorous mathematical treatment of the complete damage problem , see @xcite for further discussion .
concentrating on the damaging zone only , we divide by @xmath20 and , employing identity ( [ eq : sigma_def ] ) , rewrite it in the form of a non - linear ordinary differential equation @xmath51 where @xmath52 parameterizes the softening process and ranges from @xmath53 at the onset of localization to @xmath54 at the stress - free state of complete failure .
it is also useful to rewrite in the rate form @xmath55 which is an ordinary second - order differential equation for the damage rate @xmath56 with possibly non - constant coefficients .
note that damage irreversibility requires @xmath57 .
observe that at the onset of damage , @xmath58 and @xmath59 .
therefore , simplifies to @xmath60 leading to the general solution @xmath61 with @xmath62 and @xmath63 denoting the integration constants , the latter specifying the position of the center of the localized profile . assuming that this zone corresponds to an interval @xmath64 located sufficiently far from the boundary , the damage rate has to verify four independent conditions @xmath65 yielding @xmath66 where @xmath67 is the width of the localized zone , cf .
a. the rate of the damage variable at the onset of damage is therefore given by @xmath68 the previous expression is consistent with the constraint @xmath69 since the stress rate @xmath70 is negative in the softening regime , which implies @xmath71 .
mielke , a. ( 2005 ) .
evolution of rate - independent systems . in dafermos ,
c. and feireisl , e. , editors , _ handbook of differential equations : evolutionary equations _ , volume 2 , pages 461559 .
elsevier b.v . ,
amsterdam , the netherlands .
mielke , a. , roubek , t. , and zeman , j. ( 2007 ) .
complete damage in elastic and viscoelastic media and its energetics .
wias preprint 1285 , weierstrass institute for applied analysis and stochastics , berlin . | the contribution presents an analysis of a rate - independent non - local damage model , recently proposed by @xcite .
an analytical as well as numerical solution of a simple one - dimensional bifurcation problem is performed , demonstrating that , for the elementary localization test , the model is free of pathological features . [ [ keywords ] ] keywords + + + + + + + + damage , non - locality , energetic solution , discretization in time , bifurcation |
the motivation to research cdznte ( czt ) detectors originates from participation in the cobra experiment @xcite , a proposed massive ( several hundred kg ) array of czt crystals for double - beta decay research .
taking into account a typical mass of merely a few grams for each crystal , several tens of thousands of crystals will eventually have to be operated reliably over several years .
naturally , such a set - up would become vastly more practical by utilising simple ways to mount and operate individual crystals in the array .
single - channel readout for each crystal as opposed to coplanar grid detectors is considered to be an attractive option .
already the reduction of wiring close to the crystals by a factor of two would be highly significant in this case .
frisch collar detectors @xcite appeared to be the most practical way forward , so we modified three existing coplanar grid crystals , low - grade from ev - products . in order to have a standard to compare with
, we purchased one ev - capture plus technology detector which is of much higher quality @xcite .
crystals and preamplifier are housed together in a standard diecast enclosure featuring connectors for preamplifier power , bnc output and high - voltage input ( see figure [ photo ] ) . a single hv - power supply ( ortec 659 , nim module )
delivers both polarities up to 5kv .
two linear dc power supplies deliver @xmath0v to the preamplifier ( see figure [ preamp ] ) .
the output signal is connected directly to the data acquisition system ( daq ) using a 50@xmath1 bnc cable ( rg58 ) .
the daq consists of a 100 mhz sampling digital oscilloscope in a 3u pxi module from national instruments ( ni pxi-5112 ) mounted in a pxi crate ( ni pxi-1042 ) and is controlled by an embedded controller pc ( ni pxi-8186 ) running custom - made labview software for digital pulse acquisition .
pulses are streamed directly to disk in binary format for maximum sampling speed when using radioactive sources for detector calibration .
the dynamic range is limited to 7-bit ( 8-bit oscilloscope , both polarities measured ) , hence each acquisition needs a little fine - tuning for the vertical range to capture all important structures .
we emphasize this point since this has become an issue when calibrating with a ba-133 source , see figure [ ba133 ] .
the 81 kev line resolution is limited by digitisation noise rather than our analysis method whilst the line voltage amplitude is simply too small when the system is set - up to acquire the 356 kev line simultaneously .
the core of the data acquisition system is the preamplifier design , see figure [ preamp ] .
it represents a rather unusual method for readout of semiconductor pulses since it is not a charge - sensitive amplification system .
this preamplifier can be described as a straightforward source - follower .
the motivation to try this type initially was to learn more about signal formation in the semiconductor , i.e. to specialise the analysis and measurement on pulse - shape in contrast to pulse - height .
the preamplifier is optimised for high - speed , consistent with the data acquisition bandwidth , and highest possible signal to noise ratio .
note that both properties are almost mutually exclusive , i.e. speeding up the amplifier worsens the signal and better signal to noise ratio slows the amplifier .
however , for applications that require a higher signal gain , focussing on low - energy signals for example , it is possible to increase the gain without losing an equal factor in speed .
our application for this set - up , however , targets rather higher energy signals of up to several mev , hence the circuit in figure [ preamp ] is expected to work optimally for us . the existing three coplanar grid detectors were modified to work as frisch collar detectors .
all three are cubes of volume 1@xmath2 with gold - plated electrodes , a full area cathode and a coplanar grid anode structure .
all faces ( excluding the cathode face ) are covered with insulating paint . since we were not interested in operating the grid , paint covering the anode
was partly removed ( by dissolving it in ethanol ) and the area used to contact the anode with a wire was attached by a generous drop of silver conductive paint .
the preamplifier is ac - coupled to the crystal anode which is biased positively by the hv - power supply .
the cathode is kept at ground potential .
this mode was used to operate and test the crystals as simple planar detectors . for the frisch collar mode ,
each crystal was wrapped in two layers of thin teflon tape , covering the full height , leaving out anode and cathode , similar to devices fabricated in @xcite .
the teflon layer was wrapped in a metal foil ( aluminium kitchen foil worked best for us ) and the foil attached via a small lip to the cathode ( using silver conductive paint ) .
the metal foil height determines the performance of the frisch collar detector @xcite .
we achieved best operating performance with 9 mm - 9.5 mm foil height .
any higher shield results in stability problems when biasing the anode since the grounded shield appears to be too close , particularly at the cube corners .
finally , prepared crystals can be mounted for operation on a ground plane .
we used a small copper - clad piece of printed - circuit board connected to the preamplifier ground , see figure [ photo ] . , width=453 ] the ev - products capture plus technology crystal is essentially a frisch collar detector using a resistively coupled shield instead of a capacitively one as outlined above .
it has , however , one important feature in addition , a small anode , utilising the small pixel effect @xcite .
the best of both worlds has been combined with this type of detector in order to achieve an electron - signal - only operation with a single channel readout .
the surface area of the rectangular crystal is 1@xmath3 and its height is quoted as 5 mm .
it arrived factory certified with an energy resolution of 2.20% at the cs-137 line at 662 kev and 4.40% at the co-57 line at 122 kev . according to @xcite ,
an energy resolution of 1.5% at 662 kev has been achieved under favourable conditions .
the crystal is contacted and biased identically to our modified crystals , except that this crystal can withstand higher bias than the frisch collar crystals ( recommended bias is 1.5 kv ) .
note that an alternative design , utilising the capacitively coupled shield and a small anode has been realised in @xcite with excellent results . in case anode as well as
cathode signals are required to be measured with two preamplifiers , it is important to raise the ground plane ( pcb ) off the diecast box floor to minimise capacitance for the cathode signals
. a 1 cm insulating spacer should be sufficient .
the two - channel readout serves to gain further insight into the signal formation by picking up anode and cathode signal simultaneously for a single event .
the ratio of these signals is expected to yield depth information @xcite and its sum could improve energy resolution by boosting the total amount of charge collected .
so far , we utilised this type of readout only for the ev crystal .
further research into this mode of operation is in progress .
, width=491 ]
optimising energy resolution for frisch collar detectors is fully documented in @xcite .
performance of crystal operation in this mode was successfully re - produced . however , in this process we noticed that our absolute energy resolutions were surprisingly good compared to the literature .
note that our crystals are remarkably large for czt crystals operating as frisch collar detectors and the crystal quality was expected to be poor compared to spectroscopy - grade crystals . additionally , our readout was custom - made for pulse - shape analysis and not specialised for spectroscopy operation . obtaining energy information from integrated digitised pulses is always considered to be inferior to analogue operations ( current integration on a charge - sensitive preamplifier and spectroscopy shaping ) .
this common - knowledge can lead to interesting preamplifier designs combining charge and current sensitive readout @xcite . taking all of this into consideration
, we were surprised to see our initial energy resolutions comparable to published results . in order to achieve further improvements
the following data analysis method has been developed .
frisch collar detector fabrication inevitably leads to high input capacitances relative to planar crystals .
the metal shield wrapped around the bulk of the crystal functions by collecting induced charges from any current flowing inside the capacitively coupled crystal .
thereby , it prevents the majority of the signal appearing on the signal electrodes until the charge cloud comes close to the anode , hence achieving close to electron - only signal formation .
, width=377 ] , width=377 ] , width=377 ] a high input capacitance coupled to the source - follower preamplifier effectively smears or low - pass filters the image of the current flowing in the detector crystal ( interpreted here as a current source ) . therefore our signal ( see figure [ signal1 ] ) is not a true transient current measurement initially .
however , a cr - rc filter applied to the digitised pulse can enhance the fast components in the signal by effectively transforming the pulse to its rate - of - change image . after shaping
, the pulse more closely represents the actual transient current flowing in the crystal .
we found that for the given preamplifier settings and the data acquisition bandwidth , time - constants of 200 ns for the high - pass and 4 @xmath4s for the low - pass yield satisfactory results for all our cystals .
the result of such a shaping is shown in figure [ signal2 ] for a typical pulse .
naturally , this shaping process enhances the high - frequency noise , so we decided to apply a second step to this two - stage software filtering in the form of a moving average filter ( width 200 ns ) .
note that a moving average filter belongs to the class of optimal filters , where the moving average filter represents the optimal method to preserve any steep , fast - changing feature in a pulse while smoothing high - frequency noise .
figure [ signal3 ] shows its effect on the pulse from figure [ signal2 ] .
the final analysis step involves calculating basic pulse parameters which serve to extract the integral and therefore the energy of the pulse .
those parameters also greatly help in discriminating against corrupt pulses from pulse pile - up or short - circuit events from over - biasing crystals .
we calculate the following set of parameters : baseline , onset , peak position and amplitude , rise - time , decay - time and peak broadness ( width at 80% of amplitude left and right from maximum ) and finally the integral and the integration risetime .
for the two - channel readout we also calculate the sum signal and the ratio . , width=377 ] , width=377 ] , width=377 ] figure [ cs137 ] and figure [ ba133 ] display the energy spectra obtained from calibrations with a cs-137 and a ba-133 source , respectively , both acquired using our test standard , the capture plus crystal .
the energy resolutions are listed in table [ tab1 ] .
they have been obtained from fits to full energy peaks as shown in figure [ fit ] as an example . due to the inevitable
tailing in single - channel readout systems , we adopted the fitting function from @xcite , @xcite in order to extract reliable energy resolution values .
at this point it is worth emphasizing that no subsequent data corrections have been applied , i.e. no rise time to pulse height correlations or any other correlations have been utilised to improve energy resolution .
as shown in figure [ rtplot ] , our frisch collar crystals would benefit greatly from such an offline correction , whereas the capture plus detector does not suffer from a similarly strong tailing , mainly due to the small pixel effect .
@xcite shows an impressive effect of such a bi - parametric correction , achieving a similar resolution ( less than 1% at 662 kev ) using a custom - made frisch collar crystal , to what we achieve with the capture plus crystal .
however , we point out that our main application of czt detectors in fundamental rare event research ( see introduction ) does not allow for unspecified event efficiencies .
any data cut would have to be known with highest possible precision , hence it is generally better in such an application not to apply any cut ( or energy - dependent corrections ) at all .
our analysis merely removes clearly invalid events such as empty baselines and pulses that saturate the data acquisition system . , width=377 ] , width=377 ] .energy resolution results obtained in this work .
the outstanding results obtained with the capture detector are listed at the top .
subsequently , the frisch collar detectors are numbered as # 1,2,3 .
the bias and collar height show conditions under which the best results were obtained . re - dressing detector # 1 with various new collars over time has worsened the resolution since its best result , the 1.33% ( see figure ( 10 ) ) , was obtained as one of the first measurements . the second result at 1.5kv
is listed to indicate how over - biasing affects the result since the detector was not stable at this bias.[tab1 ] [ cols="<,<,<",options="header " , ]
a new readout and data analysis method for single channel , room - temperature semiconductor detectors is introduced which significantly improves energy resolution despite its simplicity .
the new preamplifer circuit and data analysis methods have been discussed in detail .
the application to three potentially rather poor czt crystals operated as frisch collar detectors yields energy resolutions comparable to or better than existing values from literature either for frisch collar detectors or coplanar grid detectors of similar size ( but often , where mentioned , far superior crystal quality ) .
our chosen test standard for comparison , an ev - products capture plus detector , surpasses all expectations and shows energy resolutions , to the best of our knowledge , better than any measured with a similar device so far .
energy resolutions for such a large volume czt detector of well under 1% at 662 kev without subsequent corrections , i.e. efficiency losses , have so far only been reported for much more complicated ( in terms of fabrication and readout ) pixelated crystals .
00 k. zuber , _ phys .
* b519 * ( 2001 ) 1 ; t. bloxham et al . , cobra collaboration , _ phys .
rev . _ * c76 * ( 2007 ) 025501 d.s .
mcgregor et al . , _ appl .
* 72 * ( 1998 ) 792 ; w.j .
mcneil et al . , _ appl .
* 84 * ( 2004 ) 1988 ; g. montemont et al . ,
_ ieee trans .
* 48 * ( 2001 ) 278 ; a. kargar et al .
_ * a558 * ( 2006 ) 497 ; a. kargar et al . ,
_ * a562 * ( 2006 ) 262 ; m. harrison , a. kargar and d.s . mcgregor , _ nucl
_ * a579 * ( 2007 ) 134 d.s . bale and c. szeles , _ proc .
spie _ * vol .
6319 * ( 2006 ) ; c. szeles et al . , ibid .
( 2006 ) 191 ; see also white papers at _ http://www.evproducts.com/czt_white_papers.html _
l. verger et al .
, _ ieee trans . nucl .
* 52 * ( 2005 ) 1733 zhong he et al .
_ * a380 * ( 1996 ) 228 ; _ nucl .
_ * a388 * ( 1997 ) 180 h. barrett , j. eskin and h. barber , _ phys .
* 75 * ( 1995 ) 156 a. hamrita et al .
_ * a531 * ( 2004 ) 607 st .
w. smith , _ the scientist and engineer s guide to digital signal processing _ , cal . tech .
( 1997 ) usa ; see also _ http://www.dspguide.com _ gunnink and arlt , _ nucl .
_ * a458 * ( 2001 ) 196 r.m .
keyser , _ ieee nuclear science symposium conf .
record _ ( 2001 ) 315 l. verger et al .
_ * a571 * ( 2007 ) 33 | we present methods to improve the energy resolution of single channel , room - temperature cadmium - zinc - telluride ( czt ) detectors .
a new preamplifier design enables the acquisition of the actual transient current from the crystals and straightforward data analysis methods yield unprecedented energy resolution for our test - detectors . these consist of an ev - capture plus crystal as standard and 1 cm cube frisch collar crystals created in - house from low - grade coplanar grid detectors .
energy resolutions of 1.9% for our collar detectors and 0.8% for the ev crystal at 662 kev were obtained . the latter compares
favourably to the best existing energy resolution results from pixel detectors . |
the importance and difficulty of solving models of interacting quantum particles is hard to overstate .
it is well known that the correlated motion of those particles gives rise to a wide variety of physical phenomena at different length and time scales , spanning disciplines like chemistry , condensed matter , nuclear , and high energy physics .
novel complex structures can emerge as a consequence of the competing multiple - length scales in the problem .
nonetheless , only a reduced set of interacting problems admits exact closed form solutions @xcite and the use of numerical techniques becomes essential if one is looking for accurate solutions not subjected to uncontrolled approximations . among those techniques , the statistical methods @xcite offer the potential to study systems with large number of degrees of freedom , reducing the computational complexity from exponential to polynomial growth
this scaling behavior is particularly relevant when one recognizes that most of the interesting phenomena in many - body physics occurs in the thermodynamic limit @xcite .
unfortunately , for fermions ( i.e. quantum particles obeying fermi statistics ) the sign problem plagues all useful stochastic algorithms and causes the variance of computed results to increase exponentially with increasing number of fermions @xcite . on the other hand , the growing interest in physical systems whose state functions are defined on a general metric space makes the quantum mechanics of interacting particles in curved manifolds no longer a mere intellectual exercise , but one with very practical consequences .
perhaps the most well - known examples can be found in cosmology ( e.g. , matter in strong gravitational fields , atomic spectroscopy as probe of space - time curvature @xcite ) , but the subject is certainly not exclusive to this field . in condensed matter
a very elementary case is provided by a deformed crystal .
less well - known ones are mesoscopic graphitic microtubules and fullerenes .
all these physical systems are ubiquitous in nature and the crucial role the curvature of the manifold plays has been confirmed by experimental observations ( e.g. spectrum of collective excitations @xcite ) .
therefore , the development of stable quantum methods with polynomial complexity in riemannian manifolds represents a real challenge for many - body theorists . the present manuscript deals with the ( non - relativistic ) many - particle schrdinger equation in a general metric space and its solution using stochastic techniques .
in particular , we will show how to construct approximate solutions ( wave functions ) for systems with broken time - reversal symmetry ( e.g. electrons in the presence of external electromagnetic sources ) avoiding the infamous `` phase problem '' @xcite .
the main difficulty is to define a probability measure ( semi - positive definite ) which allows one to get the complex - valued state with no asymptotic signal - to - noise ratio decay in euclidean time .
this translates into a problem of geometric complexity , which is solved approximately using constraints in the manifold where the wave function has its support . in this way
, we get stable but approximate solutions which can be systematically improved . among the large variety of problems one can attack
, we decided to choose the general problem of fermions in the presence of external gauge fields to illustrate the main ideas .
the effects of an external magnetic field on a system of electrons can be profound @xcite .
the field couples to the electron s charge and spin , modifying its spatial motion and lifting its spin degeneracies .
the field can also create spatial anisotropy , effectively reducing the dimensionality of the system from three to two .
the combination of the reduced dimension and the field itself is known to have novel consequences .
for example , in a system of non - interacting electrons hopping on a square lattice , the field transforms the energy spectrum from the simplicity of trigonometric functions to the complexity of a field - dependent self - similar structure ( hofstadter s butterfly ) whose depth mathematicians are still fathoming @xcite .
the combination of the reduced dimensionality , strong particle interactions and the field itself is known to have novel consequences , like the formation of isotropic fractional quantum hall fluids @xcite , which are incompressible states of the two - dimensional homogeneous coulomb gas .
the projector ( zero temperature ) method we will introduce uses random - walks to solve a general multidimensional partial differential equation second order in space coordinates and first order in time . whenever mention is made of a random - walk we mean a markov chain that is defined as a sequence @xmath0 of @xmath1 random variables that take values in configuration space , i.e. the space of particle positions . as usual , what characterizes a random - walk is its initial probability distribution and a conditional probability that dictates the transition from @xmath2 to @xmath3 .
this transition probability is non - unique and discretization dependent @xcite . among all the possible choices we will require a prepoint discretization of the transition probability ( short - time propagator ) because we will use monte carlo methods to generate the walkers .
the paper is organized as follows . in section [ section2 ]
we present the formulation of the general problem of fermions in curved manifolds .
in particular , for illustration purposes and to fix notation , we develop the formalism for spin-@xmath4 particles in the presence of an external electromagnetic potential .
then , we show how to project out the lowest energy state of a given symmetry in a manifold with curvature , and discuss the resulting fokker - planck equations for various distribution functions .
once the problem is precisely defined we develop , in section [ section3 ] , path - integral solutions to those multidimensional differential equations , and give an interpretation of the emergent `` quantum corrections '' in the euclidean action .
the path - integral solutions are evaluated using monte carlo techniques in section [ section4 ] .
there , we provide an stable step by step practical algorithm which emphasizes the subtle changes ( with respect to the standard diffusion monte carlo ( dmc ) technique ) due to the metric of the manifold . in section [ section5 ]
we apply such computational implementation to the problem of an electron moving on the surface of a sphere in the presence of a dirac monopole .
finally , section [ section6 ] summarizes the main findings and discusses the relevance of the stochastic method as applied to the physics of quantum hall fluids .
_ notation . _ consider a differentiable manifold @xmath5 of dimension @xmath6 ( e.g. , for the two - sphere s@xmath7 , @xmath6=2 ) with coordinates @xmath8 defined on it . if @xmath9 is a riemannian manifold , then it is a metric space , with metric tensor @xmath10 , such that the distance @xmath11 between two points in @xmath9 is @xmath12 in the usual way @xcite .
the metric tensor is positive definite and symmetric @xmath13 ( as we will see , this condition is important to define a probability density distribution ) , and is a function of the coordinates @xmath14 with the property @xmath15 .
let us consider the coordinate transformation @xmath16 : @xmath17 .
then , a generic second order contravariant ( @xmath18 ) and covariant tensor ( @xmath19 ) transform as @xmath20 respectively . throughout the paper einstein s summation convention for repeated indices
is assumed ( @xmath21 ) . _ formulation of the problem .
_ in this article we will be concerned with finite interacting fermion systems in the presence of an external electromagnetic potential @xmath22 ( @xmath23 represents a uniform field , @xmath24 and @xmath25 are the vector and scalar potentials , respectively ) whose hamiltonian for motion on the manifold , in the coordinate representation , is given by @xmath26 with @xmath27 + \frac{e^2}{2 m^ * c^2 } \sum_{i=1}^{n } a^{\mu}(i ) a_{\mu}(i ) \ , \ ] ] @xmath28 , and @xmath29 , where @xmath30 is the covariant laplace - beltrami operator and @xmath31 is a potential energy operator .
notice that we use the conventional notation where the transformation between different forms of a given tensor is achieved by using the metric tensor ( e.g. , @xmath32 , @xmath33 ) , and @xmath34 .
this hamiltonian characterizes the dynamics of @xmath35 non - relativistic indistinguishable particles of mass @xmath36 , charge @xmath37 and spin @xmath38 in a curved space with metric tensor @xmath39 , and @xmath40 .
we have assumed that the quantum hamiltonian @xmath41 in curved space has the same form as in flat space ( this amounts to a particular operator ordering prescription . )
given the previous ordering , one can rewrite the hamiltonian above in terms of the generalized ( hermitian ) canonical momentum @xmath42 @xmath43 where the kinetic momentum @xmath44 . the first term in eq .
[ eq : hami ] represents the kinetic energy of the system and is the non - relativistic approximation to the dirac operator .
@xmath31 includes the sum of one and two - body local interaction terms ( and background potential in the case of a charge neutral system ) and zeeman contribution .
the potentials are assumed to be finite almost everywhere and can only be singular at coincident points ( @xmath45 , @xmath46 ) we are interested in the stationary solutions of the resulting multidimensional schrdinger equation @xmath47 and will restrict ourselves to hamiltonians @xmath41 which are time - translation invariant . in the usual space - spin formalism the @xmath35-fermion states characterizing the system , @xmath48 , and all its first derivatives belong to the hilbert space of antisymmetric ( with respect to identical particle @xmath49-exchanges ) square - integrable functions @xmath50 , defined as @xmath51 where @xmath52 ( @xmath53 and @xmath54 are discrete spin variables ) and @xmath55 represents the permutation of the pairs @xmath56 and @xmath57 .
@xmath58 is the cartesian product manifold of dimension @xmath59 . since the system hamiltonian can be written as @xmath60 , the last term representing the zeeman coupling , the many - body wave function @xmath61 can be expressed as a tensor product of a coordinate and a spin function ( or a linear combination of such products ) , @xmath62 we want to construct @xmath35-fermion eigenstates of @xmath41 , @xmath63 , that are also eigenfunctions of the total spin @xmath64 ( @xmath65 ) , @xmath66 and this is always possible since @xmath67 $ ] = 0 .
thus , the configuration part @xmath68 must have the right symmetry in order to account for the pauli principle .
it turns out that a coordinate state @xmath69 which is symmetrized according to the young scheme @xcite and has total spin @xmath70 will be antisymmetric in the variables @xmath71 , and antisymmetric in the variables @xmath72 .
moreover , @xmath73 possesses the property of fock s cyclic symmetry , @xmath74 where , in this case , @xmath75 refers to the transposition of particle coordinates @xmath76 and @xmath77 .
this last condition is a very useful one for testing the symmetry of a given coordinate function . _
quantum projection on curved manifolds .
_ for a given total spin @xmath78 we are thus left with the task of solving the stationary many - body schrdinger equation @xmath79 , where @xmath80 satisfies the symmetry constraint discussed above . in particular
, we are interested in the zero temperature properties of this quantum system , i.e. its ground state properties . to this end
, we study the euclidean time evolution of the state @xmath73 , i.e. we analytically continue eq .
[ sch ] to imaginary time ( wick rotation , @xmath81 ) @xmath82 \phi \ , \label{scht}\ ] ] whose formal solution @xmath83 \phi_t$ ] is used to determine the limiting distribution @xmath84 which is the largest eigenvalue solution of the evolution operator @xmath85 compatible with the condition @xmath86 , where @xmath87 is a parent state and @xmath88 is a suitable ( constant ) energy that shifts the zero of the spectrum of @xmath89 .
we would like to solve the multidimensional differential equation eq .
[ scht ] using initial value random walks . in this way ,
starting with an initial population of walkers ( whose state space is @xmath90 ) distributed according to @xmath91 ( @xmath87 must be positive semi - definite ) , the ensemble is evolved by successive applications of the short ( imaginary ) time propagator @xmath92 ( @xmath93 , and @xmath94 is the number of time slices ) to obtain the limiting distribution @xmath95 .
then , we can introduce a `` pseudo partition function '' @xmath96 in terms of which we can determine the ground state energy @xmath97 as @xmath98 similarly , other ground state expectation values , e.g. , @xmath99 , can be obtained as derivatives ( with respect to a coupling constant @xmath100 ) of a modified pseudo partition function @xmath101 whose evolution operator has a modified hamiltonian , @xmath102 .
in order to reduce statistical fluctuations in the measured quantities ( i.e. , observables ) one can guide the random walk with an approximate wave function , @xmath103 , which contains as much of the essential physics as possible ( including cusp conditions at possible singularities of the potential @xmath31 ) .
then , instead of sampling the wave function @xmath104 one samples the distribution @xmath105 ( properly normalized ) with the initial time condition @xmath106 .
expectation values of operators @xmath107 ( observables ) that commute with the hamiltonian have a particularly simple form for guided walkers .
for instance , @xmath108 where the average @xmath109 stands for @xmath110 @xmath111 is the long - time stationary probability of the system , and the ( invariant ) volume element @xmath112 is given by the @xmath59-form @xmath113 \
{ \rm d}x_1 ^ 1 \wedge \cdots { \rm d}x_1^d \wedge \cdots { \rm d}x _ { n}^d \ .\ ] ] remember that in a general metric space the resolution of the identity operator with respect to the spectral family of the position operator is latexmath:[\[\bbbone = \int_{{\cal m}^{n } } \mbox{\boldmath $ \omega$ } \
that @xmath87 and the guiding function @xmath103 can , in principle , be different functions , although most of the practical calculations use the same function .
it turns out that this importance sampling procedure is decisive to get sensible results when the potential @xmath31 presents some singularities .
notice , however , that the quantum hamiltonian @xmath115 breaks explicitly time - reversal symmetry , meaning that in general @xmath73 will be a complex - valued function .
even if @xmath73 were real - valued , because it represents a fermion wave function it can acquire positive and negative values ( the case where @xmath116 is totally antisymmetric being the exception ) .
then , it is clear that we can not in principle interpret @xmath73 or @xmath117 as a probability density . for reasons that will become clear later we will be interested in sampling the probability density @xmath118
can be derived directly from eq .
[ scht ] with the result @xmath119 - ( e_l - e_t ) \bar{f } \ , \ ] ] where the drift velocity @xmath120 , and the `` local energy '' of the effective ( `` fixed - phase '' ) hamiltonian @xmath121 ( see eq .
[ fpham ] and its derivation ) is @xmath122 with @xmath123 + \widehat{v}({\cal r } ) \ , \label{fixpham}\ ] ] where @xmath124 is the phase of the many - body state @xmath73 , i.e. @xmath125 $ ] .
the differential equation satisfied by the distribution function @xmath126 is formally equivalent to the one describing brownian motion on a general manifold ( including generation and recombination processes ) , and corresponds to a kramers - moyal expansion with exactly two terms .
in fact , we can rewrite the equation above as a fokker - planck equation for @xmath59 continuous stochastic variables @xmath127 @xmath128 where the ( time - independent ) fokker - planck operator @xmath129 is given by @xmath130 \ .\ ] ] the diffusion matrix ( contravariant tensor ) @xmath131 and drift @xmath132 ( which does not transform as a contravariant vector ) are given by @xmath133 where @xmath134 is the christoffel symbol of the second kind @xmath135 and the modified local energy @xmath136 notice , however , that singularities in the `` quantum corrections '' @xcite to the local energy @xmath137 due to the metric , can induce very large fluctuations in @xmath138 . moreover , the probability density @xmath126 does not transform as a scalar function ( @xmath139 , where the primes represent the transformed coordinates and @xmath140 is a volume element in @xmath58 ) .
therefore , it is more convenient to work with a probability density that is a scalar and which is defined as @xmath141 \
\bar{f}({\cal r},t ) \ .
\label{ff}\ ] ] the differential equation @xmath142 satisfies is of the form eq .
[ fock ] with bar quantities replaced by unbar ones ( e.g. @xmath143 ) .
it turns out that @xmath144 and the drift ( which is not a tensor ) @xmath145 note that in this case the quantum correction to the local energy vanishes .
furthermore , if the metric is diagonal , i.e. @xmath146 , then the correction to the flat space drift also vanishes , i.e. @xmath147 , and @xmath148 .
this last remark is quite important , specially for @xmath149 where it is _ always _ possible to choose a coordinate system ( @xmath150 ) where the metric tensor is diagonal ( conformal gauge @xcite ) , and use the conformal parameterization ( @xmath151 ) which greatly simplifies the resulting expressions ( see section [ section5 ] ) .
the generalized fokker - planck eq . [ fock ] describes the time evolution of a distribution function @xmath142 which is completely determined by the distribution function at @xmath152 . in this sense
it describes a continuous stochastic process that is markovian . because it represents a markov process , the conditional probability that if the system is in @xmath153 at time @xmath154 it will jump to @xmath155 in time @xmath156 ( importance - sampled green s function ) @xmath157 contains the complete information about the process , and it follows that the probability densities @xmath158 and @xmath159 are connected by @xmath160 where the green s function @xmath161 is a transition probability for moving particles from @xmath162 to @xmath163 in time @xmath164 with the initial value
@xmath165 \delta({\cal r}-{\cal r}')$ ] , and is formally given by @xmath166 | { \cal r } \rangle \
\widetilde{\phi}_g^{-1}({\cal r } ) \ , \ ] ] with @xmath167 \phi_g({\cal r } ) \ .\ ] ] path integral solutions for @xmath142 may be derived from the transition probability density for small @xmath164 .
iteration of the chapman - kolmogorov equation for @xmath168 allows one to express the evolution of @xmath169 from the initial distribution @xmath170 in terms of the short - time green s function as @xmath171 where @xmath172 ; we identify @xmath173 and @xmath174 . by simple inspection
the solution of the generalized fokker - planck equation @xmath142 stays positive if it was initially positive ( i.e. , if @xmath175 . )
it is clear then that the functional integral representation of @xmath142 requires knowledge of the infinitesimal evolution operator .
it is well - known in a similar context @xcite that the integrand of the functional integral is not unique , it is discretization dependent ( compatible with the markovian property of the paths since @xmath176 is only sampled at @xmath177 and @xmath156 . ) on the other hand , it is crucial for numerically simulating those paths to use a discretization where the drift velocity and diffusion are evaluated at the * prepoint * in the integral equation . following feynman @xcite we have determined the functional form of @xmath178 to @xmath179 .
we are going to present the final result and omit the details of the calculation which are just simple ( although lengthy ) manipulations of taylor series expansions and gaussian integration .
thus , to @xmath179 the short - time conditional probability is given by @xmath180 where @xmath181}{2}-e_t \right ) \right ] \ , \label{green1}\end{aligned}\ ] ] and @xmath182 \ , \label{green2}\end{aligned}\ ] ] that is , a gaussian distribution with variance matrix @xmath183 and mean @xmath184 .
sample trajectories ( continuous but nowhere differentiable ) are generated by using the langevin equation associated with the process , i.e. @xmath185 , where @xmath186 is a gaussian random variable with zero mean .
note that @xmath187 and @xmath188 are evaluated at the * prepoint * in the integral equation .
therefore , the wick - rotated path integral for @xmath189 is @xmath190 \ \exp [ - s \left[{\cal r}(t ) \right ] ] \ , \ ] ] where the measure @xmath191 = \lim_{m \rightarrow \infty } ( 4 \pi d \tau)^{-m d/2 } \
\mbox{\boldmath $ \omega$}_1 \cdots \mbox{\boldmath $ \omega$}_{m-1}$ ] , and the euclidean action @xmath192 = \int_0^t d t ' \left \ { \frac{1}{4 d } \sum_{i=1}^n ( \dot{x}_i^{\mu } - d^{\mu}[{\cal r}(t ' ) ] ) \
g_{\mu \nu}(i ) \ ( \dot{x}_i^{\nu } - d^{\nu}[{\cal r}(t ' ) ] ) + e_l[{\cal r}(t ' ) ] - e_t \right\ } \ .\ ] ] the integrand above represents a generalized onsager - machlup function @xcite , and the dot is a short - hand for time derivatives . in closing this section , we would like to mention that functional integral solutions for @xmath126 can be obtained from previous expressions after making the replacement @xmath193 .
_ interpretation of the quantum corrections . _ the general methodology we have developed so far , can be equally applied to other situations which do not necessarily involve a curved manifold such as , for instance , particles moving in a medium with position dependent diffusion constant . in order to adapt our previous formalism
, we need to understand qualitatively the origin of the quantum corrections to the short - time propagator obtained above . to this end
, we will illustrate the general idea with the following 1@xmath6 equation ( @xmath194 , @xmath195 ) @xmath196 the `` standard '' approach to finding the green s function for this problem is simply to solve the equation @xmath197 subject to the boundary condition @xmath198 .
we can do this by taking the fourier transform in @xmath199 : @xmath200 with the boundary condition @xmath201 .
this is more complex than the usual diffusion equation because we get a convolution of @xmath202 and @xmath203 .
however , we can find an approximate solution for @xmath203 , valid for small time , by noting that the boundary condition implies that , for small times @xmath156 , @xmath204 and so we have @xmath205 or , simply , @xmath206 , where @xmath207 is @xmath208 evaluated at the * prepoint*. for small times we can ignore the order @xmath156 term and solve for @xmath168 with the result @xmath209 $ ] .
taking the inverse fourier transform we finally have @xmath210 + { \cal o}(t^2 ) \ , \ ] ] which is just the plain green s function for a 1@xmath6 random - walk with @xmath208 evaluated at the * prepoint * and with * no quantum corrections*. this is , in fact , the result that the green s function for the jacobian times @xmath126 has no quantum corrections . to make things clear let us look at a different equation with the same boundary conditions @xmath211 which characterizes a free brownian particle in 1@xmath6 .
again , taking the fourier transform we get @xmath212 making the same approximation as above , replacing @xmath213 with @xmath203 in the integrand , making an error of @xmath214 , and noticing that @xmath215 we get @xmath216 $ ] .
when we take the inverse fourier transform the terms with derivatives of @xmath208 at @xmath217 give precisely the quantum corrections for this simple case @xmath218 + { \cal o}(t^2 ) \ .\ ] ] this is the result that the green s function for @xmath126 has quantum corrections .
the general case is just as simple , and the following rule emerges . given any second order differential equation ( first order in @xmath156 ) , no matter how many dimensions , with or without curvature , with or without a position dependent diffusion constant , the rule for obtaining the short - time green s function with everything evaluated at the prepoint is as follows : * bring all derivatives in each term all the way to the left of that term*. once this is done one can simply write down the green s function for a generalized diffusion process assuming the @xmath208 , @xmath187 , @xmath188 , whatever position dependent terms they may be , are constant and evaluated at the prepoint .
the quantum corrections are then seen to be simply those extra terms we get when commuting the derivatives to the left . _
fermion - phase problem : fixed - phase method .
_ it is evident that one can not make a probability density out of a complex and/or antisymmetric wave function .
this is the reason why we decided to write down fokker - planck equations for the distribution @xmath142 ( or @xmath126 ) and not @xmath117 .
nevertheless , the phase factor associated with the original complex distribution must show up in the evaluation of the expectation values .
it is well - known that this causes the variance of the computed results to increase exponentially with increasing number of degrees of freedom .
this problem is known as the fermion - phase @xcite catastrophe , and in this section we will review a method @xcite to obtain stable , albeit approximate , path - integral solutions whose stochastic determination has a polynomial , instead of exponential , complexity .
the generalization of the ideas presented below to bosons ( with complex - valued states ) or anyons in general is straightforward @xcite . in order to avoid cumbersome notation which would obscure the main ideas , here
we will consider a simplified version of our original hamiltonian @xmath219 where , for simplicity , we introduced the vector notation , @xmath220 , and in this subsection the same convention will be used for other bold quantities .
the microscopic equations governing the imaginary - time evolution of our interacting system can be found from a variational principle of the form @xmath221 = 0 $ ] , where the euclidean action is given by @xmath222 = \int_{0}^{t } d t ' \int_{{\cal m}^{n } } \mbox{\boldmath $ \omega$ } \left\ { \frac{1}{2 m^ * } \left [ { \bf \pi } \phi \right]^ * \cdot \left [ { \bf \pi } \phi \right ] + \widehat{v } \ \phi^ * \phi + % \hbar \ \phi^ * \partial_{t ' } \phi \right\ } \ . \phi^ * \partial_{t ' } \phi \right\ } \ .\ ] ] finding the states @xmath73 and @xmath223 which minimize @xmath224 $ ] is equivalent to solving the schrdinger equation and its complex conjugate .
equivalently , we can consider as independent real fields the phase and modulus of the wave function , i.e. @xmath225 and @xmath226 such that @xmath125 $ ] , and perform independent functional variations on @xmath224 $ ] .
the resulting euler - lagrange equations are : @xmath227
\widehat{\rm i\!h}_{\cal r } \phi \right\}$ } & = & \hat{h}_{fp } \ ; |\phi|\\ & & \\ % - \hbar \
( \partial_t \chi ) \
|\phi|^2 & = & - ( \partial_t \chi ) \
|\phi|^2 & = & \mbox { $ \im m \left\ { \exp \left[-i \chi \right ] \widehat{\rm
i\!h}_{\cal r } \phi \right\}$ } |\phi| & = & - \frac{\hbar}{2 } \
\mbox{\boldmath $ { \partial}_{\mu}$ } \cdot { \bf j}^{\mu } \label{onep } \end{array } \right . , \ ] ] where @xmath228 and @xmath229 stand for the real and imaginary parts of the expressions in brackets , respectively , @xmath230 and @xmath231 + \left [ { \bf \pi}^{\mu } \phi \right]^ * \phi \right ) = \frac{|\phi|^2}{m^ * } \
\left(\hbar \mbox{\boldmath $ \partial$}_{\mu } \chi({\cal r } ) - \frac{e}{c } \mbox{\boldmath $ a$}_{\mu } \right)\ ] ] is the probability current .
the singularities of @xmath124 occur at the zeros of @xmath73 , which generically have codimension two .
so far we have simply mapped the original fermion problem into a bosonic one for @xmath225 but still coupled to its phase fluctuations .
alternatively , one can regard this as a gauge transformation of the original fermion problem , whose effect is to add a non - local gauge field potential , @xmath232 , giving rise to a _ fictitious _ magnetic field .
notice that it is this gauge field that contains information on particle statistics .
moreover , although the geometry of @xmath124 can be altered by a gauge transformation , the singularities remain invariant . the fixed - phase ( fp ) method @xcite consists in making a choice for the phase , @xmath233 , and solving the bosonic problem for @xmath225 _ exactly _ using stochastic methods .
the method is _ stable _ and has the property of providing a variational upper bound to the exact ground energy @xmath97 , @xmath234 ( the equality holds when @xmath233 is the exact ground state phase ) , and for a given @xmath233 the lowest energy consistent with this phase .
the trial phases @xmath233 should conserve the symmetries of @xmath89 and particle statistics ( for time - reversal invariant systems there is a way for systematically improving a given mean - field phase using projection techniques @xcite ) .
notice that the fp method projects out the lowest energy state of a given symmetry .
therefore , the method allows one to compute also excitations which are `` ground states '' of a particular symmetry . for ground state properties of real symmetric hamiltonian operators
the fp approach reduces @xcite to the fixed - node method @xcite .
in this section we present an algorithm for computing the ground state properties of quantum many - body systems defined on a curved manifold with general metric @xmath39 .
as mentioned in the introduction , this can be accomplished by performing all multidimensional integrals using monte carlo techniques . in this way ,
ground state expectation values are obtained by averaging over a large number of particle configurations generated according to a certain limiting probability distribution @xmath235 .
there is some freedom in the choice of this distribution @xmath236 , however , to reduce statistical fluctuations in the observables to be computed it is more efficient to use the so - called _ importance - sampled _
distribution @xmath237 , which is the product of the absolute value of the solution of the time dependent schrdinger equation @xmath238 and some positive function @xmath239 that is the best available approximation to the modulus of the ground state eigenfunction . in a curved manifold , on the other hand , it is more convenient to work with the _ modified importance - sampled _ distribution @xmath240 , which is defined as a product of the conventional importance - sampled distribution @xmath237 and the metric ( see eq .
[ ff ] ) .
the propagation of particle configurations in time @xmath164 is determined by the conditional probability ( green s function ) @xmath241 , whose separation into a diffusion ( plus drift ) and branching parts ( see eq .
[ green ] ) makes it very simple to simulate numerically .
the gaussian term represents propagation according to the equation @xmath242 , where @xmath186 is a gaussian random variable with zero mean .
the effect of the term @xmath243 is to superimpose a drift velocity on the random diffusion process so that particle configurations are directed towards regions of configuration space where @xmath244 is large .
the branching term @xmath245 in eq .
[ green1 ] , determines the creation and annihilation of configurations ( walkers ) at the point @xmath246 after a move .
if the size of the ensemble of walkers at any time @xmath156 is defined as @xmath247 then , its rate of change is given by @xmath248 f({\cal r},t ) \ .\ ] ] therefore , if the local energy @xmath249 is a smooth function of @xmath162 , and the trial energy @xmath88 is suitably adjusted , the size of the ensemble of walkers will remain approximately constant as the configurations propagate . in particular ,
if the local energy is constant and equal to @xmath88 then the fluctuations in the ensemble size will vanish . to ease notation , in the rest of the paper
we will only consider the standard situation @xmath250 .
in such a case , ground state expectation values of a generic observable @xmath107 will be computed as @xmath251({\cal r})\ .\ ] ] in the following we present a step by step computational algorithm for implementing the stochastic approach discussed above .
most parts of the algorithm follow closely the standard dmc method , described for instance in @xcite , but we believe that it is useful to present these steps in detail here because a number of straightforward but subtle modifications due to the space curvature are involved . to be more specific ( without loosing the general features ) , let us present the algorithm as it is applied to fermionic systems within the fp approach .
_ algorithm .
_ @xmath252 1 . construct a guiding wave function @xmath103 , which is the best available approximation to the exact ground state ( or lowest energy state of a given symmetry ) . in principle any choice of @xmath103 which has finite overlap with the exact state is acceptable , but the more accurate @xmath103 is the faster the convergence to the stationary solution will be , and the lower the statistical fluctuations will be as well .
recall that zero variance is obtained when the guiding wave function is equal to the desired ground state ( and the ground state is bosonic ) .
@xmath252 2 .
given the guiding function compute the quantum drift velocity @xmath253 and the local energy @xmath254 , where @xmath255 is the fp hamiltonian defined in eq .
[ fixpham ] .
@xmath256 is used to construct the drift vector @xmath257 according to eq .
[ drift ] , and the drift and local energy are used in evaluating the short - time green s function according to eqs .
[ green ] , [ green1 ] , and [ green2 ] .
when a simple guiding function can be constructed , the expressions for the drift and the local energy can be evaluated analytically , but in most cases this must be done numerically .
@xmath252 3 .
a set of @xmath258 initial configurations or walkers @xmath259 ( @xmath260 ) is created , such that particles in each walker are distributed according to the _ modified importance - sampled _
distribution @xmath261 |\phi_g|^2 $ ] ( which is equivalent to using the variational monte carlo ( vmc ) technique ) .
@xmath252 4 .
each walker @xmath262 is diffused for a time @xmath164 according to the gaussian part of the propagator @xmath263 .
this can be accomplished by moving each particle coordinate @xmath264 according to @xmath265 where @xmath186 is a gaussian random variable with a mean of zero and a variance of @xmath266 .
@xmath252 5 .
the move from @xmath262 to @xmath267 is then accepted with a probability @xmath268 where @xmath269 \left| \frac{\phi_g({\cal r}^{'})}{\phi_g({\cal r } ) } \right|^2 \frac{g({\cal r}^{\prime } \rightarrow { \cal r } ; \tau)}{g({\cal r } \rightarrow { \cal r}^{\prime } ; \tau ) } \ .\ ] ] this step ensures detailed balance in the monte carlo procedure .
a typical acceptance ratio is in excess of 99% .
notice that if @xmath270 is the exact green s function , and not its short - time approximation , then @xmath271 is unity and this step is not necessary .
this is because the green s function of an hermitian operator is symmetric , i.e. @xmath272 , but this is not the case for any importance - sampled distribution function equation .
@xmath252 6 .
after all the particles of a given walker have been diffused from the initial position @xmath262 to the position @xmath273 and the move is accepted , then the values of the drift @xmath187 , the local energy @xmath137 , and @xmath103 are updated .
we could have equally well move all particles at once in step 4 before step 5 , however , the acceptance probability for a given time step is reduced considerably .
depending upon the particular problem one can adopt one of the two strategies : single or multiparticle moves .
@xmath252 7 .
the multiplicity ( weight ) @xmath274 of a given walker , is computed from the branching part of the green s function eq .
[ green1 ] @xmath275 where @xmath276 because some of the moves can be rejected .
if the mean - squared distance the particles would diffuse in the absence of the rejection step is @xmath277 , and the actual mean - squared distance is @xmath278 then the actual time used in a branching step @xmath279 . in the case of multiparticle moves , the effective time step must be calculated separately from a computation of the accepted to attempted moves . since @xmath274 is , in general , not an integer one can use instead the integer @xmath280 where @xmath281 is a random number uniform in the range @xmath282 $ ] . in this case the average density of walkers is conserved and @xmath283 . if @xmath284 then the walker is deleted from the ensemble ; otherwise @xmath285 copies of the configuration are made and added to the ensemble .
note that fixing @xmath286 is equivalent to eliminate branching and , consequently , to perform a vmc calculation with limiting distribution @xmath287 |\phi_g|^2 $ ] .
@xmath252 8 . after diffusing and branching all walkers in time @xmath164
the mean value of the local energy over all walkers is computed from the obtained distribution .
one has to start averaging over these values , after some target ( equilibration ) time , when the configurations are sampled according to the limiting stationary distribution @xmath288 .
the target time depends upon the particular problem and how close @xmath103 is to the exact state .
@xmath252 9 . using the average value of the local energy over the whole ensemble of walkers @xmath289 ,
the trial energy @xmath88 is updated according to @xmath290 .
mixing this estimate with an old value of the trial energy , allows one to improve the convergence .
@xmath252 10 . after computing the average energy over a sufficiently long time ( @xmath291 , where @xmath292 is the number of moves per configuration ) ,
its value is stored and the first block is completed .
@xmath292 should be large enough for there to be little statistical correlation between energy subaverages obtained in different blocks .
a new ensemble of @xmath258 walkers is generated by randomly copying or deleting configurations , and steps 4 to 9 are repeated completing another block .
one has to do as many blocks as it takes in order to reach the desired statistical accuracy .
notice , however , that because the propagator is only accurate to @xmath179 , the distribution @xmath293 and the resulting estimates will have a time step bias .
the way to eliminate this bias is by extrapolating all computed expectation values to @xmath294 .
figure [ fig0 ] shows a schematic flow diagram of the algorithm presented in this section .
as an example application of the method we have developed in the previous sections we consider the problem of a single particle of charge @xmath37 , mass @xmath36 and vector position @xmath295 in @xmath296 confined to the surface of a sphere of radius @xmath297 centered at the origin ( @xmath9 = s@xmath7 , @xmath195 ) moving in the presence of the vector potential of a dirac monopole at the origin . this problem can be solved in closed form and so constitutes an ideal model system for testing the accuracy of the stochastic solutions we can derive using the formalism developed in previous sections . moreover , the case of @xmath35 interacting electrons confined to s@xmath7 in the presence of a monopole field serves as the basic model which captures the essential physics of the quantum hall effect @xcite .
the pauli hamiltonian for a spinless particle in s@xmath7 is @xmath298 where @xmath299 , and @xmath24 is the monopole vector potential ( @xmath300 , @xmath301 being the strength of the radial field . ) therefore , the total number of flux quanta @xmath302 piercing the surface of the sphere is given by @xmath303 , where @xmath304 is the elementary flux quantum .
following wu and yang @xcite we can construct angular momentum operators @xmath305 in terms of which the hamiltonian reads @xmath306 if we choose a gauge where the vector potential is @xmath307 , then the hamiltonian , eq .
[ mono ] , can be written as @xmath308 \ , \ ] ] in terms of the usual spherical angles @xmath309 and @xmath310 ( @xmath311 , @xmath312 , see fig .
[ fig1 ] . )
the eigenstates of this hamiltonian are monopole harmonics ( normalized to 1)@xcite @xmath313 \ u^{{\cal s}+m } \
v^{{\cal s}-m } \ { \cal f}(|u|,|v| ) \ ,
\nonumber \\ { \cal f}(|u|,|v| ) & = & \sum_{k=0}^{n } ( -1)^k { n\choose k } { { 2{\cal s}+n } \choose { { \cal s}+n - m - k } } ( v \bar{v})^{n - k } ( u \bar{u})^{k } \ , \nonumber \\ { \cal n}_{{\cal s}nm }
& = & \left(\frac{2{\cal s}+2n+1}{4 \pi } \
\frac{({\cal s}+n - m ) ! \ ( { \cal s}+n+m)!}{n ! \
( 2{\cal s}+n ) ! } \right)^{1/2 } \ , \end{aligned}\ ] ] where @xmath314 $ ] and @xmath315 $ ] are spinor coordinates , @xmath316 is the landau level quantum number , and @xmath317 is the ( @xmath318 ) angular momentum quantum number which labels degenerate states within the @xmath319 level . in the sum above the binomial coefficient @xmath320 vanish when @xmath321 or @xmath322 . for a given @xmath323 and @xmath324 ,
the ground state ( @xmath325 ) and first excited state ( @xmath326 ) are @xmath327 \
\psi_{gs } \ , \end{aligned}\ ] ] respectively .
the energy of a state with landau level quantum number @xmath316 is given by @xmath328 where @xmath329 is the cyclotron frequency ( @xmath330 ) .
let us now reformulate the electron - monopole problem in a way consistent with the notations introduced in the previous sections .
in this way one can compare the exact result to the numerical one obtained with the algorithm developed in this paper , thus testing the numerical technique .
first , instead of the spherical angles @xmath309 and @xmath25 we introduce new coordinates @xmath331 and @xmath332 , where @xmath333 $ ] , and @xmath332 is its complex conjugate .
geometrically , this transformation can be viewed as a stereographic projection of the sphere onto the plane , as illustrated in fig .
the hamiltonian can be rewritten as @xmath334 in terms of the ( non - hermitian ) canonical momenta @xmath335 and @xmath336 , and @xmath337 with metric tensor @xmath338 naturally , the particles moving in the projected plane are in a space with curvature , corresponding to that of the sphere .
notice that the metric tensor is diagonal when written in terms of ( @xmath339 ) , such that @xmath340 , i.e. @xmath341 ( i.e , it corresponds to the conformal gauge ) .
then , the drift is simply @xmath342 .
the stochastic method developed above allows one to obtain the exact energy eigenvalues of the electron - monopole problem in s@xmath7 iff we know the exact phase of the eigenfunctions . in other words , if the trial state is chosen such that it has the exact ground state phase , then independently of its modulus our stochastic approach will lead to the exact ground state energy .
similarly , if the trial state has a phase corresponding to an excited state eigenfunction then we will obtain the exact excited state energy eigenvalue . in the next two subsections we construct simple trial states for the ground and first excited states of the one particle problem .
their modulus are then used as guiding functions @xmath103 .
using these trial states we will apply our technique and illustrate the main ideas of our method . _ ground state _
@xmath343 : the @xmath344 degenerate ground states of the electron - monopole system are labeled by their @xmath345 $ ] angular momentum quantum numbers @xmath346 .
here we consider the @xmath347 ground state for which the exact ( unnormalized ) wave function can be written @xmath348 but instead we have constructed the following two trial states @xmath349 and @xmath350 where @xmath351 and @xmath352 are real valued constants .
the trial states @xmath353 and @xmath354 have been constructed so that for @xmath355 and @xmath356 they are both equal to the exact ground state , @xmath357 . for @xmath358 ,
the modulus of @xmath353 is no longer equal to that of the exact ground state , but the phase of the wave function is _ exact _ , i.e. @xmath359 in contrast , for @xmath360 , the modulus of @xmath354 is exact , but the phase is approximate , @xmath359 it follows that if @xmath353 is used as a trial state in a fp dmc simulation the resulting energy will be the _ exact _ ground state energy @xmath361 , while if @xmath354 is used as a trial state the simulation will not lead to the exact ground state energy , but instead will provide a variational upper bound .
as has already been emphasized the trial state used in a fp dmc simulation should be constructed to be the _
best available _ approximation to the exact eigenstate , since the quality of the trial state can greatly influence the speed of convergence and the statistical accuracy of the result of the simulation . this can be clearly illustrated by considering the trial state @xmath354 .
since the modulus of @xmath354 is exact the drift velocity @xmath362 which depends only on the modulus of the trial state will correspond to the exact drift velocity and in the absence of the branching term will lead to the exact density distribution .
it is straightforward to show that @xmath363 where @xmath364 , indicating that walkers are guided away from the regions where the wave function is small and , in this way , the particle tends to spend most of the time near the top of the sphere ( @xmath365 ) .
a potential problem appears when we consider the local energy , @xmath366 \end{aligned}\ ] ] which , of course , is not exact due to the approximate phase of the trial state . in particular
, @xmath137 _ diverges _ as @xmath367 .
as we have just shown , the drift will tend to push the particle towards @xmath368 , leading to large fluctuations in the local energy .
this in turn can lead to huge fluctuations in the population size ( number of walkers ) , since the size of the population depends exponentially on the local energy .
thus , in this particular example , one has to take small values for @xmath352 ( @xmath369 ) in order to assure fast convergence and good statistical accuracy .
figure [ fig2 ] shows the results of fp dmc simulations , using the algorithm developed in this paper , for the difference between computed and exact ground state energies for trial state @xmath353 ( circles ) and trial state @xmath354 ( squares ) using different values of the time step @xmath164 .
the @xmath370 extrapolated values are also shown . for trial state @xmath353 we used @xmath371 ,
the number of walkers was chosen to be @xmath372 , the number of monte carlo steps per walker was @xmath373 , and the acceptance rate was between @xmath374 and @xmath375 . for trial state @xmath354 we used @xmath376 ,
the number of walkers was @xmath377 , the total number of monte carlo steps per walker was @xmath378 , and the acceptance rate was the same as for @xmath353 . since the parameter @xmath352 in @xmath354 was chosen so that @xmath379 , the trial state and the corresponding green s function were nearly exact and we were able to reach reasonable statistical accuracy with a relatively small number of monte carlo steps .
as expected we find that when trial state @xmath353 is used the extrapolated energy agrees within statistical accuracy with the exact result , but when we use trial state @xmath354 , for which the phase is not exact , we obtain a variational upper bound for the exact ground state energy . in fig .
[ fig3 ] the density profiles for the exact ground state @xmath380 ( exact ) , the trial state @xmath381 ( vmc ) , the density obtained in fp dmc with trial state @xmath381 at time step @xmath382 ( fp mixed estimator ) , and the extrapolated density defined as ratio of the square of fp density to the variational density corresponding to @xmath381 , are shown . note that since the density in our dmc calculation is determined as a mixed estimate ( see eq .
[ mixede ] ) , and the density operator does not commute with the hamiltonian between the dmc solution and the trial state , the corresponding density profile ( fp mixed estimator ) improves on the variational result but still differs from the exact density .
the extrapolated estimator for the density constructed by combining both , the fp mixed estimator and the variational density makes it possible to improve on the fp density and is seen to be very close to the exact result . _
first excited state _
@xmath383 : as a further demonstration of the validity of our method we turn to the first excited state of the electron - monopole system .
again , we specify the @xmath318 angular momentum quantum number and take @xmath384 for which the exact excited state wave function is @xmath385 and @xmath386 with the property that for @xmath355 and @xmath356 they each reduce to the exact excited state , @xmath387 . as before , for @xmath358 and @xmath360 the modulus of @xmath353 is approximate and the phase is exact ( @xmath388 , @xmath389 ) and the modulus of @xmath354 is exact and the phase is approximate ( @xmath390 , @xmath391 ) .
figure [ fig4 ] shows the difference between fp dmc energies computed using trial states @xmath353 and @xmath354 and the exact excited state energy for different values of time step @xmath164 as well as the extrapolated @xmath392 result . the parameters ( number of walkers , number of monte carlo steps , etc .
) used for these simulations were the same as those used for the ground state simulations except that we took @xmath393 in @xmath354 . again
, our simulations gave the expected results when the phase of the trial state is exact we obtain the exact energy ( circles ) , @xmath394 , and when the phase is approximate we obtain a variational upper bound on that energy ( squares ) . in figure [ fig5 ] the density profiles corresponding to the trial state @xmath395 ( vmc ) , the fp density using the same state at time step @xmath396 ( fp mixed estimator ) , the extrapolated density , computed as above by taking the ratio of square of the fp density and the vmc density ( extrap .
estimator ) , and the exact density ( exact ) .
the results are qualitatively similar to those for the ground state
the fp estimator improves on the vmc result , and the extrapolated density is nearly equal to the exact excited state density .
the simulation results presented in this section provide a simple test of both the fp dmc method and the method developed in this paper for dealing with quantum corrections due to curvature .
the results clearly show that these methods can be used to study quantum systems on curved manifolds .
in this paper we have introduced a stochastic method to solve the many - body schrdinger equation on curved manifolds .
this method is essentially a generalized diffusion monte carlo ( dmc ) technique allowing one to deal with the effects arising from the space curvature .
we have shown that due to the curvature the diffusion matrix and drift vector , which appear in the green s function used as a conditional probability in dmc simulations , acquire additional terms , the so - called quantum corrections .
the explicit expressions for a general metric tensor are worked out in detail . since the presence of the curvature leads to a number of other nontrivial modifications , we have presented a step by step algorithm which can be used to implement a code dealing with dmc simulations in curved space .
it is worth emphasizing that our method can be applied to a wide variety of inhomogeneous systems ( e.g. , inhomogeneous semiconductors with a position dependent effective mass ) , not just systems on curved manifolds .
the reason for this , as discussed in section iii , is that the quantum corrections to the green s function can be interpreted as being due to those terms which appear in the generalized diffusion equation describing the system once each of the derivatives in that equation have been commuted all the way to the left of each term .
this definition of quantum corrections is quite general and can be applied to _ any _ differential equation which is second order in space and first order in time , regardless of the number of dimensions or any spatial inhomogeneity in the system . to illustrate the general methodology we have concentrated on the problem of interacting fermions in external electromagnetic potentials . in this case a variational upper bound to the exact ground state energy can be found by applying the fixed - phase approximation , where the fermionic problem is treated as a bosonic one by fixing the phase of the many - body wave function ( which is complex - valued in general ) by some trial phase . as an example , we have considered the problem of a single electron confined to the surface of a two - sphere , which has a magnetic monopole at its center .
the electron thus moves in a space with curvature in the presence of a magnetic field which breaks time - reversal symmetry .
this simple problem can be solved in closed form and , therefore , we have used it as a playground for testing our technique . in the paper we have presented two calculations , where the ground and first excited state energies are computed using the exact phases , but approximate modulus for the corresponding guiding functions .
we have shown that the exact energies are reproduced within statistical accuracy thus proving that the approach for dealing with the quantum corrections is valid .
as emphasized in the introduction , the method presented in this paper for performing dmc simulations on curved manifolds can be used to study many interesting physical systems .
an important example is the quantum hall effect , a phenomenon which occurs when a two - dimensional electron system is placed in a strong magnetic field .
as first pointed out by haldane @xcite , the electron - monopole system described in section v provides a convenient geometry for performing finite size numerical studies of quantum hall systems when many interacting electrons are placed on the sphere .
this is in part because the spherical geometry has no boundary so that finite size effects are suppressed .
in addition , the spherical geometry is conceptually simpler than the ( flat metric ) torus geometry , which also has no boundary , because the topological order exhibited by quantum hall states leads to certain nontrivial degeneracies on the torus @xcite .
recently we have used the method developed in this paper to study some of the exotic excitations which occur in quantum hall systems , specifically the fractionally charged quasiparticle excitations of the fractional quantum hall effect @xcite , and the charged spin texture excitations ( skyrmions ) of the integer quantum hall effect @xcite .
previous numerical studies of these excitations have been based on either vmc or exact diagonalization calculations which , for the most part , have assumed that the wave functions describing the excitations are confined to the lowest ( @xmath325 ) landau level .
in fact , this is a rather poor approximation for real experimental systems which can exhibit significant mixing of higher landau levels due to the electron - electron interaction . because the fp dmc method allows one to go beyond the lowest landau level approximation
it can be used to study the effect of landau level mixing on quantum hall states @xcite and , by employing the method described in this paper , we were able to perform such studies using haldane s spherical geometry .
these simulations have provided useful quantitative results for various properties of quantum hall excitations for realistic experimental parameters @xcite . along with the more rigorous test case of the electron - monopole system described in this paper
, these calculations of landau level mixing effects in quantum hall systems using the haldane sphere have shown that the method we have developed for performing fp dmc calculations on curved manifolds works well and can be used to study many other interesting physical systems .
we acknowledge stimulating discussions with david ceperley .
vm is supported by nsf grant number dmr-9725080 .
neb is supported by us doe grant number de - fg02 - 97er45639 .
work at los alamos is sponsored by the us doe under contract w-7405-eng-36 . | a generalized stochastic method for projecting out the ground state of the quantum many - body schrdinger equation on curved manifolds is introduced .
this random - walk method is of wide applicability to any second order differential equation ( first order in time ) , in any spatial dimension .
the technique reduces to determining the proper `` quantum corrections '' for the euclidean short - time propagator that is used to build up their path - integral monte carlo solutions . for particles with fermi statistics
the `` fixed - phase '' constraint ( which amounts to fixing the phase of the many - body state ) allows one to obtain stable , albeit approximate , solutions with a variational property .
we illustrate the method by applying it to the problem of an electron moving on the surface of a sphere in the presence of a dirac magnetic monopole . |
the @xmath1- meson plays a unique role in the study of the hot and dense medium created in relativistic heavy - ion collisions .
it is is the lightest bound state of hidden strangeness @xmath4 , has a small interaction with other non - strange hadrons and hence carries information from the early partonic stages of the system evolution . comparing the elliptic flow ( @xmath5 ) of @xmath1 to the @xmath5 of other multistrange hadrons ( @xmath6 and @xmath7 ) or particles composed of lighter quarks ( @xmath8 and @xmath9 ) or heavier charm quark , provides information about the partonic collectivity of the medium .
furthermore the @xmath1 can provide important information on particle production mechanisms , since it is a meson but has a mass similar to @xmath10 and @xmath11 baryons .
the measurement of its nuclear modification factor , @xmath12 adds to the picture of particle suppression and its dependence on particle mass and composition supporting hydrodynamics and recombination models .
the @xmath1 can also be sensitive to the restoration of chiral symmetry .
a certain fraction of the @xmath1 can decay inside the hot and dense media leading to a change in its spectral function@xcite .
this modification can be seen by studying the low - momentum @xmath1 decaying inside the media and reconstructed via the di - electron decay channel .
since leptons are not subject to the strong interaction , they preserve their production information .
a change in mass or width ( @xmath13 ) of @xmath1 inside the medium can lead to a change in the relative branching ratios of the @xmath14 and @xmath15 decay modes . since @xmath16
, small changes in @xmath1 or @xmath17 can induce significant changes in the branching ratio .
r0.5 the phenix detector@xcite at rhic ( relativistic heavy ion collider ) has been designed to measure both leptons and hadrons . a schematic view of the phenix detector is shown in fig .
[ fig : fig_phenix ] .
each of the two central arm spectrometers covers 90@xmath18 in azimuth and @xmath190.35 in pseudorapidity and has the capability to measure neutral and charged particles .
the high - resolution multi - wire proportional drift chambers ( dc ) together with the first layer of pad chambers ( pc1 ) provide the charged particle tracking and momentum measurement .
the typical momentum resolution is @xmath20 1.1@xmath21 .
the kaons are identified by using the timing information from a high resolution time of flight ( tof ) detector and the lead scintillator ( pbsc ) part of the electromagnetic calorimeter ( emcal ) , with good @xmath22 separation over the momentum range 0.3 - 2.5 gev/_c _ and 0.3 - 1 gev/_c _ , respectively .
the electrons are identified using a ring imaging @xmath23erenkov detector ( rich ) and by requiring the energy measured in the emcal to match the measured momentum of the charged tracks in the dc .
the zero degree claorimeters ( zdc s ) and beam beam counters ( bbc s ) are dedicated subsystems that measure global quantities such as luminosity , collision vertex and event centrality .
the minimum bias trigger is derived by a coincidence between the two bbcs ; in @xmath0 and @xmath2 the trigger requires at least one hit in each bbc arm whereas for @xmath3 at least two hits in each bbc arm and one detected neutron in zdc is needed . in order to benefit from the high luminosity in @xmath0 and @xmath24 collisions and to efficiently detect electrons , a special online emcal rich trigger ( ert ) is used .
it requires an event to have at least one track with an energy above a certain threshold in the emcal and a geometrically correlated hit in the rich .
the results presented here correspond to the data for @xmath0 ( 2005 ) , @xmath2 ( 2003 ) and @xmath3 ( 2004 ) taken at @xmath25 = 200 gev @xmath25 = 62.4 gev .
for the @xmath26 mass reconstruction , the charged tracks are combined to form pairs using three different techniques .
the first one does not require any kaon identification and assigns the kaon mass to all tracks ( `` no kaon pid '' ) .
the second method requires one track to be identified as kaon in the tof subsystem ( one kaon``pid '' ) , whereas the third method requires both the tracks to be identified as kaons either in tof or emcal subsystems ( `` two kaon pid '' ) .
the `` no kaon pid '' method allows us to go to high @xmath27 as compared to the other two , is a relatively simpler analysis , but with comparatively large b / s ratio .
the `` two kaons pid '' method allows us to go lower in @xmath27 and has small b / s ratio .
the `` one kaon pid '' method has the advantage of less background and so works with better accuracy for the low @xmath27 ( @xmath28 gev/_c _ ) region where `` no kaon pid '' method has a large background .
the @xmath0 data was analyzed using `` no kaon pid '' , and `` one kaon pid '' , @xmath24 using `` no kaon pid '' and `` two kaons pid '' and @xmath3 using `` one kaon pid''and `` two kaons pid '' .
the 62.4 gev @xmath3 data was analyzed using `` two kaons pid '' method only .
the different analysis methods have very different sources of systematic uncertainities and provide a valuable consistency check . in fig .
[ fig : fig_inv_spec_all - a ] , good agreement between the various methods can be seen .
the combined @xmath0 result using `` one kaon pid '' and `` no kaon pid '' analyses constitutes a new @xmath0 reference for @xmath1-meson , surpassing the previous one@xcite , in @xmath27 and with smaller errors . for @xmath29 ,
electrons identified using rich and emcal are combined in pairs to generate like- and unlike - sign mass spectra . however , due to the limited azimuthal angular acceptance and the strong magnetic field beginning at r=0 , the identification and rejection of @xmath30 pairs from dalitz decays and photon conversions is very difficult@xcite .
this results in a huge combinatorial background in @xmath3 , making this measurement difficult . raw yields , for both @xmath31 and @xmath30 are then extracted either by simultaneously fitting the signal and background , or by integrating the spectra in the vicinity of the peak after subtraction of the combinatorial background , estimated using an event - mixing technique@xcite .
the correction function takes into account the limited detector acceptance and resolution and reconstruction efficiency and was derived by a full single - particle monte carlo simulation , propagated through an emulator of the phenix detector and the full analysis chain .
corrections to incorporate the trigger efficiency and multiplicity effects are applied too .
[ fig : fig_inv_spec_all ] summarizes the @xmath1 measurements in @xmath0 , @xmath24 and @xmath3 collisons via di - kaon and di - electron decay modes .
the integrated yields and temperature presented in fig . [ fig : fig_dndy_t ] were extracted assuming @xmath32 dependent yields to be exponential and so fitting the @xmath32 spectra to the following exponential function with dn / dy and t as the parameters .
@xmath33 the temperatures extracted in the two decay channels are in good agreement with each other , do not change between different collision systems but grow only slighlty from 62.4 gev to 200 gev as can be seen in figs .
[ fig : fig_dndy_t - a ] & [ fig : fig_dndy_t - c ] . the integrated yields normalized to the number of participants
increase by almost a factor of two from peripheral to central @xmath3 collisions .
particle yields in the @xmath30 channel seem to be higher than the corresponding @xmath31 decay channel , but the large systematic and statistical uncertainties in di - electron measurement prevent us from making any conclusive statement .
the @xmath5 of @xmath1 was extracted following the @xmath34 method@xcite , using the @xmath35 decay mode .
a detailed description of the analysis procedure can be found in @xcite .
results for @xmath5 as a function of @xmath36 for the @xmath1-meson together with @xmath37 are shown in the left ( right ) panel of fig .
[ fig : fig_v2 ] without ( with ) scaling with the number of valence quarks .
it is clear from the left panel that despite its similar mass to the proton , @xmath38 for the @xmath1- meson follows the flow pattern of the other lighter mesons(@xmath39 and @xmath17 ) .
the right panel of the fig .
[ fig : fig_v2 ] shows universal scaling behavior of the elliptic flow of baryons and mesons when both @xmath5 and @xmath36 are scaled with the number of valence quarks .
this indicates that the elliptic flow develops during the early stage when the constituents of the flowing medium are partons and not ordinary hadrons interacting with their standard hadronic cross - sections .
vs @xmath36 for identified particles@xcite in mid - central ( 20 - 60% ) @xmath3 collisons .
( b ) @xmath40 vs @xmath41 for the same set of particles , height=288 ] the nucleus - nucleus collisions at rhic have revealed a decrease in hadron production at high @xmath27 relative to scaled @xmath0 collisions .
the nuclear medium effects on hadron production are quantified by the nuclear modification factor @xmath12 , defined as the ratio of the yield obtained from nucleus - nucleus collisions scaled down with the number of binary collisions , to the yield from elementary nucleon - nucleon collisions .
@xmath42 in the absence of any nuclear effects , the ratio should saturate at unity for high @xmath27 , where the production is dominated by hard scattering and is proportional to the number of binary collisions ( @xmath43 ) , in contrast to low @xmath27 region where soft processes dominate .
@xmath44 , which is the nuclear modification factor for @xmath2 collisions helps to differentiate initial state effects from final state effects and also to study other cold nuclear matter effects .
fig.[fig : fig_rda ] shows @xmath44 measurement of the @xmath1 for 0 - 20% and minimum bias collisions plotted together with @xmath45 and @xmath46 mesons for comparison .
whereas the @xmath44 of @xmath39 and @xmath47 are consistent with unity , the @xmath44 of @xmath1 for 0 - 20% shows some enhancement .
but with very large error bars . of @xmath48
mesons@xcite . ]
fig.[fig : fig_raa]@xcite shows the nuclear modification factor for the @xmath1-meson in central @xmath3 collisions using the new @xmath0 reference from 2005 run results , over a @xmath27 range of 2.45 - 7 gev/_c_. for comparison , the @xmath12 of direct @xmath49 are also plotted . @xmath50 and
@xmath47 follow the same suppression pattern over the entire @xmath27 range .
the @xmath1-meson at intermediate @xmath27(@xmath51 gev/_c _ ) exhibits more suppression than the protons but less suppression than @xmath47 and @xmath50 mesons , whereas at higher @xmath27 ( @xmath52 gev/_c _ ) the amount of suppression of @xmath1 and @xmath53 could be similar to that of @xmath50 and @xmath47 .
the similar suppression patterns of various mesons at high @xmath27 support the concept of meson production via jet - fragmentation at high @xmath27 outside the hot and dense medium created in the collisions . in the future
, the comparison of extended measurements of charged kaons with that of @xmath1 will give an insight into , whether or not quark flavor composition plays a role on the suppression pattern .
in central @xmath3 collisions as a function of @xmath27 for @xmath54 and direct @xmath55.,width=384 ]
the phenix experiment at rhic has measured @xmath1-meson production using both @xmath30 and @xmath31 in @xmath0 , @xmath24 and @xmath3 collisions at @xmath25 = 200 gev .
the preliminary results in @xmath0 via both channels agree to each other .
the measurements in @xmath24 suffer from low statistics in both the channels . with the high statistics @xmath24 data collected in year 2008 , results with better precision and wide @xmath27 coverage are expected . in @xmath3 , the @xmath1-meson has been measured via @xmath31 with good accuracy while @xmath30 measurements are expected to improve considerably with the newly installed hadron blind detector@xcite in phenix .
the @xmath5 of @xmath1 follows the flow pattern of other mesons as well as scales with @xmath56 = 2 , indicating partonic collectivity .
@xmath44 of @xmath1 does not show any suppression , but large error bars leave room for cronin enhancement .
the @xmath12 of @xmath1 shows a similar suppression pattern to that of @xmath50 and @xmath47 at high @xmath27 , but at intermediate @xmath27 , the suppression level is different .
the author acknowledges support by the israel science foundation , the minerva foundation and the nella and leon benoziyo center of high energy physics research .
10 s. pal , c.m .
ko and z. lin , nucl .
a707 , ( 2002 ) 525 .
d lissauer et al , 1991 phys
b253 15 - 18 .
k. adcox et al . , nucl .
instrum . meth . a 499 ( 2003 ) 469 .
v. ryabov , 2006 nucl .
a774 735 - 738 .
a. kozlov , nucl.phys .
a774 ( 2006 ) 739 - 742 d.pal et al . , nucl .
phys . a 774 , 489 ( 2006 ) .
a. kozlov , eur .
phys . j. a 31
, 836 - 841(2007 ) n. borghini and j. y. ollitrault , prc c70 , 064905(2004 ) prl 99 , 052301 ( 2007 ) prl 98 , 162301 ( 2007 ) .
v. ryabov , j.phys.g35:044030,2008 .
m. naglis , arxiv:0809.3557(nucl - ex ) i. rainovich , nucl.phys.a774:903-906,2006 a. kozlov , i. ravinovich , l. i. shekhtman , z. fraenkal , m. inzuka and i. tserruya , nim a523 , 345(2004)[arxiv : physics/0309013 ] .
z. fraenkel et al .
a546 , 466(2005 ) [ arxiv : physics/0502008 ] . | light vector mesons are among the most informative probes to understand the strongly coupled quark gluon plasma created at rhic .
the suppression of light mesons at high transverse momentum , compared to expectations from scaled @xmath0 results , reflects the properties of the strongly interacting matter formed .
the @xmath1-meson is one of the probes whose systematic measurement in @xmath0 , @xmath2 and @xmath3 collisions can provide useful information about initial and final state effects on particle production .
the mass , width and branching ratio of the @xmath1-meson decay in the di - kaon and di - electron decay channels could be modified in @xmath3 collisions due to the restoration of chiral symmetry in the qgp .
the phenix experiment at rhic has measured @xmath1-meson production in various systems ranging form @xmath0 , @xmath2 to @xmath3 collisions via both its di - electron and di - kaon decay modes .
a summary of phenix results on invariant spectra , nuclear modification factor and elliptic flow of the @xmath1-meson are presented here . |
we consider two connected and compact @xmath1-riemannian manifolds @xmath2 and @xmath3 with @xmath0-boundaries @xmath4 and @xmath5 , respectively .
we assume @xmath6 and @xmath7 . throughout of the present paper ,
we assume that * ( a-1 ) * @xmath8 and @xmath9 have a common boundary @xmath10 . @xmath11 is a disjoint union of a finite number of connected and closed components .
the metrics satisfy @xmath12 on @xmath11 .
we will add some other assumptions for @xmath13 and @xmath14 in a neighborhood of the boundary later .
note that we need our geometric assumptions only in some small neighborhoods of the boundary , in particular , we do not assume that @xmath8 and @xmath9 are diffeomorphic outside of a small neighborhood of the boundary .
let @xmath15 , @xmath16 , be the ( negative ) laplace - beltrami operator on each @xmath17 .
we consider the following interior transmission eigenvalue ( ite ) problem : @xmath18 where each @xmath19 , @xmath16 , is strictly positive on @xmath17 .
we call @xmath20 the index of refraction on @xmath17 .
if there exists a pair of non - trivial solutions @xmath21 of ( [ s1_eq_ite1])-([s1_eq_ite3 ] ) , we call corresponding @xmath22 an _ interior transmission eigenvalue_. ite problems naturally appears in inverse scattering problems for acoustic wave equations on @xmath23 with compactly supported inhomogeneity . in @xmath23 for @xmath24 , time harmonic acoustic waves satisfy the equation @xmath25 where @xmath26 is strictly positive in a bounded domain @xmath27 with a suitable smooth boundary , and @xmath28 . given an incident wave @xmath29 with an incident direction @xmath30 and energy @xmath31 , the scattered wave @xmath32 is described by the difference between the total wave @xmath33 and the incident wave @xmath34 where @xmath33 is the solution of ( [ s1_eq_scatteringeq ] ) satisfying the following asymptotic relation : as @xmath35 @xmath36 here the second term on the right - hand side is the spherical wave scattered to the direction @xmath37 .
the function @xmath38 is the scattering amplitude .
the _ s - matrix _ is given by @xmath39 where @xmath40 is an integral operator with the kernel @xmath41 .
then the s - matrix is unitary operator on @xmath42 . if there exists a non zero function @xmath43 such that @xmath44 i.e. @xmath45 , we call @xmath46 a _ non - scattering energy_. if @xmath47 is a non - scattering energy , we have that @xmath48 vanishes outside of @xmath27 from the rellich type uniqueness theorem ( see @xcite and @xcite )
. hence we can reduce to the ite problem @xmath49 with @xmath50 and @xmath51 .
if @xmath52 is a non - scattering energy , @xmath53 is also an ite of the system ( [ s1_eq_ite1euq])-([s1_eq_iteeuq ] ) .
ite problems were introduced in @xcite and @xcite in the above view point .
for the schrdinger equation @xmath54 with a compactly supported potential @xmath55 which satisfies @xmath56 in @xmath57 , we can state the ite problem similarly .
recently , @xcite introduced the ite problems on unbounded domains , considering perturbations which decrease exponentially at infinity .
the system ( [ s1_eq_ite1euq])-([s1_eq_iteeuq ] ) is some kind of non - self - adjoint problem .
moreover , we can construct a bilinear form associated with this system , but generally this bilinear form is not coercive .
note that the @xmath58-coercivity approach is valid for some anisotropic cases i.e. @xmath59 is replaced by @xmath60 where @xmath61 is a strictly positive symmetric matrix valued function and @xmath62 .
for the @xmath58-coercivity approach on this case , see @xcite .
another common approach is to reduce an ite problem to an equivalent forth - order equation . for ( [ s1_eq_ite1euq])-([s1_eq_iteeuq ] )
, we can reduce to @xmath63 which is formulated as the variational form @xmath64 for any @xmath65 .
there are also many works on this approach for acoustic wave equations and schrdinger equations . for more history , technical information and references on ite problems ,
we recommend the recent survey by cakoni and haddar @xcite . in this paper ,
we slightly generalize the isotropic type " ite problem as a system on two domains @xmath8 and @xmath9 . since we do not assume that @xmath66 and @xmath9 are diffeomorphic , it is difficult to use the forth - order equation approach ( see figure [ s1_fig_manifold ] ) . and @xmath9 with common boundary @xmath67.,width=377 ] moreover , in view of assumptions ( a-1 ) and ( a-2 ) which is added in 2.3 , the ite problem is not elliptic , and we can not construct a suitable isomorphism @xmath58 such that the system ( [ s1_eq_ite1])-([s1_eq_ite3 ] ) is @xmath58-coercive . therefore ,
neither the variational formulation approach nor the @xmath58-coercivity approach are valid for the proof of discreteness of ites in our case
. then we adopt arguments by lakshtanov and vainberg @xcite and @xcite in the present paper .
they considered isotropic cases in @xcite and anisotropic cases in @xcite in bounded domains in @xmath23 .
the approach in @xcite is based on methods of elliptic pseudo differential operators on the boundary and its application to the dirichlet - to - neumann ( d - n ) map .
we should also mention about @xcite and @xcite . recently
, they proved the weyl s asymptotics including complex ites and evaluated ite - free regions in the complex plane under various conditions .
they used the semi - classical analysis for the d - n map associated with an operator of the form @xmath68 where @xmath69 are smooth and positive valued function on a bounded domain @xmath70 . even for manifolds ,
these d - n map approaches are valid .
more precisely , in this paper , we construct the poisson operator and the associated d - n map as elliptic singular integro - differential operators and we can compute exactly their symbols . using the ellipticity of the d - n map and the analytic fredholm theory , we can prove the discreteness of the set of ites .
moreover , we consider a weyl type lower bound of the number of positive ites except for a small neighborhood of the origin .
a case which we can use the @xmath58-coercivity approach will be studied in the forthcoming paper @xcite .
the plan of the paper is as follows .
as has been mentioned above , the main stream of our argument is based on lakshtanov and vainberg @xcite . in 2 ,
we recall some basic properties of the dirichlet - to - neumann map . for our purpose , we need to study about residues and regular parts of the dirichlet - to - neumann map near its poles .
the relation between ites and non - trivial kernels of the difference of dirichlet - to - neumann maps is also introduced here .
finally , we construct an approximate solution of the dirichlet boundary value problems as a singular - integro differential operator , and we compute the symbol of the dirichlet - to - neumann map . under the preparation of 2 , we prove our main results in 3 .
we use of the analytic fredholm theory , the parameter ellipticity of pseudo differential operators and weyl type asymptotic estimates for the number of dirichlet eigenvalues on compact manifolds .
our main results are theorem [ s3_thm_discreteness ] for the discreteness of ites and theorem [ s3_thm_weylnt ] for the lower bound of the number of ites in @xmath71 with sufficiently small @xmath72 .
we use the following notations .
we put @xmath73 and @xmath74 .
for the riemannian metric @xmath75 of @xmath17 , @xmath76 and @xmath77 denote @xmath78 and @xmath79 , respectively . @xmath80 and @xmath81 denote the volume element on @xmath17 and the surface element on @xmath11 induced by @xmath82 , respectively .
we often write them as @xmath83 and @xmath84 omitting @xmath85 .
letting @xmath86 be a local coordinate of @xmath87 , @xmath88 or @xmath89 denote @xmath90 . for @xmath91 , we use the similar manner . for a multiple index @xmath92
, we write @xmath93 .
we often compute some kind of symbols @xmath94 .
for short , we use the symbol @xmath95 which means that each @xmath96 replaced by @xmath97 .
similarly , when we write @xmath98 , each @xmath99 is replaced by @xmath100 .
@xmath101 denotes the outward normal derivative on @xmath11 associated with @xmath87 . for a strictly positive valued function @xmath102 ,
@xmath103 is the @xmath104 space on @xmath17 with the inner product @xmath105 .
here we consider the following dirichlet problems : @xmath106 for @xmath16 .
we define the dirichlet - to - neumann ( d - n ) map @xmath107 by @xmath108 where @xmath109 is a solution of ( [ s1_eq_d1 ] ) . in the following , we call @xmath110 a dirichlet eigenvalue if there exists a non - trivial solution of the equation @xmath111 in fact , ( [ s2_eq_dirichleteigen ] ) is equivalent to @xmath112 which is an eigenvalue problem of the second - order self - adjoint elliptic operator @xmath113 in @xmath114 with the dirichlet boundary condition
. then its eigenvalues form an increasing sequence @xmath115 , satisfying the weyl s asymptotics which we derive in 3 .
the corresponding eigenfunctions @xmath116 can be chosen so that @xmath117 is an orthonormal basis in @xmath118 .
we denote the set of dirichlet eigenvalues by @xmath119 . for @xmath120 ,
the d - n map @xmath121 is well - defined and extends uniquely as a continuous operator @xmath122 .
let @xmath123 such that @xmath124 , and @xmath125 and @xmath126 belong to the same set @xmath127 if and only if @xmath128 .
we denote eigenvalues corresponding @xmath127 by @xmath129 .
@xmath130 means the set @xmath131 with @xmath132 @xmath133 is meromorphic with respect to @xmath22 and has first order poles at each @xmath134 .
moreover , @xmath121 has the following representations : + ( 1 ) for @xmath135 and @xmath136 , we have @xmath137 ( 2 ) in a neighborhood of @xmath138 , we have @xmath139 where @xmath140 is the residue of @xmath141 at @xmath142 given by @xmath143 and @xmath144 is analytic in a neighborhood of @xmath145 . [ s2_prop_green_meromor ] proof .
we can follow the argument of 4.1.12 in @xcite .
let @xmath146 be an extension of @xmath147 into @xmath87 satisfying @xmath148 and @xmath149 for some constants @xmath150 , and @xmath151 is in ( [ s1_eq_d1 ] ) .
then we have @xmath152 since @xmath153 is a meromorphic operator valued function with first order poles only at @xmath154 , @xmath155 is also a meromorphic @xmath156-valued function with first order poles only at @xmath154 .
next we prove ( [ s2_eq_dnrepresentation ] ) .
integrating by parts , we compute the fourier coefficients of @xmath151 with respect to the real - valued eigenfunction @xmath157 : @xmath158 from this formula and the outward normal derivative of @xmath151 , @xmath159 satisfies ( [ s2_eq_dnrepresentation ] ) .
finally we verify ( [ s2_eq_resdn ] ) and ( [ s2_eq_pk ] ) .
let @xmath160 be the projection to the eigenspace corresponding to @xmath161 i.e. @xmath162 in view of ( [ s2_eq_fouriercoefficients ] ) , we have @xmath163 and this implies ( [ s2_eq_pk ] ) . moreover , @xmath164 is analytic with respect to @xmath53 in a neighborhood of @xmath165 .
putting @xmath166 on @xmath11 , we have the proposition .
_ the formula ( [ s2_eq_pk ] ) means that the range of @xmath140 is a finite dimensional subspace spanned by @xmath167 for @xmath168 .
note that @xmath169 for all @xmath170 are linear independent since @xmath171 are orthogonal basis .
hence @xmath172 coincides with the multiplicity of @xmath165 .
we can see that the integral kernel of @xmath140 is smooth in @xmath173 by the regularity property of dirichlet eigenfunctions .
as has been in propositions [ s2_prop_green_meromor ] , @xmath174 is also meromorphic with respect to @xmath175 and has first order poles at each @xmath176 . in a neighborhood of a pole @xmath177
, we have @xmath178 where @xmath179 and @xmath180 have same properties of @xmath181 and @xmath182 , respectively . in the following
, we define the kernel of @xmath183 by @xmath184 let @xmath134 .
then the equation ( [ s1_eq_d1 ] ) has a non trivial solution if and only if @xmath185 is orthogonal to @xmath186 in @xmath187 for all @xmath188 .
[ s2_lem_singular_sol ] proof .
if @xmath185 is orthogonal to @xmath186 for all @xmath189 , there exist general solutions of the form @xmath190 for any @xmath191 .
if @xmath109 is a non trivial solution of ( [ s1_eq_d1 ] ) , we have by green s formula @xmath192 for @xmath193 . since @xmath194 , the left - hand side is equal to zero
. then @xmath195 is orthogonal to @xmath196 .
the above lemma implies a unique solvability in a subspace as follows .
let @xmath197 be a dirichlet eigenvalue .
we define @xmath198 as the eigenspace associated by @xmath177 , and @xmath199 as the subspace of @xmath200 spanned by @xmath201 for all @xmath202 .
we denote by @xmath203 and @xmath204 their orthogonal complements in @xmath205 and @xmath187 , respectively . for any @xmath206
, there exists a unique solution @xmath207 of ( [ s1_eq_d1 ] ) represented by @xmath208 [ s2_cor_singular_sol2 ] proof .
we have only to check the uniqueness .
this is trivial since the equation ( [ s2_eq_dirichleteigen ] ) has only the trivial solution in @xmath209 .
now we can state the relation between ites and the d - n map as follows .
\(1 ) suppose @xmath210 .
then @xmath22 is an ite if and only if @xmath211 . the multiplicity of @xmath53 coincides with @xmath212 .
+ ( 2 ) suppose @xmath213 .
then @xmath22 is an ite if and only if @xmath214 or the ranges of @xmath215 and @xmath216 have a non trivial intersection .
the multiplicity of @xmath53 coincides with the sum of @xmath212 and the dimension of the above intersection .
[ s2_lem_iteeqcondition ] proof .
we first prove the assertion ( 1 ) . when @xmath217 , this lemma is a direct consequence of the definition of ites .
we have only to show for @xmath218 . for @xmath219
, we have @xmath220 . from @xmath221 and ( [ s2_eq_pk ] )
, we have @xmath222 . by lemma [ s2_lem_singular_sol ] and corollary [ s2_cor_singular_sol2 ]
, the following equation has a unique non trivial solution : @xmath223 on the other hand , from @xmath224 , we have @xmath225 summarizing ( [ s2_eq_dd1 ] ) and ( [ s2_eq_dd2 ] ) and @xmath226 , @xmath227 is an ite .
conversely , if @xmath53 is an ite , from lemma [ s2_lem_singular_sol ] , the equation ( [ s1_eq_d1 ] ) , @xmath228 , with the condition @xmath229 must have a non trivial solution . in view of ( [ s2_eq_pk ] ) , we have @xmath230 , and this implies @xmath231 .
this means @xmath232 by using ( [ s2_eq_resdn ] ) . on the other hand
, @xmath233 means @xmath234 .
therefore , @xmath147 must be in @xmath235 .
we have proven the assertion ( 1 ) . for the assertion ( 2 ) , we have only to show the latter case .
in fact , if there exists a non trivial solution @xmath236 of @xmath237 with @xmath238 and @xmath239 on @xmath240 , then we have that the ranges of @xmath215 and @xmath216 have a non trivial intersection , recalling @xmath241 for @xmath242 .
conversely , if the ranges of @xmath215 and @xmath216 have a non trivial intersection , then there exists a non trivial solution @xmath236 of the above system with the condition @xmath238 and @xmath243 on @xmath240 , since @xmath196 for all @xmath244 are linear independent .
then @xmath53 is an ite .
_ in @xcite , the authors call @xmath227 _ singular spectral point _ if @xmath53 satisfies the latter condition in the assertion ( 2 ) of lemma [ s2_lem_iteeqcondition ]
. now let us compute the symbol of the d - n map .
here we construct the local regularizer for ( [ s1_eq_d1 ] ) .
as in @xcite , we follow the argument of 2 in @xcite , slightly modifying it for our case . in the following , we assume that the equation ( [ s1_eq_d1 ] ) is uniquely solvable in @xmath245 or a suitable subspace of @xmath246 .
we take a point @xmath247 and fix it .
let @xmath248 be a sufficiently small neighborhood of @xmath249 in @xmath11 . there exist small open domains @xmath250 , @xmath251 , such that @xmath252 and @xmath253 and @xmath254 are diffeomorphic to an open domain @xmath255 .
we introduce local coordinates @xmath256 in @xmath257 with the center @xmath258 such that @xmath259 , @xmath27 is given by @xmath260 , @xmath261 for a small @xmath262 , the subset @xmath263 is diffeomorphic to @xmath264 , and @xmath265 is the distance between a point @xmath266 and @xmath267 . then @xmath268 are common local coordinates of @xmath253 and @xmath269 . therefore , we have @xmath270 , \quad y ' = ( y_1 , \cdots , y_{d-1 } ) , \ ] ] in @xmath271 where @xmath272 is a smooth , positive definite and symmetric @xmath273-matrix valued function , and @xmath274 is a @xmath275-dimensional vector valued function .
a function @xmath276 with @xmath277 is homogeneous of the generalized degree @xmath278 if @xmath279 satisfies @xmath280 for any @xmath281 . for @xmath282
, we define the homogeneity by the similar manner .
taking the @xmath283-coordinate as above , we can rewrite @xmath284 as @xmath285 in @xmath286 with @xmath287 .
note that @xmath288 , @xmath289 and @xmath290 are defined by @xmath291 . in view of the assumption ( a-1 ) , we have in @xmath283-coordinates that @xmath292 , @xmath293 .
the symbol of @xmath294 is given by @xmath295 in the following , let @xmath296 be a sufficiently large integer .
now we take @xmath297 arbitrarily and fix it . using the taylor series of @xmath288 , @xmath298 , @xmath299 and @xmath300 with respect to @xmath283 centered at @xmath301
, we can expand the symbol @xmath302 of @xmath294 as the sum of following terms : @xmath303 @xmath304 and @xmath305 for @xmath306 with the remainder term which has zero of order @xmath307 at @xmath308 or @xmath309 .
we rewrite the sum of ( [ s2_eq_expand_symbol1])-([s2_eq_expand_symbol3 ] ) and the remainder term as @xmath310 then each @xmath311 is a homogeneous polynomial in @xmath312 of generalized degree @xmath313 . in particular , @xmath314 is the principal symbol of @xmath294 .
@xmath315 vanishes at @xmath316 and the order of the zero is @xmath307 . in the following arguments , we put @xmath317 we define the following differential operators : @xmath318 @xmath319 and @xmath320 let @xmath321 be a smooth function and homogeneous of the generalized degree @xmath322 with respect to @xmath265 and @xmath323
. then we have that @xmath324 is the homogeneous of the generalized degree @xmath325 with respect to @xmath265 and @xmath323 .
[ s2_prop_degree ] proof .
note that @xmath326
. then we can show that @xmath327 and @xmath328 are homogeneous of generalized degree @xmath329 and @xmath330 , respectively .
now let us construct an approximate solution of ( [ s1_eq_d1 ] ) .
suppose @xmath331 .
the system of second order ordinary differential equations @xmath332 has a unique solution @xmath333 such that each @xmath334 converges to zero as @xmath335 and satisfies @xmath336 in particular , we have @xmath337 . each solution @xmath338 is smooth and homogeneous with respect to @xmath265 and @xmath339 of generalized degree @xmath340 .
( for @xmath341 , each @xmath334 depends also on @xmath53 .
we omit @xmath53 in the notation . )
[ s2_lem_systemode ] proof . since @xmath342
, we have @xmath337 . obviously , @xmath343 is homogeneous of the generalized degree @xmath344 .
let us consider the equation @xmath345 for @xmath346 and @xmath347 with @xmath348 , @xmath349 as @xmath335 .
here we assume that @xmath350 decays exponentially as @xmath335 and is homogeneous of the generalized degree @xmath322 . extending @xmath351 and @xmath352 to be zero in @xmath353
, we have @xmath354 then , putting @xmath355 , we have @xmath356 which shows that @xmath351 is homogeneous of the generalized degree @xmath357 with respect to @xmath265 and @xmath323 . in view of proposition [ s2_prop_degree ]
, we have @xmath358 is homogeneous of the generalized degree @xmath359 .
therefore , we obtain @xmath360 is homogeneous of the generalized degree @xmath361 . repeating the similar argument inductively , we can show that @xmath334 is homogeneous of the generalized degree @xmath340 .
let @xmath362 vanish in a neighborhood of @xmath363 , and be equal to one outside a large neighborhood of @xmath364 .
taking @xmath365 with a compact support in @xmath366 , we define for @xmath367 @xmath368 and we put @xmath369 letting @xmath370 we have that @xmath371 is a distribution in @xmath372 , and @xmath373 with @xmath374 we represent @xmath294 in the form @xmath375 in the following , we consider @xmath376 let @xmath377 , @xmath378 and @xmath379 be sufficiently large .
we have @xmath380 and @xmath381 where @xmath382 and @xmath383 . [ s2_lem_regurality ]
note that @xmath384 and @xmath385 are operators which are given by sums of terms like @xmath386 up to a smooth function with @xmath387 or @xmath388 and @xmath389 .
in view of proposition [ s2_prop_degree ] , it is sufficient to show @xmath390 since the derivative @xmath391 is order zero , one or two .
now we have @xmath392 since @xmath393 is homogeneous of the generalized degree @xmath394 , using proposition [ s2_prop_degree ] , we have @xmath395 which implies ( [ s2_eq_hlm ] ) .
let @xmath396 be sufficiently large .
we have that @xmath397 is local regularizer for ( [ s1_eq_d1 ] ) i.e. @xmath398 for @xmath399 which has a compact support in @xmath400 and @xmath401 . [ s2_thm_localreg ] proof .
note that @xmath402 summing up both sides of ( [ s2_eq_sys0])-([s2_eq_sys2 ] ) , we have @xmath403 in view of lemma [ s2_lem_regurality ] and ( [ s2_eq_ar ] ) , we have that ( [ s2_eq_fourier1 ] ) and ( [ s2_eq_sys33 ] ) imply @xmath404 for @xmath405 .
we have that @xmath406 as @xmath407 . since @xmath408 , we have @xmath409 . _ remark . _
the formal sum @xmath410 is a singular integro - differential operator ( see @xcite ) . in general
, a linear operator @xmath411 on a @xmath412-dimensional compact manifold @xmath413 is a singular integro - differential operator of order @xmath414 if there exist homogeneous functions @xmath415 in @xmath91 with homogeneous degree @xmath416 such that for a function @xmath33 with support in a local coordinate neighborhood @xmath417 , @xmath418 where @xmath419 is an arbitrary function which satisfies @xmath420 for @xmath421 and @xmath422 for @xmath423 , and @xmath424 is an operator which increases the smoothness i.e. @xmath425 for any @xmath426 .
the principal symbol of @xmath411 is @xmath427 and the full symbol of @xmath411 is the formal sum @xmath428 .
then the ellipticity of @xmath411 is defined by @xmath429 for all @xmath430 .
here this means that we can construct the parametrix of @xmath411 ( see @xcite ) .
therefore , if @xmath411 is an elliptic singular integro - differential operator , @xmath411 is fredholm .
since we have @xmath431 in @xmath283-coordinates , we can show the following fact . as a consequence of corollary [ s2_cor_singular_sol2 ] and theorem [ s2_thm_localreg ] .
see also lemma 11 and theorem 14 in @xcite .
\(1 ) when @xmath53 is not a pole of @xmath141 , @xmath141 is a singular integro - differential operator on @xmath432 with the full symbol given by the following asymptotic series : @xmath433 ( 2 ) when @xmath434 is a pole of @xmath435 , the regular part @xmath182 of @xmath141 at @xmath177 is a singular integro - differential operator on @xmath436 with the full symbol given by ( [ s2_eq_dnsymbol ] ) .
[ s2_cor_dnsymbol ] we compute the principal symbol of @xmath437 . in the following ,
we denote higher order normal derivatives on @xmath11 associated with @xmath17 by @xmath438 in @xmath283-coordinates , we can locally represent @xmath439 . under the assumption ( a-1 )
, we additionally assume on @xmath240 that * ( a-2 ) * the metrics @xmath13 , @xmath14 and the indices of refraction @xmath440 , @xmath441 satisfy one of following two cases : * ( a-2 - 1 ) * for all @xmath135 , @xmath442 for @xmath443 , @xmath444 , and @xmath445 , or * ( a-2 - 2 ) * for all @xmath446 , @xmath447 for @xmath448 , @xmath444 , and @xmath449 , @xmath450 .
note that , under the assumptions ( a-1 ) with ( a-2 - 1 ) or ( a-2 - 2 ) , we can see @xmath451 for @xmath452 or @xmath453 , respectively .
when @xmath434 is a pole of @xmath454 , we define a subspace @xmath455 of @xmath200 by @xmath456 where @xmath457 if @xmath177 is a dirichlet eigenvalue of @xmath458 , and @xmath459 if otherwise .
we denote by @xmath460 the orthogonal complement of @xmath461 in @xmath187 .
when @xmath434 is a pole of @xmath174 , we call @xmath174 fredholm if its regular part @xmath462 is fredholm . in the following ,
we suppose @xmath463 .
+ ( 1 ) let @xmath53 be not a pole of @xmath174 . for the case
( a-2 - 1 ) , we have @xmath464 is an elliptic singular integro - differential operator with the principal symbol @xmath465 ( 2 ) let @xmath53 be not a pole of @xmath174 . for the case ( a-2 - 2 )
, we have @xmath466 is an elliptic singular integro - differential operator with the principal symbol @xmath467 ( 3 ) when @xmath53 is a pole of @xmath174 , the regular part of @xmath468 is singular integro - differential operator on @xmath460 with order @xmath361 for ( a-2 - 1 ) or @xmath469 for ( a-2 - 2 ) .
its principal symbol is given by ( [ s2_eq_principalsymb1 ] ) or ( [ s2_eq_principalsymb2 ] ) , respectively .
+ ( 4 ) for both of ( a-2 - 1 ) or ( a-2 - 2 ) , @xmath437 is fredholm for @xmath470 .
[ s2_lem_principalsymb ] proof .
let @xmath471 and @xmath441 satisfy ( a-2 - 1 ) . in @xmath283-coordinates
, we have @xmath472 , @xmath473 and @xmath474
. then @xmath475 , @xmath476 and @xmath477 a particular solution of this equation is @xmath478 which vanishes at @xmath479 and @xmath335
. then we can take it as @xmath480 , and @xmath481 at @xmath479 is the principal symbol of @xmath437 . in view of the assertion ( 1 ) in corollary
[ s2_cor_dnsymbol ] , we have the assertion ( 1 ) .
next we assume that @xmath471 and @xmath441 satisfy ( a-2 - 2 ) .
as above , we have @xmath482 for @xmath483 , and @xmath484 .
then we have @xmath485 hence we obtain the assertion ( 2 ) . in view of corollary [ s2_cor_singular_sol2 ] and the assertion ( 2 ) in corollary [ s2_cor_dnsymbol ] , we can show the assertion ( 3 ) by the similar way .
the ellipticity of @xmath486 implies that @xmath487 is fredholm for @xmath488 .
here we prove main results .
let us list our assumptions again : for the proof of discreteness , we need to use the analytic fredholm theory which was generalized by @xcite .
see also appendix a in @xcite .
let @xmath490 and @xmath491 are hilbert spaces .
we take a connected open domain @xmath492 .
an operator valued function @xmath493 for @xmath494 is finitely meromorphic if the principal part of the laurent series at a pole of @xmath495 is a finite rank operator . in particular
, @xmath496 is finitely meromorphic in @xmath497 as has been seen in proposition [ s2_prop_green_meromor ] .
suppose an operator valued function @xmath493 , @xmath498 , is finitely meromorphic and fredholm .
if there exists its bounded inverse @xmath499 at a point @xmath500 , then @xmath501 is finitely meromorphic and fredholm in @xmath502 . from the above theorem , if there exists a @xmath488 such that @xmath503 is invertible , @xmath503 is invertible in @xmath504 for a discrete subset @xmath505 of @xmath506 .
therefore , for the proof of the discreteness , we have only to show that @xmath503 is invertible for some @xmath507 .
we expand the symbol of @xmath508 centered at @xmath509 by the same manner in 2.2 .
however , here we change the definition of homogeneous functions with generalized degree @xmath322 by @xmath510 for @xmath488 , taking a suitable branch of @xmath511 .
we gather terms of the same generalized degree in the sense ( [ s3_def_parameterhomogeneous ] ) , and we denote the symbol in @xmath283-coordinates by @xmath512 up to the remainder term where @xmath513 is homogeneous of degree @xmath313 . in particular , putting @xmath514 , we have @xmath515 where @xmath516 is defined by ( [ s2_def_atilde2 ] ) and @xmath517 we denote by @xmath518 the solution of @xmath519 with the boundary condition @xmath520 , @xmath521 for @xmath522 and @xmath523 as @xmath335 for @xmath524 .
let @xmath525 be a pseudo differential operator ( or a singular integro - differential operator ) which is parametrized by @xmath175 such that the full symbol of @xmath526 is a asymptotic series of homogeneous functions in the sense ( [ s3_def_parameterhomogeneous ] ) . if the principal symbol does not vanish when @xmath527 , we call @xmath528 _ parameter elliptic_. let @xmath529 . + ( 1 ) we assume that ( a-2 - 1 ) holds .
then @xmath530 is parameter elliptic .
its principal symbol is @xmath531 ( 2 ) we assume that ( a-2 - 2 ) holds . then @xmath530 is parameter elliptic .
its principal symbol is @xmath532 where @xmath533 and @xmath534 .
we fix an arbitrary point @xmath535 .
suppose that ( a-2 - 1 ) holds
. obviously we have @xmath536 under the assumption , we also have @xmath537 so that @xmath538
. then the principal symbol @xmath539 of @xmath540 is given by ( [ s3_eq_symbola31 ] ) .
let us consider the case ( a-2 - 2 ) .
in view of @xmath541 on @xmath240 , we have @xmath542 so that @xmath543 since we have assumed ( a-1 ) and ( a-2 - 2 ) , we have @xmath544 then @xmath545 satisfies the equation @xmath546 precisely , we obtain @xmath547 then the principal symbol @xmath548 of @xmath540 is given by ( [ s3_eq_symbola32 ] ) . in view of lemma [ s3_lem_parametersymbol ] , we can obtain a uniform estimate in @xmath53 of @xmath549 and its inverse . in the following ,
we define the hilbert space @xmath550 for @xmath551 by the norm @xmath552 we take an arbitrary closed sector @xmath553 centered at the origin such that @xmath554 .
we put @xmath555 . for
the proof of the next result , see theorem 4.4.6 in @xcite and see also @xcite .
we assume ( a-1 ) and one of ( a-2 - 1 ) and ( a-2 - 2 ) .
the set of ites consists of a discrete subset of @xmath563 with the only possible accumulation points at @xmath344 and infinity .
there exist at most finitely many ites in @xmath564 .
[ s3_thm_discreteness ] in view of ( a-1 ) , there exists a positive integer @xmath573 such that @xmath574 where each @xmath575 is a closed and connected component of @xmath11 . taking an arbitrary point @xmath576
, we take a small neighborhood @xmath577 of @xmath249 and a sufficiently small open domain @xmath27 which is diffeomorphic to @xmath578 such that @xmath579 as has been defend in the beginning of 2.2 . then , identifying @xmath580 with the corresponding point on @xmath267 , the following function @xmath581 for @xmath582 is well - defined constant function @xmath583 or @xmath584 in @xmath264 .
obviously , the function @xmath585 or @xmath584 can be naturally extended on @xmath575 .
we define the function @xmath586 on @xmath587 by @xmath588 in the following , we use an auxiliary operator defined by @xmath589 where @xmath590 for ( a-2 - 1 ) or @xmath562 for ( a-2 - 2 ) . here
@xmath591 is given by @xmath592 where @xmath593 is the laplace - beltrami operator on @xmath11 .
then @xmath594 is a first order singular integro - differential operator when @xmath53 is not a pole of @xmath595 .
\(1 ) suppose @xmath210 .
then @xmath22 is an ite if and only if @xmath596 . the multiplicity of @xmath53 coincides with @xmath597 .
+ ( 2 ) suppose @xmath213 .
then @xmath22 is an ite if and only if @xmath598 or the range of @xmath599 and @xmath600 have a non trivial intersection .
the multiplicity of @xmath53 coincides with the sum of @xmath597 and the dimension of the above intersection .
[ s2_lem_iteeqconditionb ] let @xmath602 be not a pole of @xmath603 . + ( 1 ) then @xmath603 is a first order , symmetric and elliptic singular integro - differential operator .
its principal symbol is @xmath604 for ( a-2 - 1 ) , or @xmath605 for ( a-2 - 2 ) , where @xmath606 .
+ ( 2 ) for @xmath607 , the spectrum of @xmath603 is discrete and consists of the set of real eigenvalues @xmath608 .
[ s3_lem_symbb ] since @xmath603 has a positive principal symbol and @xmath603 is meromorphic with respect to @xmath53 , we also have the following lemma .
the proof is completely same of lemma 2.4 and 2.5 in @xcite .
note that , in view of ( [ s2_eq_resdn ] ) , we define the residue @xmath609 of @xmath610 at a pole @xmath177 by the expansion @xmath611 where @xmath612 is analytic in a small neighborhood of @xmath177 .
\(1 ) for each compact interval @xmath613 such that any pole of @xmath603 are not included in @xmath614 , there exists a constant @xmath615 such that @xmath616 for @xmath617 , @xmath618 .
+ ( 2 ) if @xmath603 is analytic in a neighborhood of @xmath177 , all eigenvalues @xmath610 are analytic in this neighborhood .
if @xmath177 is a pole of @xmath603 and @xmath619 is the rank of the residue of @xmath603 at @xmath177 , @xmath352 eigenvalues @xmath610 and its eigenfunctions have a pole at @xmath177 .
moreover , @xmath620 are eigenvalues of @xmath621 .
now we consider the relation between @xmath626 and @xmath608 for @xmath627 .
roughly speaking , we can evaluate @xmath628 by the number of the singular spectral points and the number of @xmath227 satisfying @xmath629 .
we put @xmath630 assume that @xmath631 moves from @xmath632 to @xmath633 .
since @xmath634 is meromorphic with respect to @xmath631 , @xmath635 changes only when some @xmath636 pass through @xmath344 or @xmath631 passes through a pole of @xmath637 .
when @xmath631 moves from @xmath632 to @xmath638 , we denote by @xmath639 the change of @xmath640 due to the first case , and @xmath641 as the change due to the second case i.e. @xmath642 for a pole @xmath643 of @xmath603 , we put @xmath644 for any @xmath645 . proof . in view of lemma [ s3_lem_positiveb ] , some eigenvalues
@xmath610 have a pole at @xmath177 .
if @xmath649 , we have @xmath650 as @xmath651 and @xmath652 as @xmath653 , respectively . then the number of negative eigenvalues decreases for @xmath654 and increases for @xmath655 when @xmath53 passes through @xmath177 from @xmath632 .
this implies the lemma .
we denote by @xmath662 a union of connected components @xmath575 such that @xmath663 , respectively .
let @xmath664 .
we put @xmath665 and @xmath666 , and we split @xmath667 and @xmath668 with @xmath669 and @xmath670 for a pole @xmath671 of @xmath603
. proof .
first we prove the assertion ( 1 ) .
suppose @xmath677 .
we can expand @xmath603 in a small neighborhood of @xmath177 as @xmath678 where @xmath679 is analytic . from lemma [ s3_lem_residuepositive ] , we have @xmath680 and also @xmath681 so that @xmath682 has exactly @xmath683 strictly negative eigenvalues .
then , in view of the assertion ( 2 ) in lemma [ s3_lem_positiveb ] and lemma [ s3_lem_jump ] , we have @xmath684 , so @xmath685 with @xmath686 .
for the case @xmath687 , we can see @xmath688 with @xmath689 by the similar way .
plugging these two cases , we obtain the assertion ( 1 ) .
let us prove the assertion ( 2 ) .
suppose @xmath690 for @xmath691 and @xmath692 .
then we have the following representation in a small neighborhood of @xmath177 @xmath693 with @xmath694 .
we see that @xmath695 on @xmath696 and @xmath697 on @xmath698
. then we have @xmath699 and @xmath700 . these inequalities and lemma [ s3_lem_jump ]
imply the assertion ( 2 ) .
we assume ( a-1 ) and one of ( a-2 - 1 ) and ( a-2 - 2 ) . for large @xmath701 ,
we have @xmath702 where the summation is taken over poles @xmath703 $ ] of @xmath530 . moreover , if @xmath704 or @xmath705 and @xmath706 or @xmath707 respectively where @xmath708 are defined in ( [ s3_eq_weyldirichlet ] ) , @xmath628 satisfies asymptotically as @xmath709 @xmath710 proof .
we prove for the case @xmath711 . for @xmath712 ,
the proof is similar and can be slightly simplified .
letting us recall that we call @xmath227 is a singular spectral point when @xmath227 satisfies the latter condition of the assertion ( 2 ) of lemma [ s2_lem_iteeqcondition ] , we put @xmath713 \subset { \bf r } _ { > 0 } \ } .\ ] ] here @xmath714 counts the number of singular spectral points with multiplicities taken into account .
note that @xmath715 by the definition of @xmath716 and lemma [ s2_lem_iteeqconditionb ] .
we take the summation of @xmath717 over all poles @xmath718 of @xmath719 in @xmath720 $ ] .
then we have @xmath721 see also remark of proposition [ s2_prop_green_meromor ] .
plugging this inequality and ( [ s3_eq_diffnb ] ) , we have @xmath722 since @xmath723 , we obtain ( [ s3_eq_lower1 ] ) .
if @xmath724 , we see @xmath725 and @xmath726 for any poles @xmath718 , respectively , so that we have from ( [ s3_eq_lower1 ] ) @xmath727 the asymptotic estimate ( [ s3_eq_lower2 ] ) is a direct consequence of ( [ s3_eq_lower1 ] ) and theorem [ s3_thm_weyl ] .
p. m. blekher , _ operators that depend meromorphically on a parameter _ , vestnik moskov .
mech . , * 24 * ( 1969 ) , 30 - 36 .
( in russian ) ; english transl .
: moscow univ .
math . bull .
, * 24 * ( 1969 ) , 21 - 26 . | in this paper , we consider an interior transmission eigenvalue ( ite ) problem on some compact @xmath0-riemannian manifolds with smooth boundary .
in particular , we do not assume that two domains are diffeomorphic , but we impose some conditions of riemannian metrics and indices of refraction on the boundary .
then we prove the discreteness of the set of ites , the existence of infinitely many ites , and its weyl type lower bound .
our problem is non - coercive .
however , the argument by lakshtanov and vainberg @xcite which reduces the problem to the elliptic one using the dirichlet - to - neumann map , is valid for manifolds . |
similar to monopole and quadrupole breathing modes of a gaseous bose - einstein condensate , scissors modes were first studied theoretically and subsequently observed experimentally .
the scissors modes are , however , rather special since they directly manifest the superfluid behavior of these atomic gases . moreover , the recent experimental studies appear to show a resonance behavior between two coupled scissors modes @xcite . from a theoretical point of view
this is interesting because a linear - response calculation can neither account for the coupling nor for the resonance behavior .
therefore , a first step towards an explanation of these experimental observations is to perform a calculation that goes beyond linear - response theory and accurately takes into account the mean - field interaction that couples the scissors and breathing modes . in this paper
we present a simple variational method for calculating the frequencies of these various modes and their couplings beyond the linear - response .
we perform our calculation at zero temperature and therefore do not consider the damping rates of the scissors modes . the main idea behind our method is to use a time - dependent gaussian _
ansatz _ for the groundstate wave function to derive the equations of motion of the breathing modes and the scissors modes .
then we expand the resulting equations of motion in deviations from equilibrium . in first order , i.e. , linear - response , we recover the expected uncoupled set of equations .
the second - order calculation produces a set of coupled equations which show that we need to excite all three scissors modes in order to get a nonzero coupling . at higher orders we , however , find that we only need to excite two modes to get coupling .
furthermore , we actually find under certain conditions a resonance behavior between these two modes .
the layout of the paper is as follows .
first , we rederive in sec .
[ freqs - section ] the frequencies of the scissors modes in the linear - response limit . in sec .
[ heigher ] we extend the calculation first to second , and then also to higher orders , which ultimately lead to a resonant coupling . in sec .
[ solution ] we solve the equations of motion analytically near the resonance using an envelope function approach . in sec .
[ conclusion ] we end with a discussion of our results .
we start by considering a bose - einstein condensate trapped by the following harmonic potential @xmath0 where @xmath1 , @xmath2 , and @xmath3 are the angular frequencies of the trap , and @xmath4 is the atomic mass . a scissors mode in a bose - einstein condensate
is associated with an irrotational flow with a velocity field of the form @xmath5 , if the motion is taking place in the @xmath6-plane @xcite .
similar expressions hold for the two other cartisian planes .
these kind of modes can be excited by a sudden rotation of the equilibrium axes of the trap .
to such a perturbation the condensate will respond by oscillating around the new equilibrium axes .
for example , to excite a scissors mode in the @xmath6-plane , we rotate the @xmath7 and @xmath8-axes of the trap slightly around the @xmath9-axis .
if the angle of rotation is sufficiently small , the scissors mode can be approximated by a simple oscillation of the condensate around the new equilibrium axes .
on the other hand , if the axes change through a large angle this method excites the @xmath10 quadrupole mode , where @xmath4 labels the projection of the angular momentum along the axis of symmetry . the maximum angle for which the scissors mode is defined increases with deformation of the trap @xcite . to account for all three scissors modes in the three cartisian planes we employ the following trial function for the condensate order parameter @xmath11 where @xmath12 and @xmath13 , are complex time - dependent variational parameters and @xmath14{c_{xy , r}c_{xz , r}c_{yz , r}+4b_{x , r}b_{y , r}b_{z , r}-(b_{z , r}c_{xy , r}^2+b _ { y , r}c _ { xz , r}^2+b_{x , r}c_{yz , r}^2 ) } \label{a}.\ ] ] this value of the prefactor @xmath15 guarantees the normalization of the square of the wave function @xmath16 to the total number of condensed atoms @xmath17 . here
@xmath18 and @xmath19 are the real parts of @xmath12 and @xmath13 , respectively .
the first set of parameters , @xmath12 , give rise to the well - studied breathing modes which , for axially symmetric traps , are called the monopole and quadrupole modes depending on the value of @xmath4 being equal to zero or two , respectively .
the parameters @xmath13 on the other hand determine the three scissors modes .
the equations of motion for these variational parameters can be derived from the lagrangian @xmath20={1\over2}i\hbar\int { { \rm d}{\bf r } \left(\psi^*({\bf r},t){\partial\psi({\bf r},t)\over\partial t } -\psi({\bf r},t){\partial\psi^*({\bf r},t)\over\partial t}\right ) } -e[\psi,\psi^ * ] \label{lagrangian},\ ] ] where @xmath21 $ ] is the usual gross - pitaevskii energy functional given by @xmath22=\int d{\bf r } \left [ { \hbar^2\over 2m}|{\mbox{\boldmath $ \nabla$}}\psi({\bf r},t)|^2 + v({\bf r})|\psi({\bf r},t)|^2 + { 1\over2}t^{2b}|\psi({\bf r},t)|^4 -\mu|\psi({\bf r},t)|^2 \right ] \label{functional}.\ ] ] here @xmath23 is the two - body @xmath24-matrix , which for the atomic bose - einstein condensates of interest is related to the @xmath25-wave scattering length @xmath26 through @xmath27 .
inserting our trial wave function into the lagrangian and scaling frequencies with @xmath28 and lengths with @xmath29 , it takes the dimensionless form @xmath30/n&=&\left(\alpha_{x}{\dot b}_{x , i}+\alpha_{y}{\dot b}_{y , i}+\alpha_{z}{\dot b}_{z , i}\right)/q\\\nonumber & -&{1\over2}\left [ \alpha_{x}(4|b_{x}|^2+|c_{xy}|^2+|c_{xz}|^2)\right.\\\nonumber & + & \alpha_{y}(4|b_{y}|^2+|c_{xy}|^2+|c_{yz}|^2)\\\nonumber & + & \left.\alpha_{z}(4|b_{z}|^2+|c_{xz}|^2+|c_{yz}|^2)\right]/q\\\nonumber & -&{1\over2}\left[\alpha_{x}\omega_{x}^2+\alpha_{y}\omega_{y}^2+\alpha_{z } \omega_{z}^2\right]/q\\\nonumber & -&{1\over2\sqrt{\pi}}\gamma\sqrt{q } \label{scaled lagrangian } , \end{aligned}\ ] ] where @xmath31 , @xmath32 , @xmath33 , @xmath34 , and the dot corresponds to a time derivative .
in addition @xmath35 is the dimensionless parameter that represents the strength of the mean - field interaction . minimizing the lagrangian with respect to the 12 variational parameters
, we get a set of 12 coupled equations of motion .
the resulting equations of motion are rather lengthy and complicated .
a significant simplification takes place if we expand these equations in the deviation of the variational parameters from their equilibrium values , i.e. , in @xmath36 and @xmath37 , where @xmath38 is the yet unknown eigenfrequency of the modes , and @xmath39 and @xmath40 denote the equilibrium values of @xmath12 and @xmath13 , respectively .
these latter values can simply be obtained by setting the time derivatives in the equations of motion to zero . for large condensates in the so - called thomas - fermi regime
we can ignore contributions from the kinetic energy @xcite and the equilibrium variational parameters take the simple form @xmath41 it is required that @xmath42 for the last equation to be valid .
to first order in the deviations the equations of motion read simply @xmath43 where the vector @xmath44 contains all the possible fluctuations , and the matrix @xmath45 is given by @xmath46 where @xmath47 and @xmath48 are given explicitly in appendix [ app ] .
it is clear from the last equation that to linear order the breathing modes and scissors modes are uncoupled .
the dispersion relation of these modes can be obtained by setting the determinant of @xmath49 to zero .
this results in @xmath50 where @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 , and @xmath56 .
the zeros of the first three factors in the left - hand side of the last equation give the frequencies of the scissors modes , whereas the zeros of the sixth - order polynomial give the frequencies of the breathing modes .
so far we have not produced a new result , since these frequencies have been calculated previously .
however , we have demonstrated a new way of calculating the frequencies of scissors modes in a relatively simple manner . the most important part of this paper is therefore contained in the next section , where we consider also the nonlinear effects produced by the mean - field interaction .
in this section we consider the equations of motion for the variational parameters by taking into account several higher - order terms in the deviations @xmath57 and @xmath58 .
we have calculated these equations analytically up to second order , but they turn out to be rather lengthy and contain terms that couple the breathing modes and the scissors modes . as we discuss below , in the present experiments with axially symmetric traps the coupling of the scissors modes with the quadrupole mode is always of importance , but we leave the treatment of this more complicated situation to future work . for simplicity , therefore , we focus here on triaxial traps , in which case we can ignore the breathing modes .
this is achieved by simply setting @xmath59 in the full equations of motion . up to second order
, the remaining 6 first - order equations for @xmath60 can be reduced to 3 second - order equations for @xmath61 , @xmath62 , and @xmath63 by eliminating the imaginary parts .
we find in detail @xmath64 \label{c1},\ ] ] @xmath65 \label{c2},\ ] ] and @xmath66 \label{c3}.\ ] ] these equations show that if we ignore the second - order terms we recover our previous three uncoupled scissors modes with frequencies @xmath67 , @xmath68 and @xmath69 .
it is interesting to observe that the coupling terms couple the three scissors modes such that if one mode is not excited the other two modes will not be coupled .
in other words , to get a nonzero coupling up to this order , the three modes must be excited simultaneously .
we believe that this important result is not an artefact of the gaussian approximation , but also holds for an exact calculation using the bogoliubov theory . for higher - order couplings the last conclusion is no longer true .
two modes can then be coupled , even when the third is not excited .
from now on , therefore , we assume without loss of generality that only the @xmath70 and @xmath71 modes are excited , while @xmath72 .
moreover , to investigate the possibility of a resonant coupling between these two modes , we have considered coupling terms up to ninth order .
similar to the second - order case explained above , the equations of motion can thus be expressed as @xmath73 @xmath74 where @xmath75 and the sum @xmath76 does not excced 9 , which is the order up to which we have chosen to expand the equations of motion .
the coefficients @xmath77 , @xmath78 are given in terms of the trap parameters @xmath1 , @xmath2 , @xmath3 , and the interaction parameter @xmath79 .
a resonance between these two modes takes place when the frequency of the coupling terms is equal to the frequency of the zeroth - order term , i.e. , the first two terms .
the frequency of each coupling term in the above summations is determined by substituting for @xmath80 and @xmath81 their zeroth - order solutions .
therefore , the latter frequencies will be a certain linear combination of the zeroth - order scissors modes frequencies . imposing the above resonance condition on each coupling term thus results in a relation between the two scissors modes frequencies @xmath67 and @xmath68 .
inspecting all coupling terms up to the ninth order , we found a very small number of terms for which the relation between @xmath67 and @xmath68 can be satisfied by real values of @xmath82 , @xmath83 , and @xmath84 .
ultimately , we find only a resonance when either @xmath85 or @xmath86 is satisfied .
the first resonance condition leads in the experimentally relevant axially symmetric case , where @xmath87 , to a value for the anisotropy ratio @xmath88 that is equal to 1 .
this clearly corresponds to a spherically symmetric condensate . since in this case
the scissors modes are degenerate with the quadrupole breathing modes , which we have ignored here , we focus from now on only on the second resonance condition .
it is interesting to mention that the second resonance condition is exactly the same as the one observed experimentally @xcite . for a resonance of this kind the resonant coupling terms turn out to be of seventh order , and
the equations of motion in that case read @xmath89 @xmath90 if we neglect all nonresonant terms . here
@xmath91 , @xmath92 and @xmath93 are functions of @xmath82 , @xmath83 , @xmath84 , and @xmath79 that are given explicitly in appendix [ app ] .
it is important to note here that our neglect of the nonresonant terms is justified when we are close enough to resonance .
this is similar to the rotating - wave approximation known from quantum optics .
we see that the coupling terms indeed lead to the above - mentioned resonance condition , by inserting in them the zeroth - order solutions , i.e. , @xmath94 and @xmath95 . for example , the coupling terms in eq .
( [ c1n ] ) have a total frequency of @xmath96 . separating out @xmath97 as a prefactor for the whole equation , the coupling term oscillates thus as @xmath98
therefore , a resonance takes place when @xmath99 , i.e. , when the condition in eq .
( [ cond2 ] ) is met .
similarly , the coupling term of eq .
( [ c2n ] ) is also oscillating with a frequency of @xmath100 .
sufficiently close to resonance we can write @xmath70 and @xmath71 as a product of two functions . one of them describes the slow envelope and the other the fast oscillation with the uncoupled scissors - mode frequency .
in particular , we have @xmath101 and @xmath102 where @xmath103 and @xmath104 are the slowly varying envelope functions . substituting these expressions into eqs .
( [ c1n ] ) and ( [ c2n ] ) , ignoring second - order time derivatives of @xmath104 and @xmath103 , and then eliminating @xmath103 , we obtain the following equation for @xmath104 @xmath105 where @xmath106 this equation has a solution of the form @xmath107^{1\over1-\varepsilon } \label{sol},\ ] ] where @xmath108 and @xmath109 are two constants of integration that are determined by the initial conditions .
note that the relevant quantity here is @xmath110 , which represents the actual envelope of the oscillation and is given by @xmath111^{1\over2(1-\varepsilon)}.\ ] ] in first instance we might think that the real part of @xmath104 is the relevant quantity .
however , in eqs .
( [ geq ] ) and ( [ feq ] ) we should in principle have taken the real part of the right - hand side . if we do that we automatically are lead to the condition that @xmath110 is the envelope of the oscillation . for definiteness sake
let us take the initial conditions @xmath112 and @xmath113 , where @xmath114 and @xmath115 are the real and imaginary parts of @xmath104 , respectively .
physically , this set of initial conditions corresponds to exciting the scissors modes in the @xmath116 and the @xmath6-planes simultaneously .
this should be performed experimentally by initially rotating the condensate in the @xmath116-plane by an angle @xmath117 and around the @xmath9-axis by an angle @xmath118 and then releasing the condensate .
the initial angles @xmath117 and @xmath118 are related to the constants @xmath119 and @xmath120 by @xmath121 latexmath:[\[{\dot f}_{0}={\alpha\omega_{xy}^2\over\omega_{xz } } ( 2|b_{x}^{(0)}-b_z^{(0)}|\cos{\theta_0}\;\sin{\theta_0})^3 conditions the constants @xmath108 and @xmath109 are given by @xmath123 @xmath124 using these expressions and the experimental parameters from ref .
@xcite , we give in fig . [ 2a ] the real part of @xmath125 .
this clearly shows how the energy is being exchanged between the two modes .
an interesting property of eq .
( [ f2eq ] ) is that exactly on resonance , i.e. , @xmath99 , its solution becomes nonoscillatory .
indeed we find that in this case the solution is @xmath126 where again @xmath127 and @xmath128 are constants that are determined by the initial conditions . for the above initial conditions this is a decreasing function in time .
physically , this means that , unlike the case of an off - resonant oscillation , it takes an infinite time for the energy that is transferred from the scissors mode in the @xmath116-plane to the scissors mode in the @xmath6-plane to get back to the mode in the @xmath116-plane . in fig .
[ 2b ] we show this resonance behavior .
finally , we can show that in the limit @xmath129 the oscillatory solution given in eq .
( [ sol ] ) reduces to the nonoscillatory one at resonance given by eq .
( [ sol_dec ] ) .
we have explored the role of the mean - field interaction in coupling the three scissors modes of a bose - einstein condensate . a variational approach with a gaussian trial wave function , that contains a number of variational parameters describing the scissors modes ,
provides a relatively simple way in which we can extract the main features of this coupling .
to first order in the deviations in the variational parameters from their equilibrium values we reproduce the correct frequencies of the scissors modes .
to second order we show that it is not possible to have two modes that are coupled if the third mode is not excited . instead
, all three modes need to be excited for nonlinear dynamics to occur . at higher orders we find that it is possible to couple only two modes . in this case
we find a resonance behavior if @xmath130 or @xmath131 .
up to the nineth order in the deviations , these are the only two cases of resonance that we have found .
close to resonance the equations of motion have been solved exactly using an envelope approach .
the resulting dynamics is similar to a beating between two modes with a beating frequency @xmath132 .
we notice that the observed resonance behavior @xcite occurs exactly for the same condition that we have obtained , namely @xmath130 .
also the fact that we did not find any other resonance condition up to the ninth order , indicates that the coupling terms that lead to this resonance are also responsible for the experimental resonance in the triaxial case .
furthermore , it is important to note here that , quite generally , the down - conversion process from one excitation quantum into two excitation quanta with half the energy each , i.e. , the so - called beliaev damping @xcite , vanishes for the scissors modes .
this is so because of the negative parity of the scissors modes . in terms of fluctuations of the anihilation operator @xmath133 given by @xmath134 ,
the beliaev damping process is accounted for by an interaction term proportional to @xmath135 . here
@xmath136 is again the condensate wave function . using our variational gaussian wave function given by eq .
( [ trial ] ) , we clearly see that with only two scissors modes present this integral vanishes , since the integrant is an odd function . we believe that this result is independent of our trial wave function and also true within an exact approach @xcite . in our case the seventh - order coupling terms in eqs .
( [ c1n ] ) and ( [ c2n ] ) correspond to a quadratic collisional damping process , where three excitation quanta decay into four excitation quanta with half the frequency and one excitation quantum with the same frequency .
this is shown schematically in fig .
[ schem ] .
this makes sense physically since this is the lowest - order nonvanishing process if beliaev damping is forbidden and we are forced to apply the interaction term proportional to @xmath137 twice to accomplish the down conversion .
another point worth mentioning here is that in the experiments the resonance is studied in most detail for an axially symmetric trap with an anisotropy ratio close to @xmath138 , when the @xmath6-scissors mode is degenerate with the quadrupole mode , which we have excluded in the present calculation . in that case
we may argue that one scissors - mode quantum can decay into either another scissors - mode quantum and a quadrupole - mode quantum or into two quadrupole - mode quanta .
these are in principle also beliaev damping processes , which are however again forbidden for similar reasons as before .
nevertheless , a good understanding of these experiments requires an analysis that includes not only the @xmath6 and the @xmath116 scissors modes but also the quadrupole mode , because of the degeneracy that occures in this case . however , our theory should be directly applicable to the experiments with a triaxial trap .
we hope that in the future more detailed experiments of this kind will also be performed , to make a direct comparison between our theory and experiment possible .
we would like to thank c. j. foot and p. drummond for useful discussions .
this work is supported by the stichting voor fundamenteel onderzoek der materie ( fom ) , which is financially supported by the nederlandse organisatie voor wetenschappelijk onderzoek ( nwo ) .
d. s. jin , j. r. ensher , m. r. mathews , c. e. wieman , and e. a. cornell , phys .
lett . * 77 * , 420 ( 1996 ) .
mewes , m. r. andrews , n. j. van druten , d. s. durfee , c. g. townsend , and w. ketterle , phys .
lett . * 77 * , 988 ( 1996 ) .
s. stringari , phys .
77 * , 2360 ( 1996 ) .
k. g. singh and d. s. rokhsar , phys .
lett . * 77 * , 1667 ( 1996 ) .
m. edwards , p. a. ruprecht , k. burnett , r. j. dodd , and c. w. clarck , phys
lett . * 77 * , 1671 ( 1996 ) .
y. castin and r. dum , phys .
* 77 * , 5315 ( 1996 ) .
v. m. perez - carcia , h. michinel , j. i. cirac , m. lewenstein , and p. zoller , phys .
. lett . * 77 * , 5320 ( 1996 ) .
d. gury - odelin and s. stringari , phys .
lett . * 83 * , 4452 ( 1999 ) .
m. o. da c. pires and e. j. v. de passos , j. phys .
b : at . mol .
. phys . * 33 * , 3929 ( 2000 ) .
o. m. marag , s. a. hopkins , j. arlt , e. hodby , g. hechenblaikner , and c. j. foot , phys .
* 84 * , 2056 ( 2000 ) .
e. hodby , o. m. marag , g. hechenblaikner , and c. j. foot , phys .
lett . * 86 * , 2196 ( 2001 ) .
o. m. marag , g. hechenblaikner , e. hodby , and c. j. foot , cond - mat/0101213 .
b. jackson and e. zaremba , cond - mat/0105465 . c. j. foot , private communication .
g. baym and c. j. pethick , phys .
lett . * 76 * , 6 ( 1996 ) .
f. dalfovo , s. giorgini , l. p. pitaevskii , and s. stringari , rev .
phys . * 71 * , 463 ( 1999 ) .
s. t. beliaev , sov .
jetp * 34 * , 323 ( 1958 ) .
see also g. hechenblaikner , e. hodby , s. a. hopkins , o. m. marag , and c. j. foot , cond - mat/0105394 .
the matrices @xmath47 and @xmath139 are given by @xmath140 and @xmath141 the coupling coefficients @xmath91 , @xmath92 and @xmath93 , appearing first in eqs . ( [ c1n ] ) and ( [ c2n ] ) , are given by @xmath142\\\nonumber & /&\left({\pi } ^{\frac{6}{5}}\,{{{\omega } _
x}}^4\,{{{\omega } _
y}}^2\ , \omega_{xy}^4 \omega_{xz}^4\right),\end{aligned}\ ] ] @xmath143\\\nonumber & /&\left({\pi } ^{\frac{6}{5}}\ , { { { \omega } _ x}}^4\,{{{\omega } _
y}}^2\ , \omega_{xy}^4\ , \omega_{xz}^4\right),\end{aligned}\ ] ] and @xmath144 respectively . | we explore the nonlinear coupling of the three scissors modes of an anisotropic bose - einstein condensate .
we show that only when the frequency of one of the scissors modes is twice the frequency of another scissors mode , these two modes can be resonantly coupled and a down conversion can occur .
we perform the calculation variationally using a gaussian trial wave function .
this enables us to obtain simple analytical results that describe the oscillation and resonance behaviour of the two coupled modes .
23 cm -1 cm 15 cm |
let @xmath1 with @xmath2 , be a sequence of random vectors .
a common functional of interests @xmath3 of @xmath4 is defined by @xmath5,\qquad x\in r^d,\ ] ] where @xmath6 is a certain sequence of positive constants and @xmath7 is a real function on @xmath8 .
such functionals arise in non - parametric estimation problems , where @xmath9 may be a kernel function @xmath10 or a squared kernel function @xmath11 and the sequence @xmath12 is the bandwidth used in the non - parametric regression .
the uniform convergence of @xmath3 in the situation that the @xmath4 satisfy certain stationary conditions was studied in many articles .
liero @xcite , peligrad @xcite and nze and doukhan @xcite considered the uniform convergence over a fixed compact set , while masry @xcite , bosq @xcite and fan and yao @xcite gave uniform results over an unbounded set .
these work mainly focus on random sequence @xmath13 which satisfies different types of mixing conditions . investigating a more general framework ,
andrews @xcite gave result on kernel estimate when the data sequence is near - epoch dependent on another underlying mixing sequence .
more recently , hansen @xcite provided a set of general uniform consistency results , allowing for stationary strong mixing multivariate data with infinite support , kernels with unbounded support and general bandwidth sequences .
kristensen @xcite further extended hansen s results to the heterogenous dependent case under @xmath14-mixing condition .
also see wu , huang and huang @xcite for kernel estimation in general time series settings . in comparison to the extensive results where the @xmath15 comes from a stationary time series data , there is little investigation on the the uniform convergence of @xmath3 for the @xmath15 being a non - stationary time series . in this
regard , gao , li and tjstheim @xcite derived strong and weak consistency results for the case where the @xmath15 is a null - recurrent markov chain .
wang and wang @xcite worked with partial sum processes of the type @xmath16 where @xmath17 is a general linear process . while the rate of convergence in gao , li and tjstheim @xcite is sharp , they impose the independence between @xmath18 and @xmath15 . using a quite different method , wang and wang @xcite allowed for the endogeneity between @xmath18 and @xmath15 , but their results hold only for the @xmath19 being in a fixed compact set .
the aim of this paper is to present a general uniform consistency result for @xmath3 with broad applicability . as a framework
, our assumption on the @xmath13 is only related to the conditional variance of the martingale , that is , @xmath20 $ ] .
see assumption [ assumption2.3 ] in section [ sec2 ] .
this of course is a `` high level '' condition , but it in fact is quite natural which holds true for many interesting and important examples , including stationary mixing time series , stationary iterated random function and harris recurrent markov chain .
see sections [ sec22 ] and [ sec23 ] for the identification of assumption [ assumption2.3 ] .
this condition also holds true for @xmath0 processes with innovations being a linear process , but the identification is complicated and requires quite different techniques .
we will report related work in a separate paper . by using the established result
, we investigate the uniform convergence of the nadaraya
watson estimator in a non - linear cointegrating regression model .
it confirms that the uniform asymptotics in wang and wang @xcite can be extended to a unbounded set and the independence between the @xmath21 and @xmath13 in gao , li and tjstheim @xcite can be removed .
more importantly , our result not only provides sharp convergence rate , but also the optimal range for the uniform convergence to be held .
it should be mentioned that our work on the uniform upper and lower bound estimation for a functional of harris recurrent markov chain is of independent interests .
this paper is organized as follows .
our main results are presented in next section , which includes the establishment of a framework on the uniform convergence for a class of martingale and uniform upper and lower bound estimation for a functional of harris recurrent markov chain .
an application of the main results in non - linear cointegrating regression is given in section [ sec3 ] .
all proofs are postponed to section [ sec4 ] . throughout the paper
, we denote constants by @xmath22 which may be different at each appearance .
we also use the notation @xmath23 .
we make use of the following assumptions in the development of uniform convergence for the @xmath3 defined by ( [ in1 ] ) .
recall @xmath24 where @xmath25 is an integer .
[ assumption2.1 ] @xmath26 is a martingale difference , where @xmath27 , satisfying @xmath28 , a.s .
, for some @xmath29 specified in assumption [ assumption2.4 ] below .
[ assumption2.2 ] @xmath7 is a real function on @xmath8 satisfying @xmath30 and @xmath31 for all @xmath32 and some constant @xmath33 .
[ assumption2.3 ] there exist positive constant sequences @xmath34 and @xmath35 with @xmath36 for some @xmath37 such that @xmath38 = { \mathrm{o}}_p(c_n).\ ] ] [ assumption2.4 ] @xmath39 and @xmath40 , where @xmath41 is defined as in assumption [ assumption2.3 ] and @xmath42 is defined as in assumption [ assumption2.1 ] .
we remark that assumption [ assumption2.1 ] ensures that @xmath43 is a martingale for each fixed @xmath19 and is quite weak .
clearly , assumption [ assumption2.1 ] is satisfied if @xmath21 is a sequence of i.i.d .
random variables , which is independent of @xmath44 , with @xmath45 and @xmath46 .
the lipschitz condition used in assumption [ assumption2.2 ] is standard in the investigation of uniform consistency , where we do not require the @xmath7 to have finite compact support .
assumption [ assumption2.3 ] is a `` high level '' condition for the @xmath15 .
we use it here to provide a framework . in sections
[ sec22 ] and [ sec23 ] , we will show that this condition is in fact quite natural which holds true by many interesting and important examples .
assumption [ assumption2.4 ] provides the connections among the moment condition required in assumption [ assumption2.1 ] , the condition ( [ m1 ] ) and the bandwidth @xmath12 . in many applications , we have @xmath47 , where @xmath48 and @xmath49 is a slowly varying function at infinite .
see section [ sec23 ] and examples [ example1][example3 ] in section [ sec22 ] . in the typical situation that @xmath47 , if there exists a such that @xmath50 , the @xmath42 required in assumption [ assumption2.1 ] can be specified to @xmath51 + 1 $ ] .
we have the following main result . [ th4 ] under assumptions [ assumption2.1][assumption2.4 ] ,
we have @xmath52 \biggr| = { \mathrm{o}}_p \bigl[(c_n \log n)^{1/2 } \bigr].\ ] ] if ( [ m1 ] ) is replaced by @xmath53 = { \mathrm{o}}(c_n),\qquad\mbox{a.s.},\ ] ] the result ( [ m10 ] ) can be strengthened to @xmath54 \biggr| = { \mathrm{o}}\bigl[(c_n \log n)^{1/2 } \bigr],\qquad\mbox{a.s.}\ ] ] theorem [ th4 ] can be extended to uniform convergence for the @xmath55 $ ] over unrestricted space @xmath8 .
this requires additional condition on the @xmath15 and the tail decay for the function @xmath7 .
[ th4a ] in addition to assumptions [ assumption2.1][assumption2.4 ] , @xmath56 $ ] and there exists a @xmath57 such that @xmath58.\ ] ] then , @xmath59 \biggr| = { \mathrm{o}}_p \bigl[(c_n \log n)^{1/2 } \bigr].\ ] ] similarly , if ( [ m1 ] ) is replaced by ( [ m1a9 ] ) and ( [ p09 ] ) is replaced by @xmath60,\qquad\mbox{a.s.},\ ] ] then @xmath61 \biggr| = { \mathrm{o}}\bigl[(c_n \log n)^{1/2 } \bigr],\qquad\mbox{a.s.}\ ] ] [ rem21 ] theorems [ th4][th4a ] allow for the @xmath13 to be a stationary or non - stationary time series .
see examples [ example1][example3 ] and section [ sec23 ] below .
more examples on non - stationary time series will be reported in a separate paper .
the rates of convergence in both theorems are sharp .
for instance , in the well - known stationary situation such as those appeared in examples [ example1][example3 ] , the @xmath41 can be chosen as @xmath62 .
hence , when there are enough moment conditions on the @xmath21 ( i.e. , @xmath42 is large enough ) , we obtain the optimal rate @xmath63 , by taking @xmath64 . in non - stationary situation
, the rate of convergence is different .
in particular we have @xmath65 for the @xmath13 to be a random walk given in corollary [ th3 ] .
the reason behind this fact is that the amount of time spent by the random walk around any particular point is of order @xmath66 rather than @xmath67 for a stationary time series . for more explanation in this regard
, we refer to wang and phillips @xcite .
this section provides several stationary time series examples which satisfy assumption [ assumption2.3 ] .
examples [ example1 ] and [ example2 ] come from wu , huang and huang @xcite , where more general settings on the @xmath13 are established .
example [ example3 ] discusses a strongly mixing time series .
this example comes from hansen @xcite . by making use of other related works such as peligrad @xcite , nze and doukhan @xcite , masry @xcite , bosq @xcite and andrews @xcite ,
similar results can be established for other mixing time series like @xmath68-mixing and near - epoch - dependent time series . in these examples ,
we only consider the situation that @xmath69 .
the extension to @xmath70 is straightforward and hence the details are omitted . throughout examples [ example1][example3 ] ,
we use the notation @xmath71 .
example on the harris recurrent markov chains , which allows for stationary ( positive recurrent ) or non - stationary ( null recurrent ) series , is given in section [ sec23 ] .
in the section , we also consider the uniform lower bound , which is of independent interests .
more examples on @xmath0 processes with innovations being linear processes will be reported in a separate paper .
[ example1 ] let @xmath72 be a linear process defined by @xmath73 where @xmath74 is a sequence of i.i.d .
random variables with @xmath75 and a density @xmath76 satisfying @xmath77 and @xmath78 where @xmath79 denotes the @xmath80-order derivative of @xmath81 .
suppose that @xmath82 and @xmath83 , and in addition assumption [ assumption2.2 ] , @xmath7 has a compact support .
it follows from section 4.1 of wu , huang and huang @xcite that , for any @xmath84 and @xmath85 ,
@xmath86 \biggr| = { \mathrm{o}}\biggl[\sqrt{\frac{\log n}{nh } } + n^{-1/2 } l(n ) \biggr ] , \qquad\mbox{a.s.},\ ] ] where @xmath49 is a slowly varying function .
note that @xmath13 is stationary process with a bounded density @xmath87 under the given conditions on @xmath88 .
simple calculations show that @xmath89 = { \mathrm{o}}_p(nh),\ ] ] that is , @xmath13 satisfies assumption [ assumption2.3 ]
. [ example2 ] consider the non - linear time series of the following form @xmath90 where @xmath91 is a bivariate measurable function and @xmath88 are i.i.d .
this is the iterated random function framework that encompasses a lot of popular non - linear time series models .
for example , if @xmath92 , it is the threshold autoregressive ( tar ) model ( see tong @xcite ) .
if @xmath93 , then it is autoregressive model with conditional heteroscedasticity ( arch ) model . other non - linear time series models , including random coefficient model , bilinear autoregressive model and exponential autoregressive model can be fitted in this framework similarly .
see wu and shao @xcite for details . in order to identify assumption [ assumption2.3 ] ,
we need some regularity conditions on the initial distribution of @xmath94 and the function @xmath95 . define @xmath96 denote by @xmath97 the conditional density of @xmath98 at @xmath19 given @xmath94 .
further let @xmath99 and @xmath100^{1/2 } \quad\mbox{and}\quad j(x ) = \biggl [ \int _ { { r } } \biggl| \frac{\partial}{\partial x } g'(y\mid x ) \biggr|^2 \,{\mathrm{d}}y \biggr]^{1/2},\ ] ] @xmath101 and @xmath102 can be interpreted as a prediction sensitivity measure .
these quantities measure the change in 1-step predictive distribution of @xmath98 with respect to change in initial value @xmath94 .
suppose that : there exist @xmath103 and @xmath104 such that @xmath105 < 0 \quad\mbox{and}\quad e l_{\varepsilon _ 0}^2 < 1;\ ] ] @xmath106 < \infty$ ]
; in addition to assumption [ assumption2.2 ] , @xmath7 has a compact support .
it follows from section 4.2 of wu , huang and huang @xcite that , for any @xmath84 and @xmath85 @xmath86 \biggr| = { \mathrm{o}}\biggl[\sqrt{\frac{\log n}{nh } } + n^{-1/2 } l(n ) \biggr ] , \qquad\mbox{a.s.},\ ] ] where @xmath49 is a slowly varying function .
note that @xmath13 has a unique and stationary distribution under the given condition ( i ) and ( ii ) .
see diaconis and freedman @xcite , for instance .
simple calculations show that @xmath89 = { \mathrm{o}}_p(nh),\ ] ] that is , @xmath13 satisfies assumption [ assumption2.3 ] .
[ example3 ] let @xmath107 be a strictly stationary time series with density @xmath87 .
suppose that : @xmath13 is strongly mixing with mixing coefficients @xmath108 that satisfy @xmath109 where @xmath110 and @xmath111 ; @xmath112 for some @xmath113 satisfying @xmath114 and there is some @xmath115 such that for all @xmath116 , @xmath117 where @xmath118 is the joint density of @xmath119 ; in addition to assumption [ assumption2.2 ] , @xmath7 has a compact support .
it follows from theorem 4 ( with @xmath120 ) of hansen @xcite that , for any @xmath84 and @xmath121 with @xmath122 , @xmath123 \biggr| = { \mathrm{o}}_p \biggl[\sqrt{\frac{\log n}{nh } } \biggr].\ ] ] if in addition @xmath124 , the result ( [ exa3 ] ) can be strengthened to almost surely convergence .
simple calculations show that @xmath89 = { \mathrm{o}}_p(nh),\ ] ] that is , @xmath13 satisfies assumption [ assumption2.3 ] .
let @xmath107 be a harris recurrent markov chain with state space @xmath125 , transition probability @xmath126 and invariant measure @xmath127 .
we denote @xmath128 for the markovian probability with the initial distribution @xmath129 , @xmath130 for correspondent expectation and @xmath131 for the @xmath132-step transition of @xmath107 .
a subset @xmath133 of @xmath134 with @xmath135 is called @xmath133-set of @xmath107 if for any @xmath136 , @xmath137 where @xmath138 and @xmath139 . as is well - known ,
@xmath133-sets not only exist , but generate the entire sigma @xmath140 , and for any @xmath133-sets @xmath141 and any probability measure @xmath142 on @xmath125 , @xmath143 where @xmath144 .
see nummelin @xcite , for instance .
let a @xmath133-set @xmath133 and a probability measure @xmath145 on @xmath146 be fixed .
define @xmath147 } \nu p^k(d),\qquad t\ge0.\ ] ] by recurrence , @xmath148 . by virtue of ( [ 13 ] ) , the asymptotic order of @xmath149 depends only on @xmath107 . as in chen @xcite , a harris recurrent markov chain @xmath107 is called @xmath150-regular if @xmath151 where @xmath152 .
it is interesting to notice that , under the condition ( [ d1 ] ) , the function @xmath149 is regularly varying at infinity , that is , there exists a slowly varying function @xmath153 such that @xmath154 .
this implies that the definition of @xmath150-regular harris recurrent markov chain is similar to that of @xmath150-null recurrent given in karlsen and tjstheim @xcite and gao , li and tjstheim @xcite , but it is more natural and simple .
the following theorem provides uniform upper and lower bounds for a functional of @xmath13 .
the upper bound implies that @xmath13 satisfies assumption [ assumption2.3 ] , allowing for the @xmath13 being stationary ( @xmath155 , positive recurrent markov chain ) and non - stationary ( @xmath156 , null recurrent markov chain ) .
the lower bound plays a key role in the investigation of the uniform consistency for the kernel estimator in a non - linear cointegrating regression , and hence is of independent interests .
see section [ sec3 ] for more details .
both upper and lower bounds are optimal , which is detailed in remarks [ rem22 ] and [ rem23 ] .
[ th1 ] suppose that : @xmath107 is a @xmath150-regular harris recurrent markov chain , where the invariant measure @xmath127 has a bounded density function @xmath157 on @xmath91 ; in addition to assumption [ assumption2.2 ] , @xmath158 .
then , for any @xmath159 satisfying @xmath160 for some @xmath161 , we have @xmath162 = { \mathrm{o}}_p\bigl[a(n ) h \bigr],\ ] ] where @xmath163 can be any finite integer . for a given sequence of constants @xmath164 , if there exists a constant @xmath165 such that , uniformly for @xmath67 large enough , @xmath166\ge a(n ) h / c_0,\ ] ] then , for any @xmath159 satisfying @xmath160 for some @xmath161 , we have @xmath167 \biggr\}^{-1 } = { \mathrm{o}}_p \bigl\{\bigl[a(n ) h\bigr]^{-1 } \bigr\}.\ ] ] [ rem22 ] the result ( [ 12a ] ) implies that , for any @xmath168 , there exists a constant @xmath169 such that @xmath170 \ge a(n ) h / c_\eta \biggr ) \ge1-\eta.\ ] ] this makes both bounds on ( [ 12 ] ) and ( [ 12a ] ) are optimal .
on the other hand , since the result ( [ 15 ] ) implies that @xmath171\ge a(n ) h ( 1-\eta ) /c_{\eta}\ ] ] for any @xmath168 , the condition ( [ wond ] ) is close to minimal .
note that random walk is a @xmath172-regular harris recurrent markov chain .
the following corollary on a random walk shows the range @xmath173 can be taken to be optimal as well .
[ th3 ] let @xmath174 be a sequence of i.i.d . random variables with @xmath175 , @xmath176 and the characteristic function @xmath177 of @xmath178 satisfying @xmath179 .
write @xmath180 . if in addition to assumption [ assumption2.2 ] , @xmath181 , then , for @xmath182 and @xmath183 where @xmath184 , we have @xmath185 = { \mathrm{o}}_p(\sqrt
n h)\ ] ] for any integer @xmath186 , and @xmath187 \biggr\}^{-1 } = { \mathrm{o}}_p \bigl\{(\sqrt
n h)^{-1 } \bigr\}\ ] ] for any @xmath188 .
[ rem23 ] for a random walk @xmath13 defined as in corollary [ th3 ] , it was shown in wang and phillips @xcite that @xmath189 \to_d\int f^2(s ) \,{\mathrm{d}}s l_w(1,y),\ ] ] where @xmath190 is a local time of a brownian motion @xmath191 , and @xmath192 if @xmath193 and @xmath194 if @xmath195 .
since @xmath196 for any @xmath197 , it follows from ( [ ez1 ] ) that the range @xmath198 in ( [ m15 ] ) can not be extended to @xmath199 for any @xmath200 .
[ rem24 ] as in examples [ example1][example3 ] , we may obtain a better result if @xmath201 is stationary ( positive null recurrent ) and satisfies certain other restrictive conditions .
indeed , kristensen @xcite provided such a result .
let @xmath202 be a time - homogeneous , geometrically ergodic markov chain .
denote the 1-step transition probability by @xmath203 , such that @xmath204 .
also denote the @xmath205-step transition probability by @xmath206 , such that @xmath207 . since @xmath13 is geometrically ergodic
, it has a density @xmath87 .
further suppose that : ( strong doeblin condition ) there exists @xmath208 and @xmath209 such that for all @xmath210 , @xmath211 @xmath212 exists and is uniformly continuous for all @xmath19 , for some @xmath213 , @xmath214<\infty$ ] for some @xmath215 , in addition to assumption [ assumption2.2 ] , @xmath7 has a compact support .
it follows from kristensen @xcite that , for any @xmath216 and @xmath217 , @xmath218 - g(x ) \int f^2(s ) \,{\mathrm{d}}s
\biggr| = { \mathrm{o}}_p \biggl[h^r + \sqrt { \frac{\log n}{nh } } \biggr],\ ] ] which yields ( [ 12 ] ) with @xmath219 and ( [ 12a ] ) with @xmath219 and @xmath220 , where @xmath221 is a constant such that @xmath222 .
[ rem25]it is much more complicated if @xmath13 is a null recurrent markov chain , even in the simple situation that @xmath13 is a random walk defined as in corollary [ th3 ] . in this regard
, we have ( [ ez1 ] ) , but it is not clear at the moment if it is possible to establish a result like @xmath223 - \int f^2(s ) \,{\mathrm{d}}s l_w(1,x ) \biggr| = { \mathrm{o}}_p ( c_n)\ ] ] for some @xmath224 and @xmath225 .
note that ( [ ez2 ] ) implies that @xmath226\to_p\int f^2(s ) \,{\mathrm{d}}s l_w(1,0)\ ] ] for any fixed @xmath227 .
this is a stronger convergence than that given in ( [ ez1 ] ) .
our experiences show that it might not be possible to prove ( [ ez2 ] ) without enlarging the probability space in which the @xmath13 hosts .
consider a non - linear cointegrating regression model : @xmath228 where @xmath229 is a stationary error process and @xmath230 is a non - stationary regressor .
let @xmath231 be a non - negative real function and set @xmath232 where @xmath233 .
the conventional kernel estimate of @xmath234 in model ( [ mo1 ] ) is given by @xmath235 the point - wise limit behavior of @xmath236 has currently been investigated by many authors . among them , karlsen , myklebust and tjstheim @xcite discussed the situation where @xmath230 is a recurrent markov chain .
wang and phillips @xcite and cai , li and park @xcite considered an alternative treatment by making use of local time limit theory and , instead of recurrent markov chains , worked with partial sum representations of the type @xmath237 where @xmath17 is a general linear process . in another paper ,
wang and phillips @xcite considered the errors @xmath21 to be serially dependent and cross correlated with the regressor @xmath13 for small lags . for other related works
, we refer to kasparis and phillips @xcite , park and phillips @xcite , gao _ et al . _
@xcite , marmer @xcite , chen , li and zhang @xcite , wang and phillips @xcite and wang @xcite .
this section provides a uniform convergence for the @xmath236 by making direct use of theorems [ th4 ] and [ th1 ] in developing the asymptotics . for reading convenience , we list the assumptions as follows .
[ assumption3.1 ] @xmath107 is a @xmath150-regular harris recurrent markov chain defined as in section [ sec3 ] , where the invariant measure @xmath127 has a bounded density function @xmath157 on @xmath91 ; @xmath26 is a martingale difference , where @xmath238 , satisfying @xmath239 , where @xmath240 for some @xmath241 .
[ assumption3.2 ] the kernel @xmath10 satisfies that @xmath242 , @xmath243 and for any @xmath244 , @xmath245 [ assumption3.3 ] there exists a real positive function @xmath87 such that @xmath246 uniformly for some @xmath247 and any @xmath248 , where @xmath249 can be chosen sufficiently small and @xmath250 .
assumption [ assumption3.1 ] is similar to , but weaker than those appeared in karlsen , myklebust and tjstheim @xcite , where the authors considered the point - wise convergence in distribution .
assumption [ assumption3.2 ] is a standard condition on @xmath231 as in the stationary situation .
the lipschitz condition on @xmath231 is not necessary if we only investigate the point - wise asymptotics .
see remark [ rem32 ] for further details .
assumption [ assumption3.3 ] requires a lipschitz - type condition in a small neighborhood of the targeted set for the functionals to be estimated .
this condition is quite weak , which may host a wide set of functionals .
typical examples include that @xmath251 ; @xmath252 ; @xmath253 ; @xmath254 .
we have the following asymptotic results . [ th31 ]
suppose assumptions [ assumption3.1][assumption3.3 ] hold , @xmath216 and @xmath255 where @xmath241 is given as in assumption [ assumption3.1 ] .
it follows that @xmath256^{-1/2 } \log^{1/2}n + h^{\alpha } \delta_n \bigr\},\ ] ] where @xmath257 , @xmath258 and @xmath35 satisfies that @xmath259\ge a(n ) h / c_0\ ] ] for some @xmath165 and all @xmath67 sufficiently large .
in particular , for the random walk @xmath13 defined as in corollary [ th3 ] , we have @xmath260 where @xmath261 for some @xmath188 and @xmath262 .
[ rem31 ] when a high moment exists on the error @xmath21 , the @xmath263 can be chosen sufficiently small so that there are more bandwidth choices in practice .
it is understandable that the results ( [ q1 ] ) and ( [ q1a ] ) are meaningful if only @xmath264 , which depends on the tail of the unknown regression function @xmath234 , the bandwidth @xmath12 and the range @xmath265 .
when @xmath234 has a light tail such as @xmath266 , @xmath267 may be bounded by a constant .
in this situation , the @xmath268 in ( [ q1a ] ) can be chosen to be @xmath269 for some @xmath188 .
in contrast to theorem [ th1 ] and remark [ rem23 ] , this kind of range @xmath270 might be optimal , that is , the @xmath268 can not be improved to @xmath271 , for any @xmath200 , to establish the same rate of convergence as in ( [ q1a ] ) .
[ rem32 ] both results ( [ q1 ] ) and ( [ q1a ] ) are sharp .
however , a better result can be obtained if we are only interested in the point - wise asymptotics for @xmath272 .
for instance , as in wang and phillips @xcite with minor modification , we may show that , for each @xmath19 , @xmath273 whenever @xmath13 is a random walk defined as in corollary [ th3 ] .
furthermore @xmath272 has an asymptotic distribution that is mixing normal , under minor additional conditions .
more details are referred to wang and phillips @xcite .
[ rem33 ] wang and wang @xcite established a similar result to ( [ q1a ] ) with the @xmath13 being a partial sum of linear process , but only for the @xmath19 being a compact support and imposing a bounded condition on @xmath21 .
the setting on the @xmath13 in this paper is similar to that given in gao , li and tjstheim @xcite , but our result provides the optimal range for the uniform convergence holding true and removes the independence between the error @xmath21 and @xmath13 required by gao , li and tjstheim @xcite .
proof of theorem [ th4 ] we split the set @xmath274 into @xmath275 balls of the form @xmath276 where @xmath277 $ ] , @xmath278 and @xmath279 are chosen so that @xmath280 .
it follows that @xmath281 \biggr| \nonumber \\ & & \quad\le \max_{0\le j\le m_n}\sup_{x\in a_{nj } } \sum _ { t=1}^{n } |u_t| \bigl|f\bigl[(x_t+x)/h \bigr]-f\bigl[(x_{t}+y_j)/h\bigr ] \bigr| \nonumber\\[-8pt]\\[-8pt ] & & \qquad { } + \max_{0\le j\le m_n } \biggl|\sum_{t=1}^{n } u_t f\bigl[(x_{t}+y_j)/h\bigr ] \biggr| \nonumber \\ & & \quad:= \lambda_{1n}+\lambda_{2n}.\nonumber\end{aligned}\ ] ] recalling the assumption [ assumption2.2 ] , it is readily seen that @xmath282-f \bigl[(x_{t}+y_j)/h\bigr ] \bigr| \nonumber \\ & \le & c \bigl(h m_n'\bigr)^{-1 } \sum _ { t=1}^n|u_t| \\ & \le & c ( c_n \log n)^{1/2 } \frac1n \sum _ { t=1}^n|u_t| = { \mathrm{o}}\bigl[(c_n \log n)^{1/2}\bigr ] , \qquad\mbox{a.s.}\nonumber\end{aligned}\ ] ] by the strong law of large number . in order to investigate @xmath283 ,
write @xmath284 $ ] and @xmath285 . recalling @xmath286 and @xmath287 , we have @xmath288 \biggr| \nonumber \\ & & { } + \max_{0\le j\le m_n } \biggl|\sum_{t=1}^{n } \bigl[\bigl|u_t - u_t'\bigr|+e\bigl(\bigl|u_t - u_t'\bigr| \mid{\mathcal f}_{t-1}\bigr ) \bigr ] f\bigl[(x_{t}+y_j)/h \bigr ] \biggr| \nonumber\\[-8pt]\\[-8pt ] & \le & \max_{0\le j\le m_n } \biggl|\sum_{t=1}^{n } u_t^ * f\bigl[(x_{t}+y_j)/h\bigr ] \biggr| + c \sum _ { t=1}^{n } \bigl[\bigl|u_t - u_t'\bigr|+e \bigl(\bigl|u_t - u_t'\bigr|\mid{\mathcal f}_{t-1}\bigr ) \bigr ] \nonumber \\ : \!\!&=&\lambda_{3n } + \lambda_{4n}.\nonumber\end{aligned}\ ] ] routine calculations show that , under @xmath289 and @xmath40 , @xmath290 \nonumber \\ &
\le & c \biggl(\frac{c_n}{\log n } \biggr)^{(1-{2p})/2 } \sum _ { t=1}^n \bigl[|u_t|^{2p}+e \bigl(|u_t|^{2p}\mid{\mathcal f}_{t-1}\bigr ) \bigr ] \nonumber\\[-8pt]\\[-8pt ] & \le & c ( c_n \log n)^{1/2 } \frac1n \sum _ { t=1}^n \bigl[|u_t|^{2p}+e \bigl(|u_t|^{2p}\mid{\mathcal f}_{t-1}\bigr ) \bigr ] \nonumber \\ & = & { \mathrm{o}}\bigl[(c_n \log n)^{1/2 } \bigr],\qquad\mbox{a.s.}\nonumber\end{aligned}\ ] ] by the strong law of large number again .
we next consider @xmath291 . as @xmath292\le2 ( e [ |u_t|^{2p}\mid{\mathcal f}_{t-1}])^{1/p}$ ] , a.s .
, assumptions [ assumption2.1 ] and [ assumption2.3 ] imply that @xmath293e \bigl [ \bigl(u_t^*\bigr)^2\mid{\mathcal f}_{t-1 } \bigr]={\mathrm{o}}_p(c_n).\ ] ] hence , for any @xmath294 , there exists a @xmath295 such that @xmath296 where @xmath297e [ ( u_t^*)^2\mid{\mathcal f}_{t-1}]$ ] , whenever @xmath67 is sufficiently large .
this , together with @xmath298 and the well - known martingale exponential inequality ( see , e.g. , de la pea @xcite ) , implies that , for any @xmath294 , there exists a @xmath299 ( @xmath132 is as in assumption [ assumption2.3 ] ) such that , whenever @xmath67 is sufficiently large , @xmath300 \nonumber \\ & & \quad\le p \biggl[\lambda_{3n } \ge m_0 ( c_n \log n ) ^{1/2 } , \max_{0\le j\le m_n } \sum_{t=1}^{n } \sigma_{tj}^2\le m_0 c_n \biggr]+ \eta \nonumber\\[-8pt]\\[-8pt ] & & \quad\le \sum_{j=0}^{m_n } p \biggl[\sum _ { t=1}^{n}u_t^ * f \bigl[(x_{k}+y_j)/h\bigr]\ge m_0 ( c_n \log n ) ^{1/2 } , \sum_{t=1}^{n } \sigma_{tj}^2 \le m_0 c_n \biggr ] + \eta \nonumber \\ & & \quad\le m_n \exp\biggl\{-\frac{m_0 ^ 2 c_n \log n } { 6 m_0c_n } \biggr\}+\eta\le m_nn^{-m_0/6}+\eta\le2 \eta,\nonumber\end{aligned}\ ] ] where we have used the following fact : @xmath301^d \le c_1 n^{(k+2)d}\ ] ] as @xmath302 and @xmath217
. this yields @xmath303 $ ] . combining ( [ p6])([p10 ] ) ,
we establish ( [ m10 ] ) .
to prove ( [ m11 ] ) , by checking ( [ p6])([p9 ] ) , it suffices to show that @xmath304,\qquad\mbox{a.s.}\ ] ] under the alternative condition ( [ m1a9 ] ) .
in fact , as in ( [ * ] ) , it follows from ( [ m1a9 ] ) that @xmath305e \bigl [ \bigl(u_t^*\bigr)^2\mid{\mathcal f}_{t-1 } \bigr]={\mathrm{o}}(c_n),\qquad\mbox{a.s.}\ ] ] similarly to proof of ( [ p10 ] ) , we have for sufficiently large @xmath306 ( @xmath307 , say ) , @xmath308 \nonumber \\ & & \quad= p \biggl[\lambda_{3n } \ge m_0 ( c_n \log n ) ^{1/2 } , \max_{0\le j\le m_n } \sum_{k=1}^{n } \sigma_k^2\le m_0 c_n\mbox { , i.o . } \biggr ] \nonumber\\ & & \quad\le \lim_{s\to\infty } \sum_{n = s}^{\infty } p \biggl[\lambda_{3n } \ge m_0 ( c_n \log n ) ^{1/2 } , \max_{0\le j\le m_n } \sum_{k=1}^{n } \sigma_k^2\le m_0 c_n \biggr ] \\ & & \quad\le\lim_{s\to\infty } \sum_{n = s}^{\infty } m_n \exp\biggl\ { -\frac{m_0 ^ 2 c_n \log n } { 6 m_0c_n } \biggr\ } \nonumber \\ & & \quad\le c \lim_{s\to\infty } \sum_{n = s}^{\infty } n^{(k+2)d } n^{-m_0/6}=0,\nonumber\end{aligned}\ ] ] which yields ( [ p12 ] ) . the proof of theorem [ th4 ] is now complete .
proof of theorem [ th4a ] we only prove ( [ m10a ] ) .
it is similar to prove ( [ m11a ] ) and hence the details are omitted .
we may write @xmath309\nonumber\\ & & \quad=\sum_{t=1}^{n } u_t f\bigl[(x_t+x)/h\bigr]i\bigl(\|x_t\|\le b_n/2 \bigr ) \nonumber\\[-8pt]\\[-8pt ] & & \qquad { } + \sum_{t=1}^{n } u_t f \bigl[(x_t+x)/h\bigr]i\bigl(\|x_t\|>b_n/2
\bigr ) \nonumber\\ & & \quad=\lambda_{5n } ( x)+\lambda_{6n}(x)\qquad \mbox{say}.\nonumber\end{aligned}\ ] ] it is readily seen from ( [ m10 ] ) and @xmath310\\ & & { } + \sup_{\|x\| > b_n/2 } \bigl|f(x / h)\bigr|\sum_{t=1}^{n } \\ & \le & { \mathrm{o}}_p \bigl[(c_n \log n)^{1/2 } \bigr]\end{aligned}\ ] ] as @xmath311 , a.s . by the strong law . as for @xmath312 , we have @xmath313 \\ & \le & c \sum_{t=1}^{n } p\bigl(\|x_t \|>b_n/2 \bigr ) \le c b_n^{-k_0 } \sum _ { t=1}^n e\|x_t\|^{k_0 } \\ &
= & { \mathrm{o}}\bigl [ ( c_n \log n)^{1/2 } \bigr],\end{aligned}\ ] ] which yield @xmath314 $ ] . taking these estimates into ( [ h1 ] )
, we obtain ( [ m10a ] ) .
the proof of theorem [ th4a ] is complete .
proof of theorem [ th1 ] first , assume there exists a @xmath315 such that @xmath316 for some @xmath317 and probability measure @xmath145 on @xmath125 with @xmath318 .
under this addition assumption , theorem [ th1 ] can be established by using the so - called split chain technique . to this end , define new random variables @xmath319 and @xmath320 by @xmath321 where @xmath322 .
as easily seen , @xmath323 is a harris recurrent markov chain with state space @xmath324 and @xmath325 has the same transition probability @xmath126 as those of @xmath326 .
since our result is free of the initial distribution , @xmath327 can be assumed to be identical with @xmath328 , that is , @xmath94 has the distribution @xmath145 .
further define @xmath329 , @xmath330 @xmath331 , and @xmath332,\qquad z_{jn}(x ) = \sum _ { k=\rho_{j-1}\wedge n+1}^{\rho_{j}\wedge
n } f^2 \bigl[(x_k+x)/h\bigr]\ ] ] for @xmath333 .
it is well known that the blocks @xmath334 are i.i.d .
blocks , @xmath335 having the distribution @xmath145 . hence , for each @xmath12 and @xmath19 , @xmath336 , where @xmath337 or @xmath338 is a sequence of i.i.d . random vectors .
furthermore , by recalling that @xmath127 has a bounded density function @xmath157 , @xmath339 and @xmath340 , we have @xmath341 \pi({\mathrm{d}}s ) \nonumber\\[-8pt]\\[-8pt ] & = & b h \int_{-\infty}^{\infty } f^2(s )
p(-x+sh ) \,{\mathrm{d}}s \le c^ * h\nonumber\end{aligned}\ ] ] for any @xmath342 and @xmath343 for any integer @xmath132 . see lemma 5.2 of karlsen and tjstheim @xcite or lemma b.1 of gao , li and tjstheim @xcite .
we also have the following lemma .
[ lem1]suppose that @xmath344 , where @xmath165 is a constant , and all @xmath345 , are different , where @xmath346 and @xmath347 for some @xmath348 .
then , @xmath349 \biggr| = { \mathrm{o}}_p\bigl[n^{-\varepsilon_0/4 } a(n ) h\bigr ] , \\
\label{96 } \delta_{n } & : = & \max_{0\le j\le m_n}e \biggl|\sum _ { k=0}^{d_n } \bigl[z_k^*(y_j)-ez_k^*(y_j ) \bigr ] \biggr| = { \mathrm{o}}\bigl[n^{-\varepsilon_0/4 } a(n ) h\bigr],\end{aligned}\ ] ] where @xmath263 is a constant such that @xmath350 . only consider @xmath351 , as the situation that @xmath352 is similar . to this end , write @xmath353 and @xmath354 .
we have @xmath355 \biggr| + \max_{0\le j\le m_n } \sum _ { i=1}^{d_n } \bigl [ \widehat z _
{ i}(y_j)+ e \widehat z _ { i}(y_j ) \bigr ] \nonumber\\[-8pt]\\[-8pt ] : \!\!&= & r_{1n}+r_{2n}.\nonumber\end{aligned}\ ] ] by taking @xmath356 in ( [ fi2 ] ) and noting @xmath357 , simple calculations show that @xmath358 which yields @xmath359 $ ] . as for @xmath360 , by using ( [ fi2 ] ) with @xmath361 and noting @xmath362 for any @xmath363 and some @xmath165 , the standard markov inequality implies that @xmath364 \biggr|\ge m n^{-\varepsilon_0/4}a(n)h \biggr ) \nonumber\\[-8pt]\\[-8pt ] & & \quad\le c m_n \exp\bigl(-m t
n^{-\varepsilon_0/4 } a(n)h + c_{\varepsilon } a(n)t^2h\bigr ) \nonumber \\ & & \quad\le c m_n \exp\bigl(-m n^{\varepsilon_0/4}/4 \bigr)\to0\nonumber\end{aligned}\ ] ] as @xmath365 .
hence , @xmath366 $ ] . combining ( [ b1])([l1 ] )
, we prove ( [ 95 ] ) .
the proof of ( [ 96 ] ) is similar except more simpler . indeed , by independence of @xmath367 , we obtain @xmath368 \biggr| \\ & & { } + 2\max_{0\le j\le m_n } \sum_{i=1}^{d_n}e \widehat z _ { i}(y_j ) \\
& \le & 2 \max_{0\le j\le m_n } d_n^{1/2 } \bigl[e\widetilde z _ { 1}^2(y_j)\bigr]^{1/2 } + c n^{-\varepsilon_0/2}a(n ) h \\
& \le & c n^{-\varepsilon_0/4}a(n ) h,\end{aligned}\ ] ] due to the fact : @xmath369 the proof of lemma [ lem1 ] is complete .
we are now ready to prove ( [ 12 ] ) and ( [ 12a ] ) under the additional condition ( [ 1 ] ) .
( [ 12 ] ) first . as in proof of ( [ p6 ] ) and ( [ p7 ] ) , but letting @xmath370 - 1+j / m_n',j=0 , 1,2,\ldots , m_n , $ ] where @xmath371 $ ] and @xmath372 + 1)m_n'$ ] , we have @xmath373 \le\max_{0\le j\le m_n}\sum _ { k=0}^n f^2 \bigl[(x_k+y_j)/h\bigr]+ c a(n ) h.\ ] ] note that @xmath374 \le\sum_{k=0}^{\rho _ { n(n+1 ) } } f^2\bigl[(x_k+x)/h\bigr ] = \sum _ { i=1}^{n(n+1 ) } z _ { i}(x),\ ] ] and @xmath375 is bounded in probability .
see , for example , chen @xcite . for each @xmath376
, there exist @xmath377 such that @xmath378 whenever @xmath67 is sufficiently large .
consequently , for each @xmath379 and @xmath67 large enough , @xmath380\ge a \biggr)\le p \biggl(\max_{0\le j\le m_n } \sum_{i=1}^{c_\varepsilon a(n ) } z _ { i}(y_j)\ge a \biggr)+\varepsilon.\ ] ] this , together with ( [ 95 ] ) with @xmath351 , implies ( [ 12 ] ) under ( [ 1 ] ) , since @xmath381.\ ] ] we next consider ( [ 12a ] ) under ( [ 1 ] ) .
to this regard , let @xmath382 - 1+j / m_n ' , j=0 , 1,2,\ldots , m_n , $ ] where @xmath383 $ ] and @xmath384 + 1)m_n'$ ] .
since @xmath385 } \sum_{t=1}^{n } \bigl|f^2\bigl [ ( x_t+x)/h\bigr]-f^2\bigl [ ( x_t+y_j)/h\bigr ] \bigr| \\ & & \quad \le c n h^{-1 } \max_{0\le j\le m_n-1}|y_{j+1}-y_j| \le cn^{-\varepsilon_0/2 } a(n)h,\end{aligned}\ ] ] it is readily seen that @xmath386 \ge\delta_{1n}-{\mathrm{o}}_{p } \bigl[n^{-\varepsilon_0/2 } a(n)h\bigr],\ ] ] where @xmath387 $ ] .
furthermore , by recalling ( [ a99 ] ) and noting that @xmath374 \ge\sum_{k=0}^{\rho _ { n(n)}\wedge n } f^2\bigl[(x_k+x)/h\bigr ] = \sum _ { i=1}^{n(n ) } z _ { in}(x),\ ] ] we have , for each @xmath388 and @xmath67 large enough , @xmath389 on the other hand , it follows from ( [ 95 ] ) with @xmath352 that @xmath390\\[-8pt ] & \ge & c_{1\varepsilon } a(n)\inf_{0\le j\le m_n}ez _ { 1n}(y_j ) -{\mathrm{o}}_{p}\bigl[n^{-\varepsilon_0/4 } a(n)h\bigr].\nonumber\end{aligned}\ ] ] combining ( [ 145])([147 ] ) , the result ( [ 12a ] ) under ( [ 1 ] ) will follow if we prove : there exists a @xmath391 such that @xmath392 for all @xmath67 sufficiently large . to prove ( [ 148 ] ) , first note that there exists a @xmath393 such that @xmath394 .
see lemma 3.3 of karlsen and tjstheim @xcite , for instance .
therefore , by taking @xmath395 + 1 $ ] , where @xmath396 is chosen later , we have for some @xmath391 , @xmath397 \\ & \ge&\frac{1}{d_n } \inf_{0\le j\le m_n}e \biggl(\sum _ { t=1}^ { n } f^2\bigl [ ( x_t+y_j)/h\bigr ] -i(\rho_{d_n } \le n ) \sum _ { t=1}^ { \rho_{d_n } } f^2\bigl [ ( x_t+y_j)/h\bigr ] \biggr ) \\ & \ge&\frac{1}{d_n } \biggl(\inf_{|x|\le b_n+1}e\sum _ { k=1}^nf^2\bigl [ ( x_t+x)/h \bigr]-m_n \biggr ) \\ & \ge&\frac{1}{d_n } \bigl[a(n)h / c_0-m_n \bigr ] \\ & \ge & b_0 h,\end{aligned}\ ] ] whenever @xmath67 is sufficiently large , where we have used the condition ( [ wond ] ) and the fact : it follows from ( [ 45 ] ) , ( [ 96 ] ) and @xmath398 if and only if @xmath399 that @xmath400 \biggr ] \\ & = & \max_{0\le j\le m_n } e \biggl[i(\rho_{d_n } \le
n ) \sum _ { i=1}^{d_n } z_i(y_j ) \biggr ] \\
& \le & d_n \max_{0\le j\le m_n } ez_1(y_j ) p \bigl(n(n)\ge d_n\bigr)+ \max_{0\le j\le m_n } e \biggl|\sum _ { i=1}^{d_n } \bigl[z_i(y_j)- ez_i(y_j ) \bigr ] \biggr| \\ & \le & b_2^{-1 } c^*h a^{-1}(n ) en^2(n)+{\mathrm{o}}\bigl[n^{-\varepsilon_0/4}a(n)h\bigr ] \\ & \le & c_0^{-1 } a(n)h/2\end{aligned}\ ] ] by choosing @xmath401 and @xmath67 sufficiently large .
this proves ( [ 148 ] ) and also completes the proof of ( [ 12a ] ) under ( [ 1 ] ) .
we now consider general situation .
let @xmath402 be fixed . define a transition probability @xmath403 on @xmath125 by @xmath404
let @xmath405 be an i.i.d .
bernoulli random variables with the common law @xmath406 and assume @xmath405 and @xmath202 are independent .
define a renewal sequence @xmath407 by @xmath408 with these notations , @xmath409 is a harris recurrent markov chain with the invariant measure @xmath127 .
the transition probability @xmath403 of @xmath409 satisfies the additional condition ( [ 1 ] ) and @xmath410 see chen @xcite , for instance . by virtue of these facts , it follows from the first part proof of ( [ 12 ] ) that , for any fixed @xmath186 and @xmath159 , @xmath411 = \sup_{|x|\le n^m } \sum_{k=1}^n f^2\bigl[(x_{\sigma(k)}+x)/h\bigr]= { \mathrm{o}}_p \bigl[a_t(n ) h\bigr].\ ] ] now by noting @xmath412)/n\to_{\mathrm{a.s . } } \lambda/(1-t)$ ] by the strong law and taking @xmath413 such that @xmath414 , simple calculations show that @xmath415 \le \sup_{|x|\le n^m } \sum_{k=1}^{\sigma([\lambda n ] ) } f^2\bigl[(x_{\sigma ( k)}+x)/h\bigr]={\mathrm{o}}_p\bigl[a(n ) h \bigr].\ ] ] similarly , @xmath416 = { \mathrm{o}}_p\bigl[a(n ) h\bigr]\ ] ] and hence the result ( [ 12 ] ) under general situation follows .
the proof of ( [ 12a ] ) under general situation is similar and hence the details are omitted .
proof of corollary [ th3 ] we first notice that : * @xmath417 is a harris null recurrent markov chain , satisfying ( [ 1 ] ) , @xmath418 and the invariant measure @xmath127 is the lebesgue measure . due to the fact ( f ) ,
( [ 12a1 ] ) follows immediately from theorem [ th1 ] .
to prove ( [ m15 ] ) , by theorem [ th1 ] , it suffices to show that ( [ wond ] ) holds true with @xmath419 and @xmath420 .
in fact , under the conditions of corollary [ th3 ] , @xmath421 has a density @xmath422 , satisfying @xmath423 , as @xmath424 , where @xmath425 , due to the central limit theorem .
this implies that @xmath426 for some @xmath427 and all sufficiently large @xmath132 .
hence , for @xmath428 and @xmath67 sufficiently large , we have @xmath429 & = & \inf_{|x|\le \tau_n \sqrt n+1 } \int_{-\infty}^{\infty } f^2 \bigl[(\sqrt k y+x)/h\bigr ]
p_k(y)\,{\mathrm{d}}y \\ & \ge&\frac{h}{\sqrt k}\inf_{|x|\le\tau_n \sqrt n+1}\int_{-\infty } ^{\infty } f^2(y ) p_k\bigl[(yh - x)/\sqrt k\bigr]\,{\mathrm{d}}y \\ &
\ge&\frac{h}{\sqrt k}\inf_{|x|\le3 \tau_n } p_k(x)\int _ { |y|\le m_1 } f^2(y)\,{\mathrm{d}}y \\ & \ge&\frac{a_0 h}{2\sqrt n } \int_{|y|\le m_1 } f^2(y)\,{\mathrm{d}}y,\end{aligned}\ ] ] where @xmath430 is chosen such that @xmath431 .
consequently , there exists a constant @xmath165 such that @xmath432\ge\inf_{|x|\le\tau_n \sqrt n+1 } \sum _ { k = n/2}^n e f^2 \bigl[(x_k+x)/h\bigr ] \ge\sqrt n h / c_0\ ] ] as required .
the proof of corollary [ th3 ] is now complete .
proof of theorem [ th31 ] we may write @xmath433 as @xmath434k_{h}(x_{t}-x)}{\sum_{t=1}^{n}k_{h}(x_{t}-x ) } \nonumber\\[-8pt]\\[-8pt ] : \!\!&= & \theta_{1n}(x)+\theta_{2n}(x).\nonumber\end{aligned}\ ] ] note that , for any @xmath265 , there exists a @xmath165 such that @xmath435=0 $ ] if @xmath436 .
it follows from assumption [ assumption3.3 ] that , whenever @xmath67 is sufficiently large , @xmath437}{\sum _ { t=1}^{n}k[(x_t - x)/h ] } \le ch^{\alpha } \delta_n.\ ] ] this , together with ( [ 12a ] ) [ taking @xmath438 in theorem [ th1 ] , implies that ( [ q1 ] ) will follow if we prove @xmath439 = { \mathrm{o}}_p \bigl[\bigl[a(n)h \bigr]^{1/2 } \log^{1/2}n \bigr].\ ] ] in fact , with @xmath240 and @xmath440 , we have @xmath441 since @xmath442 .
now , by recalling ( [ 12 ] ) , it is readily seen that the conditions of theorem [ th4 ] hold for @xmath443 and @xmath444 .
the result ( [ mp1 ] ) follows from ( [ m10 ] ) in theorem [ th4 ] .
the authors thank associate editor , two referees and professor jiti gao for helpful comments on the original version .
wang acknowledges the partial research support from the australian research council . | for a class of martingales , this paper provides a framework on the uniform consistency with broad applicability .
the main condition imposed is only related to the conditional variance of the martingale , which holds true for stationary mixing time series , stationary iterated random function , harris recurrent markov chains and @xmath0 processes with innovations being a linear process . using the established results , this paper investigates the uniform convergence of the nadaraya
watson estimator in a non - linear cointegrating regression model .
our results not only provide sharp convergence rate , but also the optimal range for the uniform convergence to be held .
this paper also considers the uniform upper and lower bound estimates for a functional of harris recurrent markov chain , which are of independent interests . |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.