Datasets:

ccdv
/

arxiv-summarization

Dataset card Data Studio Files Files and versions Community

Subset (2)

section · 216k rows

Split (3)

train · 203k rows

article stringlengths 0 745k	abstract stringlengths 0 92.7k
in the cores of some galaxy clusters , the hot ( @xmath710@xmath8 k ) intracluster medium ( icm ) can reach high enough densities that the gas should cool by a combination of thermal bremsstrahlung and line cooling in less than a hubble time , leading to cooling flows on the order of 100 - 1000 m@xmath3 yr@xmath4 @xcite . much effort has been devoted to detecting these cooling flows at lower temperatures using a variety of methods . intermediate temperature gas has been detected in a number of clusters via spectroscopy in the x - ray @xcite , uv ( e.g. , ; * ? ? ? * ; * ? ? ? * ) , optical ( e.g. , h@xmath9 ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , infrared ( e.g. , h@xmath10 ; * ? ? ? * ; * ? ? ? * ) , and millimeter ( e.g. , co ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? however , even if this intermediate - temperature gas originated in the hot phase , the amount of it is orders of magnitude less than predicted by simple cooling flow models . similarly , the star formation rate in the brightest cluster galaxy ( bcg ) , whether measured in the uv / optical @xcite or infrared @xcite , only accounts for a small fraction ( @xmath1110% ) of the predicted cooling flow . while there is some evidence that this intermediate - temperature gas and low - level star formation may represent a `` residual cooling flow '' ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , the fact that the majority of the x - ray - inferred cooling is unaccounted for has served as prime evidence that some form of feedback is at work in cluster cores . the exact feedback mechanism that prevents the cooling catastrophe from happening is still not fully understood . the leading hypothesis of mechanical agn - driven feedback is supported by the correlation between the total energy required to offset cooling and the amount of radio power required to inflate x - ray cavities @xcite . recently , @xcite reported the unique properties of a galaxy cluster at @xmath12 , spt - clj2344 - 4243 ( hereafter the phoenix cluster ) , which was initially discovered by the south pole telescope using the sunyaev zeldovich effect @xcite . this cluster is among the most massive ( m@xmath13 m@xmath3 ) , x - ray luminous ( l@xmath14 erg s@xmath4 ) , and strongest cooling ( m @xmath6 3820 m@xmath3 yr@xmath4 ) clusters discovered to date . the central galaxy in the phoenix cluster appears to be experiencing a 740 m@xmath3 yr@xmath4 starburst . however , it also harbors a powerful agn which makes it difficult to separate contributions to the ultraviolet , optical , and infrared emission from star - forming regions versus the agn in low - resolution imaging . further , the available ground - based data optical and space - based ultraviolet ( uv ) and infrared ( ir ) data presented in @xcite were unable to distinguish between in situ star formation or the late stages of a merger . in this letter , we present new _ hubble space telescope _ observations which improve significantly in depth and spatial resolution on the data presented in @xcite . in 2 we describe our analysis of these data , after which we present our new , detailed view of the phoenix cluster in 3 . in 4 we discuss the possible interpretations of these data , including whether or not we can separate the cooling flow scenario from a pure agn or merger scenario . we conclude with a summary of these results in 5 . throughout this letter we assume h@xmath15 = 70 km s@xmath4 mpc@xmath4 , @xmath16 = 0.27 , and @xmath17 = 0.73 . to study in detail the purported starburst in the core of the phoenix cluster , we obtained broadband imaging with the _ hubble space telescope _ _ wide field camera 3 _ ( hst wfc3 ) in five optical filters - f225w , f336w , f475w , f625w , f814w - which span rest - frame wavelengths from @xmath61000 to @xmath65500 . these observations were carried out over 2 orbits of director s discretionary time , with an average exposure time of @xmath61800s per filter ( pid # 13102 , pi mcdonald ) . in each filter , a series of two dithered exposures were obtained . the la cosmic @xcite software was run on individual exposures , generating accurate cosmic ray and bad pixel masks for each image . pairs of exposures were then combined using the pyraf astrodrizzle routine , with the aforementioned masks preceding the standard cosmic ray detection in the multidrizzle task . the final , cleaned images are presented in figure [ fig:3x2 ] . all optical and uv fluxes were corrected for galactic extinction following @xcite using a galactic reddening estimate of @xmath18 towards the cluster center , from @xcite . in figure [ fig:3x2 ] we show the newly - acquired far - uv through optical hst images of the core of the phoenix cluster , which provide a detailed picture of this system . these images show significant , extended filamentary emission at all wavelengths from @xmath610005500 , overlaid on a relatively smooth population of older , red stars . the most extended pair of filaments to the north of the bcg are @xmath66@xmath19 ( 40 kpc ) in length , similar to the most extended filaments seen in abell 1795 @xcite , and the perseus cluster @xcite . we measure a total rest - frame far - uv flux density of f@xmath20 = 1.26@xmath2110@xmath22 erg s@xmath4 @xmath23 hz@xmath4 , consistent with the galex - derived flux presented in @xcite . the fact that such complex , filamentary morphology is present in all five filters suggests that the bcg is forming stars at a prodigious rate . in the wavelength range covered , there may be contributing emission from the @xmath241549 ( f225w ) , [ ] ( f625w ) , and [ ] and h@xmath25 ( f814w ) lines . however , the f336w and f475w bands , which have similar surface brightnesses to the other three bands , should be relatively free from emission lines , suggesting that young stars , not ionized gas , is the dominant source of the observed flux in figure [ fig:3x2 ] . in figure [ fig : rgb ] we show a three - color ( f475w , f625w , f814w ) image of the cluster core . this figure shows a clear difference in the stellar populations between the young ( blue ) filaments and the underlying , smoothly - distributed , old ( red ) stars . the peak of the emission in all bands is coincident ( within the positional uncertainties ) with the x - ray point source . to the northwest and southeast of the emission peak are dark lanes , most likely due to obscuration by dust . overall , the color of the filamentary emission appears roughly constant with radius , and is reminiscent of a young , star - forming galaxy . we see no evidence for multiple bulges or tidal features , both of which would indicate that this system is the result of a recent merger of gas - rich galaxies . figure [ fig : sbprof ] shows the multi - band surface brightness profiles of the bcg ( left panel ) , which have been computed along radial cuts at four different angles ( 90@xmath26 , 120@xmath26 , 180@xmath26 , 210@xmath26 ) , chosen to avoid the blue , filamentary emission . the radial surface brightness profile follows an r@xmath1 profile , which is typical of relaxed , early - type galaxies @xcite . such r@xmath1 surface brightness distributions are also common in the final stages ( single - nucleus ) of low - redshift ( @xmath27 ) gas - rich mergers ( ulirgs ; e.g. , * ? ? * ; * ? ? ? however , with a half - light radius of @xmath617 kpc and a stellar mass of @xmath28 m@xmath3 @xcite , this bcg is a factor of @xmath64 larger in size @xcite , a factor of @xmath660 higher in stellar mass @xcite , and resides in an environment @xmath650100 times richer @xcite than normal for @xmath27 ulirgs . projecting these 1-dimensional profiles back onto the sky , we can separate diffuse , giant elliptical emission ( middle panel ) from clumpy , star - forming emission ( right panel ) . the lack of smooth , arcing tidal features and multiple bulges in the residual image ( right panel ) suggests that this complex , extended emission did not originate from a recent merger with one or more gas - rich disk galaxies . all of these factors argue that the phoenix bcg is unlike a traditional ulirg by any definition other than the high total infrared luminosity . in figure [ fig : sed ] we provide the spectral energy distribution ( sed ) in several representative regions around the bcg . the diffuse emission at large and small radii indicate a significant positive age gradient in the diffuse population . at large ( r @xmath640 kpc ) radii , the diffuse emission is qualitatively consistent with a 25 gyr old elliptical galaxy , while at smaller radii ( @xmath620 kpc ) the diffuse stellar populations appear to be much younger ( bluer ) , similar in color to a star - forming spiral galaxy . the extended , morphologically - complex filaments , after subtraction of the diffuse stellar component , show an excess of uv emission at all radii . the sed of the brightest filaments appear remarkably similar to the diffuse component in the central region , suggesting that these stars are being mixed on short timescales . in the faintest , most extended filaments , there is a substantial excess of emission in the f225w and f625w filters , at the location of the redshifted @xmath241549 and [ ] lines , respectively , suggesting that these filaments may also contain warm ( @xmath29k ) , ionized gas a scenario supported by the extended emission lines reported in @xcite . the overall flatness of the sed in the uv - bright regions is exactly what one would expect for a mix of young stars and warm , ionized gas , given the width of the broadband filters . at the peak of the optical emission , coincident with the x - ray source , the sed is qualitatively well matched by a dusty type-2 qso @xcite , which is consistent with our x - ray observations of a highly - reddened hard x - ray point source . the combination of figures [ fig:3x2][fig : sed ] paint a picture of an old , giant elliptical galaxy that is experiencing a resurgence of star formation . below we re - evaluate the sfr in this system and describe various scenarios to explain this star - forming activity , building on the discussion of @xcite . the deep , high spatial resolution hst uv and optical imaging presented in 3 have revealed an exceptionally complex system . below we utilize this improved spatial resolution to estimate a new , uv - derived sfr for the bcg in the phoenix cluster , followed by a discussion of three possible origins for the extended , filamentary uv emission . the sfr reported in @xcite , while utilizing an array of multi - wavelength data , necessarily required multiple assumptions to remove agn contamination . with the addition of high - resolution hst uv imaging , we avoid such assumptions , leaving only the ( typical ) assumptions of dust extinctions and star formation laws . @xmath21 0.2@xmath19 bins overlaid in regions with @xmath30 detections in both the f336w and f475w bands . right : smoothed reddening map , derived assuming a flat uv sed in the absence of reddening @xcite . this map allows us to perform _ local _ reddening corrections on the uv images , yielding a more tightly constrained estimate of the extinction - corrected uv luminosity.,scaledwidth=49.0% ] to estimate the reddening , we assume that the reddening - free sed is flat at 15003000 @xcite . since the f225w filter may be contaminated by emission , we use the mean @xmath31 ( rest frame @xmath32 ) flux ratio to derive a reddening correction , assuming @xmath33 for e(b - v ) = 0 . in figure [ fig : ebv ] we show the spatial distribution of the reddening , e(b - v ) , over regions with significant ( @xmath34 ) uv flux . we note that the mean reddening in this map agrees well with the reddening presented in @xcite from balmer line ratios ( e(b - v ) = 0.34 ) . using the _ local _ reddening correction and uncertainty we can accurately correct the uv luminosity , resulting in a more confident estimate of l@xmath35 = 5.7 @xmath2 0.3 @xmath21 10@xmath36 erg s@xmath4 hz@xmath4 and a uv - derived sfr of 798 @xmath2 42 m@xmath3 yr@xmath4 @xcite . given the near - ir - estimated stellar mass of m@xmath37 m@xmath3 @xcite , this high sfr implies that in 100 myr the central galaxy in the phoenix cluster will form 23% of its stellar mass . we note that this estimate is consistent with the agn - subtracted sfr quoted in @xcite of 739 @xmath2 160 m@xmath3 yr@xmath4 and with the empirical , extinction - implicit , method of @xcite , which yields sfr@xmath38 m@xmath3 yr@xmath4 . inspired by the high far - infrared luminosity , coupled with the hard x - ray and radio source , a viable explanation for this system is a dusty agn driving a large - scale outflow . while there is undoubtedly a powerful agn in the core of the phoenix cluster , the new hst data suggest that the substantial uv luminosity can not be fully attributed to a point - like agn . only a small fraction ( @xmath3910% ) of the total uv luminosity originates from a central point source , with the majority originating from extended , filamentary regions . furthermore , the high relative fluxes in the f336w and f475w bands , which should be free from line emission , suggest that the majority of the uv / optical flux in these complex filaments is continuum emission . we also find a general lack of uv / blue emission along the minor axis of the central galaxy , contrary to what is typically observed in wide - angle agn - driven outflows @xcite . thus , we argue that the filamentary uv emission does not result from an outflow . in iras 09104 + 4109 ( e.g. , * ? ? ? * ) , along with other powerful radio galaxies , much of the extended uv continuum is due to scattered , polarized light from a heavily - obscurred qso . however , the uv continuum in iras 09104 + 4109 is @xmath620 times fainter than that presented here for the phoenix cluster @xcite . considering the fact that iras 09104 + 4109 is already an extreme system , it seems unlikely that this additional factor of 20 can be accounted for with the same model . the majority of known ultraluminous infrared galaxies ( ulirgs ) appear to be late - stage mergers of gas - rich galaxies ( @xmath695% ; * ? ? ? * ) , which begs the question : is the extreme star formation in the core of the phoenix cluster fueled by a gas - rich merger ? assuming a relation between molecular gas depletion time and specific star formation rate @xcite , we estimate an h@xmath10 depletion time of 212@xmath2110@xmath40 yr , which is consistent with those measured in two nearby cooling flows : a1068 ( 4@xmath41 yr ; * ? ? ? * ) and a1835 ( 5@xmath41 yr ; * ? ? ? assuming a 1-to-1 correspondence between the total mass of stars formed and the mass of the cold gas reservoir , this timescale implies a molecular gas mass of 1.59.5@xmath42 m@xmath3 , consistent with more recent work by @xcite , yielding 1.05.0@xmath42 m@xmath3 . this cold gas mass is significantly higher than that measured for gas - rich galaxies in the virgo cluster ( e.g. , * ? ? ? further , the process of increasing the cold gas content by compressing atomic gas during a merger would be highly inefficient in the core of the phoenix cluster due to ram pressure stripping of the hi disk during the initial galaxy infall - a process observed in the virgo cluster , which has an icm density @xmath610 times lower than phoenix . our preferred explanation in @xcite was that the star formation in the cluster core is being fueled by gas cooling out of the icm . this remains the most plausible avenue for such a large amount of cold gas to reach the core of the cluster , and is supported by the exceptionally bright x - ray cool core and relatively weak radio source . such an imbalance between cooling and feedback could lead to rapid cooling of the icm , fueling bursts of star formation . following @xcite , we estimate from the x - ray data the mass deposition rate of the cooling flow , combining the x - ray cooling luminosity with the gravitational potential of the cluster in order to correct for gravitational work done as the gas falls towards the cluster center . we obtain an icm cooling rate of m = 2700 @xmath2 700 m@xmath3 yr@xmath4 which is enough to fuel a 798 m@xmath3 yr@xmath4 starburst , assuming the feedback mechanism that prevents star formation in nearby clusters is operating less efficiently in phoenix . in the future , high resolution x - ray spectroscopy of the cool core ( e.g. , * ? ? ? * ) will provide firm estimates of the icm cooling rate down to low temperatures , revealing whether or not this scenario is a plausible fuel source for this massive starburst . we report new hst observations of the phoenix cluster ( spt - clj2344 - 4243 ) with wfc3-uvis in five filters covering rest - frame wavelengths 1000 - 5500 . the high spatial resolution of hst is able to separate bright uv emission from the agn and the surrounding diffuse , extended emission , definitively confirming the presence of a starburst in the bcg . the morphology of this central galaxy is complex , with narrow filaments extending for @xmath040 kpc , reminiscent of the nearby perseus and abell 1795 clusters . we argue that the majority of the observed uv emission is due to young stars , on the basis of the complex morphology and flat sed over the wavelength range 10005500 . we confirm the high sfr presented in @xcite , measuring an extinction - corrected , agn - removed uv - derived sfr of 798 @xmath2 42 m@xmath43 yr@xmath4 . we find that merger driven scenarios would require an unreasonably large number of gas rich galaxies to supply the cold gas reservoir required to fuel the starburst , and conclude that the starburst is likely fueled by a massive cooling flow . we thank the hst director for graciously providing the data for this program . m. m. acknowledges support provided by nasa through a hubble fellowship grant from stsci . s.v . acknowledges support from a senior npp award held at nasa - gsfc and from the humboldt foundation to fund a long - term visit at mpe in 2012 . bab is supported by the national science foundation through grant ant-0638937 , with partial support provided by nsf grant phy-1125897 , the kavli foundation , and chandra award number 13800883 issued by the cxc . , e. , bianchi , s. , cortese , l. , giovanardi , c. , magrini , l. , pappalardo , c. , boselli , a. , bendo , g. j. , davies , j. , grossi , m. , madden , s. c. , smith , m. w. l. , vlahakis , c. , auld , r. , baes , m. , de looze , i. , fritz , j. , pohlen , m. , & verstappen , j. 2012 , , 542 , a32 , m. , bayliss , m. , benson , b. a. , foley , r. j. , ruel , j. , sullivan , p. , veilleux , s. , aird , k. a. , ashby , m. l. n. , bautz , m. , bazin , g. , bleem , l. e. , brodwin , m. , carlstrom , j. e. , chang , c. l. , cho , h. m. , clocchiatti , a. , crawford , t. m. , crites , a. t. , de haan , t. , desai , s. , dobbs , m. a. , dudley , j. p. , egami , e. , forman , w. r. , garmire , g. p. , george , e. m. , gladders , m. d. , gonzalez , a. h. , halverson , n. w. , harrington , n. l. , high , f. w. , holder , g. p. , holzapfel , w. l. , hoover , s. , hrubes , j. d. , jones , c. , joy , m. , keisler , r. , knox , l. , lee , a. t. , leitch , e. m. , liu , j. , lueker , m. , luong - van , d. , mantz , a. , marrone , d. p. , mcmahon , j. j. , mehl , j. , meyer , s. s. , miller , e. d. , mocanu , l. , mohr , j. j. , montroy , t. e. , murray , s. s. , natoli , t. , padin , s. , plagge , t. , pryke , c. , rawle , t. d. , reichardt , c. l. , rest , a. , rex , m. , ruhl , j. e. , saliwanchik , b. r. , saro , a. , sayre , j. t. , schaffer , k. k. , shaw , l. , shirokoff , e. , simcoe , r. , song , j. , spieler , h. g. , stalder , b. , staniszewski , z. , stark , a. a. , story , k. , stubbs , c. w. , uhada , r. , van engelen , a. , vanderlinde , k. , vieira , j. d. , vikhlinin , a. , williamson , r. , zahn , o. , & zenteno , a. 2012 , , 488 , 349 , c. p. , baum , s. a. , privon , g. , noel - storr , j. , quillen , a. c. , zufelt , n. , park , j. , edge , a. , russell , h. , fabian , a. c. , donahue , m. , sarazin , c. l. , mcnamara , b. , bregman , j. n. , & egami , e. 2008 , , 681 , 1035 , m. , tajer , m. , maraschi , l. , trinchieri , g. , lonsdale , c. j. , chiappetti , l. , andreon , s. , pierre , m. , le fvre , o. , zamorani , g. , maccagni , d. , garcet , o. , surdej , j. , franceschini , a. , alloin , d. , shupe , d. l. , surace , j. a. , fang , f. , rowan - robinson , m. , smith , h. e. , & tresse , l. 2007 , , 663 , 81 , t. d. , edge , a. c. , egami , e. , rex , m. , smith , g. p. , altieri , b. , fiedler , a. , haines , c. p. , pereira , m. j. , prez - gonzlez , p. g. , portouw , j. , valtchanov , i. , walth , g. , van der werf , p. p. , & zemcov , m. 2012 , , 747 , 29 , a. , kauffmann , g. , wang , j. , kramer , c. , tacconi , l. j. , buchbender , c. , catinella , b. , graci - carpio , j. , cortese , l. , fabello , s. , fu , j. , genzel , r. , giovanelli , r. , guo , q. , haynes , m. p. , heckman , t. m. , krumholz , m. r. , lemonias , j. , li , c. , moran , s. , rodriguez - fernandez , n. , schiminovich , d. , schuster , k. , & sievers , a. 2011 , , 415 , 61 , g. r. , odea , c. p. , baum , s. a. , clarke , t. e. , sarazin , c. l. , bregman , j. n. , combes , f. , donahue , m. , edge , a. c. , fabian , a. c. , ferland , g. j. , mcnamara , b. r. , mittal , r. , oonk , j. b. r. , quillen , a. c. , russell , h. r. , sanders , j. s. , salom , p. , voit , g. m. , wilman , r. j. , & wise , m. w. 2012 , , 424 , 1042 , r. , benson , b. a. , high , f. w. , vanderlinde , k. , ade , p. a. r. , aird , k. a. , andersson , k. , armstrong , r. , ashby , m. l. n. , bautz , m. , bazin , g. , bertin , e. , bleem , l. e. , bonamente , m. , brodwin , m. , carlstrom , j. e. , chang , c. l. , chapman , s. c. , clocchiatti , a. , crawford , t. m. , crites , a. t. , de haan , t. , desai , s. , dobbs , m. a. , dudley , j. p. , fazio , g. g. , foley , r. j. , forman , w. r. , garmire , g. , george , e. m. , gladders , m. d. , gonzalez , a. h. , halverson , n. w. , holder , g. p. , holzapfel , w. l. , hoover , s. , hrubes , j. d. , jones , c. , joy , m. , keisler , r. , knox , l. , lee , a. t. , leitch , e. m. , lueker , m. , luong - van , d. , marrone , d. p. , mcmahon , j. j. , mehl , j. , meyer , s. s. , mohr , j. j. , montroy , t. e. , murray , s. s. , padin , s. , plagge , t. , pryke , c. , reichardt , c. l. , rest , a. , ruel , j. , ruhl , j. e. , saliwanchik , b. r. , saro , a. , schaffer , k. k. , shaw , l. , shirokoff , e. , song , j. , spieler , h. g. , stalder , b. , stanford , s. a. , staniszewski , z. , stark , a. a. , story , k. , stubbs , c. w. , vieira , j. d. , vikhlinin , a. , & zenteno , a. 2011 , , 738 , 139	we present _ hubble space telescope wide field camera 3 _ observations of the core of the phoenix cluster ( spt - clj2344 - 4243 ) in five broadband filters spanning rest - frame 10005500 . these observations reveal complex , filamentary blue emission , extending for @xmath040 kpc from the brightest cluster galaxy . we observe an underlying , diffuse population of old stars , following an r@xmath1 distribution , confirming that this system is somewhat relaxed . the spectral energy distribution in the inner part of the galaxy , as well as along the extended filaments , is a smooth continuum and is consistent with that of a star - forming galaxy , suggesting that the extended , filamentary emission is not due to the central agn , either from a large - scale ionized outflow or scattered polarized uv emission , but rather a massive population of young stars . we estimate an extinction - corrected star formation rate of 798 @xmath2 42 m@xmath3 yr@xmath4 , consistent with our earlier work based on low spatial resolution ultraviolet , optical , and infrared imaging . the lack of tidal features and multiple bulges , combine with the need for an exceptionally massive ( @xmath5 m@xmath3 ) cold gas reservoir , suggest that this star formation is not the result of a merger of gas - rich galaxies . instead , we propose that the high x - ray cooling rate of @xmath62700 m@xmath3 yr@xmath4 is the origin of the cold gas reservoir . the combination of such a high cooling rate and the relatively weak radio source in the cluster core suggests that feedback has been unable to halt cooling in this system , leading to this tremendous burst of star formation .
in quantum information applications based on optically addressed atoms or ions it is often necessary to prepare these elements in specific hyperfine or zeeman spin levels . typically this involves optical pumping from one ground - state spin level to another via an excited state . the preparation of the ions into a single well - defined state ( spin polarization ) is a required initial step for quantum memory protocols based on electromagnetically induced transparency ( eit ) @xcite , raman interactions @xcite and photon echoes using controlled reversible inhomogeneous broadening ( crib ) @xcite or atomic frequency combs ( afc ) @xcite . in particular , the implementation of these protocols in rare earth ( re ) ion doped solids @xcite requires spectral tailoring of the inhomogeneous absorption in order to isolate narrow absorption peaks . among the various physical systems that have been considered for photonic quantum state storage , erbium doped solids provide a unique system where a large number of stationary atoms can coherently absorb photons at the telecommunication wavelength of 1.53 @xmath4 m . quantum memories at telecommunication wavelengths are required for a range of efficient quantum repeater protocols @xcite . in addition , erbium doped solids have exceptional optical coherence properties . an optical coherence time as long as 6.4 ms has been measured in er@xmath5:y@xmath0sio@xmath1 @xcite , which represents the longest optical coherence time measured in a solid . another interesting property of er@xmath5 is that it is a kramers ion with an odd number of electrons . this results in a large splitting between the ground state levels via a first order zeeman interaction , which leads to a larger accessible frequency bandwidth for quantum memory applications . however , the unquenched electronic spin of kramers ions results in strong spin - spin and spin - phonon interactions as compared to non kramers ions such as pr and eu . therefore the ground state population relaxation times are much shorter than in non kramers ions , usually in the range of tens to hundreds of ms @xcite . in order to achieve a high degree of population transfer via optical pumping it is necessary for the ground state lifetime to be much longer than the excited state lifetime . achieving efficient population transfer in erbium doped materials seems thus particularly challenging . the spectroscopic properties of er@xmath5:y@xmath0sio@xmath1 have been extensively studied , including optical coherence @xcite , spectral diffusion @xcite , hyperfine structure @xcite , zeeman relaxation lifetimes @xcite , zeeman g factors @xcite and erbium - host interactions @xcite . slow light has also been achieved in this material using coherent population oscillation @xcite . however , to our knowledge no study has been reported on the possibility to implement efficient population transfer between the two zeeman ground states to achieve a high degree of spin polarization . in this article we investigate optical pumping between ground state zeeman levels of erbium ions doped into a y@xmath6sio@xmath7 crystal . we first observe the limitation of standard optical pumping . for er@xmath5y@xmath0sio@xmath1 an optical relaxation time of 11 ms @xcite and a zeeman relaxation lifetime of about 130 ms ( at a magnetic field of 1.2 mt ) @xcite have been measured . the low ratio between these two relaxation lifetimes strongly limits the achievable population transfer efficiency . another limiting factor is the branching ratio between the two optical transitions connecting the two ground state zeeman levels . we then show how an enhancement of the optical pumping efficiency can be achieved by decreasing the excited state lifetime via optical stimulated emission and by improving spin branching ratios via radio frequency ( rf ) excitation . these techniques allow population transfer between the two zeeman states with less than 10@xmath2 of the total population remaining in the initial state , i.e. more than 90@xmath2 spin polarization . we also demonstrate spectral tailoring in this crystal by preparing a narrow absorption line inside wide transparency window ( spectral pit ) , as required for the crib quantum memory scheme . population transfer between the two closely spaced ground state levels of a @xmath8-system can be achieved by optical pumping via the excited state level . in an optical pumping experiment , atoms are excited by a laser in resonance with the transition connecting one of the ground state to the excited state . the excited atoms can then decay into both ground states . those which have decayed to the ground state which is not connected to the laser will in principle remain there for a time corresponding to the relaxation time between the ground states . if enough pumping cycles can be done within this time , the population from the initial ground state can be entirely transferred to the second ground state . the efficiency of this transfer is thus limited by the ratio of the excited state lifetime ( @xmath9 ) and the spin population lifetime of the ground state levels ( which we label @xmath10 , since in our case the ground state levels are zeeman levels ) as well as the branching factor @xmath11 . the latter is defined as the probability of the ion to relax into its initial state via spontaneous emission . transition ( solid line ) . the probe transitions ( dashed lines ) are labelled and the positions of the corresponding holes and antiholes in the transmission spectrum are shown below . the zeeman splitting of the ground and the excited state level , respectively , is given by @xmath12 , where @xmath13 and @xmath14 are the g factors for the ground and the excited state , respectively , @xmath15 is the bohr magneton and @xmath16 the magnetic field . note that in addition to those shown in the figure , further side- and antiholes occur that are due to the inhomogeneous broadening of the absorption line @xcite . b ) to enhance the efficiency of the population transfer the excited state lifetime is artificially lowered by stimulated emission to the short - lived second kramers doublet of the ground state ( z@xmath0 ) . ] in materials with inhomogeneous broadening , population transfer by optical pumping is usually realized and investigated with spectral holeburning ( shb ) techniques . in this scheme , the pump laser of frequency @xmath17 ( in resonance with transition 1 , fig . [ levelscheme]a ) is focused on the sample for a duration @xmath18 , the burning time . ions that are in resonance with @xmath17 are transferred to the excited state . in order to probe the transmission , a weak pulse follows after a delay @xmath19 during which the frequency of the laser is scanned around @xmath17 ( see fig . [ setup ] b ) . the inhomogeneously broadened absorption shows a spectral hole at @xmath17 which in general decays with the excited state lifetime @xmath9 . some of the ions will however relax into the second ground state zeeman level . if the temperature is low enough ( @xmath20 k ) such that zeeman levels are thermally decoupled @xcite , the population of this level will then be increased and the absorption at @xmath21 ( transition 4 in fig . [ levelscheme ] a ) will be enhanced . this enhancement is called a spectral anti - hole . it will decay with the zeeman lifetime @xmath10 . as long as the population in the two zeeman levels has not relaxed to thermal equilibrium , a part of the spectral hole at @xmath22 will persist . in case of a population transfer the hole should thus show two different decay times : @xmath9 from ions that decay from the excited back to the initial state and @xmath10 from ions that relax into the second zeeman level . the occurrence of the anti - hole and the long decay time ( @xmath10 ) for the central hole apparently provide evidence of population transfer . throughout this article we speak of this method as the standard optical pumping scheme . the intensity of the transmitted light is given by @xmath23 , where @xmath24 is the intensity of the incident radiation , and @xmath25 the length of the sample . the absorption coefficient @xmath26 , i.e. the logarithm of the transmitted intensity , gives rise to the relation between the number of ions @xmath27 in the ground and the excited state . @xmath28 is the cross - section of the transition and can be considered as a constant . since in this work rather the population than the measured intensities themselves are of interest , all curves that deal with quantities related to the absorption and transmission , respectively , ( such as the area or the depth of a spectral hole ) are given with respect to the natural logarithm of the measured intensity . to qualitatively understand the limiting factors in the population transfer process , one can use a simple rate equation model for a three level @xmath8 system . for the steady state in the case of standard optical pumping ( see appendix ) it is easy to derive that the ratio of populations in the ground states after optical pumping depends on the ratio between the zeeman lifetime @xmath10 and the excited state lifetime @xmath9 : @xmath29 where @xmath30 ( @xmath31 ) is the population fraction in the initial ( final ) zeeman state . however , this holds only for the case in which all of the excited ions relax into the desired level . in re - doped solids the selection rule for the electronic spin @xmath32 normally is only slightly lifted by the crystal field and only a small part of the ions can be found in this decay channel . this motivates to introduce an effective lifetime in the following manner : @xmath33 the branching factor @xmath34 being the probability that an ion will preserve its electron spin upon relaxation from the excited to the ground state . if @xmath11 is low , the pump rate has to be accordingly high to transfer population into the excited state . on the other hand a low branching factor is desirable if one wants the ions to relax into the second zeeman level . the ratio between ions in either of the two ground state levels is then given by @xmath35 ( see appendix ) . in previous measurements on er@xmath5:y@xmath0sio@xmath1 an excited state lifetime of @xmath36ms and a branching factor @xmath370.9 were found @xcite . this leads to an effective lifetime of @xmath38 ms which is of the same order as @xmath39 130 ms @xcite . according to the model it is thus impossible to perform efficient population transfer between the zeeman ground states of er@xmath5:y@xmath0sio@xmath1 using standard optical pumping . however , by looking at eq.[lifetimeratio ] and eq . [ efflifetimeratio ] , one can devise methods to increase the efficiency of the population transfer . a first idea is to improve the lifetime ratio by artificially increasing the decay rate from the excited state . this can be achieved by using stimulated emission to another short lived ground state level ( see fig.[levelscheme]b ) as proposed in @xcite . the corresponding results are presented in section [ stimulated ] . another possibility is to improve the branching ratio by mixing the spin level in the excited state using a radio frequency ( rf ) excitation . this technique is presented in section [ spin ] . the crystal , y@xmath0sio@xmath1 doped with er@xmath5 ( 10ppm ) , belongs to the crystal space group c@xmath40 . the er@xmath5 ions replace y@xmath5 ions and occupy two crystallographic inequivalent sites of c@xmath41 symmetry ( site 1 and site 2 ) @xcite . all measurements in this work were performed on ions at site 1 . the relevant transition is from the @xmath42i@xmath43 to the @xmath42i@xmath44 level , which are split into eight ( z@xmath45 , @xmath46 , @xmath47 labelling the lowest crystal field level ) and seven ( y@xmath45 , @xmath48 ) kramers doublets , respectively . all measurements were carried out on the z@xmath49 y@xmath50 transition having a wavelength of 1536 nm . under a magnetic field each of the crystal field levels splits into two zeeman levels ( @xmath51 . the y@xmath0sio@xmath1 crystal has three mutually perpendicular optical - extinction axis labelled @xmath52 , @xmath53 , and @xmath54 . the direction of light propagation is chosen parallel to the @xmath54-axis . the crystal was cut along these axes and its dimensions were 3 mm @xmath55 3.5 mm @xmath55 4 mm along @xmath54 , @xmath52 , and @xmath53 , respectively . the magnetic field was applied in the @xmath52-@xmath53 plane . in this case all ions of each site are magnetically equivalent @xcite . the angle of the magnetic field is defined with respect to the @xmath52-axis ( see fig . [ setup ] c ) . if not mentioned explicitly all measurements were taken at an angle of @xmath56 . nm ) and the stimulation laser ( @xmath57 nm ) could be intensity modulated independently with acousto - optic modulators ( aom ) . they where coupled into the same fiber via a wave division multiplexer ( wdm ) . the light of the stimulation laser was filtered out before detection using another wdm . an rf - wave could be applied on the crystal using a coil in helmholtz configuration which could be driven with an arbitrary function generator ( afg ) . b ) typical pulse sequence . during the readout pulse the frequency of the pump laser was scanned in order to record the absorption profile . for measurement with stimulated emission , the stimulation laser is applied for a time @xmath58 , where @xmath59 can be varied . c ) illustration of the angle of the magnetic field with respect to the crystal orientation ] . the experimental setup is shown in figure [ setup]a . the crystal was placed on the cold finger of a liquid helium cryostat ( janis st400 ) and could be cooled down to a temperature of about 2 k. all measurements , if not mentioned differently , were taken at 2.1 k. for optical pumping we used an external cavity diode laser ( toptica , @xmath60 nm ) which was operated in free run mode and its jitter within the relevant timerange was @xmath61 1 mhz . for stimulated emission experiments a tunable diode laser ( nettest , @xmath57 nm ) was at our disposal . it was amplified with an erbium doped fiber amplifier ( edfa ) and could be tuned over a wide range . both lasers could be gated independently using acousto - optical modulators ( aom ) and arbritrary function generators ( afg ) . they were coupled into the same optical fiber via a wavelength division multiplexer ( wdm ) . the light was focused into the crystal ( diameter of focus ( @xmath62 ) @xmath63 m ) . in order to avoid a large background in the detection due to fluorescence from the edfa the stimulation laser was filtered out after passing the cryostat using a second wdm . the transmitted light was measured with a photodiode ( newfocus 2011 ) . furthermore a helmholtz coil placed around the crystal was at our disposal and could be used to apply rf - pulses to the sample . the first series of measurements was made with standard optical pumping ( c.f . section [ theory ] ) in order to estimate the original efficiency of the population transfer . a spectral hole was created by sending a resonant pump pulse of 200 ms into the crystal . the dynamics of the hole was then measured by probing it at different delays . the circles in figure [ nostimnorf ] show the decay of the spectral hole using standard optical pumping . the dashed line is a fit to the data given by the sum of two exponential decays with rates of @xmath64 and @xmath65 , respectively . of 11 ms . the line is a fit to the data . only a small fraction of the ions was transferred into the second zeeman level and provides a slow component ( @xmath39100 ms ) to the decline . the origin of the slowly decaying part is confirmed by the slight enhancement in the optical depth ( anti - holes ) left and right of the wide spectral hole ( pit ) in the inset . the graph shows a zoom on the absorption line at about @xmath66 ms after the burning pulse . here the frequency of the pump laser was sweeped by 10 mhz during the burning pulse in order to widen the hole . ] the decay is clearly dominated by the excited state lifetime @xmath67=11ms . the contribution of the decay with @xmath10 is low which leads to the conclusion that only a very small fraction of the ions could be transferred to the second zeeman level of z@xmath41 . this assumption is confirmed by the hole burning spectrum shown in the inset of fig . [ nostimnorf ] . for this measurement , a 10 mhz frequency sweep was applied to the pump laser , in order to widen the spectral hole . the enhancement of absorption at @xmath68 mhz ( fig . [ levelscheme ] ) is low , in contrast to what one would expect in the case of an efficient transfer of population . these observations show that the created spectral hole is mostly due to a storage of population in the excited state . thus , as expected from the arguments given in section [ theory ] , optical pumping alone does not suffice to prepare the system into the required state . a first possible solution to improve the optical pumping efficiency is to artificially increase the decay rate from the excited state by stimulating the emission to another short lived ground state level ( see fig . [ levelscheme]b ) @xcite . for this purpose , a second laser in resonance with the transition y@xmath69 z@xmath0 ( 1546 nm , @xcite ) is applied simultaneously with the pump laser . due to a strong coupling to phonon modes , mostly due to direct phonon emission , the excited crystal field levels ( z@xmath0 , z@xmath70 , ... ) have non - radiative decay times in the range of nano- or picoseconds . thus , population from these levels will immediately relax into ground state ( z@xmath41 ) . we first characterize the stimulation process by measuring its dependence on the frequency of the stimulation laser . since the homogeneous broadening of the transition is expected to be larger than the inhomogeneous broadening @xcite , due to the short non - radiative decay times , this gives a direct measure of the homogeneous linewidth . it has been done by measuring the size of the spectral hole as a function of the frequency of the stimulation laser at 4.1 k. at this temperature the ground state zeeman levels are thermally coupled @xcite and therefore only ions in the excited state ( i.e. ions that have not been stimulated down ) contribute to the hole . the results are shown in fig . [ stimfreq ] . for this measurement , the stimulation pulse was 1 ms longer than the pump pulse . we see that when the stimulation laser is in resonance , the spectral hole almost completely vanishes , which shows that one can efficiently empty the excited state using this method . we find the linewidth to be about 14 ghz , which corresponds to a lifetime in the picosecond range . this is 9 orders of magnitude shorter than the zeeman lifetime and confirms that the z@xmath0-level is suitable for the application described above . k. the burning time was 200 ms . the stimulation pulse was 1 ms longer than the burning pulse . the delay for the readout pulse was set to 2.5 ms . the solid line is a lorentzian fit giving a full width at half maximum ( fwhm ) of 14 ghz . ] another quantity of interest is the rate at which stimulation laser moves the ions down from the excited to the ground state ( z@xmath0 ) . to measure this quantity a pump pulse with @xmath71 100ms was sent into the sample . the crystal temperature was chosen to be 4.3 k so that also in this case the size of the hole would only be given by the fraction of ions that remained in the excited state . simultaneously to the pump pulse a stimulation pulse with a duration @xmath58 was applied ( see figure [ setup]b ) . the size of the remaining hole as a function of @xmath59 was recorded and the data was fitted with an exponential decay of which the stimulation rate could be extracted . the inset of fig . [ stimulation ] shows this data for three different values of applied stimulation power . each of the lines is a fit to the respective data . note that there is an offset which has its origin in long - lived holes that occur even at this temperature for a small subset of ions @xcite . it was set to be the same for all curves . the stimulation rate as a function of the stimulation power is plotted in fig . [ stimulation ] . as one would expect the stimulation rate increases linearly with power . for this measurement , the maximal power of the stimulation laser on the sample was 20 mw and it was limited by the maximal output power of the edfa . the measurements were taken with a magnetic field of 27,4 mt at an angle of @xmath72 with the @xmath52-axis . the average beam diameter ( @xmath62 ) in the crystal was 83 @xmath73 . z@xmath0 transition vs. applied laser power . the rate at which ions are stimulated down grows linearly with the power of the stimulation laser . it was extracted from the decay curves shown in the inset and as described in the text . curves for three different values of the power of the stimulation laser are shown . each data set is normalized to its maximum value ( @xmath74 ) . the offset of 0.1 is due to persistent holes which have already been observed and discussed in @xcite.,width=340 ] let us now investigate the effect of stimulated emission on the population transfer by optical pumping . in fig . [ application ] one can see that the application of the stimulation laser simultaneously with the pump laser indeed leads to a significant enhancement of the transfer efficiency . the decay of the spectral hole at a temperature of 2.1 k is shown for the case of standard optical pumping ( open circles ) and optical pumping assisted by stimulated emission ( plain triangles ) . for standard optical pumping , the decay is dominated by the time constant @xmath9=11ms as explained in section [ standard ] . in contrast , the decay for the stimulated optical pumping is dominated by a slower decay time , which corresponds to @xmath75 ms . this indicates that a significant part of the population is now transferred to the other zeeman level by the optical pumping process . ms and the overall decay shows a much stronger contribution of this slow component than it can be observed for standard optical pumping . the enhancement in population transfer is confirmed by the strong anti - holes shown in the inset . the stimulation power the figure shows the absorption at about @xmath76 ms after the burning pulse . for this measurement , a 10 mhz sweep was applied on the pump laser and the stimulation pulse was 1 ms longer than the pump pulse and the stimulation power in front of the cryostat was 50 mw . ] this hypothesis is confirmed by directly measuring the residual absorption after the optical pumping , as shown in the inset of fig . [ application ] . in order to obtain a reference for the transmission out of resonance , the entire inhomogeneously broadened absorption line was measured ( only a zoom is shown ) . a spectral pit was created using stimulated optical pumping by sweeping the frequency of the pump laser over a range of 10 mhz while burning . for this measurement , the stimulation pulse was 1 ms longer than the pump pulse in order to de - excite population remaining in the excited state to the ground state . therefore the transparency in the central pit is mostly due to ions transferred to the other ground state . the optical depth without optical pumping is @xmath771 , while the residual optical depth after population transfer is @xmath78 0.25 . the enhanced absorption which occurs in form of wide anti - holes on both sides of the pit confirms , as explained in section [ theory ] , that the population was transferred as intended . from this measurement , it is possible to estimate the efficiency of the population transfer . the fraction of ions remaining in the initial zeeman state @xmath79 can be estimated by : @xmath80 0.25 . since initially the two zeeman states are equally populated , this means that 12.5 @xmath2 of the total number of atoms is remaining in the initial state , within the frequency range optically pumped . if one assumes that all atoms that are not in the initial state have been transferred to the other zeeman level , on can compute the ratio @xmath81=7 . this is much lower than what can be expected from the simple model presented in section [ theory ] ( see also appendix ) . this shows that , while the model presented helps to understand the process of optical pumping , it is too simplified to give quantitative estimations . although stimulated optical pumping greatly increases the population transfer efficiency , the fraction of ions remaining in the initial ground state is still high . this is in part due to the high value of the branching factor @xmath11 for this transition . in ref . @xcite , @xmath11 was estimated to be larger than 0.9 for a magnetic field applied at @xmath82 ( see fig . [ setup ] c ) , from comparison between experimental data and numerical simulations . this means that the spin is conserved to a high degree during the decay from the excited state . from the measurement of the last section , this seems to hold even for decay through the the state z@xmath6 , which involves phononic relaxation . in this section , we present a technique to improve the branching ratio for optical pumping applications . an idea to overcome this problem in general is to artificially change the spin by mixing the population of the two zeeman levels of the excited state using rf excitation . for this purpose a pair of helmholtz coils mm , length of each coil @xmath83 3 mm , distance between coils @xmath83 3 mm , 3 turns on each side . the central axis was parallel to the axis of light propagation . ] was placed around the crystal ( see fig . [ setup]a ) . note that a difference of the g - factors for the ground- and the excited state is necessary in order not to mix the spin population of the ground state levels as well . in er@xmath5:y@xmath0sio@xmath1 @xmath84 and @xmath85 ( for @xmath86 at an angle of @xmath56 , [ setup ] c ) @xcite . for electronic spins with such large g - factors , the magnetic dipole moment of the spin transition is large ( @xmath4=112 mhz / mt ) , so that significant mixing can be obtained with moderate rf - power . the rf pulse was applied simultaneously with the pump pulse ( duration 200 ms ) . a 1 @xmath4s long linear frequency sweep was continuously repeated during the burning process . due to the inhomogeneity of the external magnetic field the bandwidth , i.e. the size of the sweep , had to be chosen high ( 10 - 20 mhz ) . the required central frequency corresponds to @xmath87 in figure [ levelscheme ] and could be extracted from a hole - burning spectrum . the effect of the rf mixing was measured by estimating the fraction of atoms remaining in the initial state after optical pumping , similarly to the measurement made in section [ stimulated ] . for this measurement however , both stimulated emission and rf excitation were applied . [ rf - power ] shows the fraction of the total number of atoms remaining in the initial state as a function of the rf voltage . a significant enhancement of the transfer due to the rf - wave can be observed . this enhancement , however , is quickly saturated when increasing the rf - power . this might be explained by the achievement of an equilibrium between the population of the two zeeman levels . the inset of fig.[rf - power ] shows a zoom of the spectral pit created with and without rf mixing . for this measurement , the number of atoms in the initial state was 8 @xmath2 , as compared to 12.5 @xmath2 without rf mixing applied ( sec . [ stimulated ] ) . hence , a spin polarization of 92 @xmath2 within the optically pumped bandwith has been achieved . : y@xmath0sio@xmath1 ( black curve ) : a narrow absorption line ( fwhm 2 mhz ) could be established inside a spectral pit ( fwhm 50 mhz ) . an interesting feature is that the structure written into the absorption can as well be observed in the corresponding anti - holes , which again confirms the population transfer into the second zeeman level of the ground state . the gray curve corresponds to the inhomogeneous absorption without the state preparation sequence . for further details see text . ] the ability to tailor the absorption profile is important for quantum memory proposals such as crib or afc , as well as proposals for quantum logic gates in re doped solids @xcite . to illustrate this capacity in er@xmath5:y@xmath0sio@xmath1 , we isolated a narrow absorption line within a large transparency window ( spectral pit ) , with the help of the two methods presented above . for this purpose , we used a simple technique consisting of sweeping the laser frequency during the 200 ms long pump pulse and gating off the pump laser for a short time at the center of the frequency sweep . the total frequency range was swept in 100 @xmath4s , resulting in 2000 frequency sweeps . an example of spectral tailoring is shown in fig . the width of the pit was 50 mhz while the absorption width of the peak at the center of the pit is 2 mhz . this is much broader than the homogeneous linewidth of the transition @xcite . the width of the peak is likely to be limited by the linewidth of our free - running laser , which is larger than 1 mhz for the timescale of the burning pulse . the fact that we sweep the laser frequency by scanning the laser current might also degrade the linewidth . stabilizing the laser frequency will be required in order to obtain narrower absorption peaks . furthermore , the use of more sophisticated hole burning techniques will also be needed @xcite in order to achieve absorption width in the khz range . although the methods developed in this paper allow a significant improvement of population transfer efficiency as compared to standard optical pumping , there is still a residual absorption in the spectral pit . this will act as a passive loss for light storage protocols . further efforts will thus be needed in order to optimize light storage efficiency . note however that the width of the spectral pit ( 50 mhz ) is much larger than what can be achieved in non - kramers ions that have been used for such purposes so far , such as praesodymium @xcite or europium @xcite . this shows the potential of er doped solids for obtaining large frequency bandwidths for quantum storage applications . erbium doped solids are promising candidates for the implementation of quantum memories at telecommunication wavelengths . we have shown and discussed the difficulties for state preparation by optical pumping in er@xmath5:y@xmath0sio@xmath1 , including the ratio between the excited state and the zeeman lifetimes and the branching factor @xmath11 . with stimulated emission assisted optical pumping and spin - mixing using rf excitations , we have investigated two methods to alleviate these problems . we have shown that their application indeed leads to a significant enhancement of the optical pumping efficiency . population transfer between two zeeman levels with less than 10@xmath2 of the total population in the initial level has been demonstrated . this corresponds to a spin polarization of 90@xmath2 . note that the techniques used here can also be applied to other rare earth doped solids . in particular , the might be useful to enhance optical pumping efficiency in other kramers ions , such as neodymium @xcite . further improvements might be obtained by increasing stimulation rates with higher laser powers and using coherent rf population transfer between the two excited states . another possibility to improve the branching ratio could be to apply the magnetic field at particular angles where the spin conservation selection rule is relaxed @xcite . in order to increase the population lifetime of zeeman levels , it might be interesting to work at higher magnetic field ( hundreds of mt ) , in order to reduce spin - spin interactions . other crystals might also be explored , e.g. y@xmath6o@xmath88 , to search for longer zeeman lifetimes . finally , it would also be interesting to investigate hyperfine states , which might have longer spin population relaxation lifetimes @xcite . we also demonstrated spectral tailoring in er@xmath5:y@xmath0sio@xmath1 by isolating a narrow absorption peak within a wide transparency window , as required for the crib quantum memory scheme . although further progress remains to be done to improve the population transfer efficiency with higher optical depth , the achievement of spectral tailoring in erbium doped solids will allow proof of principle experiments of light storage with quantum memory protocols . hence , with this work a further step towards a solid state based quantum memory at telecom wavelength has been accomplished . we would like to thank marko lovric and roman kolesov for their advice concerning the design of the rf - setup . furthermore would we like to thank the group of prof . d. van der marel , mehdi brandt , d. van mechelen , and violeta guritanu for experimental assistance , as well as prof . m. chergui ( epfl ) for loaning us the cryostat . tecnical support by c. barreiro and j.d . gauthier is acknowledged . this work was supported by the swiss nccr quantum photonics and the european commission under the integrated project qubit applications ( qap ) funded by the ist directorate . the low efficiency of the population transfer was ascribed to the the ratio between the excited state lifetime @xmath9 and the ground state zeeman lifetime @xmath10 as well as the branching factor @xmath11 . the corresponding formulas ( [ lifetimeratio ] and [ efflifetimeratio ] , section [ theory ] ) were found using the following simple rate equation model : let @xmath89 , @xmath90 and @xmath91 be the population fraction in the two ground and in the excited state , respectively , as shown in figure [ popratio ] . the laser is in resonance with the transition between level 1 and 3 . for the population of the second zeeman ground state level the population dynamics is given by @xmath92 @xmath93 is the einstein - coefficient for spontaneous emission @xmath94 and @xmath95 are the spin - lattice - relaxation rates . in the case considered by gorju et al . @xcite a term which corresponds to @xmath94 does not occur since the target state is a crystal field level and the relevant phonon decay is a one way process . however , the calculations are still quite similar . in the steady state it is @xmath96 where the second equality holds for the case of sufficiently high pump power , i.e. a saturation of the considered transition . if we further assume @xmath97 , which is a good approximation since the thermal energy @xmath98 is much larger than the splitting between the two levels , it follows from equation [ dynamics ] that the ratio between the number of ions in either of the lower states is given by @xmath99 the factor 2 in the numerator of the lifetime ratio occurs since @xmath100 @xcite . so far it is assumed that all ions decay into the second zeeman level . in this case @xmath101 . however , one has to take into account that only a fraction of the ions in the excited state will do so . supposing that the excited state lifetime is given by @xmath102 with @xmath103 and with the definition of the branching factor @xmath104 ( which gives @xmath105 ) we can define the effective lifetime @xmath106 plugging this into equation [ ratio1app ] we get @xmath107 in the case of stimulated emission as presented in section [ stimulated ] the spontaneous emission rate @xmath108 has to be replaced with the stimulation rate @xmath109 . due to its extremely low lifetime , the role of the level z@xmath0 ( figure [ levelscheme ] ) can be neglected . according to equation [ ratio2app ] population ratio is now given by @xmath110 if one considers a branching factor @xmath111 and a stimulation rate of @xmath112khz , with a zeeman lifetime of 100 ms we compute a ratio of 71 , which is significantly higher ( @xmath113 one order of magnitude ) than found in the experiment . this can be explained by the fact that the model is extremely simplified as compared to reality . among others the role of other decay channels and the possibility of a different branching factor for a decay from the z@xmath0 level could not been taken into account . nevertheless , this model can help to understand the processes of optical pumping in the system qualitatively .	erbium doped solids are potential candidates for the realization of a quantum memory for photons at telecommunication wavelengths . the implementation of quantum memory proposals in rare earth ion doped solids require spectral tailoring of the inhomogeneous absorption profile by efficient population transfer between ground state levels ( spin polarization ) using optical pumping . in this article we investigate the limiting factors of efficient optical pumping between ground state zeeman levels in an erbium doped y@xmath0sio@xmath1 crystal . we introduce two methods to overcome these limiting factors : stimulated emission using a second laser and spin mixing using radio frequency excitation . both methods significantly improve the degree of spin polarization . population transfer between two zeeman levels with less than 10 @xmath2 of the total population in the initial ground state is achieved , corresponding to a spin polarization greater than @xmath3 . in addition , we demonstrate spectral tailoring by isolating a narrow absorption peak within a large transparency window .
in physical chemistry , thermodynamics and condensed matter physics , a critical point specifies the conditions ( temperature , pressure and concentration ) at which distinct phases do not exist @xcite . there are multiple types of critical points such as vapor - liquid or liquid - liquid critical point . a single fluid has a unique critical point associated with given temperature , pressure and density . for binary mixtures of fluids , in the space of temperature , pressure , concentration , critical points are represented by a curve in a convenient domain @xcite ; to each temperature we can associate a critical pressure and two critical densities corresponding to the mixture components @xcite . an important thermodynamical potential is related to the mixture volume free energy @xcite . at a given temperature , the volume free energy is associated with the _ spinodal _ curve connecting the two different phases of the binary mixture . due to conditions of equilibrium of phases , it is possible to form a general expansion of the free energy near a critical point . this form is known in the literature by means of physical chemistry considerations @xcite and is the form we use in our calculations . by calculations in molecular theories , the densities of the components fluctuate near a critical point @xcite . in the following , we use a continuous model to investigate how the average variations of densities are related to molecular interactions . two assumptions are explicit bongiorno , gouin7,rowlinson2:@xmath0 the component densities are assumed to be smooth functions of the distance from an interface layer which is assumed to be flat on the scale of molecular sizes . the correlation lengths are assumed to be greater than intermolecular distances @xcite ; this is the case when at a given temperature @xmath1 the parameters are close to the ones of a critical state @xcite.@xmath2 the binary mixture is considered in the framework of a mean - field theory . this means , in particular , that the free energy of the mixture is a classical so - called `` gradient square functional '' . this kind of landau - ginzburg model consisting of a quadratic form of the density gradients comes from maxwell and van der walls original ideas korteweg , maxwell , vdw , widom . at given critical conditions , the coefficients of the quadratic form are constant . this point of view that , in non - homogeneous regions , the mixture may be treated as bulk phase with a local free - energy density and an additional contribution arising from the non - uniformity which may be approximated by a gradient expansion truncated at the second order is most likely to be successful and perhaps even quantitatively accurate near a critical point @xcite . the approximation of mean field theory does provide a good understanding and allows one to explicitly calculate the magnitude of the coefficients of the model . these non - linear equations are able to represent interface layer and bulks and consequently allow to build a complete theory of the mixtures in non - homogeneous domains in dynamics . in section 2 we recall the equations of motion in a pure mechanical process obtained through the hamilton variational principle . section 3 is devoted to travelling waves without dissipation . due to the fact that the equations are galilean invariant , the case of equilibrium and the case of motion are analyzed together . in section 4 by means of a rescaling process taking the vicinity of a critical point into account , we integrate the equation for equilibrium as well as for motions with dissipation . two appendices present the motion equations and the mathematical reason of the choice of the free energy form near a critical point of a binary mixture of fluids obtained by a new method issued from differential geometry . we study a mixture of two fluids by a mechanical process . no assumption has to be done about composition or miscibility . the motion of a two - fluid continuum can be represented with two surjective differentiable mappings ( see fig . 1 ) @xcite : @xmath3where subscripts 1 and 2 are associated with each constituent of the mixture . term @xmath4 denotes euler variables in space - time @xmath5 and terms @xmath6 and @xmath7 denote the lagrange variables of constituents in _ reference spaces _ @xmath8 and @xmath9 respectively . in the pure mechanical case , the lagrangian density of the mixture is @xmath10where @xmath11 and @xmath12 denote the velocity vectors of each constituent , @xmath13 and @xmath14 are the densities , @xmath15 and @xmath16 are the external force potentials depending only on @xmath17 and @xmath18 is the volume energy berdichevsky , galdi.the expression of the lagrangian is in a general form . in fact dissipative phenomena imply that @xmath11 is almost equal to @xmath12 ; it is the reason why we do not take account of some kinetic energy associated with the relative velocity of the components which is of smaller order ( at least of order 2 ) and will be negligible in travelling wave behavior @xcite . because of the interaction between the constituents , the volume energy @xmath18 is not the sum of the energies of each constituent of the mixture , like for _ euler mixtures of fluids_. the mixture is assumed not to be chemically reacting . conservations of masses require @xmath19where subscript @xmath20 belongs to @xmath211,2@xmath22 . at @xmath23 fixed , the deformation gradient @xmath24 associated with @xmath25 is denoted by @xmath26 and @xmath27 is the reference specific mass in @xmath28 . eqs ( [ mass1 ] ) are equivalent to the eulerian form @xmath29the volume energy @xmath18 is given by the behavior of the mixture cahn , dunn2,eglit . in our mechanical case , for an energy depending on gradients of densities , the volume energy is @xmath30the potential @xmath31defines the _ specific free enthalpy or chemical potential _ of the constituent @xmath20 of the mixture @xcite . subscript @xmath32 corresponds to the spatial derivatives associated with gradient terms . usually , summation is made on repeated subscript @xmath32 . in practice , we consider a quadratic form with constant coefficients @xmath33 @xmath34 such that @xmath35where @xmath36 is the value of the volume energy of the homogeneous bulks . to obtain the equations of motions , we use a variational principle whose original feature is to choice variations in reference spaces ( fig . they are associated with a two - parameter family of virtual motions of the mixture ( see appendix a ) . the equation of the motion of each constituent of the mixture writes ( gouin - ruggeri and the references therein ) @xmath37where @xmath38 denotes the acceleration of the component @xmath20 ( @xmath20=1,2 ) . in applications , the motions are supposed to be isothermal ( @xmath1 denotes the common temperature value of the two components ) and correspond to strong heat exchange between components . in thermodynamics , this case corresponds to a function @xmath36 as the volume free energy of the homogeneous mixture at temperature @xmath1 . in our model , the equations of motion ( [ motions 1 ] ) yield @xmath39taking eq . ( [ mass2 ] ) into account , we can note that the two equations of system ( [ motions 1 ] ) are equivalent to the system@xmath40for which equations of components are not in divergence form , but the summation of the two equations and the fact that @xmath41 allow to obtain the total motion of the mixture in divergence form @xcite . therefore , while the global momentum equation represents a balance law , individually , the equations of system ( [ non divergence form ] ) do not . an interface in a two - phase mixture is generally schematized by a surface without thickness . far from critical conditions , this layer is of nanometre size and density and entropy gradients are very large . this is not the case in the immediate vicinity of critical points where fluctuations of densities strongly diverge . our goal is to schematize the average of these fluctuations by means of travelling waves in a continuous model using an energy in form ( [ nonlocal energy ] ) . to obtain one - dimensional travelling waves in isothermal case when we neglect diffusion and viscosity , we are looking for solutions only depending on the variable @xcite @xmath42where @xmath43 denotes the wave velocity . equations ( [ mass2 ] ) of conservation of masses write @xmath44 where @xmath45 denotes the one - dimensional velocity of component @xmath20 . equivalently @xcite , @xmath46where @xmath47 , are constant along the motion and have the dimension of a mass flow . the acceleration of mixture component _ i _ is @xmath48and system ( [ motions 2 ] ) yields @xmath49where @xmath50 and @xmath51 are two constants , superscript @xmath52 denotes the transposition and @xmath53here @xmath54 denotes the derivation along the ray . let us note that the sum of equations of system ( [ dynamical system ] ) , respectively multiplied by @xmath55 and @xmath56 , yields the first integral @xmath57where @xmath58 is a constant.consequently , in one - dimensional system , the profiles of mixture densities are associated with a _ mechanical system _ having an equivalent _ kinetic energy _ @xmath59and an equivalent _ potential of forces _ @xmath60it is possible to use all the tools of analytical dynamics as lagrange or hamilton equations , symplectic geometry or maupertuis methods . in each bulk , density gradients are null . eliminating constants @xmath61 , dynamical conditions through the interfacial layer yield @xmath62 = 0 \\ \left [ g_{,\rho _ { 2}}(\rho _ { 1},\rho _ { 2})\right ] = 0\ \\ \left [ g-\rho _ { 1}\,g_{,\rho _ { 1}}-\rho _ { 2}\,g_{,\rho _ { 2}}\right ] = 0 \,\ , , \end{array}% \right . \label{dynamic system}\ ] ] where @xmath63 = u_\beta - u_\alpha $ ] denotes the difference of values of @xmath64 between the two mixture bulks @xmath65 and @xmath66 and @xmath67 in case of equilibrium at a given temperature @xmath1 , ( @xmath68 and @xmath69 are null ) , the minimum of the total free energy , with a given total mass for each constituent , yields conditions @xmath70 = 0 \\ \left [ \mu _ { 2}(\rho _ { 1},\rho _ { 2})\right ] = 0\ \\ \left [ p(\rho _ { 1},\rho _ { 2})\right ] = 0 \,\ , , \end{array}% \right . \label{static bulks}\]]where @xmath71denotes the total thermodynamical pressure of the binary mixture . the constant @xmath72 , corresponding to the fact that @xmath73 is defined to an additive constant , can be defined as the critical pressure term for a value of @xmath74 null at the critical point . conditions of system ( [ dynamic system ] ) ( or system ( [ static bulks ] ) ) express that the bulks of gibbs surface @xmath75 associated with the dynamical volume free energy @xmath76 ( or static volume free energy @xmath73 ) correspond to the contact points with a bitangent plane ( fig . 2 ) . by adding the term @xmath77 to @xmath73 , the study of non - dissipative travelling waves of density components of a mixture turns back to an equivalent equilibrium problem . -axis is @xmath78 . in dynamical case , the intersection corresponds to the opposite of a _ dynamical pressure _ ( and defined to an additive constant @xmath72 ) . contact points @xmath79 and @xmath80 of the bitangent plane with surface @xmath75 generate a curve @xmath81 . along this _ spinodal _ curve @xmath82 , the straight line ( @xmath83 ) connecting the contact points @xmath79 and @xmath84 associated with the two mixture bulks is the _ characteristic line _ of the plane which generates a developable surface . critical point @xmath85 corresponds to the collapse of the two contact points at the given temperature @xmath1.,width=642 ] in appendix b , we propose an expansion form of the volume energy @xmath73 for the binary mixture , near its critical point at temperature @xmath1 , in the form @xmath86where @xmath87 and @xmath88 are physical constants , @xmath89 and @xmath90 are convenient linear combinations of @xmath91 and @xmath92 with @xmath93 and @xmath94 denoting the density values of the two constituents at the critical point ( for the sake of simplicity , we denote the volume free energy as @xmath95 in all systems of variables ) . with the change of variables , @xmath96 and @xmath97system ( [ dynamical system ] ) can be written@xmath98 we denote @xmath99where scalars @xmath100 are depending on the thermo - mechanical properties of the mixture near the point @xmath85 . by the new change of variables , @xmath101 and @xmath102the system ( [ ndynasyst ] ) writes in the form @xmath103or @xmath104where @xmath105 and @xmath106 . due to the fact @xmath107 is symmetric positive definite and @xmath108 is not singular , the same holds for @xmath109 . the values of constants @xmath110 and @xmath111 must correspond to the singular points of the differential system ( [ dynamical system ] ) when @xmath112 . these singular points are associated with the two bulks of the binary mixture . consequently , @xmath113 and @xmath114 ( see appendix b , for the meaning of @xmath115 as a _ `` distance '' _ to the critical point ) . we consider the following hypothesis @xmath116 and @xmath117 : @xmath118we assume that @xmath89 and @xmath90 slowly vary as a function of the one - dimensional parameter @xmath119 . the hypothesis can be expressed by a change of variable @xmath120 , where @xmath121 is an adimensional small parameter ( @xmath122 ) . with the following change of variables : @xmath123system ( [ final system develo ] ) yields @xmath124where the dot denotes the derivation with respect to @xmath125 . the solution must be a phase transition connecting singular points @xmath126 and @xmath127 . @xmath128the density profiles are of small amplitude with respect to the critical densities . consequently , we look for the solutions in the form @xmath129where @xmath130 and @xmath131 are to real positive constants . system ( premiere rescale ) yields @xmath132the form of the volume energy near the critical point must be conserved by affinity . this assumption corresponds to the universality of the form of the energy near a critical point @xcite . consequently in this rescaling process , @xmath133where @xmath134 is a positive constant with physical dimension @xmath135 . system ( [ deuxieme rescale ] ) yields the equality of same order terms , @xmath136 if we assume @xmath137 , eq . ( [ troisieme rescale])@xmath138 reduces to @xmath139 , which is not physically possible ; if we assume @xmath140 , eq . ( [ troisieme rescale])@xmath141 reduces to @xmath139 . then , to obtain non - trivial solutions , we must assume that @xmath142 and @xmath143 ; consequently , @xmath144 now , if we compare @xmath145 and @xmath146 , we must assume that @xmath147 . system ( [ troisieme rescale ] ) leads to the _ rescaled system _ @xmath148which is equivalent to @xmath149where @xmath150 is a convenient constant of integration . the first equation of system ( [ integrate recaled sytem ] ) has several possible solutions . in this case we look for phase transition , the solution must be monotonic and must go from an equilibrium point to the other one . the simplest possibility is @xmath151 . the integration of eq . ( integrate recaled sytem)@xmath152 yields@xmath153 consequently , with the change of orientation of the @xmath119-axis , we get @xmath154due to the changes of variables ( [ change 1 ] ) , ( [ change 2 ] ) , we obtain the general form of the phase transitions @xmath155depending on the two cases of boundary conditions , we obtain 1@xmath156 case : when @xmath157 and @xmath158 with @xmath159 , then @xmath160 2@xmath161 case : when @xmath162 and @xmath163 with @xmath159 , then @xmath164 let us note that the constants @xmath165 and @xmath166 depend on the thermodynamical behavior of the mixture and they may be positive or negative . the variation of @xmath167 and @xmath168 in system ( [ sol r ] ) yields the different cases . they are summarized on fig . 3 as domains delimited by the bisectrices of axes @xmath169 and @xmath170 and we represent the different possible variations of each component on fig . 4 . -axis and @xmath170-axis . points on axis @xmath165 where @xmath171 are forbidden ; this will not be the case in dynamics.,width=264 ] versus @xmath172.,width=340 ] we see , as it is known by other methods in physical chemistry that the density profiles in interfacial densities are not necessary homogeneous for mixtures of fluids @xcite . in the case of motion , we check with travelling waves submitted to viscosity ; in this case , system ( [ motions 2 ] ) can be modified to take account of the viscosity in each component of the mixture @xmath173where @xmath174 is the viscous stress tensor associated with component @xmath20 . consequently we obtain in variables @xmath175 and @xmath90 , the system of equations of motions in one - dimensional system with @xmath176 , where @xmath177 is the coefficient of viscosity of the component @xmath20 @xcite . the system can be written by using one dimensional derivation in @xmath178 by integration as in section 4.1 and with the fact that @xmath179 , @xmath180 we can make an analogous rescaling process as in section 4.1.1 . the difference only comes from scalars @xmath68 and @xmath69 . they must be of small amplitude ; consequently , we look for constants in the form @xmath181where constants @xmath182 and @xmath183 are positive . to obtain convenient solutions we must have @xmath184then , the terms @xmath185 and @xmath186 which are of order greater than the other terms in system ( [ system travelling waves ] ) , are negligible . finally , system ( [ system travelling waves ] ) yields @xmath187where @xmath188 and @xmath189 is a new constant of integration . a qualitative study of eq . ( [ trav integrated])@xmath152 can be considered . constant @xmath87 is strictly positive . the discussion comes from @xmath190 where @xmath191 represents a quartic and @xmath192 represents a straight line . if @xmath193 is a convex quartic and only periodic solutions are possible . consequently @xmath194 . two interesting cases can be considered : \a ) the straight line ( [ line ] ) is bitangent to the quartic ( [ quartic ] ) . the solution of eq . ( [ trav integrated])@xmath152 is a travelling wave of phase transition in the same form than the two first curves of fig . we must have @xmath195 then,@xmath196 we note that @xmath197 plays the role of @xmath198 in the static case . the solutions in variables @xmath13 and @xmath14 are in the same form than in section 4.1.2 and the same sort of density profiles can be observed @xmath199 we have seen that to obtain travelling waves near a critical point , the viscosity terms have to be negligible with respect to the others . this result is in accordance with dispersive waves as the ones obtained by using the korteweg - de vries equation @xcite . the equations of binary mixtures of fluids are considered in the one - dimensional case . by using an asymptotic expansion of the volume free energy near a critical point and by taking the density gradients into account , we get , in continuum mechanics , a dynamical system representing the equations of travelling waves for the two - fluid components . a rescaling process associated with the universality of the energy form near a critical point allows to obtain different sorts of travelling waves of density as well as in conservative and dissipative cases . we can discuss different possible phase transitions according to the values of phases densities of components and the characteristic of the mixture . now , we consider some physical comments . for given densities , at time @xmath200 , the velocities of the mixture components are deduced from the mass balance equations ( [ balancemasses ] ) . then , an arbitrary constant @xmath43 is introduced in the velocity fields : @xmath201 and at any time @xmath23 the relation leads to @xmath202 the initial conditions yield this arbitrary velocity @xmath43 . _ first case : interface propagation - _ by crossing the interface ( _ interface _ means the domain where the densities drastically change ) , the volume changes . this phenomenon can not occur in a closed tube with constant volume , but only in a tube with one closed end . it is similar to a phenomenon of vaporization or condensation.one imagines that the other side of the tube is closed with a piston whose displacement may be imposed . the component velocities at the fixed end are zero in bulk @xmath203 ; this determines the constant @xmath43 thanks to the condition @xmath204 due to the fact that @xmath205 and @xmath206 , we get @xmath207 and the values of @xmath68 and @xmath69 are not independent . _ second case : solitary waves - _ the volume of the interface only moves in one bulk phase and remains constant . such a wave may move in a closed tube as natterer s one @xcite . at the ends which are assumed far from the interface , the velocities in the bulk phase are zero and we still obtain eq . ( [ debits ] ) such that the values of @xmath68 and @xmath69 are not independent . hamilton s principle - variational form of the principle of virtual powers - allows us to derive the equations of motions . the variations of motions of particles are deduced from the functions @xmath208for which @xmath209 are two parameters defined in a neighborhood of zero ; they are associated with two families of virtual motions of the mixture . the real motions correspond to @xmath210 : @xmath211 . virtual material displacements associated with any variation of real motions can be written as @xcite @xmath212such a variation is dual with serrin s @xcite . this has been studied in @xcite and corresponds to the natural variation of the motion in a lagrangian representation . let us denote @xmath213 ; we deduce the two relations in lagrangian variables @xcite @xmath214where @xmath215 is the linear form associated with the gradient and @xmath216 is the divergence operator on @xmath28 . assuming that terms on the edge of @xmath217 are zero , we obtain @xmath218or@xmath219where @xmath220 @xmath221 . but , @xmath222so , the terms given by integration on the edge of @xmath28 being zero , @xmath223finally , for each constituent,@xmath224and by using the relation@xmath225we obtain @xmath226this leads to the vectorial form ( [ motions 1 ] ) @xmath227 in this appendix , thanks to a direct method of differential geometry , we obtain the form of the free energy of a binary mixture near a critical point . in physical chemistry , the form is presented thanks to the knowledge of the thermodynamic coordinates of the bulks when we are near a critical point @xcite . in this new presentation , we take only account of conditions of equilibrium of bulks when the critical point marks the limit of their coexistence . such a method implies general physical consequences as in one component fluid when the two variables are densities of matter and entropy but might also be extended in other topics when different equilibrium states collapse only in one state as in finance or biology . the volume free energy of binary mixture at temperature @xmath1 is taken in the form : @xmath228where @xmath229 are the densities of the two components . the following relations between the chemical potentials @xmath230 of the two constituents and @xmath73 hold @xmath231and the thermodynamical pressure is given by @xmath232 as the function @xmath73 is twice differentiable , the geometrical representation of @xmath73 by a surface @xmath233 admits a tangent plane at any point . the normal vector to the plane has the direction coefficients ( -@xmath234 -@xmath235 , 1 ) . the phase equilibrium of two bulks @xmath65 , @xmath66 is described by the conditions @xmath236 it is noticeable that , in this case , the third coordinate of the intersection point of the bitangent plane with the third axis is @xmath237 ( see fig . the contact points @xmath238 , @xmath239 of the bitangent plane to @xmath240 generate two curves @xmath241 , @xmath242 respectively . the curve @xmath243 is the _ spinodal _ curve of the phase transition . the case with only one contact point corresponds to a mixture with only one bulk while two contact points with a bitangent plane correspond to a two - bulk equilibrium mixture . when , at a given temperature @xmath1 , @xmath244 tends to @xmath239 , the limit point @xmath85 is a _ critical point _ corresponding to the fact that two bulks collapse into one bulk . the study of the mixture equilibrium near the critical point @xmath85 , at a given temperature @xmath1 , is closely related to the shape of surface @xmath233 in the vicinity of @xmath85 . so it is interesting to prove some theorems giving us information about the geometry of the surface @xmath233 in the vicinity of @xmath85 . let @xmath246 the equation of curve @xmath82 so that @xmath247= @xmath248 . the tangent plane at @xmath249 or @xmath250 is @xmath251or , in the @xmath82 representation , @xmath252and @xmath253it follows @xmath254when @xmath255 and @xmath256 tends to @xmath257 , the plane ( @xmath258 ) tends to the osculatory plane to @xmath82 at @xmath85 because the vector @xmath259 admits the limit @xmath260 . let @xmath263 an ordinary point of @xmath82 . in a neighborhood of @xmath263 , we have @xmath264such that @xmath265 corresponds to @xmath85 . let @xmath249 and @xmath250 two points of @xmath261 with @xmath266 . ( ordp ) lead to @xmath267and by changing @xmath256 with @xmath255 , @xmath268the two vectors in eqs.([un ] , [ du ] ) are parallel if and only if @xmath269equation ( [ sys])@xmath152 yields @xmath270 and then ( [ sys])@xmath271 is verified . this means that any relation @xmath272 different from identity is not allowed and @xmath273 is not possible . then , the critical point @xmath263 can not be an ordinary point of @xmath82 . let us now consider the simplest singular case for @xmath261 at @xmath85 @xmath274the parallelism conditions for vectors @xmath275 and @xmath276 are equivalent to @xmath277equation ( [ syss])@xmath271 implies @xmath278 which has , for @xmath279 , the unique possible solution @xmath280 and then ( [ syss])@xmath152 is satisfied . the equations of the projection @xmath281 of the curve @xmath282 on the plane @xmath283 are @xmath284and the straight line @xmath285 remains bitangent to @xmath286 . consequently , in the plane @xmath283 , the main part of the equation of @xmath281 is written in the form @xmath287 in adimensional form , @xmath288 and the main part of the equation of surface @xmath289 generated by this curve in a neighborhood of the critical point @xmath290 , when point @xmath85 is taken as origin of the axes and @xmath291 is the osculatory plane to @xmath233 at @xmath263 , is @xmath292the dimensions of @xmath119 and @xmath293 are different . in an adimensional system of axes , the equation writes@xmath294 material waves of a fluid in the vicinity of the critical point _ in _ symposium on waves in liquid / gas and liquid / vapor two - phase systems , s. morioka s. , wijngaarden l. ( eds . ) , kluwer publ . , netherlands , 1995 .	travelling waves of densities of binary fluid mixtures are investigated near a critical point . the free energy is considered in a non - local form taking account of the density gradients . the equations of motions are applied to a universal form of the free energy near critical conditions and can be integrated by a rescaling process where the binary mixture is similar to a single fluid . nevertheless , density solution profiles obtained are not necessarily monotonic . as indicated in appendix , the results might be extended to other topics like finance or biology . # 1 binary fluid mixtures ; critical points ; travelling waves ; rescaling process . 46.15.cc ; 47.35.fg ; 47.51.+a ; 64.75.ef
the main parameters that characterize any quantum resonance , are the collision energy @xmath3 , at which this state can be excited , and the width @xmath4 that determines the lifetime of the state . for a multi - channel system the total width is the sum of the partial widths , @xmath5 , where @xmath6 gives the relative probability of decaying into the @xmath7-th channel . there are many different methods for determining these parameters from a set of scattering data ( several of them are described in refs . these methods form two big groups based on two principally different approaches . + within one approach , the parameters @xmath3 , @xmath4 , @xmath8 , @xmath9 , etc are treated as the adjustable variables in a procedure of fitting the available experimental data . the simplest and most well - known example of such a method is the breit - wigner parametrization of the amplitude @xcite . a common feature of all of the methods belonging to this category , is that the number of resonances is fixed from the outset . all these methods use some parametric expression for the amplitude , or for the @xmath0-matrix , or directly for the cross section , where the resonance singularities ( or the zigzags of the cross section ) are embedded into this parametric expression by hand . these methods only differ in the method of parametrization and in the derivation of the parametric expression . + within the second approach , the resonances are considered as the poles of the @xmath0-matrix at the complex energies @xmath10 in an appropriate domain of the riemann surface of the energy . the @xmath0-matrix is written in a more general form with some adjustable parameters that do not necessarily coincide with the resonance parameters . usually , it is not known beforehand how many resonances can be found ( if any ) . after fitting the data at real collision energies , the analytic expression for the @xmath0-matrix thus obtained , is examined at complex energies where the poles ( if found ) are interpreted as the resonances . the pad approximation of the @xmath0-matrix @xcite and the laurent - pietarinen series expansion of the amplitude @xcite can be mentioned as examples . + the method we describe here belongs to the second category and is based on the rigorous semi - analytic expression for the @xmath11-channel jost matrix derived in ref . @xcite . in that expression , all the factors responsible for the `` bad behaviour '' of the jost matrix ( i.e. factors depending on the sommerfeld parameters and the channel momenta responsible for the branching of the riemann surface ) are given explicitly . the remaining unknown factors are analytic and single - valued functions of @xmath1 defined on a simple energy plane . these functions are expanded in the taylor series , and the expansion coefficients serve as the fitting parameters . @xcite , it was shown that for a non - relativistic reaction of the type @xmath12 involving charged particles , the @xmath11-channel jost matrix has the following general form : @xmath13 \nonumber & - & \left.\left [ \frac{2\eta_mh(\eta_m)}{c_0 ^ 2(\eta_m)}\pm i\right ] c_{\ell_m}(\eta_m)c_{\ell_n}(\eta_n ) k_m^{\ell_m}k_n^{\ell_n+1}{b}_{mn}(e)\right\}\ , \end{aligned}\ ] ] where @xmath14 are the channel momenta determined by the differences between the total energy @xmath1 and the channel thresholds @xmath15 , as well as by the corresponding reduced masses @xmath16 ; the channel angular momenta and the sommerfeld parameters are @xmath17 and @xmath18 ; the function @xmath19 is the coulomb barrier factor ; and @xmath20-\ln \hat{\eta}\ , \qquad \psi(z)=\frac{\gamma'(z)}{\gamma(z)}\ , \qquad \hat{\eta}=\frac{\mu e^2\|z_1z_2\|}{k\hbar^2}\ .\ ] ] it was shown that the remaining unknown matrices @xmath21 and @xmath22 in eq . ( [ multi.matrixelements ] ) are single - valued and analytic functions of the energy , defined on a simple energy plane without branching points . all the complicated topology of the riemann surface where the jost functions are defined , is determined by the coefficients of the matrices @xmath21 and @xmath22 , given in eq . ( [ multi.matrixelements ] ) explicitly . + if for a given energy @xmath1 the jost matrices ( [ multi.matrixelements ] ) are known , then the corresponding @xmath0-matrix is just their `` ratio '' , @xmath23^{-1}\ , \ ] ] and the scattering cross section for the channel @xmath24 can be found as @xmath25 the resonances are the points @xmath26 on the riemann surface of the energy , where @xmath27 and therefore where the @xmath0-matrix has poles . + the energy surface has a square - root branching point at every channel threshold @xmath15 . this is because the jost matrices depend on the energy @xmath1 via the channel momenta ( [ chmom ] ) and for each of them there are two possible choices of the sign in front of the square root . the resonance spectral points are located on the so called non - physical sheet of this riemann surface , i.e. such a layer of the surface where all the channel momenta have negative imaginary parts . in the numerical calculations , the choice of the sheet is done by an appropriate choice of the signs in front of the square roots ( [ chmom ] ) . + since the matrices @xmath21 and @xmath22 are analytic , they can be expanded in the taylor series around any complex point @xmath2 . near this point , they can therefore be approximated by the first @xmath28 terms of these series : @xmath29 \label{b_series } b(e ) & \approx & \sum_{i=0}^m b_i(e_0)(e - e_0)^i\ , \end{aligned}\ ] ] where the expansion coefficients @xmath30 and @xmath31 are ( @xmath32)-matrices . these matrices depend only on the choice of the point @xmath2 . after finding them , the jost matrices ( [ multi.matrixelements ] ) can be used at any complex energy @xmath1 within a circle around @xmath2 where the approximations ( [ a_series ] , [ b_series ] ) are satisfactory . + we treat the elements of the matrices @xmath30 and @xmath31 as the adjustable parameters in the procedure of fitting experimental cross section . after finding the optimal values for them , we look for the roots of eq . ( [ spectral ] ) and thus find the resonance parameters @xmath3 and @xmath4 . as to the partial widths @xmath33 , they can easily be found following the procedure described in ref . indeed , we know their sum @xmath34 and we can find their ratios ( see ref . @xcite ) : @xmath35 at a resonance energy , @xmath36 , all the elements of the @xmath0-matrix are singular because all of them have the same singular factor @xmath37 . however , in the ratio ( [ ggratio ] ) this factor cancels out . therefore , if we explicitly invert the matrix @xmath38 and use it in eq . ( [ s_matrix ] ) without common factor @xmath39 , then we can avoid numerical evaluation of the limit ( [ ggratio ] ) . in the simplest case of a two - channel problem , we obtain : @xmath40 it should be noted that such a simple and numerically stable procedure for calculating the partial widths is only possible when the jost matrices are parametrized . if we were parametrizing the @xmath0-matrix directly , then the limits ( [ ggratio ] ) would have to be calculated numerically as the ratios of singular functions . we assume that there are sets of experimental data available for at least one channel @xmath24 ( or perhaps for several channels ) , i.e. the cross sections @xmath41 with the corresponding experimental errors ( standard deviations ) @xmath42 , measured at the collision energies @xmath43 . the center @xmath2 of the expansions ( [ a_series ] , [ b_series ] ) can be chosen somewhere within the interval covered by these collision energies ( where we expect to find a resonance ) . the optimal values of the expansion parameters are found by minimizing the following @xmath44 function @xmath45 ^ 2\\[3 mm ] \nonumber & + & \displaystyle \sum_{j=1}^{n^{(m'n')}}\left[\frac{\sigma_{m'n'}(e^{(m'n')}_j)- \sigma^{\mathrm{fit}}_{m'n'}(e^{(m'n')}_j ) } { \delta^{(m'n')}_j}\right]^2 \ + \ \cdots\\[3 mm ] \nonumber & + & \displaystyle \sum_{m < n , j}\left\|s_{mn}(e^{(mn)}_j)-s_{nm}(e^{(mn)}_j)\right\|^2\ , \end{aligned}\ ] ] where the fitting cross section @xmath46 depends on the expansion coefficients via eqs . ( [ cross_section_sigma_nm ] ) , ( [ s_matrix ] ) and ( [ multi.matrixelements ] ) . the last sum in the above @xmath44 function makes the fitted @xmath0-matrix symmetric in accordance with the detailed balance theorem ( see a more detailed discussion in ref . the minimization is done using the minuit code @xcite . + the number of the adjustable parameters depends on the number @xmath11 of the existing channels ( the data do not have to be available for all of them ) and on the number @xmath28 of the terms in the taylor series ( [ a_series ] , [ b_series ] ) . generally speaking , the expansion coefficients @xmath47 and @xmath48 are the @xmath32 matrices of complex elements . however , as was shown in ref . @xcite , they become real matrices if the point @xmath2 is on the real axis . therefore , if @xmath2 is on the interval covered by the experimental energies @xmath43 , the the total number of real fitting parameters is @xmath49 . in order to demonstrate the efficiency of the proposed method , we choose a model two - channel problem where the parameters of the resonances can be determined in an exact way . for this model , we generate artificial data points with a typical distribution of errors . in addition to the error - bar for each pseudo - data point , we introduce a random shift ( up or down ) from the exact cross section curve , i.e. a typical experimental `` noise '' . using these points , we extract the resonance parameters and compare them with the corresponding exact values . + our artificial data points are generated using the following two - channel potential having a coulomb tail : @xmath50 the short - range term in this potential is the same as in the famous noro - taylor model @xcite . the units in eq . ( [ pot ] ) are therefore the same , namely , they are such that the reduced masses for both channels are equal to one , @xmath51 , with @xmath52 , and both angular momenta are zero , @xmath53 . the threshold energies for the channels are @xmath54 and @xmath55 . + .the resonance energies and widths generated by the potential ( [ pot ] ) . [ cols="^,^,^,^,^ " , ] the main idea of the proposed method has its roots in the so called `` effective - range expansion '' widely used in nuclear and atomic physics ( see , for example , ref . @xcite ) . within this approach , a certain function of the scattering phase - shift is expanded in the power series of the collision energy , and the expansion coefficients are used as the adjustable parameters to fit experimental data . traditional effective - range expansion is very useful but is limited to low energies . moreover , it is difficult to apply it to multi - channel processes . + we generalize this approach and thus remove all these limitations . first of all , we expand a more fundamental quantity : the jost matrix . when the jost - matrix expansion in the power series is obtained , one can easily derive the corresponding expansion of the @xmath0-matrix , the phase - shift , or any other quantity that is needed . secondly , it is not necessary to do the expansion near the point @xmath56 . actually , the expansion can be done around any complex value of variable @xmath1 . concerning the multi - channel problems , for the jost - matrices , this does not pose any difficulty . simply , the expansion coefficients become matrices . + in order to expand a function in the taylor series of @xmath1 , one has to be sure that this function is an analytic and single - valued function of @xmath1 . however , all the quantities describing the scattering processes ( amplitude , @xmath0-matrix , jost matrix , etc ) are multi - valued functions defined on a complicated riemann surface of the energy with the number of branch points equal to the number of channels . in order to circumvent this difficulty , we use earlier derived general semi - analytic expression for a multi - channel jost matrix , where all the factors responsible for its `` bad behaviour '' , are given explicitly . the remaining unknown functions are more simple , more smooth , and ( more importantly ) are analytic and single - valued functions defined on a simple energy plane . we only do the expansion for these functions . + in our earlier publication @xcite , we demonstrated how the proposed method works for the reactions involving neutral particles . in the present paper , we consider a more general case when the potential has both a short - range part and a coulomb tail . similarly to the simple effective - range theory , presence of the coulomb potential makes the explicit coefficients in the semi - analytic expression for the jost matrix more complicated . however , the remaining functions of @xmath1 are still smooth and can be approximated by just a few terms of the taylor series . we have demonstrated this using a two - channel model . + we have shown that even with rather inaccurate experimental data ( @xmath57 in our model ) the resonances can still be found . for the narrow resonance , we managed to reproduce not only the energy and total width , but the partial widths as well , when the data points had rather big deviations from the exact values . for a wide resonance ( the resonance number 2 ) with such a high experimental noise , we still obtained reasonable parameters . and even the energy of the extremely wide resonance ( number 3 ) was obtained not far from the exact value . this shows that the proposed method is accurate and stable . + one of the advantages of the proposed method is that the fitting procedure involves all matrix elements of the jost matrix and therefore all elements of the @xmath0-matrix even if the data are available just in one channel . the resulting @xmath0-matrix should therefore be correct in all channels . this means that by fitting accurately - measured data in one or two channels , we could in principle obtain a reasonable estimate for the cross section in the other channels where the measurements are difficult or impossible . + it should be noted that the proposed method is non - relativistic and therefore can not be directly used in high - energy physics . there are however a wide range of problems in atomic and low - energy nuclear physics , where it could find applications . in principle , one can try the same parametrization for high energies as well , if the relativistic relation between the energy and momentum , @xmath58 , is used in all the formulae . such `` intuitive '' inclusion of relativistic kinematics into non - relativistic operators is very common for various parametrizations of scattering data in particle physics . in our case , however , this would mean that the mathematical rigor and substantiation are lost . s. a. rakityansky , s. a. sofianos , n. elander , _ `` pad approximation of the s - matrix as a way of locating quantum resonances and bound states '' _ , j. phys . a : math . 40 , pp . 1485714869 ( 2007 ) . a. @xmath60varc , m. had@xmath61imehmedovi , h. osmanovi , j. stahov , l. tiator , l. r. workman , _ `` introducing the pietarinen expansion method into the single - channel pole extraction problem '' _ , phys . rev . c 88 , 035206 ( 2013 ) . rakityansky , n. elander , _ `` a method for extracting the resonance parameters from experimental cross - sections '' _ , international journal of modern physics e vol . 22 , no . 5 1350032 ( 17 pages ) , ( 2013 ) .	a new parametrization of the multi - channel @xmath0-matrix is used to fit scattering data and then to locate the resonances as its poles . the @xmath0-matrix is written in terms of the corresponding `` in '' and `` out '' jost matrices which are expanded in the taylor series of the collision energy @xmath1 around an appropriately chosen energy @xmath2 . in order to do this , the jost matrices are written in a semi - analytic form where all the factors ( involving the channel momenta and sommerfeld parameters ) responsible for their `` bad behaviour '' ( i.e. responsible for the multi - valuedness of the jost matrices and for branching of the riemann surface of the energy ) are given explicitly . the remaining unknown factors in the jost matrices are analytic and single - valued functions of the variable @xmath1 and are defined on a simple energy plane . the expansion is done for these analytic functions and the expansion coefficients are used as the fitting parameters . the method is tested on a two - channel model , using a set of artificially generated data points with typical error bars and a typical random noise in the positions of the points . published in : + international journal of modern physics e , vol . 25 , no . 2 ( 2016 ) 1650014 ( 12 pages ) + doi : 10.1142/s0218301316500142
an ordered state known as the @xmath1-density wave ( ddw ) has been proposed as the origin of the pseudogap phase of the cuprates.@xcite a variety of experiments have been explored from this perspective . these include the superfluid density and the resonance peak in neutron scattering,@xcite the hall number,@xcite angle resolved photoemission spectroscopy ( arpes),@xcite the specific heat,@xcite the quasiparticle charge,@xcite and the direct signature of ddw in polarized neutron scattering.@xcite in addition , it has been explored how the notion of this competing order , when combined with interlayer tunneling , and the doping imbalance of the multilayered cuprates , can result in the striking systematics of the layer dependence of the superconducting transition temperature @xmath4.@xcite in all cases , the theory is consistent with the existing observations . in this paper , we will address the zero - temperature infrared ( ir ) hall angle @xmath5 as a function of the hole - doping @xmath3 , because we are encouraged by the recent measurements of rigal _ et al._@xcite there are two specific reasons : ( 1 ) the ddw state predicts hole pockets as fermi surfaces in the underdoped cuprates , which should have important experimental consequences . arpes experiments can only detect half of each of these pockets,@xcite which therefore appear as fermi arcs.@xcite thus an important prediction of our theory remains untested , except through its indirect signature in the doping dependence of the superfluid density . a measurement of @xmath5 can , in principle , clarify this issue , and we believe that it has.@xcite ( 2 ) the ddw theory also predicts a quantum critical point at a doping @xmath6 within the superconducting dome and it has been argued that this should be visible in the hall number , @xmath7,@xcite if superconductivity is destroyed by applying a magnetic field . there is some experimental evidence of this effect.@xcite the difficulty with this experiment is that it needs to performed in a field as high as 60 t , which is experimentally quite demanding . we believe that a measurement of @xmath8 at high frequencies in the pseudogap state above @xmath0 should have a similar behavior at @xmath9 as @xmath7 does . we expect that the high frequency behavior at @xmath10 will be similar to the @xmath11 behavior with superconductivity destroyed by a magnetic field if both experiments probe the same underlying state which we believe is the ddw state which causes the pseudogap and coexists with superconductivity in the underdoped superconducting state . given that the ddw state is a broken symmetry state with a local order parameter , it should be describable by a mean field hartree - fock theory and its consequent elementary excitations . this is precisely the approach we shall assume in the present paper . the mean field hamiltonian for the ddw state is @xmath12,\ ] ] where @xmath13 is the annihilation operator for an electron of spin @xmath14 in the @xmath15-direction and momentum @xmath16 , @xmath17 is the chemical potential , and the vector @xmath18 . the lattice spacing will be set to unity . we ignore the residual interactions between quasiparticles ; the principal effect of electron - electron interactions is to produce non - zero @xmath19 . the single particle spectrum on the square lattice with nearest - neighbor hopping @xmath20 and next - neighbor hopping @xmath21 is @xmath22 the @xmath1-wave order parameter of the ddw state is @xmath23 where the amplitude @xmath24 is a function of doping . we can express the hamiltonian in terms of a two - component quasiparticle operator : @xmath25 , and then diagonalize this @xmath26 hamiltonian to get @xmath27 & 0 \\ 0 & [ e_{-}({\bf k})-\mu ] \end{array } \right ) \chi_{{\bf k},\alpha } . \label{eq : twobytwo}\end{aligned}\ ] ] the two - component quasiparticle operator @xmath28 is unitarily related to @xmath29 , and the sum is over the reduced brilloun zone ( rbz ) . @xmath30 are the two bands of the ordered ddw state , with @xmath31 and @xmath32 . for a system of ddw quasiparticles in the presence of a magnetic field @xmath33 in the @xmath15-direction , and an electric field @xmath34 in the @xmath35 plane , @xmath5 is the angle between @xmath34 and the current @xmath36 : @xmath37 . we will compute the necessary conductivities , @xmath38 and @xmath39 in the framework of boltzmann theory@xcite applied to the ddw mean - field hamiltonian . since we consider a non - interacting model , this semiclassical approach easily generalizes to finite frequencies as well . a number of comments regarding the validity of our boltzmann approach are in order . 1 . in a normal metal , it is well known ( see ref . ) that the external frequency @xmath40 and wavevector @xmath41 must satisfy @xmath42 and @xmath43 , where @xmath44 is the fermi wavevector . although we must have @xmath45 for localization effects to be neglected ( @xmath46 is the mean free path ) , there are no further restrictions on the product @xmath47 , where @xmath48 is the lifetime due to impurity scattering . 2 . in a superconductor , the same conditions apply at high frequencies , unless we want to capture interesting order parameter disequilibrium effects , such as charge imbalance etc . , whence we must satisfy @xmath49 , where @xmath50 is the superconducting gap.@xcite 3 . for a particle - hole condensate , such as ddw , the condition for the validity of the boltzmann equation should be the same as in a normal metal . the diagonalization in eq . ( [ eq : twobytwo ] ) does not mix particles and holes and , therefore , we can apply the boltzmann formalism to ddw quasiparticles , which have relatively simple , particle - number conserving scattering terms . we assume that ddw quasiparticles have only one scattering time , though it may vary along the fermi surface.@xcite this assumption is clearly supported by experiments , at least in the pseudogap regime of ybco@xmath51 for @xmath52.@xcite it surprisingly appears to be true for even very lightly doped sample of @xmath53.@xcite the alternate view that for each @xmath54 , there are two scattering times@xcite @xmath55 and @xmath56 appears to be untenable in this regime . ( above the ddw ordering temperature , the situation may , of course , be more complicated . ) further complications from interband transitions will be neglected , because , to a first approximation , the effect of these high frequency processes will be simply to renormalize the effective single band parameters . the longitudinal and hall conductivities , @xmath57 and @xmath58 , are : @xmath59 \delta(e_{+}(k)-\mu ) + ( e_{+ } \rightarrow e_{-})\\ & = & \frac{2e^{3}h}{(2\pi)^{2}\hbar^{4}c}i_{xy}[\omega,\tau_{\bf k } ] , \label{eq : xy } \\ \sigma_{xx}(\omega ) & = & \frac{2e^{2}}{\hbar^{2}}\int_{\text{rbz}}\frac{d^{2}k}{(2\pi)^{2 } } \left(\frac{\tau_{\bf k}}{1-i\omega\tau_{\bf k}}\right)\left(\frac{\partial e_{+}(k)}{\partial k_{x}}\right)^{2 } \delta(e_{+}(k)-\mu ) + ( e_{+ } \rightarrow e_{- } ) = \frac{2e^{2}}{(2\pi)^{2}\hbar^{2 } } i_{xx}[\omega,\tau_{\bf k } ] . \label{eq : xx}\end{aligned}\ ] ] here , we have defined @xmath60 $ ] and @xmath61 $ ] for later reference . in the equation for @xmath62 , we have made the approximation @xmath63 , which is very reasonable so long as @xmath64 is large and varies smoothly . thus , the finite frequency hall angle is given by , @xmath65}{i_{xx}[\omega,\tau_{\bf k}]}. \label{eq : tanw}\ ] ] at finite frequency @xmath40 , @xmath66 becomes complex . in the limit that @xmath67 , the imaginary part can be determined without the complications@xcite of the unknown anisotropic @xmath64 . thus , @xmath68=-\frac{\omega}{\omega_h } , \label{eq : cot}\ ] ] where @xmath2 , the hall frequency , is defined as @xmath69 where @xmath70 and @xmath71 are the same as the integrals @xmath60 $ ] and @xmath61 $ ] , except that the factor @xmath72\right)$ ] is replaced by unity . the imaginary part of @xmath73 can therefore be determined in a largely model - independent manner in this limit , it is essentially a measure of fermi surface geometry @xcite while the real part of @xmath74 involves the unknown parameter @xmath64 which can depend on many details . if we had replaced @xmath75 by its @xmath54 average over the fermi surface , @xmath48 , the expression for @xmath76 $ ] would have been exactly the same as in eq . ( [ eq : cot ] ) , regardless of the magnitude of the product @xmath47,@xcite but now there would have been a dissipative real part containing @xmath48 , that is , within this approximation : @xmath77 with eq . ( [ eq : wh ] ) in hand , we can now calculate @xmath2 as a function of @xmath3 in the ddw state . we choose a representative set of values for the needed parameters . in keeping with our analysis for the related quantity , @xmath78 , @xcite we choose @xmath79 ev , @xmath80 . for a comparison with experimental data , it is necessary to choose an appropriate relation between the chemical potential @xmath17 and the doping @xmath3 . physically , this relation can be exceedingly complex in the underdoped regime , where a plethora of competing charge and spin ordered states can intervene as @xmath81 . we , therefore , do not discuss the behavior in this heavily underdoped regime , although in the past we have attempted to describe this regime by arguing that the chemical potential is perhaps pinned to zero.@xcite between the overdoped ( @xmath82 ) and the moderately underdoped regime ( @xmath83 ) , we make the simplest possible assumption that @xmath17 is a smooth function of @xmath3 . the actual function is not very significant , but to be concrete , we choose the relation implied by the band structure . to discuss the nonanalyticity close to @xmath84 , we neglect the @xmath3 dependence of the chemical potential for simplicity , as the doping dependence of @xmath24 is much more important . for illustrative purposes , we take @xmath85 \ ; \text{ev } \label{eq : w0}\ ] ] with @xmath86 . this form gives mean - field - like nonanalyticity at @xmath9 . ( the exponent @xmath87 can be replaced if , for instance , @xmath88 ising behavior is preferred . ) it is also a reasonable representation of the form suggested in ref . . we believe , however , that the final result is not strongly dependent on this particular detailed form of eq . ( [ eq : w0 ] ) . as the ddw gap is varied ( implicitly as a function of doping @xmath3 ) . proceeding clockwise from the left panel on the top , @xmath89 ; these numbers are for illustrative purposes only , and the @xmath17 was chosen so as to focus on the lower band only . ] before doing an explicit calculation , it is revealing to plot the hole pockets as the amplitude of the ddw gap @xmath24 is increased for @xmath17 varying smoothly as a function of doping @xmath3 , in particular we show the results for a _ fixed _ value of @xmath17 in fig . [ fig : cplot ] . one can see that the hole pockets become less elliptical as @xmath90 is increased at constant @xmath17 . so , even though @xmath17 is kept constant , with increasing @xmath90 below @xmath9 , the curvature of the holepockets increases where the fermi velocity is largest , @xcite consequently @xmath91 increases , and the perimeter decreases , so that @xmath92 decreases . @xcite the net result is an increase of @xmath93 . this is a robust explanation of the increase of @xmath93 as the system is underdoped in agreement with rigal _ et al._@xcite as a function of @xmath3 . the choices of the ddw gap amplitude and the chemical potential are described in the text . ] to be quantitative , we explicitly calculate @xmath93 using eq . ( [ eq : w0 ] ) and the band structure parameters given above . concomitantly , as mentioned above , @xmath17 was determined from the band structure . the results are shown in fig . [ fig : omega ] . the results are clearly consistent with the experiment of rigal _ et al._@xcite the enhancement , as the system is underdoped , is significant , even though its actual magnitude is perhaps a factor of two smaller . moreover , the absolute magnitudes are well captured . beyond this , it is difficult to compare in detail . the experiment was performed on thin films of ybco for which we do not have the precise knowledge of the doping levels , nor do we have a good criterion to relate the doping with the chemical potential . to complicate matters further , the chain contributions in ybco are not included in our calculation , and these contributions were not subtracted in their experimental results . the parameters used here are generic ; it is possible to improve the agreement with the experiment by adjusting them , but we do not find this to be a very meaningful exercise . we then explore more closely the signature of the quantum critical point in the infrared hall angle , which is already evident in fig . [ fig : omega ] . to demonstrate the robustness of the quantum critical point , we set @xmath17 to be a constant , equal to @xmath94 and use the mean field ansatz @xmath95 . for a representative value , we take @xmath96 and @xmath86 . these are identical to our previous parametrizations.@xcite the results are shown in fig . [ fig : omega2 ] . close to the quantum critical point @xmath97 , with @xmath17 set to a constant , equal to @xmath98 ev . ] by examining the integrals @xmath91 and @xmath92 in more detail , we find that , close to the critical doping , @xmath99 , where @xmath100 and @xmath101 are constants , and the slope diverges at the transition . ( at finite temperature , this will be rounded . ) we emphasize that although the hot spots determine the critical singularity close to @xmath84,@xcite the increase of @xmath93 is determined by the evolution of the hole pockets and the gapping of the fermi surface as the doping is decreased . the hall angle measurements of rigal , _ et al . _ @xcite are strong evidence for the existence of hole pockets in the underdoped cuprates . we believe that ddw order is the simplest explanation which is also consistent with the absence of hole pockets in arpes . consistency with both of these experiments ( and the hall number measurement of ref . ) is a strong challenge of other proposals for the pseudogap . our analysis opens a number of interesting directions for future research . our calculation could be extended using the kubo formalism , which would have a wider region of validity . a more careful comparison between theory and experiment could be made with a better model for the chemical potential at the precise doping levels of the experiment . finally , further exploration of the putative quantum critical point at @xmath102 at which ddw order vanishes could more firmly establish its existence and its properties . we thank h. d. drew for drawing our attention to their experiments and also for useful discussions . we are also grateful to n. p. armitage , d. basov , and s. kivelson for helpful comments . c. n. has been supported by the nsf under grant no . dmr-9983544 and the a.p . sloan foundation . s. c. has been supported by the nsf under grant no . s. t. has been supported in part by the nsf under grant no . dmr-9971138 , and in part by dmr-9971138 . j. f. was supported by the funds from the david saxon chair at ucla . 99 s. chakravarty and h. -y . kee , phys . rev . b * 61 * , 14821 ( 2000 ) . s. chakravarty , r. b. laughlin , d. k. morr , and c. nayak , phys . b * 63 * , 094503 ( 2001 ) . s. tewari , h. -y . kee , c. nayak , and s. chakravarty , phys . b * 64 * , 224516 ( 2001 ) . s. chakravarty , c. nayak , s. tewari , and x. yang , phys . lett . * 89 * , 277003 ( 2002 ) . k. hamada and d. yoshioka , phys . b * 67 * , 184503 ( 2003 ) ; s. chakravarty , c. nayak , and s. tewari , phys . rev . b * 68 * , 100504(r ) ( 2003 ) . h. -y . kee and y. b. kim phys . b * 66 * , 012505 ( 2002 ) . s. chakravarty , phys . b * 66 * , 224505 ( 2002 ) . h. -y . kee and y. b. kim , phys . b * 66 * , 052504 ( 2002 ) . s. chakravarty , h. -y . kee , c. nayak , int . b * 15 * , 2901 ( 2001 ) . s. chakravarty , h. -y . kee , c. nayak , in _ physical phenomena at high magnetic fields - iv _ , eds . g. boebinger , z. fisk , l. p. gorkov , a. lacerda , j. r. schrieffer ( world scientific , singapore , 2002 ) . h. a. mook , pengcheng dai , s. m. hayden , a. hiess , j. w. lynn , s. -h . lee , and f. dogan , phys . b * 66 * , 144513 ( 2002 ) . s. chakravarty , h. -y . kee and k. voelker , nature * 428 * , 53 ( 2004 ) . one might be concerned that our model is in contradiction with claims that the nodal fermi velocity is independent of doping in copper oxide superconductors [ x. j. zhou _ et al . _ , nature * 423 * , 398 ( 2003 ) ] . an incorrect , naive argument can be constructed to show that the fermi velocity will vary as @xmath103 with doping @xmath3 for any model with hole pockets thereby supposedly ruling out such models . if we assume that the hole pockets are circular , then it might be argued that the @xmath44 associated with the hole pocket will lead to the relation @xmath104 ; hence , the nodal fermi velocity should be given by @xmath105 , assuming that the effective mass is unchanged . that this is simply an incorrect argument can be seen from ref . . it is trivially seen that , along the nodal direction , @xmath44 seen in arpes is the same as that predicted by the band structure because of the special properties of the coherence factors.@xcite this is approximately true even away from this direction , as was verified by explicit calculations . in any case , we also do not find the experimental evidence of the doping independence of @xmath106 convincing : a collage of measurements of chemically diverse materials , in each of which @xmath107 has significant , but conflicting doping dependence , was used to argue that statistically there is no overall doping dependence . finally , in the same experimental paper mentioned in this footnote , it was also argued that the ` high - frequency fermi velocity ' diverges as @xmath108 , as the doping @xmath109 . this is not a very meaningful statement , as at higher energies there are no quasiparticles to speak of in the arpes spectra . f. f. balakirev , j. b. betts , a. migliori , s. ono , y. ando , g. s. boebinger , nature * 424 * , 912 ( 2003 ) . see , for example , j. m. ziman , _ electrons and phonons _ ( oxford university press , oxford , 1960 ) . r. peierls , _ surprises in theoretical physics _ ( princeton university press , princeton , 1979 ) ; see also j. rammer and h. smith , rev . phys . * 58 * , 323 ( 1986 ) . see , for example , m. tinkham , _ introduction to superconductivity _ , 2nd . ed . ( mc - graw - hill , new york , 1996 ) . a. carrington , a. p. mackenzie , c. t. lin , and j. r. cooper , phys . * 69 * , 2855 ( 1992 ) ; h. kontani , k. kanki , and k. ueda , phys . b * 59 * , 14723 ( 1999 ) . e. abrahams and c. m. varma , phys . b * 68 * , 094502 ( 2003 ) k. segawa and y. ando , phys . rev . lett . * 86 * , 4907 ( 2001 ) . y. ando , y. kurita , s. komiya , s. ono , and k. segawa , preprint . p. w. anderson , phys . lett . * 67 * , 2092 ( 1991 ) ; p. coleman , a. schofield , and a. m. tsvelik , phys . rev . lett . * 76 * , 1324 ( 1996 ) ; g. kotliar , a. sengupta , and c. m. varma , phys . b * 53 * , 3573 ( 1996 ) . see , for example , t. p. beaulac , f. j. pinski , and p. b. allen , phys . b * 23 * , 3617 ( 1981 ) . it might be argued from the one - component analysis of optical measurements that @xmath110 for energies above the pseudogap scale [ for a review , see d. b. tanner and t. timusk , in _ physical properties of high temperature superconductors iii _ , d. m. ginsberg editor , ( world scientific , singapore , 1992 ) ] , which is consistent with the marginal fermi liquid phenemenology.@xcite we find that for underdoped materials , relevant to the present discussion , a better description may be in terms of a two - component analysis , as described in the same review article mentioned in this footnote . in a two - component analysis , where a drude part and a mid - infrared part are separated out , there are no such restrictions on @xmath47 . the appropriateness of the two - component analysis in the underdoped regime has been emphasized recently [ w. j. padilla _ et al . _ , preprint , submitted to phys . moreover , the hall scattering rate @xmath111 and the transport scattering rate @xmath48 behave differently , as emphasized in ref . , and the former does not obey this inequality . these complications are beyond the scope of the present paper .	infrared hall measurements in the pseudogap phase of the high-@xmath0 cuprates are addressed within the framework of the ordered @xmath1-density wave state . the zero - temperature hall frequency @xmath2 is computed as a function of the hole - doping @xmath3 . our results are consistent with recent experiments in absolute units . we also discuss the signature of the quantum critical point in the hall frequency at a critical doping inside the superconducting dome , which can be tested in future experiments .
schematic of an interferometer which implements the 2-qubit deutsch - jozsa algorithm ( a ) . all beam splitters are 50/50 . with beam splitters c1 and c2 in place , the standard algorithm ( b ) is performed . in this work , we show that an alternate encoding ( c ) is preferable in the presence of random noise as indicated on rails 2 and 3 ; replacing beam splitters c1 and c2 with beam splitters d1 and d2 implements this modified algorithm . fig 2 . experimental setup . variable - reflectivity beam splitters are implemented using a pair of polarizing beam splitters ( pbs ) and a half waveplate . the `` preparation '' portion of the interferometer produces the same superposition as the pair of hadamards in fig . the oracle consists of two variable beam splitters which can each be set to swap two rails or leave them unchanged , plus two half - waveplate used to induce @xmath14 phase shifts on two of the outputs . the random noise is generated by inducing turbulent airflow under rails 2 and 3 while they are spatially superposed . fig 3 . experimental data : normalized intensity is a measure of the fraction of photons reaching each detector , pd1 through pd4 . data are shown for both the dfs and standard encodings , for each of the four oracles ( 00 , 01 , 10 , and 11 ) ; c indicates `` constant '' oracles while b indicates `` balanced '' oracles . the bottom plot shows the same data in the presence of noise . note that the noise has a much more significant effect in the case of the standard encoding . the probability of the algorithm returning a 0 or a 1 for each of the oracles , in each encoding , with and without the addition of phase noise . the data are extracted by summing the normalized intensities from fig . 3 for pd1 and pd2 ( dashed line , indicating constant oracles ) and for pd3 and pd4 ( solid line , indicating balanced oracles ) . note that the success rate is close to 1 even in the presence of noise when the dfs encoding is used .	for a practical quantum computer to operate , it will be essential to properly manage decoherence . one important technique for doing this is the use of `` decoherence - free subspaces '' ( dfss ) , which have recently been demonstrated . here we present the first use of dfss to improve the performance of a quantum algorithm . an optical implementation of the deutsch - jozsa algorithm can be made insensitive to a particular class of phase noise by encoding information in the appropriate subspaces ; we observe a reduction of the error rate from 35% to essentially its pre - noise value of 8% . [ theorem]acknowledgement [ theorem]algorithm [ theorem]axiom [ theorem]claim [ theorem]conclusion [ theorem]condition [ theorem]conjecture [ theorem]corollary [ theorem]criterion [ theorem]definition [ theorem]example [ theorem]exercise [ theorem]lemma [ theorem]notation [ theorem]problem [ theorem]proposition [ theorem]remark [ theorem]solution [ theorem]summary one of the great stumbling blocks to building quantum computers , with their oft - touted ability to resolve certain problems more efficiently than any classical algorithm @xcite is the ubiquity of decoherence . coupling of any element of a quantum computer to an environment destroys its unitary evolution , and introduces uncontrollable noise ; at first , it was thought by many @xcite that these errors would make quantum computation impossible in practice . since then , a variety of techniques for correcting errors and/or building in immunity to certain classes of decoherence have been developed @xcite and it has been proved that if errors are kept below a certain constant threshold , arbitrarily large quantum computers are possible@xcite . one important technique involves computing within certain subspaces of the full system s hilbert space known as decoherence - free subspaces ( dfss ) @xcite which remain unaffected by the interaction with the environment . such dfss exist when the interaction hamiltonian has an appropriate symmetry property . dfss have been demonstrated in a linear - optical experiment @xcite and in nmr @xcite and recently to help circumvent the technical noise which had previously limited the operation of ion - trap quantum computers @xcite . to date , no demonstration has been made of the usefulness of dfss in the context of the implementation of an actual quantum - computing algorithm . in this paper , we present a linear - optical implementation of the two - qubit deutsch - jozsa algorithm dj92,nchuang , and demonstrate that when a certain class of noise is introduced into the system , greatly increasing the error rate of the algorithm , it is possible to ` encode ' one logical qubit into two physical qubits and take advantage of dfss , reducing the error rate to close to zero . linear optics is well known to be an extremely powerful arena for the transportation and manipulation of quantum information , benbrassard , reckzeilinger . although it is also well known that due to the linearity of optics , this arena does not allow for scalable construction of quantum gates @xcite , recent work has shown that the incorporation of detection and post - selection may in fact render all - optical quantum computers an attractive possibility @xcite . work also proceeds on development of nonlinearities which would allow for the development of natural two - qubit gates in optics@xcite . while we do not yet have access to a truly scalable optical quantum - computer architecture , many of the elements of any such system would be identical to those used in simple linear - optical geometries @xcite . for this reason , linear optics remains an important domain for the study of quantum coherence and error correction , even while the ultimate fate of optical quantum computing is uncertain . recently , striking demonstrations of quantum search algorithms @xcite have been carried out in linear - optical systems , as has the first verification of dfss @xcite . additionally , it is already clear that even if quantum computation never becomes truly practical , quantum information processing may have a great effect on the practice of communications and cryptography@xcite . although some information - processing will be necessary in this area as well , the question of scalability is not crucial , and linear - optical quantum computation could well prove applicable for elements such as quantum repeaters repeatersbriegel . in this context , we have chosen to study the applicability of dfss to a linear - optical implementation of the quantum deutsch - jozsa algorithm , despite the non - scalable nature of the present architecture . the deutsch - jozsa algorithm is designed to distinguish between two classes of functions ( `` oracles '' ) on n - bit binary inputs . `` constant '' functions return the same value ( 0 or 1 ) for all @xmath0 possible inputs , while `` balanced '' functions return 0 for half the possible inputs and 1 for the other half . clearly , a classical algorithm would on some occasions require as many as @xmath1 queries to unambiguously determine to which class a given oracle belongs . by contrast , deutsch and jozsa showed @xcite that a quantum algorithm requires only one such query . in the 2-qubit deutsch - jozsa algorithm @xcite , the oracle is a function on a single bit . it takes as input a query bit @xmath2 and a signal bit @xmath3 ; its action is to perform the unitary mapping @xmath4 _ _ @xmath5 _ _ @xmath6 to perform the algorithm , the input is prepared in @xmath7\otimes \lbrack \left\| 0\right\rangle -\left\| 1\right\rangle ] $ ] which the oracle maps to @xmath8=\frac{1}{2}[\left\| 0\right\rangle e^{i\pi f(0)}+\left\| 1\right\rangle e^{i\pi f(1)}]\otimes h\left\| 1\right\rangle $ ] . a hadamard on the query qubit then transforms it into @xmath9 which is equal to 0 for constant and 1 for balanced functions . thus measurement in the computational basis allows one to determine a global property of @xmath10 , namely @xmath11 in a single evaluation of the function . furthermore , the signal qubit is in fact superfluous after the oracle @xcite . thus only one logical qubit is needed after the operation of the oracle . if some source of decoherence is present during the propagation from the oracle to the final hadamard , one may consider encoding this logical qubit in some decoherence - free subspace of the two physical qubits . in this experiment we represent the four basis states of two logical qubits ( @xmath12 and @xmath13 , where the first bit corresponds to the query and the second to the signal ) by a photon traveling down one of four optical rails numbered 1 , 2 , 3 and 4 respectively . it is possible to implement a universal set of one- and two - qubit operations in a four - rail representation @xcite . for example a not gate on the query qubit can be realized by simultaneously swapping rails 1 and 3 and rails 2 and 4 . a cnot gate on the signal qubit is implemented by swapping rails 3 and 4 . to perform a hadamard gate on the query qubit , we combine rails 1 and 3 and rails 2 and 4 at two 50/50 beam - splitters ; a @xmath14 phase shift is also needed on two of the arms.analogous gates can be constructed for the other qubit . the transformations introduced by the four possible functions can also be implemented in this representation by four different settings of an oracle operating as follows : if @xmath15 is 1 , rails 1 and 2 are swapped ; if @xmath16 is 1 , rails 3 and 4 are swapped . thus the task of distinguishing balanced from constant oracles reduces to that of determining whether the number of swaps was odd or even . the schematic diagram of the interferometer is shown in fig . 1 . each photon is sent along rail 2 corresponding to the logical state @xmath17 . the two pairs of 50/50 beam splitters @xmath18 , @xmath19and @xmath20 , @xmath21 implement the two hadamard gates on the query and signal qubits respectively , preparing the qubits for the oracle s action . the last two 50/50 beam splitters @xmath22 and @xmath23 realize the hadamard gate on the query qubit after the oracle . rails 1 - 4 illuminate photodiodes @xmath24 . a photon reaching pd1 or pd2 indicates that the value of the query qubit after the algorithm , @xmath25 , is 0 . this constitutes a determination that the oracle is constant , while the other two detectors indicate balanced oracles . one source of decoherence in such systems is the phase noise introduced by fluctuating optical path lengths between the different sections of the apparatus , created either by variations in distance or by temperature variations and turbulent air flow . in real optical systems , the stability of certain path - length differences may be larger than that of others , either because of the physical proximity of certain paths to one another or because of the particular sources of mechanical or thermal noise . this may lead to a situation where the dominant source of decoherence exhibits a particular symmetry which can be exploited for computing within dfss . to simulate the effects such processes could have in larger - scale , distributed quantum - information systems , we introduced a high degree of turbulence by placing the tip of a hot soldering iron below two of the optical rails . these two optical paths ( rails 2 and 3 ) were spatially superposed in this region , distinguished only by their polarisation ; for this reason , they experienced essentially the same random phase shifts under the influence of the turbulent air flow , relative to the other optical rails . since the outputs of the optical deutsch - jozsa setup are the outputs of two parallel interferometers , which measure the phase of rail 2 with respect to that of rail 4 and rail 3 with respect to rail 1 , this phase noise destroys the interference on which the success of the algorithm relies . on the other hand , inspection of the optical schematic makes the physical process behind the algorithm evident : rails 1 and 3 are prepared in phase with one another , while rails 2 and 4 are also prepared in phase , but @xmath26out of phase with the former pair . thus , constructive interference is observed either between 1 and 3 or between 2 and 4 . if a single pair ( 1 and 2 or 3 and 4 ) is swapped by the oracle , destructive interference is instead observed at both interferometers , while if an even number of swaps occurs , constructive interference is restored . in other words , so long as each interferometer compares an output of each of the potential swap regions in the oracle with one from the other , it is possible to distinguish a balanced oracle ( one swap ) from a constant oracle ( zero or two swaps ) . the strategy to deal with phase noise impressed symmetrically on paths 2 and 3 now becomes clear : instead of interfering 2 with 4 and 1 with 3 , one can accomplish the same task by interfering 2 with 3 and 1 with 4 . in this way , the random phase appears at both inputs to the same interferometer , and has no effect on the measured results . this modification can be expressed as an encoding of the data into a pair of dfss . since our engineered phase noise has identical effects on the two states of odd parity ( @xmath27 and @xmath28 stored on rails 2 and 3 respectively ) and on the two states of even parity ( @xmath29 and @xmath13 , stored on rails 1 and 4 ) , each fixed - parity subspace can store a single logical qubit in a decoherence - free fashion . the action of the soldering iron tip may be modelled by the evolution operator @xmath30 , where @xmath31 is a random , fluctuating phase . in a subspace with a definite eigenvalue of @xmath32 , the random phase , @xmath33 does nothing but impress an overall global phase on the quantum state , leaving the information within the subspace unaffected . since the 2-qubit deutsch - jozsa algorithm relies on a single qubit ( query qubit ) after the oracle has completed its action , this single qubit may be encoded in either of these dfss , providing immunity to parity - dependent phase noise which occurs between the oracle and the final hadamard gates . as shown in fig 1c , a cnot after the oracle encodes the query qubit into these dfss , and a second cnot after the final hadamard can be used for decoding . the decoding cnot is unnecessary since measurements are only performed on the query qubit . swapping rails 3 and 4 performs the encoding ; or equivalently , beam - splitters @xmath22 and @xmath23 may be replaced by @xmath34 and @xmath35 . the actual experimental setup is shown in fig . 2 . the light source was a diode laser operating at 780 nm . to implement the four different oracle settings a specific kind of variable beam splitter ( vbs ) was designed . this variable beam splitter consists of a half - waveplate between two polarizing beam splitters ( pbs ) ; any desired reflectivity can be obtained with this optical arrangement . for realizing our oracles , a pair of these vbss was used , and each was adjusted either for maximum or minimum reflectivity , essentially acting as a swap or the identity . hadamards were constructed using similar vbss . after the oracle rails 2 and 3 were combined into the same spatial mode in a pbs to guarantee the collective phase shift for these beams in presence of decoherence , and then separated out by another pbs . the transformation between two different encodings was performed by applying another vbs to either swap rails 3 and 4 or not . the experimental setup was designed such that in all of these interferometers the spatial path lengths are always balanced . the average fringe visibility for all four output ports and all possible settings of oracle and encoding was measured to be about 95% . this setup consists of 16 different possible mach - zehnder interferometers , of which two are in operation for any given oracle / encoding combination . the experiment was performed by measuring the signals at detectors @xmath36 through @xmath37 as the half waveplates were adjusted to cycle through all four oracles and both encodings . the intensities at detectors pd1 through pd4 were normalized to their sum , to yield the probabilities of a photon reaching each of the detectors . these normalized intensities are plotted in fig . 3 for all 4 oracle settings , in both the standard algorithm and the dfs encoding . ideally , all the photons should arrive at detectors @xmath36 and @xmath38 for constant functions and at detectors @xmath39 and @xmath37 for balanced functions . in figure 4 , we plot the probability of a photon reaching _ either _ pd1 or pd2 , and the probability of a photon reaching _ either _ pd3 or pd4 . the average error rates were measured to be about 8% in the absence of added noise . the sources of errors in this experiment were mostly due to imperfect visibility , ( due to alignment and waveplate setting ) , and uncertainty and drift in the optical phases setting , when a @xmath40 phase error on one beam correspond to a 2% error rate . the drift of the interferometer during measurement was kept low by balancing all path lengths as mentioned above and enclosing the interferometer . introduction of the turbulent airflow increased the average error rates to 35% for the standard algorithm . when the dfs encoding was used in the presence of turbulence , however , the error rates dropped to 7% , essentially equal to the value in the absence of noise . this work demonstrates that a simple modification of a quantum algorithm may be used to encode information into dfss , and significantly reduce the error rate introduced by realistic , physical noise sources , provided that these sources have certain symmetry properties . phase noise is an everpresent issue in coherent optical systems , and often exhibits certain correlations which should be exploitable in this manner . we also note that since the noise characteristics are intimately tied to the particular physical realization of a quantum circuit , it may often prove easier to design the decoherence - free process by direct consideration of the multiple interferometers which constitute the optical device , than by contemplation of the very general quantum circuits . this work was supported by the us air force office of scientific research ( f49620 - 01 - 1 - 0468 ) , nserc , and photonics research ontario . we would like to acknowledge useful discussions with daniel lidar and also thank guillaume foucaud and chris ellenor for technical assistance . r. feynman , _ int . j. theor . phys . _ * 21 * , 467 ( 1982 ) ; d. deutsch , _ proc . r. soc . london . ser a_. * 400 * , 1818 , 97 ( 1985 ) ; p. w. shor , in _ proceedings of the 35th annual symposium on foundations of computer science _ , s. goldwasser , ed . ( ieee computer society , los alamitos , ca , 1994 ) , p. 124 ; l. k. grover , phys . rev . lett . * 79 * , 325 ( 1997 ) . m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _ , cambridge university press ( 2000 ) . r. landauer , _ philos . t. roy . soc . a_. * 353 * , 367 ( 1995 ) . p.w . shor , phys . rev . a * 52 * , 2493 ( 1995 ) ; e. knill , r. laflamme , phys . rev . a * 55 * , 900 ( 1997 ) ; d. gottesman , phys . rev . a * 54 * , 1862 ( 1996 ) l .- m . duan , g .- c . guo , phys . rev . lett . * 79 * , 1953 ( 1997 ) ; p. zanardi , m. rasetti , phys . rev . lett . * 79 * , 3306 ( 1997 ) d. a. lidar , i. l. chuang , k. b. whaley , phys . rev . lett . * 81 * , 2594 ( 1998 ) ; j. kempe , _ et al . _ , phys . rev . a. * 63 * , 042307 ( 2001 ) ; d. bacon , _ et al . _ , phys . rev . lett . * 85 * , 1758 ( 2002 ) . e knill , r laflamme , wh . zurek , science , vol * 279 * , 342 ( 1998 ) . p.g . kwiat , a.j . berglund , j.b . altepeter , a.g white , science , vol * 290 * , 498 ( 2000 ) . em . fortunato , _ et al . _ new . j. phys * 4 * , 5 ( 2002 ) . d. deutsch , r. jozsa , _ proc . r. soc . london a _ , * 439 * , 553 ( 1992 ) . d. kielpinski , _ et al_. , science , vol * 291 * , 1013 ( 2001 ) . c. h. bennett and g. brassard , in _ proceedings of the ieee international conference on computers , systems & signal processing , bangalore , india _ ( ieee , new york , 1984 ) , p. 175 m. reck , _ et al . _ , phys . rev . lett . * 73 * , 58 ( 1994 ) . n.j . cerf , c. adami , p.g . kwiat , phys . rev . a , * 57 * , r1477 ( 1998 ) . e.knill , r.laflamme , and g.milburn , nature , vol * 409 * ( 2001 ) . d. gottesman , a. kitaev , and j. preskill , phys . rev . a , * 64 * , 012310 ( 2001 ) . q. a. turchette , _ et al . _ , phys . rev . let . * 75 * , 4710 ( 1995 ) ; a. rauschenbeutel , _ et al . _ , phys . rev . lett . * 83 * , 5166 ( 1999 ) ; j.d . franson , phys . rev . lett , * 78 * , 3852 ( 1997 ) , e. harris and l.v . hau , phys . rev . lett . * 82 * , 4611 ( 1999 ) ; k.j . resch , j.s . lundeen , and a.m. steinberg , phys . rev . lett * 87 * , 123603 ( 2001 ) ; k.j . resch , j.s . lundeen , and a.m. steinberg , phys . rev . lett . * 89 * , 037904 ( 2002 ) . j. ahn , t.c . weinacht , p.h . bucksbaum , science , vol * 287 * , 5452 ( 2000 ) . n bhattacharya , hbv van den heuvell , rjc spreeuw , phys . rev . let . * 88 , * 137901 ( 2002 ) . d. bouwmeester , _ et al_. , nature , * 390 * , 575 ( 1997 ) . j. brendel , _ et al_. , phys . rev . lett . * 82 * , 2594 ( 1999 ) . h .- j . briegel , _ et al_. , phys . rev . lett . * 81 * , 5932 ( 1998 ) . d. collins , k.w . kim , w.c . holton , phys . rev . a * 58 * , 1050 ( 1998 ) .
the interstellar medium ( ism ) of elliptical galaxies is thought to be considered to be an accumulation of stellar mass loss and supernova ejecta . while standard supernova rates predict the ism to have a high metallicity of several times the solar value ( e.g. @xcite ; @xcite ) , the observed ism metallicity in fact fell below half a solar , even lower than the stellar metallicity ( e.g. @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . this strong discrepancy calls into question our current understanding of supernova enrichment and chemical evolution of galaxies . ngc 4636 is one of the relatively isolated yet most luminous elliptical galaxies , both in optical ( @xmath7=@xmath8 assuming a distance of 17 mpc ) and in x - ray ( @xmath9= 3.8@xmath10 ; _ einstein _ ) band ( e.g. @xcite ; @xcite ) . using rosat , trinchieri et al . ( 1994 ) discovered a very extended x - ray emission surrounding this galaxy , out to @xmath11 . the first asca observation of ngc 4636 in the pv ( performance verification ) phase yielded the very low ism metallicity ( @xcite ) , together with abundance and temperature gradients ( @xcite ) . we have re - observed ngc 4636 with asca ( @xcite ) for an extremely long time ; over 200 ks . this has allowed matsushita ( 1997 ) to perform a much deeper study of the extended x - ray component after rosat . in this paper , we utilize the overall asca data including this long exposure , to study the spectral properties of ngc 4636 and look into the abundance problem . ngc 4636 has been observed from asca twice . following the pv observation ( 1993 july 22 , with the sis in 4ccd mode ) , the second much longer observation was conducted in the ao-4 phase from 1995 december 28 through 1996 january 4 , with the sis in the 1ccd mode . we discarded the data taken under cut - off rigidities less than 6 gev c@xmath12 , or elevation angle less than @xmath13 and @xmath14 from night and day earth respectively . this has yielded exposure times of 36 ks ( with the gis ) and 39 ks ( with the sis ) for the pv observation , and those for the ao-4 observation are 172 ks ( gis ) and 215 ks ( sis ) . we accumulated on - source spectra within 4 times the effective radius , @xmath15 ( @xcite ) , centered on ngc 4636 . since the sis response has changed with time , we treat the sis spectra from the two observations separately . the background spectrum was obtained by integrating the blank - sky data over the same region of the detector . figure [ fig1 ] shows the background - subtracted sis spectrum for the ao-4 data . we jointly fit the two sis ( pv and ao-4 ) spectra and one gis spectrum , with a standard two component model ( @xcite ; @xcite ; @xcite ) . the model consists of a thin thermal emission from the ism with free temperature @xmath16 and free metallicity , and a thermal bremsstrahlung with temperature fixed at 10 kev representing the contribution from low - mass x - ray binaries . both components are subjected to a common interstellar absorption @xmath17 . although the gis ( @xcite ; @xcite ) is less sensitive than the sis to the low - energy ( e.g. @xmath18 1 kev ) atomic lines , it can constrain the hard component better than the sis , and is fully sensitive to the si - k and s - k lines . therefore , the joint use of the two instruments is essential . in this paper we adopt for the solar iron abundance the ` meteoritic ' value , fe / h@xmath19 by number ( @xcite ) as the first - cut spectral analysis , we represented the ism component by the plasma emission model of raymond & smith ( 1977 ; hereafter r - s model ) with solar abundance ratios . the best - fit model parameters turned out to be consistent with those of awaki et al . ( 1994 ) , but the fit was totally unacceptable ( table 1 ) . we then allowed the abundances to deviate from the solar ratios , by dividing heavy elements into two groups to estimate relative contributions from type ia and type ii supernovae ( sne ) . one group consists of so - called @xmath1-elements , o , ne , mg , si and s , which are assumed to have a common abundance @xmath20 . the other group includes fe and ni , with a common abundance @xmath21 . abundance of he is fixed to be 1 solar . the abundances of the other elements are assumed to be the same as @xmath1-elements , although their effect on @xmath22 kev spectrum is negligible . the fit incorporating 6 parameters ( @xmath23 , @xmath21 , @xmath16 , @xmath17 , and two normalizations ) still remained far from acceptable , as shown in table 1 . we further replaced the r - s model with meka model ( @xcite ; @xcite ; @xcite ) , mekal model ( @xcite ) , or masai model ( @xcite ) , but none of them were successful . multi - temperature ism models did not improve the fit , either . in these fits , the data - to - model discrepancy is always largest around the fe - l complex region ( 0.8 @xmath24 kev ) . in addition , the fit residuals and the derived physical quantities both depend significantly on the plasma emission codes . these reconfirm the serious problems in the theoretical fe - l line modeling , as pointed out by fabian et al . ( 1994 ) and arimoto et al . furthermore , we have found strong false couplings between @xmath20 and @xmath21 , arising from the following two effects . on one hand , the fitting algorithm tries to reduce the fe - l discrepancy , by adjusting intensities of the o - k and ne - k lines which overlap the fe - l complex ; this strongly affects @xmath20 , since the data have the highest statistics in this energy range . in turn @xmath20 affects @xmath25 , because the bound - free emission from oxygen and neon acts nearly as a continuum to the fe - l lines and controls their equivalent widths . these effects make both @xmath21 and @xmath23 highly unreliable . in order to avoid these problems , we tentatively restricted the energy range for the sis spectral fit to @xmath26 kev which is unaffected by the fe - l , o - k and ne - k lines . for the gis spectrum , we used an energy range above 1.6 kev , because of its poorer energy resolution . we re - arranged heavy elements into 4 groups ; mg ; si ; s ; and other elements including o , ne , fe and ni ( hereafter @xmath27 ) . elements in the last group contribute to the @xmath28 kev spectrum mainly via free - bound continua ( fe - k and ni - k lines are undetectable ) . then , the fit has become acceptable with the reduced @xmath29 of 1.06 for 220 dof .. the derived temperature , 0.73 kev ( 0.670.76 kev with 90% confidence for a single parameter ) , agrees with those derived using the whole energy range . we have further found that the mg to si abundance ratio always agrees with the solar ratio ( fig.2 ) , even though their abundances correlate considerably with @xmath27 . the s abundance may be similar , or slightly higher by about 30 % . on the other hand , the stellar and supernova nucleosysnthesis predicts o , ne and mg to follow the solar abundance ratios ( @xcite ) . therefore , it is reasonable to assume that all the @xmath1-elements ( o , ne , mg , si and s ) have the same abundance in solar units . we can now safely return to the original element grouping into @xmath23 and @xmath21 , and refit the spectra above 1.6 kev ( sis ) and 1.7 kev ( gis ) . the fit is acceptable with the reduced @xmath29 of 1.06 for 210 dof . , and the derived temperature is 0.75 ( 0.720.78 ) kev . now , @xmath23 is determined relatively well ( fig.3 ) , because it relies upon the si - k and s - k lines instead of using the unreliable o - k and ne - k lines that are now ignored . unlike the full - range fit , @xmath23 depends on @xmath21 only weakly through fe contribution to the free - bound continuum . these results unambiguously show that @xmath23 is roughly one solar , unless @xmath21 is extremely high . our next task is to better constrain @xmath25 . this requires us to utilize the fe - l line information , but simply reviving the energy range below 1.6 kev would bring us back to the original problems described in 3.1 . in order to relax the too strong constraints imposed by the data in the fe - l region , we have decided to assign 20% systematic errors to each spectral data in the 0.41.6 kev range . this is because the data - to - model discrepancy and the model - to - model differences are both @xmath30 20 % around the fe - l region ( @xcite ) . we have thus fitted the sis / gis spectra jointly with the two - component model , incorporating the increased systematic error and again employing the element grouping into @xmath23 and @xmath25 . the fit is acceptable with either of the plasma models ( table 1 ) , and the confidence regions on the @xmath23 vs. @xmath21 plane for different plasma emission codes have come to a similar region ( fig.3 ) . as a consequence , we can now conclude with confidence that @xmath23 and @xmath21 are both consistent with @xmath30 1 solar , within a factor of 2 . in addition , the ratio between @xmath23 to @xmath21 coincides with the solar ratio within @xmath31% . in order to study the metal distribution , we accumulated the gis and sis spectra in three concentric annular regions with radius range , 04@xmath32 , 4@xmath328@xmath32 , and 8@xmath3212@xmath32 centered on the galaxy . we assume that the ism is spherically symmetric . this `` ring - sorted analysis '' is however limited by the spectral mixing effect among different sky regions , due to the extended point - spread - function of the asca x - ray telescope ( xrt ; @xcite ) . this tends to smear out any radial gradient in the ism properties . following the analysis in 3.3 , we include 20% systematic error in the 0.4 - 1.6 kev data . we fit only with the r - s model , since the difference due to plasma emission code is only several tens of % after inclusion of the systematic error . an acceptable fit is obtained in all three regions ( table 2 ) . as shown in table 2 and fig.4 , the derived @xmath23 and @xmath21 are @xmath2 1 solar within @xmath33 , and drop to @xmath34 solar at @xmath35 . this confirms the existence of strong abundance gradient in the ism in ngc 4636 , and reveals that both @xmath1-elements and fe equally show the abundance gradient . their ratio is close to the solar value at all radii . the absolute values of the abundances are higher than those derived in mushotzky et al . ( 1994 ) , who assumed a solar ratio in the metals , but the temperatures and hydrogen column densities are consistent . by carefully examining the abundance ratios and uncertainties in the fe - l complex , we have concluded that the ism abundances of fe and @xmath1-elements in the x - ray luminous galaxy , ngc 4636 , are in fact as high as about 1 solar within @xmath33 . the abundance ratio between @xmath23 and @xmath21 is very close to the solar ratio within @xmath31% . this means that both type ia and type ii sne ejecta make significant contribution in the ism . the ism iron abundance is expected to be given by @xmath36 ( @xcite ; @xcite ) , where @xmath37 is the average iron abundance of the stars in units of the solar abundance , @xmath38 the hubble constant in units of 50 km s@xmath39 , @xmath40 is the present rate of sn ia in ellipticals in units of the rate as estimated by tammann ( 1982 ) , and @xmath41 is the iron yield per sn ia event . the stellar fe abundance , @xmath37 , averaged over the whole galaxy is 0.74 solar ( @xcite ) . since our revised value of @xmath21 is fairly close to these stellar values , the previously - quoted severe problem that the ism abundance was lower than the stellar metallicity has been mostly removed . we can solve for the relative contribution from sn ia and sn ii on @xmath1-elements and fe . if we assume that sn ii products have 3 times higher value of @xmath23 than @xmath21 ( @xcite ) and sn ia produces only fe , the sn ii contribution necessary to produce the observed @xmath23 ( 1 solar ) can enrich fe to @xmath21 = 1/3 solar . the rest of @xmath21 , i.e. 2/3 of the best - fit fe abundance , has to be explained by sn ia . using equation ( 1 ) , we can estimate the sn ia rate @xmath42 . in order to derive an upper limit for @xmath42 , let us assume that stars contain no fe from sn ia . this gives an equation only for the sn ia contribution , @xmath43 . the left hand side is now estimated to be @xmath2 0.7 solar , so that the condition for @xmath42 is @xmath44 . therefore , the present abundance result suggests that sn ia rate would be less than 1/7 of tammann s rate for @xmath45 . considering the uncertainties in @xmath23 and @xmath21 , this is consistent with the recent estimation of @xmath46 by capperallo et al . ( 1993 ) based on the supernova counting . a systematic study of the ism abundance involving many galaxies would be neccessary to further look into this problem . also , fe - l diagnostics need to be closely examined in order to reach consistent answers among the models ( @xcite ) . we have confirmed and refined the strong abundance gradient feature in the ism discovered by mushotzky et al . in addition , we have for the first time discovered strong abundance gradient in @xmath1-elements , which is very similar to that of @xmath21 . there may be an internal metallicity gradient in the ism within the galaxy , as suggested from the significant gradient in the stellar metallicity within @xmath47 ( @xcite ) . however , no information is available about the stellar metallicity in larger scales . there exists an extended x - ray emission around ngc 4636 with a radius @xmath48 100 kpc , as discovered by trinchieri et al . ( 1994 ) , and confirmed with asca , by matsushita ( 1997 ) . since the hot gas responsible for the extended emission has a very low metallicity as seen in the data in r=812@xmath32 , the abundance gradient can be explained in terms of the growing contribution from the metal poor gas in the outer region . the similarity in the slope of of @xmath23 and @xmath21 gradient suggests common dilution due to external gas . dum anders , e. , & grevesse , n. , 1989 , , 53,197 arimoto , n. , matsushita , k. , ishimaru , y. , ohashi , t. , & renzini , a. 1997 , , 477 , 128 awaki , h. et al . 1994 , , 46 , l65 cappellaro , e. , turatto , m. , benetti , s. , tsvetkov , d. yu . , bartunov , o. s. , & makarova , i. n. 1993 , , 273 , 383 davies , r.l . , sadler , e.m . , peletier , r.f . 1993 , , 262 , 650 faber , s.m . , wegner , g. , burstein , d. , davies , r.l . , dresller , a. , lynden - bell , d. , terlevich , r.j . 1989 , 69,763 fabian , a.c . , arnaud , k.a . , bautz , m.w . , & tawara , y. 1994 , , 436 , l63 fabbiano , g. , kim , d .- w . , & trinchieri , g. , 80 , 531 forman , w. , jones , c. , & tucker , w. 1985 , , 293 , 102 kaastra , j.s . 1992 , an x - ray spectral code for optically thin plasmas ( internal sron - leiden report , updated version 2.0 ) liedahl , d. a. , osterheld , a.l . , & goldstein , w.h . 1995 , , 438 , l115 loewenstein , m. , & mathews , w.g . 1991 , , 373 , 445 makishima , k. et al . 1996 , , 48 , 171 masai , k. 1984 , , 98 , 367 masai , k. 1997 , , in press matsumoto , h. , koyama , k. , awaki , h. , tsuru , t. , loewenstein , m. , & matsushita , k. 1997 , , 482 , 133 matsushita , k. et al . 1994 , , 436 , l41 matsushita , k. 1997 , ph.d . thesis , university of tokyo mewe , r. , gronenschild , e.h.b.m . , & van den oord , g.h.j . 1985 , a&as , 62,197 mewe , r. , lemen , j.r . , & van den oord , g.h.j . 1986 , a&as , 65,511 mushotzky , r.f . , loewenstein , m. , awaki , h. , makishima , k. , matsushita , k. , & matsumoto , h. 1994 , , 436 , l79 ohashi , t. et al . 1996 , , 48 , 157 raymond , j.c . , & smith , b.w . 1977 , , 35 , 419 renzini , a. , ciotti , l. , dercole , a. , & pellegrini , s. 1993 , , 419 , 52 serlemitsos , p.j . , 1995 , , 47 , 105 tammann , g. 1982 , in supernovae : a survey of current research , ed . m. rees & r. stoneham ( dordrecht : reidel ) , p. 371 tanaka , y. , inoue , h. , & holt , s. 1994 , , 46 , l37 thielemann , f - k . , nomoto , k. , & hashimoto , m. 1996 , , 460 , 408 trinchieri , g. , kim , d .- w . , fabbiano , g. , & canizares , c.r . 1994 , , 428 , 555 lllllr code & @xmath16 & @xmath23 & @xmath21 & @xmath49 & @xmath50 + & ( kev ) & ( solar ) & ( solar ) & ( @xmath51 ) & + r - s & 0.76 & 0.48 & @xmath52@xmath23 & 5 & 1112/373 + r - s & 0.75 & 0.54 & 0.50 & 4 & 1093/372 + meka & 0.59 & 1.66 & 1.49 & 9 & 1235/372 + mekal & 0.66 & 0.76 & 0.71 & 10 & 727.5/372 + masai & 0.74 & 1.50 & 1.68 & 1 & 786.6/372 + + r - s & 0.76 @xmath53 & 0.95 @xmath54 & 0.75 @xmath55 & 1 @xmath56 & 223.0/330 + meka & 0.63 @xmath53 & 1.10 @xmath57 & 1.12 @xmath58 & 6 @xmath59 & 337.4/330 + mekal & 0.72 @xmath60 & 1.13 @xmath61 & 0.88 @xmath62 & 4 @xmath63 & 247.4/330 + masai & 0.72 @xmath53 & 1.17 @xmath64 & 1.33 @xmath65 & 2 @xmath66 & 271.5/330 + cllllr r & @xmath16 & @xmath23 & @xmath21 & @xmath49 & @xmath50 + ( arcmin ) & ( kev ) & ( solar ) & ( solar ) & ( @xmath51 ) & + 04 & 0.75 @xmath67&1.19 @xmath68 & 0.93 @xmath69 & 2 @xmath70 & 182.5/264 + 48 & 0.81 @xmath71&0.53 @xmath72 & 0.48 @xmath73 & 1 @xmath74 & 202.3/259 + 8 - 12 & 0.77 @xmath75&0.37 @xmath76 & 0.25 @xmath77 & 0 @xmath78 & 149.0/194 +	high quality x - ray spectra of ngc 4636 are obtained with asca . theoretical models are found unable to reproduce the data in the fe - l line region . spectral data above 1.4 kev indicate that mg to si abundance ratio is @xmath0 solar . assuming that the abundance ratios among @xmath1-elements are the same with the solar ratios , spectral fit with increased systematic error in the 0.41.6 kev range gives abundances of @xmath1-elements and fe to be both @xmath2 1 solar by taking one solar of fe to be 3.24@xmath3 by number . these new abundance results solve discrepancy between stellar and hot - gas metallicity , but still a low supernova rate is implied . we also detect strong abundance gradients for both @xmath1-elements and iron in a similar fashion . the abundance is @xmath0 solar within @xmath4 , and decreases outerwards down to @xmath5 solar at @xmath6 from the galaxy center . dilution due to an extended hot gas is suggested .
perovskite chromites @xmath8cro@xmath0 , where @xmath8 is a rare earth element or yttrium are revisited in the recent years as possible multiferroic materials in which multiple ferroic orders such as ferroelectricity and antiferromagnetism coexist as discussed below @xcite . compared to perovskite manganites , which are well studied in light of multiferroicity , the microscopic physical properties of chromites are not explored in detail and the mechanism for multiferroicity in hocro@xmath0 is still under debate . the coexistence of ferroelectric and magnetic orders in rare - earth orthochromites was first suggested by subba rao @xmath10 @xcite . based on dielectric studies it is reported that the heavy rare earth chromites , @xmath8cro@xmath0 ( @xmath8=ho , er , yb , lu ) undergo a ferroelectric transition in the temperature range @xmath11 k @xcite . in a recent article , electrical polarization and magnetodielectric effect studies are reported for polycrystalline lucro@xmath0 and ercro@xmath0 @xcite . although both lucro@xmath0 and ercro@xmath0 showed the presence of a polar state induced by magnetic ordering below @xmath12 , polarization was not affected by applied magnetic fields , so that the magnetoelectric coupling was not evident in these compounds . further , the magnetodielectric effect observed in the case of ercro@xmath0 is one order of magnitude higher compared with lucro@xmath0 reflecting the role of different magnetism of rare - earth cations in ferroelectricity @xcite . recently , ghosh _ et al_. , have studied the ferroelectric properties of polycrystalline hocro@xmath0 by measuring the thermal variation of pyroelectric current @xcite . it was found that pyroelectric current exhibits its maximum value around the antiferromagnetic transition temperature @xmath13 k , nevertheless , the ferroelectric order temperature which associated with the emergence of spontaneous electric polarization is observed at a higher temperature of @xmath14 k. the atypical multiferroic behavior observed in hocro@xmath0 is argued to be a result of ho displacements and oxygen octahedral rotations in the non centrosymmetric @xmath15 space group @xcite . the role of the rare earth ion in determining the physical properties of chromites @xmath8cro@xmath0 was revealed in a recent communication where the origin of ferroelectricity in orthochromites has been attributed to the instability of the symmetric position of the rare earth ion @xcite . the interaction between magnetic rare earth and week ferromagnetic cr@xmath7 ions is the driving force for the breaking of symmetry , and thus the emergence of multiferroic behavior in these systems @xcite . despite a debatable multiferriocity , orthochromites possess a plethora of physical phenomena , providing excellent opportunities to study and understand the basic interactions in materials . the detailed knowledge on the properties of rare earth ion is of particular importance to understand the multiferroicity in rare earth chromites @xmath8cro@xmath0 . the rare earth orthochromites crystallize in a distorted orthorhombic perovskite structure with four formula units per unit cell @xcite . in hocro@xmath0 the exchange coupling between the cr@xmath7 nearest neighbors is predominantly antiferromagnetic and they order magnetically below the nel temperature of @xmath16 k @xcite . on the other hand , earlier reports differ on the aspect of ho - ordering . cooperatively induced ordering of ho in hocro@xmath0 was reported at 12 k @xcite whereas no ordering was observed by hornreich @xmath17 , down to @xmath18 k @xcite . ferroelectricity is observed in @xmath8cro@xmath0 systems only when the @xmath8 ion is magnetic . this directly suggests the exchange interaction between cr@xmath7 and @xmath8 is very important in inducing polarization and warrants the study of the local distortions around the @xmath8 ion as well as its magnetic properties . hence , we have chosen hocro@xmath0 as our subject to investigate the role of rare earth in the magnetic and thermodynamic properties of chromites . polycrystalline hocro@xmath0 was synthesized by solid state reaction of ho@xmath19o@xmath20 ( 3n ) and cr@xmath19o@xmath20 ( 4n ) in stoichiometric ratio . the precursors were mixed intimately and subsequently heat treated at 1100 @xmath21c for @xmath22 h. then , the material was reground and annealed again at @xmath23c for @xmath24 h. the phase purity of the synthesized powder sample was confirmed by powder x - ray diffraction ( pxrd ) with cu-@xmath25 ( @xmath26 ) radiation , using a huber x - ray diffractometer ( huber g@xmath27 ) in transmission guinier geometry . the profiles of the pxrd data were analyzed using the rietveld method @xcite implemented in the @xmath28 software suit @xcite . this confirmed the formation of orthorhombic single phase . the powder was then pressed into pellets and sintered at @xmath29c for @xmath30 h for further magnetic and thermal characterization . the heat capacity was measured in the temperature range from @xmath31 mk to @xmath32 k using a commercial quantum design physical property measurement system equipped with a dilution insert . the heat capacity values were extracted using the relaxation method @xcite . the background heat capacity of the microcalorimeter and the apiezon @xmath33 grease used for thermal conduction was measured before the sample measurement and subtracted from the raw data to obtain the absolute heat capacity of the sample . inelastic neutron scattering ( ins ) measurements were carried out on the basis backscattering spectrometer of the spallation neutron source ( sns ) , oak ridge national laboratory , usa @xcite . more details of this measurement is presented in reference @xcite . inelastic neutron scattering experiments were also carried out on the high resolution time of flight spectrometer focus at the spallation neutron source sinq at psi , villigen in switzerland . the polycrystalline sample was enclosed in an aluminum cylinder ( @xmath34 mm diameter , @xmath5@xmath35 mm height ) and placed into a he cryostat and the spectrum was collected using an incident energy of @xmath36 mev . additional experiments were performed for the empty container as well as for vanadium to allow the correction of the raw data with respect to background , detector efficiency , absorption and detailed balance according to standard procedures . inelastic data reduction and analysis was carried out using the software dave @xcite . k. the vertical bars denote the position of bragg reflections . a spurious peak indicated by asterisk symbol could not be indexed . ( b ) graphical representation of the crystal structure of hocro@xmath0 with cr - o@xmath37 octahedra in @xmath38 space group . blue , gray and red spheres indicate ho , cr and o atoms , respectively . ] the rietveld refinement of room temperature pxrd data is presented in fig . [ hco_xrd_refstru](a ) , and the results are tabulated in table 1 . using shannon radii values for the ions @xcite the value of the tolerance factor @xmath39 for hocro@xmath0 is found to be @xmath5@xmath40 . accordingly hocro@xmath0 is an orthorhombically distorted perovskite with the space group @xmath38 @xcite . the crystal structure at @xmath41 k is in good agreement with the previously determined orthorhombic structure with similar lattice parameters @xcite . the unit cell parameters obey the relationship , @xmath42 which is characteristic of @xmath43type orthorhombic structures . a buckling of the network of octahedra corresponding to cooperative rotation about a @xmath44$]axis leads to the @xmath43type orthorhombic structure . the clinographic view of the cro@xmath37 octahedra in hocro@xmath0 is presented in fig . [ hco_xrd_refstru](b ) . in perovskite manganites @xmath8mno@xmath0 ( @xmath8 = la , pr , nd , sm , eu , gd , tb and dy ) , in addition to the distortion due to buckling of the mno@xmath37 octahedron , a second distortion also arises because of the jahn teller effect . this is because the mn@xmath7 in @xmath8mno@xmath0 with four unpaired electrons the @xmath45shell in high spin state is jahn teller active . on the other hand cr@xmath7 in chromites with three unpaired electrons in @xmath45-orbitals is jahn teller inactive . thus a contribution to the lattice distortion in @xmath8cro@xmath0 due to the jahn teller effect is ruled out . [ hco_xrd_roomtempstrparam ] @ccccc + atoms & x & y & z & b@xmath46 ( @xmath47 ) + & & & & + + cr & @xmath48 & 0 & 0 & 0.815(2 ) + ho & @xmath49 & @xmath50 & @xmath51 & 1.031(3 ) + o1 & @xmath52 & @xmath53 & @xmath51 & 1.037(5 ) + o2 & @xmath54 & @xmath55 & @xmath56 & 1.152(1 ) + + @cccc + @xmath57 & @xmath58 & @xmath59 & @xmath60 + + @cccc + @xmath61 & @xmath62 & @xmath63 & @xmath64 + the variation of the specific heat ( @xmath65 ) of hocro@xmath0 with temperature is presented in fig . [ hco_specheat_full ] . the low temperature features in @xmath65 can be visualized clearly in a log - log plot as shown in the inset of fig . [ hco_specheat_full ] . the direct inspection of the specific heat curve evidences the presence of three main contributions or features : ( 1 ) a sharp increase in specific heat below @xmath66 k with a maximum at @xmath5@xmath67 k due to anomalously large hyperfine interaction between the electronic and nuclear spins of @xmath68ho@xmath7 leading to a nuclear - schottky specific heat ( @xmath69 ) ; ( 2 ) the electronic schottky contribution ( @xmath70 ) from thermal depopulation of the @xmath71 ground state multiplet of ho@xmath7 with a maximum at @xmath5@xmath72 k ; and ( 3 ) the @xmath73like anomaly with a peak at @xmath74@xmath75142 k due to the magnetic ordering of the cr@xmath7 moments . to determine different contributions to @xmath76 , a detailed analysis was performed in two steps , first @xmath76 was modeled in the temperature range @xmath77 k and then @xmath78 k. the specific heat below @xmath80 k has two broad features with maxima around @xmath5@xmath67 k and @xmath5@xmath72 k. from the low temperature specific heat measurements of ho metal it was found that anomalously large hyperfine interaction between the electronic and nuclear spins of ho commonly leads to a nuclear schottky anomaly with a maximum at @xmath75@xmath67 k @xcite . the hamiltonian for hyperfine interactions can be written in the form @xcite : @xmath81\ ] ] where , @xmath82 is the magnetic hyperfine constant , which is a measure of the strength of the hyperfine interaction between the nuclear moment and the magnetic moment associated with the 4@xmath83 electrons . @xmath84 is the quadrupolar coupling constant . the field is applied in the @xmath85 direction . since the projection @xmath86 can take @xmath87 values , i.e. @xmath88 , the hyperfine specific heat @xmath76 will be a schottky type specific heat , associated with the @xmath87 hyperfine levels . the nuclear spin of @xmath68ho with 100% natural abundance is @xmath89 , while @xmath90cr with @xmath91 has a natural abundance of 9.5% . one can calculate the mean square of effective nuclear moment for the natural abundance of active isotopes of ho and cr , @xmath92 , yielding @xmath93 for @xmath68ho and @xmath94 for cr . with these effective nuclear moments one can conclude that the cr hyperfine contribution is about three orders of magnitude smaller than the ho one . thus only the ho contribution is taken into account in the calculation of nuclear hyperfine contribution to the specific heat . due to the hyperfine interaction the holmium nucleus has @xmath95 possible spin orientations relative to an effective field @xmath96 . the energies @xmath97 of various nuclear spin states , i.e. the eigenvalues of the hamiltonian in equation ( [ hco_lt_hc_hamiltonian ] ) are @xmath98\ ] ] where , @xmath99 . information about @xmath82 and @xmath84 can be obtained by measuring the heat capacity at sufficiently low temperatures . in case of holmium , the quadrupolar coupling contribution is small and can be neglected @xcite . therefore for @xmath100 the equation ( [ hco_nuclear_energy ] ) reduces to , @xmath101 the specific heat in the temperature range @xmath102 k is modeled by taking into consideration the contributions from nuclear specific heat @xmath69 , an electronic schottky term @xmath70 and a lattice term @xmath103 . thus at low temperatures the @xmath76 of hocro@xmath0 is given by , @xmath104 where , @xmath69 and @xmath70 are given by the general expression for an n - level schottky specific heat term given by @xcite : @xmath105}}{{\sum\limits_i { \sum\limits_j { \exp \left [ { - \left ( { { \delta _ i } + { \delta _ j } } \right)/t } \right ] } } } } \ ] ] in this expression , @xmath106 and @xmath107 jmol@xmath3k@xmath3 is universal gas constant . to calculate the nuclear specific heat @xmath69 , a schottky curve for @xmath68ho with @xmath108=7/2 , eight equally spaced energy levels with splitting energy , @xmath109 , where @xmath110 , is used . to calculate the electronic schottky specific heat @xmath70 , a simple two level schottky term with energy splitting @xmath111 is used , as the contribution from higher energy terms is negligible in the temperature range @xmath102 k. in this temperature range the lattice contribution is expressed by a single debye term . at low temperatures , when @xmath112 , the debye temperature , the debye specific heat can be represented by well - known debye t@xmath113law as @xcite , @xmath114 where , @xmath115 is the number of atoms per molecule . + the fit results to equation ( [ hco_ltspecheat ] ) are given in fig . [ hco_lt_spheat_nuc_entropy ] ( a ) . from the fit the values of electronic schottky splitting energy @xmath116 and the debye temperature @xmath117 are found to be , 1.379(5 ) mev and @xmath118 k , respectively . the schottky energy 1.379(5 ) mev is consistent with earlier reports based on specific heat measurements @xcite and optical absorption zeeman spectroscopy @xcite . the value of @xmath82 from our analysis is found to be 0.2615(6 ) k which is slightly smaller compared to the value found in metallic ho , @xmath119 k @xcite , in inter - metallic hoco@xmath120 , @xmath121 k @xcite and in paramagnetic salts , @xmath121 k @xcite . the energy difference between two adjacent nuclear levels due to hyperfine field thus calculated from equation ( [ hco_nuclear_energy_approximate ] ) is @xmath122 @xmath1ev . from inelastic neutron scattering measurements it is possible to observe these energy levels directly , as an inelastic peak centered at @xmath5@xmath123 @xmath1ev . similar observations were made in spin ice compound ho@xmath120ti@xmath120o@xmath124 which shows a nuclear schottky peak around @xmath67 k in the specific heat data @xcite , which was later observed in inelastic neutron scattering measurements as a peak at @xmath5@xmath125 @xmath1ev @xcite . the effective magnetic ( hyperfine ) field at the holmium nuclei can be computed by writing @xcite , @xmath126 where , @xmath127 for @xmath68ho and @xmath128 is the nuclear magneton ( @xmath129 jt@xmath3 ) . using the value of @xmath82 in equation ( [ hco_hypconst ] ) the hyperfine field , @xmath96 is found to be @xmath130 t. this value is comparable to the values reported for ho@xmath120ti@xmath120o@xmath124 ( 720 t ) @xcite , metallic ho ( 770 t ) @xcite based on specific heat data and in ercro@xmath0 ( 530 t ) @xcite based on mssbauer spectroscopy . the nuclear specific heat @xmath69 was obtained by subtracting the lattice and electronic contributions from @xmath76 . the entropy @xmath131 associated with the nuclear specific heat was calculated by the numerical integration of @xmath132 . @xmath131 is given by the expression , @xmath133 the @xmath132 versus @xmath74 and the computed @xmath131 are presented in fig . [ hco_lt_spheat_nuc_entropy](b ) . the nuclear entropy @xmath131 reaches a maximum value of @xmath5@xmath6 jmol@xmath3k@xmath3 at @xmath5@xmath134 k. the theoretical limiting value of entropy for @xmath68ho with nuclear spin @xmath89 is calculated as @xmath135 jmol@xmath3k@xmath3 . it is shown as an horizontal dashed line in fig . [ hco_lt_spheat_nuc_entropy](b ) . an excellent agreement between the experimental @xmath131 with the theoretical value suggests that only contribution to the low temperature peak in the specific heat is from the nuclear schottky term due to hyperfine interactions . the non - magnetic contribution to the specific heat @xmath137 in the temperature region @xmath138 k is fitted assuming the contributions from an electronic schottky term @xmath139 , lattice term @xmath140 and a linear term @xmath141 . hocro@xmath0 has 5 atoms per formula unit , which implies that 15 vibrational modes to the phononic specific heat exist . taking in to account this constraint , we approximate the lattice contribution to the specific heat ( @xmath140 ) as sum of a debye term ( @xmath142 ) and two distinct einstein terms @xmath143 . the specific heat associated with the magnetic ordering , ( @xmath144 ) results in a @xmath73like transition with a maximum at @xmath5@xmath145 k. @xmath144 is obtained by subtracting @xmath137 from the experimental data ( @xmath76 ) . the magnetic entropy @xmath146 associated with @xmath144 is obtained by the numerical integration of @xmath147 . the @xmath137 and @xmath144 in the temperature range @xmath138 k can be written as @xcite , @xmath148 here @xmath139 is purely electronic schottky term as we are fitting only above 2 k , where the nuclear contribution is negligible . the non - magnetic contribution to the specific heat in this temperature is obtained by equations ( [ hco_gen_sch_term ] ) , ( [ non - magnetic ] ) , ( [ ht_specheat ] ) and , @xmath149}\label{hco_einstein}\\ & & { c_{linear } } = \gamma t\label{hco_linear}\end{aligned}\ ] ] in these expressions , @xmath8 is the gas constant , @xmath150 = @xmath151 , @xmath152 = @xmath153 , @xmath154 is boltzmann constant , @xmath155 is the coefficient of the linear term and @xmath156 is the number of atoms per molecule . the fitting was performed excluding the data in the temperature range 60200 k. the different contributions to the measured specific heat obtained from fitting are presented in fig . [ hco_cp_hight_fit_hco_cm_entropy](a ) . the values of debye temperature @xmath157 and two einstein temperatures ( @xmath158 , @xmath159 ) obtained from the fit are @xmath160 k ( 46(1 ) mev ) and @xmath161 k ( 67(1 ) mev ) and @xmath162 k ( 15.17(2 ) mev ) , respectively . this contribution to @xmath140 was parametrized by using 3 debye modes with debye temperature ( @xmath157 ) , 7 einstein modes with einstein temperature @xmath158 , and another 5 einstein modes with @xmath159 . although the used parametrization over simplify the phonon spectrum , the obtained key results are not influenced by subtleties in the choice of the modeled lattice contribution , i.e. , by the number of debye and einstein contribution or by the used absolute values within reasonable error bars @xcite . based on optical absorption spectroscopy @xcite , magnetization and magnetic susceptibility @xcite , specific heat @xcite , elastic neutron diffraction @xcite and inelastic neutron scattering ( ins ) @xcite experiments five electronic schottky levels were observed in hocro@xmath0 with the fifth level being at @xmath5@xmath163 k ( 23.4 mev ) . to calculate @xmath139 , a schottky curve for five energy levels is used which contribute to @xmath139 below @xmath41 k. the ground state energy level is assumed to be zero i.e. , @xmath164 . from the low temperature specific heat analysis , where a simple two level schottky model was used to calculate @xmath139 , the energy splitting is found to be 1.379(5 ) mev . we fixed this value as @xmath165 in the present calculations . the higher energy levels are determined from the fit as @xmath166=@xmath167 mev , @xmath168=@xmath169 mev and @xmath170=@xmath171 mev . these values are in excellent agreement with the reported crystal field energy values in hocro@xmath0 ( within 0.5@xmath172 ) . after subtracting the contributions , @xmath139 , @xmath140 and @xmath141 from the total specific heat @xmath76 , the magnetic specific heat @xmath144 is obtained , which can be seen as the deviation from the total fit in fig . [ hco_cp_hight_fit_hco_cm_entropy](a ) in the temperature range @xmath173 k. the magnetic entropy , @xmath146 was calculated by the numerical integration of @xmath147 analogous to the expression ( [ hco_nucentropy ] ) , by replacing @xmath69 by @xmath144 . the @xmath147 versus @xmath74 and computed @xmath146 is presented in fig . [ hco_cp_hight_fit_hco_cm_entropy](b ) . the experimental magnetic entropy value reaches a maximum value of @xmath534 jmol@xmath3k@xmath3 around 180 k , above @xmath12 . the theoretical limiting value of @xmath146 was calculated by adding the contributions due to ordering of both ho@xmath7 ( @xmath174 ) and cr@xmath7 ( @xmath175 ) moments , i.e. , @xmath176 jmol@xmath3k@xmath3 , indicated as a horizontal dotted line in fig . [ hco_cp_hight_fit_hco_cm_entropy](b ) . the experimental magnetic entropy is very close to the theoretical magnetic entropy around 180 k @xmath177@xmath178 . the good agreement between the experimentally found value for the magnetic entropy and that calculated for the spin only component of cr@xmath7 ions and orbital magnetic moment of ho@xmath7 ions , allows for concluding the following : first , the orbital moments of cr@xmath7 ions appear to be quenched while keeping the full spin moment . second , the lattice contribution seems to be described sufficiently well by the debye and einstein models . the value of linear coefficient @xmath155 from the fit is found to be @xmath179 mjmol@xmath3k@xmath4 . this value of @xmath155 is comparable to the reported values , @xmath180 mjmol@xmath3k@xmath4 which is associated with conduction electrons in some doped manganites @xcite . the linear coefficient is usually attributed to charge carriers , and is proportional to the density of states at the fermi level . however , hocro@xmath0 is an electrical insulator , thus the origin of linear term should be interpreted with caution . several electrical insulators have been reported with large values of @xmath155 , viz . , lamno@xmath181 ( @xmath182 mjmol@xmath3k@xmath4 ) @xcite , bavs@xmath0 ( @xmath183 mjmol@xmath3k@xmath4 ) @xcite and la@xmath184yca@xmath185mn@xmath120o@xmath124 ( @xmath186 mjmol@xmath3k@xmath4 ) @xcite . the origin of a linear contribution in these electrical insulators was attributed to a variety of magnetic phenomena . the most plausible explanation for the appearance of a linear term in the heat capacity in hocro@xmath0 is due to disordered ho@xmath7 as in the case of insulating ho@xmath187y@xmath188mno@xmath0 . for this compound it was observed that with increasing y content @xmath155 is reduced and drops to zero at @xmath189 , indicating the dependence of @xmath155 on the ho content @xcite . it was interpreted that the appearance of the linear term is due to high degeneracy of disordered ho@xmath7 spins above the ordering temperature . from high resolution neutron spectroscopy a huge quasielastic scattering was observed in hocro@xmath0 , which was understood as due to fluctuating disordered ho electronic moments @xcite . this supports our interpretation of disordered ho electronic moments as a possible source of a linear term in the observed specific heat . it is worth noting at this point that , to model the low temperature specific heat data using equation ( [ hco_ltspecheat ] ) , a linear term was not required which can be understood due to spin ordering of ho@xmath7 at these low temperatures @xcite . these observations confirm that the main origin of the linear term in hocro@xmath0 is disordered ho@xmath7 spins . inelastic neutron scattering spectra measured at the basis back scattering spectrometer are presented in fig . [ inelestic_hyperfine_crystalfield](a ) . at low temperatures two clear inelastic signals are observed on both energy gain and energy loss sides . a detailed study of the hyperfine spectra of hocro@xmath0 is published elsewhere @xcite . the inelastic spectra was modeled with the equation , @xmath190 \otimes r\left(\omega\right)+b\ ] ] where , delta function @xmath191 and @xmath192 represent elastic and inelastic peaks , respectively . these terms are convoluted numerically with the experimentally determined resolution function , @xmath193 , which is asymmetric due to the neutron pulse shape . @xmath194 is a flat background term and @xmath195 are scaling factors . the average energies of the inelastic peaks as obtained from the fits are @xmath196 @xmath1ev , this is in excellent agreement with the hyperfine splitting energy , @xmath122 @xmath1ev determined from our low temperature specific heat data . the fitting result to the 1.5 k data is presented in fig . [ inelestic_hyperfine_crystalfield](b ) . as can be seen from fig . [ inelestic_hyperfine_crystalfield](a ) a strong quasielastic scattering signal arises with increasing temperatures , broadening the elastic peak . an additional lorentzian term was required to describe this intensity . the temperature evolution of the quasielastic term was attributed to fluctuating electronic moments of the ho ions , which get increasingly disordered at higher temperatures . this reassures the validity of the large linear coefficient obtained by fitting the specific heat data at higher temperatures . from our earlier detailed report on the temperature dependence quasielastic scattering in hocro@xmath0 @xcite we found that the intensity decreases sharply below 40 k. further the temperature dependence of ordered magnetic moment of ho obtained from our recent neutron powder diffraction measurements @xcite , varies inversely as the temperature dependence of quasielastic scattering intensity and shows a sharp increase below 40 k. this confirms that the origin of quasielastic scattering is indeed fluctuating ho moments which are short range in nature . a similar phenomenon was also observed in ho@xmath120ti@xmath120o@xmath124 , the authors have interpreted it as being due to the fluctuating electronic moments of the ho @xcite . wan _ et al_. , @xcite showed both analytically and numerically that indirect magnetic exchange , which is short - range in nature is another driving force for the off - center atomic motion and ferroelectricity . in the present case , proposed short - range magnetic exchange interactions could cause the off - center atomic motion leading to ferroelectricity . we will report a detailed nuclear and magnetic structure studies as a function of temperature elsewhere , which should shed more light on atomic displacements and short range magnetic order in hocro@xmath0 . a typical inelastic spectrum measured at the time of flight instrument focus is presented in fig . [ inelestic_hyperfine_crystalfield](c ) . the non - kramer s ho@xmath7 ions in hocro@xmath0 are at sites of point group symmetry @xmath197 ( c@xmath198 ) , which typically leads to a singlet ground state . thus the ground multiplet of the ho@xmath7 ion , @xmath199 , split into @xmath200 singlets by the crystalline field generated by surrounding ions . the energy range of our inelastic data , limited to only three crystal field levels , makes a crystal field calculation inadequate using a point charge model . despite that , the observed inelastic peaks , fitted with gaussian peak functions as shown in fig . [ inelestic_hyperfine_crystalfield](c ) , are centered at energies , @xmath201 mev , @xmath202 mev and @xmath203 mev are in excellent agreement with those determined from heat capacity data and reported values @xcite . the hyperfine excitation signal which is masked gradually by quasielastic scattering upon increasing temperature , in contrast , the crystal electric field ( cef ) excitation signal is still visible in high temperature range . the temperature evolution of inelastic peak associated with the cef excitation between ground state ( @xmath204 ) and the second excitation level ( @xmath205 ) is presented in fig . [ figure6](a ) . the inelastic peak due to transitions between first excitation level ( @xmath165 ) and the second excitation level ( @xmath205 ) can also be seen ( above 10 k ) . two inelastic peaks were simultaneously fitted with gaussian peak function , both the peaks are well separated in all temperatures below 100 k above which the peak due to the transition between first excitation level ( @xmath165 ) and the second excitation level ( @xmath205 ) vanishes . the integrated peak intensity as a function of temperature for the peak corresponding to crystal field excitation @xmath205 is plotted in fig . [ figure6](b ) . it exhibits a typical behavior of van vleck contribution below @xmath12 and a clear anomaly at @xmath12 = 140 k. given the fact that the cef interaction reflects directly the electrical and magnetic potential created by neighboring ions , the anomaly of peak intensity around @xmath12 indicated the change of local environment surrounding the ho@xmath7 ion in hocro@xmath0 . because both cr and ho moments order below @xmath12 , the anomaly on peak intensity can be mainly attributed to the effect of exchange field on the ho@xmath7 ion from the long range order of cr@xmath7 magnetic sublattice . et al_. , @xcite proposed that the multiferroicity in @xmath8cro@xmath0 is caused by the interaction between magnetic rare earth and weak ferromagnetic cr@xmath7 ions following the breaking of symmetry . the observed anomaly in cef signal strengthens the importance of ho - cr exchange striction . the anomaly in [ figure6](b ) also implies the possible distortion of ho ions and their surroundings since the pyroelectric current exhibits its maximal at the same temperature @xcite . the temperature dependence of @xmath205 peak position is shown in fig . [ figure6](c ) . in the temperature range 1.5 60 k , the peak position remains unchanged at 10.1 mev and starts moving gradually to lower energy , above @xmath560 k , which is well below @xmath12 . the change of peak position of cef excitation at @xmath560 k hints at the shift of cef levels accompanied with the change of local crystallographic symmetry of ho@xmath7 ion . however , so far there is no report on the observation of structural distortion for hocro@xmath0 at 60 k. it is noticed that a tiny peak is observed at the same temperature in pyroelectric current curve , as shown in fig . 5(a ) in the ref . @xcite . therefore , the change of cef peak position at 60 k is related to the change of ferroelectric properties . a detailed study on the temperature dependent crystal structure is needed to understand the unusual behavior of cef excitation and thus the mechanism of multiferroicity in hocro@xmath0 . high quality polycrystalline hocro@xmath0 was prepared by solid state reaction method and characterized by means of x - ray powder diffraction , heat capacity and inelastic neutron scattering measurements . from the structural analysis we could establish the consistency of the observed crystal structure and theoretical predictions based on goldschimdt s tolerance factor rule . from the low temperature nuclear contribution to the specific heat results we obtained the first cef excitation energy for ho@xmath7 , 1.379(5 ) mev and hyperfine field of 600(3 ) t with a hyperfine splitting energy , 22.5(2 ) @xmath1ev for @xmath68ho with @xmath108=@xmath206 . the entropy ( @xmath207 ) associated with nuclear hyperfine specific heat ( @xmath208 ) was also estimated from the low temperature specific heat , which is in excellent agreement with theory . the hyperfine splitting energy determined from specific heat data is then confirmed from the peak observed in inelastic back scattering data . the large linear term @xmath2 mjmol@xmath3k@xmath4 in the specific heat was understood as due to disordered ho@xmath7 spins , which is further supported by strong quasielastic scattering observed in inelastic backscattering data at high temperatures . from the analysis of high temperature specific heat , by fixing the first cef excitation level to 1.379(5 ) mev , obtained from low temperature specific heat analysis , we determined three more crystal field transitions at 10.37(4 ) mev , 15.49(9 ) mev and 23.44(8 ) mev . the lower energy crystal field levels observed from the inelastic neutron scattering measurements are in excellent agreement with those determined from specific heat data . the magnetic entropy ( @xmath209 ) associated with the magnetic specific heat ( @xmath144 ) is obtained , which is consistent with the theoretical prediction . the linear term in specific heat and also quasielastic scattering observed in inelastic neutron spectra , further adds another possible driving force for the observed ferroelectricity in the form of short range exchange interactions in this compound as proposed by wan _ et al_. , @xcite . further , from the temperature evolution of crystal field spectra we confirm a direct correlation between the magnetic ordering and the ferroelectricity in this compound , as predicted by rajeswaran _ et al_. @xcite . in addition to aforementioned mechanisms , the asymmetry driven ferroelectricity as proposed by ghosh _ et al_. @xcite and indra _ et al_. , @xcite should also be considered as one of the driving force for ferroelectricity . more detailed temperature dependent structural studies are required for the quantitative analysis of distortions induced ferroelectricity . thus our study and recent reports on the ferroelectricity confirm that hocro@xmath0 and @xmath8cro@xmath0 in general , possesses more than one ingredient which can drive ferroelectricity , suggesting that these materials are potential multiferroic candidates for device applications . our study warrants , more detailed temperature dependent nuclear and magnetic structure studies to establish a most favorable mechanism for multiferroicity in hocro@xmath0 and rare - earth orthochromites in general . we thank the expert assistance of t. strssle , sinq , paul scherrer institute . part of the research conducted at sns was sponsored by the scientific user facilities division , office of basic energy sciences , us department of energy . this work is partially based on the experiments performed at the swiss spallation neutron source sinq , instrument focus ( proposal i d 20090536 ) . financial support from the european project eu nmi3 is acknowledged . 10 url # 1#1urlprefix[2][]#2 serrao c r , kundu a k , krupanidhi s b , waghmare u v and rao c n r 2005 _ phys . b _ * 72 * 220101 schpers m , wolter a u b , drechsler s l , nishimoto s , mller k h , abdel - hafiez m , schottenhamel w , bchner b , richter j , ouladdiaf b , uhlarz m , beyer r , skourski y , wosnitza j , rule k c , ryll h , klemke b , kiefer k , reehuis m , willenberg b and sllow s 2013 _ phys . rev . b _ * 88*(18 ) 184410	we report a comprehensive specific heat and inelastic neutron scattering study to explore the possible origin of multiferroicity in hocro@xmath0 . we have performed specific heat measurements in the temperature range 100 mk290 k and inelastic neutron scattering measurements were performed in the temperature range 1.5200 k. from the specific heat data we determined hyperfine splitting at 22.5(2 ) @xmath1ev and crystal field transitions at 1.379(5 ) mev , 10.37(4 ) mev , 15.49(9 ) mev and 23.44(9 ) mev , indicating the existence of strong hyperfine and crystal field interactions in hocro@xmath0 . further , an effective hyperfine field is determined to be 600(3 ) t. the quasielastic scattering observed in the inelastic scattering data and a large linear term @xmath2 mjmol@xmath3k@xmath4 in the specific heat is attributed to the presence of short range exchange interactions , which is understood to be contributing to the observed ferroelectricity . further the nuclear and magnetic entropies were computed to be , @xmath5@xmath6 jmol@xmath3k@xmath3 and @xmath534 jmol@xmath3k@xmath3 , respectively . the entropy values are in excellent agreement with the limiting theoretical values . an anomaly is observed in peak position of the temperature dependent crystal field spectra around 60 k , at the same temperature an anomaly in the pyroelectric current is reported . from this we could elucidate a direct correlation between the crystal electric field excitations of ho@xmath7 and ferroelectricity in hocro@xmath0 . our present study along with recent reports confirm that hocro@xmath0 , and @xmath8cro@xmath0 ( @xmath9 rare earth ) in general , possess more than one driving force for the ferroelectricity and multiferroicity .
the high statistics and accuracy of experimental measurements at lep and slc result in the outstanding level of precision in definition of the standard model parameters ( see ref . @xcite and references therein ) . in this situation effects of radiative corrections are of great importance . one of particular contributions to the process of electron positron annihilation is the radiation of secondary pairs . in comparison with the photon radiation , it is relatively small , because it appears only starting from the @xmath0 order . nevertheless , the total effect of pair production can reach several permil ( or even e few percent in particular cases ) and should be taken into account . we will consider here the reactions of the following type : e^+ + e^- f + \|f + ( f\|f ( ) ) , where @xmath1 is the so - called primary fermionic pair , @xmath2 is the secondary one , and @xmath3 stands for the usual photon radiation , which can take place in our case starting from the third order @xmath4 . the primary pair is the one registered in the particular measurement . the secondary one can be either real or virtual , and has to be accounted as a radiative correction according to certain experimental conditions ( with some restrictions ( cuts ) , if required ) . in principle , the pairs can be of the same flavour , like in the process @xmath5 . when both pairs are seen in the detector , one has to apply an algorithm to choose the primary pair . the complete second order qed calculation for production of the initial state secondary pairs ( @xmath6 , @xmath7 , and hadrons ) was performed in refs . the corresponding final state effect was considered in ref . @xcite . the interference between amplitudes for the initial and final state pair production has not been calculated systematically . an estimate of this effect will be presented in the conclusions . it is observed , that the @xmath0 approximation is not adequate to provide the desirable precision , especially in the region close to the @xmath8-peak . from another side , at lep2 , when the radiative return is allowed , qed radiative corrections reach several hundred percent and should be evaluated with care . in order to achieve the precision , required by the experiments ( see ref . @xcite ) , one should consider higher orders of the perturbation theory . another important point is that in the @xmath8-peak region the sub leading corrections are comparable in size with the leading ones . so , one has to go beyond the leading log approximation ( lla ) . the goal of the present paper is to describe and discuss the treatment of pair corrections , implemented in the semi analytical code zfitter v.6.30 @xcite . the main attention will be paid to the third order corrections . some numerical illustrations will be given . the uncertainty of the approach will be analyzed in the conclusions . starting from the @xmath9 order , one has amplitudes with simultaneous emission of pairs and photons . we treat them as a part of pair corrections , because they have not been taken into account in the traditional photonic rc . the main subject of this article is to analyze this effect . as usually in @xmath6 annihilation , the most important contributions are from the initial state ( is ) radiation ( real and virtual ) . in a typical high energy experiment , the conditions of the kinoshita lee nauenberg theorem are fulfilled for the final state ( fs ) radiation . that is why the large logarithms ( mass singularities ) cancel out in the fs correction , and the latter becomes relatively small . from the other hand , a specific energy dependence of the kernel cross section can provoke an enhancement of the is correction , like it happens at the @xmath8-peak or in the radiative return to the resonance at lep2 . pair corrections to the process of @xmath10 become ambiguous , when one observes @xmath11-fermion final states . nevertheless , the bulk of the pair rc to the @xmath12-fermion observables is coming from the kinematical situations , when the secondary pair is either virtual or escapes registration ( in the beam pipe or because of a small energy ) . on could try to subtract from the experimental data all real secondary pairs by means of a monte carlo generator . but that would destroy the cancellation between the real and virtual pairs , and the corrections would become even more pronounced and dangerous . one can get a serious numerical instability , trying to subtract both the virtual and the real pairs by means of different codes . even if the subtraction of the pure second order pairs is evaluated perfectly , we have to keep in mind the higher order corrections due to instantaneous emission of pairs and photons . the cancellation between the virtual and real pair rc is important not only in the @xmath4 , but in higher orders as well . indeed , we would end up with a completely wrong third order corrections ( the error would exceed the entire effect ) , if we were using for the subtraction of virtual and real pair contributions two different codes , one with the third order pair - photon effect and another one without it . here we will follow the definition of signal and background according to feynman diagrams , as discussed in ref . @xcite . in figure 1 we show several representatives of feynman graphs for real pair production processes . only the diagrams fig . 1@xmath13 and fig . 1@xmath14 ( plus the ones with pair production via a virtual photon emitted from the initial state positron of the final state anti fermion @xmath15 ) are to be accounted as the signal . these diagrams describe the non singlet ( ns ) mechanism of pair production . the diagrams of the type @xmath16 and @xmath17 with the production of a secondary pair via a virtual @xmath8-boson are treated as background . the so - called multi peripheral process is described by fig . diagrams of type @xmath19 represent the singlet mechanism of pair production . their contribution is considered as a background too . the pure electroweak four fermion processes with @xmath20-boson intermediate states fall into a separate class @xcite from both the theoretical and the experimental points of view . the relevant ( signal ) feynman graphs for virtual pair corrections are given in figure 2 . in other words we will count only the non singlet secondary pairs produced via virtual photons fig . 1@xmath21 ( some numbers for the singlet pairs will be presented ) . the contribution from the so - called background diagrams , like multi peripheral , double resonant @xmath22 , and singlet pair production , are supposed to be subtracted from the data . interferences between the _ signal _ and _ background _ amplitudes are typically small , but some of them , like the ones between the is singlet and non singlet pairs , should be taken into account . we will not use any direct cut on the phase space volume of real secondary pairs ( an inclusive treatment of radiative events ) . only the condition ( see eq . ( [ cuts ] ) ) on the primary pair invariant mass @xmath23 will be applied ( the dependence on the secondary pair invariant mass will be shown for the fs correction ) . in the lowest order for the fs pair corrections we use the results of ref . our aim here is to estimate the effect of higher orders . concerning the interplay with the is photonic corrections , it is claimed , that the final state pair corrections should be realized in the multiplicative way : [ multip ] = _ ( 1+_)(1 + _ ) , where @xmath24 stands for the initial state photonic correction , and @xmath25 gives the final state pair one . at lep2 energies , when the radiative return to the @xmath8-peak is allowed , we have huge values of @xmath24 ( typically above 100 percent ) , and the additive summation of the corrections @xmath26 fails completely . in principle , the initial state pair correction should participate in formula ( [ multip ] ) together with the photonic one , but , as one can see from numerical illustrations , an additive treatment of the is and fs pair corrections is acceptable as well . the effect of simultaneous photon and pair radiation from the final state can be estimated by multiplying the corresponding contributions : @xmath27 fs photon rc and the @xmath4 fs pair rc . one can see that this source does not give more than 0.05 permil in the accepted signal definition . this estimate was confirmed also by a computation ( analogous to eq . ( [ sig3p ] ) ) of the @xmath28 lla result for the fs non singlet pairs . it is worth to note here , that additional strong cuts on the final state photons and pairs can increase the fs corrections and revive the problem of the corresponding missing higher order terms . .radiative correction @xmath29 in permil for different cuts . [ cols="^,^,^,^,^,^,^",options="header " , ] first we have to remind about the effect of the initial final interference ( ifi ) in pair corrections . we suggest to estimate it by using the known effect for the ifi in photonic corrections . the bulk of the ifi pair correction is provided by the terms of the order @xmath30 ( the sub leading logarithms ) . the terms are coming from the kinematical region of small invariant mass real pairs ( and from the corresponding region of integration over virtual pairs ) . thus , one can use the factorization of the correction factors , corresponding to the photon emission and the pair production . so , we just multiply the ifi photonic correction by the factor @xmath31 and receive the estimate for the ifi pair rc @xmath32 . the typical value of the photonic ifi rc at lep energies is a few permil , therefore one can safely neglect the corresponding pair contribution . note , that in this way we considered only the non singlet mechanism of pair production . the interference of the initial state singlet pair production amplitudes with the final state ns ones is expected to be of the same magnitude . let us discuss now the uncertainty in the pair corrections . we mean just the indefiniteness in the treatment of hadronic pairs and of the @xmath28 ( and higher order ) leptonic pair corrections . other sources , relevant for concrete experiments , like the problem to distinguish between the primary and secondary pair or the proper subtraction of the background , have to be discussed elsewhere @xcite . first we suppose , that 15% is a reasonably safe estimate for the error in the second order hadronic pair contribution itself . in this case we use a phenomenological approach based on the measurements of electron positron annihilation into hadrons at low energies . the absence on the market of a _ plug and play _ parameterization of the @xmath33-function ( ratio of the annihilation cross sections into hadrons and muons ) does not allow to put a better precision level . for our computations of the fs hadronic pairs a simple parameterization of the function was constructed . the dispersion relations were used to look at the corresponding vacuum polarization effect and check the function . the stability in respect of the fluctuations in involved data was verified . a good agreement was observed with the published numbers on hadronic pair corrections @xcite in both the initial and the final states . the uncertainty of the higher order leptonic pairs is derived as follows . we assume that the leading logs @xmath34 , especially the ones multiplied by a power of @xmath35 , are taken into account completely . as stated above , all sub - leading terms can not be generated neither by convolution nor by exponentiation . therefore , we can estimate the error of the third order contribution to be about @xmath36 ( actually we take @xmath37 ) of its own size . for the third order hadronic contribution , derived by means of convolution , we should take into account also the intrinsic uncertainty in hadronic pairs , discussed above . as concerning the fourth and higher orders , we suggest to take the error just of the order @xmath38 @xmath6 contribution ( for all flavours together ) . the situation is better for the case of the final state pair corrections : the phenomenological treatment of hadronic pairs and the omitted effect of the interplay between the fs photons and pairs can not lead to more than 10 percent error in the fs pair rc , which is already small itself . remind also the interference between the initial and final pair production amplitudes , discussed above . collecting the sources of theoretical uncertainties , involved into our approach , we conclude : _ i ) _ at lep1 energies the error is about 0.4 permil on the wings and 0.2 permil on the top of the resonance ; _ ii ) _ at lep2 , when the radiative return to the @xmath8-peak is forbidden , we have a reasonably good error of about 0.1 permil ( for the realistic cut @xmath39 ; _ iii ) _ when the radiative return is allowed , the size of the error can reach 1 permil . the approach to pair corrections described here is implemented into the semi analytical code zfitter v.6.30 @xcite . a treatment of higher order pairs ( with pair photon interplay ) is available also in topaz0 @xcite , geantle @xcite , koralw @xcite , and some other codes . some comparisons of the approaches were done in ref . @xcite . bardin , m. grunewald and g. passarino , hep - ph/9902452 . m. kobel _ et al . _ [ two fermion working group ] , hep - ph/0007180 . berends , w.l . van neerven and g.j.h . burgers , nucl . b * 297 * ( 1988 ) 429 [ erratum - ibid . b * 304 * ( 1988 ) 921 ] . kniehl , m. krawczyk , j.h . kuhn and r.g . stuart , phys . b * 209 * ( 1988 ) 337 . hoang , j.h . kuhn and t. teubner , nucl . b * 455 * ( 1995 ) 3 [ hep - ph/9507255 ] . bardin , p. christova , m. jack , l. kalinovskaya , a. olchevski , s. riemann and t. riemann , comput . commun . * 133 * ( 2001 ) 229 [ hep - ph/9908433 ] . kuraev and v.s . fadin , sov . j. nucl . * 41 * ( 1985 ) 466 . s. catani and l. trentadue , jetp lett . * 51 * ( 1990 ) 83 . s. jadach , m. skrzypek and m. martinez , phys . b * 280 * ( 1992 ) 129 . m. skrzypek , acta phys . * b23 * ( 1992 ) 135 . arbuzov , phys . * b470 * ( 1999 ) 252 [ hep - ph/9908361 ] . d. decamp _ et al . _ [ aleph collaboration ] , z. phys . * c 48 * ( 1990 ) 365 . s. jadach , m. skrzypek and b. f. l. ward , phys . lett . * b 257 * ( 1991 ) 173 . g. montagna , o. nicrosini , f. piccinini and g. passarino , comput . commun . * 117 * , 278 ( 1999 ) [ hep - ph/9804211 ] . d. bardin , j. biebel , d. lehner , a. leike , a. olchevski and t. riemann , comput . commun . * 104 * ( 1997 ) 161 [ hep - ph/9612409 ] .	radiative corrections due to pair production in process of electron positron annihilation are considered . the main attention is paid to the corrections of the third order . simultaneous emission of photons and pairs is taken into account . the leading log approximation and convolution procedure are used . exponentiation is discussed . theoretical uncertainty in description of higher order secondary pairs is estimated . + pacs : 12.20.m quantum electrodynamics , 12.20.ds specific calculations * _ dipartimento di fisica teorica , universit di torino ; + infn , sezione di torino , + via giuria 1 , i-10125 torino , italy _
the superb image quality of the eso vlt makes possible the study of spatial structure in unprecedented detail . integral - field - unit ( ifu ) spectroscopy is a very powerful technique , allowing complex processes , both physical , chemical and kinematic , be probed with the same angular resolution as with conventional imaging data . as well as the ability to take spectra of contiguous areas , the now well - known advantages of ifu spectroscopy also include the elimination of slit losses and the relative ease of target acquisition . in the near - infrared , it is key to understanding the distribution , excitation and kinematics of gas and/or stars in objects ranging from obscured protostellar complexes to morphologically peculiar high - redshift galaxies . since many important advances will come from the study of statistical samples of faint astrophysical objects , simultaneous integral - field spectroscopy of several sources in a single field will greatly increase the efficient use of the precious vlt resource . the 2k @xmath4 2k pixels available in foreseeable infrared arrays means that fully covering the focal plane of an 8m - class telescope with an ifu capable of fine spatial sampling will remain impracticable for many years . to allow high - spatial - resolution integral - field spectroscopy of several targets within a field , we therefore need the ability to position a number of small - field ifus anywhere within a larger patrol field , a non - trivial task . the capability to work at wavelengths longer than 1.7@xmath1 m , where thermal emission dominates the background , is crucial . for example , the study of h@xmath7 emission in galaxies at @xmath8 , or using co , hei and heii lines to probe heavily obscured proto - stars all require integral - field spectroscopy with @xmath9@xmath1 m . the challenges of cryogenically cooling the optics and the deployment mechanism therefore have to be solved . we have developed a concept for a modular instrument with deployable optics to position 32 ifu fields over a 10@xmath0 field . we summarise the science drivers to estimate the required number of ifu fields and present strawman functional requirements for the instrument . finally , we then describe the results of our design study . extremely deep optical and infrared surveys the hubble deep fields ( williams et al . 1996 ) and their ilk are revolutionising the study of galaxy evolution by providing large samples of galaxies at a range of redshifts for which detailed comparison of the distribution , excitation and kinematics of the gas and stars with nearby analogues may be made . radio , x - ray and submm surveys ( using scuba , mambo , vla , _ xmm _ and _ chandra _ , and future missions such as blast , _ sirtf _ and _ herschel _ ) are providing large , complementary samples of x - ray- and radio - loud agn and distant starbursts , many severely obscured by dust ( e.g. smail et al . 1997 ; richards 2000 ) . @xmath6-band surveys find around @xmath10 @xmath11 @xmath12degree@xmath13 galaxies to @xmath14 19.5 @xmath11 22 ( huang et al . a fraction of these typically fulfil the selection criteria for an astrophysical application so that a programme will typically select a few tens of objects distributed over a 510@xmath0 field ( see table 1 ) . radio galaxy , 4c41.17 , as imaged in @xmath6 , with scuba 850-@xmath1 m contours superimposed over a small region ( ivison et al . eros with @xmath15 , some associated with the scuba galaxies , are circled . ] one of the defining characteristics of highly - obscured agn and star - forming galaxies is their red colours in the rest - frame optical and near - infrared . by their very nature these galaxies are difficult to identify in optical surveys . a few examples of such extremely red objects ( eros hu & ridgway 1994 ) turned up through deep near - infrared surveys , but these were viewed as little more than curiosities . a number of recent developments have forced a reassessment of this situation and it now appears that highly - obscured sources may host power sources which dominate the extragalactic background light in the hard x - ray and submm bands ( smail et al . 1999 ) . under half of the ero population has been identified with obscured starbursts , the rest with massive elliptical galaxies at @xmath16 12 whose red colours arise from their evolved stellar populations . the long - term goals of these studies are to constrain the early formation of massive galaxies at high redshifts and to derive the relative histories of obscured star formation and accretion activity to test models of galaxy and black hole evolution . knowledge of the redshift distribution is crucial @xmath17 breaks degeneracies in the models and allows the mass and luminosity of the sources to be explored , e.g. via observations of co. near - infrared ifu spectroscopy can also provide extinction estimates , independent star - formation rates , and probe for signs of agn activity ( e.g. broad h@xmath7 emission , ivison et al . 2000 ) . at @xmath18 the surface density of @xmath15 eros is @xmath2360degree@xmath13 , and so we immediately see the benefit of spectroscopic multiplexing . the surface density of submm galaxies above 2mjy at 850@xmath1 m is @xmath19degree@xmath13 with at least half expected to be associated with eros . a multiplexing factor of @xmath230 may thus be appropriate to take advantage of upcoming surveys . it is difficult , however , to predict this factor precisely , which is an excellent reason to build a modular instrument ( 3.3 ) . studies of galaxy evolution ultimately require full two - dimensional spectroscopy in the @xmath20 , @xmath21 and @xmath6 bands for meaningful samples of sources , to cover the key rest - frame optical diagnostic lines at @xmath22 . virtually all galaxies are spatially resolved in the best vlt @xmath6-band seeing , and many will have complex morphologies the result of interactions and mergers for the obscured systems , the result of rotation for some of the field galaxy population . integral - field spectroscopy will be essential to determine their detailed internal properties . other obvious astrophysical applications for near - infrared multiple - ifu spectroscopy include investigating the build - up of spiral disks , exploiting gravitational amplification of distant galaxies by foreground cluster lenses probing and investigating young globulars in starburst galaxies . .estimated source densities . [ cols= " < , < , < " , ] [ tab1a ] our instrument concept can provide a maximum number of pick - offs of about 32 . to use all 32 pick - offs , the mean surface density of randomly - distributed sources needs to be 0.52arcmin@xmath13 for a patrol field of 510arcmin , i.e. about 40 per field . this is because the pick - off mechanism inevitably vignettes a small fraction of the field . from table 1 it is obvious that about 30 deployable ifus are needed for the science there is no strong case for hundreds of simultaneous measurements , and too few ifus would not provide optimum efficiency . based on the science drivers , we concluded that the instrument should have the following capabilities : * 12.5@xmath1 m coverage , fully cryogenic for good @xmath6-band performance ; * patrol field of 510arcmin diameter ; * approximately 30 individual deployable fields ; * spatial sampling in the ifus should be 0.2arcsecpixel@xmath5 to capitalise on the best seeing conditions at @xmath6 ; * field of an ifu should be around 4 @xmath4 4arcsec so that for most sources some sky area is included ; * a spectral resolution of @xmath23 3000 to work effectively between oh lines ; * 2k @xmath4 2k detectors were assumed , with up to 8 needed for a full set of 32 deployable units ; * the deployment mechanism must be capable of positioning an ifu to within 1/5 of an ifu field . since the field of the ifu and the size of the array mean that there is no cross dispersion , if accurate relative astrometry is needed across the full spectral range then the @xmath24 spectra must be obtained before reconfiguring the field . there are two basic approaches to deploying reflective image slicers physically move them in the focal plane , or tilt / move a pick - off mirror to direct the desired field into a fixed image slicer . moving the ifu means that the position of the spectra on the detector will vary depending on the position of the ifu in the field , and the design of a cryogenic mechanism which does not obstruct the moving modules is very difficult . a simpler method is to use a pick - off mirror system to steer the desired field area into a fixed array of image slicers . this also has the advantage of producing a fixed spectrum on the detector(s ) . additional goals for the design of the pick - off mechanism were to use proven technologies , stepper motors and gears for the moving parts , and so a small pick - off mirror is preferred to a heavier image slicer . an articulated arm carries relay optics to feed small portions of the focal plane to each ifu which are at fixed locations in the instrument . the focal plane ( where the pick - offs move ) should ideally be flat to remove the need for a focus mechanism for each and every pick - off . it is also telecentric , which avoids pupil motion dependent on field position . field correction is achieved by lenses that form the cryostat window . the principle of operation of the pick - off arm concept is shown in figure 2 . a potential difficulty of using steerable mirrors the need to compensate for changes in the optical path length between the focal plane and the spectrograph is avoided because a collimated beam is passed around the elbow and shoulder and so no other mechanisms are required . although simple in theory , the articulated arm has a number of complexities in its engineering : it must be light and stiff , compact and protected against control failure . the end result , shown in figure 3 , is therefore quite complex . both the necessary rotations are produced by stepping motors which drive worm and wheel mechanisms . in the event of a malfunction in the control chain , a stopwork prevents the shoulder from turning more than @xmath2555@xmath26 by stalling the motor . the elbow rotation has two stages of gear reduction an initial worm and wheel identical to ( and concentric with ) the shoulder drive , followed by a gear and sector reduction to give a fine motion to the prism at the far end of the main arm . an additional benefit is that the weight of the elbow drive motor is not carried by the shoulder bearings or any of the moving support structures . this layout does mean , however , that a rotation of the shoulder induces a rotation of the elbow . to keep the main arm parallel to its original orientation as the shoulder moves , both shoulder and elbow drives must operate . if the elbow drive fails for any reason , it must be possible to drive the shoulder ( if only to retract the arm from the field of view ) without damaging the elbow joint . consequently , the main arm is mounted so that it is driven positively in one direction but via a spring in the other direction . this gives the freedom to deal with such a failure of the elbow motor , or a clash between an arm and its neighbours . a stopwork similar to that on the shoulder drive prevents excessive rotation of the elbow . rather than building a single large spectrograph , the instrument is instead designed to be highly modular with a number of identical spectrographs each fed by only a small number of ifus . there are several variations possible , ranging from 1 ifu 1 spectrograph 1 array to 4 ifus 1 spectrograph 1 array . the selection of one these options will ultimately depend on the number of arrays available and a more detailed science trade - off between the ifu fields and number of spectral elements . since much of the mass and space envelope for the instrument is occupied by the ifu deployment mechanisms , the spectrograph modules must be compact and simple . the maxiumum number of pick - off fields of view is provided by the solution in which 4 integral fields feed 1 spectrograph , as shown in figure 4 . small image slicers such as those described in content et al . 1997 or wells et al.2000 are packaged in groups of four to provide a compact unit at the end of the feed arms an optical design has been developed for a compact 2 mirror plus 1 grating spectrograph that has 2 slits at the entrance aperture . each slit is covers 2k pixels and is formed by two of the four image slicers ; the spectra cover 1k pixels . if only half the arms are used , it would be possible in principal to obtain spectra covering 2k pixels . to provide 32 simultaneously addressable integral fields of 4@xmath3 @xmath4 4@xmath3 covering a total 10-arcmin field , the instrument consists of 8 unit spectrographs , as described above , each of which has 4 movable arms and 4 image slicers . the overall dimensions of the instrument are 1.5 m diameter and 2.5 m long for the packaging shown in figure 5 . the mass of all the cold hardware is 700 kg , and the total instrument mass will be about 2tonnes , including cryostat , electronics , etc . this very modular concept allows considerable freedom in the detailed design . for example , it would be possible for one of the unit spectrographs to have a different spectral resolution , or to dedicate some arms to @xmath6-only spectroscopy and others exclusively to @xmath20 . equally important , the modular design would enable a modular development of the whole instrument . to build and test one unit spectrograph with 4 deployable fields is not a very big engineering step from existing cryogenic ir spectrographs . this unit could , of course , be used for science while the others were being built . the instrument concept described here was based on design studies carried out at the uk atc and the university of durham . several people participated or provided comments as the ideas progressed . we d like to thank ian parry , eli atad , keith taylor , roger haynes , suzanne ramsay - howatt , ian egan , paul van der werf , antonio chrysostomou , arjun dey for their scientific or technical contributions . m. wells , p. hastings , s. k. ramsay - howat : ` design and testing of a cryogenic image slicing ifu for ukirt and ngst ' . in : _ optical and ir telescope and instrumentation and detectors _ , ed . by m. iye , a. f. moorwood , proc . . 4008 , pp.456 ( 2000 )	integral - field spectroscopy is the most effective method of exploiting the superb image quality of the eso vlt , allowing complex astrophysical processes to be probed on the angular scales currently accessible only for imaging data , but with the addition of information in the spectral dimension . we discuss science drivers and requirements for multiple deployable integral fields for spectroscopy in the near - infrared . we then describe a fully modular instrument concept which can achieve such a capability over a 510@xmath0 field with up to 32 deployable integral fields , each fully cryogenic with 12.5@xmath1mcoverage at a spectral resolution of @xmath23000 , each with a 4@xmath3 @xmath4 4@xmath3 field of view sampled at 0.2arcsecpixel@xmath5 to take advantage of the best @xmath6-band seeing .
lattice study of heavy quark physics is indispensable for determining the standard model parameters such as the quark masses and ckm matrix elements , and for finding signals of new physics beyond it . obtaining accurate results for heavy quark observables , however , is a non - trivial task . since lattice spacings of order @xmath11 currently accessible is comparable or even larger than the compton wavelength of heavy quark given by @xmath12 for charm and bottom , a naive lattice calculation with conventional fermion actions suffers from large uncontrolled systematic errors . for this reason , effective theory approaches for heavy quark have been pursued . one of the approaches is the lattice version of the non - relativistic qcd ( nrqcd ) , which is applicable for @xmath13 @xcite . since the expansion parameter of nrqcd is the quark velocity squared @xmath14 , lattice nrqcd works well for sufficiently heavy quark such as the the bottom ( @xmath15 ) , and the bottomonium spectrum @xcite and the @xmath16 hybrid spectrum @xcite have been studied successfully using lattice nrqcd . an serious constraint with the approach , however , is that the continuum limit can not be taken due to the condition @xmath13 . thus scaling violation from the gauge and light quark sectors should be sufficiently small . in practice it is often difficult to quantify the magnitude of systematic errors arising from this origin . another difficulty is that there are a number of parameters in the nrqcd action which have to be determined . since in the present calculations the tuning of parameters is made at the tree level ( or tadpole improved tree level ) of perturbation theory , the accuracy achieved is rather limited . another approach for heavy quark uses a space - time asymmetric quark action , aiming to implement the @xmath17 improvement for arbitrary quark mass @xcite . with appropriate parameter tunings , this action is unitarily equivalent to the nrqcd action up to higher order corrections for @xmath13 , and goes over into the light quark sheikholeslami - wohlert ( sw ) action @xcite for @xmath18 . this approach has been originally proposed by the fermilab group and the action is hence called the `` fermilab action '' , whose first application is found in @xcite . since the necessary tuning of mass - dependent parameters is in general difficult , in practice one uses the usual sw quark action even for @xmath13 , where the sw action is unitarily equivalent to nrqcd . this simplified approach , called the `` non - relativistic interpretation '' for the sw quark , has been widely used in current lattice simulations of heavy quark , such as the calculation of the @xmath19 meson decay constant @xcite . toward the continuum limit @xmath20 the lattice action approaches the usual @xmath17-improved action and the systematic error becomes smaller as @xmath21 . however , the @xmath22 dependence at @xmath23 is quite non - linear , and it is not trivial how the systematic error could be controlled . recently use of the anisotropic lattice for heavy quark simulations has been proposed @xcite as a possible alternative to solve the difficulties of the effective approach . on an anisotropic lattice , where the temporal lattice spacing @xmath24 is smaller than the spatial one @xmath5 , one can achieve @xmath25 while keeping @xmath26 . therefore , using anisotropic lattices , one can reduce @xmath27 ( @xmath28 ) discretization errors while the computer cost is much less than that needed for the isotropic lattice at the same @xmath24 . naively it is expected that the reduction of @xmath27 errors entails the reduction of most of discretization errors due to large quark mass , since the on - shell condition ensures that the large energy scale flows only into the temporal direction as far as one considers the static particle , _ with zero or small spatial momentum_. if such a naive expectation is correct , the discretization error is controlled by a small parameter @xmath29 as it is for light quarks , and one can achieve even better accuracy by taking a continuum limit . however , it is not obvious that one can eliminate all @xmath30 errors at the quantum level , even if it is possible at the tree level . another advantage of the anisotropic lattice , which is more practical , is that a finer temporal resolution allows us to determine large masses more accurately . this has been already demonstrated in simulations of the glueball @xcite and the hybrid meson @xcite . klassen calculated the charmonium spectrum in the quenched approximation , employing lattices with the ratio of the temporal and spatial lattice spacings @xmath31 and 3 , as a feasibility study of the anisotropic approach @xcite . he tuned the ratio of the temporal and spatial hopping parameters @xmath32 non - perturbatively by demanding the relativistic dispersion relation for mesons . for the spatial clover coefficient @xmath33 , he adopted two choices , the tree level tadpole improved value correct for any mass ( @xmath34 ) and that correct only in the massless ( @xmath35 ) limit , in order to make a comparison . he mainly studied the spin splitting of the spectrum , and obtained an unexpected result that two different choices of the clover coefficients lead to two different values of the s - state hyperfine splitting even in the continuum limit @xcite . the continuum limit is of course unique , and clearly , at least one of the two continuum extrapolations is misleading . since the hyperfine splitting is sensitive to the clover coefficients , it is plausible that the disagreement is due to a large discretization error arising from the choice of the clover coefficients . in an unpublished paper @xcite , he pointed out the possibility that the @xmath36 errors still remain with his choice of the parameters , which we review in the next section . a similar statement can be found in some recent studies @xcite . in fact , he adopted rather coarse lattice spacings @xmath370.30 fm where @xmath26 . it is then questionable whether the reliable continuum extrapolation is performed at such coarse lattice spacings . using the same anisotropic approach as klassen , chen has recently calculated the quenched charmonium spectrum @xcite . she employed @xmath38 and finer ( @xmath390.25 fm ) lattices , and adopted the tree level tadpole improved clover coefficient @xmath33 correct for any mass , which is expected to be better than the other choice that is correct only in the massless limit . she computed not only the ground state masses but also the first excited state masses , and extrapolated them to the continuum limit . her results at @xmath38 are consistent with klassen s results at @xmath38 and 3 with the same choice of the clover coefficients . since chen s calculation was performed only at @xmath38 , similar calculations at different values of @xmath40 using fine lattices are needed to check the reliability of the continuum limit from the anisotropic approach . in addition , the complete p - state fine structure has not yet obtained in this approach so far , since the mass of @xmath41 state has not been measured in previous studies . in this work , we present a detailed study of the charmonium spectrum from the anisotropic lattice qcd . we perform simulations in the quenched approximation at @xmath42 , employing fine lattice spacings in the range @xmath430.2 fm . we attempt to determine the ground state masses of all the s- and p - states ( including @xmath44 ) as well as their first excited state masses . to estimate the systematic errors accurately , we adopt both the tree level tadpole improved value and non - perturbative one for @xmath6 , and both the pole mass and kinetic mass for @xmath45 which is tuned to the experimental value . we focus on the lattice spacing dependence and continuum limit of the mass splittings . we compare our results with the previous anisotropic results by klassen and chen to check the consistency , and with experimental values @xcite to estimate the quenching effect . in addition , to understand the discrepancy of the hyperfine splitting mentioned above , we make a leading order analysis using the potential model . to examine the effect of clover coefficients , we estimate the hyperfine splitting at leading order . comparing the leading order estimates with numerical results for the hyperfine splitting , we attempt to find a probable solution for this problem . our preliminary results are already reported in refs . @xcite . this paper is organized as follows . in section [ sec : ansioqcdact ] , we summarize and discuss the theoretical aspect of the anisotropic lattice qcd . in section [ sec : simulations ] , we give details of our simulation . our results for the charmonium spectra are shown in section [ sec : results ] , where we attempt to take the continuum limit and estimate the quenching effect . we address the problem of the discrepancy of the hyperfine splitting and study the effect of clover coefficients in section [ sec : hfsrevisit ] . section [ sec : conclusion ] is devoted to our conclusions . in this section we first define the anisotropic lattice action used in this work and fix notations . we then derive the tree level values of bare parameters in our massive quark action , and discuss effects of the anisotropy . although it was already discussed in earlier papers @xcite , we briefly describe the outline of derivations in order to be self - contained . we also consider the tadpole improvement of bare parameters and see how tree level values are modified . in this work , we use the standard wilson gauge action defined on an anisotropic lattice : @xmath46 \label{gaugeact}\ ] ] where @xmath47 is the gauge coupling , and @xmath48 and @xmath49 are the spatial and temporal plaquettes with @xmath50 . the anisotropy is introduced by the parameter @xmath51 and we call this the `` bare anisotropy '' . we denote spatial and temporal lattice spacings as @xmath5 and @xmath24 and define the `` renormalized anisotropy '' @xmath52 . we have @xmath53 at the tree level , and the @xmath54 at finite @xmath55 can be determined non - perturbatively by wilson loop matching @xcite . in numerical simulations , there are two methods for anisotropy tuning , either varying @xmath51 to keep @xmath40 constant or _ vice versa_. since the former is more convenient for keeping the physical size constant , and easier for performing the continuum extrapolation , we adopt it in this work . for the quark action , we employ the space - time asymmetric clover quark action on an anisotropic lattice proposed in ref . @xcite : @xmath56 , \label{massform}\end{aligned}\ ] ] where @xmath57 and @xmath58 is the bare quark mass , and @xmath59 and @xmath60 with @xmath61 . the wilson operator @xmath62 is defined by @xmath63 with the wilson coefficients @xmath64 and @xmath65 , \\ d^2_\mu \psi_x & \equiv & \frac{1}{a_\mu^2 } \left [ u_{\mu , x } \psi_{x+\hat{\mu } } + u^{\dagger}_{\mu , x-\hat{\mu } } \psi_{x-\hat{\mu } } -2\psi_x \right ] . \end{aligned}\ ] ] for the field tensor @xmath66 , we adopt the standard clover leaf definition . note that , in eq . ( [ massform ] ) , the factors in front of spatial wilson and clover operators include @xmath51 rather than @xmath40 . this is merely a convention and there is no deep theoretical reason . this action is essentially the same as the one employed by klassen@xcite and chen@xcite . in chen s work , however , @xmath67 was a tuning parameter with @xmath68 fixed . the two parameterizations are related to each other by a field rescaling @xmath69 . therefore @xmath70 instead of @xmath70 ] corresponds to @xmath72 in our convention . among these six parameters @xmath73 , at least one is redundant , so that we take @xmath74 as a redundant parameter and use it to remove the fermion doublers . although @xmath2 may not be taken arbitrary in the @xmath17 improved anisotropic quark action @xcite for the matrix elements , it can be taken arbitrary for the hadron mass calculation . therefore we always set @xmath75 and leave @xmath2 free in this work . the remaining parameters @xmath76 are used to tune the quark mass and reduce the lattice discretization error . for convenience in numerical simulations , we also present the quark action in a different form . rescaling the fields @xmath77 , the quark action can be transformed into a form given by @xmath78 \nonumber\\ & - & k_s \sum_i [ \bar{\psi}_x ( r_s- \gamma_i ) u_{i , x } \psi_{x+\hat{i } } + \bar{\psi}_x ( r_s+\gamma_i ) u^{\dagger}_{i , x-\hat{i } } \psi_{x-\hat{i } } ] \ } \nonumber\\ & + & i k_s c_s \sum_{x , i < j } \bar{\psi}_x \sigma_{ij } \hat{f}_{ij}(x ) \psi_x + i k_s c_t \sum_{x , i } \bar{\psi}_x \sigma_{0i } \hat{f}_{0i}(x ) \psi_x , \label{fnal}\end{aligned}\ ] ] where @xmath79 and @xmath1 are the spatial and temporal hopping parameters and the clover coefficients , respectively . the hopping parameters @xmath79 are related to the bare quark mass @xmath80 through @xmath81 the form eq . ( [ fnal ] ) on the anisotropic lattice is the same as that on the isotropic lattice in ref . note however that ref . @xcite uses the inverse of our definition for @xmath6 . we refer to their definition as @xmath82 . using eq . ( [ mqzeta ] ) one can convert @xmath83 to @xmath84 . in our convention , the relation between @xmath85 and @xmath86 is given by @xmath87 or equivalently , @xmath88 following ref . @xcite , we call the quark action eq . ( [ massform ] ) as the `` mass form '' and eq . ( [ fnal ] ) as the `` hopping parameter form '' . to derive the tree level value of bare parameters , we follow the fermilab method and calculate the lattice hamiltonian@xcite . after some algebra ( see appendix [ sec : hamiltonian ] for details ) , we obtain the lattice hamiltonian eq . ( [ contlikeh ] ) . using the fwt transformation eq . ( [ fwt ] ) , we then transform it to the non - relativistic form , in which the upper components of the dirac spinor completely decouple from the lower ones ( _ i.e. _ eliminate @xmath89 and @xmath90 ) . the transformed hamiltonian is given by @xmath91}{8m_e^2 } + \cdots \right)\hat{\psi } \label{nrhamiltonian}\ ] ] with @xmath92 , \label{electronicmass}\end{aligned}\ ] ] where @xmath93 , @xmath94 and @xmath95 are defined in eq . ( [ dashedparam ] ) . the @xmath96 term gives the leading order contribution to the hyperfine splitting , while @xmath97 $ ] term yields the fine structure splitting . the matching condition @xmath98 is equivalent to @xmath99 this yields the tree level value of bare parameters for massive quark : @xmath100 @xmath101 @xmath102 we note that @xmath33 is independent of the quark mass , while @xmath103 and @xmath104 have complicated mass dependences . the term @xmath105 seems to exist in eq . ( [ treezeta ] ) and ( [ treece ] ) . to see this explicitly , we expand @xmath103 and @xmath104 in @xmath106 . this gives @xmath107 @xmath108 the @xmath109 term , which is @xmath110 for heavy quarks at currently accessible lattice spacings of @xmath111 gev , appears in @xmath103 and @xmath104 even at the tree level . since @xmath112 is always multiplied with the spatial wilson coefficient @xmath2 in eqs . ( [ nu_smallm ] ) and ( [ ct_smallm ] ) , one can eliminate the @xmath109 term at the tree level by choosing @xmath113 however , this choice has the disadvantage that the mass splitting between unphysical doubler states and the physical state decreases as @xmath51 increases . moreover , the hopping terms in the quark action are no longer proportional to the @xmath114 projection operators . it is also doubtful that , beyond the tree level , the @xmath109 term can be still eliminated by this choice . if one adopts the conventional choice @xmath115 the @xmath109 term remains , but the unphysical doubler states decouple . this choice also has the practical merit that the quark action has the full projection property , so that the coding is easier and the computational cost is lower . the tree - level full mass dependences of @xmath103 and @xmath104 for @xmath116 and @xmath117 are shown in figs . [ fig : zeta_ce_tree_rxiinv ] and [ fig : zeta_ce_tree_r1 ] . in order to compare at the same @xmath5 , we choose @xmath118 as the horizontal axis instead of @xmath119 where @xmath120 is the pole mass . since @xmath121 gev and @xmath122 gev in current typical simulations , we plot results for @xmath123 . for @xmath116 shown in fig . [ fig : zeta_ce_tree_rxiinv ] , both @xmath103 and @xmath104 are monotonic functions in mass , and they converge to their massless values as @xmath51 increases at any fixed values of @xmath118 . hence , the @xmath109 dependence can be controlled by increasing @xmath51 . at @xmath124 the mass dependences of @xmath103 and @xmath104 completely disappear with the cost that the physical and unphysical states are almost degenerate . in actual simulations with @xmath116 , taking @xmath125 to decouple unphysical doublers , one is allowed to use the massless values for @xmath103 and @xmath104 , since their mass dependences are monotonic and very weak . in this case mass dependent parameter tuning can be avoided even at @xmath126 . for @xmath117 , on the other hand , the mass dependences of @xmath103 and @xmath104 are complicated and non - negligible even for large @xmath51 . indeed @xmath103 and @xmath104 do not converge to their massless values as @xmath51 increases at fixed @xmath118 , as shown in fig . [ fig : zeta_ce_tree_r1 ] . the deviation from the massless values at @xmath127 is smaller than the one at @xmath128 , but it becomes larger again as @xmath51 increases . therefore , taking @xmath127 - 3 in simulations with @xmath117 , one needs to perform a mass dependent parameter tuning . for both choices of @xmath2 , it is better to use a moderate value of @xmath51 , rather than excessively large values . in our numerical study of the charmonium spectra , we adopt the choice @xmath117 , and make a mass dependent parameter tuning , due to the practical reasons mentioned above . finally we show the tree level value of the parameters in the massless limit . by taking @xmath129 in eqs . ( [ treezeta])-([treece ] ) , one obtains @xmath130 in the mass form , or @xmath131 in the hopping parameter form . note that there is an ambiguity in the tree level value of @xmath132 , since @xmath133 at the tree level but @xmath134 in the simulation . fortunately , this ambiguity almost disappears after the tadpole improvement , as shown in the next subsection . in this section we apply the tadpole improvement @xcite to the parameters of the anisotropic lattice action at tree level in order to partially include higher order corrections . one first rewrites the lattice action using a more continuum - like link variable @xmath135 , where @xmath136 is the expectation value of the spatial or temporal link variable , _ i.e. _ one replaces @xmath137 and then repeats the tree - level calculations . we will show below how the tree - level values of bare parameters are modified . by the replacement of eq . ( [ tadpoleu ] ) , the anisotropic gauge action eq . ( [ gaugeact ] ) becomes @xmath138 , \label{tadpolegauge}\ ] ] where @xmath139 , and @xmath140 and @xmath141 are given by @xmath142 requiring space - time symmetry for the action eq . ( [ tadpolegauge ] ) in the classical limit , one obtains the tree - level tadpole - improved value of the anisotropy ( denoted by an index ` ti ' ) , @xmath143 in practice @xmath144 in eq . ( [ tixi0 ] ) agrees with the renormalized anisotropy @xmath40 within a few % accuracy at @xmath145 . therefore one can replace the factor @xmath146 by @xmath40 in the following equations . this simplifies the tree level expression . moreover , the arbitrariness for the choice of anisotropy disappears . when the fermion action is rewritten in terms of @xmath147 and @xmath148 instead of @xmath149 and @xmath150 , the action keeps the same form with @xmath151 @xmath152 then @xmath153 and the bare quark mass @xmath154 are modified to @xmath155 @xmath156 using parameters with the tilde , one can repeat the derivation in the previous subsection . for a massless quark , one obtains @xmath157 therefore , tadpole - improved ( ti ) tree - level estimates are @xmath158 which indicates that non - perturbative @xmath6 at @xmath159 is closer to @xmath51 than to @xmath40 , and @xmath160 as can be seen in eqs . ( [ tizeta ] ) and ( [ ticst ] ) , the tadpole improvement eliminates the uncertainty of choice of anisotropy ( _ i.e. _ whether to chose @xmath51 or @xmath40 ) at tree level . converting to the @xmath85 convention , one obtains @xmath161 note that @xmath162 is normalized to 1 since @xmath103 equals @xmath163 and not @xmath164 , hence the former definition is practically more convenient than the latter one . note also that tadpole factors in @xmath165 and @xmath166 are different because @xmath167 equals @xmath168 and not @xmath169 . similarly , for massive quarks , tadpole - improved tree - level estimates become @xmath170 with @xmath171 , and @xmath172 @xmath173 where @xmath174 . we proceed to calculate the charmonium spectrum in the quenched approximation as our first numerical study using the anisotropic lattice . in this section we describe the computational details of our quenched charmonium calculation . for the gauge sector , we use the anisotropic wilson gauge action given in eq . ( [ gaugeact ] ) . throughout this paper , we employ @xmath175 , where @xmath40 is the renormalized anisotropy . in order to achieve @xmath42 , we tune the bare anisotropy @xmath51 , using the parameterization of @xmath176 given by klassen@xcite : @xmath177 where @xmath178 , @xmath179 and @xmath180 we perform simulations in the quenched approximation , at four values of gauge coupling @xmath181 , 5,90 , 6.10 and 6.35 . these couplings correspond to @xmath1820.2 fm and @xmath1830.48 for @xmath184 gev . the spatial lattice size @xmath185 is chosen so that the physical box size is about 1.6 fm , while the temporal lattice size @xmath186 is always set to be @xmath187 . for the charm quark , we use the anisotropic clover quark action eq . ( [ fnal ] ) with the conventional choice of the spatial wilson coefficient , @xmath188 , as mentioned in sec.[sec : massivetree ] . we take two values for the bare quark mass @xmath189 at each @xmath55 in order to interpolate ( or extrapolate ) results in @xmath106 to the charm quark mass @xmath190 . the charm quark mass @xmath190 is fixed from the experimental value of the spin averaged @xmath191 meson mass . in this procedure , we use both the pole mass @xmath192 and kinetic mass @xmath193 for the @xmath191 meson . for @xmath6 , the ratio of the hopping parameters , we adopt both the tree - level tadpole - improved value @xmath194 , and a non - perturbative value @xmath195 determined from the meson dispersion relation . we describe our method of tuning @xmath6 in detail in sec.[sec : mztuning ] . for the spatial clover coefficient @xmath196 , we employ the tree - level tadpole - improved value for massive quarks eq . ( [ timasscb ] ) . note that @xmath196 has no mass dependence at the tree level . on the other hand , we adopt the tree - level tadpole - improved value in the massless limit eq . ( [ ticst ] ) for the temporal clover coefficients @xmath197 . we discuss possible systematic errors arising from our choice of the parameters @xmath6 and @xmath1 in sec.[sec : syserr ] . the tadpole factors @xmath198 in eqs . ( [ ticst ] ) and ( [ timasscb ] ) are estimated by the mean plaquette prescription : @xmath199 if we adopted the alternative definition @xmath200 instead , @xmath201 would be greater than 1 . we use @xmath40 instead of @xmath146 in eq . ( [ ticst ] ) . gauge configurations are generated by a 5-hit pseudo heat bath update supplemented by four over - relaxation steps . these configurations are then fixed to the coulomb gauge at every 100 - 400 sweeps . on each gauge fixed configuration , we invert the quark matrix by the bicgstab algorithm to obtain the quark propagator . we always perform the iteration of the bicgstab inverter by @xmath186 times , where @xmath186 is the temporal lattice size . by changing the stopping condition for the quark propagator , we have checked that this criterion is sufficient to achieve the desired numerical accuracy . we accumulate 400 - 1000 configurations for hadronic measurements . our simulation parameters are compiled in tables [ tab : simparam ] and [ tab : simparam2 ] . in table [ tab : simparam3 ] , we compare some of the parameters used in our simulation ( labeled by `` set a '' ) with those in the previous studies by klassen ( `` set b '' and `` set d'')@xcite and by chen ( `` set c'')@xcite for later references . in this work , we calculate all of s- and p - state meson masses of charmonia , namely @xmath202(@xmath203 ) , @xmath204(@xmath205 ) , @xmath206(@xmath207 ) , @xmath208(@xmath209 ) , @xmath210(@xmath211 ) and @xmath44(@xmath212 ) . for this computation , we measure the correlation function of the operators which have the same quantum number as one of above particles . in table [ tab : spope ] we give the operators for the s- and p - state mesons . there are two types of operators , those of the form @xmath213 and of @xmath214 , where @xmath215 represents a combination of @xmath216 matrices and @xmath217 the spatial lattice derivative . we call them the @xmath215 operator and the @xmath218 operator , respectively . the latter appears only for the p - state mesons . note that there are two lattice representations for the @xmath44 state ( e and t representations ) due to breaking of rotational symmetry . we measure the correlation functions of the @xmath215 operators @xmath219 where @xmath220 is a source smearing function , and we always adopt a point sink . we employ the point source ( @xmath221 ) with @xmath222 and an exponentially smeared source ( @xmath223 ) with @xmath224 , where @xmath225 and @xmath226 are smearing parameters . therefore we have three source combinations , @xmath227 , 01 and 11 , for the @xmath215 operators . the smearing parameters @xmath225 and @xmath226 at each @xmath55 are chosen so that the effective mass of the @xmath191 meson for @xmath228 has a wide plateau . to obtain the correlation functions of the @xmath218 operators , we measure @xmath229 where @xmath230 is the discretized derivative at the sink , and we employ a smeared derivative source ( @xmath231 ) given by @xmath232 with @xmath225 and @xmath226 the same as those for @xmath223 . for the @xmath208 state , for example , we calculate @xmath233 with @xmath234 . for the @xmath218 operators , we have two source combinations , @xmath235 and 12 . in total , s - state mesons have @xmath227 , 01 and 11 source combinations , and p - state mesons have 00 , 01 , 11 , 02 and 12 source combinations except for @xmath44 . since there is no @xmath215 operator for @xmath44 , it has only 02 and 12 source combinations . to calculate the dispersion relation of s - state mesons , we measure correlation functions for four lowest non - zero momenta , @xmath236 in addition to those at rest . correlation functions with the same value of @xmath237 but different orientations are averaged to increase the statistics . let us describe our method of tuning @xmath6 and @xmath106 in detail . we determine the input parameters @xmath238 and @xmath239 as follows . first we fix @xmath240 and choose @xmath241 and @xmath242 where the 1s meson mass roughly agrees with the experimental value . then we determine both the tree - level tadpole - improved value @xmath194 and the nonperturbative value @xmath195 at @xmath243 and @xmath242 . to obtain @xmath194 at fixed @xmath106 , we use eqs . ( [ tibearemass ] ) and ( [ timasszeta ] ) . we replace the factor @xmath244 in eq . ( [ timasszeta ] ) with @xmath245 , using eq . ( [ tixi0 ] ) . on the other hand , @xmath195 is obtained by demanding that the relativistic dispersion relation is restored at small momenta for the 1s meson . the dispersion relation on a lattice is given by @xmath246 where @xmath247 is called the ` speed of light ' , and @xmath192 and @xmath193 are the pole and kinetic masses of the 1s meson . throughout this paper , a capital letter @xmath248 denotes the meson mass , while a small one @xmath249 the quark mass . generally @xmath247 is not equal to one due to lattice artifacts . we extract the speed of light @xmath247 by fitting @xmath250 linearly in @xmath251 for three or four lowest momenta , since the linearity of @xmath250 in @xmath251 is well satisfied . we identify @xmath195 with a point where @xmath252 or equivalently @xmath253 for the 1s meson . to determine @xmath195 , we perform preparatory simulations and calculate @xmath247 for @xmath254 , 3.0 and 3.2 at @xmath243 and @xmath242 using 100 - 200 gauge configurations . then we find @xmath255 , where @xmath256 , from an interpolation of @xmath6 . as shown in table [ tab : simparam2 ] , the speed of light @xmath247 at @xmath195 is indeed equal to 1 within 1% , which is roughly the size of the statistical error . production runs for the charmonium spectrum described in sec.[sec : choiceparam ] are performed at @xmath189 and @xmath257 for each @xmath55 . accidentally , for @xmath258 and 6.35 , @xmath259 holds within our numerical accuracy , so we use the same data for the analysis at these @xmath55 . finally we linearly interpolate or extrapolate results at @xmath189 to those at @xmath260 , with fixed @xmath6 ( @xmath261 or @xmath195 ) . as already mentioned , we identify @xmath190 with a point where the spin - averaged 1s meson mass @xmath45 in units of a physical quantity @xmath262 is equal to the corresponding experimental value : @xmath263 with @xmath264 mev for charmonium . in this work , we adopt the sommer scale @xmath10 and the spin - averaged mass splittings @xmath265 and @xmath266 as the scale quantity @xmath267 . the spin - averaged masses are defined by @xmath268/4 , \label{spinaverages}\\ m(n\bar{p } ) & = & [ 3m(n^1p_1)+5m(n^3p_2)+3m(n^3p_1)+m(n^3p_0)]/12 \label{spinaveragep}\end{aligned}\ ] ] with @xmath269 the radial quantum number . the experimental values of the mass splittings @xmath270 and @xmath271 are 457.9 mev and 595.4 mev , respectively . the experimental values of @xmath10 is not known , and we use a phenomenological estimate @xmath272 fm . for the definition of the lattice meson mass @xmath273 in eq . ( [ tunem0 ] ) , we have two choices in the case of @xmath274 : one is the pole mass @xmath192 and the other is the kinetic mass @xmath193 . on the other hand , in the case of @xmath275 , @xmath276 should hold by definition . in practice , there can be small deviations due to the statistical error . therefore we have 4(=2@xmath2772 ) choices for @xmath278 in total . from meson correlation functions we extract the meson mass ( energy ) by standard @xmath279 fitting with a multi - hyperbolic - cosine ansatz ( termed @xmath280-cosh fit below ) @xmath281 , \label{fitansatz}\ ] ] where @xmath282 represents the source combination ( 00 , 01 , etc . ) , @xmath283 is the time separation from the source , and @xmath280 is the number of states included in the fit . we determine the mass of the ground state and the first radial excited state for each particle , and the mass splittings such as @xmath284 and @xmath285 , from a 2-cosh fit using several correlation functions with different source combinations simultaneously . here we use the correlation functions of @xmath227 , 01 and 11 sources for s - states , while 00 , 11 , 02 and 12 sources are used for p - states except for @xmath286 . for @xmath286 , we use the correlation functions of 02 and 12 sources . the 2-cosh fit for each s - state always gives the ground state mass consistent with that from the 1-cosh fit . on the other hand , for the p - state , the 2-cosh fit is preferred over the 1-cosh fit because the @xmath287 mass from the 1-cosh fit using the correlation function of 11 and 12 sources occasionally disagrees by a few @xmath288 , due to excited state contaminations . to determine the mass of the first excited state accurately , it is better to adopt results from the 3-cosh fit . however , we do not perform the 3-cosh fit systematically because of the instability of it , and adopt results from the 2-cosh fit for the first excited state mass . this may cause an overestimation of the first excited state mass due to a contamination from higher excited states . to determine the spin - averaged @xmath191 mass and the @xmath191 energy at @xmath289 , and the spin mass splittings such as @xmath290 and @xmath291 , we perform a 1-cosh fit ( @xmath292 ) using the source combination which gives the widest plateau in the effective mass . we use the 01 source for the s - state , and the 12 source for the p - state . we always check that the spin mass splitting from a simultaneous 2-cosh fit mentioned above agrees with that from the 1-cosh fit within 1 - 2 @xmath288 . we also check that the splitting @xmath291 from a 1-cosh fit using the 11 source agrees with that using the 12 source . in these analyses , we perform both the uncorrelated fit , and the correlated fit which takes account of the correlation between different time slices and different sources . the uncorrelated fit is always stable and gives @xmath293 . the correlated fit with 1-cosh ansatz is also stable and produces results consistent with those from the uncorrelated fit . however , the correlated 2-cosh fit is often unstable , either failing to invert the covariance matrix or giving large @xmath294 even if it converges . therefore we adopt the uncorrelated fit for our final analysis . the fitting range @xmath295 $ ] for the final analysis is determined as follows . from an inspection of the effective mass plot , we determine @xmath296 which roughly has the same physical length independent of @xmath55 . we repeat the 1- and 2-cosh fits for each @xmath55 , varying @xmath297 with fixed @xmath296 , and find a range of @xmath297 where the ground state mass and the first excited state mass ( for 2-cosh fit ) are stable against @xmath297 . we also check that it has reasonable value of @xmath298 . the final @xmath297 is then chosen from the region accepted above so that its physical length is roughly equal independent of @xmath55 . typical examples of the effective mass plot and @xmath297-dependence of the fitted mass are shown in figs . [ fig : seffmb590mq144]-[fig : peffmb590mq144 ] , and in fig . [ fig : tminb590mq144 ] , respectively . our final fitting ranges are summarized in table [ tab : frange ] . statistical errors of masses and mass splittings are estimated by the jack - knife method . the typical bin size dependences of jack - knife errors for the ground state masses are shown in figs . [ fig : bindeppi ] and [ fig : bindep1p1 ] . we always adopt a bin size of 10 configurations , _ i.e. _ 1000 - 4000 sweeps . we discuss scaling violation for our action and how the results at finite @xmath5 are extrapolated to the continuum limit @xmath299 . since we use the anisotropic wilson gauge action with nonperturbatively tuned @xmath51 , the scaling violation from the gauge sector starts at @xmath300 . for the quark sector , we use the anisotropic clover quark action with tadpole - improved clover coefficients @xmath1 , and either the tadpole - improved value @xmath194 or nonperturbative value @xmath195 for @xmath6 . since we adopt the tree - level tadpole - improved value of @xmath196 for massive ( @xmath301 ) quarks , the scaling violation arising from the choice of @xmath196 is @xmath300 and @xmath302 . on the other hand , for @xmath197 , we adopt the tree - level tadpole - improved value correct only in the massless ( @xmath303 ) limit , which generates an additional @xmath304 error . recall that the @xmath305 ( not only @xmath29 ) dependence of the parameter remains with our choice of the spatial wilson coefficient @xmath4 at the tree level , as discussed in sec.[sec : ansioqcdact ] . in the case of @xmath255 , therefore , the scaling violations are @xmath300 and @xmath306 at leading order , and @xmath302 at next - to - leading order . the size of these errors are estimated to be @xmath3071% , @xmath3084% and @xmath3091% for @xmath3106.35 corresponding to @xmath3112.8 gev . here we took @xmath312 mev ( @xmath313 ) and @xmath314 gev ( @xmath315 ) , and the renormalized coupling constant @xmath316 is estimated from eq . ( [ tigauge ] ) . it is expected that the @xmath302 errors are largely eliminated by the tadpole improvement . when the tree level tadpole improved value @xmath194 is used instead of @xmath195 , we have additional @xmath317 and @xmath318 errors , since the kinetic term is a dimension four operator . the size of the additional errors is estimated to be @xmath31912% and @xmath3206% . again we expect that the dominant part of this error is eliminated by the tadpole improvement . in this work we adopt an @xmath321-linear extrapolation for the continuum limit , because the leading order scaling violation is always @xmath322 irrespective of the choice of @xmath6 . we also perform an @xmath5-linear extrapolation to estimate systematic errors . in practice we use results at three finest lattice spacings _ i.e. _ @xmath258 - 6.35 ( @xmath323 ) for the continuum extrapolation , excluding results at @xmath181 ( @xmath324 ) , which appear to have larger discretization errors as expected from the naive order estimate . performing such extrapolations for all sets of @xmath325 and @xmath326 , we adopt the choice which shows the smoothest scaling behavior for the final value , and use others to estimate the systematic errors . now we present our results of the quenched charmonium spectrum obtained with the anisotropic quark action . in this section , we first compare results of @xmath195 with @xmath194 . second , we determine the lattice scale , and study the effect of @xmath278 tuning . we then show the results of charmonium masses and mass splittings , and estimate their continuum limit . in fig . [ fig : cenpz ] , we plot a typical example of the dispersion relation and the speed of light . as shown in the left figure , the linearity of @xmath327 in @xmath251 is satisfied well . indeed the ` effective speed of light ' , defined by @xmath328 has a wide plateau as shown in the right figure . therefore we employ the linear fit in @xmath251 to extract the speed of light @xmath247 from @xmath327 . this figure also illustrates that the speed of light @xmath247 for @xmath202 agrees well with that for @xmath204 within errors . this is indeed the case for all data points as observed in table [ tab : simparam2 ] . the speed of light @xmath247 seems universal for all mesons as pointed out in ref.@xcite . the nonperturbative value of @xmath6 , @xmath195 , is obtained by demanding that the speed of light @xmath247 is equal to 1 within 1% . on the other hand , the tree - level tadpole - improved value , @xmath194 , gives @xmath247 deviating from 1 by @xmath3294% _ @xmath3294 @xmath288 at most , which is much smaller than the size of the @xmath330 error ( @xmath33115% , @xmath33222% ) estimated in the previous section . this suggests that @xmath330 errors associated with @xmath194 are almost eliminated by the tadpole improvement , as expected . in fig . [ fig : tunedzeta ] , @xmath333 and @xmath334 at @xmath243 and @xmath242 are plotted as a function of @xmath335 . we find that @xmath336 ( circles ) and @xmath162 ( squares and solid line ) agree within errors at @xmath337 but deviate from each other at @xmath338 ( @xmath339 ) . the latter is one of the reasons why we exclude this point in the continuum extrapolation . one also notices that the slope of @xmath103 approaching the value @xmath68 in the continuum limit is steep , and in addition , the difference @xmath340 for our data does not have a smooth dependence in @xmath341 . as discussed in sec.[sec : hfsrevisit ] , these features of @xmath336 bring complications in the scaling behavior of the hyperfine splitting . in this work , we determine the lattice spacing via the sommer scale @xmath10@xcite , the @xmath342 meson mass splitting , and the @xmath343 splitting . we compare the results obtained with these different scales , in order to estimate the quenching errors . in order to calculate the static quark potential needed for the extraction of @xmath10 , additional pure gauge simulations listed in table [ tab : r0 ] are performed . using @xmath344 fm lattices , we measure the smeared wilson loops at every 100 - 200 sweeps at six values of @xmath55 in the range @xmath3106.35 . details of the smearing method @xcite are the same as those in ref.@xcite . we determine the potential @xmath345 at each @xmath55 from a correlated fit with the ansatz @xmath346 where @xmath347 and @xmath348 are the spatial and temporal extent of the wilson loop in lattice units . the fitting range of @xmath349 is chosen by inspecting the plateau of the effective potential @xmath350 a correlated fit to @xmath345 is then performed with the ansatz @xmath351 \right ) , \label{potfit}\ ] ] where @xmath288 is the string tension and @xmath352 $ ] is the lattice coulomb term from one - gluon exchange @xmath353 = 4\pi \int_{-\pi}^{\pi } \frac{d^{3}\mathbf{k}}{(2\pi)^{3 } } \frac{\cos ( \mathbf{k } \cdot \hat{r } ) } { 4 \sum_{i=1}^{3 } \sin^{2 } ( k_{i}a_s/2)}_. \label{gxchg}\ ] ] we extract @xmath354 from the condition that @xmath355 , _ i.e. _ @xmath356 with @xmath357 . the error of @xmath354 is estimated by adding the jack - knife error with bin size 5 and the variation over the fitting range of @xmath358 . keeping to the ansatz eq . ( [ potfit ] ) , we attempt three different fits : ( i ) 2-parameter fit with @xmath359 and @xmath360 fixed , ( ii ) 3-parameter fit with @xmath359 fixed , and ( iii ) 4-parameter fit . we check that @xmath354 from these three fits agree well within errors ( see fig . [ fig : r0fit ] ) . we adopt @xmath354 from the 2-parameter fit as our final value . results of @xmath354 at each @xmath55 are summarized in table [ tab : r0 ] . next we fit @xmath354 as a function of @xmath55 with the ansatz proposed by allton @xcite , @xmath361 where @xmath362 and @xmath363 is the two - loop scaling function of su(3 ) gauge theory , @xmath364 and @xmath365 parameterize deviations from the two - loop scaling . from this fit , we obtain that @xmath366 with @xmath367 . as shown in fig . [ fig : r0fit ] , the fit curves reproduce the data very well . we use eq . ( [ alltonrst ] ) in our later analysis . finally , we obtain @xmath5 from the input of @xmath368 fm . the values of @xmath5 at each @xmath55 are given in table [ tab : simparam ] . the quarkonium @xmath369 and @xmath370 splittings are often used to set the scale in heavy quark simulations since the experimental values are well determined and they are roughly independent of quark mass for charm and bottom . here we take the spin average for @xmath191 , @xmath287 and @xmath371 masses , so that the most of the uncertainties from the spin splitting cancel out . the lattice spacing at @xmath260 is given by @xmath372 where @xmath373 denotes the value in the temporal lattice unit . we use the data of @xmath374 and check that other choices do not change @xmath375 sizably . in table [ tab : m0ch - as ] we summarize the values of @xmath190 and @xmath375 for all @xmath267 including @xmath10 , and plot the @xmath55-dependence of @xmath375 in fig . [ fig : beta - as ] . we observe that @xmath376 holds for @xmath3106.35 . to show this explicitly , on the right we also plot the ratio @xmath377 and @xmath378 as a function of @xmath379 . deviations from unity are about @xmath3805% for @xmath377 , + ( @xmath38115)% for @xmath378 and hence + ( @xmath38125)% for @xmath382 at our simulation points . the major source of discrepancy among the lattice spacings from different observables is the quenching effect . another source is the uncertainty of input value of @xmath383 fm , which is only a phenomenological estimate . other systematic errors are expected for @xmath384 for the following reasons . our fitting for @xmath371 masses may be contaminated by higher excited states . in addition , the lattice size @xmath81.6 fm may be too small to avoid finite size effects for @xmath371 masses . on the other hand , the fitting for @xmath287 masses are more reliable , and we have checked that the finite size effects are negligible for @xmath270 in preparatory simulations ( see also ref.@xcite ) . for these reasons , we consider the scale @xmath385 to be the best choice for physical results on the spectrum . we present the results for three scales in the following , however , to show the dependence of the spectrum on the choice of the input for the lattice spacing . in order to make a comparison with the results by klassen and chen , who employ @xmath10 to set the scale , we use the results with @xmath379 . in fig . [ fig : comptuning ] , we plot the results of spin - averaged mass splittings and spin mass splittings for each choice of @xmath278 . the upper two figures show the spin - averaged splittings @xmath270 and @xmath271 , while the lower two show the s - state hyperfine splitting @xmath290 and the p - state fine structure @xmath291 . numerical values for each choice at @xmath386 are given in table [ tab : comptuning ] . here we set the scale with @xmath10 because it has the smallest statistical error . for all of mass splittings in fig . [ fig : comptuning ] , the results for @xmath387 well agree with those for @xmath388 , suggesting that the mass splittings are independent of the choice of @xmath6 whenever the @xmath193 tuning is adopted . this can be understood as follows@xcite . setting the measured kinetic mass to the experimental value @xmath389 for the meson roughly corresponds to setting @xmath390 for the quark , where the kinetic mass for the quark @xmath391 is given by eq . ( [ kineticmass ] ) at the tree level . since the spin - averaged splitting is dominated by @xmath391 , setting @xmath390 for each @xmath6 results in the same value for this splitting . with our choice of the spatial clover coefficient @xmath392 , @xmath393 also holds independent of @xmath6 at the tree level . hence the spin splitting takes approximately the same value because it is dominated by the magnetic mass @xmath394 given by eq . ( [ magneticmass ] ) . as a result , we practically have only two choices for @xmath278 , _ i.e. _ @xmath374 and @xmath395 . as observed in fig . [ fig : comptuning ] , however , the results for @xmath374 agree with those for the other choices at three finest @xmath5 , within a few @xmath288 for the hyperfine splitting and 1@xmath288 for other mass splittings . this shows that the choice @xmath374 is as acceptable as any other , with our numerical accuracy , for the lattices we adopted . since the hyperfine splitting for the choice @xmath374 has a smoother lattice spacing dependence ( at @xmath396 ) and a smaller error than that for other choices in fig . [ fig : comptuning ] , we decide to use the data with @xmath374 for the continuum extrapolations . the results for other choices are used to estimate the systematic errors . a slight bump in the lattice spacing dependence of the hyperfine splitting for @xmath397 is in part ascribed to the statistical error of @xmath195 itself , as discussed in sec.[sec : hfsrevisit ] . the results for charmonium spectrum , obtained for @xmath374 , for the three choices of scale are plotted in fig . [ fig : spectrum ] together with the experimental values , and numerical values are listed in tables [ tab : cc_r0]-[tab : cc_2s1s ] . as observed in fig . [ fig : spectrum ] , the gross features of the mass spectrum are consistent with the experiment . for example , the splittings among the @xmath398 states are resolved well and with the correct ordering ( @xmath399 ) . statistical errors for the 1s , 1p and 2s state masses are of 1 mev , 10 mev and 30 mev , respectively . when we set the scale from the @xmath400 ( @xmath401 ) splitting , the spin structure and the @xmath401 ( @xmath400 ) splittings are predictions from our simulations . we now discuss our results for the s - state hyperfine splitting @xmath290 , which is the most interesting quantity in this work . the hyperfine splitting ( hfs ) , arising from the spin - spin interaction between quarks , is very sensitive to the choice of the clover term , as noticed from eqs . ( [ nrhamiltonian ] ) and ( [ magneticmass ] ) . since the clover term also controls the lattice discretization error of the fermion sector , the calculation of the hfs is a good testing ground for the lattice quark action . in fig . [ fig : hfs ] we plot our results for the s - state hfs with @xmath374 for each scale input by filled symbols . from the @xmath321-linear continuum extrapolation using 3 points at @xmath258 - 6.35 , we obtain @xmath402 where the first error is the statistical error . the second error represents the ambiguity in the continuum extrapolation , estimated as the difference between the @xmath321-linear and the @xmath5-linear fits . the third error is the systematic error associated with the choice of @xmath278 . we estimate it from the maximum difference at the continuum limit between the choice of @xmath374 and the other three choices . our estimate of the s - state hfs is smaller than the experimental value by 27 % if the @xmath400 splitting is used to set the scale . a probable source for this large deviation is quenching effects . in this figure , we also plot previous anisotropic results by klassen ( set b in table [ tab : simparam3])@xcite and chen ( set c)@xcite at @xmath38 and 3 with the _ same _ choice of the clover coefficients @xmath1 and using @xmath10 to set the scale . the difference between our simulation and theirs is the choice of @xmath6 and the tadpole factor for @xmath1 , as noted in table [ tab : simparam3 ] . we use @xmath194 and the tadpole factor estimated from the plaquette @xmath403 , while they used @xmath195 and tadpole estimate from the mean link in the landau gauge @xmath404 . as shown in this figure , our result in the continuum limit with @xmath10 input agrees with the results by klassen@xcite and chen@xcite . the results with a _ different _ choice of the clover coefficients @xmath1 by klassen ( set d ) will be shown in sec.[sec : hfsrevisit ] , where we will study the effect of @xmath196 to the hfs . results for the p - state fine structure are shown in figs . [ fig : psp ] and [ fig : psp2 ] . the value of the p - state fine structure in the continuum limit and the systematic errors are estimated in a similar manner to the case of the s - state hfs . for @xmath405 splitting , we obtain @xmath406 note that the systematic errors from the choice of the fit ansatz ( second error ) are rather large here , due to the large scaling violation seen in fig . [ fig : psp ] . the result with the @xmath342 input yields a 17% ( 2.5@xmath288 ) smaller value than the experiment . our result with the @xmath10 input is consistent with the previous results by klassen@xcite and chen@xcite . for @xmath407 splitting , we obtain @xmath408 where we use the result from the e representation operator for @xmath44 . as observed in tables [ tab : cc_r0]-[tab : cc_2s1s ] , the mass difference @xmath409 is always consistent with zero , suggesting that the rotational invariance for this quantity is restored well in our approach . the value of @xmath410 is smaller than the experimental one by 23% ( @xmath411 ) with the @xmath342 input . there is no lattice result from the anisotropic relativistic approach to be compared with . next we consider the ratio of the two fine structures , @xmath412 . in fig . [ fig : ratp ] , we plot the lattice spacing dependence of this ratio . as shown in this figure , the scaling violation of the ratio is smaller than that for the individual splittings ( figs . [ fig : psp ] and [ fig : psp2 ] ) . moreover , results are always consistent with the experimental value within errors . presumably this is in part due to a cancellation of systematic errors such as the discretization effect and the quenching effect in the ratio . our continuum estimate of this ratio is @xmath413 our results agrees well with the experimental value . we omit the systematic error arising from the choice of @xmath278 , which is found to be much smaller than others . another interesting quantity is the p - state hyperfine splitting , @xmath414 , where @xmath415/9 $ ] . this should be much smaller than the s - state hyperfine splitting because the p - state wavefunction vanishes at the origin . the lattice spacing dependence is shown in fig . [ fig:1p13pavnl ] and the continuum estimate is @xmath416 the sign is always negative at finite @xmath5 and in the continuum limit , but within errors the continuum value is consistent with the experimental value . we do not observe sizable differences between results using different scale inputs for this quantity . the mass splittings between the orbital ( radial ) exited state and the ground state such as the @xmath369 ( @xmath370 ) splitting are dominated by the kinetic term in the non - relativistic hamiltonian eq . ( [ nrhamiltonian ] ) . since the dependence on the choice of @xmath278 is small compared to the statistical error , as shown in fig . [ fig : comptuning ] , we ignore the systematic error from the choice of @xmath278 in this and next subsections . results of the spin - averaged and spin - dependent @xmath369 splittings are shown in figs . [ fig : psavdife ] and [ fig : psdif ] . in the continuum limit , the spin - averaged @xmath369 splitting is @xmath417 the spin - dependent @xmath369 splitting deviates from the experimental value by 0 - 10% ( 1 - 5@xmath288 ) with the @xmath10 input and 15 - 25% ( 3 - 5@xmath288 ) with the @xmath370 input , as shown in fig . [ fig : psdif ] . the result of the @xmath418 splitting with the @xmath10 input agrees with the result by chen within a few @xmath288 in the continuum limit . in figs . [ fig : exdifsav ] and [ fig : exdifs ] , we show the results of the spin - averaged and spin - dependent @xmath370 splittings . in the continuum limit , these splittings deviate from the experimental values by @xmath820% ( 2.5@xmath288 ) with the @xmath10 input and @xmath830% ( 4@xmath288 ) with the @xmath342 input . for the spin - averaged @xmath370 splitting , we obtain @xmath420 besides quenching effects , possible sources of the deviations are finite size effects and the mixing of the @xmath371 with higher excited states . figure [ fig : exdifp ] shows the result for @xmath419 splittings . note that there is no experimental value for this splitting at present . our results of @xmath370 and @xmath419 splittings are consistent with previous results by chen . we also calculate mass splittings such as @xmath421 and @xmath422 , but these suffer from large statistical and systematic errors . we leave accurate determinations of the excited state masses for future studies . we summarize the continuum results for the charmonium spectra obtained with the data of @xmath374 and the @xmath321-linear fit ansatz in fig . [ fig : spectrumcont ] , where the scale is set by @xmath400 splitting . numerical values for three scales are listed in tables [ tab : cc_r0]-[tab : cc_2s1s ] , where the errors are only statistical . among three different scales , results with the @xmath369 input are the closest to the experimental value for the ground state masses . the spin splittings such as the hyperfine splitting @xmath290 and the fine structure @xmath291 are always smaller than the experimental values irrespective of the choice of the scale input , which is considered to be quenching effects . we now come back to the issue of the hyperfine splitting . in sec.[sec : hfs ] , we have shown that our result of the hfs ( set a in table [ tab : simparam3 ] ) agrees with previous results by klassen ( set b ) and chen ( set c ) in the continuum limit , with the same choice of the clover coefficients eqs . ( [ timasscb ] ) and ( [ ticst ] ) . however , as mentioned in the introduction , when klassen made a different choice of the clover coefficients ( set d ) , he obtained apparently different values of the hfs in the continuum limit . this choice is given by in the mass form notation eq . ( [ massform ] ) , while the correct choice @xmath423 corresponds to @xmath424 . ] @xmath425 where the tilde denotes the tadpole improvement , @xmath426 . since @xmath427 as @xmath428 , it agrees with the correct choice @xmath423 in the limit @xmath299 with fixed @xmath429 , but is incorrect at finite @xmath5 . the quark action then generates an additional @xmath306 error . even with such a choice , _ if _ @xmath305 is small enough , the result should converge to a universal value after the continuum extrapolation . however , in refs.@xcite , klassen obtained @xmath430 mev with @xmath425 , which is much larger than the result @xmath431 mev with @xmath423 both by klassen and in the present work . a possible source of this discrepancy is a large mass - dependent error of @xmath432 ( @xmath28 ) for the results with @xmath425 . in fact , klassen adopted rather coarse lattices with @xmath4332 , for which such errors may not be negligible . because the hfs is sensitive to the spatial clover term , the choice of @xmath425 may then result in a non - linear @xmath5 dependence for the hfs . in the following , in order to study the effect of the choice of the spatial clover coefficient @xmath33 to the hfs , we make a leading order analysis motivated by the potential model @xcite and compare it with numerical results , which will give us a better understanding of the above problem of the hfs . the potential model predicts that , at the leading order in both @xmath316 and @xmath12 , @xmath434 where @xmath435 for the quarkonium , @xmath436 are quark and anti - quark spins , and @xmath437 is the wavefunction at origin . @xmath438 is the hyperfine splitting in the continuum quenched ( @xmath439 ) theory , which is not necessarily equal to the experimental value . in non - relativistic qcd , the @xmath440 interaction arises from the @xmath96 term for quark and anti - quark . giving the non - relativistic interpretation to our anisotropic lattice action , we expect that the lattice hfs is effectively given by @xmath441 where @xmath394 is the magnetic mass eq . ( [ magneticmass ] ) in the effective hamiltonian . therefore , in our approach , hfs is dominated by the magnitude of @xmath442 , which depends on the spatial clover coefficient @xmath33 . the ratio , @xmath443 generally deviates from 1 at finite @xmath5 , and should approach 1 as @xmath299 . at the leading order in @xmath316 , @xmath444 , while @xmath445 with @xmath391 the kinetic mass eq . ( [ kineticmass ] ) . since @xmath391 does not depend on the spatial clover coefficient @xmath33 at the tree level , we neglect the lattice artifact for @xmath446 and set @xmath447 in the following , which is sufficient for the present purpose . now we define @xmath448 as a measure of lattice artifacts for the hfs , where the tilde denotes the tadpole improvement . in the continuum limit , @xmath449 . since @xmath429 is constant independent of @xmath5 , we identify @xmath429 with @xmath450 for the pole mass tuning ( _ i.e. _ when setting the measured pole mass to the experimental value @xmath451 for the meson ) , and with @xmath452 for the kinetic mass tuning ( @xmath389 ) . at the tree level with the tadpole improvement , the pole mass @xmath450 , the kinetic mass @xmath452 and the magnetic mass @xmath453 for the quark are given by @xmath454 where @xmath455 , @xmath426 , and @xmath456 is given by eq . ( [ tibearemass ] ) . to obtain eqs . ( [ tikineticmass ] ) and ( [ timagneticmass ] ) , we use the formula @xmath457 . in the following we present the @xmath305 dependence of @xmath458 in the case of @xmath459 ( set a , b , c ) and @xmath460 ( set d ) , and compare them with the corresponding numerical data for the s - state hfs . for the definition of @xmath6 ( or @xmath103 ) , there are two choices adopted so far , the tree level tadpole improved value @xmath194 and nonperturbative one @xmath195 . at @xmath461 , @xmath462 for the quark , but @xmath463 for the measured meson . on the other hand , at @xmath464 , @xmath465 though @xmath276 . thus in the case of @xmath464 , _ i.e. _ @xmath276 tuning , the identification of @xmath429 ( = @xmath450 or @xmath452 ) in @xmath458 eq . ( [ effhfs ] ) mentioned above is ambiguous . although such an ambiguity should vanish in the continuum limit , we present @xmath458 with both @xmath466 and @xmath467 to check consistency . for actual numerical data of the hfs , we focus on the results with the @xmath10 input because klassen has adopted @xmath10 for the scale setting . first we consider the case of @xmath425 ( set d ) , which is correct only for @xmath468 at the tree level . in fig . [ fig : effhfs_znp_tk ] we plot the @xmath469 dependence of @xmath458 at @xmath42 and 2 for @xmath425 with @xmath470 . numerical values of @xmath336 were taken from ref.@xcite . because of the ambiguity for @xmath429 mentioned above , we show the results with @xmath466 and @xmath467 ; the difference between them decreases as @xmath299 , as expected . we have checked that plotting @xmath458 as a function of @xmath321 , instead of @xmath469 , does not change the figure qualitatively . we also plot the results with @xmath425 but @xmath471 , where @xmath462 holds , as a dotted line ( @xmath42 ) and a dashed line ( @xmath38 ) for a guide to the eye . as shown in this figure , @xmath458 has a non - linear @xmath321 dependence toward the continuum limit ( = 1 ) , indicating that the mass dependent error is large for the region @xmath4722 . @xmath458 is larger than 1 even at @xmath473 , which suggests that the actual hfs should rapidly decrease toward @xmath474 , and data at @xmath475 are needed for a reliable continuum extrapolation for the hfs . now let us compare @xmath458 with numerical results of hfs . in fig . [ fig : hfsklassenlat98 ] , we plot corresponding results of hfs by klassen for @xmath425@xcite . the results at @xmath476 for @xmath425 are clearly larger than the results for @xmath423 ( see filled circles in fig . [ fig : hfs ] ) , and the results at @xmath476 and 2 _ appear _ to converge to @xmath477 95 mev in the continuum limit with an @xmath321-linear scaling . however , comparing fig . [ fig : effhfs_znp_tk ] and fig . [ fig : hfsklassenlat98 ] , we find that the lattice spacing dependence of the numerical data of hfs qualitatively agrees with that of @xmath458 : for both hfs and @xmath458 , data at @xmath476 are larger than data at @xmath478 , and the difference between @xmath476 and 2 decreases as @xmath299 . from an @xmath321-linear extrapolation of @xmath458 using the finest three data points , we obtain @xmath4791.3 at @xmath480 . because the correct continuum limit of @xmath458 is 1 , this suggests a @xmath48130% overestimate from the neglect of non - linear dependence of @xmath458 on @xmath321 . hence the result with @xmath425 , hfs(@xmath480 ) @xmath477 95 mev , reported in refs.@xcite is likely an overestimate by @xmath48130% . these analyses indicate that the origins of this overestimate are , first , the choice for the spatial clover coefficient @xmath482 , and second , the use of coarse lattices with @xmath324 . as shown in fig . [ fig : tunedzeta ] , @xmath103 ( @xmath483 in this case ) should eventually start to move up to 1 linearly around @xmath484 , which corresponds to @xmath485 in fig . [ fig : effhfs_znp_tk ] , but klassen s data of @xmath336 ( open diamonds ) do not reach such a region . we conclude that the continuum extrapolation for the hfs should not be performed using the data on such coarse lattices , and results at finer lattice spacing are required . next we consider the case of @xmath423 ( set a , b and c ) , which is correct for any @xmath305 at the tree level . in this case , there are two choices for @xmath6 , @xmath194 and @xmath195 . as mentioned in sec.[sec : effecttune ] , @xmath486 holds for both choices of @xmath6 , with @xmath423 . in the case of @xmath274 , which is adopted only in our work ( set a ) so far , @xmath487 is always satisfied , since @xmath488 by definition . this suggests that the scaling violation of hfs for @xmath423 should be much smaller than that for @xmath425 . the numerical result for the hfs with the pole mass tuning has already been shown in fig . [ fig : hfs ] and re - plotted in fig . [ fig : comp_hfs ] by filled circles , which gives our best estimate , hfs(@xmath480 ) @xmath489 mev . we next consider the case of @xmath275 , where @xmath253 for the measured meson . when we identify @xmath467 , @xmath487 is always satisfied again because @xmath490 even at @xmath275 . when we identify @xmath466 , @xmath491 in general , due to the deviation of @xmath195 from @xmath194 . the results of @xmath458 with @xmath466 at @xmath275 are shown in fig . [ fig : effhfs_znp ] , and corresponding numerical results for the hfs are shown in fig . [ fig : comp_hfs ] . comparing fig . [ fig : effhfs_znp ] with fig . [ fig : comp_hfs ] we again note that the lattice spacing dependence of the hfs qualitatively agrees with that of @xmath458 , _ i.e. _ , for both hfs and @xmath458 , data at @xmath476 by klassen ( open diamonds , set b ) and those at @xmath478 by chen ( open triangles , set c ) are close to each other and larger than our data at @xmath492 . an @xmath321-linear extrapolation using the finest three data points gives hfs @xmath49375 mev and @xmath4941.0 at @xmath480 . the latter confirms that a continuum estimate of hfs with @xmath423 is more reliable than that with @xmath425 . concerning our results at @xmath476 , as shown in fig . [ fig : effhfs_znp ] , @xmath458 for @xmath255 ( stars ) does not scale smoothly around @xmath495 , while that for @xmath492 ( filled circles ) is always unity . this behavior is caused by the fact that the difference , @xmath496 , is not monotonic in @xmath305 ( see fig . [ fig : tunedzeta ] ) . correspondingly the numerical value of the hfs , displayed in fig . [ fig : comp_hfs ] , also shows a slightly non - smooth lattice spacing dependence near @xmath497 , which qualitatively agrees with the @xmath469 dependence of @xmath458 in this region . a possible source of this behavior is the statistical error of @xmath195 itself , because hfs ( @xmath458 ) is also sensitive to the value of @xmath6 as well as @xmath33 . due to this reason , we have not used the results with @xmath275 for our main analysis in sec.[sec : results ] . in this article , we have investigated the properties of anisotropic lattice qcd for heavy quarks by studying the charmonium spectrum in detail . we performed simulations adopting lattices finer than those in the previous studies by klassen and chen , and made a more careful analysis for @xmath30 errors . in addition , using derivative operators , we obtained the complete p - state fine structure , which has not been addressed in the previous studies . from the tree - level analysis for the effective hamiltonian , we found that the mass dependent tuning of parameters is essentially important . in particular , with the choice of @xmath4 for the spatial wilson coefficient , an explicit @xmath498 dependence remains for the parameters @xmath6 and @xmath197 even at the tree level . moreover we have shown in the leading order analysis that , unless the spatial clover coefficient @xmath499 is correctly tuned , the hyperfine splitting has a large @xmath30 errors , which can explain a large value of the hyperfine splitting in the continuum limit from rather coarse lattices in the previous calculation by klassen . on the other hand , if @xmath499 is mass - dependently tuned , the continuum extrapolation is expected to be smooth for the hyperfine splitting . based on these observations , we employed the anisotropic clover action with @xmath4 and tuned the parameters mass - dependently at the tree level combined with the tadpole improvement . we then computed the charmonium spectrum in the quenched approximation on @xmath500 lattices with spatial lattice spacings of @xmath501 . a fine resolution in the temporal direction enabled a precise determination of the masses of s- and p- states which is accurate enough to be compared with the experimental values . our results are consistent with previous results at @xmath9 obtained by chen@xcite , and the scaling behavior of the hyperfine splitting is well explained by the theoretical analysis . we then conclude that the anisotropic clover action with the mass - dependent parameters at the tadpole - improved tree level is sufficiently accurate for the charm quark to avoid large discretization errors due to heavy quark . we note , however , that @xmath501 is still necessary for a reliable continuum extrapolation . we found in our results that the gross features of the spectrum are consistent with the experiment . quantitatively , however , the s - state hyperfine splitting deviates from the experimental value by about 30% ( 7@xmath288 ) , and the p - state fine structure differs by about 20% ( 2.5@xmath288 ) , if the scale is set from the @xmath400 splitting . we consider that a major source for these deviations is the quenched approximation . certainly further investigations are necessary to conclude that the anisotropic qcd can be used for quarks heavier than the charm . in particular it is important to determine the clover coefficients as well as other parameters non - perturbatively , since the spin splittings are very sensitive to the clover coefficients . it is also interesting to calculate the spectrum with @xmath502 and compare the result with the current one in this paper , since the notorious @xmath498 dependence vanishes from the parameters with this choice at the tree level . finally full qcd calculations including dynamical quarks are needed to establish the theoretical prediction without systematic errors for an ultimate comparison with the experimental spectrum . this work is supported in part by grants - in - aid of the ministry of education ( nos . 10640246 , 10640248 , 11640250 , 11640294 , 12014202 , 12304011 , 12640253 , 12740133 , 13640260 ) . vl is supported by jsps research for the future program ( no . jsps - rftf 97p01102 ) . kn and m. okamoto are jsps research fellows . m. okamoto would like to thank t.r . klassen for giving us his manuscript of ref.@xcite . mo also would like to thank a.s . kronfeld for useful discussions . the lattice hamiltonian @xmath503 is identified with the logarithm of the transfer matrix @xmath504 : @xmath505 @xmath504 and @xmath503 for the asymmetric clover quark action on the isotropic lattice have been derived in ref.@xcite . an extension to the anisotropic lattice is straightforward . using the fields @xmath506 and @xmath507 which satisfy canonical anti - commutation relations , the hamiltonian in temporal lattice units @xmath503 for the anisotropic quark action is given by @xmath508\hat{\psi}+o({\bf p}^3a_s^3),\ ] ] where @xmath509 , @xmath510 and @xmath511 @xmath512 and @xmath513 therefore the lattice hamiltonian in physical units is given by @xmath514\hat{\psi}+o({\bf p}^3a_s^2)\\ & = & \hat{\bar{\psi } } \left [ m_{1 } - \frac{\zeta_{\rm f}^{\prime } a_t}{2(1+m_0)}(r_s^{\prime } { \bf d}^2 + ic_s^{\prime}{\bf\sigma}\cdot{\bf b } ) -i\zeta_{\rm f}^{\prime } f_1(m_0)\theta - { \zeta_{\rm f}^{\prime}}^2 f_2(m_0)a_t\theta^2 \right]\hat{\psi}+o({\bf p}^3a_s^2),\label{anisoh}\end{aligned}\ ] ] where @xmath515 note that eq . ( [ anisoh ] ) for the anisotropic lattice is the same as that for the isotropic lattice except for use of \{@xmath516 } instead of \{@xmath517}. thus one can repeat the derivation of the tree level value of bare parameters ( @xmath518 and @xmath1 ) in ref.@xcite even for the anisotropic lattice , after replacing \{@xmath517 } by \{@xmath516}. when the lattice hamiltonian is expressed in more continuum - like form @xmath519 + \cdots ] \hat{\psi } , \label{contlikeh}\ ] ] the coefficients @xmath520 are given by @xmath521 in order to determine tree level parameters , the lattice hamiltonian should be matched to the continuum one to the desired order in @xmath5 . the continuum hamiltonian to which the lattice one is matched is either the dirac hamiltonian @xmath522 , or the non - relativistic hamiltonian @xmath523 . both choices give the same tree level parameters . in the hamiltonian formalism , the unitary transformation @xmath524 is possible because the eigenvalues of @xmath503 are invariant under it . for example , consider a unitary transformation , @xmath525 with @xmath526 where @xmath527 and @xmath528 are parameters . this is called the foldy - wouthuysen - tani ( fwt ) transformation , whose element is a spin off - diagonal matrix . after this transformation the coefficients @xmath520 become @xmath529 the transformed hamiltonian @xmath530 with @xmath531 is matched to either @xmath532 or @xmath533 so as to obtain tree level parameters . b. a. thacker and g. p. lepage , phys . d * 43 * , 196 ( 1991 ) . g. p. lepage , l. magnea , c. nakhleh , u. magnea and k. hornbostel , phys . d * 46 * , 4052 ( 1992 ) [ arxiv : hep - lat/9205007 ] . c. t. davies , k. hornbostel , a. langnau , g. p. lepage , a. lidsey , j. shigemitsu and j. h. sloan , phys . d * 50 * , 6963 ( 1994 ) [ arxiv : hep - lat/9406017 ] . c. t. davies , k. hornbostel , g. p. lepage , a. lidsey , p. mccallum , j. shigemitsu and j. h. sloan [ ukqcd collaboration ] , phys . d * 58 * , 054505 ( 1998 ) [ arxiv : hep - lat/9802024 ] . h. d. trottier , phys . d * 55 * , 6844 ( 1997 ) [ arxiv : hep - lat/9611026 ] . n. h. shakespeare and h. d. trottier , phys . d * 58 * , 034502 ( 1998 ) [ arxiv : hep - lat/9802038 ] . t. manke , i. t. drummond , r. r. horgan and h. p. shanahan [ ukqcd collaboration ] , phys . rev . d * 57 * , 3829 ( 1998 ) [ arxiv : hep - lat/9710083 ] . t. manke _ et al . _ [ cp - pacs collaboration ] , phys . lett . * 82 * , 4396 ( 1999 ) [ arxiv : hep - lat/9812017 ] . t. manke _ et al . _ [ cp - pacs collaboration ] , phys . d * 62 * , 114508 ( 2000 ) [ arxiv : hep - lat/0005022 ] . t. manke _ et al . _ [ cp - pacs collaboration ] , phys . d * 64 * , 097505 ( 2001 ) [ arxiv : hep - lat/0103015 ] . a. x. el - khadra , a. s. kronfeld and p. b. mackenzie , phys . d * 55 * , 3933 ( 1997 ) [ arxiv : hep - lat/9604004 ] . b. sheikholeslami and r. wohlert , nucl . phys . b * 259 * , 572 ( 1985 ) . z. sroczynski , nucl . * 83 - 84 * , 971 ( 2000 ) [ arxiv : hep - lat/9910004 ] . s. aoki _ et al . _ [ jlqcd collaboration ] , phys . rev . * 80 * ( 1998 ) 5711 . a. x. el - khadra , a. s. kronfeld , p. b. mackenzie , s. m. ryan and j. n. simone , phys . d * 58 * , 014506 ( 1998 ) [ arxiv : hep - ph/9711426 ] . c. w. bernard _ et al . _ , * 81 * , 4812 ( 1998 ) [ arxiv : hep - ph/9806412 ] . a. ali khan _ et al . _ [ cp - pacs collaboration ] , phys . d * 64 * , 034505 ( 2001 ) [ arxiv : hep - lat/0010009 ] . t. r. klassen , nucl . * 73 * , 918 ( 1999 ) [ arxiv : hep - lat/9809174 ] . klassen , unpublished . c. j. morningstar and m. j. peardon , phys . d * 56 * , 4043 ( 1997 ) [ arxiv : hep - lat/9704011 ] . c. j. morningstar and m. j. peardon , phys . d * 60 * , 034509 ( 1999 ) [ arxiv : hep - lat/9901004 ] . j. harada , a. s. kronfeld , h. matsufuru , n. nakajima and t. onogi , phys . rev . d * 64 * , 074501 ( 2001 ) [ arxiv : hep - lat/0103026 ] . s. aoki , y. kuramashi and s. tominaga , arxiv : hep - lat/0107009 . p. chen , phys . d * 64 * , 034509 ( 2001 ) [ arxiv : hep - lat/0006019 ] . the experimental values are taken from c. caso _ et al . _ [ particle data group collaboration ] , eur . j. c * 3 * , 1 ( 1998 ) . a. ali khan _ et al . _ [ cp - pacs collaboration ] , nucl . suppl . * 94 * , 325 ( 2001 ) [ arxiv : hep - lat/0011005 ] . s. aoki _ et al . _ [ cp - pacs collaboration ] , arxiv : hep - lat/0110129 . m. fujisaki _ et al . _ [ qcd - taro collaboration ] , nucl . . proc . suppl . * 53 * , 426 ( 1997 ) [ arxiv : hep - lat/9609021 ] . t. r. klassen , nucl . b * 533 * , 557 ( 1998 ) [ arxiv : hep - lat/9803010 ] . j. engels , f. karsch and t. scheideler , nucl . b * 564 * , 303 ( 2000 ) [ arxiv : hep - lat/9905002 ] . g. p. lepage and p. b. mackenzie , phys . d * 48 * , 2250 ( 1993 ) [ arxiv : hep - lat/9209022 ] . r. sommer , nucl . b * 411 * , 839 ( 1994 ) [ arxiv : hep - lat/9310022 ] . g. s. bali and k. schilling , phys . d * 46 * , 2636 ( 1992 ) . g. s. bali and k. schilling , phys . d * 47 * , 661 ( 1993 ) [ arxiv : hep - lat/9208028 ] . s. aoki _ et al . _ [ cp - pacs collaboration ] , phys . d * 60 * , 114508 ( 1999 ) [ arxiv : hep - lat/9902018 ] . c. r. allton , arxiv : hep - lat/9610016 . for example , see , w. lucha , f. f. schoberl and d. gromes , phys . * 200 * , 127 ( 1991 ) .	we present a detailed study of the charmonium spectrum using anisotropic lattice qcd . we first derive a tree - level improved clover quark action on the anisotropic lattice for arbitrary quark mass by matching the hamiltonian on the lattice and in the continuum . the heavy quark mass dependence of the improvement coefficients , _ i.e. _ the ratio of the hopping parameters @xmath0 and the clover coefficients @xmath1 , are examined at the tree level , and effects of the choice of the spatial wilson parameter @xmath2 are discussed . we then compute the charmonium spectrum in the quenched approximation employing @xmath3 anisotropic lattices . simulations are made with the standard anisotropic gauge action and the anisotropic clover quark action with @xmath4 at four lattice spacings in the range @xmath5=0.07 - 0.2 fm . the clover coefficients @xmath1 are estimated from tree - level tadpole improvement . on the other hand , for the ratio of the hopping parameters @xmath6 , we adopt both the tree - level tadpole - improved value and a non - perturbative one . the latter employs the condition that the speed of light calculated from the meson energy - momentum relation be unity . we calculate the spectrum of s- and p - states and their excitations using both the pole and kinetic masses . we find that the combination of the pole mass and the tadpole - improved value of @xmath6 to yield the smoothest approach to the continuum limit , which we then adopt for the continuum extrapolation of the spectrum . the results largely depend on the scale input even in the continuum limit , showing a quenching effect . when the lattice spacing is determined from the @xmath7 splitting , the deviation from the experimental value is estimated to be @xmath830% for the s - state hyperfine splitting and @xmath820% for the p - state fine structure . our results are consistent with previous results at @xmath9 obtained by chen when the lattice spacing is determined from the sommer scale @xmath10 . we also address the problem with the hyperfine splitting that different choices of the clover coefficients lead to disagreeing results in the continuum limit . making a leading order analysis based on potential models we show that a large hyperfine splitting @xmath895 mev obtained by klassen with a different choice of the clover coefficients is likely an overestimate .
magnetohydrodynamic ( mhd ) mechanisms are often invoked to explain the launching , acceleration and collimation of jets from young stellar objects , x - ray binaries , active galactic nuclei ( agns ) , microquasars , and quasars ( see , _ e.g. _ , * ? ? ? * and references therein ) . strongly magnetized jets , particularly those with a strong toroidal field encircling the collimated flow , are often referred to as `` current - carrying '' or `` poynting flux - dominated '' ( pfd ) jets . a large current flowing parallel to the jet flow is responsible for generating a strong , tightly wound helical magnetic field . the global picture of a current - carrying jet with a closed current system linking the magnetosphere of the central engine and the hot spots was introduced by @xcite and applied to agn double radio sources . this closed current system includes a pair of current circuits , each containing both a forward electric current path ( the jet flow itself , with its toroidal magnetic field , toward the lobe ) , and a return electric current path ( along some path back to the agn core ) . theory of magnetically driven outflows in the electromagnetic regime has been proposed by @xcite and @xcite and subsequently applied to rotating black holes @xcite and to magnetized accretion disks @xcite . an underlying large - scale poloidal field for producing the magnetically driven jets is almost universally assumed in many theoretical / numerical models . however , the origin and existence of such a galactic magnetic field are still poorly understood . in contrast with the large - scale field models , lynden - bell @xcite examined the `` magnetic tower '' ; expansion of the local force - free magnetic loops anchored to the star and the accretion disk by using the semi - analytic approach . global magnetostatic solutions of magnetic towers with external thermal pressure were also computed by @xcite using the grad - shafranov equation in axisymmetry . full three - dimensional mhd numerical simulations of magnetic towers have been performed by @xcite . recent x - ray and radio observations have revealed the dynamical interaction between the outbursts driven by agn and the background igm / icm , such as x - ray `` cavities '' with corresponding radio " bubbles @xcite . the cavities are believed to be filled with very low density relativistic plasma , inflated from the electromagnetic jets that are being powered by agns . this paper describes nonlinear dynamics of propagating magnetic tower jets in galaxy cluster scales ( @xmath0 tens of kpc ) based on three - dimensional mhd simulations to argue the jet / lobe transition @xcite and the stability properties @xcite . we solve the nonlinear system of time - dependent ideal mhd equations numerically in a 3-d cartesian coordinate system @xmath1 . the basic numerical treatments are introduced in @xcite . we assume an initial hydrostatic equilibrium in the gravitationally stratified medium , adopting an iso - thermal king model @xcite to model the magnetic towers from agns in galaxy cluster systems . agn accretion disk can not be resolved in our computational domain , and thus the magnetic flux and the mass are steadily injected in a central small volume during a certain time period . since the injected magnetic fields are not force - free , they will evolve as a `` magnetic tower '' and interact with the ambient medium . in the present paper , we present two different runs : one is called the `` unperturbed case '' in the following discussion , which is a run without any initial perturbation to the background initial profiles @xcite . the other is called the `` perturbed case '' , where a finite amplitude perturbation ( a few percent of the background sound speed ) is given to the velocities of the background gas @xcite . the total computational domain is taken to be @xmath2 corresponding to a ( 160 kpc:@xmath3 to @xmath4 kpc)@xmath5 box . the numerical grids are @xmath6 in the unperturbed case and @xmath7 in the perturbed case . normalizing factors are a length @xmath8 kpc , a sound speed @xmath9 cm s@xmath10 , a time @xmath11 yr , a density @xmath12 g @xmath13 . the corresponding unit pressure @xmath14 as @xmath15 dyn @xmath16 , and the unit magnetic field @xmath17 as @xmath18 @xmath19 g . the initial sound speed in the simulation is constant , @xmath20 , throughout the computational domain , which give a sound crossing time @xmath21 , corresponding to a typical time scale @xmath22 myr . therefore , @xmath23 is equivalent to the unit time scale @xmath24 myr . in the king model we use here , we adopt the cluster core radius @xmath25 to be @xmath26 ( i.e. , 20 kpc ) and the slope @xmath27 to be @xmath28 in the unperturbed case and @xmath29 in the perturbed case . magnetic fluxes and mass are continuously injected into a central volume of the computational domain for @xmath30 , after which the injection is turned off . a magnetic energy injection rate is @xmath31 ergs s@xmath10 , a mass injection rate is @xmath32/yr , and an injection time is @xmath33 myr . distribution of density in the @xmath34 plane along with the poloidal velocity ( _ left _ ) and three - dimensional view of selected magnetic field lines ( _ right _ ) for the perturbed case . ] during the dynamical evolution of magnetic tower jet , the narrow `` jet body '' and the expanded `` lobe '' are formed as shown in fig . 1 ( _ left _ ) . the 3d view of magnetic field lines as illustrated in fig . 1 ( _ right _ ) indicates that the magnetic tower jet has a well - ordered helical field configuration showing that a tightly wound central helix goes up along the central axis and a loosely wound helix comes back at the outer edge of the magnetic tower . the profiles of underlying external gas plays an important role in the transition of jet / lobe . the jet body is confined jointly by the external pressure and the gravity inside @xmath25 , while it expands radially beyond @xmath25 to form the lobe in a steeply decreasing ambient pressure . radial profiles of the forces ( _ left _ ) and the magnetic field components ( _ right _ ) along the @xmath35-axis for the unperturbed case . the position of the magnetic tower edge is shown . ] the jet axial current and the ambient gas pressure can determine the radius of the magnetically dominated lobes @xcite , that is , the internal gas pressure plays a minor role in the lobes which are typically seen in fr i radio galaxies @xcite . our numerical result shows the magnetically dominated lobes expand subsonically which can be seen in some observations @xcite . a preceding hydrodynamic shock wave beyond the tower , which may be identified as an agn - driven shock in recent x - ray observations @xcite . distribution of axial current density in the @xmath34 ( _ left _ ) for the perturbed case . time variation of the azimuthal fourier power of the current density with the axial ( @xmath36 ) wavelengths @xmath37 ( _ middle _ ) and @xmath38 ( _ right _ ) . ] the outer edge of the magnetic tower may be identified as a tangential discontinuity without the normal field component . the interior of tower ( lobe ) is separated from the non - magnetized external gas via this discontinuity . figure 2 shows the distributions of forces ( _ left _ ) and the magnetic field components ( _ right _ ) along the radial directions . at the tower edge , the outwardly directed magnetic pressure gradient force is roughly balanced with the inwardly directed thermal pressure gradient force . on the other hand , at the core part of jet body , the quasi - force free equilibrium is achieved . in the context of magnetically controlled fusion systems , the helical field in the magnetic tower can be regarded as the reversed field pinch ( rfp ) profile . the current - carrying magnetic tower jet , which possesses a highly wound helical field , is subject to the current - driven instability ( cdi ) . although the destabilizing criteria will be modified by the ambient gas and the rfp configuration of the tower , we find that the propagating magnetic tower jets can develop the non - axisymmetric cdi modes . both the internal elliptical ( @xmath39 ) mode ( fig . 3 _ right _ ) like the `` double helix '' and the external kink ( @xmath40 ) mode ( fig . 3 _ middle _ ) grow to produce the wiggles at the different parts ( fig . 3 _ left _ ) @xcite . this morphological feature ( the jet is suddenly disrupted at the lobe region with visible [ external ] wiggles ) can be observed in hercules a @xcite . by performing 3d mhd simulations , we have investigated the nonlinear dynamics of magnetic tower jets in the galaxy cluster system . under the gravitationally stratified atmosphere , the magnetic tower jet forms a global current closure system having the reverse field pinch magnetic configuration . the magnetic tower jet is well confined by the external gas pressure to form the collimated `` jet body '' inside the cluster core radius . the narrow jet expands radially to form the magnetically dominated `` lobe '' beyond the cluster core radius . thus , the profiles of underlying external gas play a important role in the transition of jets / lobes . the current - carrying tower jet , which possesses a highly wound helical magnetic field , is subject to the current - driven instability . in the gravitationally stratified atmosphere , the effect of thermal confinement on the magnetic tower jet will get gradually weaker as it grows . as a result , a separation of current flowing path between the jet axial current and the return current occurs . therefore , the edge of the jet axial current becomes a free boundary against the external kink mode to produce the visible wiggles as seen in some observations . this work was carried out under the auspices of the national nuclear security administration of the u.s . department of energy at los alamos national laboratory under contract no . de - ac52 - 06na25396 . it was supported by the laboratory directed research and development program at lanl and by igpp at lanl . benford , g. 1978 , , 183 , 29 benford , g. 2006 , , 369 , 77 blandford , r. d. 1976 , , 199 , 883 blandford , r. d. , & znajek , r. l. 1977 , , 179 , 433 blandford , r. d. , & payne , d. g. 1982 , , 199 , 883 croston , j. h. , hardcastle , m. j. , birkinshaw , m. , & worrall , d. m. 2003 , , 346 , 1041 fabian , a. et al . 2000 , , 318 , l65 fabian , a. et al . 2006 , , 366 , 417 forman , w. et al . 2005 , , 635 , 894 gitti , m. , mcnamara , b. r. , nulsen , p. e. j. , & wise , m. w. 2007 , , 660 , 1118 gizani , n. a. b. & leahy j. p. 2003 , , 342 , 399 kato , y. , mineshige , s. , & shibata , k. 2004 , , 605 , 307 kraft r. p. , azcona , j. , forman , w. r. , hardcastle m. j. , jones , c. , & murray , s. s. 2006 , , 639 , 753 king , i. 1962 , , 67 , 471 li , h. , lovelace , r. v. e. , finn , j. m. , & colgate , s. a. 2001 , , 561 , 915 li , h. , lapenta , g. , finn , j. m. , li , s. , & colgate , s. a. 2006 , , 643 , 92 lovelace , r. v. e. 1976 , nature , 262 , 649 lynden - bell , d. 1996 , , 279 , 389 meier , d. l. , koide , s. , & uchida , y. 2001 , science , 291 , 84 nakamura , m. , li , h. , & li , s. 2006 , , 652 , 1059 nakamura , m. , li , h. , & li , s. 2007 , , 656 , 721 wise , m. w. , mcnamara , b. r. , nulsen , p. e. j. , houck , j. c. , & david , l. p. 2007 , , 659 , 1153	we argue the behavior of poynting flux - dominated outflows from agn in the galactic cluster systems by performing three - dimensional mhd simulations within the framework of the `` magnetic tower '' model . of particular interests are the structure of mhd waves , the cylindrical radial force balance , the ( de)collimation , and the stability properties of magnetic tower jets . transition between the jet / lobe and the formation of wiggling jet by growing current - driven instability are discussed .
it is by now well understood that the presence of a small concentration of magnetic impurities into an otherwise non - magnetic metallic host can affect dramatically the low - temperature properties of the system . the prototype of these interesting phenomena is the kondo effect.@xcite at the single - impurity level , there is a non - perturbative crossover between a curie - weiss law behavior ( in which the impurity behaves like a non - interacting localized magnetic moment ) at high temperature , and a strongly interacting regime in which the magnetic impurity and the band electrons form a singlet ground state . in other words , in this low - temperature ( strong - coupling ) regime , the band electrons conspire to screen out the spin ( and magnetic moment ) of the impurity . at zero temperature a similar crossover occurs as a function of an external magnetic field . this picture has been developed by the concerted use of the renormalization group @xcite , exact " numerical simulations and scaling @xcite , exact solutions via the bethe - ansatz @xcite and large-@xmath2 expansions @xcite . from the point of view of scaling , the kondo problem is a typical situation in which a trivial fixed point , which describes band electrons decoupled from the magnetic impurity , is destabilized by a _ marginally relevant _ perturbation , the coupling to the magnetic impurity . this leads to an _ asymptotically free _ renormalization group flow with a @xmath5-function which is quadratic in the coupling constant . marginal perturbations appear in critical systems at a _ critical dimension_. the standard @xmath6 kondo hamiltonian is effectively a model of one dimensional chiral fermions coupled to a single magnetic impurity through the forward scattering channel @xcite . clearly , in this case we are at the _ lowest _ critical dimension . this is a direct consequence of the fact that the band electrons have a fermi surface where the density of states is finite and essentially constant . thus , the kondo effect is ultimately due to the availability of states in the electron band which can efficiently screen the impurity spin no matter how weak the exchange coupling constant may be . some time ago withoff and fradkin ( wf ) @xcite considered a genaralization of the kondo problem to systems in which the density of band electron states may actually go to zero at the fermi energy . they showed that if the density of states of the electron band vanishes at the fermi energy as a positive power of the energy , the kondo effect is suppressed for small values of the exchange constant and that the kondo screening only happens beyond a critical value of this coupling . in fact , it is easy to see that the exponent @xmath7 of the one particle density of states , @xmath8 plays a role here quite analogous to the distance to the lower critical dimension @xmath9 in critical phenomena . wf showed , using a combination of a poor man s scaling argument " and a large-@xmath2 limit , that at least for small values of the exponent @xmath7 , this is the correct picture . quite generally , if the exponent @xmath10 this is a non - marginal kondo system . there are a number of systems of physical interest where this situation does arise . a simple example are the fermionic excitations of a quantum antiferromagnet in a _ flux phase _ @xcite . more importantly , the normal state excitations of a @xmath11-wave superconductor ( with symmetry @xmath1 ) behave precisely in this fashion @xcite . in the vicinity of each _ node _ of the gap function ( hence , the use of the term node " hereafter ) , the dispersion relation for the normal quasiparticles is linear in the momentum . thus , sufficiently close to the node , the quasiparticles have an effective relativistic - like dispersion . in the theory of superconductivity ( isotropic or not ) @xcite the dynamics of the quasiparticles is usually pictured in terms of nambu spinors . for the case of a d - wave superconductor , nersesyan , tsvelik and wenger @xcite have shown that this approach leads to effective hamiltonian for the quasiparticles which takes the form of a massless dirac hamiltonian for each node of the gap , with the speed of light " equal to the fermi velocity . ( naturally , relativistic massless dirac fermions themselves always have this property . ) in this paper we consider a model which describes properly the coupling of flux - phase fermions to a magnetic impurity . it turns out that this model can also be used to describe the coupling of a local magnetic impurity to a d - wave superconductor , including pair - breaking effects . in a separate publication we will discuss in detail the problem of a magnetic impurity in a @xmath11-wave superconductor in more detail @xcite . in this paper we reconsider the kondo effect for non - marginal systems . the model has fermions coupled locally to a magnetic impurity . the fermions are assumed to obey a relativistic - like dispersion law and hence a density of states vanishing linearly with the energy . for simplicity we consider models with just one species of relativistic " fermions . we will refer to them as having a single node " . in particular we will discuss the case of an impurity coupled to an electron band with a density of states that vanishes _ linearly _ with the distance to fermi energy @xmath12 . this case was not examined by wf who found that the singularity structure changed as soon as @xmath13 . we will show in this paper that at @xmath14 the additional singularities conspire to give _ simple scaling laws _ modified by _ logarithmic corrections _ @xcite . this picture is strongly reminiscent of a critical system at an _ upper _ critical dimension . as in the conventional kondo problem , here too we can construct an effective one - dimensional theory . however , when one carefully reduces the @xmath4-dimensional fermions with a relativistic - like energy - momentum dispersion to an effective @xmath15-dimensional model , one finds that there are _ at least _ two angular momentum channels that are _ always _ coupled . thus , these are all _ inherently _ multichannel kondo systems . in particular , it is always the @xmath16 channel coupled either to @xmath17 channel or to @xmath18 channel . which pair of angular momentum channels are actually coupled depends on the _ parity _ of the node ( or cone ) to which the channels belong . we find two equivalent ways to represent the dimensionally reduced model . we have the freedom to choose the effective @xmath15-dimensional fermions to have a local kinetic energy and , hence , to behave like conventional right movers . however , in this picture , the effective interaction with the impurity becomes non - local . the other , alternative picture , is to have a local coupling between right moving fermions which now have a non - local kinetic energy . in fact , due to phase space factors , the density of states of the effective @xmath15-dimensional model goes to zero linearly with energy . the linearity of the density of states reflects the fact that in the original problem the fermions move in two - space dimensions . it is this feature what drives the system out of marginality and what generates a critical coupling constant below which the width of the kondo resonance vanishes . in all cases of physical interest , _ parity _ is an exact symmetry . this means that the number of nodes ( cones ) is _ even _ and that there should be as many cones with positive parity as there are with negative parity . in the case of both the @xmath11-wave superconductors and of the flux - phase fermions , this property follows from the alternating signs of the gap function . in contrast , if either parity or time reversal were broken , all four cones would have the same properties under parity . however , in that case there would always be a non - zero energy gap on the entire fermi surface . thus , in general , the effective kondo hamiltonians always have an exact degeneracy ( angular momentum channels ) . the spin symmetry is the usual @xmath19 spin rotation invariance ( which here we call _ color _ ) . here we will work with the @xmath20 generalization of this symmetry , with the physical @xmath21 . the angular momentum degeneracy leads to and @xmath22 _ flavor _ degeneracy . for a problem with one node , @xmath23 . when more than one node is considered , @xmath24 . we will also consider impurity scattering amplitudes which may change the angular momentum of the fermions . the basic and simplest model is worked out in section [ sec : toy ] . in the second part of the paper we use large-@xmath2 methods to investigate the behavior of these systems near the critical coupling @xcite . we find that there is still a kondo scale once we go over the critical coupling . however , the position and the width of the resonance are not related anymore in the very simple way they are in the usual kondo effect . the details are worked out in section [ sec : scaling ] . the most salient feature of our results is the presence of logarithmic corrections to simple scaling in all quantities of physical interest , including the zero - field paramagnetic suceptibility ( at @xmath25 ) , the kondo scale and the @xmath3-matrix for bulk fermions . the large-@xmath26 theory that we present here suggests the following scenario . for small values of the coupling constant @xmath27 , the impurity is free and effectively paramagnetic . in contrast , for large values of @xmath27 , the impurity is screened by the fermions . the critical coupling @xmath28 is dimensionful and scales with the energy cutoff . however , the fact that these systems are inherently multichannel kondo systems suggests that for @xmath29 the impurity is actually _ overscreened _ @xcite . in the language of the renormalization group this requires a flow with two finite fixed points : one _ infrared unstable _ fixed point for the paramagnetic - kondo phase transition and an _ infrared stable _ to describe the multichannel behavior of the kondo phase . as in the conventional kondo problem , the large-@xmath26 theory can only describe the formation of the kondo singlet which in a non - marginal system can only take place at finite coupling . in marginal kondo systems multichannel behavior is found as a next - to - leading order effect in the @xmath30 expansion or to third order in perturbative scaling @xcite . in the systems of physical interest mentioned above where this model applies , the number of channels @xmath31 is always larger than the number of colors . in such a situation one can imagine that there exists a critical number of channels such that both fixed points actually coincide and below this critical number the transition dissapears altogether @xcite . we should emphasize here that this physics can not be accessed by a straightforward use of the large-@xmath26 expansion . the paper is organized as follows . in section [ sec : toy ] we derive the effective @xmath15-dimensional impurity models for systems of fermions in two space dimensions with relativistic - like dispersion ( flux phase ) . in section [ sec : scaling ] we study these models in the large-@xmath2 limit and determine their critical behavior . in section [ sec : prop ] we calculate the propagator for the band fermions in the @xmath32 limit and use it to derive the @xmath3-matrix . in section [ sec : genimp ] we consider a model with the most general type of scattering process for fermions with only one node coupled to a single impurity and determine the phase diagram . section [ sec : conc ] is devoted to the conclusions and to the discussions of the similarities and differences between the problem we discuss here and the conventional kondo problem . in appendix a we sketch the calculations of a few integrals . in this section we want to study the s - d hamiltonian a la read and newns @xcite , with the difference that we are not assuming a constant density of states for the conduction fermions . we are coupling an f - impurity to a bath of relativistic " electrons , _ ie . , _ electrons with a linear dispersion relation . we assume the chemical potential for the electrons ( the fermi energy ) is zero . this is a very special case but it will turn out to be quite important for the case of a d - wave superconductor as it will be discussed in the following sections . we start with a two - dimensional free fermion hamiltonian with a linear spectrum and a fermi velocity @xmath33 which may correspond to the linearization of a fermion band of width @xmath34 , where @xmath35 is an energy cutoff . we will consider a model of band fermions with @xmath36 spin components with @xmath37 for the physically relevant case of spin @xmath19 . in general , this model will have a global @xmath20 color " ( or spin ) symmetry . this model has a well defined large-@xmath2 limit . the free part of the hamiltonian is @xmath38 where @xmath39 and @xmath40 are pauli matrices , so that @xmath41 eq.([eq : hdirac ] ) can be diagonalized by expanding the fields into a linear combination of spinor eigenstates . let s call them @xmath42 . this eigenmodes satisfy @xmath43 a particular ( and convenient ) choice of eigenstates is given by @xmath44 the fields @xmath45 s carry a spin index @xmath46 apart from the spinor index @xmath47 and @xmath48 which we ignore at this point but we shall put it back in when dealing with the interaction with the impurity term . @xmath49 so that @xmath50 \nonumber \\ \psi_2(p ) & = & \frac{1}{\sqrt{2 } } \left [ \zeta_{+}(p ) - \zeta_{-}(p ) \right ] e^{i\theta}\end{aligned}\ ] ] now we can expand the @xmath51 fields in a basis of angular momentum eigenmodes @xmath52 where @xmath53 is an integer and @xmath54 indicates the positive / negative energy spinor eigenmodes of the dirac equation given by eq.([eq : h0 ] ) . we then have @xmath55 \label{eq : ke}\ ] ] in real space we have @xmath56 to reduce to an equivalent one - dimensional problem we use polar coordinates where @xmath57 , @xmath58 and @xmath59 . by making use of the expansion @xmath60 where @xmath61 are the bessel functions . they satisfy the relations @xmath62 using these in eq.([eq : psir ] ) one gets @xmath63 \nonumber \\ \psi_2(r , \phi ) & = & \int_0^{\infty } \frac{pdp}{2\pi } \frac{1}{\sqrt{2 } } \sum_{m=-\infty}^{\infty } i^{m+1 } { j}_{m+1}(pr ) \ e^{-i(m+1)\phi } \left [ \zeta_{m , + } ( p ) - \zeta_{m , -}(p ) \right]\end{aligned}\ ] ] at the impurity site , _ i.e. , _ for @xmath64 , @xmath65 and @xmath66 for @xmath67 . so , we are left with @xmath68 \nonumber \\ \psi_2(0 ) & = & \int_0^{\infty } \frac{pdp}{2\pi } \frac{1}{\sqrt{2 } } \left [ \zeta_{-1 , + } ( p ) - \zeta_{-1 , -}(p ) \right ] \label{eq : psi0}\end{aligned}\ ] ] this is telling us that the effect of having a cone - like dispersion relation for the fermions instead of a flat band will not only reduce the available density of states but it will also induce an angular momentum mode mixing . even though we may regard the impurity as having a @xmath69-function spatial form factor as in the case of the usual kondo effect , the coupling to the fermions will no longer be restricted only to the s - wave , or the spherically symmetric angular momentum channel . this feature will become evident in the following step . we want to rewrite the theory by introducing effective one - dimensional fermion operators . we use the radial component of incoming and outgoing waves . we find two flavors " of right - movers given by @xmath70 in other words , @xmath71 are the fermion states with angular momentum @xmath16 while @xmath72 are the fermion states with angular momentum @xmath18 . we summarize these mappings as @xmath73 for a system with an @xmath20 symmetry , the s - d interaction can be written in the form @xmath74 where @xmath75 is the impurity spin coupled to the electrons at the real - space position @xmath76 . for a system with an @xmath20 symmetry , @xmath77 . we will use the standard representation of spin operators ( here , the generators of @xmath20 defined in terms of the fermion operators @xmath78 as @xmath79 . if the constraint @xmath80 is satisfied , the impurity is in the lowest ( fundamental ) representation of @xmath20 . other representations can be constructed by changing the filling fraction " of the impurity spin . notice that , for @xmath19 , the only possible value of the filling fraction is equal to one . once again we stress the fact that the angular momentum channels given by @xmath81 and @xmath82 couple to the impurity . this is an important difference with respect to the case of constant density of states , where only the @xmath81 channel couples to the impurity . we shall single out these two channels from the free part of the hamiltonian to obtain an effective theory by replacing eq . ( [ eq : psi0 ] ) into eq . ( [ eq : s - d ] ) @xmath83 \nonumber \\ & + & \ \int_{0}^{\infty } \frac{p \ dp}{2\pi } v_f \ \left [ \zeta^{\dagger}_{-1 , + } ( p)\zeta_{-1 , + } ( p ) - \zeta^{\dagger}_{-1 , - } ( p ) \zeta_{-1 , - } ( p ) \right ] \nonumber \\ & + & \frac{g}{2 } \int_{0}^{\infty } \frac{p \ dp}{2\pi } \int_{0}^{\infty } \frac{q \ dq}{2\pi } \big [ \zeta^{\dagger}_{0 , + } ( p ) \tau^ { a } \zeta_{0 , + } ( q ) + \zeta^{\dagger}_{0 , - } ( p ) \tau^ { a } \zeta_{0 , - } ( q ) \nonumber \\ & & \qquad \qquad+ \ \ \zeta^{\dagger}_{0 , + } ( p ) \tau^ { a } \zeta_{0 , - } ( q ) + \zeta^{\dagger}_{0 , - } ( p ) \tau^ { a } \zeta_{0 , + } ( q ) \big ] \ s^ { a } \nonumber \\ & + & \frac{g}{2 } \int_{0}^{\infty } \frac{p \ dp}{2\pi } \int_{0}^{\infty } \frac{q \ dq}{2\pi } \big [ \zeta^{\dagger}_{-1 , + } ( p ) \tau^ { a } \zeta_{-1 , + } ( q ) + \zeta^{\dagger}_{-1 , - } ( p ) \tau^{a } \zeta_{-1 , - } ( q ) \nonumber \\ & & \qquad \qquad - \ \ \zeta^{\dagger}_{-1 , + } ( p ) \tau^{a } \zeta_{-1 , - } ( q ) - \zeta^{\dagger}_{-1 , - } ( p ) \tau^{a } \zeta_{-1 , + } ( q ) \big ] \ s^ { a}\end{aligned}\ ] ] by using the correspondence between the @xmath51 s and @xmath84 s defined above we may write the hamiltonian in a more compact form . since we found that this model contains effectively two flavors " of right movers , we define a flavor index @xmath85 with @xmath23 . the system has an effective @xmath19 flavor " symmetry which originates in the unavoidable mixing of angular momentum waves by the impurity . similarly , we will write down the spin index @xmath86 explicitly . the effective hamiltonian is ( repeated indices are summed ) @xmath87 \tau^{a}_{\sigma \omega } \left [ \int^{\infty}_{-\infty } \frac{dq}{2\pi } \sqrt{\|q\| } \ c^{\dagger}_{l \omega}(q ) \right ] f^{\dagger}_{\alpha } \tau^{a}_{\alpha\beta } f_{\beta } \label{eq : heff}\ ] ] where @xmath88 . the single particle excitation energy is @xmath89 . ( [ eq : heff ] ) shows that the coupling between the impurity spin and the band electrons is momentum dependent . we note here that there is an alternative representation of this model . it may be obtained by , instead of rescaling the fields by a factor of @xmath90 as we have done here , defining a new momentum variable @xmath91 ( an later extending @xmath92 to @xmath93 ) . in this form , the effective hamiltonian becomes @xmath94 \tau^{a}_{\sigma \omega } \left [ \int^{\infty}_{-\infty } \frac{dk ' } { 2\pi } \ c^{\dagger}_{l \omega}(k ' ) \right ] f^{\dagger}_{\alpha } \ \tau^{a}_{\alpha\beta } f_{\beta } \label{eq : halt}\ ] ] in this case @xmath95 . in this representation , the interaction becomes local at the expense of a non - local kinetic energy . in fact these two representations are equivalent and are the only ones compatible with local ( anti)-commutation relations for the @xmath45-fields . this is the representation used by withoff and fradkin @xcite . in this paper we will not use this representation of the model . we may now proceed to study this theory with a functional integral formalism@xcite . the lagrangian , in imaginary time , is @xmath96 \tau^{a}_{\sigma\omega } \left [ \int^{\infty}_{-\infty } \frac{dq}{2\pi } \sqrt{\|q\| } \ c_{l \omega } ( q ) \right ] f^{\dagger}_{\alpha } \ \tau^ { a}_{\alpha\beta } \ f_{\beta } \nonumber \\ & & \qquad + f^{\dagger}_{\sigma } \left ( \frac{\partial}{\partial\tau } - { \hat h}_{\sigma } \right ) f_{\sigma } + \epsilon_f(\tau ) \left ( f^{\dagger}_{\sigma } \ f_{\sigma } - q_f \right ) \label{eq : lag}\end{aligned}\ ] ] in eq.([eq : lag ] ) we have defined @xmath97 . we have also included a magnetic field @xmath98 which we will choose to be a diagonal matrix @xmath99 and it will be defined below . the lagrange multiplier field @xmath100 has been introduced to enforce the constraint of charge ( occupancy ) @xmath101 at the impurity site . in principle , the partition function separates into a sum of subsectors each of which is characterized by an impurity occupancy @xmath101 . following the standard large-@xmath26 decoupling approach of read and newns @xcite , we write the spin operators in terms of fermions and find an effective four fermi interaction . this interaction can be written in a simple form by making use of the well known identity , which holds for the generators of @xmath20 @xmath102 a hubbard - stratonovich ( h - s ) transformation is now introduced to decouple the fermionic quartic term which arises from this expansion . up to an integration over the h - s fields @xmath103 and @xmath104 , eq.([eq : lag ] ) is equivalent to @xmath105 eq.([eq : h - slag ] ) can be rearranged , by field shifting and completing squares , in the form @xmath106 \ \left ( \frac{\partial}{\partial\tau } + e(p ) \right ) \nonumber \\ & & \kern2.5 in \left [ c_{l \sigma } ( p ) + \left ( \frac{\partial}{\partial\tau } + e(p ) \right)^{-1}\kern-.1 in \sqrt{\|p\| } \varphi_{l}(\tau ) \ f_{\sigma } ( \tau ) \right ] \nonumber \\ & + & f^{\dagger}_{\sigma } ( \tau ) \left [ \frac{\partial}{\partial\tau } - { \hat h}_{\sigma } + \epsilon_f(\tau ) - \varphi^{}_{l}(\tau ) \int_{-\infty}^{\infty } \frac{dp}{2\pi } \ \|p\| \left ( \frac{\partial}{\partial\tau } + e(p ) \right)^{-1 } \kern-.1 in \varphi_{l}(\tau ) \right ] f_{\sigma } \nonumber \\ & & \kern2 in + \ \frac{n_c}{j_0 } \sum_{l=1}^{n_f } - q_f \epsilon_f(\tau ) \label{eq : q - lag}\end{aligned}\ ] ] we now obtain an effective theory for the fields @xmath107 and @xmath108 by integrating out the fermions . the partition function is given by @xmath109 where @xmath110\ ] ] is the partition function of free fermions . the impurity part of the effective action is @xmath111 \nonumber \\ & & \kern2 in \ + \ \int d\tau \left ( \frac{n_c}{j_0 } \left ( \sum_{l=1}^{n_f } \|\varphi_{l}\|^2\right ) - q_f \epsilon_f \right ) \label{eq : seff}\end{aligned}\ ] ] here we stress that the model of physical interest has @xmath21 and @xmath23 . the effective action of eq . ( [ eq : seff ] ) has the standard form of reference @xcite . the key difference here is the form of the free fermion green s function which in this problem has a relativistic form . for the usual kondo problem the magnetic impurity is coupled to system of band electrons with a constant density of states at the fermi surface . in the model that we discuss here , the density of states of the effective fermions ( the right movers " ) is still constant but the interaction with the magnetic impurity has an explicit momentum dependence . this momentum dependence is such that the effective coupling at low momenta becomes arbitrarily small . we will show below that up to a critical value of the coupling constant @xmath112 there is no kondo effect . this is in fact the result of reference @xcite . notice that the momentum dependence of the interaction , is a direct consequence of the relativistic dispersion . should any finite density of states arise , either by effects of a chemical potential or induced by disorder , a crossover to a conventional kondo effect will occur . there is an important physical case in which a finite density of states is precluded by reasons of symmetry , and the fermi energy has to be locked at zero . this is the case of the d - wave superconductors which we will discuss elsewhere @xcite . we now consider the @xmath113 limit . here , _ @xmath26 means the limit in which the rank of the group of spin rotations becomes large instead of being @xmath19 . to proceed with the @xmath30 expansion , we look first for static solution for @xmath114 . from the @xmath22 flavor symmetry of the effective hamiltonian , there is a manifold of solutions which span the group @xmath22 . clearly , all the solutions break @xmath22 _ spontaneously_. as in all impurity problems , it is impossible to break spontaneously a continuous symmetry of bulk fermions by coupling them to an impurity , which has a finite hilbert space . thus , this apparent spontaneous symmetry breaking is an artifact of the @xmath113 limit . in fact , we expect that it will already be restored by the leading @xmath30 correction . this is precisely what happens in the large-@xmath2 approach to the conventional kondo problem @xcite . thus , quantities which exhibit this apparent spontaneous symmetry breaking will get strongly corrected already in the next order in @xmath30 . in what follows we will seek a static , symmetric , solution of the form @xmath115 . we will now derive the form of the saddle point equations which will determine @xmath116 as a function of @xmath112 and of the filling fraction of the impurity @xmath101 . first we need to compute @xmath117 by working in imaginary frequency we automatically get the time - ordered expression form for eq.([eq : seff ] ) . to compute eq.([eq : g0 ] ) we introduce a lorentzian cutoff function @xmath118 and extend the integration over @xmath119 to @xmath120 . we have @xmath121 we can go back to eq.([eq : seff ] ) which now reads @xmath122 + \beta \left ( \frac { n_c n_f\|\varphi_0\|^2}{j_0 } - q_f \epsilon_f \right)\ ] ] where @xmath123 and the magnetic field @xmath124 has been set to zero . the saddle point equations are @xmath125 where we should understand the integral as computed using some convenient adiabatic cutoff . the non - trivial solution for @xmath126 is given by @xmath127 it will prove useful to define the dimensionless variables @xmath128 and @xmath129 . the saddle point equations eq.([eq : phasesft ] ) and eq.([eq : j0 ] ) now read @xmath130 ^ 2}\ ] ] and @xmath131 where @xmath132 . there exists a critical value for @xmath112 , which we define as the value of @xmath112 at the point where @xmath126 departs from zero , for vanishing @xmath107@xcite . it is given by @xmath133 in the next section we use the spe to extract the critical behavior of the system near @xmath28 . in the previous section we obtained the saddle - point equations ( spe ) for this theory ; however , these spe have a singular behavior around the point @xmath134 , @xmath135 . in particular , an expansion in powers of @xmath136 for small @xmath136 is not possible for @xmath137 . in this section we want to investigate in further detail the behavior of the spe and the scaling behavior of @xmath136 as we approach the critical point . we will find that , opposite to the situation in the usual kondo problem , there are now two independent energy scales for @xmath136 and @xmath138 . we go back to the spe expressed in their original form @xmath139 and @xmath140 where @xmath141 and @xmath142 . in eq.([eq : scaling1 ] ) , the integrand is the equal time propagator for the impurity , with the integration variable @xmath143 being the imaginary frequency scaled by the band electrons cutoff @xmath144 . this equation sets up a scale for a crossover between two different behaviors , in much the same way as in the usual kondo effect the imaginary part of the band electrons green function sets the scale of the kondo temperature . in the case of the usual kondo effect , the expression equivalent to eq.([eq : scaling1 ] ) is @xmath145 the case @xmath146 is , in a sense , a limiting case as can be seen from eq.([eq : scaling3 ] ) . on one hand , the r. h. s. of this equation is a positive function . this fact forces @xmath147 to be negative in order to have a solution . however , even in this case one gets @xmath148 where @xmath144 is some electron band cutoff . this suggests that for @xmath149 , the impurity level @xmath150 has to approach zero . however , if @xmath151 , there is no solution , since @xmath152 is a strictly positive number . now , if @xmath26 is an integer bigger than 2 , and letting @xmath153 , we find @xmath154 here we can distinguish two different regimes : @xmath155 which corresponds to the case @xmath156 , and @xmath157 which corresponds to @xmath158 . however , for the system being discussed here , the situation is a little different and , actually , more complex since , as it turns out , now there are two different scales involved . in principle there is a scale set on @xmath143 by @xmath136 at the point where @xmath159 if the value of the frequency ( and therefore of @xmath143 ) is small enough so that @xmath160 , we can approximate @xmath161 for @xmath136 small enough , the approximation is consistent . working on this idea , we split eq.([eq : scaling1 ] ) as @xmath162 where we have used the fact that the only solution consistent with @xmath163 has @xmath164 . the spe for @xmath136 may be treated using a similar approach . eq.([eq : scaling2 ] ) can be split into @xmath165 and @xmath166 , where the first one is the corresponding integral up to @xmath167 , and the second one takes over from that point to infinity . again we are interested in the small @xmath136 and small @xmath138 regime . the detailed computation of the integrals @xmath168 , @xmath169 , and @xmath165 , @xmath166 is given in appendix [ sec : a ] . we will be interested in the following limiting cases : ) @xmath170 : + in the regime in which @xmath171 , with @xmath172 , the contribution from @xmath169 is neglegible and @xmath173 or @xmath174 . thus , in this regime , @xmath170 . in this limit , the leading term from the other spe is going to be ( see appendix [ sec : a ] , eq.([eq : applast ] ) ) @xmath175 \label{eq : nu3}\ ] ] this gives @xmath176 } \ - \ \frac { \log \log \left [ \frac { \left ( \pi \frac{q_f}{n_c}\right)^2 } { \frac{\pi^2}{4 } - \frac{1}{g_0 } } \right ] } { \log^2 \left [ \frac { \left ( \pi \frac{q_f}{n_c}\right)^2 } { \frac{\pi^2}{4 } - \frac{1}{g_0 } } \right ] } \ + \dots \label{eq : delta2}\ ] ] and @xmath177 } \label{eq : nu2}\ ] ] @xmath178 : + we now consider the opposite regime , @xmath179 , where we obtain @xmath180 in this regime , clearly @xmath181 and the other spe gives @xmath182 we can get a solution for @xmath138 by iteration on eq.([eq : scaling22 ] ) , @xmath183 } \ \frac{2}{\pi } \left ( \frac{\pi^2}{4}-\frac{1}{g_0}\right ) \label{eq : nu1}\ ] ] using eq.([eq : scaling16 ] ) we get @xmath138 which , after being replaced into eq.([eq : scaling22 ] ) gives a scaling form for @xmath136 in this regime @xmath184^{1/3 } \approx \ e^{-1/\delta } \log \left [ \frac{\pi}{2\left(\frac{\pi^2}{4 } - \frac{1}{g_0}\right ) } \right ] \label{eq : scaling23}\ ] ] thus , @xmath185 \right]^{-1}\kern-5pt \left ( 1 - \frac{\log \log \left [ \frac{2}{\pi } \left ( \frac{\pi^2}{4 } - \frac{1}{g_0 } \right ) \right]^{-1}}{\log \left [ \frac{2}{\pi } \left ( \frac{\pi^2}{4 } - \frac{1}{g_0 } \right ) \left(\frac{\pi}{3}\left(\frac{1}{2}-\frac{q_f}{n_c}\right)\right)^{1/3 } \right]^{-1 } } \right ) \label{eq : delta1}\ ] ] notice that , for the physical case @xmath21 , the two regimes @xmath186 and @xmath187 are identical . only as @xmath26 grows large it becomes possible to distinguish one case from the other . however , in the case @xmath188 the magnitude of @xmath136 appears to go to zero as @xmath189 even though the critical coupling is independent of the value of @xmath101 . in contrast , in the case @xmath186 such unwanted feature is not present . it is worth to note here that , in spite of this apparent difficulty , we will show at the end of this section that the zero - temperature suceptibility at zero field has the same behavior in both regimes . thus , this difficulty is not physically relevant . for simplicity , we will use the case @xmath186 to extrapolate to the physically meaningful case of @xmath21 . we can use these results to derive the @xmath5-function for the coupling constant @xmath190 in the limit @xmath113 . in order to do this we may use eq.([eq : scaling22 ] ) for the regime in which @xmath191 , or eq.([eq : nu2 ] ) for the case @xmath192 . starting from eq.([eq : nu3 ] ) we may replace @xmath136 in terms of @xmath138 to obtain @xmath193 where @xmath194 . the @xmath5-function keeps track of the flow of @xmath190 as the cutoff @xmath144 is decreased from very large values ( infinite bandwidth ) . hence , @xmath195 in both limits @xmath196 and @xmath197 we find @xmath198 where @xmath199 . we immediately see that there are two fixed points : ( a ) a stable fixed point at @xmath200 and ( b ) an unstable fixed point at @xmath201 . the stable fixed point represents the weak coupling phase in which the fermions are decoupled from the impurity and the impurity spin is unscreened and there is no kondo effect . the non - trivial fixed point at @xmath201 separates the weak coupling phase from a strong coupling phase that , in principle , should exhibit a kondo effect . this unstable fixed point can be regarded as the usual marginally unstable fixed point of the standard kondo problem , now pushed to a finite value of the coupling constant by effect of the reduction of the density of band states . hence , in the present case , the interaction has to become strong enough so as to overcome the effect of the depletion in the number of states available in order to produce the kondo screening . thus , for systems with a small coupling to the impurity , there will be no kondo screening and we should expect a decoupled impurity behavior . if @xmath202 , the system is in the kondo screeening phase . we may define a _ kondo scale _ @xmath203 for the regime @xmath204 . from eq.([eq : nu3 ] ) through eq.([eq : nu2 ] ) we see that , in terms of the physical parameters of the theory we have @xmath205\right)^{-1}\ ] ] the results found here generalize the work of wf @xcite . in that work the case of a density of states which vanishes linearly was not discussed on account that the structure of the singularities was not smoothly connected with the case of a constant density of states . the results of this section show that the main difference is the presence of logarithmic corrections to scaling in all the physical quantities . the case of a magnetic impurity coupled to a fermi system with a linear density of states seems to be analogous to the behavior of critical systems at their upper critical dimension . in contrast , the case of the constant density of states is the analog of a critical system at the lower critical dimension ( marginal instability ) . the closest analog to this problem is the critical behavior of the gross - neveu model in @xmath206 ( space - time ) dimensions @xcite . the results that we find here hold for the simple regime of large @xmath26 which does not describe correctly the dynamics of the system with more than one fermion flavor . at this leading order in @xmath30 , the existence of several channels is only reflected into a trivial degeneracy and it does not lead to any new physics . however , as has been shown by blandin and nozires @xcite , already the leading corrections in @xmath30 will lead to non - trivial behavior in the screened " phase , @xmath207 . in the standard kondo problem , this behavior was confirmed by both the bethe - ansatz solution @xcite and conformal field theory @xcite . we note here that , due to the non - local character of the effective one - dimensional theory discussed in section [ sec : toy ] , it is not possible to use neither the bethe - ansatz nor conformal field theory to study the non - critical kondo problem that we discuss here . nevertheless , we expect that the physics of the phase with @xmath204 should be similar to that of the conventional overscreend kondo problem . we also note that , in a more realistic model , one expects _ four _ nodes with two nodes of each chirality and , hence with more flavors and a more pronounced overscreened behavior . we close this section with a discussion of the zero - field susceptibility for this model . the coupling of a magnetic field only to the impurity spin has the form of a zeeman term @xmath208 here , @xmath209 are the elements of a suitably chosen @xmath20 diagonal generator , which also includes the appropriate normalization factors so as to set the bohr magneton to one . we will need to ensure the tracelessness of the generator , and we can do so by keeping all the elements of order one with alternating signs . in this way we prevent the case of overpopulating one of the states that the zeeman term splits with respect to the others , _ i.e. , _ we explicitly avoid cases like the one with @xmath210 elements being @xmath211 and the last element being @xmath210 . this is an important consideration to care about if we want to study the magnetic field crossover , _ i.e. , _ if we were interested in obtaining the suceptibility at finite field . this was discussed in reference @xcite . however , we will be interested only in the zero - field suceptibility and it will be sufficient to assume all the @xmath209 to be @xmath212 , so that the sum of their squares gives a factor of @xmath26 . the magnetization @xmath213 is given by @xmath214 the susceptibility at zero field is @xmath215 once again , we consider two regimes : @xmath196 and @xmath216 . we discuss only the regime @xmath204 since the suceptibility is infinite below @xmath217 , where the impurity is unscreened . ) @xmath218 : + for @xmath219 , we get @xmath220 \label{eq : suscb}\ ] ] @xmath191 : + for @xmath221 , we have @xmath222 \left\ { 1 - 5 \left ( \frac{1}{2 } - \frac{q_f}{n_c } \right ) + \ldots\right\ } \label{eq : susca}\ ] ] once again , we find logarithmic corrections to scaling . we see from eq.([eq : susca ] ) and eq.([eq : suscb ] ) that the zero - field susceptibility @xmath223 , at zero temperature , has a behavior consistent with the general picture described above . it is finite for @xmath224 , but diverges as @xmath190 approaches @xmath217 . this is consistent with the impurity being screeened ( kondo effect ) for a value of the coupling constant bigger than the critical value . however , as we approach the critical coupling constant from the screening ( kondo ) phase , the susceptibility must diverge , since for values of the coupling lower than the critical value , @xmath136 vanishes and the impurity effectively decouples from the band . in this regime one expects the impurity will behave like a free spin in a magnetic field and , since we are at zero temperature , the susceptibility should diverge . in this section we derive the propagator for the band electrons to order @xmath225 . we will use this expression to calculate the @xmath3-matrix in @xmath32 limit . we will consider processes in which an electron with a given energy is initially in a state with well defined _ angular momentum _ , and it is scattered by the impurity into a state with possibly a different angular momentum . thus , we will parametrize the @xmath3-matrix by the magnitude of the incoming and the outgoing momenta ( namely , the energy of the state ) and by the angular momenta of channels involved in the scattering process . this description is natural since at most only a pair of angular moementum states ( @xmath226 with our choice of node ) are actually mixed by the impurity . in contrast , the computation of the correlation function @xmath227 , for @xmath228 and @xmath229 away from the impurity will involve a mixing of all the angular momentum channels . however , in principle , we could add a source term and also decompose the source field in angular momentum modes , and keep only the modes that get mixed by the interaction with the impurity . in other words , we are considering the scattering of band electrons among angular momentum channels produced by the interaction with the impurity . in the case of a coupling to a flat band ( usual kondo problem ) , the only angular momentum channel that gets involved is the @xmath230 . thus , in practice , we only need to consider the partial waves that actually get mixed by scattereing from the impurity . therefore , it will be sufficient to work with the equivalent 1-dimensional fermions @xmath71 and @xmath72 while keeping in mind the correspondences defined in section [ sec : toy ] . the relevant part of the action is @xmath231 we add sources @xmath232 and @xmath233 for the fields @xmath234 and @xmath235 by adding the following source term to the action @xmath236 after completing squares and integrating over the band fermion fields , we are left with ( appart from a normalization factor ) an action which depends only on the impurity fermion fields and the sources , @xmath237 \left ( \frac{1}{-i\omega + e(p)}\right ) \nonumber \\ & & \kern2truein \left [ \int \frac{d\omega ' ' } { 2\pi } \varphi_l^ ( \omega - \omega '' ) \sqrt{\|p\| } f_\sigma ( \omega '' ) + \eta_{l \sigma } ( p,\omega ) \right ] \nonumber \\ & + & \int_{-\infty}^{\infty } \frac{d\omega}{2\pi } \int_{-\infty}^{\infty } \frac{dp}{2\pi } \eta^{\dagger}_{l \sigma } ( p,\omega ) \left ( \frac{1}{-i\omega + e(p)}\right ) \eta_{l \sigma } ( p,\omega)\end{aligned}\ ] ] after a few fairly straightforward manipulations and another integration now over the impurity fermion fields , we can re - write the action as @xmath238 \ = \ - \ \sum_{\sigma=1}^{n_c } \int \frac{d\omega}{2\pi } \ \int \frac{d\omega ' } { 2\pi } j^{\dagger}_\sigma ( \omega ) \ k^{-1 } ( \omega , \omega ' ) \ j_\sigma(\omega ' ) \ ] ] where we have introduced the operators @xmath239 s and the kernel @xmath240 which are given by @xmath241 and @xmath242 the generating functional for the green s functions ( at zero temperature ) can be written as @xmath243 \ = \ \exp \left ( - \ s_f \left ( \eta , \eta^{\dagger } \right ) \right)\ ] ] so that correlation functions of the band fermion fields can be computed by differentiation over @xmath244 @xmath245 \right>_{\eta = \eta^{\dagger } = 0}\end{aligned}\ ] ] we restrict the calcultation to the static solution case where @xmath246 . then @xmath247 where the @xmath3-matrix @xmath248 is given by @xmath249 where @xmath250 was given in eq . ( [ eq : g0 ] ) and eq . ( [ eq : g00 ] ) . the @xmath3-matrix of eq . ( [ eq : tmatrix ] ) exhibits several important features : ) eq . ( [ eq : tmatrix ] ) gives the @xmath3-matrix at _ imaginary _ frequency . the @xmath3-matrix for _ real _ frequency ( at zero temperature ) is found by an analytic continuation to the real axis , @xmath251 . if @xmath252 and @xmath253 are non - zero , there exists a finite matrix element for the scattering of a band electron from a state with angular momentum @xmath254 into a state with angular momentum @xmath255 ( and viceversa ) . in contrast , for a flat band , the @xmath3-matrix is diagonal in angular momentum states . this matrix element is not invariant under @xmath22 since it mixes the angular momentum channels . as we pointed out above , this is a consequance of the unphysical spontaneous breaking of the flavor ( channel ) symmetry . thus only the _ trace _ of this @xmath3-matrix is physical . all other matrix elements will be suppressed by infrared divergent corrections to next order in the @xmath30 expansion . from the analytic continuation of the trace of the @xmath3-matrix one can find the position of the kondo resonance @xmath256 , the width @xmath257 and the phase shift @xmath258 . using the fact that @xmath259 , we can write the trace in the form @xmath260 which is valid for frequencies low compared to the bandwidth @xmath261 but comparable to the dynamically generated scales . the phase shift @xmath258 is @xmath262 the imaginary part of the @xmath3-matrix is @xmath263 the function given by eq.([eq : imt ] ) has a peak at the kondo resonance @xmath256 , which is the solution of the equation @xmath264 one easily obtains the asymptotic solution @xmath265}}+\ldots \label{eq : omegak}\ ] ] where we have kept only the leading logarithmic corrections . we see immediately that we have a resonance centered at @xmath256 , which is basically the kondo scale @xmath107 that we found in section [ sec : scaling ] but modified by the ubiquitous logarithmic corrections . in particular , the position of the kondo resonance scales with the distance to the critical coupling constant and it goes to zero ( the fermi energy ) at @xmath28 . the resonance width @xmath257 can be read off from eq.([eq : imt ] ) @xmath266 from eq . ( [ eq : phaseshift ] ) it is clear that the dimensionless parameter @xmath267 sets the scale for the range of the scattering amplitude . thus , as far as transport properties such as the resistivity are concerned , @xmath136 is the important parameter , whereas thermodynamic properties such as the suceptibility , depend entirely on the impurity energy scale @xmath107 , _ i.e. , _ the kondo scale @xmath203 . the analysis done in the last section shows that , due to the strong logarithmic corrections , these two scales are completely different . once again , in the conventional kondo effect , @xmath203 controls both effects . hence , in contrast with conventional kondo behavior , the interaction with the magnetic impurity induces a non - trivial renormalization of the propagator of the band fermions . this will happen even for @xmath268 , although this effect only occurs in higher order in @xmath30 . finally we stress that , due to the @xmath22 flavor symmetry , there can not be any channel mixing scattering processes in this model . however , in the following section , we will present a model which describes the most general impurity scattering processes for systems with only one node . in that model the channel ( or flavor ) symmetry @xmath22 is broken explicitly already at the level of the impurity hamiltonian . thus we expect to find processes which will mix the various channels . in the past sections we considered a situation in which the fermions coupled to the impurity spin only through their own spin density @xmath269 . scattering processes of this type have the simplifying feature that particles and holes interact with the impurity independently from each other and exactly in the same way . however , in practice , more general scattering processes will be present . we also made the assumption that only one node of the two - dimensional fermions is present . here , we will consider all possible scattering processes involving _ only one node_. the coupling through the spin density has the very special feature that it is _ diagonal _ in the components of the spinors , _ i.e. , _ the impurity does not mix particles with holes . in terms of the effective hamiltonian of eq . ( [ eq : halt ] ) , this diagonal coupling implies that there is no _ explicit _ mixing of angular momentum channels . hence , the model has an @xmath22 symmetry ( with @xmath23 ) . any process which mixes particles and holes will break the flavor symmetry explicitly . we will include now these processes . we will also incorporate the correct node structure . for example , in the case of a flux phase say , on a square lattice , there will be four different _ nodes _ corresponding to the symmetry points @xmath270 of the first brillouin zone . as we discussed in section [ sec : intro ] , a @xmath11-wave superconductor has an analogous node structure , at the points where the gap has nodes . we will discuss of this very interesting case in a separate publication @xcite . it is straightforward to see that the main effect of a more general node structure is to _ increase _ the number of _ flavors _ in the effective model . in the one - node model of the previous section we found that there were two flavors and an @xmath19 flavor symmetry . the existence of two flavors can be traced back to the spinor structure of the original problem . when the two - dimensional fermions have more than one node , the number of flavors becomes _ twice the number of nodes _ , @xmath271 . thus , for a flux phase we will have @xmath272 flavors . however , flux phases are _ even _ under parity . hence , there are two pairs of nodes with opposite parity . the parity of the node is given by the _ relative sign _ of the two terms in the free fermion hamiltonian of eq . ( [ eq : h0 ] ) . thus , a change of parity is equivalent to the mapping @xmath273 , in other words to a reflection across the @xmath274 axis . in section [ sec : toy ] we showed that , with the choice of parity we made there , the channels with angular momenta @xmath16 and @xmath18 get mixed by the impurity . hence , if nodes with the other parity are also included , we will also get mixing between @xmath16 and @xmath275 , but no direct mixing between @xmath275 and @xmath276 . in section [ sec : toy ] we also found that only the mapping between the components of the original fermion and the fermions of the effective one - dimensional theory carried information about which angular momentum components are mixed . this feature is always present . thus , the net effect of having a multinode structure is to increase the number of flavors . consequently , although the thermodynamic properties will be insensitive to parity assignments , the scattering amplitudes of the original fermions with the impurity will carry the information of the parity of the nodes . this feature can be used to determine the nature of a superconducting state . in what follows , we will consider a system with four nodes , two with positive parity and two with negative parity , interacting with a single magnetic impurity . for simplicity we will assume that the impurity form factor is strongly peaked at small momentum , so that inter - node scattering processes can be ignored . we will keep , however , processes in which particles and holes may interact differently with the impurity and/or mix with each other . we parametrize these processes with four coupling constants @xmath277 , with @xmath278 . the generalized impurity hamiltonian is given by @xmath279 where the indices @xmath254 labels the nodes and @xmath280 and @xmath5 the spin components . @xmath281 is the @xmath282 identity matrix and @xmath283 ( @xmath284 ) are the three pauli matrices . here @xmath112 is the coupling constant that was used in the model of section [ sec : toy ] . the free fermion hamiltonian for each node @xmath285 is invariant under the symmetry transformation @xmath286 . the effect of this transformation on on @xmath287 is to reverse the sign of both @xmath288 and @xmath289 . this implies that the sign of @xmath288 and @xmath289 is irrelevant to the properties of the theory , which should depend only on the absolute values of these two coupling constants . in the expressions for the fermi field @xmath290 we have @xmath291 and @xmath292 . in what follows we use a single flavor index @xmath293 with @xmath271 however , we will keep in mind that the impurity does not mix different nodes but it does mix the flavor components associated with the same node . the model of equivalent one - dimensional right - moving excitations has an @xmath19 spin ( color ) symmetry and an @xmath294 flavor symmetry ( associated with the channels ) broken down to @xmath295 . the free fermion kinetic term has the form @xmath296 and an impurity interaction term with the following structure @xmath297 where @xmath298 is the coupling matrix which has a block diagonal form . for @xmath299 associated with the same node @xmath300 has the form @xmath301 in other terms the matrix has the form @xmath302 where @xmath3 is the @xmath303 matrix of eq . ( [ eq : tij ] ) and @xmath304 is the @xmath305 identity matrix . we have also let the spin indices @xmath306 to run from @xmath307 and used the identity of eq . ( [ eq : suntrace ] ) . the form of the impurity hamiltonian eq . ( [ eq : himp ] ) shows that , if the coupling constants @xmath277 are all different , the @xmath19 flavor symmetry of each node is broken by the interactions but the symmetry involving different nodes remains intact . we will now proceed as in the previous sections and solve this model in the limit of @xmath113 . after a hubbard - stratonovich transformation , the quartic term of the euclidean action becomes ( repeated indices are summed ) @xmath308 where @xmath309 . since @xmath3 is of the form @xmath302 , then @xmath213 has the same form , _ i.e. , _ @xmath310 . in the following , @xmath213 stands for the @xmath311 matrix of eq . ( [ eq : hsgen ] ) . it can be easily shown that @xmath312 where @xmath313 and @xmath314 . in the case considered in section [ sec : i.e. , _ when @xmath315 , @xmath316 , we have a full @xmath317 flavor symmetry between the @xmath46-fields . in other words , the two channels play exactly the same role . when the coupling constants are all different , this symmetry is broken down to a @xmath318 , where one of the @xmath319 symmetries is generated by the identity and the other by @xmath320 . the @xmath113 limit is taken in the standard fashion . the hubbard - stratonovich fields @xmath321 will be chosen to be an arbitrary vector for each node and the same vector for all nodes . let @xmath322 be a unitary transformation which diagonalizes the @xmath282 matrix @xmath320 , and let @xmath323 , so that @xmath324 . the eigenvalues of @xmath320 can be rewritten as @xmath325 , where @xmath326 the modified effective action can be written as @xmath327 - q_f \epsilon_f \nonumber \\ & = & - n_c \int_{-\infty}^{\infty } \frac{d\omega}{2\pi } \ \log\left(-i\omega + \epsilon_f + { \frac{1}{2 } } n_{nodes } \left ( \|{\tilde \sigma}_1\|^2 + \|{\tilde \sigma}_2\|^2 \right ) g_0 ( i\omega)\right ) \nonumber \\ & & \qquad \qquad + \ { \frac{1}{2 } } n_{nodes } n_c \left ( \frac{\|{\tilde \sigma}_1\|^2}{g_0 - \sqrt{{\vec g}^2 } } + \frac{\|{\tilde \sigma}_2\|^2 } { g_0 + \sqrt{{\vec g}^2 } } \right ) - q_f \epsilon_f \label{eq : modeff}\end{aligned}\ ] ] the new saddle point equations are @xmath328 and @xmath329 and @xmath330 we now solve these new spe s . the solution will be cast in the form of a phase diagram which can be plotted in a @xmath331 plane ( see figure [ figure ] ) . in principle , we find three different solutions of the spe s : ) @xmath332 . this is the region of the phase diagram below the line @xmath333 . this is the weak - coupling phase . the magnetic impurity is effectively decoupled from the band electrons . to leading order in @xmath30 the impurity does not interact with the band fermions and behaves like a free magnetic moment . consequently the impurity spin suceptibility is infinite , there is no resonance and there is no kondo screening . both @xmath334 and @xmath335 . there is no consistent solution of this form unless @xmath336 . in other words , this case is possible only if the only non - vanishing coupling constant is @xmath112 , with @xmath337 , in which case we do have the @xmath317 symmetry . this is the line on the axis @xmath190 of the figure [ figure ] for @xmath207 ; @xmath338 , @xmath339 . a solution of this form satisfies the equations @xmath340 the action of this solution is given by @xmath341 for a solution of this type , @xmath342 and @xmath343 are determined by @xmath344 through the relations @xmath345 this solution is only allowed above the line @xmath346 in the phase diagram . this is the kondo or screening phase . the results of the previous sections apply to this phase . @xmath338 , @xmath347 . the spe s also allow for solutions of this form . in principle such solutions are allowed above the line @xmath348 . the spe s for this solution are exactly the same ones found in ( c ) with @xmath349 replaced by @xmath350 . clearly this solution competes with the solution of case ( c ) over a significant region of the phase diagram . in fact its allowed region is completely included within the allowed region for ( c ) . however , we argue that this solution is always metastable and never occurs . in fact , for any finite value of @xmath351 , the interaction between the impurity and the band electrons will be stronger in case ( c ) since @xmath352 , favoring the first case in much the same way in which the bonding state is preferred to the antibonding one when the degeneracy in a two - level system is lifted by a perturbation . it is easy to check that this is indeed correct by looking for solutions of both types in the vicinity of the phase transition , for @xmath190 close to @xmath217 and @xmath351 small . in addition , since there is no actual symmetry change between these two phases " , we do not expect a phase transition . we conclude that the generalized impurity model has just two phases : a phase with an unscreened impurity and a phase with a kondo effect . the phase boundary is at the line @xmath346 . the physics of the kondo phase is almost identical to what was described in the previous sections . in fact , at the level of the @xmath32 theory , the physical observables of the generalized impurity model can be calculated using the formulas for the single node model of the previous sections . the main difference is that , since the @xmath317 flavor symmetry at each node is now broken by the explicit form of the impurity hamiltonian , operators with non - trivial matrix elements in that sector are now allowed . in particular , there will be non - vanishing , finite , off - diagonal matrix elements of the band fermion @xmath3-matrix . in other words , the impurity will mix band fermion states with different angular momentum . in contrast , processes which mix different nodes are still strictly forbidden . finally we note that scattering processes of the type described here effectively reduce the number of independent channels . if it were not for the existence of several nodes , we would expect to find a kondo effect in this phase with a completely screened impurity . once again , the existence of additional channels associated with the multinode structure turns this into a multichannel kondo system if inter - node scattering processes are not allowed . in this paper we have considered the problem of flux - phase fermions coupled locally to a single magnetic impurity . this is formally the same as the problem of fermions with a relativistic dispersion in two space dimensions couped to a local spin . we derived explicitely an effective theory in one - space dimension . the physics of the effective one - dimensional theory for this problem differs from the conventional radial picture of the kondo problem in several important ways . firstly , it is always a multichannel system . the channels reflect the spinor structure of the nodes and the multiplicity of nodes . secondly , the relativistic dispersion implies a density of states which vanishes linearly at the fermi energy . consequently , unlike the conventional kondo model , the effective one - dimensional problem has a non - local coupling between the effective right movers and the impurity . this is the feature that drives the kondo effect away from marginality and it is responsible for the phase transition between an unscreened impurity phase and a phase with a kondo effect . this is consistent with earlier results of withoff and fradkin . however , for the case of the linearly vanishing density of states that we discussed in this paper , the nature of the scaling in the vicinity of this zero - temperature phase transition is drastically changed . we find that all physical quantities exhibit very simple scaling laws modified by logarithmic corrections . we found this behavior in the kondo scale and in the impurity spin suceptibility . this behavior is strongly reminscent of critical phenomena at an upper critical dimension . one important consequence of the logarithmic corrections is that there are more dynamical scales and that all physical quantities are no longer controlled by the kondo scale @xmath203 alone . this is particularly clear if one compares the amplitude of the @xmath3-matrix and the position and width of the kondo resonance . the model we studied and solved here using large-@xmath2 methods is an interesting problem on its own right . in a subsequent publication we will report on a study of a similar model for magnetic impurities in @xmath11-wave superconductors where we will draw heavily on the ideas that we developed here . it is interesting to compare the effective one - dimensional model that we derived here with the conventional one - dimensional models for the conventional kondo problem . the standard kondo problem is equivalent to a model in one space dimension with a single right mover which interacts locally with the impurity spin through the fermion spin density . this coupling through a density is crucial for the physics of the kondo problem to be correctly described by the model of right moving fermions . the fact that the fermions are chiral ( namely , only right movers are present ) means that , up to a fermi velocity , the fermion density and current are the same observable . this model can be described entirely in terms of a conserved current . this is the starting point of the conformal field theory approach of affleck and ludwig@xcite . however , it is also crucial for the success of the approach of anderson , yuval and hamman@xcite . in fact , the equivalence that exists between kondo systems and problems of macroscopic quantum coherence@xcite rely heavily on bosonization of models of fermion coupled locally to impurities through a density . the fact that this density is associated to a conserved current means that it can not acquire anomalous dimensions . thus , the fixed points of the conventional kondo contains only marginal operators which are made marginally relevant by quantum fluctuations . from this analysis it is clear that it is not possible to describe the phase transition that we discuss here in terms of a conformal field theory coupled to local boundary operators representing the impurity . the non - locality of the effective one - dimensional theory is essential . it is because the model is non - local that the operator that couples to the impurity can ( and does ) acquire an anomalous dimension . this is the mechanism which drives the phase transition . these models are not equivalent to any standard macroscopic quantum tunneling ( mqt ) model . it may appear that , because the density of states vansihes like a power of the energy , these models could be related to a subohmic mqt system which are known not to have phase transitions . however , subohmic mqt models of quantum impurities coupled to a macroscopic _ bosonic _ system , with subohmic spectral density . the models that we discussed here are fermionic and are not equivalent to a subohmic bosonic theory . we are grateful to k. ingersent for useful discussions and for bringing to our attention the possible physical scenario discussed at the end of sec . [ sec : intro ] . this work was supported in part by nsf grants no . nsf dmr94 - 24511 at the department of physics of the university of illinois at urbana - champaign , and dmr89 - 20538/24 at the materials research laboratory of the university of illinois . in section [ sec : scaling ] and from eq.([eq : scaling9 ] ) we have @xmath353 and @xmath354 in the region of integration where @xmath355 , @xmath169 can be approximated by @xmath356 on the other hand , when we are in the regime @xmath357 , @xmath358 the following inequality holds @xmath359 we will be interested in the following cases . @xmath360 @xmath361 for @xmath362 , we can expand the denominator in eq.([eq : app4 ] ) to get @xmath363 we can now get the behavior of eq.([eq : scaling1 ] ) in these two different regimes : for @xmath179 , @xmath364 the other limit corresponds to @xmath365 ; consequently the contribution from @xmath169 is neglegible and @xmath366 now we want to work out similar approximations for the other spe . we have @xmath367 in which we made use of the fact that over the integration interval , the variable @xmath143 is very small , so that the approximations @xmath368 and @xmath369 are consistent . for the other portion of the integral ( _ i.e. , _ for @xmath166 ) we are in a situation where @xmath370 and then @xmath371 , so we neglect the logaritmic term against 1 . we need here the following result @xmath372 then @xmath373 the function @xmath374 verifies that , for @xmath375 , @xmath376 and for @xmath377 is given by @xmath378 \nonumber \\ & & \qquad \qquad \approx \ z \left ( \log z - 1 \right ) - \frac{1}{3 } \ z^3\left ( \log z - \frac{1}{3 } \right ) + \dots \label{eq : app14}\end{aligned}\ ] ] getting everything together , in the limit @xmath379 and @xmath380 , where @xmath381 one can write the spe in the form @xmath382 which can be solved by iteration ( see eq.([eq : nu1 ] ) in section [ sec : scaling ] ) . the other limiting case corresponds to @xmath383 , with @xmath138 , @xmath384 . here @xmath385 the spe now is @xmath386 \label{eq : applast}\ ] ] since in this case , @xmath387 . the cubic term can be dropped rightaway . as for the other terms , as we approach the transition , @xmath388 remains fixed while @xmath389 . therefore , the leading term is going to be the one with the factor @xmath390 . this immediately gives the implicit expression for @xmath136 in terms of the impurity occupancy @xmath391 and the distance to the critical point @xmath392 that we used in eq.([eq : nu3 ] ) in section [ sec : scaling ] . the usual logarithmic divergencies of the high temperature expansion of the standard kondo problem are absent for this non - marginal system . the logarithmic corrections that we discuss have a completely different and new character . after this paper was submitted for publication , k. ingersent ( k. ingersent , private communication ) has kindly pointed out to us that in a recent work of his he has investigated similar issues using a wilson rg approach . ingersent s rg suggests that our phase transition may actually be suppressed .	we consider a band of fermions in two space dimensions with a flux phase ( relativistic ) dispersion relation coupled to a local magnetic impurity via an @xmath0 interaction . this model describes spinons of a flux phase and it is also a qualitative model of the quasiparticles in a @xmath1 superconductor . we find a zero - temperature phase transition at a finite coupling constant between a weak coupling unscreened impurity state and a strong coupling regime with a kondo effect . we use large-@xmath2 methods to study the phase transition in this kondo system away from marginality . the kondo energy scales linearly with the distance to the transition . the zero - field magnetic suceptibility at zero temperature diverges linearly . similar behavior is found in the @xmath3-matrix which shows a resonance at the kondo scale . however , in addition to this simple scaling , we always find the presence of logarithmic corrections - to - scaling . such behavior is typical of systems at an upper critical dimension . we derive an effective fermion model in one space dimension for this problem . unlike the usual kondo problem , this system has an intrinsic multichannel nature which follows from the spinor structure of @xmath4-dimensional relativistic fermions .
the forward - backward ( fb ) asymmetry in top - quark pair production in proton - antiproton collisions is an observable which originates from the difference in the production rates for top quarks in the forward and backward hemispheres @xcite . the total fb asymmetry was measured by the cdf and d0 collaborations at the tevatron @xcite . the measurement can be carried out in the laboratory frame ( @xmath2 frame ) as well as in the center - of - mass frame of the top - quark pair ( @xmath3 frame ) . the asymmetries in the two frames are defined as @xmath4 where @xmath5 is the number of events , @xmath6 indicates the laboratory frame ( @xmath3 frame ) , and @xmath7 is the top - quark rapidity in frame @xmath8 . the measurements obtained by the cdf collaboration using 5.3 fb@xmath9 of data are @xcite @xmath10 the quoted uncertainties are derived from a combination of statistical and systematic errors . the production of top - quark pairs at hadron colliders is dominated by qcd . at the tevatron , the charge conjugation invariance of the strong interaction implies that the difference in the production of top quarks in the forward and backward hemispheres is equivalent to the difference in the production of top and antitop quarks in the forward hemisphere . therefore , in qcd the fb asymmetry is equivalent to the charge asymmetry . qcd predicts a non - vanishing contribution to the fb asymmetry starting at order @xmath11 in the squared amplitude . a contribution to the asymmetry arises if , in the interference of one - loop and tree - level diagrams , the top - quark fermionic line and the light - quark fermionic line are connected by three gluons @xcite . the same is true in the case of the interference of two tree - level diagrams with three particles in the final state . the dominant effect is from the quark - antiquark annihilation channel , while a further , numerically small contribution to the asymmetry at order @xmath11 originates from the flavor excitation channel @xmath12 , where @xmath13 indicates additional partons in the final state . the gluon fusion channel does not contribute to the fb asymmetry at any order in perturbation theory , due to the fact that the gluon distribution is the same for protons and antiprotons . the total fb asymmetry predicted by qcd at the first non - vanishing order ( which , for reasons discussed later in this paper , we will indicate as next - to - leading order ( nlo ) ) is lower than the one measured at the tevatron . two recent evaluations using the formulas in @xcite report the following values @xcite , obtained using mstw2008 nlo parton distribution functions ( pdfs ) @xcite , @xmath14 the central values quoted in ( [ eq : nlo ] ) refer to the choice @xmath15 , where @xmath16 is the factorization scale and @xmath17 the top - quark mass , which is set to @xmath18 gev . the errors originate from the scale variation in the range @xmath19 . these choices will also be adopted in the rest of the paper . electroweak corrections enhance the prediction for the asymmetry by less than 10% of the central value @xcite , and to separate uncertainties coming from electroweak calculations we do not include these corrections in our numerical results . as one can see , the discrepancy between the theory prediction and the experimental measurement in the @xmath2 frame is less than two standard deviations ( 2@xmath20 ) , while in the @xmath3 frame the two values agree within @xmath21 , although the central value of the experimental measurement is a bit higher . in @xcite , the cdf collaboration measured the fb asymmetry in the @xmath3 frame as a function of the top - pair invariant mass @xmath22 . after grouping the events in two bins corresponding to @xmath23 gev and @xmath24 gev , they found the asymmetry in the latter bin to be @xmath25 which is more than 3@xmath20 higher than the stated theoretical nlo prediction in @xcite of @xmath26 obtained using the mcfm program @xcite . a measurement of the fb asymmetry in two bins of the rapidity difference @xmath27 was also performed and again in that case the higher bin shows a tension with the nlo qcd prediction , although with larger experimental errors . many attempts to explain these results in terms of new physics scenarios have been made , see e.g. @xcite . the task is complicated by the fact that the new physics contributions should not spoil the good agreement between theory and measurements for the total pair - production cross section and the differential distribution in the pair invariant mass . recently , we have calculated the top - pair invariant - mass distribution and the top - quark rapidity and transverse - momentum distributions in renormalization - group ( rg ) improved perturbation theory , which incorporates the resummation of logarithmic corrections that become numerically large in the limit of soft - gluon emission @xcite . these calculations were carried out at next - to - next - to - leading - logarithmic ( nnll ) accuracy . the resummed distributions were then matched to the nlo results to obtain predictions which have nlo+nnll accuracy . alternatively , the resummed results at nnll order can be employed to generate approximate next - to - next - to leading ( nnlo ) predictions @xcite . we note that working at approximate nnlo accuracy does not alter in a fundamental way any of the results obtained within the nlo+nnll framework . on the other hand , we believe that in certain kinematic regions , for instance at large pair invariant mass , resummation is important . for these reasons , and for the sake of simplicity , in the following we focus on nlo+nnll calculations . by integrating the differential distributions it is then straightforward to calculate the top - quark fb asymmetry both in the laboratory frame and in the @xmath3 frame . starting from the double differential cross section in the pair invariant mass and scattering angle , we can also compute the @xmath22 and rapidity - dependent asymmetries , which can be compared to experimental measurements . the main goal of this letter is to present the result of these calculations of the total and differential fb asymmetries in a systematic way . at the large hadron collider ( lhc ) , the symmetry of the @xmath28 initial state dictates that the rapidity distributions of the top and antitop quarks are symmetric and that the fb asymmetry vanishes . however , it was observed in @xcite that at the lhc top quarks are preferably produced at larger rapidities than antitop quarks in the laboratory frame . like the fb asymmetry at the tevatron , this rapidity - dependent charge asymmetry is generated at order @xmath11 in the squared amplitude , mainly through the asymmetric part of the quark - antiquark annihilation channel . therefore , potential new physics contributions would effect these two quantities in a correlated way , and the higher collider energy at the lhc gives it better access to distortions at higher rapidities . as a final application of our formalism , we evaluate at nlo+nnll order the partially integrated charge asymmetry at the lhc , giving results as a function of a lower cut - off on the top and antitop rapidities in the laboratory frame . the fb asymmetry in the laboratory frame can be calculated starting from the top - pair production cross section differential with respect to the top - quark transverse momentum @xmath29 and rapidity @xmath30 . to do so , it is convenient to first define a total and differential asymmetric cross section via @xmath31 \nonumber \\ & \equiv \int_0^{y^+_t } dy_t \left [ \left ( \frac{d\sigma}{dy_t } \right)_f - \left ( \frac{d\sigma}{dy_t } \right)_b \right]\equiv \int_0^{y^+_t } dy_t \frac{d \delta\sigma^{p\bar{p}}_{\text{fb}}}{dy_t } \,.\end{aligned}\ ] ] here @xmath32 where @xmath33 is the square of the hadronic center - of - mass energy . to obtain the fb asymmetry in the laboratory frame one needs to calculate the ratio of the asymmetric cross section in ( [ eq : fbasympp ] ) to the total cross section : @xmath34 in our phenomenological analysis we will consider two levels of perturbative precision for the asymmetric cross section and fb asymmetry . the first involves the differential cross section at nlo in fixed - order perturbation theory , the second the nlo calculation supplemented with soft - gluon resummation to nnll order . in the laboratory frame , the resummed calculations are carried out using the @xmath35 scheme introduced in @xcite . such predictions resum a class of logarithms which become large in the limit where @xmath36 is small ( @xmath37 indicates the momentum of the antitop quark , and @xmath38 the total momentum of additional partons in the final state ) , in which case real gluon emission is soft although the top pair is not produced at rest . while the phase - space integrals in ( [ eq : fbasympp ] ) are in general sensitive to regions where @xmath39 is not small , the threshold region is dynamically enhanced due to the sharp fall - off of the pdfs away from small values of @xmath39 . in fact , it was shown in @xcite that the leading terms in the @xmath35 threshold limit reproduce essentially the entire nlo correction to the differential cross section in the quark channel , which implies that resumming the corrections which become large in this limit is an improvement on the fixed - order calculation . while in the fixed - order counting the asymmetric cross section first arises from the nlo calculation of the differential cross section , in the resummed counting it first appears at nll order . the nlo+nnll calculation is thus a refinement on the leading term , and will be considered our best prediction . before illustrating our results for the fb asymmetry , we need to clarify an important point concerning our convention for counting orders in the perturbative expansion . we calculate @xmath40 itself as a perturbative expansion in @xmath41 , using a fixed - order or logarithmic counting as appropriate . for example , the first non - vanishing contribution to the asymmetry in fixed - order perturbation theory is obtained by calculating the numerator in ( [ eq : fbaa ] ) to order @xmath11 and the denominator to order @xmath42 . the resulting asymmetry is of order @xmath41 , which we will refer to as nlo , with reference to the order at which the differential distributions in ( [ eq : fbasympp ] ) are calculated relative to @xmath42 . similarly , in rg - improved perturbation theory , the first non - vanishing contribution to @xmath40 is obtained by calculating both the numerator and the denominator at next - to - leading - logarithmic ( nll ) order ; the resulting asymmetry is then counted as nll . this convention was already adopted in @xcite . there are two counting schemes in the literature ( in fixed - order perturbation theory ) which are different from ours . the first one also treats @xmath40 itself as a perturbative expansion , but counts the order @xmath41 contribution as lo , and so on , as in @xcite . the second one treats the numerator and the denominator as separate perturbative series and does not further expand the ratio , as adopted in the quoted mcfm results in @xcite . we note that while the first scheme differs from ours only by name , the second scheme leads to different numerical results . as discussed in @xcite , the nlo+nnll results are considerably more stable with respect to the choice of scheme than the nlo results . .[tab : ppbar ] the asymmetric cross section and fb asymmetry in the @xmath2 frame . the first error refers to perturbative uncertainties estimated through scale variations as explained in the text , and the second error in the mstw2008 case is the pdf uncertainty . [ cols="<,^,^,^,^,^,^ " , ] in this section we study the simplest realization of a charge asymmetry at the lhc , namely the rapidity - dependent quantity in the laboratory ( @xmath28 ) frame . in particular , we focus on the partially integrated charge asymmetry and charge - asymmetric cross section , where we impose the restriction @xmath43 on the differential cross section . we define these through @xmath44 with @xmath45 as in ( [ eq : yplimits ] ) . we can study the partially integrated charge asymmetry and asymmetric cross section ( [ eq : lhcasym ] ) at nlo and nlo+nnll order using the results for the differential cross section in 1pi kinematics obtained in @xcite . the results generated with mstw2008 pdfs are shown in figure [ fig : lhcasym ] , where the bands reflect uncertainties related to scale variations carried out with the same procedure as at the tevatron . for the charge - asymmetric cross section shown in the left - hand panel of the figure , the main effect of the resummation is to decrease the scale dependence of the result to a relatively small region of the nlo error band . for the partially integrated asymmetry shown in the right - hand panel of the figure , the resummation is a mild effect up to @xmath46 , but substantially reduces the asymmetry at higher values of the cut . however , due to large k factors in the gluon channel at the lhc , the uncertainty band for the nlo curve is very sensitive to whether one consistently expands the asymmetry ratio in ( [ eq : lhcasym ] ) . if we had deviated from our normal procedure and had not expanded the ratio , instead evaluating the denominator at nlo order , the nlo band would actually overlap quite well with the nlo+nnll results shown in figure . the nlo+nnll result is largely insensitive to this change of systematics the result where both the numerator and denominator are evaluated at nlo+nnll order is within the error band shown in the figure . the partially integrated charge asymmetry vanishes for @xmath47 , and becomes progressively larger at higher values of the cut . however , the charge asymmetric cross section shown in the left - hand side of the figure is very small at higher rapidity values and the experimental measurement is difficult . a reasonable way to compare theory and experiment in this case would be to perform a measurement in a high - rapidity bin with @xmath48 . in such a bin the standard model charge asymmetry is predicted to be only slightly different from zero , less than 2% depending on the exact choice of the cut , so any appreciable charge asymmetry in an experimental measurement would be a clear sign of a new physics contribution to the high - rapidity region of the distribution , which is already hinted at by the tevatron measurements . note that other partially integrated differential distributions can be used to study the phase - space dependence of the charge asymmetry . for instance , one can consider the so - called central charge asymmetry introduced in @xcite , which imposes a cut @xmath49 along with a restriction to relatively high pair invariant mass in order to reduce the contribution from the @xmath50 channel to the symmetric part of the partially integrated cross section . unfortunately , it is beyond the scope of the paper to perform an nlo+nnll analysis for such an observable , because it requires a combination of pim and 1pi kinematics which is not possible to derive starting directly from the results of @xcite . the total top - quark fb asymmetry measured at the tevatron is not in good agreement with the predictions obtained at the first non - vanishing order in perturbative qcd . the tension between theory and experiment is about two standard deviations for the asymmetry measurement in the laboratory frame , and approximately 1@xmath20 in the top - pair rest frame . moreover , the measurement of the asymmetry in the large pair invariant - mass region ( @xmath24 gev ) by the cdf collaboration @xcite differs from the leading - order prediction in qcd by more that three standard deviations , and there is also a discrepancy between theory and experiment for the fb asymmetry at values of the top - pair rapidity difference @xmath51 . the calculations of the double differential distribution for the top - quark pair production in two different kinematic schemes , carried out in @xcite , allow us to evaluate the fb asymmetry in both the laboratory frame and in the top - pair rest frame . the predictions obtained in this way include the full nlo corrections to the total and asymmetric cross sections , whose ratio determines the fb asymmetry , as well as the resummation of soft - gluon emission effects up to nnll accuracy . such nlo+nnll results represent the most accurate determination of the qcd contribution to the asymmetry that can be obtained to date . studies of the nlo results indicated that power corrections to our results obtained in the soft limit within the effective theory approach are small , due to the mechanism of dynamical threshold enhancement . the full nnlo calculation would be very useful in quantifying this at the next order in perturbation theory , but we do not expect such power corrections to change our predictions in any significant way . tables [ tab : ppbar ] and [ tab : ttbar ] show that the numerical impact of the nnll corrections on the total asymmetry is modest , less than @xmath52 of the central value in all cases considered . it must be observed that while the total and asymmetric cross sections calculated at nlo+nnll order have smaller scale uncertainties than the corresponding quantities calculated at nlo , the same is not always true for the fb asymmetry , indicating that the errors at nlo are affected by accidental cancellations and do not reflect the true perturbative uncertainties . after the inclusion of the nnll corrections , the predicted values of the asymmetry remain lower than the measured ones by @xmath53 in the @xmath54 frame and @xmath55 in the @xmath56 frame . we have also shown that pdf uncertainties are much reduced for the fb asymmetries compared to those for the asymmetric or total cross sections . for instance , the total asymmetry calculated with mstw2008 pdfs differs from the one obtained from cteq6.6 or nnpdf2.1 pdfs by only a few percent . the nlo+nnll calculation of the asymmetry in the high invariant - mass bin @xmath57 gev increases the nlo prediction only by a very small amount , about 3% of the central value . the perturbative and pdf uncertainties of the calculation are much smaller than the experimental errors . hence , if the tevatron measurement of the fb asymmetry in the high invariant - mass bin is confirmed as the experimental error is reduced , it will become a compelling signal for physics beyond the standard model . given the situation at the tevatron , it is well motivated to measure observables sensitive to the top quark asymmetry also at the lhc . since the fb asymmetry vanishes at the lhc , one of the possible choices consists in measuring the charge asymmetry partially integrated in a given rapidity region . starting from the nlo+nnll resummed rapidity distributions for the top and antitop quarks , we calculated the charge asymmetry as a function of a lower rapidity cut in the laboratory frame . a careful choice of the lower bound on the rapidity integration region can allow one to use this observable to test for the presence of new physics in the top - quark sector at the lhc , thus providing complementary information to the tevatron measurements . _ acknowledgments : _ we would like to thank rikkert frederix for providing us a private version of madfks . this research was supported in part by the state of rhineland - palatinate via the research center _ elementary forces and mathematical foundations _ , by the german federal ministry for education and research under grant 05h09ume , by the german research foundation under grant ne398/3 - 1 , by the european commission through the _ lhcphenonet _ initial training network pitn - ga-2010 - 264564 , and by the schweizer nationalfonds under grant 200020 - 124773 .	we make use of recent results in effective theory and higher - order perturbative calculations to improve the theoretical predictions of the qcd contribution to the top - quark pair production forward - backward asymmetry at the tevatron . in particular , we supplement the fixed - order nlo calculation with higher - order corrections from soft gluon resummation at nnll accuracy performed in two different kinematic schemes , which allows us to make improved predictions for the asymmetry in the @xmath0 and @xmath1 rest frames as a function of the rapidity and invariant mass of the @xmath1 pair . furthermore , we provide binned results which can be compared with the recent measurements of the forward - backward asymmetry in events with a large pair invariant mass or rapidity difference . finally , we calculate at nlo+nnll order the top - quark charge asymmetry at the lhc as a function of a lower rapidity cut - off for the top and antitop quarks .
supernova ( sn ) 2006jc is the prototype of the class of hydrogen - deficient sne that exhibit prominent narrow lines of he i in their spectra , the so - called type ibn " sne @xcite . sn 2006jc also has the distinction of being the first sn for which a pre - explosion outburst was detected ; in the case of sn 2006jc an outburst was previously identified at the sn position in 2004 , hence 2 years prior to the sn itself @xcite . outbursts of this sort are usually associated with luminous blue variables ( lbvs ) for which eruptions are considered a natural part of their evolution , despite the exact physics responsible for them is poorly known . @xcite first confirmed the presence of a bright , massive ( @xmath7 ) lbv - like progenitor in pre - explosion observations of the type iin sn 2005gl . lbvs provide a natural connection between pre - explosion outbursts and the creation of dense h - rich circum - stellar envelopes with which subsequent sne interact yielding type iin sne . in contrast , type ibn sne require the presence of an he - rich and , generally , h - poor circumstellar medium ( csm ) arising from a wolf - rayet ( wr ) progenitor undergoing an lbv - like eruption , although this interpretation has been somewhat questioned with the discovery of a type ibn sn in the outskirts of an elliptical galaxy ( hence in a likely old stellar population environment ; * ? ? ? * ) . so far , about two dozen type ibn sne have been discovered ( see , e.g. , the samples of * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , and no other example has shown an lbv - like outburst similar to that observed before the explosion of sn 2006jc . in addition , no progenitor of a type ibn sn has ever been seen in quiescence to unequivocally prove the wr nature of their progenitors . for all these reasons , the stellar configuration that produced the sequence of events observed at the location of sn 2006jc is still debated . the most common interpretation for sn 2006jc and its precursor outburst , is that the progenitor was a wr star with residual lbv - like instability ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? . an alternative scenario , however , was proposed by @xcite for sn 2006jc , invoking a massive binary system to explain the chain of events that occurred in 2004 - 2006 : an lbv companion erupted in 2004 , and while it was a wr star that exploded in 2006 as a normal stripped - envelope sn . nonetheless , this scenario did not comfortably explain the narrow he i emission features observed in the spectrum of sn 2006jc . furthermore , if sn 2006jc occurred in a massive star forming region there was the possibility that the precursor outburst may have arisen from an lbv - like star in close proximity to the progenitor but otherwise unrelated to the explosion . here we present late - time observations of the site of sn 2006jc , to explore the nature of the progenitor systems responsible for type ibn sne . sn 2006jc occurred in ugc 4904 , for which the corrected recessional velocity is @xmath8 . assuming @xmath9 , we adopt a distance of @xmath10 . sn 2006jc was discovered by k. itagaki @xcite on 2006 oct 9 , and analysis by @xcite suggested it was discovered only a few days after maximum . more recently an earlier observation of ugc 4904 acquired by k. itagaki on 2006 oct 3 was found , in which sn 2006jc was detected at @xmath11 . comparisons of the lightcurve of sn 2006jc with other sne with similar decline rates suggests that sn 2006jc may have exploded on the order of @xmath12 days before maximum @xcite . we assume a metallicity for the site of sn 2006jc of half - solar , or large magellanic cloud , abundances following the measurement of @xcite . lcccc date & instrument & filter & exposure & program + ( ut ) & & & time ( s ) & + 2008 nov 19.9 & @xmath13 & @xmath14 & 460 & 10877@xmath15 + 2008 nov 19.9 & @xmath13 & @xmath16 & 700 & 10877 + 2008 nov 22.0 & @xmath13 & @xmath17 & 800 & 10877 + 2008 nov 22.0 & @xmath13 & @xmath18 & 360 & 10877 + + 2010 apr 30.5 & @xmath19 & @xmath20 & 1380 & 11675@xmath21 + 2010 apr 30.6 & @xmath19 & @xmath22 & 897 & 11675 + 2010 apr 30.6 & @xmath19 & @xmath14 & 868 & 11675 + 2010 apr 30.6 & @xmath19 & @xmath23 & 868 & 11675 + + @xmath15 pi : w. li + @xmath21 pi : j. maund + the site of sn 2006jc was observed at two separate epochs using the hubble space telescope ( hst ) with the wide field planetary camera 2 ( wfpc2 ) and the advanced camera for surveys ( acs ) , and a log of these observations is presented in table [ tab : obs ] . + the wfpc2 observations from 2008 ( or 776 days post - maximum ) were retrieved from the space telescope science institute hst archive , having been processed through the on - the - fly - recalibration pipeline . photometry of the wfpc2 observations was conducted using the dolphot package @xcite , with the wfpc2 specific module . the position of sn 2006jc fell on the planetary camera chip , which has a pixel scale of 0.05 arcsec . + the 2010 observations ( 1303 days post - maximum ) were acquired using the @xmath24 subarray of the acs wide field channel ( wfc ) 1 . the observations , in each filter , were composed of four separate dithered exposures to aid in improving the sampling of the point spread function ( psf ) . the individual exposures were subject to bias striping noise " , leading to obvious horizontal stripes across each image @xcite . the horizontal noise features were almost completely removed using the acs_destripe_plus package , running in the pyraf environment , however at low levels some evidence of these stripes is just perceivable in the corrected images . the observations were then processed and combined using the astrodrizzle package , which also corrects for the geometric distortion of the acs wfc cameras . we found that attempts to decrease the output pixel size to @xmath25 arcsec resulted in obvious aliasing patterns in the final combined images for each filter ; and , therefore , we only drizzled the observations to a final pixel size of @xmath26 arcsec , matching the original acs wfc pixel scale . photometry of the acs observations was conducted using dolphot with the acs specific module . + for both the wfpc2 and acs observations , photometric limits were determined using artificial star tests in which fake stars were inserted into the images and recovery was attempted using the same algorithm used for detecting and conducting photometry on real sources . an artificial star was deemed to have been successfully detected if it was recovered within 1 pixel of the location at which it was inserted and its measured brightness was within @xmath27 of the input brightness . the detection probability , as a function of magnitude , was parameterised as a cumulative normal distribution . + in order to determine the position of sn 2006jc on the hst observations , we followed the procedure of @xcite . we measured the position of sn 2006jc in a deep 2008 gemini near infrared imager ( niri ) @xmath28-band image ( with a precision of @xmath29 pixels ) . the gemini niri image was aligned with the 2010 acs wfc @xmath14 observation , using @xmath30 stars , on which we were able to determine the sn location with a precision of @xmath31 pixels or 0.035 arcsec . the position was further mapped ( using @xmath32 stars ) to the 2008 wfpc2 @xmath14 observation , with a final uncertainty of @xmath33 pixels ( or @xmath34 ) . the site of sn 2006jc in the late - time hst acs observations is shown in figure [ fig : res : image ] . although there is a large concentration of bright sources to the south east of the position of sn 2006jc , the sn position itself is remarkably isolated . a source is recovered close to the transformed sn position ( with an offset @xmath35 arcsec or @xmath36 pixels ) in the 2010 acs observations with @xmath37 ( @xmath38 ) , @xmath39 ( 4.7 ) and @xmath2 ( 5.8 ) mags . the values for sharpness and @xmath40 for the psf - fitting photometry conducted by dolphot suggested , for each of the three filters , that the late - time source was point - like and not extended . the source is , however , not significantly detected ( @xmath41 ) in the @xmath20 observation , which would include @xmath3 emission at the redshift of the host galaxy , to a limit of @xmath42 mags . no source is recovered at the sn position in the 2008 wfpc2 observations , to limits of @xmath43 , @xmath44 , @xmath45 and @xmath46 mags . a comparison of the photometry of the late - time source with the colour - colour sequence for supergiants , derived using atlas9 synthetic spectra ( * ? ? ? * with the parameters for supergiants suggested by @xcite ) , and for stellar clusters using starburst99 @xcite is shown on fig . [ fig : res:2col ] . + determining the exact nature of the late - time source at the position of sn 2006jc is complicated by the source s detection at only one epoch , and at very faint levels . there are a number of possible scenarios for the origin of this flux , if it is actually associated with sn 2006jc : brightness arising from the sn itself at late - times ; a light echo ; a host cluster ; an associated star ; or a spatially coincident , but unassociated star . + , @xmath14 and @xmath22 observations and the position of sn 2006jc is indicated by the cross - hairs . the image is oriented with north up and east to the left.,width=302 ] ) , with respect to the two - colour sequences ( @xmath47 ) for supergiants ( solid grey ) in the temperature range @xmath48 and starburst99 cluster models ( dotted dark grey line ) for the age range @xmath49 . points along the supergiant two - colour sequence delineate @xmath50 intervals ( as labeled ) . also shown , as the arrow , is the reddening vector corresponding to @xmath51 mag . the light grey point shows the last complete @xmath52 photometric measurement made by @xcite at 109.7 days post - maximum . the dark triangle ( @xmath53 ) indicates the colour predicted for a light echo , derived from integrated colour of the observed light curve of sn 2006jc.,width=283 ] @xcite , @xcite , @xcite , @xcite , @xcite and @xcite presented optical lightcurves of sn 2006jc covering up to 180d after maximum light , showing the sn 2006jc exhibited a precipitous decline in brightness ( @xmath54 in the @xmath55-band ) . if the light curve decay suddenly flattened at this time , for sn 2006jc to reach the brightness of the source observed in 2010 would require the decline rate to drop dramatically to @xmath56 ( or even lower if the flattening of the lightcurve occurred later ) . the lack of a detection of sn 2006jc in the 2008 wfpc2 observations suggests that the rate of decline observed at 180d continued at much later epochs . the last reported three - colour photometry of sn 2006jc in 2007 reported by @xcite , in comparable bands to the 2010 observations , also shows that sn 2006jc becoming redder in the @xmath57 colour compared to the source recovered in 2010 ( see fig . [ fig : res:2col ] ) . some caution , however , is required due to the large uncertainties associated with the later photometry of sn 2006jc and the photometry of the 2010 source . based on the red late - time colour of sn 2006jc and the severe implications for a very sudden flattening of the light curve after @xmath58 we consider it unlikely that the source recovered in the 2010 observations is sn 2006jc . the point - like nature of the late - time source ( fwhm @xmath59 pix , corresponding to 15.5 pc at the distance of ugc 4904 ) is evidence against its origin in a light echo arising in an interstellar dust sheet , as it would require a very specific configuration of a sheet of scattering dust located immediately behind sn 2006jc . observations by @xcite , @xcite and @xcite indicated the presence of significant dust formation after @xmath60 , with @xcite measuring @xmath61 mags , arising from dust within @xmath62 of the point of the explosion , 230 days post - explosion . based on observations at mid - infrared wavelengths @xcite also reported the presence of a pre - existing shell of dust extending to @xmath63 around the progenitor of sn 2006jc in which a light echo might arise and which , at its maximum extent , would remain unresolved in these late - time observations . at the time of these observations , however , the light echo would arise _ behind _ sn 2006jc , requiring scattering through angles @xmath64 which is inefficient @xcite . in addition , the contribution of an optical light echo is expected to produce a sed which is very close to that of the sn at early epochs . to assess the likely colour of a light echo , we considered the colour of sn 2006jc integrated over the entire light curve reported by @xcite and @xcite . we conservatively assume that the brightness from the time of explosion to the first photometric measurements , which were only acquired after maximum , is constant . we find the total integrated colour , under these conditions , to be @xmath65 and @xmath66 , which is consistent with the earliest observed colours of @xmath67 and @xmath68 mags ( see , e.g. , * ? ? ? * ) and bluer than the photometry of the 2010 source ( see fig . [ fig : res:2col ] ) . the evolution of the observed colour of sn 2006jc and other type ibn sne ( e.g. sn 2010al ; * ? ? ? * ) suggest , however , the sn may have been even bluer prior to the first photometric measurements , implying that our integrated colour is in fact a red limit for the colour of a possible light echo . as evident in fig . [ fig : res : image ] , sn 2006jc occurred in a relatively sparse area of ugc 4904 . in an area within 1 arcsec of the position of sn 2006jc , we find 10 sources recovered with @xmath69 , including the source at the sn position . given the density of sources within the 1 arcsec area , we estimate the probability of finding one bright source inside the @xmath27 error radius on the position of sn 2006jc to be 0.01 , implying that it is unlikely to find by chance an unrelated source to be spatially coincident with the sn position . + overall , given the brightness , colours and spatial extent of the late - time source derived from the 2010 acs observations , the simplest explanation is that this source is a star and is likely to be associated with sn 2006jc . future observations with hst will be required to confirm the continued presence of this source if it is indeed a star . on the colour - colour plane ( see fig . [ fig : res:2col ] ) , the position of the late - time source is coincident the supergiant two - colour sequence for limited reddening . assuming a galactic @xmath70 @xcite reddening law and the total extinction measured towards sn 2006jc at early times by @xcite of @xmath71 mag , the corresponding probability density function for the position of the source on the hertzsprung - russell ( hr ) diagram , in conjunction with cambridge stars models @xcite , is shown on fig . [ fig : res : hrd ] . if no prior is assumed for the reddening , the colours of the late - time source , assuming a stellar origin , limit the total reddening to @xmath72 mag . while the filter combination used for the 2010 acs wfc observations can be used to place a constraint on the degree of reddening ( and , therefore , extinction ) , it is not a sensitive diagnostic of stellar temperature at high temperatures . the range of temperatures and luminosities that can be accommodated by the observed photometry primarily reflects the large range in possible bolometric corrections , rather than being due to uncertainties in reddening . despite the extent of possible locations for this source on the hr diagram , we note this star is outside of the regions associated with s doradus variables in either quiescence or outburst @xcite . if the late - time source is an unresolved cluster , which might have hosted the progenitor of sn 2006jc , we note that absolute magnitude of the source , excluding considerations of extinction , is @xmath73 , which is 3 mags fainter than the suggested brightness at which unresolved clusters may be confused for individual stars @xcite . in comparison with starburst99 models @xcite the colours of the late - time source are consistent with an unresolved cluster for @xmath74 mags . if sn 2006jc occurred in a host cluster , any extinction that the cluster was subject to would also have affected the sn at early times , which was not observed . in addition , there is insufficient time for any dust formed in sn 2006jc to be distributed on spatial scales similar to the typical sizes of clusters to extinguish a host cluster by the time of the 2010 observations . based on the properties of the late - time source we cautiously conclude that the most likely explanation is that it is a star and the companion to the progenitor that exploded as sn 2006jc . the exact nature of the star is difficult to constrain given the available late - time hst photometry . the fact that the locus of the star does not correspond exactly with the two locations in which s doradus variables are found on the hr diagram would argue against this star being an lbv , but rather a normal supergiant . this conclusion is further supported by the absence of strong @xmath3 emission at late - times as might be expected of classical lbvs ; however we note that ugc2773 - 2009ot and lmc r71 also do not show strong @xmath3 at late - times afters outbursts and that the limited depth of the late - time @xmath20 observation might not probe weak @xmath75 if it were present . + an important question raised by the possible binary nature of the progenitor system of sn 2006jc is which component was responsible for the 2004 precursor outburst . the late - time hst photometry , at the lowest luminosity limit , is consistent with a @xmath76 a - f supergiant ( see fig . [ fig : res : hrd ] ) , which would make it comparable to properties of the progenitors identified for faint outburst events such as sn 2008s and ngc 300-ot @xcite , but still fainter and less massive than the a - to - f - type hypergiant progenitors of more classical sn impostors ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in direct contrast to the former two events , however , there is no evidence that the late - time source at the position of sn 2006jc is enshrouded in significant quantities of dust . furthermore , apart from the precursor and sn 2006jc , at the sn position , there is no evidence for any significant outburst events after the explosion of the sn . long term monitoring of the site of sn 2006jc , before and after the 2004 outburst and the 2006 sn explosion , is shown in fig . [ fig : disc : monit ] . the limited history of outbursts at the location of sn 2006jc would suggest , therefore , that any significant variability ended with the sn . + it is interesting to note that the location of sn 2006jc is offset ( by @xmath77 or a projected distance of @xmath78 ) from the nearest obvious sites containing young massive stars ( see fig . [ fig : res : image ] ) . similar conditions have been observed for the hydrogen - rich interacting type iin sne and some lbvs @xcite ; although @xcite note that classical lbvs are found in close proximity to late - type o stars while non - classical lbvs are found in relative isolation ( this matter was further debated in the exchange between @xcite and @xcite ) . the absence of significant @xmath3 emission in the environment of sn 2006jc is also consistent with the observations of @xcite for the positions of interacting sne with respect to hii regions . this could support the interpretation of the properties of the late - time source as a lower mass star ( @xmath79 ) , that is not associated with recent massive star formation . the companion would then have to have been sufficiently close to the progenitor to interact with it to induce the outburst ; however , the absence of strong h features from the subsequent sn spectra would require the companion to be sufficiently distant at the time of explosion ( e.g. @xmath80 of h in type iib sne ; although we note that @xcite , @xcite and @xcite report the emergence of weak @xmath81 emission at @xmath8251 days which could be related to the putative companion star ) . we compared the position of the late - time source on the hr diagram with the locus of binary companions , at the time of the explosion of the primary , predicted by binary population and spectral synthesis ( bpass ) models @xcite , as shown on figure [ fig : res : hrd ] ( see also * ? ? ? * ) . the locus for the companions , generally following the main sequence and its turn - off , intersects with the contours for the late - time source at higher masses ( @xmath83 ) . if the companion was indeed massive , this would favour a high mass wr progenitor . we caution , however , that this family of models do not predict the progenitor and binary companion for sn 1993j , given the observed constraints on the binary progenitor system determined by @xcite . @xcite present a model whereby an outburst can be triggered by the interaction of a main sequence companion star with the extended envelope of an evolved primary , with eccentric orbits making the outbursts more extreme . @xcite proposes a similar model , however it requires a collision between the two stars to produce an outburst ; which can not be the case for sn 2006jc due to the persistence of the source at the sn position after the explosion . an interesting consequence of repeated periastron passages noted by @xcite is the removal of h envelope of the primary and the increasing eccentricity of the orbit , with @xcite specifically noting the case of wr 140 . the dependence of the mass loss history of the primary on the orbital properties of the binary system might also explain the evidence for episodic mass loss in the form of shells of material discernible in early - time spectra of sn 2006jc @xcite . given the importance of binary interactions in the evolution of massive stars @xcite previous encounters with the companion may also have responsible for the h - poor nature of the progenitor and not just the single observed pre - explosion outburst . an additional possible progenitor scenario for sn 2006jc is suggested by the successful identification @xcite and confirmation , through disappearance , of the low mass progenitor of the type ib sn iptf 13bvn @xcite . in this scenario the progenitor of sn 2006jc need not have been a massive star capable of evolving into a wr star or an lbv , but was rather a lower mass star ( @xmath84 ) that was stripped of its h envelope and that then evolved as a helium giant . the luminosities of such stars would be insufficient to drive strong winds , comparable to higher mass wr stars , but would instead have extended he - rich envelopes @xcite . due to the structural differences between helium giants and massive wr stars , the interaction between a low mass helium giant and a binary companion could be different to that expected for a binary system containing a wr star , and might be more amenable to the production of pre - explosion outbursts and the laying down of a dense csm . detailed calculations of such a binary configuration are , however , beyond the scope of this paper . as noted above , however , the predicted locus for the low mass binary companion to the progenitor of iptf 13bvn presented by @xcite lies at higher temperatures than we have derived for the source recovered in the 2010 observations . ultimately , confirmation with future hst observations of the sn location to required to test if the source is still present or if it has disappeared and to identify which of the above scenarios is responsible for the late - time brightness of sn 2006jc . the research of jrm is supported through a royal society university research fellowship . ap is partially supported by the prin - inaf 2014 with the project `` transient universe : unveiling new types of stellar explosions with pessto '' . we thank noam soker for very useful discussions during the preparation of the manuscript . we thank the team members of the italian supernovae search project ( issp : http://italiansupernovae.org/ ) , in particular fabio briganti and fabio martinelli of the astronomical centre , lajatico , pisa ( http://www.astronomicalcentre.org/ ) and fabrizio ciabattari of the astronomical observatory of monte agliale , borgo a mozzano , lucca ( http://www.oama.it/ ) for sharing their images of ugc 4904 with us . we are also grateful to t. yusa for his help in collecting data from amateur astronomers . this research has made use of the nasa/ ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .	we present late - time observations of the site of the type ibn supernova ( sn ) 2006jc acquired with the hubble space telescope advanced camera for surveys . a faint blue source is recovered at the sn position with brightness @xmath0 , @xmath1 and @xmath2 mags , although there is no detection in a contemporaneous narrow - band @xmath3 image . the spectral energy distribution of the late - time source is well fit by a stellar - like spectrum ( @xmath4 and @xmath5 ) subject to only a small degree of reddening consistent with that estimated for sn 2006jc itself at early - times . the lack of further outbursts after the explosion of sn 2006jc suggests that the precursor outburst originated from the progenitor . the possibility of the source being a compact host cluster is ruled out on the basis of the source s faintness , however the possibility that the late - time source maybe an unresolved light echo originating in a shell or sphere of pre - sn dust ( within a radius @xmath6 ) is also discussed . irrespective of the nature of the late - time source , these observations rule out a luminous blue variable as a companion to the progenitor of sn 2006jc .
the primordial inflation explorer ( pixie)@xcite is a space - based polarizing fourier transform spectrometer ( fts ) designed to measure the polarization and intensity spectra of the cosmic microwave background ( cmb ) . as for previous fts - based instruments flown to measure the cmb @xcite , pixie s design and experimental approach represent a significant departure from the focal plane imagers most commonly used for these measurements . this is especially true for the detectors . instead of requiring several thousand diffraction - limited , ultra - low - noise detectors , pixie can achieve nk - scale sensitivity across 2.5 decades in frequency with just four multimode polarization - sensitive bolometers based on silicon thermistors . with a large etendue @xmath1 of @xmath2 @xmath3 sr per detector , the detectors are designed for a high optical load ( @xmath4 pw ) , but their noise equivalent power ( nep ) is near the thermodynamic limit and is subdominant to photon noise from the cmb . using high impedance silicon thermistor - based bolometers allows the use of simple and mature junction field effect transistor ( jfet)-based voltage amplifiers . as a nulling experiment , where the signal is a small modulated component in a bright background , the detectors will always operate where assumptions of linearity are strong . a series of discrete symmetries built into the instrument enable multiple detectors to measure the same signal , allowing measurement and control of detector - sourced systematic effects . developing detectors for a fts with a large but mechanically robust absorbing area ( @xmath0 times that of planck s spider - web bolometers @xcite ) , large enough bandwidth and appropriate geometry to measure optical frequencies from 15 ghz to 5 thz , and sufficiently low nep ( @xmath5 w@xmath6 ) requires meeting a unique set of design , fabrication , and performance criteria . we describe these in this paper . complete descriptions of the pixie instrument are available@xcite , but here we reproduce the highlights that drive bolometer design and performance requirements . a cartoon of the pixie fts is shown in figure [ fig : fts ] . light is directed into the fts by the primary mirrors , folding flats , and secondary mirrors . polarizer a defines the instrument s polarization basis , transmitting horizontal ( @xmath7 ) polarization and reflecting vertical ( @xmath8 ) polarization . polarizer b is oriented at 45@xmath9 relative to polarizer a and mixes the beams . the moving mirror injects an optical phase delay . polarizer c has same orientation as polarizer b and sorts the beams . polarizer d has the same orientation as polarizer a and again splits polarizations . light is then directed into polarization - maintaining receiver horns@xcite and onto the focal planes . each focal plane consists of two bolometers mounted back - to - back with their polarization axes orthogonal , allowing a simultaneous measurement of both linear polarization states . the low frequency cutoff of the instrument ( @xmath10 ghz ) is set by the etendue , and the high frequency cutoff ( @xmath11 thz ) is set by a series of filters and the polarizer grid spacing . is @xmath12 mm , which corresponds to an optical path difference @xmath13 of @xmath14 mm . the mirror completes a stroke from @xmath15 mm to @xmath16 mm in 3 seconds . the optical path difference between beams @xmath13 is related to the frequency of incident radiation @xmath17 by the relationship @xmath18 , where @xmath19 is the speed of light . the frequency of the mirror movement @xmath20 is related to the frequency of incident radiation by @xmath21 . where @xmath22 is the moving mirror s velocity . the cmb signal is largely confined to acoustic frequencies below 15 hz . the dust signal is confined to acoustic frequencies below 100 hz . these constraints drive the bolometer bias and bandwidth requirements . right : simulated time domain signal ( interferogram ) incident on a pixie bolometer . this shows the fourier transform of the polarized cmb.,scaledwidth=33.0% ] light incident on the instrument is represented by @xmath23 . if both beams are open to the sky , the power measured by the detectors is @xmath24 where the superscripts @xmath25 and @xmath26 indicate whether the detector is on the left of right side of the fts ( see figure [ fig : fts ] ) , the subscripts @xmath27 and @xmath28 indicate whether the detector is measuring @xmath7 or @xmath8 polarization , @xmath17 is the frequency of incident radiation , @xmath29 is the mirror position , and @xmath19 is the speed of light . each power expression given in equation [ eq : p ] consists of a dc term and a term modulated by the movement of the mirror . the former represents the total intensity of incident light , and the latter is the fourier transform of the difference spectrum between orthogonal incident polarizations ( stokes @xmath30 in instrument - fixed coordinates ) . given the small linearly polarized fraction of the microwave sky , the signal measured by the detectors thus consists of a small modulated component on top of near - constant optical bias . taking the inverse fourier transform of the measured signals removes the dc component and we are left with the spectra of polarized light measured by each detector : @xmath31 where the subscript @xmath17 indicates that we are working in the frequency domain . the pixie bolometers are designed to measure this signal . of the endbank s gold thermalization bar and the effective thermal conductance @xmath32 ( which accounts for electrothermal feedback gain ) of the silicon legs between the endbank and the chip frame . these can be independently tuned . with indium bumps the pair of chips will be stood off from each other by @xmath33 @xmath34m.,title="fig:",scaledwidth=45.0% ] of the endbank s gold thermalization bar and the effective thermal conductance @xmath32 ( which accounts for electrothermal feedback gain ) of the silicon legs between the endbank and the chip frame . these can be independently tuned . with indium bumps the pair of chips will be stood off from each other by @xmath33 @xmath34m.,title="fig:",scaledwidth=52.0% ] a cartoon of a pixie bolometer and a cross sectional view of a hybridized pair of bolometers is shown in figure [ fig : bolo_cart ] . the bolometer fabrication and ion implantation are performed in the detector development lab ( ddl ) at nasa / gsfc . the bolometer consists of an optical absorber structure coupled to silicon thermistor - based thermometers that operate at @xmath35 mk under optical and electrical bias . the thermometers are coupled to the chip frame held at the bath temperature ( @xmath36 mk ) . the absorber structure ( section [ sec : abs ] ) is a large - area ( 1.61 @xmath3 ) grid of freestanding , micromachined , degenerately doped silicon wires . the thermistors are located on silicon membranes ( `` endbanks '' , section [ sec : eb ] ) at either end of the absorber structure . each endbank consists of a gold bar for thermalization and two thermistors that are independently biased and read out for laboratory characterization but will be wired in series or parallel in flight . the endbanks are coupled to the chip frame ( section [ sec : frame ] ) through a series of silicon legs . on the chip frame are wirebond pads for bias and readout of the thermistors and an array of indium bumps enabling the hybridization of a pair of bolometer chips . when hybridized , orthogonal absorbers will be separated by @xmath33 @xmath34 m , such that the two bolometers will sample orthogonal polarizations of nearly the same electric field . -direction ( the direction of absorption ) , the absorbers are flat to less than 2 @xmath34 m . the periodic structure visible along the @xmath8-direction is set by the endbank leg placement , but with an amplitude of @xmath37 @xmath34 m and a period of several mm , they absorbers are essentially flat . the chip frame is visible around the edge of the image . since the strings never protrude more than @xmath38 @xmath34 m above the frame , a hybridized pair can be offset by fewer than 20 @xmath34 m without risk of strings colliding during launch.,title="fig : " ] -direction ( the direction of absorption ) , the absorbers are flat to less than 2 @xmath34 m . the periodic structure visible along the @xmath8-direction is set by the endbank leg placement , but with an amplitude of @xmath37 @xmath34 m and a period of several mm , they absorbers are essentially flat . the chip frame is visible around the edge of the image . since the strings never protrude more than @xmath38 @xmath34 m above the frame , a hybridized pair can be offset by fewer than 20 @xmath34 m without risk of strings colliding during launch.,title="fig : " ] shown in figure [ fig : abs ] , the bolometer absorber structure consists of an array of parallel micromachined silicon wires that absorbs one linear polarization@xcite . they are 3 @xmath34 m wide , 1.35 @xmath34 m thick , 15.8 mm long , and are on a 30 @xmath34 m pitch , making them effective absorbers for frequencies up to @xmath11 thz . the wires are degenerately doped with phosphorous to be metallic at all temperatures . the absorber structure s effective sheet resistance is designed to be 377 @xmath39 , impedance matching it to free space . the string width and thickness are highly uniform across the array . the thickness is set by the device layer thickness on the starting silicon - on - insulator ( soi ) substrate , and they are etched to width with an inductively - coupled plasma ( icp ) reactive ion etch ( rie ) process . in previous generations of devices , we found that the doping process induced compressive stress in absorber strings along the axis of absorption , causing individual strings to buckle and protrude @xmath40 @xmath34 m from the chip frame . this is problematic for a hybridized pair of chips which are separated by @xmath33 @xmath34 m , since absorbers could collide and break , particularly during vibrations at launch . to prevent this , and to raise the resonant frequencies of the strings to well above the frequencies they will experience at launch , we deposited a highly - tensile al@xmath41o@xmath42 film on individual strings outside the active absorbing area . this feature effectively pulls the strings taught ( see figure [ fig : abs ] ) , and based on our elastic model should keep the amplitude of any string vibrations to less than @xmath43 @xmath34 m rms over the band of excitation frequencies expected during launch ( 20 hz - 10 khz ) . we also estimated the bolometer sensitivity to cosmic ray hits . to minimize the impact of cosmic ray hits , the time constant of the absorber structure should be much shorter than the hit rate . based on planck s measured hit rate @xcite , and scaling for our geometry , we expect one cosmic ray hit every two seconds . measurements and thermal models of our absorber structure yield a time constant of a few ms , so particle hits can easily be flagged and removed during data processing without corrupting an entire interferogram . each bolometer chip has two endbanks , one at each end of the absorber structure ( figure [ fig : abs ] ) . the endbanks consist of a gold bar for thermalization and two doped silicon thermistors on a crystalline silicon membrane . the gold bar also sets the heat capacity of the endbank . the endbank membrane is the same thickness ( 1.35 @xmath34 m ) as the absorber strings , and is also formed from the device layer of the soi substrate . endbanks are connected to the chip frame through eight silicon legs . the thermistors are doped to operate below their metal - insulator transition . the electron transport mechanism in this regime is variable range hopping ( vrh)@xcite , which yields the following resistance - temperature relationship : @xmath44 where @xmath45 and @xmath46 are constants determined largely by the device geometry and doping concentration , respectively . we also developed a detailed thermal model of the endbank that takes into account heat flow between each thermal element ( absorber strings , gold bar , thermistor and legs ) . this will be discussed in greater detail in section [ sec : perf ] . the bolometer frame is maintained at the refrigerator bath temperature . under normal operation , the bath will be kept at @xmath36 mk . the chip is maintained at this temperature through a series of gold wire bonds connecting gold heat sink areas on the chip to the package . the frame is designed such that any two bolometer chips can be indium bump hybridized together , though we can also characterize individual bolometers in the light or dark ( figure [ fig : packaged ] ) . the frame has several arrays of indium bumps . some are designed for mechanical purposes , and others are electrical , allowing the thermistors on one chip to be read out through wirebond pads on the other chip . -length absorbers , enabling it to fit in a package designed for an x - ray microcalorimeter array . right : basic bias and readout circuit used for dark detector characterization . this is not a complete schematic , but rather a representation of what is used.,title="fig:",scaledwidth=50.0% ] -length absorbers , enabling it to fit in a package designed for an x - ray microcalorimeter array . right : basic bias and readout circuit used for dark detector characterization . this is not a complete schematic , but rather a representation of what is used.,title="fig:",scaledwidth=49.0% ] we measured detector load curves and noise in a dark cryogenic environment at a range of temperatures ( @xmath47 mk ) . a photograph of a packaged bolometer and a schematic of the basic bias and readout circuit are shown in figure [ fig : packaged ] . the package , bias and readout system , and dewar were originally built to characterize the silicon thermistor - based x - ray microcalorimeters developed for the astro - e2/suzaku mission @xcite . the thermistor is wired in series with a bias resistor . the bias resistor s resistance is much greater than the thermistor s operating resistance , so the thermistor is under current bias . tensioned leads connect the bolometer to a cryogenic jfet amplifier , mitigating capacitive microphonic contamination of the signal band . for pixie , we use interfet nj14al16 jfets@xcite that operate at @xmath48 k in a source - follower configuration . these are the same model jfets that were flown as part of the xrs instrument on astro - e2/suzaku@xcite and the sxs instrument on astro - h / hitomi@xcite . they convert the high source impedance of the thermistors ( m@xmath49-scale ) to the low output impedance of the jfets ( @xmath50 k@xmath49 ) . the jfets have excellent noise performance ( @xmath51 @xmath52 at 100 hz ) . the low impedance signal is then ac coupled to a room temperature amplifier . as a function of bias power @xmath53 , where @xmath54 is the voltage drop across the thermistor and @xmath55 is the thermistor bias current . data are shown for a pair of thermistors wired in series on one of the two endbanks . the data from the other endbank are similar . because the bias circuit is optimized for low temperature , high impedance operation , it could not supply enough current for us to observe thermistor self - heating at the higher bath temperatures . from the low bias resistances , where the thermistor is isothermal with the bath , we can extract the vrh parameters . right : average thermal conductance @xmath56 as a function of thermistor temperature . the legend shows the bath temperature corresponding to the measurement . shown in circles is the model prediction.,title="fig:",scaledwidth=49.0% ] as a function of bias power @xmath53 , where @xmath54 is the voltage drop across the thermistor and @xmath55 is the thermistor bias current . data are shown for a pair of thermistors wired in series on one of the two endbanks . the data from the other endbank are similar . because the bias circuit is optimized for low temperature , high impedance operation , it could not supply enough current for us to observe thermistor self - heating at the higher bath temperatures . from the low bias resistances , where the thermistor is isothermal with the bath , we can extract the vrh parameters . right : average thermal conductance @xmath56 as a function of thermistor temperature . the legend shows the bath temperature corresponding to the measurement . shown in circles is the model prediction.,title="fig:",scaledwidth=50.0% ] a series of measured load curves is shown in figure [ fig : bias_noise ] , plotted to show thermistor resistance as a function of bias power . from the measured resistances under low electrical bias , where the thermistor is isothermal with the bath , we determine the vrh parameters @xmath45 and @xmath46 ( equation [ eq : vrh ] ) . for this device , fitting to the data gives @xmath57 k and @xmath58 @xmath49 . the expected operating resistance under optical bias is @xmath59 m@xmath49 , matching well the input impedance of the jfets . the average thermal conductance @xmath56 between the thermistors and the bath is calculated from the high bias end of the load curves where self - heating is observed . in a steady state , the bias power @xmath60 is equal to the heat flow from the thermistor to the bath , so @xmath56 is given by @xmath61 where @xmath62 is the thermistor temperature ( calculated from equation [ eq : vrh ] ) , and @xmath63 is the bath temperature . this is plotted for the 100 mk data in figure [ fig : bias_noise ] . we then fit to the measured @xmath56 with a function @xmath64 , also shown in figure [ fig : bias_noise ] . the fit is close to the expected value for phonon conduction at these temperatures ( @xmath65 ) . the dc electrical responsivity @xmath66 is also determined from the load curves . following the notation of jones @xcite , the responsivity is given by the following expression : @xmath67 where @xmath26 is the resistance @xmath68 and @xmath69 is is the dynamic resistance @xmath70 , determined from the load curve . the measured responsivity @xmath66 is plotted as a function of current @xmath55 for a few temperatures in figure [ fig : se_noise ] . as a function of bias current for a few bath temperatures . right : measured noise spectrum along with the fit from our thermal model . note that the measurement was ac coupled , so both measured signals drop off at low frequency . running this model for the optical and electrical bias conditions in expected in flight , we calculate a bolometer nep of @xmath71 w/@xmath72 . this is well below the expected photon noise from the cmb across the entire pixie bandwidth ( see figure [ fig : nep]).,title="fig:",scaledwidth=49.0% ] as a function of bias current for a few bath temperatures . right : measured noise spectrum along with the fit from our thermal model . note that the measurement was ac coupled , so both measured signals drop off at low frequency . running this model for the optical and electrical bias conditions in expected in flight , we calculate a bolometer nep of @xmath71 w/@xmath72 . this is well below the expected photon noise from the cmb across the entire pixie bandwidth ( see figure [ fig : nep]).,title="fig:",scaledwidth=49.0% ] we compared the measured bolometer characteristics to predictions from a thermal model that takes into account heat exchanges between the gold bar , the thermistors , and the legs . given the endbank geometry , we solve for the etendue between each thermal component , then determine the heat flow between each of them . heat flow between components on the endbank is via ballistic phonon conduction , where the phonon mean free path @xmath73 is much greater than the endbank dimensions . this assumption is based on thermal transport measurements made on far infrared microwave kinetic inductance detectors ( mkids)@xcite , where @xmath73 in silicon at 100 mk was measured to be @xmath74 mm@xcite . for heat flowing out the legs , we account for a mean free path reduction due to phonon scattering off the leg edges . the model is then used to predict the dc thermal conductance ( figure [ fig : bias_noise ] ) and the noise ( section [ sec : noise ] ) . noise spectra were measured at a range of temperatures and bias currents . when pixie observes the cmb the bolometers will operate under high optical bias ( @xmath4 pw ) . this results in a large temperature difference between the thermistors and the bath . it is therefore necessary to take into account non - equilibrium effects in the noise @xcite . accounting for non - equilibrium effects , the nep of a simple bolometer in the dark is given by @xmath75 where the constants @xmath76 , @xmath77 , and @xmath78 account for non - equilibrium effects , @xmath79 and @xmath80 are the amplifier s voltage and current noise spectral densities , and @xmath81 accounts for other sources of noise , for example parasitic resistance in the leads or stray light . the factor of @xmath82 , where @xmath83 is the bolometer s electrical responsivity , refers the johnson , amplifier , and excess noise to the bolometer s input . multiplying the nep@xmath84 by @xmath85 gives the voltage power spectral density of the bolometer . with relevant parameters taken from load curve analysis ( section [ sec : loadcurves ] ) , and accounting for non - equilibrium effects , we calculated the expected bolometer noise under a few bias conditions , and compared them to measured noise spectra . an example noise spectrum and fit is shown in figure [ fig : se_noise ] . fits are consistently good for multiple measured bias conditions . running the model for the bias conditions expected during flight , we expect to be photon noise limited across the entire pixie bandwidth ( figure [ fig : nep ] ) . we designed , fabricated , and characterized large area polarization - sensitive bolometers for the pixie experiment . multimode bolometers designed for a fts like pixie require a different optimization from bolometers developed for focal - plane imagers . in particular , they must handle a large , near - constant optical bias ( @xmath4 pw ) , but still operate with sensitivity near the thermodynamic limit across the relevant bandwidth . they also must have a large and mechanically robust absorbing area that is sensitive to all optical frequencies of interest , but relatively insensitive to particle hits . mechanical characterization of the fabricated pixie bolometers with scanning electron microscopy and white light interferometry shows that our tensioning scheme successfully flattens the absorber strings , enabling indium bump hybridization of a pair of bolometer chips . the resonant frequencies of the tensioned absorbers are also expected to be well above the vibrational frequencies the detectors will experience at launch . we will confirm this in the near future by subjecting a hybridized pair of bolometers to environmental testing . the dark data provide significant insight into the performance of the pixie bolometers , particularly regarding the thermal behavior of the endbanks . our thermal model agress well with the data , and the results indicate that the pixie bolometers satisfy the sensitivity and bandwidth requirements of the space mission ( see figure [ fig : nep ] ) . in the immediate future , we will further characterize the bolometers , specifically targeting the absorber structure . these measurements will include thermal transport measurements of absorbers in the dark , ac impedance measurements , and optical measurements with a cryogenic blackbody source . this work was supported by nasa / gsfc irad funding . we are especially grateful to the x - ray microcalorimeter group at nasa / gsfc for lending the astro - e2/suzaku test platform for pixie detector characterization . , a. , fixsen , d. j. , chuss , d. t. , dotson , j. , dwek , e. , halpern , m. , hinshaw , g. f. , meyer , s. m. , moseley , s. h. , seiffert , m. d. , spergel , d. n. , and wollack , e. j. , `` the primordial inflation explorer ( pixie ) : a nulling polarimeter for cosmic microwave background observations , '' _ _ * 7 * , 025 ( july 2011 ) . , a. , chuss , d. t. , dotson , j. , dwek , e. , fixsen , d. j. , halpern , m. , hinshaw , g. f. , meyer , s. , moseley , s. h. , seiffert , m. d. , spergel , d. n. , and wollack , e. j. , `` the primordial inflation explorer ( pixie ) , '' in [ _ space telescopes and instrumentation 2014 : optical , infrared , and millimeter wave _ ] , _ _ * 9143 * , 91431e ( aug . 2014 ) . , j. c. , cheng , e. s. , cottingham , d. a. , eplee , jr . , r. e. , fixsen , d. j. , hewagama , t. , isaacman , r. b. , jensen , k. a. , meyer , s. s. , noerdlinger , p. d. , read , s. m. , rosen , l. p. , shafer , r. a. , wright , e. l. , bennett , c. l. , boggess , n. w. , hauser , m. g. , kelsall , t. , moseley , jr . , s. h. , silverberg , r. f. , smoot , g. f. , weiss , r. , and wilkinson , d. t. , `` measurement of the cosmic microwave background spectrum by the cobe firas instrument , '' _ _ * 420 * , 439444 ( jan . 1994 ) . , g. s. , gush , h. p. , halpern , m. , shinkoda , i. , and towlson , w. , `` anisotropy in the microwave sky : results from the first flight of the balloon - borne anisotropy measurement ( bam ) , '' _ _ * 475 * , l73l76 ( feb . 1997 ) . , w. a. , bock , j. j. , crill , b. p. , koch , t. c. , jones , w. c. , lange , a. e. , and paine , c. g. , `` initial test results on bolometers for the planck high frequency instrument , '' _ _ * 47 * , 59966008 ( nov . 2008 ) . , a. , wollack , e. j. , and stevenson , t. r. , `` angular and polarization response of multimode sensors with resistive - grid absorbers , '' _ journal of the optical society of america a _ * 31 * , 1557 ( july 2014 ) . , ade , p. a. r. , aghanim , n. , armitage - caplan , c. , arnaud , m. , ashdown , m. , atrio - barandela , f. , aumont , j. , baccigalupi , c. , banday , a. j. , and et al . , `` planck 2013 results . x. hfi energetic particle effects : characterization , removal , and simulation , '' _ _ * 571 * , a10 ( nov . 2014 ) . , r. l. , mitsuda , k. , allen , c. a. , arsenovic , p. , audley , m. d. , bialas , t. g. , boyce , k. r. , boyle , r. f. , breon , s. r. , brown , g. v. , cottam , j. , dipirro , m. j. , fujimoto , r. , furusho , t. , gendreau , k. c. , gochar , g. g. , gonzalez , o. , hirabayashi , m. , holt , s. s. , inoue , h. , ishida , m. , ishisaki , y. , jones , c. s. , keski - kuha , r. , kilbourne , c. a. , mccammon , d. , morita , u. , moseley , s. h. , mott , b. , narasaki , k. , ogawara , y. , ohashi , t. , ota , n. , panek , j. s. , porter , f. s. , serlemitsos , a. , shirron , p. j. , sneiderman , g. a. , szymkowiak , a. e. , takei , y. , tveekrem , j. l. , volz , s. m. , yamamoto , m. , and yamasaki , n. y. , `` the suzaku high resolution x - ray spectrometer , '' _ _ * 59 * , 77112 ( jan . 2007 ) . , f. s. , adams , j. s. , brown , g. v. , chervenak , j. a. , chiao , m. p. , fujimoto , r. , ishisaki , y. , kelley , r. l. , kilbourne , c. a. , mccammon , d. , mitsuda , k. , ohashi , t. , szymkowiak , a. e. , takei , y. , tashiro , m. , and yamasaki , n. , `` the detector subsystem for the sxs instrument on the astro - h observatory , '' in [ _ space telescopes and instrumentation 2010 : ultraviolet to gamma ray _ ] , _ _ * 7732 * , 77323j ( july 2010 ) . brown , a .- hsieh , w .- t . , moseley , s. h. , stevenson , t. r. , kongpop , u. , wollack , e. j. , et al . , `` fabrication of an absorber - coupled mkid detector and readout for sub - millimeter and far - infrared astronomy , '' in [ _ spie astronomical telescopes+ instrumentation _ ] , 77410p77410p , international society for optics and photonics ( 2010 ) . , abergel , a. , ade , p. a. r. , aghanim , n. , alves , m. i. r. , aniano , g. , armitage - caplan , c. , arnaud , m. , ashdown , m. , atrio - barandela , f. , and et al . , `` planck 2013 results . all - sky model of thermal dust emission , '' _ _ * 571 * , a11 ( nov . 2014 ) .	the primordial inflation explorer ( pixie ) is an explorer - class mission concept designed to measure the polarization and absolute intensity of the cosmic microwave background . in the following , we report on the design , fabrication , and performance of the multimode polarization - sensitive bolometers for pixie , which are based on silicon thermistors . in particular we focus on several recent advances in the detector design , including the implementation of a scheme to greatly raise the frequencies of the internal vibrational modes of the large - area , low - mass optical absorber structure consisting of a grid of micromachined , ion - implanted silicon wires . with @xmath0 times the absorbing area of the spider - web bolometers used by planck , the tensioning scheme enables the pixie bolometers to be robust in the vibrational and acoustic environment at launch of the space mission . more generally , it could be used to reduce microphonic sensitivity in other types of low temperature detectors . we also report on the performance of the pixie bolometers in a dark cryogenic environment .
the friendship paradox is a phenomenon observed in various social networks . the term was coined by feld @xcite . it has been empirically observed that people s perception of their own popularity is self - aggrandizing ; most people believe that they are more popular than their friends on average @xcite . however , feld observed that in reality , most people have fewer friends than their friends do . in @xcite , this phenomena is used for the early detection of flu outbreaks among college students . in @xcite , it is utilized to efficiently sample early - warning sensors during catastrophic events such as hurricanes . in addition to degree , the same paradox has been observed about other individual attributes ( called the _ generalized friendship paradox _ @xcite , or gfp ) . for example , in @xcite it has been observed that on twitter , for most people , their friends share , on average , more viral content and also tweet more . in @xcite , it has been observed that in scientific collaboration networks , one s co - authors have , on average , more citations , more publications and more co - authors . in this paper , we consider a network growth model which is a generalization of the preferential attachment scheme @xcite . in our model , nodes are endowed with ` qualities ' ( ak.a . ` fitness ' or ` attractiveness ' in the literature @xcite ) . qualities are discrete positive numbers drawn from a given distribution @xmath0 and assigned to a node upon its birth ( remaining the same thenafter ) . we assume that the probability that node @xmath1 with degree @xmath2 and quality @xmath3 receives a link from subsequent nodes is proportional to @xmath4 . we obtain two statistical measures of this model : one is the degree - quality joint distribution , which is the fraction of nodes that have degree @xmath5 and quality @xmath6 in the steady state . the second quantity is the nearest - neighbor distribution of quality and degree : it gives the fraction of nodes with degree @xmath7 and quality @xmath8 that are connected to a node with degree @xmath5 and quality @xmath6 . equipped with these distributions , we can quantify the paradox and study how it depends on the underlying quality distribution @xmath0 . to our knowledge , no similar theoretical result is available in the literature for any network growth model ( either purely preferential @xcite , or fitness - based @xcite ) . we show that employing the above scheme as the attachment mechanism renders the occurrence of the gfp contingent upon the underlying distribution of node qualities . we then employ measures defined in the literature for assessing the gfp on the network level , and we investigate the dependence of these measures on the model parameters and the quality distribution . we demonstrate that , in the proposed model , the network exhibits a quality paradox at the network level for any quality distribution . we contend that this is indicative of a positive correlation between degree and quality ; i.e. , those with higher qualities are more likely to have higher degrees , and vice versa . in the growth model considered in this paper , nodes are added successively to the network . the initial network has @xmath9 nodes and @xmath10 links . at each time step , one new node is added to the network . we assume that each node has an intrinsic quality , which is drawn from a given distribution @xmath0 . the quality is assigned to each new incoming node upon birth , and will remain the same thenafter . the mean of the distribution @xmath0 is denoted by @xmath11 . a node of degree @xmath5 and quality @xmath6 is also referred to as _ a @xmath12 node _ throughout . each new incoming node attaches to @xmath13 existing nodes in the network . we consider the simplest additive model that incorporates both degree ( popularity ) and quality in the dynamics of connection formation : the probability that an existing node with degree @xmath5 and quality @xmath6 receives a link from the new node is proportional to @xmath14 . this means that , for example , a paper that is new and has very few citations can compensate for its small degree with having a high quality . or in the social context , a newcomer who does not have many friends in the new social milieu but is gregarious and sociable can elevate the chances of making new friends . the new node is called the _ child _ of the existing nodes that it connects to , and they are called its _ parents_. by _ a @xmath15-@xmath12 child - parent pair _ , we mean a node with degree @xmath7 and quality @xmath8 that is connected to a parent node of degree @xmath5 and quality @xmath6 . the probability that an existing node @xmath1 receives a new link is @xmath16 , where the normalization factor @xmath17 is given by @xmath18 . the sum over all node degrees at time @xmath19 , which equals twice the number of links at time @xmath19 , is equal to @xmath20 $ ] . for long times , the sum over the quality values of all the nodes will converge to the mean of the quality distribution times the number of nodes , that is , we can replace @xmath21 by @xmath22\mu$ ] . so at time @xmath19 , the probability that node @xmath1 receives a link equals @xmath23 . throughout the present paper , the steady - state joint distribution of quality and degree is denoted by @xmath24 . the expected number of nodes with degree @xmath5 and quality @xmath6 at time @xmath19 is denoted by @xmath25 . we denote by @xmath26 the expected number of @xmath15-@xmath12 child - parent pairs . we seek the steady - state fraction of nodes who have degree @xmath5 and quality @xmath6 . in appendix [ app : sol_1 ] we derive the following expression for this quantity : @xmath27 note that in the special case of a single permitted value for the quality ( that is , when @xmath28 $ ] ) this model reduces to the shifted - linear preferential attachment model analyzed , for example , in @xcite . the solution in this special case simplifies to @xmath29 this coincides with the degree distribution of shifted - linear kernels given in @xcite and ( * ? ? ? * equation d.9 ) . furthermore , when @xmath30 , all nodes will have zero quality and attachments will be purely degree - proportional , synonymous with the conventional preferential - attachment model proposed initially in @xcite . for the special case of @xmath31 we obtain @xmath32 this is equal to the degree distribution of the conventional ba network ( see , e.g. , @xcite ) . let us also examine the behavior of in the limit of large @xmath5 . in this regime , we can use the asymptotic approximation that for large values of @xmath1 , the function @xmath33 . then we replace @xmath34 with @xmath35 , independent of @xmath6 . therefore , the steady - state joint degree - quality distribution @xmath24 is proportional to @xmath35 . marginalizing out @xmath6 to recover the degree distribution , we obtain the well - known power law , @xmath36 . to quantify how qualities and degrees of adjacent nodes correlate , we need to go beyond the quality - degree distribution obtained in the previous section . the closed - form expression for the nearest - neigbor correlations under the preferential attachment model is derived in @xcite ; that work only considers degrees and does not address qualities . we would like to quantify the conditional distribution @xmath37 , the fraction of neighbours of a given node with degree @xmath5 and quality @xmath6 that have degree @xmath7 and quality @xmath8 . we refer to this as the _ nearest - neighbor quality - degree distribution _ ( nnqdd ) . in appendix [ app : sol_2 ] we study the rate equation describing how the distribution @xmath37 evolves as nodes are added to the network . this gives rise to a system of difference equations which we solve to obtain that , in the steady - state , @xmath38.$ } \label{nnqdd } \end{aligned}\ ] ] in order to obtain the nearest - neighbor quality distribution @xmath39 , one needs to perform the calculations @xmath40 , which requires knowledge of @xmath41 . in turn we have @xmath42 , which according to , yields different sums for different quality distributions @xmath0 . as discussed in section [ sec : intro ] , gfp refers to an average alter superiority in arbitrary aspects ( e.g. , number of citations , exposure to viral online content ) . in this paper , we use the ` quality ' dimension that is incorporated in the model as the subject of the gfp . our objective is to compare the degrees and qualities of nodes with their neighbors . we say that a node experiences the friendship paradox if the degree of that node is less than the average of the degrees of its neighbors . similarly , we say that a node experiences the quality paradox if the quality of the node is less than the average of the qualities of its neighbors . the above - mentioned definitions characterize individual - level paradoxes . our primary interest is to what fraction of nodes experience the friendship and quality paradoxes . to this end , we compare the average degree of the nodes with the average degree of the neighbors of all nodes ( and similarly for quality ) . comparing these two average values yields a macro measure for the system , indicating whether it exhibits paradoxes on average . we call these as the _ network - level friendship paradox _ and _ network - level quality paradox_. our measure of the network - level quality paradox is defined as @xmath43 . the summations are performed over all nodes in the network . note that the numerator of the first sum is actually the sum of the qualities of the neighbors of all nodes . node @xmath44 is repeated @xmath45 times in this sum , once for each of its neighbors . focusing on the limit as @xmath46 , we can use the law of large numbers and express the nqp as follows @xmath47 the greater nqp becomes , the more strongly the paradox holds . negative nqp is indicative of the absence of a quality paradox at the network level . undertaking similar steps to above , we can measure the network - level friendship paradox via @xmath48 note that the numerator is the variance of the degree distribution , so it is positive . the denominator is the average degree and is also positive . so the nfp is always positive , which means that by this definition : _ any network exhibits the friendship paradox at the network level . _ so the task of the present paper with regard to the nfp is to investigate its magnitude , i.e. , to measure how strongly the paradox holds . for example , in the conventional barabasi - albert scale - free model , where the degree variance diverges , the nfp also diverges , which is a result of the presence of macro hubs . to study the nfp and the nqp in concrete settings , we confine ourselves to two quality distributions @xmath0 for illustrative purposes . we consider a finite support for @xmath6 , so that @xmath49 . for each distribution , we are going to consider four different values @xmath50 , and four different values of @xmath51 . the first distribution we consider is the bernoulli case , where nodes can either have quality zero or quality @xmath52 . the probability of quality zero is @xmath53 and the probability of quality @xmath52 is @xmath54 , where @xmath55 . the second distribution we consider is the discrete exponential distribution with decay factor @xmath56 . the probability that the quality is @xmath6 is proportional to @xmath57 . note that in the case of @xmath58 , one recovers a uniform distribution as a special case . we consider both @xmath59 and @xmath60 , yielding decreasing and increasing distributions in @xmath6 , respectively . these distributions are depicted in figure [ distributions ] . .48 0.48 the results for the bernoulli quality distribution are depicted in figure [ bernoulli ] . as depicted in figure [ nqp_fixed_theta ] , for a fixed @xmath52 , the nqp decreases as @xmath50 ( the initial degree of nodes ) increases . also , it is observable that the sensitivity of the nqp to the variations of the quality distribution diminishes for larger values of @xmath50 . as illustrated in figure [ nfp_fixed_theta ] , the nfp increases as @xmath50 ( the initial degree of nodes ) increases . hence , according to _ the variance of the degree distribution grows faster than the mean degree , as @xmath50 increases . _ on the other hand , for a given @xmath50 , increasing @xmath52 ( which is tantamount to increasing @xmath11 ) , increases the nqp . this means that according to _ as @xmath52 increases , the mean of the qualities of the neighbors increases faster than the mean of the qualities of the nodes_. figure [ nqp_fixed_beta ] pertains to this case . observe that as @xmath52 increases , the nqp becomes more sensitive to the distribution of qualities . finally , figure [ nfp_fixed_beta ] represents the nfp for a fixed @xmath50 and different values of @xmath52 . from figures [ nqp_fixed_theta ] , [ nfp_fixed_theta ] , [ nqp_fixed_beta ] and [ nfp_fixed_beta ] , a general observable pattern is that as @xmath53 increases , the nfp increases ( monotonically for almost all values of @xmath53 ) , whereas the nqp is concave and unimodal ( it increases at first , achieves maximum , and then decreases ) . .5 0.5 + 0.5 0.5 now we focus on the exponential quality distribution with the decay factor denoted by @xmath56 . as depicted in figure [ nqp_fixed_theta_decay ] , for a given @xmath52 , the nqp decreases as @xmath50 increases . also , it is observed that as @xmath50 increases , the sensitivity of the nqp to the quality distribution diminishes . these are both similar to the results of the bernoulli distribution . as can be seen in figure [ nfp_fixed_theta_decay ] , the nfp increases as @xmath50 increases . so similar to the bernoulli case , the variance of the degree distribution grows faster than the mean degree , as @xmath50 increases . from figure [ nqp_fixed_beta_decay ] we observe that for a fixed @xmath50 , increasing @xmath52 increases the nqp . we observe that as @xmath52 increases , nqp becomes more sensitive to the changes in the decay factor . finally , figure [ nfp_fixed_beta_decay ] represents the nfp for a fixed @xmath50 and different values of @xmath52 . we observe that increasing @xmath52 increases the nfp for positive decays . also , for very small decay factors ( which generate right - skewed distributions that are highly unequal ) , changing @xmath52 has scant effect on the nfp . this is reasonable because when the decay factor is small , all large values of @xmath6 have small chances of occurrence . consequently , changing @xmath52 minimally changes the shape of the distribution for small decay factors . a trend is discernible from figures [ nqp_fixed_theta_decay ] , [ nfp_fixed_theta_decay ] , [ nqp_fixed_beta_decay ] and [ nfp_fixed_beta_decay ] : as @xmath56 increases , the nfp decreases ( monotonically for all values of @xmath56 ) , whereas nqp is concave and increases up to a point around @xmath61 , and then decreases . since @xmath58 yields a uniform distribution , we can qualitatively conclude that the probability of the network - level quality paradox is higher when qualities are heterogeneous , as compared to when qualities are similar . finally , to verify our results , we run monte carlo simulations to synthesize networks that grow under the prescribed quality - based preferential attachment mechanism , and then calculate the desired quantities by averaging over nodes in the synthesized network . due to computational limitations , we restrict this validation to the case where @xmath62 and @xmath63 for the bernoulli quality distribution and the case where @xmath62 and @xmath64 for the exponential quality distribution . these results are shown in figures [ nqp_fixed_theta ] , [ nfp_fixed_theta ] , [ nqp_fixed_theta_decay ] and [ nfp_fixed_theta_decay ] . the markers show the results of simulations , averaging over 100 monte carlo trials , and the solid curves correspond to our theoretical expressions . we have tested the results on various other quality distributions and observed similar results ; these additional simulations not reported here due to space limitations . in general , we observe that for a fixed @xmath52 , increasing @xmath50 increases the nfp and decreases the nqp regardless of the quality distribution . also , for a fixed @xmath50 , increasing @xmath52 increases the nqp and decreases the nfp . .5 0.5 + 0.5 0.5 note that in all cases the nqp is nonnegative . this has roots in the correlation between degree and quality of single nodes ( intra - node correlation , rather than inter - node correlation ) . let us denote the correlation between degree and quality for a node by @xmath65 , which is the pearson correlation coefficient obtained from the joint distribution @xmath24 . from , we have : @xmath66 this implies that the sign of nqp is the same as the sign of @xmath65 ( since @xmath67 and @xmath68 are nonnegative ) . the observation that nqp is always nonegative indicates that @xmath65 is also always nonegative . we conclude that _ the quality - dependent preferential attachment model generates networks in which degree and quality of a node are always positively correlated . _ this is what we intuitively expect the model to exhibit ; increasing quality increases degree . for example , in citation networks , papers with higher qualities receive more citations . conversely , a paper with many citations is more likely to have a high quality . in the case of friendship networks , a person that is more sociable ends up with more friends than an anti - social person , and conversely , a popular person is more likely to be friendly than an isolated person . we also observe that in all cases , @xmath11 ( equivalently , @xmath52 ) and @xmath50 have opposite effects on both the nfp and the nqp . that is , the effect of increasing @xmath50 is akin to that of decreasing @xmath11 , and vice versa . we observed similar trends for other quality distributions ; these results are omitted here due to space limitations . what causes this disparity is the following : as can be seen in and , @xmath11 only appears in the distributions in the form of @xmath69 . thus increasing @xmath11 and decreasing @xmath50 have the same effect on this variable , and consequently , on the distribution . the aim of the present paper was to put in crisp theoretical focus the seemingly prevalent phenomena of the friendship paradox and the generalized friendship paradox . we proposed a network growth model that incorporates quality . in this model , the probability that a node receives a link increases with both its degree and quality . we analysed the model theoretically in the steady - state ( large size limit ) , and found two theoretical quantities that characterize the interrelation between quality and degree . the first quantity is @xmath24 , which is the joint degree - quality distribution , and equals the fraction of nodes who have degree @xmath5 and quality @xmath6 . the second quantity characterizes nearest - neighbor correlations , and is the nearest - neighbor quality - degree distribution , denoted by @xmath70 . we then defined two network - level measures for the quality and friendship paradoxes and computed them for two particular examples of quality distributions . we observed that for a fixed @xmath52 , increasing @xmath50 increases the nfp and decreases the nqp regardless of the quality distribution . we also observed that for a fixed @xmath50 , increasing @xmath52 increases the nqp and decreases the nfp . we also observed that @xmath11 and @xmath50 have opposite effects on the nfp and also on the nqp . we also tested these results on various other quality distributions , and they proved robust ; the effects of @xmath50 and @xmath11 on paradoxes are opposite regardless of the quality distribution . there are many interesting extensions of this work to pursue . in addition to the network - level paradox , we can also study the individual - level paradox , which would require the utilization of the nnqdd to compare the degrees and qualities of nodes with those of their neighbors . the individual - level paradox has empirical implications which enable us to assess the quality distribution of real networks . 4 feld , s.l . : why your friends have more friends than you do . american journal of sociology , 96(6 ) , 146477 ( 1991 ) ezar , w. , zuckerman , j.t . : what makes you think you re so popular ? self - evaluation maintenance and the subjective side of the friendship paradox " . social psychology quarterly , 64(3 ) , 207223 ( 2001 ) gracia - herranz , m. , moro , e. , cebrian , m. , christakis n. , fowler , j. : using friends as sensors to detect global - scale contagious outbreaks . plos one , 9 , e92413 ( 2014 ) kryvasheyeu , y. , chen , h. , moro , e. , van hentenryck , p. , cebrian , m. : performance of social network sensors during hurricane sandy . arxiv preprint arxiv:1402.2482 ( 2014 ) eom , y.h . , jo , h.h . , generalized friendship paradox in complex networks : the case of scientific collaboration , nature scientific reports , 4 , ( 2014 ) hodas , n.o . , kooti , f. , lerman , k. : friendship paradox redux : your friends are more interesting than you . in : 7th international aaai conference on weblogs and social media , pp . 225233 . ( 2013 ) barabsi , a.l , albert , r. , emergence of scaling in random networks , science , 286 , 509512 ( 1999 ) bianconi , g. , barabsi , a.l . : competition and multiscaling in evolving networks . europhysics lettres , 54(4 ) , 436442 ( 2001 ) caldarelli , g. , capocci , a. , de los rios , p. , muoz , m.a . : scale - free networks from varying vertex intrinsic fitness , physical review letters , 89(25 ) , 258702(2002 ) servedio , v.t.p . , caldarelli , g. , butta , p. : vertex intrinsic fitness : how to produce arbitrary scale - free networks , physical review e , 70(5 ) , 056126 ( 2004 ) smolyarenko , i.e. , hoppe , k. rodgers , g.j . , network growth model with intrinsic vertex fitness . physical review e , 88(1 ) , 012805 ( 2013 ) dorogovtsev , s.n . , mendes , j.f.f . , samukhin , a.n . : structure of growing networks with preferential linking , physical review letters , 85(21 ) , 4633 - 4436 ( 2000 ) fotouhi , b. , rabbat , m.g . : degree correlation in scale - free graphs , european physical journal b , 86(12 ) ( 2013 ) krapivsky , p.l . , redner , s. : organization of growing random networks . physical review e , 63(6 ) , 066123 ( 2001 ) we seek the fraction of nodes who have degree @xmath5 and have quality @xmath6 . we begin by writing the rate equation which quantifies the temporal evolution of @xmath25 . suppose that a node with quality @xmath6 and degree @xmath71 at time @xmath72 , receives a link from the new incoming node . consequently , its degree will become @xmath5 and @xmath25 increments . conversely , if a node with quality @xmath6 and degree @xmath5 at time @xmath72 , receives a link from the new incoming node , @xmath25 decrements . finally , each new incoming node increments @xmath73 with probability @xmath0 . the rate equation thus reads @xmath74 replacing @xmath25 by @xmath22 p_t(k,\theta)$ ] , this can be expressed in terms of @xmath75 as follows : @xmath76 \big [ p_{t+1}(k,\theta)-p_t(k,\theta ) \big ] + p_{t+1}(k,\theta)= \nonumber \\ & \resizebox{0.93\linewidth}{!}{${\ , \displaystyle \frac { \beta(k-1+{\,\theta } ) [ n(0)+t ] p_t(k-1,\theta ) } { { \,2l(0)+n(0)}+(2\beta+\mu ) t } } - { \ , \displaystyle \frac { \beta(k + { \,\theta } ) [ n(0)+t ] p_t(k , \theta ) } { { \,2l(0)+n(0)}+(2\beta+\mu ) t } } + \rho({\,\theta } ) \delta_{k,\beta } .$ } \label{pkdot_1_0 } \end{aligned}\ ] ] in the limit as @xmath77 , the transients vanish . so , we drop the @xmath19 in the arguments and rewrite as : @xmath78 this can be rearranged and expressed equivalently as follows : @xmath79 multiplying both sides by @xmath80 and rearranging the terms , this can be recast as follows @xmath81 setting @xmath82 , this yields @xmath83 . for all @xmath84 , the second term on the right hand side vanishes , and this equation reduces to a straightforward recursion @xmath85 , whose solution is @xmath86 we begin by writing the rate equation to quantify the evolution of @xmath88 , which is the number of nodes with degree @xmath7 and quality @xmath8 who are connected to a parent node of degree @xmath5 and quality @xmath6 . upon introduction of a new node , regardless of its quality , the following is true : if it attaches to a node of degree @xmath7 and quality @xmath8 who is the child of a parent of degree @xmath5 and quality @xmath6 , then the degree of the receiving node increments and consequently @xmath89 decrements . also , @xmath89 decrements if the new node attaches to the parent node in such a pair of nodes . another way that @xmath88 can increment is if either there is a child - parent pair of @xmath90 or @xmath91 . if the new node attaches to the child node in the former case or to the parent node in the latter case , then @xmath92 increments . finally , with probability @xmath93 , the new node will have quality @xmath8 , and if the new node attaches to an existing node of degree @xmath71 and quality @xmath6 , then @xmath89 increments . the rate equation reads @xmath94 $ } \nonumber \\ & \resizebox{0.93\linewidth}{!}{$+ \beta \left [ { \ , \displaystyle \frac { ( k-1+\theta ) n_t ( k-1,\theta,\ell,\phi ) - ( k+\theta ) n_t ( k,\theta,\ell,\phi ) } { { \,2l(0)+n(0)}+(2\beta+\mu ) t } } \right ] + \rho(\phi ) \delta_{\ell,\beta } { \ , \displaystyle \frac { \beta ( k-1+\theta)n_t(k-1,\theta ) } { { \,2l(0)+n(0)}+(2\beta+\mu ) t}}$ } \label{rate_big_1 } \end{aligned}\ ] ] undertaking the same steps that let us transform into , and denoting the fraction @xmath95 by @xmath96 , this can be re - written in terms of @xmath97 instead of @xmath26 . in the limit as @xmath77 , we can drop the @xmath19 subscript and obtain : @xmath98 let us define the new sequence @xmath99 . using this substitution and applying the properties of the gamma function as well as the delta function , we can rewrite equivalently as @xmath100 using the expression in to rewrite the last term on the right hand side of this equation , we can express it equivalently as follows @xmath101 now define the generating function @xmath102 . multiplying both sides of by @xmath103 , summing over all values of @xmath104 and rearranging the terms , we arrive at @xmath105 ( the lower bound of the sum is @xmath106 because @xmath107 is zero for @xmath108 . ) the inverse transform of the factor @xmath109 in the summand can be taken through the following steps : @xmath110 so we can invert term by term . we get @xmath111 from this , we readily obtain @xmath112 the last step is to abridge this quantity and the desired nnqdd distribution , that is , @xmath87 . remember that the nnqdd is the fraction of @xmath15 nodes among the neighbors of a @xmath12 node . to obtain this fraction , we first need to obtain the total number of neighbors of @xmath12 nodes , then find the number of @xmath15 nodes among these nodes , and divide the latter by the former . the total number of neighbors of @xmath12 nodes is simply @xmath113 . the number of @xmath15 nodes among them equals @xmath114n}$ ] , because the @xmath15 node can both be the parent or the child of the a @xmath12 node to be connected to it . so we have @xmath115 . inserting the results of and into this expression and simplifying the results , we obtain @xmath38.$ } \label{nnqdd } \end{aligned}\ ] ]	the friendship paradox refers to the sociological observation that , while the people s assessment of their own popularity is typically self - aggrandizing , in reality they are less popular than their friends . the generalized friendship paradox is the average alter superiority observed empirically in social settings , scientific collaboration networks , as well as online social media . we posit a quality - based network growth model in which the chance for a node to receive new links depends both on its degree and a quality parameter . nodes are assigned qualities the first time they join the network , and these do not change over time . we analyse the model theoretically , finding expressions for the joint degree - quality distribution and nearest - neighbor distribution . we then demonstrate that this model exhibits both the friendship paradox and the generalized friendship paradox at the network level , regardless of the distribution of qualities . we also show that , in the proposed model , the degree and quality of each node are positively correlated regardless of how node qualities are distributed .
in this paper we establish a relation between the size of the fatou components of a semi - hyperbolic rational map , and the hausdorff dimension of the julia set . before formulating the results , we first discuss some background . a rational map @xmath0 of degree at least @xmath1 is _ semi - hyperbolic _ if it has no parabolic cycles , and all critical points in its julia set @xmath2 are _ non - recurrent_. we say that a point @xmath3 is non - recurrent if @xmath4 , where @xmath5 is the set of accumulation points of the orbit @xmath6 of @xmath3 . in our setting , we require that the julia set @xmath2 is not the whole sphere , it is connected and , in addition , there are infinitely many fatou components . let @xmath7 be the sequence of fatou components , and define @xmath8 . since @xmath2 is connected , it follows that each component @xmath9 is simply connected , and thus @xmath10 is connected . we say that the collection @xmath11 is a _ packing _ @xmath12 and we define the _ curvature distribution function _ associated to @xmath12 ( see below for motivation of this terminology ) by @xmath13 for @xmath14 . also , the _ exponent _ @xmath15 of the packing @xmath12 is defined by @xmath16 where all diameters are in the spherical metric of @xmath17 . in the following , we write @xmath18 if there exists a constant @xmath19 such that @xmath20 . if only one if these inequalities is true , we write @xmath21 or @xmath22 respectively . we denote the hausdorff dimension of a set @xmath23 by @xmath24 ( see section [ section - minkowski ] ) we now state our main result . [ theorem - main ] let @xmath0 be a semi - hyperbolic rational map such that the julia set @xmath2 is connected and the fatou set has infinitely many components . then @xmath25 where @xmath26 is the curvature distribution function of the packing of the fatou components of @xmath27 and @xmath28 . in particular @xmath29 . it is remarkable that the curvature distribution function has polynomial growth . as a consequence , we have the following corollary . [ corollary ] under the assumptions of theorem [ theorem - main ] we have @xmath30 where @xmath26 is the curvature distribution function , and @xmath15 is the exponent of the packing of the fatou components of @xmath27 . this essentially says that one can compute the hausdorff dimension of the julia set just by looking at the diameters of the ( countably many ) fatou components , which lie in the complement of the julia set . the study of the curvature distribution function and the terminology is motivated by the _ apollonian circle packings_. an apollonian circle packing is constructed inductively as follows . let @xmath31 be three mutually tangent circles in the plane with disjoint interiors . then by a theorem of apollonius there exist exactly two circles that are tangent to all three of @xmath31 . we denote by @xmath32 the outer circle that is tangent to @xmath31 ( see figure [ fig : apollonian ] ) . for the inductive step we apply apollonius s theorem to all triples of mutually tangent circles of the previous step . in this way , we obtain a countable collection of circles @xmath11 . we denote by @xmath33 the apollonian circle packing constructed this way . if @xmath34 denotes the radius of @xmath10 , then @xmath35 is the curvature of @xmath10 . the curvatures of the circles in apollonian packings are of great interest in number theory because of the fact that if the four initial circles @xmath36 have integer curvatures , then so do all the rest of the circles in the packing . another interesting fact is that if the curvatures of the circles are relatively prime integers , then there are infinitely many circles in the packing with curvature that is a prime number . for a survey on the topic see @xcite . in order to study the curvatures of an apollonian packing @xmath12 one defines the _ exponent _ @xmath15 of the packing by @xmath37 and the _ curvature distribution function _ associated to @xmath38 by @xmath39 for @xmath14 . we remark here that the radii @xmath34 are measured with the euclidean metric of the plane , in contrast to where we use the spherical metric . let @xmath9 be the open ball enclosed by @xmath10 . the _ residual set _ @xmath40 of a packing @xmath12 is defined by @xmath41 . the set @xmath40 has fractal nature and its hausdorff dimension @xmath42 is related to @xmath43 and @xmath15 by the following result of boyd . [ theorem - boyd ] if @xmath12 is an apollonian circle packing , then @xmath44 recently , kontorovich and oh proved the following stronger version of this theorem : [ theorem - kontor - oh ] if @xmath12 is an apollonian circle packing , then @xmath45 where @xmath46 . in particular , @xmath47 . apollonian.png ( 82,82 ) @xmath32 ( 55,85 ) @xmath48 ( 30,68 ) @xmath49 ( 75,60 ) @xmath50 [ fig : apollonian ] in @xcite , merenkov and sabitova observed that the curvature distribution function @xmath43 can be defined also for other planar fractal sets such as the sierpiski gasket and the sierpiski carpets . more precisely , if @xmath11 is a collection of topological circles in the plane , and @xmath9 is the open topological disk enclosed by @xmath10 , such that @xmath51 contains @xmath10 for @xmath52 , and @xmath9 are disjoint for @xmath52 , one can define the _ residual set _ @xmath40 of the _ packing _ @xmath33 by @xmath53 . a fundamental result of whyburn implies that if the disks @xmath54 , @xmath52 are disjoint with @xmath55 as @xmath56 and @xmath40 has empty interior , then @xmath40 is homeomorphic to the standard sierpiski carpet @xcite . in the latter case we say that @xmath40 is a sierpiski carpet ( see figure [ fig : test2 ] for a sierpiski carpet julia set ) . one can define the curvature of a topological circle @xmath10 as @xmath57 . then the _ curvature distribution function _ associated to @xmath12 is defined as in by @xmath58 for @xmath14 . similarly the exponent @xmath15 of @xmath12 is defined as in . in general , the limit @xmath59 does not exist , but if we impose further restrictions on the geometry of the circles @xmath10 , then we can draw conclusions about the limit . to this end , merenkov and sabitova introduced the notion of _ homogeneous planar sets _ ( see section [ section - homogeneous ] for the definition ) . however , even these strong geometric restrictions are not enough to guarantee the existence of the limit . the following theorem hints that a self - similarity condition on @xmath40 would be sufficient for our purposes . [ theorem - merenkov - sabitova ] assume that @xmath27 is a hyperbolic rational map whose julia set @xmath2 is a sierpiski carpet . then @xmath60 where @xmath26 is the curvature distribution function and @xmath15 is the exponent of the packing of the fatou components of @xmath27 . the authors made the conjecture that for such julia sets we actually have an analogue of theorem [ theorem - kontor - oh ] , namely @xmath61 , where @xmath28 . note that theorem [ theorem - main ] partially addresses the issue by asserting that @xmath47 . however , we believe that the limit @xmath62 does not exist in general for julia sets . observe that the conclusion of theorem [ theorem - main ] remains valid if we alter the metric that we are using in the definition of @xmath43 in a bi - lipschitz way . for example , if the julia set @xmath2 is contained in the unit disk of the plane we can use the euclidean metric instead of the spherical . on the other hand , the limit of @xmath63 as @xmath64 is much more sensitive to changes of the metric . the following simple example of the standard sierpiski carpet provides some evidence that the limit will not exist even for packings with very nice " geometry . the standard sierpiski carpet is constructed as follows . we first subdivide the unit square @xmath65 ^ 2 $ ] into @xmath66 squares of equal size and then remove the interior of the middle square . we continue subdividing each of the remaining @xmath67 squares into @xmath66 squares , and proceed inductively . the resulting set @xmath40 is the standard sierpiski carpet and its hausdorff dimension is @xmath68 . the set @xmath40 can be viewed as the residual set of a packing @xmath33 , where @xmath32 is the boundary of the unit square , and @xmath69 are the boundaries of the squares that we remove in each step in the construction of @xmath40 . using the euclidean metric , note that for each @xmath70 the quantity @xmath71 is by definition the number of curves @xmath10 that have diameter at least @xmath72 . thus , @xmath73 ( note that we also count @xmath32 ) . since @xmath74 , we have @xmath75 on the other hand , it is easy to see that @xmath76 , since there are no curves @xmath10 with diameter in the interval @xmath77 . thus , @xmath78 , and this shows that @xmath79 does not exist . in general , if one can show that there exists some constant @xmath80 such that @xmath81 for large @xmath3 , then the limit will not exist . [ fig : test1 ] carpet_julia.jpg ( 70,90 ) @xmath82 ( 44,48 ) @xmath9 [ fig : test2 ] we also note that in theorem [ theorem - main ] one might be able to weaken the assumption that @xmath27 is semi - hyperbolic , but the assumption that @xmath27 has connected julia set is necessary , since there exist rational maps whose fatou components are nested herman rings , and in fact in this case there exist infinitely many fatou components with large " diameter ( see ( * ? ? ? * proposition 7.2 ) ) . thus , if @xmath43 is the number of fatou components whose diameter is at least @xmath83 , we would have @xmath84 for large @xmath3 . the proof of theorem [ theorem - main ] will be given in two main steps . in section [ section - minkowski ] , using the self - similarity of the julia set we will establish relations between the hausdorff dimension of the julia set and its _ minkowski dimension _ ( see section [ section - minkowski ] for the definition ) . then in section [ section - homogeneous ] we will observe that the julia sets of semi - hyperbolic maps are homogeneous sets , satisfying certain geometric conditions ( see section [ section - homogeneous ] for the definition ) . these conditions allow one to relate the quantity @xmath63 with the _ minkowski content _ of the julia set . using these relations , and the results of section [ section - minkowski ] , the proof of theorem [ theorem - main ] will be completed . before proceeding to the above steps , we need some important distortion estimates for semi - hyperbolic rational maps that we establish in section [ section - conformal elevator ] , and we will refer to them as the _ conformal elevator_. these are they key estimates that we will use in establishing geometric properties of the julia set . similar estimates have been established for sub - hyperbolic rational maps in ( * ? ? ? * lemma 4.1 ) . the author would like to thank mario bonk for many useful comments and suggestions , and for his patient guidance . the heart of this section is lemma [ lemma - elevator ] and the whole section is devoted to proving it . let @xmath0 be a semi - hyperbolic map with @xmath85 ; in particular , by sullivan s classification and the fact that semi - hyperbolic rational maps do not have siegel disks or herman rings ( ( * ? ? ? * corollary ) ) , @xmath27 must have an attracting or superattracting periodic point . conjugating @xmath27 by a rotation of the sphere @xmath17 , we may assume that @xmath86 is a periodic point in the fatou set . furthermore , conjugating again with a euclidean similarity , we can achieve that @xmath87 , where @xmath88 denotes the unit disk in the plane . note that these operations do not affect the conclusion of theorem [ theorem - main ] , since a rotation is an isometry in the spherical metric that we used in the definition of @xmath43 , and a scaling only changes the limits by a factor . furthermore , since the boundaries @xmath10 of the fatou components @xmath9 have been moved away from @xmath86 , the diameters of @xmath10 in spherical metric are comparable to the diameters in the euclidean metric . this easily implies that the conclusion of theorem [ theorem - main ] is not affected if we define @xmath89 using instead the euclidean metric for measuring the diameters . in this section the euclidean metric will be used in all of our considerations . by semi - hyperbolicity ( see ( * ? ? ? * theorem ii(b ) ) ) and compactness of @xmath2 , there exists @xmath90 such that for every @xmath91 and for every connected component @xmath92 of @xmath93 the degree of @xmath94 is bounded by some fixed constant @xmath95 that does not depend on @xmath96 . furthermore , we can choose an even smaller @xmath97 so that the open @xmath97-neighborhood of @xmath2 that we denote by @xmath98 is contained in @xmath88 , and avoids the poles of @xmath27 that must lie in the fatou set . then @xmath27 is uniformly continuous in @xmath99 in the euclidean metric , and in particular , there exists @xmath100 such that for any @xmath101 with @xmath102 we have @xmath103 . let @xmath104 , @xmath105 be arbitrary , and define @xmath106 . since for large @xmath107 we have @xmath108 ( e.g. see ( * ? ? ? * corollary 14.2 ) ) , there exists a largest @xmath70 such that @xmath109 . by the choice of @xmath110 , we have @xmath111 . using the uniform continuity and the choice of @xmath112 , it follows that @xmath113 , thus @xmath114 we now state the main lemma . [ lemma - elevator ] there exist constants @xmath115 independent of @xmath116 ( and thus of @xmath110 ) such that : 1 . if @xmath117 is a connected set , then @xmath118 2 . @xmath119 3 . for all @xmath120 we have @xmath121 , with good distortion estimates . for _ hyperbolic _ rational maps ( i.e. , no parabolic cycles and no critical points on the julia set ) the map @xmath122 would actually be bi - lipschitz and part ( c ) of the above lemma would be true with @xmath123 instead of @xmath124 . however , in the semi - hyperbolic case , the presence of critical points on the julia set prevents such good estimates , but part ( a ) of the lemma restores some of them . in order to prepare for the proof we need some distortion lemmas . using koebe s distortion theorem ( e.g. see ( * ? ? ? * theorem 1.3 ) ) one can derive the following lemma . [ lemma - koebe ] let @xmath125 be a univalent map and let @xmath126 . then there exists a constant @xmath127 that depends only on @xmath128 , such that @xmath129 for all @xmath130 . we will be using the notation @xmath131 . we also need the next lemma . [ lemma - simply connected components ] let @xmath132 be a semi - hyperbolic rational map with @xmath133 and assume that @xmath134 is connected . then there exists @xmath135 such that for all @xmath136 , each component of @xmath137 is simply connected , for all @xmath138 . as before , by conjugating , we may assume that @xmath86 is a periodic point in the fatou set , and the julia set is far " from the poles of @xmath139 . by semi - hyperbolicity ( see ( * ? ? ? * theorem ii(c ) ) ) , for each @xmath136 and @xmath140 , there exists @xmath135 such that each component of @xmath137 has euclidean diameter less than @xmath141 , for all @xmath138 . by compactness of @xmath134 , we may take @xmath135 to be uniform in @xmath3 . we choose a sufficiently small @xmath141 such that the @xmath142-neighborhood @xmath143 of @xmath134 does not contain any poles of @xmath139 . we claim that each component of @xmath137 is simply connected . if this was not the case , there would exist an open component @xmath92 of @xmath144 , and a family of compact components @xmath145 of @xmath146 . thus @xmath147 for @xmath148 . assume that @xmath149 is the smallest such integer . note that @xmath92 intersects the julia set @xmath134 , because @xmath150 does so . hence , we have @xmath151 . since @xmath152 and @xmath92 share at least one common boundary point , it follows that @xmath153 , and in particular @xmath154 does not contain any poles of @xmath139 , i.e. , @xmath155 for all @xmath148 . by the choice of @xmath156 the set @xmath157 is a simply connected set in the @xmath141-neighborhood of @xmath134 . note that @xmath158 can not be entirely contained in @xmath159 , otherwise @xmath92 would not be a component of @xmath160 . thus , there exists some @xmath161 and a point @xmath162 . we connect the point @xmath163 to @xmath86 with a path @xmath164 , and then we lift @xmath165 under @xmath139 to a path @xmath166 that connects a preimage @xmath167 of @xmath163 to a pole of @xmath139 ( see ( * ? ? ? * lemma a.16 ) for path - lifting under branched covers ) . the path @xmath168 can not intersect @xmath92 , so it stays entirely in @xmath169 . this contradicts the fact that @xmath169 contains no poles . now we are ready to start the proof of lemma [ lemma - elevator ] . since @xmath170 , for @xmath171 we have @xmath172 , and for the component @xmath173 of @xmath93 that contains @xmath174 we have that the degree of @xmath175 is bounded by @xmath176 . lemma [ lemma - simply connected components ] implies that we can refine our choice of @xmath97 such that @xmath173 is also simply connected . let @xmath177 be the riemann map that maps the center @xmath178 of @xmath174 to @xmath179 , and @xmath180 be the translation of @xmath3 to @xmath179 , followed by a scaling by @xmath181 , so we obtain the following diagram : @xmath182{d}{\psi } & b(x,\varepsilon_0 ) \arrow{d}{\phi } \\ { \mathbb d}\arrow{r } { } & { \mathbb d}\end{tikzcd}\ ] ] diagram.jpg ( 40,70 ) @xmath122 ( 13,35 ) @xmath183 ( 80,35 ) @xmath184 ( 30,45 ) @xmath173 ( 20,60 ) @xmath174 ( 91,45 ) @xmath185 ( 72,59 ) @xmath186 ( 10,70 ) @xmath178 ( 75,70 ) @xmath3 ( 18.5,5 ) @xmath187 ( 69,5 ) @xmath188 [ fig : diagram ] the proof will be done in several steps . first we prove that @xmath187 is contained in a ball of fixed radius smaller than @xmath189 . second , we show a distortion estimate for @xmath183 , namely it is roughly a scaling by @xmath190 . in the end , we complete the proofs of ( a),(b),(c ) , using lemmas that are generally true for proper maps . we claim that there exists @xmath191 , independent of @xmath174 such that @xmath192 this will be derived from the following modulus distortion lemma . if @xmath193 is a jordan region , and @xmath194 is a connected subset of @xmath193 with @xmath195 , we denote by @xmath196 the modulus of the curve family that separates @xmath194 from @xmath197 . [ lemma - modulus ] let @xmath198 be jordan regions , and @xmath199 be a proper holomorphic map of degree @xmath200 . 1 . if @xmath201 is a jordan region with @xmath202 , and @xmath194 is a component of @xmath203 , then @xmath204 2 . if @xmath194 is a jordan region with @xmath205 , and @xmath206 , then @xmath207 a particular case of this lemma is ( * ? ? ? * lemma 5.5 ) , but we include a proof of the general statement since we were not able to find it in the literature . we first show @xmath208 . using a conformal map , we map the annulus @xmath209 to the circular annulus @xmath210 , and by composing with @xmath139 , we assume that we have a proper holomorphic map @xmath211 , of degree at most @xmath200 . we divide the annulus @xmath210 into nested circular annuli centered at the origin @xmath212 such that each @xmath213 does not contain any critical value of @xmath139 in its interior . note that @xmath214 where we denote by @xmath215 the modulus of curves that separate the complementary components of the annulus @xmath213 . we fix @xmath135 . by making the annuli @xmath213 a bit thinner , we can achieve that @xmath216 does not contain any critical value of @xmath139 , and @xmath217 let @xmath218 be a preimage of @xmath213 , so that @xmath219 are nested annuli separating @xmath194 from @xmath197 , and avoiding the critical points of @xmath139 . note that @xmath220 is a covering map of degree @xmath221 , thus @xmath222 . this implies that @xmath223 to see the first inequality , note that an admissible function @xmath128 for @xmath196 yields admissible functions @xmath224 for @xmath225 . combining and we obtain @xmath226 letting @xmath227 concludes the proof . the inequality in @xmath228 follows from poletski s inequality ( * ? ? ? * chapter ii , section 8) . since holomorphic maps are @xmath189-quasiregular , we have @xmath229 for all path families @xmath230 in @xmath193 . first we shrink the regions @xmath193 and @xmath231 as follows . consider a jordan curve @xmath232 very close to @xmath233 such that @xmath232 encloses a region @xmath234 that contains @xmath201 and all critical values of @xmath139 . then @xmath235 is a jordan region that contains @xmath194 and all critical points of @xmath139 . let @xmath236 be the family of paths in @xmath237 that connect @xmath238 to @xmath239 and avoid preimages of critical values of @xmath139 , which are finitely many . also , note that @xmath240 , where @xmath241 is the family of paths in @xmath242 that connect @xmath243 to @xmath244 , and avoid the critical values of @xmath139 . to see this , observe that any such path @xmath245 has a lift @xmath246 that starts at @xmath238 and ends at @xmath247 . using monotonicity of modulus and we have @xmath248 . if @xmath249 is the family of all paths in @xmath237 that connect @xmath238 to @xmath239 , then @xmath249 differs from @xmath230 by a family of zero modulus . the same is true for the corresponding family @xmath250 in @xmath251 . thus , we have @xmath252 . by reciprocality of the modulus and monotonicity , it follows that @xmath253 finally , observe that the path family separating @xmath194 from @xmath197 can be written as an increasing union of families separating @xmath194 from sets of the form @xmath254 , where @xmath255 gets closer and closer to @xmath193 . writing @xmath196 as a limit of moduli of such families , one obtains the desired inequality . we now return to the proof of . applying lemma [ lemma - modulus]@xmath208 to @xmath256 , and using the fact that @xmath257 along with monotonicity of modulus we obtain @xmath258 since modulus is invariant under conformal maps , we have @xmath259 if @xmath260 is such that @xmath261 then by grtzsch s module theorem ( see @xcite ) we have @xmath262 , where @xmath263 is a strictly decreasing bijection . thus , by @xmath264 is uniformly bounded below , and by monotonicity there exists @xmath265 such that @xmath266 . hence , @xmath267 , which proves . now , the version of koebe s theorem in lemma [ lemma - koebe ] yields latexmath:[\ ] ] the minkowski dimension is another useful notion of dimension for a fractal sets @xmath369 . for @xmath135 we define @xmath370 to be the maximal number of disjoint open balls of radii @xmath135 centered at points @xmath371 . we then define the upper and lower minkowski dimensions respectively as @xmath372 if the two numbers agree , then we say that their common value @xmath373 is the minkowski , or else , box dimension of @xmath363 . it is easy to see that the definition of the minkowski dimension is not affected if @xmath370 denotes instead the smallest number of open balls of radii @xmath135 centered at @xmath363 , that cover @xmath363 . the important difference between the hausdorff and minkowski dimensions is that in the hausdorff dimension we are taking into account coverings @xmath366 with different weights @xmath374 attached to each set , but in the minkowski dimension we are considering only coverings of sets with equal diameters . it easily follows from the definitions that we always have @xmath375 from now on , @xmath370 will denote the maximal number of disjoint open balls of radii @xmath376 , centered at points @xmath371 . based on the distortion estimates that we developed in section [ section - conformal elevator ] , and using results of @xcite and @xcite we have the following result that concerns the hausdorff and minkowski dimensions of julia sets of semi - hyperbolic maps . [ theorem - hausdorff measure positive ] let @xmath0 be a semi - hyperbolic rational map and @xmath377 . we have 1 . @xmath378 2 . @xmath379 3 . there exists a constant @xmath19 such that for all @xmath135 @xmath380 where @xmath370 is the maximal number of disjoint open balls of radii @xmath376 ( in the spherical metric ) , centered in @xmath2 . by considerations as in the beginning of section [ section - conformal elevator ] , we may assume that @xmath381 , and use the euclidean metric which is comparable to the spherical metric . this will only affect the constant in part ( c ) of the theorem . the parts ( a ) and ( b ) follow from ( * ? ? ? * theorem 1.11(e ) and ( g ) ) . also , if @xmath382 are disjoint balls of radius @xmath135 centered at @xmath2 then the collection @xmath383 covers @xmath2 , where @xmath384 has the same center as @xmath385 but twice the radius . thus , we have @xmath386 taking limits , and using ( a ) , we obtain @xmath387 which shows the left inequality in ( c ) . for the right inequality in ( c ) , we use the following result of falconer . [ theorem - falconer ] let @xmath388 be a compact metric space with @xmath389 . suppose that there exist @xmath390 such that for any ball @xmath391 of radius @xmath392 there is a mapping @xmath393 satisfying @xmath394 for all @xmath395 . then @xmath396 . it remains to show that this theorem applies in our case . to show the existence of @xmath183 we will carefully use the distortion estimates of lemma [ lemma - elevator ] . let @xmath343 be so small that for @xmath392 and @xmath104 the conclusions of lemma [ lemma - elevator ] are true for the ball @xmath116 . in particular , there exists @xmath340 , independent of @xmath174 , such that @xmath397 for some @xmath70 . for each ball @xmath398 , there exists @xmath138 such that @xmath399 is surjective ( e.g. see ( * ? ? ? * corollary 14.2 ) ) . compactness of @xmath2 allows us to choose a uniform @xmath138 , independent of @xmath400 . by the analyticity of @xmath401 , there exists a constant @xmath402 such that for all @xmath403 we have @xmath404(c ) , there exists @xmath405 independent of @xmath174 such that latexmath:[\[\begin{aligned } \label{theorem - falc- f^n } for @xmath120 . now , we can construct the desired @xmath407 . let @xmath408 be any right inverse of the surjective map @xmath409 . also , the inclusion allows us to define a right inverse @xmath410 of @xmath122 , restricted on a suitable subset of @xmath174 . now , let @xmath411 , and observe that by , and we have @xmath412 thus , the hypotheses of theorem [ theorem - falconer ] are satisfied with @xmath413 . let @xmath414 be a packing , as defined in the introduction , where @xmath10 are topological circles , surrounding topological open disks @xmath9 ( in the plane or the sphere ) such that @xmath415 contains @xmath10 for @xmath52 , and @xmath416 are disjoint . then the set @xmath417 is the residual set @xmath40 of the packing @xmath12 . in the following , one can use the euclidean or spherical metric , but it is convenient to consider @xmath418 as the boundary of the unbounded component of the packing @xmath12 ( see figures [ fig : apollonian ] and [ fig : test2 ] ) , and use the euclidean metric to study the other disks @xmath419 . thus , we will restrict ourselves to the use of the euclidean metric in this section . 1 . each @xmath419 is a _ uniform quasi - ball_. more precisely , there exists a constant @xmath423 such that for each @xmath9 there exist inscribed and circumscribed , cocentric circles of radii @xmath34 and @xmath424 respectively with @xmath425 2 . there exists a constant @xmath426 such that for each @xmath427 and @xmath428 there exists a circle @xmath10 intersecting @xmath429 such that @xmath430 3 . the circles @xmath10 are _ uniformly relatively separated_. this means that there exists @xmath431 such that @xmath432 for all @xmath433 . the disks @xmath419 are _ uniformly fat_. by definition , this means that there exists @xmath434 such that for every ball @xmath429 centered at @xmath9 that does not contain @xmath9 , we have @xmath435 where @xmath436 denotes the @xmath1-dimensional lebesgue measure . ( here one can use the spherical measure for packings on the sphere . ) condition @xmath437 means that the sets @xmath9 look like round balls , while @xmath438 says that the circles @xmath10 exist in all scales and all locations in @xmath40 . condition @xmath421 forbids two large " circles @xmath10 to be close to each other in some uniform manner . note that this only makes sense when @xmath439 are disjoint , e.g. in the case of a sierpiski carpet . finally , @xmath422 is used to replace @xmath421 when we are working with fractals such as the sierpiski gasket , or generic julia sets regarded as packings , where @xmath54 are not disjoint . we now summarize some interesting properties of homogeneous sets , that are not needed though for the proof of theorem [ theorem - main ] . a set @xmath440 is said to be _ porous _ if there exists a constant @xmath441 such that for all sufficiently small @xmath442 and all @xmath443 , there exists a point @xmath444 such that @xmath445 a jordan curve @xmath446 is called a @xmath351-_quasicircle _ if for all @xmath447 there exists a subarc @xmath448 of @xmath165 joining @xmath3 and @xmath449 with @xmath450 . the _ ( ahlfors regular ) conformal dimension _ of a metric space @xmath360 , denoted by @xmath451 , is the infimum of the hausdorff dimensions among all ahlfors regular metric spaces that are quasisymmetrically equivalent to @xmath360 . for more background see chapters @xmath452 and @xmath453 in @xcite . by @xmath437 , each @xmath9 contains a ball of diameter comparable to @xmath457 . thus , summing the areas of the sets @xmath9 , and noting that they are all contained in @xmath458 , we see that for each @xmath135 , there can only be finitely many sets @xmath9 with @xmath459 . we conclude that @xmath40 is locally connected ( see ( * ? ? ? * lemma 19.5 ) ) . condition @xmath438 implies that for @xmath460 , every ball @xmath429 centered at @xmath40 intersects a curve @xmath10 of diameter comparable to @xmath461 . let @xmath462 and consider the ball @xmath463 . then @xmath464 intersects a curve @xmath10 of diameter comparable to @xmath465 , and if @xmath466 is sufficiently small but uniform , then @xmath467 . thus @xmath429 contains a curve @xmath10 of diameter comparable to @xmath461 . by @xmath437 , @xmath9 contains a ball of radius comparable to @xmath457 and thus comparable to @xmath461 ( note that here we use the euclidean metric ) . hence , @xmath468 contains a ball of radius comparable to @xmath461 . this completes the proof that @xmath40 is porous . for our last assertion we will use a criterion of mackay ( * theorem 1.1 ) which asserts that a doubling metric space which is _ annularly linearly connected _ has conformal dimension strictly greater than @xmath189 . a connected metric space @xmath363 is annularly linearly connected ( abbr . alc ) if there exists some @xmath469 such that for every @xmath470 , and @xmath471 in the annulus @xmath472 there exists an arc @xmath473 joining @xmath3 to @xmath449 that lies in a slightly larger annulus @xmath474 . it suffices to show that @xmath40 is alc . the idea is simple , but the proof technical , so we only provide a sketch . let @xmath475 , and consider a path @xmath476 ( not necessarily in @xmath477 ) that joins @xmath3 and @xmath449 . the idea is to replace the parts of the path @xmath165 that lie in the complementary components @xmath9 of @xmath40 by arcs in @xmath455 and then make sure that the resulting arc stays in a slightly larger annulus @xmath474 . the assumption that the curves @xmath10 are quasicircles guarantees that the subarcs that we will use are not too large " , and condition @xmath421 guarantees that the large " curves @xmath10 do not block the way from @xmath478 to @xmath449 , since these curves are not allowed to be very close to each other . using @xmath421 , we can find uniform constants @xmath479 such that there exists at most one curve @xmath480 with @xmath481 that intersects @xmath482 . we call a curve @xmath10 _ large _ if its diameter exceeds @xmath483 , and otherwise we call it _ small_. we enlarge slightly the annulus ( maybe using a larger @xmath484 ) to an annulus @xmath485 so that @xmath486 contains all small curves @xmath10 that intersect @xmath165 . we now check all different cases . if @xmath165 meets the large @xmath480 that intersects @xmath487 , using the fact that @xmath480 is a quasicircle , we can enlarge the annulus to an annulus @xmath488 with a uniform @xmath489 , so that @xmath3 can be connected to @xmath449 by a path in @xmath490 . we call the resulting path @xmath165 . note that here we have to assume that @xmath491 , so that the path @xmath165 does not lie in the unbounded component of the packing and it passes through several curves @xmath10 on the way from @xmath3 to @xmath449 . the case @xmath492 , which occurs only when @xmath493 , is similar and in the previous argument we just have to choose a path @xmath165 that lies in @xmath494 . we still assume that @xmath495 contains all small curves @xmath10 that intersect @xmath165 . if @xmath165 meets a small @xmath10 that does not intersect @xmath496 , then we can replace the subarcs of @xmath165 that lie in @xmath9 with arcs in @xmath10 that have the same endpoints . the resulting arcs will lie in the annulus by construction . next , if @xmath165 meets a small @xmath10 that does intersect @xmath496 , we follow the same procedure as before , but now we have to choose the sub - arcs of @xmath10 carefully , so that they do not approach @xmath178 too much . this can be done using the assumption that the curves @xmath10 are uniform quasicircles . the resulting arcs will lie in a slightly larger annulus @xmath497 , where @xmath498 a is uniform constant . finally , if @xmath165 intersects a large @xmath10 which does not meet @xmath499 we can use the assumption that @xmath10 is a quasicircle to replace the subarcs of @xmath165 that lie in @xmath9 with subarcs of @xmath10 that have diameter comparable to @xmath461 . thus , a larger annulus @xmath500 will contain the arcs of @xmath10 that we obtain in this way . we need to ensure that this procedure indeed yields a path that joins @xmath3 and @xmath449 inside @xmath501 . this follows from the fact that @xmath502 . the latter fact follows from the assumption that the curves @xmath10 are uniform quasicircles , which in turn implies that each @xmath9 contains a ball of radius comparable to @xmath457 , i.e. , @xmath437 is true ( for a proof of this assertion see ( * ? ? ? * proposition 4.3 ) ) . as we will see in lemma [ lemma - n - n ] , a homogeneous set has the special property that there is some important relation between the curvature distribution function @xmath43 and the maximal number of disjoint open balls @xmath370 , centered at @xmath40 . thus , considerations about the residual set @xmath40 , which are reflected by @xmath370 , can be turned into considerations about the complementary components @xmath9 , which are comprised in @xmath43 . [ lemma - mereknov - sabitova ] assume that @xmath40 is the residual set of a packing @xmath33 that satisfies @xmath437 and @xmath421 ( or @xmath437 and @xmath422 ) . for any @xmath503 , there exist constants @xmath504 depending only on @xmath505 and the constants in @xmath506 ( or @xmath507 ) such that for any collection @xmath508 of disjoint open balls of radii @xmath442 centered in @xmath40 we have the following statements : 1 . [ lemma - m - s-1]there are at most @xmath509 balls in @xmath508 that intersect any given @xmath10 with + @xmath510 2 . [ lemma - m - s-2 ] there are at most @xmath511 curves @xmath10 intersecting any given ball in @xmath512 and satisfying + @xmath513 using this lemma one can prove a relation between the curvature distribution function @xmath514 ( using the euclidean metric ) and the maximal number @xmath370 of disjoint open balls of radius @xmath376 , centered at @xmath40 . namely , we have the following lemma . [ lemma - n - n ] assume that the residual set @xmath40 of a packing @xmath12 satisfies @xmath420 and @xmath421 or @xmath420 and @xmath422 . then there exists a constant @xmath19 such that for all small @xmath135 we have @xmath515 where @xmath505 is the constant in @xmath438 . let @xmath508 be a maximal collection of disjoint open balls of radius @xmath376 , centered at @xmath40 . for each ball @xmath516 , by condition @xmath438 there exists @xmath10 such that @xmath517 and @xmath518 . on the other hand , lemma [ lemma - mereknov - sabitova](a ) implies that for each such @xmath10 there exist at most @xmath509 balls in @xmath508 that intersect it . thus @xmath519 conversely , note that by the maximality of @xmath508 , it follows that @xmath512 covers @xmath40 . hence , if @xmath10 is arbitrary satisfying @xmath520 , it intersects a ball @xmath521 in @xmath512 . for each such ball @xmath521 , lemma [ lemma - mereknov - sabitova](b ) implies that there exist at most @xmath511 curves @xmath10 with @xmath520 that intersect it . thus @xmath522 by considerations as in the beginning of section [ section - conformal elevator ] , we assume that @xmath523 , and we will use the euclidean metric since this does not affect the conclusion of the theorem . let @xmath32 be the boundary of the unbounded fatou component , @xmath419 be the sequence of bounded fatou components , and @xmath455 . then @xmath524 can be viewed as a packing , and @xmath525 is its residual set . note , though , that the sets @xmath10 need not be topological circles in general . this , however , does not affect our considerations , since it does not affect the conclusions of lemmas [ lemma - mereknov - sabitova ] and [ lemma - n - n ] , as long as the other assumptions hold for @xmath10 and the simply connected regions @xmath9 enclosed by them . we will freely use the terminology curves " for the sets @xmath10 . by theorem [ theorem - hausdorff measure positive ] we have that the quantity @xmath526 is bounded away from @xmath179 and @xmath86 as @xmath527 , where @xmath28 . if we prove that @xmath2 is a homogeneous set , satisfying @xmath420 and @xmath422 , then using lemma [ lemma - n - n ] , it will follow that @xmath63 is bounded away from @xmath179 and @xmath86 as @xmath64 , and in particular @xmath528 which will complete the proof . julia sets of semi - hyperbolic rational maps are locally connected if they are connected ( see ( * ? ? ? * proposition 10 ) ) , and thus for each @xmath135 there exist finitely many fatou components with diameter greater than @xmath376 ( see ( * ? ? ? * problem 19-f ) ) . first we show that condition @xmath437 in the definition of homogeneity is satisfied . the idea is that the finitely many large fatou components are trivially quasi - balls , as required in @xmath437 , so there is nothing to prove here , but the small fatou components can be blown up with good control to the large ones using lemma [ lemma - elevator ] . the distortion estimates allow us to control the size of inscribed circles of the small fatou components . let @xmath529 , where @xmath530 are the constants appearing in lemma [ lemma - elevator ] . we also make @xmath531 even smaller so that for @xmath532 and @xmath104 the conclusions of lemma [ lemma - elevator ] are true . since there are finitely many curves @xmath10 with @xmath533 , for these @xmath10 there exist cocentric inscribed and circumscribed circles with radii @xmath34 and @xmath424 respectively , such that @xmath534 , for some @xmath535 . this implies that @xmath536 . if @xmath10 is arbitrary with @xmath537 , then for @xmath538 and @xmath539 , by lemma [ lemma - elevator](c ) there exists @xmath70 such that @xmath540 note that the fatou component @xmath9 is mapped under @xmath122 onto a fatou component @xmath541 . since @xmath122 is proper , the boundary @xmath10 of @xmath9 is mapped onto @xmath542 . then the above inequality can be written as @xmath543 hence , @xmath544 is one of the large " curves , for which there exists a inscribed ball @xmath545 such that @xmath546 . observe that @xmath547 . let @xmath548 be a preimage of @xmath549 under @xmath122 , and @xmath550 be the component of @xmath551 that contains @xmath552 . for each @xmath553 , by lemma [ lemma - elevator](c ) one has @xmath554 thus @xmath555 letting @xmath556 , and @xmath557 , one obtains @xmath558 , so @xmath437 is satisfied with @xmath559 . similarly , we show that condition @xmath438 is also true . let @xmath340 be the constant in lemma [ lemma - elevator](b ) and consider @xmath560 so small that the conclusions of lemma [ lemma - elevator ] are true for @xmath104 and @xmath532 . note that by compactness of @xmath2 there exists @xmath503 such that for @xmath561 and @xmath104 there exists @xmath10 such that @xmath562 and @xmath563 indeed , one can cover @xmath2 with finitely many balls @xmath564 of radius @xmath565 centered at @xmath2 , such that each ball @xmath566 contains a curve @xmath567 . this is possible because every ball @xmath566 centered in the julia set must intersect infinitely many fatou components , otherwise @xmath122 would be a normal family in @xmath566 . in particular , by local connectivity most " fatou components are small , and thus one of them , say @xmath568 , will be contained in @xmath566 . now , if @xmath429 is arbitrary with @xmath569 , we have that @xmath570 for some @xmath571 , and thus @xmath572 . since @xmath573 $ ] lies in a compact interval , easily follows , by always using the same finite set of curves @xmath574 that correspond to @xmath575 , respectively . we may also assume that @xmath576 for each of these curves . now , if @xmath577 , @xmath104 , by lemma [ lemma - elevator](b ) we have @xmath578 for some @xmath70 . by the previous , @xmath579 intersects some @xmath580 with @xmath581 , thus @xmath582 . hence , @xmath429 contains a preimage @xmath10 of @xmath544 , and by lemma [ lemma - elevator](a ) , ( c ) we obtain @xmath583 however , @xmath544 was one of the finitely many curves that we chose in the previous paragraph . this and the above inequalities impliy that @xmath584 with uniform constants . this completes the proof of @xmath438 . finally , we will prove that condition @xmath422 of homogeneity is satisfied . this follows easily from the fact that the fatou components of a semi - hyperbolic rational map are _ uniform john domains _ in the spherical metric ( * ? ? ? * proposition 9 ) . since we are only interested in the bounded fatou components , we can use instead the euclidean metric . a domain @xmath585 is a @xmath586-john domain ( @xmath587 ) if there exists a basepoint @xmath588 such that for all @xmath589 there exists an arc @xmath590 connecting @xmath591 to @xmath320 such that for all @xmath592 we have @xmath593 where @xmath594 . in our case , the bounded fatou components @xmath595 are uniform john domains , i.e. , john domains with the same constant @xmath596 . let @xmath429 be a ball centered at some @xmath9 that does not contain @xmath9 . we will show that there exists a uniform constant @xmath434 such that @xmath597 if @xmath598 , then @xmath599 , so is true with @xmath600 . otherwise , @xmath601 intersects @xmath455 at a point @xmath591 . we split in two cases : the basepoint @xmath320 satisfies @xmath602 . then consider a path @xmath603 from @xmath320 to @xmath178 , as in the definition of a john domain , such that @xmath604 for all @xmath592 . in particular let @xmath605 be a point such that @xmath606 , thus @xmath607 . since @xmath608 , we have @xmath609 , hence @xmath610 the basepoint @xmath320 lies in @xmath611 . consider a point @xmath612 close to @xmath613 such that @xmath614 . then , by the definition of a john domain for @xmath615 , we have @xmath616 . hence , @xmath617 , so @xmath618 even when the julia set of a semi - hyperbolic map is a sierpiski carpet , the uniform relative separation of the peripheral circles @xmath10 in condition @xmath421 need not be true . in fact , it is known that for such julia sets condition @xmath421 is true if and only if for all critical points @xmath620 , @xmath621 does not intersect the boundary of any fatou component ; see ( * ? ? ? * proposition 3.7 ) . recall that @xmath621 is the set of accumulation points of the orbit @xmath622 . in ( * ? ? ? * proposition 3.5 ) it is shown that if the boundaries of fatou components of a semi - hyperbolic map @xmath27 are jordan curves , then they are actually uniform quasicircles . if , in addition , they are uniformly relatively separated ( i.e. , condition @xmath421 ) , proposition [ proposition - homogeneous ] implies that @xmath623 . on the other hand , if @xmath27 is a semi - hyperbolic _ polynomial _ with connected julia set , then not all boundaries of fatou components are jordan curves . in fact , @xmath2 coincides with the boundary of a single fatou component @xmath624 which is a john domain , and is called the _ basin of attraction of @xmath86_. according to a recent result of kinneberg ( ( * ? ? ? * theorem 1.1 ) ) , boundaries of planar john domains have conformal dimension equal to @xmath189 , if they are connected . therefore , @xmath625 , in contrast to the previous case . . by theorem [ theorem - main ] there exists a constant @xmath19 such that @xmath626 for all @xmath14 . taking logarithms , one obtains @xmath627 letting @xmath628 yields @xmath629 which completes part of the proof . recall that the exponent @xmath15 of the packing of the fatou components of @xmath27 is defined by @xmath630 and it remains to show that @xmath631 . note that for @xmath632 the sum @xmath633 diverges . also , since for semi - hyperbolic rational maps there are only finitely many large " fatou components , if @xmath634 , then @xmath635 for all @xmath636 . if @xmath637 , using , one has @xmath638 this implies that @xmath639 . conversely , assume that @xmath640 . since there are only finitely many large " fatou components , we only need to take into account the sets @xmath10 with @xmath641 in the sum @xmath642 . using again we have @xmath643 hence @xmath644 , which completes the proof .	recently mereknov and sabitova introduced the notion of a homogeneous planar set . using this notion they proved a result for sierpiski carpet julia sets of hyperbolic rational maps that relates the diameters of the peripheral circles to the hausdorff dimension of the julia set . we extend this theorem to julia sets ( not necessarily sierpiski carpets ) of semi - hyperbolic rational maps , and prove a stronger version of the theorem that was conjectured by merenkov and sabitova .
the study of the phase behaviour of simple and complex fluids by computer simulation is a subject of much current research activity @xcite . of particular interest are the critical point properties of such systems @xcite . in the vicinity of a critical point , the correlation length grows extremely large and may exceed the linear size of the simulated system . when this occurs , the singularities and discontinuities that characterise critical phenomena in the thermodynamic limit are shifted and smeared out @xcite . unless care is exercised , such finite - size effects can lead to serious errors in computer simulation estimates of critical point parameters . to cope with these problems , finite - size scaling ( fss ) techniques have been developed @xcite . fss methods enable one to extract accurate estimates of infinite - volume thermodynamic quantities from simulations of finite - sized systems . to date , their application to fluid criticality has been principally in conjunction with simulations in the constant-@xmath0vt or grand - canonical ensemble ( gce ) . the principal merit of this ensemble is that the particle density fluctuates on the scale of the system as a whole , thus enabling direct measurement of the large - scale density fluctuations that are the essential feature of fluid criticality . the gce has proven its worth in fss studies of criticality in a variety of fluid systems including the lennard - jones ( lj ) fluid @xcite and a 2d spin fluid model @xcite . notwithstanding its wide utility , however , there exist many complex fluids for which use of the gce ensemble is by no means efficient . systems such as semi - dilute polymer solutions are difficult to simulate in the gce due to excluded volume effects which hinder chain insertions . while smart insertion techniques go some way to ameliorating this difficulty , the long chain lengths of greatest interest are currently inaccessible @xcite . similarly , electrolyte models such as the restricted primitive model show very poor acceptance rates for particle insertions due to the formation of bound ion clusters @xcite . thus it is interesting to ask whether one can deal with the near - critical density fluctuations in such systems _ without _ having to implement inefficient particle transfer operations . the approach we consider here , is to employ an ensemble wherein the total particle number is fixed , but the density is allowed to fluctuate by virtue of _ volume _ transitions . specifically , we consider how the fss ideas , hitherto only applied to systems with constant volume , may be generalised to an isothermal - isobaric ( npt - ensemble ) simulation . since finite - size scaling usually rests on the idea of comparing the correlation length with the ( fixed ) linear dimensions of the systems , the generalisation to systems whose linear dimensions are dynamically fluctuating is not completely obvious . we make a scaling _ ansatz _ for the near - critical scaling operator distributions and scaling fields , expressed in terms of powers of the particle number . this is then tested via a simulation study of the critical lennard - jones fluid , in which it found that the fss predictions are indeed consistent with the simulation results . finally we discuss the relative merits of the npt- and @xmath0vt- ( gce ) ensembles for simulation studies of fluid criticality . we consider a classical single component fluid , whose configurational energy ( which we write in units of @xmath1 ) resides in a sum of pairwise interactions amongst the @xmath2 particles it contains @xmath3 where @xmath4 is the interparticle potential which , for this work , we assign the lennard - jones form : @xmath5\label{eq : ljdef}\ ] ] where @xmath6 is the dimensionless well depth and @xmath7 serves to set the length scale . within the isothermal - isobaric ensemble , the partition function is given by @xmath8 } \label{eq : z}\ ] ] where @xmath9 is the pressure , and @xmath10 is the homogeneously fluctuating system volume . the associated ( helmholtz ) free energy is @xmath11 in the vicinity of the liquid - vapour critical point , the coarse - grained properties of the system are expected to exhibit scaling behaviour . for simple fluids with short - ranged interactions , this behaviour is ising - like @xcite and is controlled by two relevant scaling fields @xmath12 and @xmath13 that measure deviations from criticality . in general ( in the absence of the special ising ` particle - hole ' symmetry ) , these scaling fields comprise linear combinations @xcite of the reduced chemical potential @xmath0 and well depth @xmath6 : @xmath14 where subscripts denote critical values , and @xmath15 and @xmath16 are non - universal field mixing parameters , the values of which depend on the specific system under consideration . the respective conjugate operators are defined by @xcite @xmath17 whereupon , utilizing equation [ eq : z ] , and noting that @xmath18 ( with @xmath19 , the entropy density ) , one finds @xmath20 , \hspace{1 cm } { \cal m } = \frac{1}{1-sr}[\rho - su]\ ] ] where @xmath21 is the particle density and @xmath22 is the energy density . we term @xmath23 the ordering operator , while @xmath24 is termed the energy - like operator . for the finite - size scaling behaviour of the joint distribution @xmath25 we make the following _ ansatz _ : @xmath26 where @xmath27 and @xmath28 are non - universal metric factors and @xmath29 is a universal function in the sense that it is identical for all members of a given universality class _ and _ a given choice of boundary conditions . here we have chosen simply the particle number @xmath2 rather than the volume @xmath30 as a measure of the ` finite - size ' of the systems , using now suitable powers of @xmath2 in the scaling assumption . the unspecified exponents @xmath31 and @xmath32 in equation [ eq : ansatz ] can be related to the standard critical indices @xmath33 and @xmath34 by an argument analogous to that invoked by binder @xcite for the ising model in the canonical ensemble . consider the generalised isothermal compressibility , @xmath35 , defined by @xmath36 this may be reexpressed as a fluctuation relation by appeal to equations [ eq : z ] and [ eq : g ] , whence one finds @xmath37 from equation [ eq : ansatz ] , the scaling of the fluctuation in the ordering operator @xmath23 at finite @xmath12 is given by @xmath38 where @xmath39 is an unspecified scaling function . now in the critical ( @xmath40 ) and thermodynamic limits ( @xmath41 ) , one requires for consistency that @xmath42 . accordingly @xmath43 , and matching powers of @xmath2 in equation [ eq : kappa ] and [ eq : scalem ] then yields @xmath44 a relation that is satisfied by the assignments @xmath45 where we have employed the hyperscaling relation @xmath46 to eliminate @xmath47 from equation [ eq : reln ] . thus precisely at criticality , where @xmath48 , equation [ eq : ansatz ] can be written : @xmath49 where @xmath50 is a universal function characterising the ising fixed point . we note that this scaling behaviour is simply related to that for the @xmath0vt - ensemble @xcite by the substitution @xmath51 . in what follows we will test the validity of equation [ eq : ansatz1 ] via simulations of the 3d lennard - jones fluid at its liquid - vapour critical point . monte - carlo simulations of the lennard - jones fluid were carried out within the npt - ensemble using a metropolis algorithm @xcite . in common with many other simulation studies of the model , the potential was cutoff at a distance @xmath52 , and was not shifted . in the course of the simulations , three separate periodic systems comprising @xmath53 and @xmath54 particles were studied . in order to minimise fluctuations in the acceptance rate for volume moves , the method of eppenga and frenkel @xcite was employed . within this scheme the system performs a random walk in @xmath55 rather than in @xmath30 itself , thereby automatically adjusting the step size in the volume to be larger at lower densities than higher densities . this technique is particularly useful in the coexistence region region , due to the large volume fluctuations . from a previous detailed simulation study of the lj fluid in the grand - canonical ensemble @xcite , the critical temperature is known to lie at @xmath56 . using this value of @xmath6 , a number of short runs were performed in which the pressure @xmath57 was tuned in order to locate approximately the coexistence curve . once a near - critical value of @xmath57 had been obtained , a long run comprising @xmath58 volume change attempts and an equal number of displacements attempts per particle were performed . during this run , the histogram of the joint distribution @xmath59 was accumulated . in order to locate more precisely the critical pressure , the ordering operator distribution @xmath60 was studied for each value of @xmath2 . precisely at criticality , and for the correct choice of the field mixing parameter @xmath15 , @xmath60 is expected to be a symmetric function by virtue of the fact that the field mixing transformation restores the ising symmetry of the critical fluid @xcite . the value of the field mixing parameter @xmath15 and the pressure @xmath57 were therefore tuned simultaneously using the histogram reweighting scheme @xcite , until a symmetric function was obtained . this occurred for values @xmath61 and @xmath62 . this value of @xmath15 is in accord with that obtained previously in the gce study of the same model @xcite . the resulting form of @xmath60 for each system size , expressed in terms of the scaling variable @xmath63 , is plotted in figure [ fig : pom ] . the non - universal scale factor @xmath64 was chosen so that the data for the @xmath65 system has unit variance , while the exponent ratio @xmath66 was assigned the ising estimate @xmath67 of reference @xcite . also shown for comparison ( solid line ) is the scaled critical magnetisation distribution of the 3d ising model obtained in a separate study @xcite . clearly the data for each system size collapse well on to one another and are thus consistent with the proposed scaling form . they also agree well with the universal form of the ising magnetisation distribution . we attribute the small deviations to corrections - to - scaling associated with the rather small numbers of particles . these discrepancies are comparable in size and form with those observed in reference @xcite . unfortunately , owing to these corrections - to - scaling , as well as to the somewhat poor statistical quality ( a point to which we return in section [ sec : discuss ] ) , it was not possible to obtain a reliable independent estimate of the exponent ratio @xmath66 . a procedure similar to that outlined above was also carried out for the energy - like operator @xmath68 . in this case , however , there is no symmetry requirement for the critical form of the function @xcite . however , we found that by assigning the field mixing parameter @xmath16 , the value @xmath69 found in the previous gce study of the lj fluid @xcite , the function matched closely the form of the critical ising model energy distribution . again the scaling of the @xmath68 with @xmath2 was found to be consistent with the scaling form eq . [ eq : ansatz1 ] . the aim of this work was to develop a fss formalism for simulations of near - critical fluids at constant pressure . a scaling form was proposed and tested for the lennard - jones fluid , whose critical point parameters are known to high precision . good consistency with the scaling prediction was found . additionally , it was shown that the forms of the critical scaling operator distributions are the _ same _ as those for the @xmath0vt - ensemble ( and hence the canonical ensemble of the critical ising model ) , despite the very different nature of the simulation ensembles . as is well known , finite - size effects _ may _ differ in various ensembles of statistical mechanics : the microcanonical ensemble , canonical and grand canonical ensembles of fluids are known to have distinct finite - size properties , and are characterized by different scaling functions . our interpretation of the equivalence of the constant - npt and constant-@xmath70 ensemble scaling functions , is that in the thermodynamic limit , @xmath71 , and therefore it is immaterial whether one uses @xmath72 or @xmath73 as the pair of given intrinsic thermodynamic variables . all that matters is that both ensembles have a single extensive variable ( @xmath2 or @xmath30 ) . different scaling properties are obtained when two extensive variables are used , such as in the constant - nvt ( canonical ) ensemble of fluids . it follows that the gce - based fss techniques developed in reference @xcite for accurately locating fluid critical point parameters , carry over directly to the npt - ensemble . it is hoped that use of this latter ensemble will prove beneficial in situations where the gce is highly inefficient , such as for long - chain polymer or electrolyte models . we add the further comment that finite - size scaling with @xmath2 rather than @xmath30 is well known in mean - field spin systems , when every spin interacts with every other spin , and geometry and space have no meaning @xcite . while use of the npt - ensemble is likely to be much more efficient than the gce in cases where particle insertions have a low acceptance probability , we believe that for simple fluids , use of the grand canonical ensemble is much to be preferred . the present study of the lj fluid consumed circa @xmath74 hours cpu time on a cray - ymp , using a vectorized program . by contrast , the previous gce study of the same model was performed with considerably less computational effort using workstations . moreover , it was possible to study much larger systems ( comprising up to @xmath75 particles ) , with considerably greater statistical accuracy than attainable here . the reasons for this difference in efficiency seem to be manifold . owing to the fluctuating volume , it is not possible to easily implement a cell structure for efficiently locating neighbouring particles within the cutoff range . a calculation of all particle separations prior to a trial volume transition involves an @xmath76 operation , which can only be avoided if no potential cutoff is employed @xcite . by contrast , use of a cell structure in the gce leads to a @xmath77 growth in computational complexity . a second disadvantage of the npt - ensemble , is that both volume changes and particle moves must be performed . for each particle move , two partial energy calculations are required , while for a volume change ( with finite interaction cutoff ) , a total energy calculation is required . by contrast , only particle insertions need be implemented in a gce scheme @xcite , each of which requires only a single partial energy calculation . additionally it seems likely that for a given @xmath78 the random walk in @xmath30 required to pass from one phase to the other at coexistence is longer than that in @xmath2 for the gce , even if one employs a dynamically adjusting volume step size such as that of reference @xcite . consequently the correlation time in the npt - ensemble is considerably greater than that of the gce . finally we note that another recent study has also attempted to apply finite - size scaling methods to the lj fluid in the npt - ensemble @xcite . in this study the authors focused on the distribution of the specific volume @xmath79 and attempted to locate the critical point by requiring scale invariance in @xmath80 , subject to the requirement that @xmath80 has two peaks of _ equal _ height at coexistence . however , as we have shown in this work , the volume distribution is not the scaling counterpart for fluids of the ising magnetisation , rather it is the operator @xmath23 . moreover , the limiting ( large @xmath2 ) critical point form of @xmath80 does not have two peaks of equal heights , and is in fact highly asymmetic in form , as figure [ fig : pv ] shows . we therefore ascribe the rather low accuracy in the estimates for @xmath81 and the difficulties in obtaining the surface tension exponent in reference @xcite to this incorrect choice of the scaling operator .	we consider the application of finite - size scaling methods to isothermal - isobaric ( constant - npt ) simulations of pure continuum fluids . a finite - size scaling _ ansatz _ is made for the dependence of the relevant scaling operators on the particle number . to test the proposed scaling form , constant pressure simulations of the lennard - jones fluid at its liquid - vapour critical point are performed . the critical scaling operator distributions are obtained and their scaling with particle number found to be consistent with the proposed behaviour . the forms of these scaling distributions are shown to be identical to their ising model counterparts . the relative merits of employing the constant - npt and grand canonical ( constant-@xmath0vt ) ensembles for simulations of fluid criticality are also discussed .
spacetime singularities are one of the most problematic features of general relativity . physics breaks down there and unpredictability appears to be unavoidable . among all the predictions of general relativity , black holes ( bhs ) are usually considered one of the most fascinating objects which populate our universe , and are frequently used to test different attempts to unify general relativity with quantum mechanics . after the singularity theorems by hawking and penrose @xcite ( an excellent overview of these theorems and subsequent extensions can be found in @xcite ) , bhs are known to have a singularity inside them . these theorems can be circumvented and regular bhs , that is , solutions of einstein equations that have horizons but are regular everywhere , can be constructed . in particular , charged regular bh solutions exist in the framework of einstein nonlinear electrodynamics ( nled ) theory . the interest in these theories is twofold . first , quantum corrections to maxwell theory can be described by means of non linear effective lagrangians that define nleds as , for example , the euler heisenberg lagrangian @xcite , which is effectively described by born infeld ( bi ) theory @xcite . even more , higher order corrections give place to a sequence of effective lagrangians which are polynomials in the field invariants @xcite . and second , in case of dealing with open bosonic strings , the resulting tree level effective lagrangian is shown to coincide with the bi lagrangian @xcite . these nled theories , when coupled to gravity , give place to very interesting phenomena as , for instance , the appereance of generalized reissner nordstrm geometries in the form of bi like solutions @xcite . interestingly , exact regular bh geometries in the presence of nled were obtained in @xcite . in particular , the ayn beato and garca solution @xcite , further discussed in @xcite , extended the preliminary attempt of bardeen @xcite to obtain regular bh geometries . moreover , bhs with the euler heisenberg effective lagrangian as a source term were examined in @xcite , and a similar type of solutions with lagrangian densities that are powers of maxwell s lagrangian were analyzed in @xcite . the plausibility of these solutions is usually checked with the help of energy conditions . in fact , if a bh is regular , the strong enery condition is violated somewhere inside the horizon @xcite but the weak or dominant energy conditions could be satisfied everywhere @xcite . moreover , as pointed out in @xcite , regular bhs that satisfy the weak energy condition ( wec ) and their energy momentum tensor is such that @xmath0 have a de sitter behaviour at @xmath1 . regular bh solutions possessing this symmetry , some of them satisfying the wec and with an asymptotically reissner nrdstrom behaviour , have been constructed in the framework of einstein nleds @xcite . in a recent work @xcite , several black hole metrics corresponding to nonlinearly charged black holes which were shown to be consistent with a logarithmic correction to the bekenstein hawking entropy formula were constructed . the main drawback of this work was that the wec was shown to be perturbatively violated at order @xmath2 . therefore , as stated in @xcite , we think that it would be interesting to investigate whether or not is possible to obtain effective regular bh geometries with reproduce the logarithmic correction without violating this energy condition . in this work we tackle this problem and construct a new and very simple static and spherically symmetric regular bh solution , obtained within einstein nled theory . our result will be based on a useful formula relating the electric field , which will be imposed to be coulomb like , with the curvature invariants @xmath3 and @xmath4 . this bh will be shown to be reissner nrdstrom like at infinity . as stated before , the wec will be shown to be satisfied everywhere . moreover , after a careful analysis of the horizons , the entropy and heat capacity will reveal that our model realizes some quantum gravity predictions which add a logarithmic correction to the bh entropy and which give place to a remnant . finally , some conclusions are established regarding a possible realization of a quadratic generalized uncertainty principle by nled . in geometrized units , einstein s equations ( @xmath5 ) read @xmath6 where @xmath7 is the energy momentum tensor . let us form the following curvature invariants : @xmath8 as pointed out in @xcite in the four dimensional case , the non weyl part of the curvature determined by the matter content can be separated by showing that @xmath9 where @xmath10 is the trace of the energy momentum tensor and @xmath11 is the dimension of the spacetime . for simplicity let us take spherically symmetric and static solutions given by ( @xmath12 ) @xmath13 for the matter content we choose certain nled . assuming that the corresponding lagrangian only depends on one of the two field invariants , a particular choice for an energy momentum tensor for nled is @xmath14\ ] ] where @xmath15 is the corresponding lagrangian , @xmath16 and @xmath17 . on one hand , in the electrovacuum case , and considering only a radial electric field as the source , that is , @xmath18 maxwell equations read @xmath19 thus , @xmath20 on the other hand , the components of the einstein tensor and the curvature invariants are given by @xmath21}{r^2 } \nonumber \\ g_{rr}&= & \frac{-1+f(r)+r f'(r)}{r^2 f(r)}\nonumber \\ g_{\theta \theta}&= & \frac{1}{2 } r \left[2 f'(r)+r f''(r)\right ] \nonumber \\ g_{\phi\phi}&=&\sin^{2}\theta\ , g_{\theta \theta}.\end{aligned}\ ] ] and @xmath22}{r^4}+f''(r)^2 \nonumber \\ & & r^{\mu\nu}r_{\mu\nu}= \frac{4 + 4 f(r)^2 + 8 r^2 f'(r)^2 + 8 f(r ) \left[-1+r f'(r)\right]}{2 r^4 } \nonumber \\ & & + \frac{r^4 f''(r)^2 + 4 r f'(r ) \left[-2+r^2 f''(r)\right]}{2 r^4 } \nonumber \\ & & r = g^{\mu\nu}r_{\mu\nu}=-\frac{2 \left[-1+f(r)+2 r f'(r)\right]}{r^2}-f''(r),\end{aligned}\ ] ] respectively . + + therefore , taking into account that @xmath23 and using eqs . ( [ max ] ) , ( [ compe ] ) and ( [ inv ] ) , we arrive to the following expression for the electric field : @xmath24 \nonumber \\ & = & \frac{r^{2}}{4q}\sqrt{4 r^{\mu\nu}r_{\mu\nu } -r^{2}}.\end{aligned}\ ] ] + + it is important to point out that eqs . ( [ identity ] ) and ( [ electric ] ) are also valid for @xmath25 , as can be easily shown by direct calculation . moreover , eq . ( [ electric ] ) follows directly from eqs . ( [ main ] ) , ( [ nlt ] ) and ( [ eqmaxwell ] ) . + + as a simple example of the applicability of eq . ( [ electric ] ) , let us consider the case of a reissner nrdstrom bh whose relevant curvature invariants are given by @xmath26 and @xmath27 . therefore , @xmath28 and @xmath29 , as expected . as a bypass of eqs . ( [ max ] ) and ( [ electric ] ) , the following ordinary differential equation for @xmath30 can be derived : @xmath31 in the maxwell case ( @xmath32 ) , the solutions are given by @xmath33 . the solution with positive sign coincides with the reissner nrdstrom one if @xmath34 and @xmath35 ( we think that this is a simple way of obtaining this well known solution ) . the negative sign corresponds to the quasicharged einstein rosen bridge @xcite , which is usually assumed to be an _ ad hoc _ modification of the theory corresponding to a negative electromagnetic energy density . this is a clear example of how the einstein equations do not say anything about energy conditions , which have to be imposed apart from the dynamics . + + in the general case , taking @xmath36 , we get @xmath37 where @xmath38 and @xmath39 are arbitrary constants . it can be seen that only the ( first ) positive sign choice gives place to the correct @xmath40 case . within this choice , eq . ( [ long ] ) simplifies to @xmath41 therefore , if a nled model depending on @xmath42 is given , then eq . ( [ long ] ) can be used to calculate the corresponding geometry , which solves einstein equations . we note that the opposite way is usually considered when regular bh solutions are looked for . that is , one starts from certain metric and tries to look for some nled model such that the coupled einstein non linear theory is satisfied . this underlying nled theory can be obtained using the @xmath43 framework @xcite , which is somehow dual to the @xmath44 framework . after introducing the tensor @xmath45 together with its invariant @xmath46 , one considers the hamiltonian like quantity @xmath47 as a function of @xmath43 , which specifies the theory . therefore , the lagrangian can be written as a function of @xmath43 as @xmath48 finally , by reformulating the coupled einstein nled equations in terms of @xmath43 , @xmath49 is shown to be given by @xcite @xmath50 where the mass function @xmath51 is such that @xmath52 . let us assume a nled theory such that , when dealing only whith static and spherically symmetric sources , the corresponding electric field is regular everywhere . let us consider a simple ansatz for the electric field , which we express as @xmath53 where @xmath54 and @xmath55 . from this choice it is clear that @xmath56 . as commented before , the corresponding geometry can be obtained with the help of eq . ( [ electric ] ) . + after a long but straightforward calculation we get ( for simplicity only @xmath57 will be considered ) : + * @xmath58 . in this case the first two curvature invariants diverge at @xmath59 . @xmath60 can be made regular everywhere after choosing @xmath61 . * @xmath62 . in this case the three curvature invariants diverge at @xmath59 . * @xmath63 . in this case @xmath60 diverges at @xmath59 . * @xmath64 . in this case the corresponding geometry is given by ( choosing the appropriate sign ) @xmath65 where @xmath66 are arbitrary constants . moreover , the curvature invariants are regular provided @xmath61 . let us analize this case in more detail . after considering the schwarzschild de sitter metric , we get that @xmath67 and @xmath68 . therefore , @xmath69 and we can write @xmath70 which is of the form of eq . ( [ longbis ] ) . moreover , this solution is asymptotically reissner nrdstrom de sitter , @xmath71 and , when @xmath1 , the metric function behaves as de sitter space as @xmath72 the curvature invariants are given by @xmath73 therefore , the spacetime is regular everywhere , as stated before . the corresponding electric field is given by @xmath74 note that , as shown in eq . ( [ field ] ) , the electric field , which is regular everywhere , is given by a very simple expression , in contrast to other works ( see , for example , refs . @xcite ) . moreover , as in the case of a bi electric field , it develops a maximum , but this time not at @xmath59 . this feature could be related with the fact that a bh with bi electric field is not regular ( see , for example , ref . @xcite ) , in contrast with our case . in other words , it appears as if the bi field was shifted to make our solution regular everywhere . in this sense , the spirit of bi nled theory is recovered . up to this point , nothing has been said about the plausibility of the obtained solution , except that it seems to be related to the bi one . we can shed more light on the corresponding matter content by computing the structural function @xmath49 for the nled theory . + + for our @xmath64 solution , the mass function is given by @xmath75 we point out that this mass functions tends to @xmath76 both in the @xmath40 and in the asymptotic spatial limit . from this , the structural function as a function of @xmath77 is given by @xmath78 now , @xmath77 and @xmath43 are related by means of @xmath79 then we obtain @xmath80 it is interesting to perform a series expansion of the structural function around @xmath81 to compare our electrodynamics with other previous results . after introducing the parameter @xmath82 , we have @xmath83 a couple of comments are in order here . first of all , let us note that maxwell s theory , @xmath84 , is recovered for small fields . in addition , a quadratic bi like term appears ( fifth term in the rhs of eq . ( [ ourh ] ) ) . this quadratic term is easy to interpret in light of the cutoff field which is an essential ingredient of bi theory . therefore , the difficulty of interpreting this nled theory can be adscribed to the other terms which appear in the rhs of eq . ( [ ourh ] ) . in spite of this , let us note that similar terms have appeared since the discovery of the first exact regular bh solution by ayn beato and garca @xcite . in fact , the hamiltonian function presented in ref . @xcite can be expanded for weak fields to give @xmath85 which except for some constants coincide with our eq . ( [ ourh ] ) . in spite of some drawback related with the interpretation of some terms which appear in eq . ( [ ourh ] ) , the nled we are considering has a nice feature because the wec is satisfied . in particular , with the help of @xmath51 , this wec can be expressed as @xcite @xmath86 where the prime denotes derivative with respect to @xmath77 . therefore , using eq . ( [ theeq ] ) whith @xmath87 we get @xmath88 + and , therefore , the wec is satisfied everywhere ( as commented in the introduction , given the regularity of the bh solution , the strong enery condition is violated somewhere inside the horizon @xcite . however , we left this calculation to the interested reader because we think it does not contribute to the subsequent discussion ) . note that this agrees with the existence of the de sitter behaviour @xcite for the metric function when @xmath1 , given by eq . ( [ limit ] ) . in order to study the existence of event horizons in this geometry we have to find the values of @xmath77 where @xmath89 - i.e : the set of values of @xmath77 for which @xmath90 . hereafter , only the case @xmath91 will be considered . the @xmath92 function appearing in eq . ( [ theeq ] ) can be written as @xmath93 where we have defined the auxiliary function @xmath94 as @xmath95 clearly the horizons are located at @xmath96 . however , in order to have a physically meaningful model , we are interested in the case of three real roots . this can be achieved by forming the discriminant from the corresponding polynomial equation and then analysing when it is negative @xcite . that is , @xmath97 where @xmath98 and @xmath99 are certain functions of the mass @xmath76 and charge @xmath100 @xcite . when solving for small values of @xmath100 and @xmath76 , the previous condition has a graphic solution given by the shaded region of fig . ( [ fig1 ] ) . + however , we note that the linear tendency is maintained for bigger values both of @xmath76 and @xmath100 . as fig . ( [ fig1 ] ) suggests , there exists a region where the polynomial ( [ p(r ) ] ) has three real roots . from this general case of three real roots we are interested in the particular case of two positive and one negative as it would be desirable to resemble the reissner - nrdstrom case and to avoid degeneracies in the horizons . for this particular case the actual limit for the mass of the bh would be @xmath101 to get a glimpse of what is happening with this new limit it suffices to use the following properties for the roots of a cubic equation @xcite . if @xmath102 and @xmath103 are the roots of an arbitrary cubic equation , then + @xmath104 where in our specific case : @xmath105 , @xmath106 and @xmath107 . if one assumes that @xmath108 , which is the limit with no pathologies in the sign and existence of the three different roots , then it is straightforward to show that , at first order in @xmath100 , @xmath109 which is a good symptom of our theory since this case would describe a schwarzschild bh with a charge perturbing it . finally , if we use again the relations given by eq . ( [ relations ] ) , this time relaxing the condition to @xmath110 but still having @xmath111 , then the error made to the other roots @xmath112 and @xmath113 is about @xmath114 which is a very small number . from this we conclude that the limit @xmath115 has a solid footing . to summarize , in this discussion we have shown that for any combination of @xmath116 , @xmath94 has two real positive roots which are very close to the horizons of a perturbed schwarzschild bh , i.e : @xmath117 and @xmath118 . the perturbation is performed by the charge @xmath100 and it has the additional and very important feature of regularising the geometry . therefore , following the previous reasoning , in the next section we are going to work in the regime @xmath108 , which guarantees two horizons and no pathologies since it fulfills the condition given by eq . ( [ secondr ] ) . we also note that this regime avoids the case of having a naked singularity that would defy the penrose cosmic censorship conjecture , which is believed to be true @xcite . within this limit , the two horizons of the regular bh , @xmath119 and @xmath120 are given by @xmath121 + \mathcal{o } \left(\frac{q}{2m}\right)^3\nonumber \\ r_-&=&\frac{\|q^3\|}{12\sqrt3 m^2 } + \mathcal{o}\left(\frac{q}{2m}\right)^3\end{aligned}\ ] ] ( remember that the third root of @xmath30 is given by @xmath122 but is not of physical interest as it can not be interpreted as a horizon ) . given the expressions for the horizons as expressed by eqs . ( [ horizons ] ) , in this section a discussion of the thermodynamics of this regular bh will be presented . as in the the reissner - nordstrm case , the surface gravity is given by @xmath123 therefore , the temperature is readily obtained using the equation @xmath124 . we obtain @xmath125 and , therefore , using the second principle in the form , @xmath126 , the entropy of the bh takes the form @xmath127 where @xmath128 is the schwarzschild bh area and @xmath129 is a constant . in addition , an interesting result obtained from this particular solution is the emergence of a remnant mass , which can be obtained by calculating the heat capacity , defined as @xmath130 . after some calculations we get @xmath131 + therefore , as the bh evaporation stops when @xmath132 , a remnant mass can be obtained . from eq . ( [ hc ] ) , this mass is shown to be given by @xmath133 there is an intriguing similitude between the entropy of the regular bh , given by eq . ( [ entropy ] ) , and that of a schwarzschild bh with quantum corrections . specifically , different approaches to quantum gravity predict corrections to the bekenstein hawking entropy in the form @xcite @xmath134 where the @xmath135 coefficients are parameters which depend on the specific model we are considering . among all the models employed to extract conclusion from the merging between gravity and the quantum world , a minimal length scale is thought to be a essential ingredient @xcite . this minimum length can be implemented , for example , as a generalized uncertainty principle ( gup ) , which has been very recently derived from first principles @xcite . this gup can be implemented using deformed commutation relations which include terms linear ( or quadratic ) in the particle momenta as @xmath136 where @xmath137=i \hbar \delta_{ij}$ ] and @xmath138 ( @xmath139 ) and @xmath140 is a dimensionless constant . these gup effects have been used to compute the bh entropy @xcite , which is given by @xmath141 therefore , given the similarity between eqs . ( [ entropy ] ) and ( [ gup ] ) , both the charge and the gup parameter can be related by @xmath142 where only a quadratic gup is considered . and , therefore , @xmath143 . ] moreover , the existence of a remnant in the charged case here considered is in complete agreement with similar conclusions obtained within the quadratic gup formalism @xcite . thus , a quadratic gup can be clearly realized by the nled model here employed which , in this sense , acquires more importance . in this work , we have managed to derive a regular black hole solution which solves the einstein equations when certain non linear electrodynamics model is invoked . this model here considered satisfies the weak energy condition . moreover , although the matter content is shown to be similar , for small fields , to the ayn beato and garca solution , the electric field here considered has a very simple mathematical expression , in contrast to other cases . we have computed the thermodynamic properties of this regular black hole by a careful analysis of the horizons and we have found that the usual bekenstein hawking entropy gets corrected by a logarithmic term . therefore , in this sense our model realizes some quantum gravity predictions which add a logarithmic correction to the black hole entropy . in particular , we have established some similitudes between our model and a quadratic generalized uncertainty principle by showing that the charge plays the role of the @xmath140parameter which enters into the later . this similitude has been confirmed by the existence of a remnant , which prevents complete evaporation , in complete agreement with the quadratic generalized uncertainty principle case . we note that , although it has been shown that the @xmath64 case represents a regular black hole , other regular solutions with different @xmath99 , together with their energy conditions , remain to be studied . + + the authors acknowledge support from the faculty of science and vicerrectora de investigaciones of universidad de los andes , bogot , colombia , under fapa project enredamiento , gravitacin y sistemas cunticos " . r. penrose , phys . * 14 * , 57 ( 1965 ) . s. w. hawking and g. f. r. ellis , the large scale structure of space time , cambridge university press , cambridge ( 1979 ) . j. m. senovilla , arxiv:0605007 . w. heisenberg and h. euler , z. phys . * 120 * , 714 ( 1936 ) . j. schwinger , phys . rev . * 82 * , 664 ( 1951 ) . m. born and l. infeld , proc . london a * 144 * , 425 ( 1934 ) ; ibid . * 143 * , 410 ( 1934 ) and * 147 * , 522 ( 1934 ) . z. bialynicka birula , i. bialynicki birula , phys . d * 2 * , 2341 ( 1970 ) . e. s. fradkin and a. a. tseytlin , phys . lett b. * 163 * , 123 ( 1985 ) . a. a. tseytlin , nucl . b * 501 * , 41 ( 1997 ) . a. garca d. , h. salazar i. and j. f. plebaski , nuovo cimento ser . b * 84 * , 65 ( 1984 ) . n. breton , phys . rev . d * 67 * , 124004 ( 2003 ) . s. hendi , ann . physics * 333 * , 282 ( 2013 ) . e. ayon beato and a. garcia , phys . * 80 * , 5056 ( 1998 ) . beato , a. garca , phys . b * 464 * , 25 ( 1999 ) . beato , a. garca , gen . . gravit . * 31 * , 629 ( 1999 ) . beato , a. garca , phys . b * 493 * , 149 ( 2000 ) . beato , a. garca , gen . relativ . gravit . * 37 * , 635 ( 2005 ) . i. dymnikova , class . quantum gravity , * 21 * , 4417 ( 2004 ) . f. baldovin , m. novello , s.e . perez bergliaffa and j. salim , classical quantum gravity * 17 * , 3265 ( 2000 ) . k. a. bronnikov , phys . d * 63 * , 044005 ( 2001 ) . j.m . bardeen , proceedings of gr5 , tbilisi , ussr , p. 174 h. yajima and t. tamaki , phys . rev . d * 63 * , 064007 ( 2001 ) . m. hassaine and c. martinez , classical quantum gravity * 25 * , 195023 ( 2008 ) . o. b. zaslavskii , phys . b , * 688 * , 278 ( 2010 ) . s. ansoldi , p. nicolini , a. smailagic and e. spallucci , phys . b * 645 * , 261 ( 2007 ) . l. balart and e. c. vagenas , phys . b * 730 * , 14 ( 2014 ) . l. balart and e. c. vagenas , phys . d * 90 * , 124045 ( 2014 ) . e. contreras , f. d. villalba and p. bargueo , epl ( accepted , 2016 ) . s. cherubini . d. bini , s. capozziello and r. ruffini , int . journal mod . d * 11 * , 827 ( 2002 ) . a. einstein and n. rosen , phys * 48 * , 73 ( 1935 ) . h. salazar i. , a. garca d. and j. plebaski , j. math . * 28 * , 2171 ( 1987 ) . k. a. bronnikov , phys . d * 63 * , 044005 ( 2001 ) . j. g. graves , d. brill , physical review . * 120 * , 4 ( 1960 ) . bronshtein , k.a semendyayev , g.musiol , h.muehlig , handbook of mathematics , springer , leipzig ( 2005 ) . r. k. kaul and p. majumdar , phys . lett * 84 * , 5255 ( 2000 ) . a. j. medved and e. c. vagenas , phys . rev . * d70 * , 124021 ( 2004 ) . g. a. camelia , m. arzano and a. procaccini , phys . d * 70 * , 107501 ( 2004 ) . k. a. meissner , class . * 21 * , 5245 ( 2004 ) . s. das , p. majumdar and r. k. bhaduri , class . quant . grav . * 19 * , 2355 ( 2002 ) . m. domagala and j. lewandowski , class . * 21 * , 5233 ( 2004 ) . a. chatterjee and p. majumdar , phys . * 92 * , 141301 ( 2004 ) . m. m. akbar and s. das , class . * 21 * , 1383 ( 2004 ) . y. s. myung , phys . b579 * , 205 ( 2004 ) . a. chatterjee and p. majumdar , phys . rev . * d71 * , 024003 ( 2005 ) . r. aros , d.e . daz and a. montecinos , jhep , 012 ( 2010 ) . a. sen , jhep , 156 ( 2013 ) . r. aros , d.e . daz and a. montecinos , phys . d * 88 * , 104024 ( 2013 ) . s. hossenfelder , living rev . relativity , * 16 * , 2 ( 2013 ) . r. n. costa filho , j. p. m. braga , j. h. s. lira and j. s. andrade jr . lett b , * 755 * , 367 ( 2016 ) . r. j. adler , p. chen and d. i. santiago , gen . gravit . * 33 * , 2101 ( 2001 ) . g. amelino camelia , m. arzano , y. ling and g. mandanici , class . * 23 * , 2585 ( 2006 ) . b. majumder , phys . lett * b703 * , 402 ( 2011 ) . b. majumder , gen . gravit . * 45 * , 2403 ( 2013 ) . p. bargueo and e. c. vagenas , phys . b * 742 * , 15 ( 2015 ) . r. banerjee and s. ghosh , phys . b688 * , 224 ( 2010 ) . a. dutta and s. gangopadhyay , gen . . gravit . * 46 * , 1747 ( 2014 ) .	a simple regular black hole solution satisfying the weak energy condition is obtained within einstein non linear electrodynamics theory . we have computed the thermodynamic properties of this black hole by a careful analysis of the horizons and we have found that the usual bekenstein hawking entropy gets corrected by a logarithmic term . therefore , in this sense our model realizes some quantum gravity predictions which add this kind of correction to the black hole entropy . in particular , we have established some similitudes between our model and a quadratic generalized uncertainty principle . this similitude has been confirmed by the existence of a remnant , which prevents complete evaporation , in agreement with the quadratic generalized uncertainty principle case .
the development of the theory of brownian motion played a pivotal role and continues to do so in the development of statistical mechanics and thermodynamics @xcite . thermodynamics together with relativity and quantum theory form three pillars on which much of the entire structure of physics rests . tampering with the axioms in either of those theories is not a good idea ; doing so may well lead to contradictions with the other theories . in particular , the field of thermodynamics bears consequences for many branches of physics . its four laws are well - known @xcite : the zeroth law guarantees that states of thermal equilibrium exist which can be characterized by a temperature @xmath0 . the first law provides a balance among the various contributions that make up the internal energy of a system while the second law introduces the concept of thermodynamic entropy @xmath1 , which notably is extensive and never decreases for a closed physical system . in addition , the second law tells us that there exists an absolute zero of temperature . the third law is attributed to walther hermann nernst ( 1864 - 1941 ) and arose as the result of his seminal idea being guided by his critical analysis of chemical and electrochemical reactions at lower temperatures that at low temperatures there occurs for isothermal processes a perfect correspondence between the enthalpy and the gibbs free energy . thereby , the approximate rule hypothesized earlier by marcelin berthelot and julius thomson , becomes a `` law '' at zero temperature . nernst announced this result already in his lectures in 1905 , terming it `` mein wrmesatz '' ( my law of heat ) @xcite . he took this result even further : he also studied _ how _ fast the difference between the changes in the enthalpy @xmath2 and the gibbs free energy @xmath3 , @xmath4 tends to zero @xcite . in fact , this difference vanishes faster than linear in temperature implying that the change of entropy itself must vanish at absolute zero . this in turn implies identical , generally vanishing initial slopes for the corresponding quantities @xmath2 and @xmath3 as a function of temperature . an elucidating account of the history of the third law and the controversies surrounding its acceptance is presented in the books by dugdale @xcite and by wilks @xcite . in its strict form , as given by max planck @xcite , the third law reads : the entropy @xmath5 per particle approaches at absolute zero a constant value @xmath6 that possibly depends on the chemical composition of the system . this limiting entropy constant @xmath6 can generally be set equal to zero . the third law carries prominent consequences for quantum mechanics and the field of low - temperature physics . first , the fact that at absolute zero temperature the isotherm coincides with the isentrope ( adiabat ) immediately implies that this absolute zero temperature is unattainable by use of a sequence of isothermal and adiabatic reversible operations @xcite . therefore , it has the consequence that the efficiency of a carnot engine , a heat engine that cyclically operates between two heat baths of different temperatures which are never brought into contact with each other , can never reach @xmath7 for any finite upper temperature . moreover , the constant value of the entropy at absolute zero is given by the degeneracy @xmath8 of the @xmath9-particle system in the corresponding quantum ground state , @xmath10 , where @xmath11 is the boltzmann constant . the limiting value of the intensive quantity @xmath12 in the thermodynamic limit of particle number @xmath13 will typically be zero , so long as the degeneracy @xmath14 does not grow with @xmath9 faster than exponentially @xcite . a well - known exception is the case of noninteracting , independent particles carrying a non - vanishing spin @xmath15 , yielding @xmath16 for the limiting entropy per particle . moreover , the third law also implies that thermal quantities such as specific heats , the isobar thermal coefficient of expansion , the isochor coefficient of tension , etc . , all approach zero as @xmath17 . likewise , the ( magnetic ) susceptibility becomes constant as @xmath17 , so that the classic curie law must loose its validity at very small temperatures . are there known exceptions of systems not obeying the third law ? it is known that many classical systems do not obey the third law . in particular , noninteracting classical systems with their constant values for the specific heat clearly violate the third law . a well - known case is the classical ideal gas for which the entropy @xmath1 assumes the form @xmath18 $ ] , where @xmath19 is the volume and @xmath20 denotes the entropy constant . it clearly does not fulfill the third law because it diverges logarithmically with temperature @xmath0 for a constant specific heat @xmath21 . even when we use for @xmath21 the physically relevant low - temperature result , namely that quantum mechanically the specific heat @xmath21 vanishes faster than @xmath22 , we still find a dependence on the particle density which is not compatible with the formulation of planck . this observation that the classical gas does not obey the third law led nernst to speculate that the classical gases must undergo a `` degeneracy '' , which has been resolved with the quantum statistics for the ideal fermi gas and the ideal bose - einstein gas , which indeed do obey the third law in the strict formulation by planck . are there yet other remaining open problems with the third law ? apparent difficulties with the third law occur for metastable states that do not necessarily guarantee the sufficiently fast relaxation within a finite time scale towards thermal equilibrium , the latter being a prerequisite for the validity of the third law . in this context , glasses provide a system class that can provide detectable deviations from the third law at low temperatures which likely are the result of frozen - in ordered excited states that have not yet fully relaxed . according to common wisdom the known deviations from the third law will all be cured by quantum mechanics , quantum statistics , and interactions among particles . in the following we shall investigate the thermodynamic low temperature properties for open quantum systems that are coupled to a heat bath of _ finite _ dissipation strength . because both , the thermodynamics of classical open systems and the quantum statistical mechanics of open systems are strictly valid only for systems that are only infinitely weakly coupled to a bath , it is _ a priori _ not obvious how the third law lives up to the sub - system in presence of finite quantum dissipation @xcite . the effect of the finite coupling of a sub - system to an environment in fact induces several subtleties for quantum brownian motion @xcite . for example , the equilibrium density matrix is no longer given by its standard canonical form @xmath23 where @xmath24 denotes the system hamiltonian and @xmath25 $ ] is the partition function . therefore , in clear contrast to the classical case with dissipation , this reduced density matrix becomes also a function of the interaction strength with the environment . this being so , taking a gas of free independent quantum brownian particles that are coupled to a heat bath with finite dissipation strength , one may speculate that the role of the interactions of the free particle with the abundant bath degrees of freedom will be sufficient to cure the shortcomings stemming from a classical gas of free brownian particles . to start out , we first recall the standard results for a single harmonic quantum oscillator that is infinitely weakly coupled to a bath that establishes the temperature @xmath0 . this situation is reminiscent of the famous treatment of the specific heat of a solid by albert einstein in 1907 @xcite , where he found an exponential suppression of the specific heat as @xmath17 . this finding impressed the berlin school so immensely , in particular nernst and his collaborators ( who in 1910 experimentally confirmed this salient first prediction of quantum theory ) , that nernst together with planck were able to bring the `` new copernicus '' @xcite into the exclusive circle of berlin physicists in 1913 . let us consider an oscillator degree of freedom of mass @xmath26 and force constant @xmath27 , its hamiltonian @xmath24 reads @xmath28 in terms of the angular frequency @xmath29 the quantum mechanical energy eigenvalues read @xmath30 , yielding for the partition function @xmath31 the well - known expression @xmath32}\ ] ] where @xmath33 is the inverse temperature . using familiar relations we find that the internal energy @xmath34 reads @xmath35 and , correspondingly , the entropy is given by @xmath36\\ & = k_\text{b}\left[\frac{\hbar\beta\omega_0}{\exp(\hbar\beta\omega_0)-1}- \ln\big(1-\exp(-\hbar\beta\omega_0)\big)\right]\ , . \label{hose}\end{aligned}\ ] ] for low temperatures , @xmath37 , the entropy approaches zero like @xmath38 the specific heat can now be derived either from ( [ hoe ] ) as @xmath39 or from ( [ hos ] ) as @xmath40 in both cases , one obtains for the specific heat @xmath41 its low - temperature behavior @xmath42 is not analytic in temperature and corresponds to einstein s result for the low - temperature behavior of the specific heat of a solid @xcite . for high temperatures one finds @xmath43\,.\ ] ] as this result shows , the specific heat for a free particle can not simply be obtained by taking the limit @xmath44 of the harmonic oscillator . such a procedure will not properly account for the reduced number of degrees of freedom which within the equipartition theorem will lead to a high - temperature specific heat of only @xmath45 for the free particle . we now couple the harmonic oscillator of the previous section to an environment consisting of an infinite number of harmonic oscillators forming a heat bath . in contrast to the previous section , the coupling strength will not be kept negligible here . in addition , system and bath are infinitely weakly coupled to a superbath which has the purpose to provide the temperature @xmath0 . the total hamiltonian @xmath46 does not need to account for the superbath and therefore consists of three parts @xcite @xmath47 where @xmath24 is given by ( [ ho ] ) , the bath hamiltonian reads @xmath48 and the coupling is bilinear in the coordinates @xmath49 we note that quantum systems that are coupled to an environment of finite strength are rarely exactly solvable . the dissipative quantum oscillator becomes exactly solvable with its bilinear coupling to a bath because of the inherent quadratic structure of the total hamiltonian in ( [ total ] ) . this fact holds true even for the case of time - dependent , parametrically driven dissipative quantum harmonic systems @xcite . in order to describe the influence of the environment on the system oscillator , it is sufficient to know the spectral density of bath oscillators defined by @xcite @xmath50 for later purposes , we introduce the laplace transform of the damping kernel , which generally depends on frequency , thereby causing memory - friction @xcite , , @xmath51 the important special case of strictly ohmic dissipation is characterized by @xmath52 and @xmath53 which leads to a memoryless damping of strength @xmath54 . we next discuss the specific heat of the damped harmonic oscillator by following two routes . first , we start from the energy @xmath34 and employ the common relation in ( [ ce ] ) . as an alternative route we shall in section [ evspfdho ] determine the entropy from the partition function by means of ( [ hos ] ) from which ( [ cs ] ) allows one to evaluate the specific heat in the case of strictly ohmic damping . the energy of the damped harmonic oscillator is given by @xmath55 where the expectation value of an operator @xmath56 acting in the hilbert space of the system is defined with respect to the canonical density matrix of system plus environment as @xmath57 } { \tr\left[\exp(-\beta h)\right]}\,.\ ] ] for ohmic damping , the second moments of position and momentum can be expressed as @xcite @xmath58 and @xmath59 where we have introduced a temperature - dependent function @xmath60 for @xmath61 , when evaluating @xmath62 , the integrand decreases only with @xmath63 and a finite value can only be obtained by introducing a high - frequency cutoff in the damping kernel @xmath64 . however , this divergent term gives rise only to a temperature - independent contribution to @xmath62 and thus to the energy ( [ eho ] ) . when evaluating the specific heat according to ( [ ce ] ) , this constant term will disappear and a finite result is obtained even for ohmic damping . after some algebra , one finds for the specific heat @xmath65 where @xmath66\ ] ] and @xmath67 is the trigamma function . at low temperatures , the specific heat thus assumes the form @xmath68\left(\frac{k_\text{b}t}{\hbar\omega_0}\right)^3+o(t^5)\,.\ ] ] this result differs significantly from the expression ( [ clt ] ) in the absence of dissipation . while in the latter case , the presence of an energy gap led to an exponential suppression of the specific heat , we now find a linear increase with temperature . this behavior indicates the existence of a finite density of states even at small excitation energies @xcite . even at high temperatures the effect of dissipation can be detected , albeit in a less spectacular manner . the leading correction in the high - temperature expansion reads @xmath69 thus , the leading correction depends on the damping strength @xmath54 . this finding is in clear contrast to the behavior of the quantum escape rate @xcite : there , the leading quantum correction to the escape rate always enhances the classical result and is _ independent _ of the dissipation strength . another prescription to obtain the specific heat starts out from the canonical partition function @xmath70 } { \tr_b\left[\exp(-\beta h_\text{b})\right]}\,,\ ] ] where @xmath71 denotes the partial trace in the hilbert space of the bath . in the absence of a system - bath coupling , this expression would correspond to the partition function of the system alone . for a damped harmonic oscillator , the partition function becomes @xcite @xmath72 with the matsubara frequencies @xmath73 . in view of the divergence for strictly ohmic damping mentioned above , we allow here for a possible frequency dependence of @xmath74 . following the standard procedure of statistical mechanics , we can obtain the energy from the partition function by means of @xmath75 inserting ( [ zhoe ] ) , one obtains @xmath76\ ] ] which in general differs from the expression @xmath77\ ] ] obtained by evaluating the integral ( [ fnt ] ) by residues . the only exception is the special case of strictly ohmic damping where @xmath78 is constant . this generally non - vanishing difference does not come as a surprise if one only takes a closer look at the partition function ( [ zho ] ) : the evaluation of ( [ ezz ] ) yields @xmath79 \end{aligned}\ ] ] where the index `` b '' denotes an average with respect to the bath hamiltonian @xmath80 only . this result differs from the energy @xmath81 by the term in the brackets which , generally , vanishes only in the absence of a system - bath coupling . the coincidence between @xmath82 and @xmath83 for the harmonic oscillator subject to strictly ohmic damping should therefore be considered as exceptional . nevertheless , we briefly sketch how one would obtain the specific heat from the partition function because this will give us as a by - product an expression for the entropy of the damped harmonic oscillator . for strictly ohmic damping the product ( [ zhoe ] ) does not converge and in principle a high - frequency cutoff for @xmath74 should be introduced . however , this divergence can again be traced back to an infinite energy shift due to the environmental coupling . we may shift the energy by an arbitrary amount @xmath84 by multiplying the partition function by @xmath85 without changing the entropy or the specific heat . after performing an appropriate energy shift , we arrive at an expression of the partition function valid even for strictly ohmic damping @xcite , reading @xmath86 where @xmath87 is the gamma function , @xmath88 are defined in ( [ lambda ] ) , and @xmath89 is a constant whose precise value is irrelevant for the following . by virtue of ( [ hos ] ) we obtain for the entropy the result @xmath90 \,,\ ] ] where we introduced the abbreviation @xmath91-z\psi(1+z)\ ] ] with the digamma function @xmath92 . in the absence of damping , @xmath93 , this reproduces the result ( [ hose ] ) for the entropy of an uncoupled harmonic oscillator . for very low temperatures , the entropy ( [ sdho ] ) vanishes like @xmath94 as required by the third law of thermodynamics . by means of ( [ cs ] ) , the expression ( [ cho ] ) for the specific heat is recovered identically . the low - temperature behavior ( [ slt ] ) is in agreement with the expression derived by ford and oconnell on the basis of the free energy @xcite . for the damped harmonic oscillator these authors found that the entropy vanishes in the limit of zero temperature also for more general forms of the bath density of states . according to the equipartition theorem the specific heat of a free particle is @xmath95 . in the limit of an infinite `` box '' this represents even the correct quantum value , because the classical and the quantum partition function become equal . in view of the fact that the specific heat of an ideal gas thus remains non - zero down to the lowest temperatures , one becomes curious to investigate the specific heat of a free particle coupled to an environment when the coupling strength is not assumed to vanish : does quantum dissipation help to restore the third law ? free quantum brownian motion has been addressed in earlier work @xcite wherein the main focus centered on the role of free quantum diffusion @xcite . interestingly enough , for ohmic dissipation the quantum diffusion remains `` classical '' , being proportional to time @xmath96 , except at zero temperature itself , where one finds a logarithmic behavior in time @xmath96 @xcite . at finite temperatures this quantum behavior is observable at intermediate times only @xcite . one is thus tempted to conclude that finite quantum dissipation will not be sufficient to cure the classical behavior for the specific heat . therefore , we shall next investigate the behavior of the specific heat for free quantum brownian motion in closer detail . as already mentioned at the end of section [ pfe ] , simply taking the limit @xmath44 in the results for the damped harmonic oscillator is not without problems . we therefore start our calculation from the energy @xmath97\ ] ] which can be obtained from the second moment of momentum @xmath98 by evaluating the integral ( [ fnt ] ) by residues . proceeding as in section [ cdho ] , one derives the specific heat of the free particle in the presence of strictly ohmic damping @xmath99 which for either @xmath100 or @xmath101 yields @xmath45 , as expected from the equipartition theorem . on the other hand , for low temperatures the specific heat tends to zero as @xmath102 in agreement with the third law of thermodynamics . the specific heat ( [ cfpo ] ) together with its linear low - temperature behavior are depicted in the main part of fig . [ cfree ] as full and dashed line , respectively . temperatures @xmath103 much larger than the damping strength are required in order to restore the classical result @xmath104 . of a free quantum brownian particle is shown as a function of the temperature @xmath0 for strictly ohmic friction of strength @xmath54 . the dashed line indicates the linear low - temperature behavior . in the inset , the modification due to a finite cutoff frequency @xmath105 is depicted for different cutoff scales , @xmath106 from the upper to the lower full line . the dashed line indicates again the linear low - temperature behavior.,scaledwidth=80.0% ] as the damping strength already serves to set the temperature scale , it is instructive to introduce a cutoff in the density of states ( [ jom ] ) of the bath oscillators in order to study how a reduction of the environmental influence changes the dependence of the specific heat on temperature . to this end , we introduce a high - frequency cutoff @xmath105 by choosing the drude model where @xmath107 this environment leads to a specific heat @xmath108 \,,\ ] ] where we have introduced the abbreviations @xmath109 independently of the value of the cutoff frequency @xmath105 , the specific heat will go to zero linearly as stated in ( [ cfplin ] ) . however , for temperatures larger than the cutoff frequency , the suppression of the specific heat due to the environmental coupling will be ineffective . a reduction of the environment thus tends to restore the classical value of @xmath95 for the specific heat . this effect is depicted in the inset in fig . [ cfree ] , where the cutoff frequency takes the values @xmath110 and @xmath111 from the upper to the lower curve . the dashed curve represents the linear low - temperature behavior which still dominates at temperatures below @xmath112 . in contrast to many phenomena in quantum dissipation , here an increase of the coupling to the environment does not render the system more classical . on the contrary , a stronger environmental coupling makes the dissipative quantum system behave more quantum mechanically and thus helps to ensure the validity of the third law of thermodynamics , the vanishing of the specific heat with decreasing temperatures . with this work we have explored the behavior of the specific heat for a quantum system that is coupled to a heat bath with finite coupling strength . our findings are contrasted with the third law of thermodynamics which generically predicts a vanishing of the specific heat at low temperatures . for a harmonic oscillator the presence of quantum dissipation changes the well - known einstein - like behavior of an exponentially fast approach towards zero specific heat into a power - law behavior with a slope that increases with increasing coupling strength . even more intriguing is the behavior for a freely moving quantum particle : while the quantum treatment in absence of dissipation simply coincides with the classical behavior , the specific heat takes a constant value @xmath113 , the role of finite quantum dissipation is able to restore the third law , yielding a leading linear temperature dependence . quite counterintuitively , its approach to the classical value occurs the faster the weaker is the dissipation strength . in contrast to common quantum statistical mechanics which intrinsically is based on a vanishingly small coupling to the environment , the finite coupling strength between the sub - system and the bath causes some subtleties that must be recognized . as made explicit in section [ evspfdho ] , the thermodynamic quantities depend on the procedure invoked in their definition : the commonly used expression based on the partition function provides results that generally do _ not _ agree with the result obtained from the corresponding quantum expectation value . interestingly enough , in the strict ohmic limit ( in the absence of a high - frequency cutoff ) the specific heat for the damped harmonic oscillator does not depend on the prescription employed . for the case of memory friction , however , where a finite cutoff frequency is present , the thermodynamic quantities depend both on the value of this cutoff and the prescription used in their evaluation . the results obtained for the low - temperature behavior of the specific heat of simple quantum systems are not only of academic interest , but may turn out to be relevant for experiments in nanoscience where one tests the quantum thermodynamics of small systems @xcite that are coupled to an environment with a finite coupling strength . this work has been supported by the deutsche forschungsgemeinschaft ( dfg ) ( ph , sfb 486 ) . both authors like to congratulate peter talkner for his first 60 years and his fine scientific career . we both have heavily profited repeatedly and continue to strongly profit from his insight and breadth of knowledge . may our present ongoing fruitful collaborations with him blossom further and even intensify . the authors of reference @xcite introduced two different partition functions for a damped harmonic oscillator , see their eqs . ( 4.19 ) and ( 5.11 ) . the one which corresponds to our eq . ( [ zhoe ] ) is denoted as @xmath114 and is defined in their eq .	the quantum thermodynamic behavior of small systems is investigated in presence of finite quantum dissipation . we consider the archetype cases of a damped harmonic oscillator and a free quantum brownian particle . a main finding is that quantum dissipation helps to ensure the validity of the third law . for the quantum oscillator , finite damping replaces the zero - coupling result of an exponential suppression of the specific heat at low temperatures by a power - law behavior . rather intriguing is the behavior of the free quantum brownian particle . in this case , quantum dissipation is able to restore the third law : instead of being constant down to zero temperature , the specific heat now vanishes proportional to temperature with an amplitude that is _ inversely _ proportional to the ohmic dissipation strength . a distinct subtlety of finite quantum dissipation is the result that the various thermodynamic functions of the sub - system do not only depend on the dissipation strength but depend as well on the prescription employed in their definition .
theory of measurements has a special status in quantum mechanics . unlike classical mechanics , in quantum mechanics it can not be assumed that the effect of the measurement on the system can be made arbitrarily small . it is necessary to supplement quantum theory with additional postulates , describing the measurement . one of such additional postulate is von neumann s state reduction ( or projection ) postulate @xcite . the essential peculiarity of this postulate is its nonunitary character . however , this postulate refers only to an ideal measurement , which is instantaneous and arbitrarily accurate . real measurements are described by the projection postulate only roughly . the important consequence of von neumann s projection postulate is the quantum zeno effect . in quantum mechanics short - time behavior of nondecay probability of unstable particle is not exponential but quadratic @xcite . this deviation from the exponential decay has been observed by wilkinson _ _ @xcite . in 1977 , mishra and sudarshan @xcite showed that this behavior when combined with the quantum theory of measurement , based on the assumption of the collapse of the wave function , leaded to a very surprising conclusion : frequent observations slowed down the decay . an unstable particle would never decay when continuously observed . mishra and sudarshan have called this effect the quantum zeno paradox or effect . the effect is called so in allusion to the paradox stated by greek philosopher zeno ( or zenon ) of elea . the very first analysis does not take into account the actual mechanism of the measurement process involved , but it is based on an alternating sequence of unitary evolution and a collapse of the wave function . the zeno effect has been experimentally proved @xcite in a repeatedly measured two - level system undergoing rabi oscillations . the outcome of this experiment has also been explained without the collapse hypothesis @xcite . later it was realized that the repeated measurements could not only slow the quantum dynamics but the quantum process may be accelerated by frequent measurements as well @xcite . this effect was called a quantum anti - zeno effect by kaulakys and gontis @xcite , who argued that frequent interrogations may destroy quantum localization effect in chaotic systems . an effect , analogous to the quantum anti - zeno effect has been obtained in a computational study involving barrier penetration , too @xcite . recently , an analysis of the acceleration of a chemical reaction due to the quantum anti - zeno effect has been presented in ref . @xcite . although a great progress in the investigation of the quantum zeno effect has been made , this effect is not completely understood as yet . in the analysis of the quantum zeno effect the finite duration of the measurement becomes important , therefore , the projection postulate is not sufficient to solve this problem . the complete analysis of the zeno effect requires a more precise model of measurement than the projection postulate . the purpose of this article is to consider such a model of the measurement . the model describes a measurement of the finite duration and finite accuracy . although the used model does not describe the irreversible process , it leads , however , to the correct correlation between the states of the measured system and the measuring apparatus . due to the finite duration of the measurement it is impossible to consider infinitely frequent measurements , as in ref . the highest frequency of the measurements is achieved when the measurements are performed one after another , without the period of the measurement - free evolution between two successive measurements . in this paper we consider such a sequence of the measurements . our goal is to check whether this sequence of the measurements can change the evolution of the system and to verify the predictions of the quantum zeno effect . the work is organized as follows . in section [ sec : mod ] we present the model of the measurement . a simple case is considered in section [ sec : id ] in order to determine the requirements for the duration of the measurement . in section [ sec : meas ] we derived a general formula for the probability of the jump into another level during the measurement . the effect of repeated measurements on the system with a discrete spectrum is investigated in section [ sec : discr ] . the decaying system is considered in section [ sec : dec ] . section [ sec : concl ] summarizes our findings . we consider a system which consists of two parts . the first part of the system has the discrete energy spectrum . the hamiltonian of this part is @xmath0 . the other part of the system is represented by hamiltonian @xmath1 . hamiltonian @xmath1 commutes with @xmath0 . in a particular case the second part can be absent and @xmath1 can be zero . the operator @xmath2 causes the jumps between different energy levels of @xmath0 . therefore , the full hamiltonian of the system equals to @xmath3 . the example of such a system is an atom with the hamiltonian @xmath0 interacting with the electromagnetic field , represented by @xmath1 . we will measure in which eigenstate of the hamiltonian @xmath0 the system is . the measurement is performed by coupling the system with the detector . the full hamiltonian of the system and the detector equals to @xmath4 where @xmath5 is the hamiltonian of the detector and @xmath6 represents the interaction between the detector and the system . we choose the operator @xmath6 in the form @xmath7 where @xmath8 is the operator acting in the hilbert space of the detector and the parameter @xmath9 describes the strength of the interaction . this system detector interaction is that considered by von neumann @xcite and in refs . @xcite . in order to obtain a sensible measurement , the parameter @xmath9 must be large . we require a continuous spectrum of operator @xmath8 . for simplicity , we can consider the quantity @xmath10 as the coordinate of the detector . the measurement begins at time moment @xmath11 . at the beginning of the interaction with the detector , the detector is in the pure state @xmath12 . the full density matrix of the system and detector is @xmath13 where @xmath14 is the density matrix of the system . the duration of the measurement is @xmath15 . after the measurement the density matrix of the system is @xmath16 and the density matrix of the detector is @xmath17 where @xmath18 is the evolution operator of the system and detector , obeying the equation @xmath19 with the initial condition @xmath20 . since the initial density matrix is chosen in a factorizable form , the density matrix of the system after the interaction depends linearly on the density matrix of the system before the interaction . we can represent this fact by the equality @xmath21 where @xmath22 is the superoperator acting on the density matrices of the system . if the vectors @xmath23 form the complete basis in the hilbert space of the system we can rewrite eq . ( [ eq : supop ] ) in the form @xmath24 where the sum over the repeating indices is supposed . the matrix elements of the superoperator are @xmath25 due to the finite duration of the measurement it is impossible to realize the infinitely frequent measurements . the highest frequency of the measurements is achieved when the measurements are performed one after another without the period of the measurement - free evolution between two successive measurements . therefore , we model a continuous measurement by the subsequent measurements of the finite duration and finite accuracy . after @xmath26 measurements the density matrix of the system is @xmath27 further , for simplicity we will neglect the hamiltonian of the detector . after this assumption the evolution operator is equal to @xmath28 where the operator @xmath29 obeys the equation @xmath30 with the initial condition @xmath31 . then the superoperator @xmath22 is @xmath32 in order to estimate the necessary duration of the single measurement it is convenient to consider the case when the operator @xmath33 . in such a case the description of the evolution is simpler . the measurement of this kind occurs also when the influence of the perturbation operator @xmath34 is small in comparison with the interaction between the system and the detector and , therefore , the operator @xmath34 can be neglected . we can choose the basis @xmath35 common for the operators @xmath0 and @xmath1 , @xmath36 where @xmath37 numbers the eigenvalues of the hamiltonian @xmath0 and @xmath38 represents the remaining quantum numbers . since the hamiltonian of the system does not depend on @xmath39 we will omit the parameter @xmath11 in this section . from eq . ( [ eq : supmatr ] ) we obtain the superoperator @xmath40 in the basis @xmath35 @xmath41 where @xmath42 and @xmath43 represent the kronecker s delta in a discrete case and the dirac s delta in a continuous case . ( [ eq : s1 ] ) can be rewritten using the correlation function @xmath44 we can express this function as @xmath45 . since vector @xmath46 is normalized , the function @xmath47 tends to zero when @xmath48 increases . there exists a constant @xmath49 such that the correlation function @xmath50 is small if the variable @xmath51 . then the equation for the superoperator @xmath40 is @xmath52 using eqs . ( [ eq : rho1 ] ) and ( [ eq : s2 ] ) we find that after the measurement the non - diagonal elements of the density matrix of the system become small , since @xmath53 is small for @xmath54 when @xmath55 is large . the density matrix of the detector is @xmath56 from eqs . ( [ eq : evol1 ] ) and ( [ eq : det1 ] ) we obtain @xmath57 where @xmath58 the probability that the system is in the energy level @xmath37 may be expressed as @xmath59 introducing the state vectors of the detector @xmath60 we can express the density operator of the detector as @xmath61 the measurement is complete when the states @xmath62 are almost orthogonal . the different energies can be separated only when the overlap between the corresponding states @xmath62 is almost zero . the scalar product of the states @xmath62 with different energies @xmath63 and @xmath64 is @xmath65 the correlation function @xmath50 is small when @xmath51 . therefore , we have the estimation for the error of the energy measurement @xmath66 as @xmath67 and we obtain the expression for the necessary duration of the measurement @xmath68 where @xmath69 since in our model the measurements are performed immediately one after the other , from eq . ( [ eq : tt ] ) it follows that the rate of measurements is proportional to the strength of the interaction @xmath9 between the system and the measuring device . the operator @xmath2 represents the perturbation of the unperturbed hamiltonian @xmath70 . we will take into account the influence of the operator @xmath34 by the perturbation method , assuming that the strength of the interaction @xmath9 between the system and detector is large . the operator @xmath2 in the interaction picture is @xmath71 in the second order approximation the evolution operator equals to @xmath72 using eqs . ( [ eq : supmatr ] ) and ( [ eq : evol2 ] ) we can obtain the superoperator @xmath73 in the second order approximation , too . the expression for the matrix elements of the superoperator @xmath73 is given in the appendix ( eqs.([eq : ap1 ] ) , ( [ eq : ap2 ] ) and ( [ eq : ap3 ] ) ) . the probability of the jump from the level @xmath74 to the level @xmath75 during the measurement is @xmath76 . using eqs.([eq : ap1 ] ) , ( [ eq : ap2 ] ) and ( [ eq : ap3 ] ) we obtain @xmath77 the expression for the jump probability can be further simplified if the operator @xmath34 does not depend on @xmath39 . we introduce the function @xmath78 changing variables we can rewrite the jump probability as @xmath79 introducing the fourier transformation of @xmath80 @xmath81 and using eq . ( [ eq : w ] ) we obtain the equality @xmath82 where @xmath83 from eq . ( [ eq : pp ] ) , using the equality @xmath84 , we obtain @xmath85 the quantity @xmath86 equals to @xmath87 we see that the quantity @xmath88 characterizes the perturbation . let us consider the measurement effect on the system with the discrete spectrum . the hamiltonian @xmath0 of the system has a discrete spectrum , the operator @xmath89 , and the operator @xmath2 represents a perturbation resulting in the quantum jumps between the discrete states of the system @xmath0 . for the separation of the energy levels , the error in the measurement should be smaller than the distance between the nearest energy levels of the system . it follows from this requirement and eq . ( [ eq : tt ] ) that the measurement time @xmath90 , where @xmath91 is the smallest of the transition frequencies @xmath92 . when @xmath9 is large then @xmath93 is not very small only in the region @xmath94 . we can estimate the probability of the jump to the other energy level during the measurement , replacing @xmath47 by @xmath95 in eq . ( [ eq : w1 ] ) . then from eq . ( [ eq : w1 ] ) we obtain @xmath96 we see that the probability of the jump is proportional to @xmath97 . consequently , for large @xmath98 , i.e. for the strong interaction with the detector , the jump probability is small . this fact represents the quantum zeno effect . however , due to the finiteness of the interaction strength the jump probability is not zero . after sufficiently large number of measurements the jump occurs . we can estimate the number of measurements @xmath26 after which the system jumps into other energy levels from the equality @xmath99 where @xmath100 is the largest matrix element of the perturbation operator @xmath101 . this estimation allows us to introduce the characteristic time , during which the evolution of the system is inhibited @xmath102 we call this duration the inhibition time ( it is natural to call this duration the zeno time , but this term has already different meaning ) . the full probability of the jump from level @xmath74 to other levels is @xmath103 . from eq.([eq : jumpprob ] ) we obtain @xmath104 if the matrix elements of the perturbation @xmath101 between different levels are of the same size then the jump probability increases linearly with the number of the energy levels . this behavior has been observed in ref . @xcite . due to the unitarity of the operator @xmath29 it follows from eq.([eq : supmatr ] ) that the superoperator @xmath22 obeys the equalities @xmath105 if the system has a finite number of energy levels , the density matrix of the system is diagonal and all states are equally occupied ( i.e. , @xmath106 where @xmath107 is the number of the energy levels ) then from eq . ( [ eq : prop1 ] ) it follows that @xmath108 . such a density matrix is the stable point of the map @xmath109 . therefore , we can expect that after a large number of measurements the density matrix of the system tends to this density matrix . when @xmath98 is large and the duration of the measurement is small , we can neglect the non - diagonal elements in the density matrix of the system , since they always are of order @xmath97 . replacing @xmath47 by @xmath95 in eqs . ( [ eq : ap1 ] ) , ( [ eq : ap2 ] ) and ( [ eq : ap3 ] ) and neglecting the elements of the superoperator @xmath73 that cause the arising of the non - diagonal elements of the density matrix , we can write the equation for the superoperator @xmath73 as @xmath110 where @xmath111 then for the diagonal elements of the density matrix we have @xmath112 , or @xmath113 if the perturbation @xmath101 does not depend on @xmath39 then it follows from eq . ( [ eq : deriv ] ) that the diagonal elements of the density matrix evolve exponentially . as an example we will consider the evolution of the measured two - level system . the system is forced by the perturbation @xmath101 which induces the jumps from one state to another . the hamiltonian of this system is @xmath114 where @xmath115 here @xmath116 are pauli matrices and @xmath117 . the hamiltonian @xmath0 has two eigenfunctions @xmath118 and @xmath119 with the eigenvalues @xmath120 and @xmath121 respectively . the evolution operator of the unmeasured system is @xmath122 where @xmath123 if the initial density matrix is @xmath124 then the evolution of the diagonal elements of the unmeasured system s density matrix is given by the equations @xmath125 let us consider now the dynamics of the measured system . the equations for the diagonal elements of the density matrix ( eq . ( [ eq : deriv ] ) ) for the system under consideration are [ eq : eqs1 ] @xmath126 where the inhibition time , according to eq . ( [ eq : inht ] ) , is @xmath127 the solution of eqs . ( [ eq : eqs1 ] ) with the initial condition @xmath128 is @xmath129 from eq . ( [ eq : prop1 ] ) it follows that if the density matrix of the system is @xmath130 then @xmath131 . hence , when the number of the measurements tends to infinity , the density matrix of the system approaches @xmath132 . = .6 = .6 = .6 we have performed the numerical analysis of the dynamics of the measured two - level system ( [ eq : ham1])([eq : ham3 ] ) using eqs.([eq : rho1 ] ) , ( [ eq : supmatr ] ) and ( [ eq : evol4 ] ) with the gaussian correlation function ( [ f ] ) @xmath133 from the condition @xmath134 we have @xmath135 . the initial state of the system is @xmath119 . the matrix elements of the density matrix @xmath136 and @xmath137 are represented in fig . [ fig1 ] and fig . [ fig2 ] , respectively . in fig . [ fig1 ] the approximation ( [ eq : aprox1 ] ) is also shown . this approach is close to the exact evolution . the matrix element @xmath136 for two different values of @xmath9 is shown in fig . we see that for larger @xmath9 the evolution of the system is slower . the influence of the repeated non - ideal measurements on the two level system driven by the periodic perturbation has also been considered in refs . similar results have been found : the occupation of the energy levels changes exponentially with time , approaching the limit @xmath138 . we consider the system which consists of two parts . we can treat the first part as an atom , and the second part as the field ( reservoir ) . the energy spectrum of the atom is discrete and the spectrum of the field is continuous . the hamiltonians of these parts are @xmath0 and @xmath1 respectively and the eigenfunctions are @xmath23 and @xmath139 , @xmath140 there is the interaction between the atom and the field represented by the operator @xmath34 . so , the hamiltonian of the system is @xmath141 the basis for the full system is @xmath142 . when the measurement is not performed , such a system exhibits exponential decay , valid for the intermediate times . the decay rate is given according to the fermi s golden rule @xmath143 where @xmath144 and @xmath145 is the density of the reservoir s states . when the energy level of the atom is measured , we can use the perturbation theory , as it is in the discrete case . the initial state of the field is a vacuum state @xmath118 with energy @xmath146 . then the density matrix of the atom is @xmath147 or @xmath148 , where @xmath149 is an effective superoperator @xmath150 when the states of the atom are weakly coupled to a broad band of states ( continuum ) , the transitions back to the excited state of the atom can be neglected ( i.e. , we neglect the influence of emitted photons on the atom ) . therefore , we can use the superoperator @xmath149 for determination of the evolution of the atom . since the states in the reservoir are very dense , one can replace the sum over @xmath151 by an integral over @xmath152 @xmath153 where @xmath154 is the density of the states in the reservoir . the density matrix of the field is @xmath155 . the diagonal elements of the field s density matrix give the spectrum . if the initial state of the atom is @xmath156 then the distribution of the field s energy is @xmath157 . from eqs . ( [ eq : ap1]),([eq : ap2 ] ) and ( [ eq : ap3 ] ) we obtain @xmath158 where @xmath159 is given by the equation ( [ eq : pp ] ) . from eq . ( [ eq : spectr ] ) we see that @xmath160 is the measurement - modified shape of the spectral line . the integral in eq . ( [ eq : pp ] ) is small when the exponent oscillates more rapidly than the function @xmath161 . this condition is fulfilled when @xmath162 . consequently , the width of the spectral line is @xmath163 the width of the spectral line is proportional to the strength of the measurement ( this equation is obtained using the assumption that the strength of the interaction with the measuring device @xmath9 is large and , therefore , the natural width of the spectral line can be neglected ) . the broadening of the spectrum of the measured system is also reported in ref . @xcite for the case of an electron tunneling out of a quantum dot . the probability of the jump from the state @xmath164 to the state @xmath165 is @xmath166 . from eqs . ( [ eq : sef ] ) it follows @xmath167 using eq . ( [ eq : resw ] ) we obtain the equality @xmath168 where @xmath169 the expression for @xmath88 according to eq . ( [ eq : g ] ) is @xmath170 the quantity @xmath171 is the reservoir coupling spectrum . the measurement - modified decay rate is @xmath172 . from eq . ( [ eq : decprob2 ] ) we have @xmath173 the equation ( [ eq : result ] ) represents a universal result : the decay rate of the frequently measured decaying system is determined by the overlap of the reservoir coupling spectrum and the measurement - modified level width . this equation was derived by kofman and kurizki @xcite , assuming the ideal instantaneous projections . we show that eq . ( [ eq : result ] ) is valid for the more realistic model of the measurement , as well . an equation , similar to eq.([eq : result ] ) has been obtained in ref . @xcite , considering a destruction of the final decay state . depending on the reservoir spectrum @xmath88 and the strength of the measurement the inhibition or acceleration of the decay can be obtained . if the interaction with the measuring device is weak and , consequently , the width of the spectral line is much smaller than the width of the reservoir spectrum , the decay rate equals the decay rate of the unmeasured system , given by the fermi s golden rule ( [ eq : goldrule ] ) . in the intermediate region , when the width of the spectral line is rather small compared with the distance between @xmath174 and the nearest maximum in the reservoir spectrum , the decay rate grows with increase of @xmath98 . this results in the anti - zeno effect . if the width of the spectral line is much greater compared both with the width of the reservoir spectrum and the distance between @xmath174 and the centrum of the reservoir spectrum , the decay rate decreases when @xmath98 increases . this results in the quantum zeno effect . in such a case we can use the approximation @xmath175 where @xmath176 is defined by the equality @xmath177 and @xmath178 is the centrum of @xmath88 . then from eq . ( [ eq : result ] ) we obtain the decay rate @xmath179 . from eq . ( [ eq : pp ] ) , using the condition @xmath180 and the equality @xmath181 we obtain @xmath182 therefore , the decay rate is equal to @xmath183 the obtained decay rate is insensitive to the spectral shape of the reservoir and is inverse proportional to the measurement strength @xmath98 . in this work we investigate the quantum zeno effect using the definite model of the measurement . we take into account the finite duration and the finite accuracy of the measurement . the general equation for the probability of the jump during the measurement is derived ( [ eq : resw ] ) . the behavior of the system under the repeated measurements depends on the strength of measurement and on the properties of the system . when the the strength of the interaction with the measuring device is sufficiently large , the frequent measurements of the system with discrete spectrum slow down the evolution . however , the evolution can not be fully stopped . under the repeated measurements the occupation of the energy levels changes exponentially with time , approaching the limit of the equal occupation of the levels . the jump probability is inversely proportional to the strength of the interaction with the measuring device . in the case of a continuous spectrum the measurements can cause inhibition or acceleration of the evolution . our model of the continuous measurement gives the same result as the approach based on the projection postulate @xcite . the decay rate is equal to the convolution of the reservoir coupling spectrum with the measurement - modified shape of the spectral line . the width of the spectral line is proportional to the strength of the interaction with the measuring device . when this width is much greater than the width of the reservoir , the quantum zeno effect takes place . under these conditions the decay rate is inversely proportional to the strength of the interaction with the measuring device . in a number of decaying systems , however , the reservoir spectrum @xmath88 grows with frequency almost up to the relativistic cut - off and the strength of the interaction required for the appearance of the quantum zeno effect is so high that the initial system is significantly modified . when the spectral line is not very broad , the decay rate may be increased by the measurements more often than it may be decreased and the quantum anti - zeno effect can be obtained . we obtain the superoperator @xmath73 in the second order approximation substituting the approximate expression for the evolution operator ( [ eq : evol2 ] ) into eq . ( [ eq : supmatr ] ) . thus we have @xmath184 where @xmath185 is the superoperator of the unperturbed measurement given by eq . ( [ eq : s2 ] ) , @xmath186 is the first order correction , @xmath187 and @xmath188 is the second order correction , @xmath189 where @xmath190 j. von neumann , _ mathematisch grundlagen der quanten - mechanik _ ( springer , berlin , 1932 ) . english translation : _ mathematical foundations of quantum mechanics _ ( princeton university press , princeton , nj , 1955 ) .	in 1977 , mishra and sudarshan showed that an unstable particle would never be found decayed while it was continuously observed . they called this effect the quantum zeno effect ( or paradox ) . later it was realized that the frequent measurements could also accelerate the decay ( quantum anti - zeno effect ) . in this paper we investigate the quantum zeno effect using the definite model of the measurement . we take into account the finite duration and the finite accuracy of the measurement . a general equation for the jump probability during the measurement is derived . we find that the measurements can cause inhibition ( quantum zeno effect ) or acceleration ( quantum anti - zeno effect ) of the evolution , depending on the strength of the interaction with the measuring device and on the properties of the system . however , the evolution can not be fully stopped .
among the many possibilities for coherent manipulation of quantum systems , stimulated raman adiabatic passage ( stirap ) is one of the most widely used and studied @xcite . this technique transfers population adiabatically between two states @xmath0 and @xmath1 in a three - state quantum system , without populating the intermediate state @xmath2 even when the time - delayed driving fields are on exact resonance with the respective pump and stokes transitions . the technique of stirap relies on the existence of a dark state , which is a time - dependent coherent superposition of the initial and target states only , and which is an eigenstate of the hamiltonian if states @xmath0 and @xmath1 are on two - photon resonance . because stirap is an adiabatic technique , it is robust to variations in most of the experimental parameters . in the early applications of stirap in atomic and molecular physics its efficiency , most often in the range 90 - 95% , has barely been scrutinized because such an accuracy suffices for most purposes . because stirap is resilient to decoherence linked to the intermediate state ( which is often an excited state ) this technique has quickly attracted attention as a promising control tool for quantum information processing @xcite . the latter , however , demands very high fidelity of operations , with the admissible error at most @xmath3 , which is hard to achieve with the standard stirap because , due to its adiabatic nature , it approaches unit efficiency only asymptotically , as the temporal pulse areas increase . for usual pulse shapes , e.g. , gaussian , the necessary area for the @xmath3 benchmark is so large that it may break various restrictions in a real experiment . several scenarios have been proposed to optimize stirap in order to achieve such an accuracy . because the loss of efficiency in stirap derives from incomplete adiabaticity , unanyan _ @xcite , and later chen _ et al . _ @xcite , have proposed to annul the nonadiabatic coupling by adding a third pulsed field on the transition @xmath4 . however , this field must coincide in time with the nonadiabatic coupling exactly ; its pulse area , in particular , must equal @xmath5 , which makes the pump and stokes fields largely redundant . an alternative approach to improve adiabaticity is based on the dykhne - davis - pechukas formula @xcite , which dictates that nonadiabatic losses are minimized when the eigenenergies of the hamiltonian are parallel . this approach , however , prescribes a strict time dependences for the pump and stokes pulse shapes @xcite , or for both the pulse shapes and the detunings @xcite . another basic approach to robust population transfer , which is an alternative to adiabatic techniques , is the technique of composite pulses , which is widely used in nuclear magnetic resonance ( nmr ) @xcite , and more recently , in quantum optics @xcite . this technique , implemented mainly in two - state systems , replaces the single pulse used traditionally for driving a two - state transition by a sequence of pulses with appropriately chosen phases ; these phases are used as a control tool for shaping the excitation profile in a desired manner , e.g. , to make it more robust to variations in the experimental parameters intensities and frequencies . recently , we have proposed a hybrid technique composite adiabatic passage ( cap ) which combines the techniques of composite pulses and adiabatic passage via a level crossing in a two - state system @xcite . cap can deliver extremely high fidelity of population transfer , far beyond the quantum computing benchmark , and far beyond what can be achieved with a single frequency - chirped pulse . recently , the cap technique has been demonstrated experimentally in a doped solid @xcite . to state @xmath1 via a sequence of pump - stokes pulse pairs . on one - photon resonance ( top ) , the order of the pump and stokes pulses is reversed from pair to pair , while off single - photon resonance it is the same for all pulse pairs . ] in this paper , we combine the two basic techniques of composite pulses and stirap into a hybrid technique , which we name _ composite stirap_. this technique , which represents a sequence of an odd number of forward and backward ordinary stiraps , @xmath6 , adds to stirap the very high fidelity of composite pulses . each individual stirap can be very inaccurate , the affordable error being as much as 20 - 30% , but all errors interfere destructively and cancel in the end , thereby producing population transfer with a cumulative error far below the quantum computing benchmark of @xmath3 . we derive an analytical formula for the composite phases , applicable to an arbitrary odd number of pulse pairs @xmath7 ; the phases do not depend on the shape of the pulses and their mutual delay . the dynamics of a three - state @xmath8 system ( fig . [ tog ] ) is described by the schrdinger equation , [ schr ] i_t ( t ) = ( t)(t ) , where the vector @xmath9^t$ ] contains the three probability amplitudes . the hamiltonian in the rotating - wave approximation and on two - photon resonance between states @xmath0 and @xmath1 is [ h ] ( t ) = 2 , where @xmath10 and @xmath11 are the rabi frequencies of the pump and stokes fields , @xmath12 is the one - photon detuning between each laser carrier frequency and the bohr frequency of the corresponding transition , and @xmath13 is the population loss rate from state @xmath2 ; we assume @xmath14 . states @xmath0 and @xmath1 are coupled by @xmath10 , while states @xmath2 and @xmath1 are coupled by @xmath11 . the evolution of the system is described by the propagator @xmath15 , which connects the amplitudes at the initial and final times , @xmath16 and @xmath17 : @xmath18 . the mathematics is substantially different when the pump and stokes fields are on resonance or far off - resonance with the corresponding transition : therefore we consider these cases separately . first , we will consider the one - photon resonance , @xmath19 . then there is a mapping between the three - state problem and a corresponding two - state problem described by the hamiltonian ( t ) = 2 . ( in this correspondence , @xmath10 and @xmath11 are assumed real . ) in general , if the two - state propagator is parameterized in terms of the complex cayley - klein parameters @xmath20 and @xmath21 ( @xmath22 ) as = , we can write the propagator of stirap as [ u - ab ] = . if @xmath10 and @xmath11 are reflections of each other , @xmath23 [ e.g. , if @xmath24 and @xmath25 are identical symmetric functions of time ] , where @xmath26 is the pulse delay , then it is easily shown that @xmath27 . we use this property to parameterize the stirap propagator as @xmath28 in the adiabatic limit , @xmath29 ; hence @xmath30 . .pump and stokes phases for different number of pulse pairs @xmath7 for resonant composite stirap . [ cols="<,<",options="header " , ] the fidelity of the nonresonant composite stirap is illustrated in fig . [ c - stirap ] ( bottom frames ) and fig . [ contoffres ] . again , composite stirap greatly outperforms single stirap in terms of fidelity and robustness . as a function of the pulse delay and the peak rabi frequency for a single off - resonant stirap ( top ) and a sequence of five stiraps ( bottom ) , with phases given by eq . , for @xmath31 pulse shapes , eqs . . the detuning is @xmath32 . ] composite stirap may be affected by several sources of errors . in the first place , errors in the composite phases should be held low in order to keep the high fidelity . we found that an error below 1% in the phases , which is relatively easy to achieve in the lab , can be tolerated . in fig . [ c - stirap ] we have added a curve , which demonstrates the fidelity of composite stirap for @xmath33 and a standard deviation of 0.01 radians in the composite phases ; despite this error , the technique still has ultrahigh fidelity , with an error below @xmath3 . stirap owes much of its great popularity to the fact that it can operate , unlike other techniques , in the presence of population losses from the middle state @xmath2 . however , when ultrahigh fidelity is aimed the presence of such losses can reduce the fidelity and they can not be very large . ( the decay can be harmful only in the resonant case , while off - resonant composite stirap is much more resilient to them . ) we have found that in the resonant case , if the decay rate is sufficiently low , or if the pulse duration is sufficiently short ( @xmath34 ) , composite stirap still maintains high fidelity and outperforms the standard stirap , as seen in fig . [ decay ] . as @xmath13 increases above @xmath35 , stirap behaves better but the fidelities of both stirap and composite stirap drop rapidly and are inadequate for quantum computing purposes . the presence of losses can be compensated with higher rabi frequency ; as a rough estimate the scaling law @xmath36 applies . it is also important to note that in the presence of decay the pulse pairs should be as close to each other as possible , as in the inset of fig . [ decay ] . this is readily achieved with microsecond and nanosecond pulses , e.g. , produced by acoustooptic modulators , as has been demonstrated recently in a doped - solid experiment @xcite . because composite stirap involves @xmath7 pulse pairs , its duration is longer than stirap by the same factor , given that there are no gaps between the pulse pairs , as shown in the inset of fig . [ decay ] . in return , composite stirap gives a fidelity which can not be achieved with ordinary stirap , even with the much higher pulse areas . thus the main advantage of composite stirap over ordinary stirap is the ultrahigh fidelity . the main advantage of composite stirap over other variations of stirap , which provide `` shortcuts '' to adiabaticity by eliminating or reducing the nonadiabatic coupling @xcite is the simplicity of implementation , which requires just the control of the relative phases between the pulse pairs , and the preserved robustness of stirap with respect to variations in the interaction parameters . the `` shortcuts '' techniques use less pulse area , and therefore are faster than composite stirap ( although still slower than resonant techniques which use areas of just @xmath37 @xcite ) , but they give away most of the robustness of stirap by imposing strict restrictions on the pulse shapes , and some of them on the detunings too ; some of them even place considerable transient population in the intermediate state . pulse shapes with @xmath38 and @xmath19 . ] the hybrid technique proposed here combines two popular methods for manipulation of quantum systems stirap and composite pulses . it greatly outperforms the standard stirap in terms of fidelity due to cancelation of the nonadiabatic errors by destructive interference . the greatly enhanced fidelity , well beyond the quantum computing benchmark , while preserving stirap s robustness against variations in the interaction parameters , makes composite stirap a promising technique for quantum information processing . this work is supported by the bulgarian nsf grants d002 - 90/08 and dmu-03/103 , and the alexander von humboldt foundation . k. bergmann , h. theuer , and b. w. shore , rev . phys . * 70 * , 1003 ( 1998 ) ; n. v. vitanov , m. fleischhauer , b. w. shore , and k. bergmann , adv . at . , , opt . phys . * 46 * , 55 ( 2001 ) ; n. v. vitanov , t. halfmann , b. w. shore , and k. bergmann , annu . phys . chem . * 52 * , 763 ( 2001 ) . m. hennrich , t. legero , a. kuhn , and g. rempe , phys . * 85 * , 4872 ( 2000 ) ; a. kuhn , m. hennrich , and g. rempe , ibid . * 89 * , 067901 ( 2002 ) ; t. wilk , s.c . webster , h.p . specht , g. rempe , and a. kuhn , ibid . * 98 * , 063601 ( 2007 ) ; j.l . srensen , d. mller , t. iversen , j.b . thomsen , f. jensen , p. staanum , d. voigt , and m. drewsen , new j. phys . * 8 * , 261 ( 2006 ) . g. dridi , s. gurin , v. hakobyan , h.r . jauslin , and h. eleuch , phys . a * 80 * , 043408 ( 2009 ) . levitt and r. freeman , j. magn . * 33 * , 473 ( 1979 ) ; r. freeman , s. p. kempsell , and m. h. levitt , j. magn * 38 * , 453 ( 1980 ) ; m.h . levitt , prog . nmr spectrosc . * 18 * , 61 ( 1986 ) ; r. freeman , _ spin choreography _ ( spektrum , oxford , 1997 ) . h. hffner , c. f. roos , and r. blatt , phys . rep . * 469 * , 155 ( 2008 ) ; n. timoney , v. elman , s. glaser , c. weiss , m. johanning , w. neuhauser , and c. wunderlich , phys . a * 77 * , 052334 ( 2008 ) ; b.t . torosov and n.v . vitanov , phys . a * 83 * , 053420 ( 2011 ) ; s.s . ivanov and n.v . vitanov , opt . lett . * 36 * , 7 ( 2011 ) ; g.t . genov , b.t . torosov , and n.v . vitanov , phys . a * 84 * , 063413 ( 2011 ) .	we introduce a high - fidelity technique for coherent control of three - state quantum systems , which combines two popular control tools stimulated raman adiabatic passage ( stirap ) and composite pulses . by using composite sequences of pairs of partly delayed pulses with appropriate phases the nonadiabatic transitions , which prevent stirap from reaching unit fidelity , can be canceled to an arbitrary order by destructive interference , and therefore the technique can be made arbitrarily accurate . the composite phases are given by simple analytic formulas , and they are universal for they do not depend on the specific pulse shapes , the pulse delay and the pulse areas .
different helioseismic instruments , both from ground and space , observe the sun in different observables . due to the different techniques used by these instruments , it is possible to measure the solar oscillations simultaneously either as variations in the photospheric velocity or as intensity fluctuations . it is therefore important to confirm that the oscillation mode parameters measured from both the intensity and velocity agree with each other to a high degree of precision . however , the initial measurement of low degree _ p_-mode frequencies from doppler velocity ( @xmath0 ) and continuum intensity ( i@xmath1 ) observations from michelson doppler imager ( mdi ) instrument on board _ solar and heliospheric observatory _ ( soho ) showed systematic differences . a comparison of 108-day power spectra between @xmath0 and i@xmath1 showed a weighted mean difference of @xmath2hz for @xmath3 , and @xmath4hz for @xmath5 modes @xcite . since the apparent frequency shift between an oscillation observed in velocity and intensity can not be a property of the mode , it must arise from systematic errors while calculating the frequencies from the observed power spectra . hence it was argued that the source of the systematic difference could be due to the opposite asymmetry effect seen between the velocity and intensity power spectra @xcite . @xcite also presented a similar evidence using virgo and soi / mdi data . around the same time @xcite reported that the intermediate degree modes observed in @xmath0 and total spectral intensity also show different central frequencies and observed that the apparent differences could be as large as 50 @xmath6hz close to the acoustic cut - off frequency . however , the analysis of @xcite , using an asymmetric line profile - fitting formula , illustrated that the frequency difference between @xmath0 and i@xmath1 in the intermediate degree range is much smaller compared to that obtained by fitting a symmetric lorentzian profile . using the same asymmetric line profile - fitting formula , @xcite re - analyzed the data from mdi and showed that the frequency differences between @xmath0 and i@xmath1 are considerably reduced . @xcite have also analyzed data from different instruments and have argued that the reported frequency shift is merely an artifact of the reduction technique . renewed interest in the topic began when local helioseismic techniques were developed to study the properties of high - degree modes in localized regions . @xcite analyzed azimuthally averaged ( 2-d ) power spectra and inferred that the eigenfrequencies obtained using the asymmetric peak profiles agree reasonably well with each other while the use of symmetric profiles gives significant differences between frequencies computed using continuum intensity and velocity or continuum intensity and line - depth spectra . in order to gain further information for high - degree and high - frequency modes , @xcite analyzed the high - resolution gong+ data . these authors also compared the azimuthally averaged power spectra of a region centered on the equator and reported that the frequency dependence of the frequency shift between @xmath0 and i is negligible below the acoustic cutoff frequency around 5.3 mhz and substantial above the cutoff . these results supported the finding of @xcite . however , the conclusion is based on the visual comparison of the peak frequency of the power spectra and may not necessarily be a true measure of the shift due to the reversal of the asymmetry and different background power between @xmath0 and i@xmath1 spectra . it is now well established that line asymmetry of the solar power spectra alters the peak frequencies that are obtained under the assumption that the lines are symmetric ( e.g. @xcite . however , the cause of the opposite asymmetry between the velocity and intensity spectra still remains inconclusive . the current understanding is that the reversal in the sign of asymmetry between the @xmath0 and i@xmath1 spectra is due to the influence of the solar background noise that is correlated with the source @xcite and the level depends on the characteristic granulation . on the other hand , the numerical simulation @xcite indicates that the reversal is produced by the radiative transfer effects . since the physics of the correlated noise is not yet fully understood and the spatial leakage signature for @xmath0 and i is different due to their center - to - limb variations , our objective is to examine the frequency dependence of the observed asymmetry and differences in eigenfrequencies between velocity and intensity observations as a function of the radial distance from the disk center to the limb . a preliminary investigation of a similar nature using azimuthally averaged power spectra is reported in @xcite . however the present analysis differs from all earlier ones since here we use the three - dimensional ( 3-d ) power spectrum , which is associated with flow fields , while the azimuthally averaged spectrum has no flow fields associated with it . the rest of the paper is organized as follows : we briefly describe the data and analysis technique in section 2 , while the results are described in section 3 . finally , we summarize the main conclusions in section 4 . we use the dopplergrams and continuum intensity images obtained by the mdi instrument during the period of 1997 may 19 21 when solar activity was near minimum . we have chosen 4 regions centered at heliographic longitudes of @xmath7 , @xmath8 , @xmath9 , and @xmath10 ; all centered at the equator . the spatial extent of each of the localized region covers @xmath11 pixels in heliographic longitude and latitude and gives a resolution of 0.01616 mm@xmath12 . each region is tracked for 4096 minutes , which gives a frequency resolution of 4.07 @xmath6hz . the standard ring diagram technique @xcite was used to obtain the power as a function of ( @xmath13 , @xmath14 , @xmath15 ) . to extract the frequencies and other mode parameters , the three - dimensional power spectrum is fitted with a model with asymmetric peak profiles of the form @xmath16 where @xmath17 here @xmath18 is the oscillation power for a mode with a temporal frequency @xmath15 and the total wavenumber @xmath19 . the peak power in the mode is represented as @xmath20 while the coefficients @xmath21 to @xmath22 account for the variation in power with @xmath23 and along the ring . the terms involving @xmath24 and @xmath25 define the background power and @xmath26 is the mode width . the term @xmath27 gives the mean frequency , and this form is chosen as it gives satisfactory fits to the mean frequency over the whole fitting interval . the parameters @xmath28 controls the asymmetry in the peaks , and the form of asymmetry is the same as that used by @xcite i.e. the parameter is positive for positive asymmetry and negative for negative asymmetry . by setting @xmath29 we can also fit symmetric lorentzian profiles . a more detailed description of the parameters and fitting procedure is described in @xcite and @xcite . the 13 parameters ( @xmath30 , @xmath21 , @xmath31 , @xmath22 , @xmath24 , @xmath25 , @xmath32 , @xmath33 , @xmath28 , @xmath34 , @xmath35 , @xmath26 , @xmath36 ) are determined by fitting the spectra using the maximum likelihood approach @xcite . in this work we express @xmath23 in units of @xmath37 so that @xmath23 is identified with the degree @xmath38 of the spherical harmonic of the corresponding global mode . figure [ fig1 ] shows a characteristic @xmath39 diagram obtained from the fits to the power spectra at the disk center using asymmetric profile corresponding to eq . in general we obtain about 750 and 600 modes from the velocity and intensity spectra respectively . intensity spectra normally yields less number of modes due to the inherently lower signal - to - noise ratio . the number of fitted modes also decreases for regions that are away from the disk center . comparing the number of modes between symmetric and asymmetric fits we find that the symmetric fit yields more modes . this implies that the inclusion of the asymmetry in the fitting formula reduces the number of modes that are successfully fitted . this could be due to some cross - correlation between @xmath28 and other parameters of the model , particularly , the background . although , we fit the spectra up to radial order of @xmath40 , we restrict the analysis up to @xmath41 modes since higher order radial modes have large estimated fitting errors . physically , asymmetry is a result of an interaction between an outward - directed wave from the source and a corresponding inward - directed wave that passes through the region of wave propagation @xcite . to illustrate the peaks in the power spectra and to visualize the asymmetry associated with different observables , we show an example of the azimuthally averaged power for @xmath42 corresponding to the disk center spectra for @xmath0 and i@xmath1 in figure [ fig2 ] . it is apparent that the line profiles are asymmetric . at low frequency , the asymmetry agrees with the known results ; the velocity spectrum has negative asymmetry i.e. more power on the lower frequency side of the peak while the continuum intensity spectrum show positive asymmetry i.e. more power on the higher frequency side of the peak . this reversal of asymmetry is believed to arise from the addition of correlated noise to the amplitude spectra @xcite . in addition , we notice that the asymmetry associated with velocity spectra appears to be reduced or even reversed at higher frequencies . visual inspection of the azimuthally averaged power spectra show a comprehensible frequency shift of the central frequency between @xmath0 and i@xmath1 above the cutoff frequency of 5.3 mhz . we also observe that the velocity power is higher near the disk center and decreases toward the limb in agreement with the predominately radial nature of the oscillatory velocity field . figure [ fig3 ] shows the asymmetry of the line profiles as a function of frequency obtained from fitting the three dimensional power spectra at disk center and 45 e longitude . as expected , this parameter is predominantly negative for all modes in the velocity spectra ( upper panels ) indicating that there is more power on the low frequency side of the peak . the asymmetry is minimum around 3 mhz while at low end of the frequency range the asymmetry of _ p_@xmath43 and higher order radial modes is more likely to be zero within the estimated errors . thus , in this frequency range the mode frequencies obtained from asymmetric and symmetric profiles should be nearly identical . surprisingly , we do not observe the asymmetry to be reversed at higher frequencies as noted above and also in @xcite . the asymmetry for the intensity spectra ( lower panels of figure [ fig3 ] ) appears to be more complex and in general these are larger than the velocity asymmetry . for _ f _ and _ p@xmath44_-modes , this parameter is primarily positive ( more power on the higher frequency end of the peak ) and increases with frequency . for higher order _ p_-modes , asymmetry is negative both at low- and high - end of the frequency range and is likely to be insignificant or zero at other frequencies . thus the asymmetry is different than what is observed for low - degree modes @xcite , intermediate degree modes @xcite , and high - degree modes obtained from the azimuthally averaged spectra @xcite . although most of these differences may be due to frequency , we do not rule out the possibility that the form of asymmetry used in the model are simplified and may not represent the real profiles adequately at all frequencies . the asymmetry may also weakly depend on @xmath38 @xcite and this dependence has been neglected in the fitting formula . finally , we note that the asymmetry of both velocity and intensity spectra do not show any significant changes at different locations on the solar disk . it is interesting to compare the asymmetries from 3-d spectra with those obtained from azimuthally averaged spectra . figure [ fig4 ] illustrates the asymmetries obtained from both the fits . while the asymmetry observed in intensity spectra obtained from 2-d fits are entirely positive , the behavior is more complex in the 3-d spectra . on the other hand , asymmetry associated with 3-d velocity spectra is entirely negative while for 2-d spectra , it reverses sign at high end of the frequency range . thus significant differences are seen between the asymmetry obtained from fitting the three - dimensional and azimuthally averaged power spectra . although it is not easy to understand these differences due to the different functional form of the model profiles to fit different spectra ( six parameters are fitted in azimuthal spectra as opposed to 13 in the three - dimensional spectra ) , we have made a preliminary attempt by constructing synthetic 3d spectra with @xmath45 . this spectra are then azimuthally averaged and fitted to calculate the asymmetry in peak profiles . because of non - zero values of velocities ( @xmath46 ) the process of azimuthal averaging also introduces some asymmetry and hence the value of @xmath28 in azimuthally averaged spectra is not the same as that used in preparing the 3d spectra . the effect of varying the flow velocity ( @xmath34 & @xmath35 ) and background term , @xmath24 is shown in fig [ fig5 ] . the synthetic spectra were constructed with a uniform amplitude @xmath47 , width @xmath48hz and included only the f - mode ridge . the background in synthetic spectra included only the first term with different values of @xmath24 . the effect of velocity increases with @xmath23 and hence the shift in asymmetry from input value also increases with @xmath23 . further , since @xmath30 and @xmath24 are constant , the peak to background ratio increases with @xmath23 . it is clear that even when the flow velocities are increased beyond their normal values , the maximum change in @xmath28 is about than 0.02 at the highest value of @xmath38 and rather large velocity is required to make significant differences . similar results are obtained when spectra with different values of @xmath28 were used . in all cases azimuthal averaging tends to increase the value of @xmath28 , though the effect may be small . peak asymmetry is believed to have a contribution from the acoustic source location and a contribution from the effects of correlated background noise from the source . @xcite have argued that the line asymmetries occur due to the modes correlated with velocity background while to reach the same effect , @xcite have proposed a small amount of intensity - correlated background . the amount of coherent correlated background , as well as the need of this component to model both @xmath0 and i@xmath1 , has also been debated ( e.g. @xcite ) . in order to understand the role of the background power in the reversal of the asymmetry between @xmath0 and i@xmath1 , we plot the background power as a function of degree @xmath38 of the mode in figure [ fig6 ] . in the same plot , we also show the background power obtained from the symmetric fits ( open symbols ) . it is clear that the background power obtained from symmetric and asymmetric fitting of the velocity spectra ( upper panel ) closely agree with each other . in contrast the intensity spectrum returns a significantly higher background power when fitted with the asymmetric line profile compared to the symmetric profile . this indicates that the line asymmetry is reversed in the intensity . since granulation contrast decreases from the disk center to the limb , we also expect the background power to decrease near the limb . but no significant variation in power is noticed from the disk center to the limb which supports the idea that only a part of the background is correlated with the intensity oscillations @xcite . the frequency differences from the symmetric fits to both @xmath0 and i@xmath1 spectra at four different longitudes are plotted in figure [ fig7 ] . the frequencies obtained from the intensity spectrum are systematically higher than the frequencies from the velocity spectrum , as one would expect due to the reversal of the line asymmetry . for the disk center spectra , the differences , in general , are on the order of 10 @xmath6hz or less up to @xmath49 mhz ; the minimum difference of about 23 @xmath6hz is seen for @xmath50 and higher radial order modes in the 5min band . since near the limb we primarily observe the horizontal velocity field which is dominated by granulation , it is expected that the frequency difference between intensity and velocity should decrease near the limb due to a decrease in correlation with oscillations . but we find that the difference increases systematically from the disk center to the limb . this again supports the idea that only a part of the background is correlated with the intensity oscillations . the use of asymmetric profiles to fit the power spectra results in a better determination of the mode frequencies and as a result the frequency differences between velocity and intensity is significantly diminished ( figure [ fig8 ] ) . similar to fits with symmetric profiles , the differences are mostly positive as the frequencies obtained from intensity fluctuations are higher than those from the velocity . however , the _ f_-modes show some exception at the high end of the frequency where the peak frequencies observed in velocity are marginally higher ( and clearly larger than the estimated errors ) than those observed in intensity . the differences between higher order radial modes , mainly @xmath51 and @xmath52 , are slightly bigger but these modes also have a large uncertainty in the fitted frequencies as seen from the error estimates . although subtle variations are evident between different longitudes , the frequency differences do not reveal any systematic trend between the disk center and 45 longitude . although the asymmetric fitting formula reduces the frequency differences between the velocity and intensity spectra , there still remains some systematic differences of the order of 510 @xmath6hz , which is much higher than the uncertainty in estimated errors . @xcite have reported a similar result based on the analysis of azimuthally averaged spectra and have argued that the assumed profiles does not adequately describe the velocity spectra where the signal - to - noise ratio is high . however , our result indicates that the form of asymmetry used for intensity spectra may not be a correct representation of the real profiles . to demonstrate this , we show the difference in eigenfrequencies between identical modes obtained from the power spectra at different longitudes with respect to the spectra at the disk center in figure [ fig9 ] . the frequency difference in @xmath0 shows little variation in longitude for frequencies less than 3.5 mhz . variations between intensity spectra are higher than seen in velocity . moreover , the shapes of the curve in figure [ fig9 ] for velocity are similar as are the shapes of the curves in intensity , leading to the frequency differences shown in figure [ fig8 ] . therefore it appears that the use of an asymmetric line profile provides more accurate estimates of the eigenfrequencies of solar oscillation for velocity rather than for intensity . in addition , fitting the intensity spectra is more challenging than fitting the velocity spectra due to low amplitudes and low signal - to - noise ratio . the number of modes fitted from the intensity spectra clearly supports this conjecture . using the local helioseismic technique of ring diagram analysis we have studied the high - degree _ f_- and _ p_-mode frequency differences between velocity and continuum intensity data obtained from the mdi instrument during a quiet period . since the spectra are known to be asymmetric , we fitted the three - dimensional spectra with an asymmetric model profile based on the known form of the asymmetry observed in velocity and intensity . the asymmetry obtained from fitting the velocity spectra agrees with the previous results while that of intensity shows a frequency dependent behavior . we further notice significant differences between the asymmetry obtained from fitting the three - dimensional and azimuthally averaged power spectra . fitting the three - dimensional disk center power spectra with a symmetric lorentzian profiles leads to frequency differences on the order of 10 @xmath6hz or less up to @xmath53 4 mhz between intensity and velocity . this difference is found to be increasing systematically from the disk center to the limb . the use of asymmetric profiles leads to frequency differences that are smaller than the differences resulting from the symmetric fits . however , systematic differences still remain at the high end of the frequency range , mostly above 4 mhz . we demonstrate that this difference arises from the fitting of the intensity rather than the velocity spectra . we speculate that the form of asymmetry used in the model for the intensity spectra is over - simplified and does not adequately represent the real profile at all frequency ranges . we also conclude that the frequency differences between velocity and intensity do not vary significantly from the disk center to the limb . however , variations between intensity spectra are higher than seen in velocity . this supports the idea that only a part of the background is correlated with the intensity oscillations . in this context several authors ( e.g. @xcite ) have advocated the importance of fitting the intensity - velocity cross spectrum along with a multicomponent model of the coherent background . this provides an impetus for future work using the high - resolution observations from _ solar dynamics observatory_. harvey , j. w. , hill , f. , komm , r. w. , leibacher , j. , pohl , b. , & the gong 1998 , in new eyes to see inside the sun and stars , eds . deubner , j. christensen - dalsgaard , & d. kurtz , ( iau publication , the netherlands ) , 49	using the local helioseismic technique of ring diagram we analyze the frequencies of high degree _ f_- and _ p_-modes derived from both velocity and continuum intensity data observed by mdi . fitting the spectra with asymmetric peak profiles , we find that the asymmetry associated with velocity line profiles is negative for all frequency ranges agreeing with previous observations while the asymmetry of the intensity profiles shows a complex and frequency dependent behavior . we also observe systematic frequency differences between intensity and velocity spectra at the high end of the frequency range , mostly above 4 mhz . we infer that this difference arises from the fitting of the intensity rather than the velocity spectra . we also show that the frequency differences between intensity and velocity do not vary significantly from the disk center to the limb when the spectra are fitted with the asymmetric profile and conclude that only a part of the background is correlated with the intensity oscillations .
connectivity augmentation is a classical problem in combinatorial optimization . given a graph @xmath0 and a parameter @xmath9 , add a set of new edges @xmath3 of minimum cardinality or weight such that the augmented graph @xmath10 is @xmath11-connected ( resp . , @xmath11-edge - connected ) . efficient algorithms are known for both connectivity and edge - connectivity augmentation over abstract graphs and constant @xmath11 @xcite . in this paper we consider weighted connectivity augmentation for planar straight - line graphs ( pslgs ) . the vertices are points in euclidean plane , the edges are noncrossing line segments between the corresponding vertices , and the weight of an edge is its euclidean length . the edge- and node - connectivity of a planar graph is at most 5 by euler s theorem . further , not every pslg can be augmented to a 3-connected ( resp . , 4-edge - connected ) pslg ; see @xcite for feasibility conditions . finding the minimum _ number _ of edges to augment a given pslg to @xmath11-connectivity or @xmath11-edge - connectivity is np - complete @xcite for @xmath12 ; the reduction requires the input graph @xmath5 to be disconnected ( the np - hardness claim for connected input ( * ? ? ? * corollary 2 ) turned out to be flawed ) . worst case bounds are known for the most important cases : every pslg @xmath5 with @xmath13 vertices can be augmented to 2-edge - connectivity with at most @xmath14 edges @xcite ; and at most @xmath15 new edges if @xmath5 is already connected @xcite . at most @xmath16 suffice for 2-connectivity , where @xmath17 is the number of 2-blocks in @xmath5 @xcite . all these bounds are the best possible . * our results . * we show that every connected pslg @xmath0 with @xmath13 vertices in general position can be augmented to a 2-connected pslg @xmath10 by adding new edges of total euclidean length @xmath2 ( section [ sec : augment ] ) . a set @xmath3 that minimizes @xmath18 can be computed in @xmath4 time ( section [ sec : dp ] ) ; however the problem becomes np - hard when @xmath5 is disconnected ( section [ sec : hardness ] ) . further , there is a sequence of edge insertions and deletions that transforms a connected pslg @xmath5 into a planar straight - line cycle @xmath6 such that the graph at all stages remains connected with the sum of euclidean edge lengths below @xmath8 , and , at termination , @xmath7 , where @xmath19 denotes the euclidean minimum spanning tree of @xmath20 ; these bounds are the best possible ( section [ sec : dynamic ] ) . our proof is constructive , and yields a polynomial - time algorithm for computing such sequences . * related previous work . * biconnectivity augmentation over _ planar _ graphs ( where no embedding of @xmath5 is given ) is also np - complete @xcite . over planar graphs _ with fixed combinatorial embedding _ , biconnectivity augmentation remains np - hard for disconnected graphs ; but there is a near - linear time algorithm when the input is already connected @xcite . frederickson and jaja @xcite show that the _ weighted _ augmentation of a tree to be 2-connected or 2-edge - connected ( without planarity constraint ) is np - complete even if the weights are restricted to @xmath21 . the problem is apx - hard , and breaking the approximation ratio of 2 is a major open problem @xcite . in the _ geometric _ setting , we show ( in section [ sec : dp ] ) that the minimum - weight augmentation of a planar straight - line tree to a 2-connected ( 2-edge - connected ) pslgs can be computed efficiently . the length of the edges in planar connectivity augmentation was studied only recently in the context of wireless networks . given a plsg @xmath22 where the vertices induce a 2-edge - connected unit disk graph , dobrev et al . @xcite compute a 2-edge - connected pslg by adding edges of length at most 2 . kranakis et al . @xcite studied the combined problem of adding the minimum _ number _ of edges of _ bounded length _ : a 2-edge - connected augmentation is possible such that @xmath23 is at most the number of bridges in @xmath5 and @xmath24 . however , finding the minimum number of new edges of bounded length is np - hard . let @xmath0 be a planar straight - line graph ( pslg ) , where @xmath20 is a set of @xmath13 points in the plane , no three of which are collinear , and @xmath25 is a set of open line segments between pairs of points in @xmath20 . the length of an edge @xmath26 , denoted @xmath27 , is the euclidean distance between @xmath28 and @xmath29 ; the total length of the edges is @xmath30 . denote by @xmath31 the set of faces of @xmath5 . the _ faces _ of @xmath5 are the connected components of the complement of all vertices and edges of @xmath5 , that is , @xmath32 . if @xmath5 is connected , then every bounded face is simply connected . a _ walk _ is an alternating sequence of points ( vertices ) and line segments ( edges ) whose consecutive elements are incident , and hence it is uniquely described by the sequence of its vertices @xmath33 . a walk is _ closed _ if @xmath34 . a walk is called a _ path _ if no vertex appears more than once . every face @xmath35 defines a closed walk @xmath36 that contain all edges on the boundary of @xmath31 , called _ facial walk _ , where every edge @xmath37 is incident to the face @xmath38 , and consecutive edges in the path , @xmath37 and @xmath39 , are also consecutive in the counterclockwise rotation of all edges incident to @xmath40 ( see fig . [ fig:2](a ) ) . note that every edge @xmath41 occurs twice in the facial walks of the faces of @xmath5 : a _ cut vertex _ ( resp . , _ bridge _ ) occurs twice in some facial walk . a walk @xmath42 is _ convex _ if @xmath43 for @xmath44 ( for closed walks , @xmath45 ) , where @xmath46 is the measure of the minimum counterclockwise angle that rotates the ray @xmath47 into @xmath48 . a vertex of @xmath5 is called _ convex _ if @xmath43 for each pair @xmath37 and @xmath39 of consecutive edges in the counterclockwise rotation of edges incident to @xmath40 ; and _ reflex _ otherwise . a convex walk @xmath49 is _ safe _ if not all of its vertices are collinear , and the vertices lie on the boundary of the convex hull of @xmath50 with the possible exception of the first or last vertex . let @xmath49 be a walk contained in the boundary walk of a face @xmath35 . the shortest walk from @xmath51 to @xmath52 homotopic to @xmath53 , denoted @xmath54 , is called the _ geodesic _ between @xmath51 and @xmath52 . for a walk @xmath53 , let @xmath55\rightarrow \mathbb{r}^2 $ ] be a piecewise linear arc from @xmath51 to @xmath52 that traverses the edges of @xmath53 in the given order . a _ homotopy _ between two walks , @xmath49 and @xmath56 , is a continuous function @xmath57\times[0,1]\rightarrow \mathbb{r}^2 $ ] such that @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 for all @xmath63 . intuitively , the walk @xmath53 can be continuously deformed into @xmath64 in the face @xmath38 . the walks @xmath53 and @xmath64 are _ homotopic _ if such a homotopy exists . ( 1,0.25453794 ) ( 0,0 ) with vertices in convex position . ( b ) a convex path @xmath65 , where @xmath66 . ( c ) a convex path @xmath67 with four edges , where @xmath54 contains edge @xmath68.,title="fig : " ] ( 0.28871011,0.04562865)(0,0)[lb ] ( 0,0 ) with vertices in convex position . ( b ) a convex path @xmath65 , where @xmath66 . ( c ) a convex path @xmath67 with four edges , where @xmath54 contains edge @xmath68.,title="fig : " ] ( 0.12182641,0.00579987)(0,0)[lb ] ( 0,0 ) with vertices in convex position . ( b ) a convex path @xmath65 , where @xmath66 . ( c ) a convex path @xmath67 with four edges , where @xmath54 contains edge @xmath68.,title="fig : " ] ( 0.48033978,0.00580222)(0,0)[lb ] ( 0,0 ) with vertices in convex position . ( b ) a convex path @xmath65 , where @xmath66 . ( c ) a convex path @xmath67 with four edges , where @xmath54 contains edge @xmath68.,title="fig : " ] ( 0.82863433,0.00580222)(0,0)[lb ] ( 0.2537617,0.23149408)(0,0)[lb ] ( 0.0668696,0.03822856)(0,0)[lb ] ( 0.05432268,0.13912942)(0,0)[lb ] ( 0.60850681,0.04221873)(0,0)[lb ] ( 0.51319131,0.24042385)(0,0)[lb ] ( 0.38191831,0.03798715)(0,0)[lb ] ( 0.43719944,0.15077211)(0,0)[lb ] ( 0.81877382,0.24583347)(0,0)[lb ] ( 0.68479637,-0.00034599)(0,0)[lb ] ( 0.96493903,0.04462578)(0,0)[lb ] ( 0.75089298,0.0516515)(0,0)[lb ] ( 0.8487234,0.13190323)(0,0)[lb ] given a walk @xmath49 and polygonal environment @xmath69 with @xmath13 vertices , the geodesic can be computed in @xmath70 time @xcite . it is known @xcite that all interior vertices of @xmath54 are reflex vertices of @xmath5 ; and if @xmath53 is a convex chain , then so is @xmath54 ( it may be a straight - line segment ) . geodesics play a crucial role in our worst - case bounds , since @xmath71 by definition , and the edges of @xmath54 do not cross any existing edges of @xmath5 . we show the following . [ lem:1 ] let @xmath0 be a pslg , and let @xmath72 be a safe convex walk contained in some facial walk of @xmath5 . then @xmath54 is a simple path that does not contain any vertices of @xmath53 except for @xmath51 and @xmath52 at its endpoints . let @xmath73 be the convex hull of @xmath74 . by assumption , the vertices of @xmath53 lie on @xmath75 with the possible exception of @xmath51 and @xmath52 , which may lie in the interior of @xmath73 . by construction , @xmath54 lies in @xmath76 with the possible exception of its endpoints @xmath51 and @xmath52 . hence , @xmath54 does not contain any interior vertex of @xmath53 . [ lem:1 + ] let @xmath0 be a pslg , and let @xmath72 be a convex walk contained in some facial walk of @xmath5 . if @xmath77 is a simple path that contains none of the vertices @xmath51 , @xmath78 , then @xmath54 is also a simple path that does not contain any of the vertices @xmath79 . by lemma [ lem:1 ] , @xmath80 is a simple cycle . let @xmath81 be the first edge of @xmath77 . since @xmath77 is homotopic to @xmath82 , the edge @xmath81 lies in the angular domain @xmath83 , and @xmath84 . consequently , the edge @xmath68 lies in the exterior of the cycle @xmath85 . note that @xmath86 . since @xmath77 is a convex chain , the interior of triangle @xmath87 lies in the exterior of @xmath85 . on the other hand , @xmath88 lies in @xmath87 , and so it is disjoint from the vertices @xmath89 . we conclude that @xmath54 is a simple path that does not contain any of the vertices @xmath79 . let @xmath5 be a pslg with no three collinear vertices . denote by @xmath90 the set of maximal convex walks contained in the facial walks of @xmath5 ( note that @xmath91 can be computed with a graph traversal in @xmath92 time ) . partition @xmath91 into three subsets : @xmath93 contains the convex walks that consist of a single edge ; @xmath94 contains the closed convex walks @xmath95 , i.e. , @xmath34 ; and @xmath96 contains all open convex walks of two or more edges . we define a _ dual graph _ @xmath97 where the nodes correspond to the convex walks in @xmath98 , and two nodes are adjacent if and only if the corresponding convex chains share an edge in @xmath25 . [ lem:2 ] let @xmath0 be a connected pslg with @xmath99 . then 1 . every edge in @xmath25 is part of a convex chain in @xmath98 , 2 . the dual graph @xmath97 is connected , \(a ) let @xmath100 . since @xmath5 has at least three vertices and is connected , we may assume that @xmath17 is incident to two or more edges . assume that @xmath101 and @xmath102 are the edges preceding and following @xmath103 in the clockwise rotation of edges around @xmath17 respectively ( possibly @xmath104 ) . then the boundary walks of the faces incident to @xmath17 contain the paths @xmath105 and @xmath106 . since @xmath17 is the apex of at most one reflex angle , we may assume that @xmath107 or @xmath108 is convex ; and so @xmath109 is part of a maximal convex walk of 2 or more edges . this walk is in either @xmath94 or @xmath96 . \(b ) for every vertex @xmath110 , every angle @xmath111 formed by consecutive incident edges @xmath26 and @xmath112 is convex with the possible exception of one reflex angle ( since the sum of these angles is @xmath113 . consequently , the walks @xmath114 where @xmath115 are part of distinct convex chains in @xmath98 that induce a path or a cycle in the dual graph . now consider two chains @xmath116 , and two arbitrary edges @xmath117 and @xmath118 from them . since @xmath5 is connected , there is a path @xmath119 such that @xmath120 and @xmath121 . since every two consecutive edges of @xmath64 are part of convex cycles in the same component of @xmath97 , the chains @xmath122 and @xmath123 are also part of the same component of @xmath97 . the proof of lemma [ lem:2 ] is provided in the appendix . when we modify a given pslg with edge insertion operations , we prove the following . [ thm:1 ] let @xmath124 be a connected pslg with @xmath125 and no three collinear vertices . then @xmath5 can be augmented to a 2-edge - connected pslg @xmath10 such that @xmath126 , and this bound is the best possible . we prove the upper bound constructively , augmenting a connected pslg @xmath0 incrementally into a 2-edge - connected pslg @xmath10 . then decompose every convex walk in @xmath91 of two or more edges into edge - disjoint convex walks of two or three edges . let @xmath127 be the set of all resulting convex paths of two or three edges ( that is , we discard convex walks that consist of a singe edge and convex cycles of 3 edges ) . see fig . [ fig:2](b ) for an illustration . ( 1,0.19059586 ) ( 0,0 ) is its boundary walks . ( b ) dashed lines indicate convex paths of two or three edges in @xmath127 ; dotted lines indicate maximal convex paths of a single edge , and triangles . ( c ) the 2-edge - connected pslg @xmath10 produced by our algorithm.,title="fig : " ] ( 0.1414397,0.00392041)(0,0)[lb ] ( 0.49326912,0.00392041)(0,0)[lb ] ( 0.84646753,0.00392041)(0,0)[lb ] ( 0,0 ) is its boundary walks . ( b ) dashed lines indicate convex paths of two or three edges in @xmath127 ; dotted lines indicate maximal convex paths of a single edge , and triangles . ( c ) the 2-edge - connected pslg @xmath10 produced by our algorithm.,title="fig : " ] for each convex path @xmath128 , augment @xmath5 with the edges of @xmath54 ( refer to fig . [ fig:2](c ) ) , and denote by @xmath129 the resulting graph . note that @xmath129 is 2-edge - connected since every edge in @xmath25 is part of a cycle by lemma [ lem:2](a ) : each cycle is either a triangle in @xmath5 , or a cycle @xmath130 for some @xmath131 ; and every edge in @xmath3 is part of a cycle by construction . next , we derive an upper bound for @xmath18 . every @xmath41 appears twice in the boundary walks of the faces of @xmath5 , and so it appears in at most two convex paths in @xmath132 . by definition , @xmath71 for every @xmath131 . overall , we have @xmath133 ( 1,0.18503062 ) ( 0,0 ) with three solid edges , where the augmentation to 2-edge - connectivity requires the addition of the dashed edges.,title="fig : " ] ( 1.00869647,0.10597043)(0,0)[lb ] ( 0,0 ) with three solid edges , where the augmentation to 2-edge - connectivity requires the addition of the dashed edges.,title="fig : " ] ( -0.04272756,-0.02722992)(0,0)[lb ] ( 0.46769892,-0.02384202)(0,0)[lb ] ( -0.0440314,0.22161412)(0,0)[lb ] ( 0.96983538,-0.0114394)(0,0)[lb ] ( 0.96983538,0.22090027)(0,0)[lb ] ( -0.04964884,0.09439476)(0,0)[lb ] we now show a matching lower bound . for every @xmath134 , let @xmath135 be defined on four vertices @xmath136 , @xmath137 , @xmath138 , and @xmath139 with edge set @xmath140 ; refer to figure [ fig:3 ] . since @xmath122 and @xmath141 are leaves in @xmath5 , and @xmath142 would cross @xmath143 , both @xmath144 and @xmath145 have to be added . we have @xmath146 , and the ratio @xmath147 is the best possible . we strengthen theorem [ thm:1 ] to vertex - connectivity . [ thm:2 ] let @xmath124 be a connected pslg with @xmath125 and no three collinear vertices . then @xmath5 can be augmented to a 2-connected pslg @xmath148 such that @xmath126 , and this bound is the best possible . we prove the upper bound constructively . consider the convex walk in @xmath98 , defined above . we augment @xmath5 into @xmath10 such that the vertex set of each convex walk in @xmath98 induces a 2-connected subgraph in @xmath129 ; and every new edge in @xmath3 is part of one of these subgraphs . note that this implies that @xmath129 is 2-connected : by lemma [ lem:2](a ) , every vertex is part of a 2-connected subgraph ; if two 2-connected subgraphs share two vertices , then their union is 2-connected . by lemma [ lem:2](b ) the union of the subgraphs induced by the convex chains is 2-connected . in the remainder of the proof , we consider a single convex walks @xmath149 . * case 1 : @xmath150 . * if the vertices @xmath151 are distinct , then @xmath53 is a cycle , and all vertices of the walk are part of a 2-connected component . otherwise @xmath152 . in this case , @xmath153 forms a convex polygon , whose interior contains @xmath34 but no other vertices . consequently , @xmath154 , and the only cut vertex along the walk is @xmath122 . add the edge @xmath155 , where @xmath156 by the triangle inequality ( fig . [ fig:2-con - upperbound](a ) ) . as a result , the vertices of @xmath53 induce a 2-connected subgraph . * case 2 : @xmath157 . * we decompose @xmath53 into edge - disjoint walks recursively as follows . if @xmath54 does not contain any interior vertex of @xmath53 , then we are done . otherwise , let @xmath73 be the convex hull of @xmath74 and let @xmath158 be the subchain along @xmath75 . by lemma [ lem:1 ] , the set of interior vertices of @xmath159 does not contain any vertex of @xmath160 . starting from @xmath160 , successively append the edges of @xmath53 proceeding @xmath40 or following @xmath161 while the path @xmath160 maintains the property that @xmath160 contains no interior vertices of @xmath159 . then recurse on any prefix or suffix path in @xmath162 . we obtain a decomposition of @xmath53 into subpaths @xmath160 such that @xmath163 is a simple cycle ( fig . [ fig:2-con - upperbound](b ) ) . by lemma [ lem:1 + ] , any two such consecutive paths share two vertices . consequently , the union of the cycles @xmath163 is a 2-connected graph . analogously to the proof of theorem [ thm:1 ] , we have @xmath2 , and the same lower bound construction shows that this bound is the best possible . ( 1,0.2037577 ) ( 0,0 ) ( 0.14552631,0.0058292)(0,0)[lb ] ( 0.72101208,0.0058292)(0,0)[lb ] let @xmath31 be a face of a pslg @xmath5 , and let @xmath164 , @xmath165 be the ( closed ) facial walk of @xmath31 . we define the graph @xmath166 , where @xmath167 is the set of vertices in @xmath168 and @xmath169 . given a walk @xmath170 contained in @xmath168 , a vertex @xmath171 is a _ cut vertex relative to @xmath172 _ if it appears more than once in @xmath172 . for every pair @xmath173 , we introduce a weight function , corresponding to the feasibility of an edge @xmath174 . let @xmath175 if the line segment @xmath174 does not cross any edge in @xmath176 , lies in the face @xmath31 and in the wedges @xmath46 and @xmath177 ; and let @xmath178 otherwise . note that two feasible edges , @xmath179 and @xmath180 , do not cross if their indices do not interleave ( i.e. , if @xmath181 ) . we present a dynamic programming algorithm @xmath182 that finds a set @xmath3 of edges of minimum total weight such that @xmath183 is a 2-connected pslg . we call an optimal solution @xmath184 . [ rem : separation ] every edge in @xmath184 lies in the face @xmath31 . indeed , suppose @xmath183 is a 2-connected pslg and @xmath185 is outside @xmath31 . then @xmath40 and @xmath161 are part of a simple cycle formed by some edges of @xmath186 , and @xmath187 is also 2-connected , showing that @xmath3 is not optimal . by remark [ rem : separation ] , each face of the input can be treated independently . if we insert an edge in face @xmath31 , then it decomposes @xmath31 into two faces @xmath188 and @xmath189 that can be considered independently . however , defining subproblems in terms of faces might generate an exponential number of subproblems . instead we define our subproblems in term of continuous intervals of the facial walk of @xmath31 . we characterize an optimal solution @xmath184 for @xmath186 in terms of local properties of the subproblems @xmath172 . let @xmath190 be a cut vertex with respect to a walk @xmath172 . the vertices between two consecutive occurrences of @xmath190 in @xmath172 are called _ descendants _ of @xmath190 . non - descendant _ of @xmath190 in @xmath191 is a vertex in @xmath168 that is neither a descendant in @xmath191 nor equal to @xmath190 . the _ @xmath192-th group _ of descendants is the set of vertices between the @xmath192-th and @xmath193-st occurrence of @xmath190 . a set @xmath194 of feasible edges _ satisfies _ a group if there is a cycle in the graph @xmath195 that contains a descendant in that group and a non - descendant of @xmath190 ; and @xmath194 satisfies a cut vertex @xmath190 if it satisfies all of its groups . if @xmath183 is a 2-connected pslg , then @xmath3 satisfies all cut vertex in @xmath196 . indeed , suppose @xmath3 does not satisfy a vertex @xmath190 , then the deletion of @xmath190 would disconnect one of its groups of descendants from the rest of @xmath186 , hence @xmath190 would be a cut vertex in @xmath183 . let @xmath197 $ ] , @xmath198 , be the minimum weight of an edge set @xmath194 that satisfies all groups of all cut vertices relative to @xmath172 and such that @xmath199 , @xmath200 and @xmath201 is a pslg . algorithm @xmath182 uses the following recursive relation to compute subproblems @xmath197 $ ] and returns @xmath202 $ ] . 1 . if @xmath172 does not contain any cut vertex relative to @xmath172 , then @xmath197=0 $ ] . 2 . if @xmath203 and @xmath204 , then @xmath197=\infty$ ] . 3 . if @xmath205 is not a cut vertex relative to @xmath172 , then @xmath197=\min\{c[s+1,t]$ ] , + and @xmath206+c[k , t]+f(s , k)\}\}$ ] . if @xmath205 is a cut vertex relative to @xmath172 , then let + @xmath207 . + set @xmath197=\min _ { ( p_i , p_j)\in x}\ { c[s , i]+c[i , j]+c[j , t]+f(i , j)\}$ ] . ( 1,0.22557258 ) ( 0,0 ) ( 0.09403438,0.16538865)(0,0)[lb ] ( 0.20893672,0.16538865)(0,0)[lb ] ( 0,0 ) ( 0.50768916,0.19887505)(0,0)[lb ] ( 0.43349221,0.05705054)(0,0)[lb ] ( 0,0 ) ( 0.45647476,0.13978147)(0,0)[lb ] ( 0.66884885,0.18311674)(0,0)[lb ] ( 0,0 ) ( 0.84649214,0.19887505)(0,0)[lb ] ( 0.77229519,0.05705054)(0,0)[lb ] ( 0,0 ) ( 0.8697669,0.08856708)(0,0)[lb ] ( 0,0 ) ( 0.95808963,0.0373527)(0,0)[lb ] ( 0.16095829,0.00207535)(0,0)[lb ] ( 0.5352174,0.00207535)(0,0)[lb ] ( 0.87008081,0.00207535)(0,0)[lb ] * correctness . * we show that @xmath182 correctly computes @xmath197 $ ] and that @xmath202 $ ] corresponds to an optimal edge set @xmath194 . the base case ( i ) is trivial . in case ( ii ) , @xmath203 is a cut vertex of @xmath168 , and all other vertices in @xmath208 $ ] are descendants of @xmath205 . therefore , there is no edge incident to a non - descendant of @xmath205 in @xmath172 , and @xmath205 can not be satisfied in @xmath172 ( fig . [ fig : dynamicprog](a ) ) . in case ( iii ) , since @xmath205 does not have descendants , the set @xmath194 corresponding to @xmath197 $ ] either has no edge incident to @xmath205 or it has an edge between @xmath205 and some descendant @xmath209 of a cut vertex @xmath190 in @xmath172 . in the former case , we have @xmath197=c[s+1,t]$ ] since @xmath194 satisfies all cut vertices relative to @xmath210 . in the latter , the edge @xmath211 creates a cycle satisfying the group of descendants containing @xmath212 for every cut vertex in @xmath213 . it divides @xmath31 into two faces : @xmath188 ( resp . , @xmath189 ) whose facial walk contains @xmath214 ( resp . , @xmath215 ) . every group that still needs to be satisfied is either in @xmath188 or @xmath189 . if @xmath216 ( resp . , @xmath217 ) is the set with minimum weight between vertices in @xmath214 ( resp . , @xmath215 ) that satisfies the groups of descendants in @xmath188 ( resp . , @xmath189 ) , then @xmath218 . in case ( iv ) , all non - descendants of @xmath205 in @xmath172 are in the set @xmath219 , where @xmath209 is the last occurrence of @xmath205 in @xmath172 . @xmath194 must contain at least one edge in @xmath220 in order to satisfy @xmath205 . let @xmath221 be an edge that minimizes @xmath222 , breaking ties by maximizing @xmath223 . then , every edge @xmath224 such that @xmath225 is between vertices in @xmath191 . in particular , there exist no edge in @xmath194 between a vertex in @xmath226 and @xmath227 . we can partition @xmath194 into @xmath216 , @xmath217 , and @xmath228 such that they each contain only edges between vertices in @xmath229 , @xmath230 , and @xmath231 respectively , each of them being the minimum - weight set that satisfies the groups of descendants in their respective subproblems . the edges in @xmath216 , @xmath217 , @xmath228 can not cross since they correspond to edge - disjoint walks in @xmath168 . thus , we have @xmath197=c[s , i']+c[i',j']+c[j',t']+f(i',j')$ ] . let @xmath194 be the edge set corresponding to @xmath202 $ ] . since @xmath194 satisfies all cut vertices in @xmath196 , which corresponds to the cut vertices of @xmath186 , the graph @xmath201 is 2-connected . since @xmath194 is a minimum - weight edge set that satisfies all groups of @xmath196 by definition , we have @xmath232 . * running time . * the feasible edge weights @xmath233 can be precomputed in @xmath234 time @xcite for all @xmath235 . there are @xmath234 subproblems , and each can be computed in @xmath234 time since it depends on @xmath234 smaller subproblems . hence algorithm @xmath182 takes @xmath236 time . [ thm : alg - pvca ] for a connected pslg @xmath237 , an edge set @xmath3 of minimum weight such that @xmath1 is 2-connected can be computed in @xmath4 time . we identify the faces of @xmath237 and run algorithm @xmath182 in all faces . by remark [ rem : separation ] , the union of the optimal solutions for all faces is the optimal solution for @xmath5 . since the size of the union of all facial walks is @xmath238 , algorithm @xmath182 takes @xmath4 time . [ thm : alg - peca ] for a connected pslg @xmath237 , an edge set @xmath3 of minimum weight such that @xmath1 is 2-edge - connected can be computed in @xmath4 time . we modify algorithm @xmath182 into algorithm @xmath239 , which computes the set @xmath3 . let an edge @xmath240 be a _ bridge relative to @xmath172 _ , @xmath241 , if it appears twice in @xmath172 ( recall that an edge can only appear up to two times in a facial walk ) . assume that @xmath240 and @xmath242 are the first and second occurrences . call @xmath243 the descendants of @xmath240 in @xmath172 , and all other vertices non - descendants . replace the recurrence relation as follows . 1 . if @xmath172 does not contain a bridge relative to @xmath172 , then @xmath197=0 $ ] . 2 . if @xmath244 is not a bridge relative to @xmath172 , then @xmath197=\min\{c[s+1,t],$ ] @xmath206+c[k , t]+f(s , k)\}\}$ ] . 3 . if @xmath244 is a bridge relative to @xmath172 , then let @xmath245descendants of @xmath244 in @xmath246 @xmath244 in @xmath247 . set @xmath197=\min_{(p_i , p_j)\in x}\ { c[s , i]+c[i , j]+c[j , t]+f(i , j)\}$ ] . the proof of correctness and runtime analysis are similar to algorithm @xmath182 . theorems [ thm : alg - pvca ] and [ thm : alg - peca ] extend to any nonnegative weight function . in particular , a minimum cardinality edge set @xmath3 can also be computed in @xmath4 time . [ thm : hardness ] given a ( disconnected ) pslg @xmath0 and a positive integer @xmath192 , deciding whether there exists an edge set @xmath3 such that @xmath248 and @xmath1 is a 2-edge - connected pslg is np - hard . we reduce from planar - monotone-3sat , which is np - complete @xcite . an instance of this problem is given by a plane bipartite graph between @xmath13 _ variables _ and @xmath249 _ clauses _ such that the variables are embedded on the @xmath250-axis , no edge crosses the @xmath250-axis and every clause has degree 2 or 3 . a clause is called _ positive _ if it is embedded on the upper half - plane and _ negative _ otherwise . planar - monotone-3sat asks if there is an assignment from @xmath251 to variables such that each positive ( resp . , negative ) clause is adjacent to at least one ` true ` ( resp . , ` false ` ) variable . given such an instance we build a pslg @xmath0 as follows . we divide the reduction into variable , wire and clause gadgets . ( 1,0.67445368 ) ( 0,0 ) ( 0.15789859,0.66727207)(0,0)[lb ] ( 0.7739534,0.396809)(0,0)[lb ] ( 0,0 ) ( 0.29932747,0.3118339)(0,0)[lb ] ( 0.87515493,0.3118339)(0,0)[lb ] ( 0.29097977,0.00297177)(0,0)[lb ] ( 0.77164448,0.00297177)(0,0)[lb ] ( 0,0 ) ( 0.88375018,0.4835537)(0,0)[lb ] ( 0,0 ) ( -0.00070631,0.48362432)(0,0)[lb ] ( 0,0 ) ( 0.6339468,0.48362432)(0,0)[lb ] ( 0,0 ) ( 0.82544268,0.15465722)(0,0)[lb ] * variable gadget . * a gadget of a variable adjacent to two positive and one negative clause is shown in fig . [ fig : reduction](a ) . the gray boxes and small disks represent 2-connected components and leaves respectively . each leaf requires at least one edge for the augmentation and the closest node from each of them is 1 unit apart . a pair of leaves can possibly share an edge and the @xmath222-th gadget for each @xmath252 has an even number of leaves @xmath253 , hence the gadget requires at least @xmath254 length . there are exactly two possible ways to achieve this bound and they encode the ` true`/`false ` value of the variable . [ fig : reduction](a ) can be generalized to other variables by repeating , omitting and changing the length and number of leaves of the component highlighted in the figure . * wire gadget . * the wire gadget that connects the @xmath222-th variable to the @xmath255-th clause , denoted as the @xmath256-wire , share the edge @xmath257 with the @xmath222-th variable gadget and its first leaf is called @xmath258 ( fig . [ fig : reduction](b ) ) . turns , if needed , can be done as in fig . [ fig : reduction](c ) . * clause gadget . * the @xmath255-th clause gadget contains a special leaf called @xmath259 shown in red in fig . [ fig : reduction](d ) that is located at an odd distance @xmath260 from at most three leaves , say @xmath258 , @xmath261 , and @xmath262 . if the clause is incident to two variables , use the turn shown in fig . [ fig : reduction](c ) as a clause gadget , naming the upper left leaf @xmath259 . if the @xmath255-th clause is positive , we place all @xmath258 at an even distance from each other , which makes the position of @xmath259 always realizable . use reflections through the @xmath250-axis for negative clauses . let @xmath263 denote the number of leaves of the @xmath256-wire . we set @xmath264 assume that the planar - monotone-3sat instance have a satisfying assignment . we build @xmath3 as follows . for each variable assigned ` false ` ( resp . , ` true ` ) add the edges shown as dashed ( resp . , dotted ) lines of fig . [ fig : reduction](a ) to @xmath3 . for each @xmath256-wire add to @xmath3 all dashed lines shown in fig . [ fig : reduction](b ) , ( c ) and the closest dashed line in fig . [ fig : reduction](d ) from the @xmath256-wire if the @xmath222-th variable is assigned ` false ` and the @xmath255-th clause is positive ; or if the @xmath222-th variable is assigned ` true ` and the @xmath255-th clause is negative . add the dotted lines otherwise . we obtain @xmath3 such that @xmath265 and the graph @xmath1 is 2-edge - connected . assume that there exists @xmath3 such that @xmath265 and the graph @xmath1 is 2-edge - connected . we call _ @xmath256-leaves _ the set of all leaves in the @xmath256-wire and the two leaves in the clause gadget adjacent to it . since the number of @xmath256-leaves is odd and the closest point from any such leaf is at least 1 unit away , the minimum length required to have at least an edge in @xmath3 incident to each leaf is @xmath266 . since @xmath192 is the sum of all such lower bounds , then : ( i ) the subset of @xmath3 that is incident to a @xmath256-leaf must have exactly @xmath266 unit length edges ; and ( ii ) the subset of @xmath3 that is incident to the @xmath222-th variable gadget must be either the set of dashed or dotted lines in fig . [ fig : reduction](a ) . assume that @xmath3 contains the dotted lines in fig . [ fig : reduction](a ) in the @xmath222-th variable gadget . then , for a positive clause @xmath255 , @xmath3 must contain an edge between @xmath258 and a point in the maximal 2-connected component of @xmath5 that contains @xmath257 , or else the bridge in @xmath5 that connects such 2-connected component will remain a bridge in @xmath1 . therefore , all other @xmath256-leaves must be matched in order to satisfy ( i ) . since @xmath1 is 2-edge - connected , @xmath259 is connected to some other leaf . then , if it is connected to a @xmath256-leaf , the @xmath222-th variable gadget uses the dotted edges ( ` true ` ) . by applying the symmetric argument for negative clauses , all such clauses must be incident to a variable gadget using dashed edges ( ` false ` ) . then we have also a satisfying assignment for the planar - monotone-3sat instance . given a ( disconnected ) pslg @xmath0 and @xmath267 , finding a set @xmath3 such that @xmath248 and @xmath1 is a 2-connected pslg is np - hard . the same reduction used in the proof of theorem [ thm : hardness ] also works for 2-connectivity . notice that the length required by leaves is the same and any @xmath3 , @xmath248 , that augments @xmath5 to 2-edge - connected also achieves 2-connectivity . [ thm : dynamic ] let @xmath124 be a connected pslg with @xmath125 and no three collinear vertices . then there exists a sequence of edge insertion and deletion operations that transforms @xmath5 into a planar straight - line cycle @xmath268 such that @xmath7 and that every intermediate graph is a connected planar straight - line graph of weight at most @xmath8 . these bounds are the best possible . [ [ phase2-from - an - arbitrary - tree - to - a - tree - in - the - delaunay - triangulation . ] ] phase 2 : from an arbitrary tree to a tree in the delaunay triangulation . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + lawson @xcite showed that every triangulation of a point set @xmath20 can be transformed into the delaunay triangulation of @xmath20 , denoted @xmath274 , through a sequence of edge flips . if an edge @xmath109 in a triangulation is adjacent to two triangles , @xmath275 and @xmath276 , where @xmath277 is in convex position , then an _ edge flip _ replaces edge @xmath109 by a new edge @xmath278 , and produces a new triangulation on @xmath20 . lawson @xcite also showed that there is a sequence of so - called _ delaunay flips _ , where point @xmath279 lies in the interior of the circumcircle of @xmath280 . the length of the flip sequence is @xmath234 ; this bound is the best possible @xcite , but finding the minimum length of flip - sequence between two triangulations is apx - hard @xcite . let @xmath281 be a sequence of triangulations on @xmath20 , where @xmath282 and @xmath283 is obtained from @xmath284 by an edge flip described above for @xmath285 . we now describe how to construct a sequence of spanning trees @xmath286 , where @xmath287 is a spanning tree in the triangulation @xmath283 ; and @xmath287 is obtained from @xmath288 by an edge insertion followed by an edge deletion ( such that the intermediate graph is also a connected pslg ) . suppose that @xmath283 is obtained from @xmath284 by an edge flip that replaces @xmath109 by @xmath278 . we distinguish two cases : assume that @xmath293 . since edge @xmath109 violates the delaunay condition , point @xmath279 lies in the interior of the circumcircle of @xmath280 . consequently , @xmath294 ( equality would imply that @xmath295 , @xmath17 , @xmath296 , and @xmath279 are cocircular ) . without loss of generality , assume @xmath297 , that is @xmath298 is an obtuse triangle . by the sine theorem , @xmath109 is the longest side of @xmath298 , and so @xmath299 . the deletion of edge @xmath109 disconnects the spanning tree @xmath288 into two trees in which @xmath295 and @xmath17 are in different components . vertex @xmath296 is in one of the two components . if @xmath296 and @xmath295 are in the same component , then @xmath296 and @xmath17 are in different components and @xmath300 is a spanning tree . otherwise @xmath296 and @xmath295 are in different components , and @xmath301 is a spanning tree . since the edge @xmath109 was replaced by a shorter edge , @xmath302 or @xmath303 , we have @xmath304 . note also that the intermediate graph , @xmath305 or @xmath305 , is connected and its length is bounded above by @xmath306 . [ [ phase3-from - an - arbitrary - tree - in - the - delaunay - triangulation - to - rm - mstv . ] ] phase 3 : from an arbitrary tree in the delaunay triangulation to @xmath19 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to transform @xmath307 into the @xmath19 , we add the @xmath19 edges to @xmath307 one at a time as described below . it is well known that @xmath274 contains the euclidean spanning tree @xmath19 of @xmath20 as a subgraph . let @xmath308 be an edge in @xmath19 that is not in @xmath307 . add @xmath308 to @xmath307 creating a connected graph @xmath309 whose total weight is at most @xmath310 . since @xmath307 is a spanning tree , there exist exactly one cycle in @xmath309 and it contains @xmath308 . delete a longest edge of such cycle which results in a connected graph @xmath311 that weighs at most @xmath312 . by repeating this procedure at most @xmath313 times , we obtain @xmath314 . given a graph @xmath315 , pick an arbitrary edge @xmath26 of the convex hull of @xmath20 that is not present in @xmath316 . let @xmath53 be the unique path between @xmath28 and @xmath29 in @xmath316 , where @xmath317 by the triangle inequality . augment @xmath316 with the edge @xmath26 into a pslg @xmath272 , and let @xmath318 be the planar straight - line cycle formed by @xmath26 and @xmath319 . we apply a sequence of edge insertion and deletions to @xmath272 . in each step @xmath222 , we maintain a pslg @xmath320 , a weakly simple polygon @xmath321 whose edges are contained in @xmath287 , and an ordering among the multi - edges of @xmath322 ( any multi - edge of @xmath322 is present as a single edge in @xmath287 ) , such that all edges induced by @xmath323 are in @xmath324 and @xmath325 ( where the weight of any multi - edges of @xmath322 are counted with multiplicity , all other edges in @xmath326 are counted once ) . the vertex set @xmath323 will grow monotonically until @xmath327 . given @xmath287 and @xmath322 , we construct @xmath311 and @xmath328 as follows . assume that @xmath329 . since @xmath287 is connected , there is a vertex @xmath330 in @xmath322 adjacent to some vertex outside of @xmath322 . in the counterclockwise order of edges incident to @xmath330 there must exist at least two transitions between edges to vertices in @xmath323 and not in @xmath323 , that is , a pair of consecutive edges @xmath331 and @xmath332 such that @xmath333 and @xmath334,i.e . , @xmath335 and @xmath336 . since no three vertices are collinear , at least one of such pairs forms a convex walk @xmath337 . without loss of generality , assume that @xmath338 and @xmath339 . an example is shown in fig . [ fig : phase4 - 5](left ) . construct @xmath328 from @xmath322 by replacing edge @xmath340 with the path @xmath341 . similarly , we construct @xmath311 from @xmath287 by adding the geodesic path @xmath342 ( if an edge of @xmath342 is already present in @xmath287 , we increment its multiplicity by one ) , and then deleting ( one copy of ) the edge @xmath340 , and any other edges that are induced by @xmath343 but not present in @xmath344 . by the triangle inequality we have @xmath345 . we then have @xmath346 , since we added @xmath342 and removed ( one copy of ) @xmath340 . each step strictly increased the number of vertices in @xmath323 . consequently , phase 4 executes at most @xmath347 times and the resulting graph @xmath348 contains the weakly simple polygon @xmath349 @xmath350 . [ [ phase5-from - a - weakly - simple - polygon - c - to - a - simple - polygon - p . ] ] phase 5 : from a weakly simple polygon @xmath85 to a simple polygon @xmath319 . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + let @xmath351 be a weakly simple polygon on vertex set @xmath352 and edge multiset @xmath353 . ( since no three vertices are collinear , no vertex lies in the relative interior of an edge ) . we construct a sequence of edge insertions and deletions that transform @xmath351 into a simple polygon on @xmath20 , while monotonically decreasing the total edge length ( counted with multiplicity ) of intermediate weakly simple polygons @xmath322 until @xmath322 becomes hamiltonian . this monotonicity ensures that the algorithm terminates . we describe a generic step of the algorithm , where @xmath354 , @xmath355 , is a weakly simple polygon with vertex set @xmath327 . if every vertex has multiplicity one , then @xmath322 is a simple polygon , and our algorithm is complete . otherwise , let @xmath110 be a vertex of multiplicity two or higher . see fig . [ fig : phase4 - 5 ] for examples . since no three vertices are collinear , any two consecutive edges of @xmath322 define a convex and a concave angle @xmath356 and @xmath357 . let @xmath223 , @xmath358 , be an index such that @xmath359 and @xmath360 is minimal . construct @xmath328 from @xmath322 by replacing the walk @xmath361 with @xmath362 . the minimality of the angle guarantees that @xmath328 is a weakly simply polygon , and the triangle inequality yields @xmath363 , as required . we now show our lower bounds , proving that the upper bound @xmath364 on the length of the output and @xmath8 on the length of all intermediate graphs are the best possible . for every @xmath134 , let @xmath365 be a graph with 4 vertices @xmath136 , @xmath137 , @xmath138 , and @xmath139 , and edge set @xmath140 ; depicted in figure [ fig:3 ] . we have @xmath366 . an @xmath367 of these four points is the path @xmath368 , with @xmath369 . the graph @xmath135 is a path , which is not 2-connected . no edge of @xmath135 can be deleted without disconnecting the graph , and only @xmath144 or @xmath145 can be inserted without introducing crossings , each of which has length 1 . hence the length of the 2nd graph in the sequence leading to 2-connectivity is @xmath370 , which tends to @xmath8 when @xmath371 goes to 0 . every 2-connected graph @xmath6 contains at least two edges between @xmath372 and @xmath373 , and the length of any edge between these vertex sets is at least 1 . if @xmath129 contains exactly two edges between @xmath372 and @xmath373 , then it must contain the edges @xmath374 and @xmath375 . consequently , every 2-connected graph @xmath6 satisfies @xmath376 , and this bound is attained for the cycle @xmath368 . we have @xmath377 , and the ratio @xmath378 is the best possible . s. dobrev , e. kranakis , d. krizanc , o. morales ponce , and l. stacho , approximating the edge length of 2-edge connected planar geometric graphs on a set of points . in _ proc . 10th latin _ , lncs 7256 , springer , 2012 , pp .	we consider edge insertion and deletion operations that increase the connectivity of a given planar straight - line graph ( pslg ) , while minimizing the total edge length of the output . we show that every connected pslg @xmath0 in general position can be augmented to a 2-connected pslg @xmath1 by adding new edges of total euclidean length @xmath2 , and this bound is the best possible . an optimal edge set @xmath3 can be computed in @xmath4 time ; however the problem becomes np - hard when @xmath5 is disconnected . further , there is a sequence of edge insertions and deletions that transforms a connected pslg @xmath0 into a planar straight - line cycle @xmath6 such that @xmath7 , and the graph remains connected with edge length below @xmath8 at all stages . these bounds are the best possible .
the authors are very much grateful to v. i. anisimov for helpful advise on the lsda+u method and for a fruitful discussion on the spin alignment of ndnio@xmath0 , and also to t. mizokawa for useful discussion about their work @xcite . this work is supported by grant - in - aid for coe research `` spin - charge - photon '' .	electronic structures of ndnio@xmath0 and ynio@xmath0 are calculated by using lsda+u method with rotational invariance . the jahn - teller distortion is not allowed under the observed magnetic ordering . no orbital order on ni sites can not be observed in the calculation of both systems but different types of charge ordering . in a small distorting system ndnio@xmath0 , all ni ions are trivalent and oxygen sites have a particular ordering , both in charge and orbital . in a large distorting system ynio@xmath0 , the charge disproportionation occurs , 2ni@xmath1 ni@xmath2ni@xmath3 . therefore , the charge ordering stabilizes the asymmetry of the arrangement of ni magnetic moments in both systems . 2 perovskite transition metal oxides are now of high interest because of the large variety and the possible controllability of their physical properties . perovskite nickelates rnio@xmath0 ( r = an trivalent rare earth or y ions ) can be classified into three different categories , according to their tolerance factor ( @xmath4 ) . @xcite one is those whose tolerance factor is much smaller ( @xmath5 ) and has a larger distortion , such as lunio@xmath0 or ynio@xmath0 . they are antiferromagnetic insulators at low temperatures . this class of rnio@xmath0 undergoes transition to paramagnetic insulator at the neel temperature @xmath6 and also has another phase transition from paramagnetic insulator to paramagnetic metal ( m - i transition ) at very high transition temperature . second is those whose tolerance factor is intermediate ( @xmath7 ) , such as ndnio@xmath0 or prnio@xmath0 . these materials are also antiferromagnetic insulators at low temperatures and above @xmath6 they are paramagnetic metals . @xcite third is lanio@xmath0 where @xmath8 and is paramagnetic metal . low temperature phase of above mentioned first and second classes of rnio@xmath0 have the unique magnetic structures . @xcite its magnetic diffraction peak is characterized by the propagation vector @xmath9 and the magnetic unit cell is identified as the @xmath10 supercell of the crystallographic lattice . @xcite the magnetic order is specified to be alternating ferromagnetic ( fm ) and antiferromagnetic ( afm ) couplings @xmath11 along three crystallographic directions . on the contrary to the asymmetric arrangement of magnetic bonds ( fm / afm ) around each nickel site , ni - o distances around each ni ion are almost equal to each other . for instance , the difference between the longest and the shortest ni - o bonds in one nio@xmath12 octahedron is at most 2% in rnio@xmath0 . @xcite the jahn - teller distortion is absent in rnio@xmath0 , even in the largely distorted system ynio@xmath0 , whose tolerance factor is 0.88 . experimentally there is only one crystallographic site for ni ions in ndnio@xmath0 and two different sites for ni ions in ynio@xmath0 . therefore , one can expect trivalent ions ni@xmath13 in ndnio@xmath0 . on the contrary , ni ions in ynio@xmath0 are divalent and quadrivalent , _ i.e. _ the charge disproportionation 2ni@xmath1 ni@xmath2ni@xmath3 occurs . in this letter , we study the electronic properties of the afm insulating phase of two perovskite nickelates , ndnio@xmath0 and ynio@xmath0 , which are typical ones of above mentioned two respective classes . we use the lsda+u method with the rotational invariance in conjunction with the lmto - asa method . @xcite the lsda+u method counts the electron - electron interaction between localized orbitals by the hartree - fock type interaction term . the ionic positions and lattice parameters used in the present calculations are imported from the diffraction experiments . @xcite the unit cell here is the @xmath14 crystallographic supercell containing 16 nd or y ions , 16 ni ions , and 48 o ions and , in addition , 32 empty spheres , totally 112 atomic spheres . sixty - four k - points in the brillouin zone are sampled in the calculation of the density of states . ndnio@xmath0 is antiferromagnetic up to @xmath15 , but the nd spin moment vanishes at about 30k . @xcite therefore , three 4f electrons of nd ion are counted in the frozen core . we also calculated the electronic structure of ndnio@xmath0 with 4f electrons in valence states , and the results show no significant difference . the coulomb and exchange parameters @xmath16 and @xmath17 of ni ions are fixed to be 7.0 ev and 0.88 ev , respectively , through all the lsda+u calculations . these values are consistent with photoemission experiments and the results of the lsda calculations . calculated results are not changed much over a large range of values of @xmath16 and @xmath17 . the crystallographic space groups of ndnio@xmath0 and ynio@xmath0 are orthorhombic @xmath18 and monoclinic @xmath19 , respectively . the group theoretical analysis shows that , using the projection operator method , the jahn - teller distortion of one nio@xmath12 octahedron can not be transfered over the whole enlarged @xmath20 supercell with @xmath21 . @xcite therefore , the jahn - teller distortion actually can not appear in the observed spin ordered state . ndnio@xmath0 : there are two possible spin configurations in @xmath22 satisfying the observed transfer vectors @xmath21 . one is that the spin magnetic moments align ferromagnetically on the ( 101 ) plane and the planes are stacked with a doubled period as @xmath23 , whose magnetic space group is monoclinic @xmath24 . @xcite this spin configuration is assumed in the present letter . another possible spin configuration with @xmath25 spin order could be the doubled period checkerboard stacking along crystallographic @xmath26-axis , and the magnetic space group is orthorhombic @xmath27 . @xcite the state of @xmath27 shows orbital ordering of e@xmath28 states on ni sites , but the resultant calculated total energy is higher by 0.18 ev per @xmath29 cell than that of @xmath24 . because there is no possibility of subsidiary jahn - teller distortion to lower the total energy , @xmath27 could not be the symmetry to be considered . the total energy is also calculated in a fictitious antiferromagnetic state of @xmath30 whose magnetic unit cell is identical to the crystallographic one and is higher by 2.0 ev per @xmath31 cell . figure [ fig - ndnio3-pdos ] shows the projected density of states on the ni 3d orbitals in ndnio@xmath0 . the system is insulator with a gap @xmath32ev . each nickel ion has the local magnetic moment @xmath33 within the atomic sphere of a radius @xmath34 , contrary to the observed value of @xmath35 . @xcite it should be noticed that the local magnetic moment can not be uniquely defined . furthermore , ndnio@xmath0 is not a simple antiferromagnetic insulator but a dynamical effect is essential , which may be an origin of this discrepancy . @xcite there is no distinctive variation in partial spin density of states of ni site in each magnetic sublattice . an e@xmath28 band lies on the energy range @xmath36 . the numbers of states in the energy ranges @xmath37 and @xmath38 are 16 , respectively , which are the one occupied and one vacant e@xmath28 states per ni ion . actually , these e@xmath28 orbitals extend over surrounding oxygen sites from ni ions due to strong hybridization between ni e@xmath28 and o p orbitals . the extended occupied e@xmath28 orbital has a 60% weight on p orbitals on surrounding six oxygens , a 10% weight on an individual oxygen . therefore , one would establish a model where one occupied e@xmath28 state with majority spin locates at the top of the valence bands , and it hybridizes strongly with the p states on nearby o ions . this is the molecular orbital @xmath39 state . @xcite then one can assign all nickel ions in ndnio@xmath0 to be trivalent ni@xmath13 ( t@xmath40e@xmath41 ) , even though the ni ion is not truly ionized by + 3 charge . in the projected density of states of ni ion site , one observes a large amount of e@xmath28 states at the bottom of the d bands , which are the bonding states between ni d and o p , corresponding to the @xmath42 states in the molecular orbital picture . @xcite the e@xmath28 band in the range @xmath43 does not show any orbital ordering . in fact , off - diagonal elements within the e@xmath28 subblock of the occupation matrix @xmath44 are zero and the diagonal elements are identical . the absence of the orbital ordering may be consistent with the fact that the @xmath45 $ ] axis has three - fold rotational symmetry in the present spin configuration , once the distortion is neglected . due to this pseudo three - fold rotational symmetry , the basis orbitals of e@xmath28 representation of the trigonal group d@xmath46 is a good basis set and those derived from e@xmath28 orbitals are @xmath47 and @xmath48 . therefore , there is no difference between the occupancies of @xmath49 and @xmath50 . figures [ fig - ndnio3-sd ] is the spatial profiles of spin densities @xmath51 in the energy range @xmath52 ( @xmath39 state ) . only one spin component of o p orbitals on the fm bond is bridging between two ni e@xmath28 orbitals , while both spin components on the afm bond couple with ni e@xmath28 orbitals of respective spins . consequently , oxygen ions on the afm bond have more charge in this energy range than oxygen on the fm bond . this is the realization of oxygen - site charge - ordered state discussed by mizokawa _ et al . _ in the framework of the hartree - fock calculation . @xcite however , the charge difference between oxygen sites of ni@xmath53-@xmath54-ni@xmath53 and ni@xmath53-@xmath55-ni@xmath56 in @xmath39 states is mostly compensated by hybridized @xmath42 state at lower energies . besides , all oxygen ions have no local magnetic moment . more significant results seen in fig . [ fig - ndnio3-sd ] may be the p orbital ordering on oxygen sites . the magnetic space group is @xmath24 and its unitary part is @xmath57 . the unit cell and lattice primitive vectors are not identical to those in @xmath18 . the spin density in fig . [ fig - ndnio3-sd ] is consistent with this magnetic group @xmath24 , neither higher nor lower than this . the symmetry lowering of the unitary part @xmath57 is the origin of the opening the band gap at @xmath58 in the majority spin band , despite to the absence of the orbital ordering on ni sites ( _ i.e. _ symmetry driven band gap ) . therefore , one can conclude that the origin of the insulator phase at low temperatures in ndnio@xmath0 is the characteristic spin density on the oxygen sites or , equivalently , the orbital ordering there . the energy gap @xmath59 ev is due to the symmetry of charge order . in fictitious ideal cubic structure without distortion or tilting of nio@xmath12 octahedra , the system becomes metal whose valence and conduction bands touch at points with each other . @xcite the structure of @xmath39 bands is insensitive to the value of @xmath16 . the value of the band gap is unchanged down to @xmath60 ev . this is because the gap is driven by the symmetry . ynio@xmath0 : one should expect larger distortion in ynio@xmath0 because ynio@xmath0 is the typical system with the small tolerance factor . @xcite there are two different crystallographic ni sites , and the distances from each ni ion to surrounding o ion are different by 3@xmath614% from one type of ni ion to the other type . @xcite calculated self - consistent solution is that with the apparent charge disproportionation and no orbital polarization . since ni@xmath3 site has no spin magnetic moment , the two spin configurations discussed in ndnio@xmath0 become identical and the spin configuration of ynio@xmath0 is uniquely determined . the magnetic space group is @xmath62 if the spins of ni@xmath3 ions are non vanishing and , on the contrary , @xmath63 if the spins of ni@xmath3 ions are zero . the latter symmetry @xmath63 is actually the case . figure [ fig - ynio3-pdos ] shows the projected density of states at ni ion sites . the system is insulator with a gap @xmath64ev . the resultant magnetic moments for half of ni ions are @xmath65 within the atomic sphere of a radius @xmath66 , namely divalent ions ni@xmath67 ( t@xmath40e@xmath68 ) , of larger nio@xmath12 octahedron and zero for another half of ni ions within the atomic sphere of a radius @xmath34 , namely quadrivalent ni@xmath3 ( t@xmath40 ) , of smaller octahedron . the experimentally observed magnetic moments are @xmath69 for ni@xmath67 ions and @xmath70 for ni@xmath3 ions . @xcite the discrepancy may be due to a possible non - collinear spin order . the number of states in the energy range @xmath71 is 16 and dominant weight on ni@xmath67 . because the number of ni@xmath67 in the @xmath14 cell is 8 , these states are assigned as two e@xmath28 states mainly on ni@xmath67 and surrounding oxygens . the e@xmath28 orbitals of ni@xmath3 is lifted in the higher energy range ( @xmath72 ) without spin polarization . a large amount of e@xmath28 orbitals in ni@xmath67 ions locates at the bottom of the d bands , in the range @xmath73 for ni@xmath67 , and in the range @xmath74 for ni@xmath3 , stabilizes the bonding between ni@xmath3 ions and oxygen ions . from these facts , one can establish a model that deep @xmath42 molecular orbitals ( one per both ni@xmath67 and ni@xmath3 ) stabilize the system , and other two e@xmath28 states ( @xmath39 states ) per one ni@xmath67 ion located at @xmath75 . the charge disproportionation is mainly due to the crystal field effect . the low spin state in ni@xmath3 or ni@xmath13 ion is energetically unstable in small @xmath76 case and the ground state multiplet of ni@xmath67 is @xmath77a@xmath78 ( t@xmath40e@xmath68 ) for all arbitrary values of @xmath76 . @xcite therefore , the small tolerance factor causes two different ni sites , compressed ni@xmath3 ( large @xmath76 ) and dilated ni@xmath67 ( small @xmath76 ) , rather than uniformly dilated ni@xmath13 ionic states . @xcite the standard values of @xmath79 ( @xmath80 is the racah parameter ) for ni@xmath67 is presumably around 1.0 . once one estimate the crystal field effects from the ni - o bond lengths @xmath81 , the difference of @xmath82 on two ni ion sites is presumably about 20% . two narrow e@xmath28 bands can be seen in the energy range @xmath83 and @xmath84 in fig . [ fig - ynio3-pdos]a . two ni@xmath67 e@xmath28 states are spatially extending over wide area and not only hybridizing with the nearest neighbor o ions but also extending over the nearest @xmath85 ions . this situation is well depicted in the spin density . figure [ fig - ynio3-sd ] shows the isometric surfaces of the spin density in the range of @xmath83 . the d - wavefunctions on ni@xmath67 are extending over the ( 101 ) plane , and antiferromagnetically coupled with other ni@xmath67 ions . charge in each ni ion is compensated in ynio@xmath0 as in ndnio@xmath0 and the difference between total charges in the muffin tin spheres on ni@xmath67 and ni@xmath3 sites is very small , equals to 0.03 . this variation is the same order as that of oxygen ions in ndnio@xmath0 . however , charge disproportionation is coupled with the lattice distortion , where larger oxygen octahedron is surrounding ni@xmath67 , and stabilizes the lattice system in ynio@xmath0 . therefore , diffraction experiment can detect the charge ordering in the yttrium system easier than in the neodymium system . in conclusion , we have studied two typical antiferromagnetic insulating phase of rnio@xmath0 , ndnio@xmath0 and ynio@xmath0 , by using the lsda+u method . a possibility of the jahn - teller distortion is excluded by the group theoretical consideration . no orbital order on ni sites is observed in both systems , but two different types of ordering are observed . in small distorting rnio@xmath0 such as ndnio@xmath0 , oxygen sites shows the orbital ordering and this is the origin of the gap opening in ndnio@xmath0 . in large distorting rnio@xmath0 such as ynio@xmath0 , the charge disproportionation 2ni@xmath86ni@xmath2ni@xmath3 occurs . the charge ordering mechanism can explain the stabilization of the asymmetric alignment of the local magnetic moments around each nickel site in both systems . we may add the final comment about effects of electron - electron correlation in rnio@xmath0 . the widths of the calculated e@xmath28 bands @xmath87 by the lda are @xmath88(ynio@xmath0 ) , @xmath89(ndnio@xmath0 ) , and @xmath90(lanio@xmath0 ) . then , the ratio of @xmath91 may be estimated as @xmath92 , assuming the value of the coulomb repulsion @xmath16 is common for all , and one can see a large reduction of @xmath91 in lanio@xmath0 , which may be the key parameter for the difference of the ground states of these perovskite nickelates . ndnio@xmath0 shows the anomalous m - i transition . @xcite lanio@xmath0 is presumably an anomalous metal of strongly correlated electrons @xcite and could not be treated within the framework of the lsda+u method .
according to the latest analysis@xcite of extended data records of the pioneer @xmath6 spacecraft , the small frequency drift@xcite ( blue - shift ) observed analyzing the navigational data of both the spacecraft , known as pioneer anomaly ( pa ) , may present a further time - dependent component in addition to the well known constant one . both linear and exponential models were proposed@xcite for the pa ; according to the authors of ref . , the exponential one is directly connected to non - gravitational effects@xcite since it takes into account the possible role of the on - board power generators suffering a radioactive decay . in this letter we work out the orbital effects of such a new term in the hypothesis that the time - dependent pa component is due to some sort of long - range modification of the known laws of gravitation resulting in an additional anomalous acceleration with respect to the nearly sunward constant one , having magnitude@xcite \|a_pio\|=(8.741.33)10 ^ -10 m s^-2,in terms of which the constant part of the pa has often been interpreted . indeed , in this case it should act on the major bodies of the solar system as well , especially those whose orbits lie in the regions in which the pa manifested itself in its presently known form . in this respect , we will not consider the exponential model . recent studies@xcite , partly preceding the one in ref . , pointed towards a mundane explanation of a large part of the pa in terms of non - gravitational effects pertaining the spacecraft themselves . since the anomalous acceleration is@xcite a_pio-210 ^ -11 m s^-2 yr^-1,the time - dependent linear component of the postulated pa - type acceleration@xcite a= ( t - t_0)a_piocan be treated as a small perturbation of the dominant newtonian monopole @xmath7 over timescales of the order of an orbital period @xmath8 for all the planets of the solar system . table [ pertu ] explicitly shows this fact for saturn , uranus , neptune and pluto which move just in the spatial regions in which the pa perhaps started to appear ( saturn ) , or fully manifested itself ( uranus , neptune , pluto ) in its presently known form . @cccccc@ planet & @xmath0 ( au ) & @xmath1 & @xmath9 ( m s@xmath5 ) & @xmath8 ( yr ) & @xmath10 ( m s@xmath5 ) + saturn & @xmath11 & @xmath12 & @xmath13 & @xmath14 & @xmath15 + uranus & @xmath16 & @xmath17 & @xmath18 & @xmath19 & @xmath20 + neptune & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 + pluto & @xmath26 & @xmath27 & @xmath28 & @xmath29 & @xmath30 + thus , the gauss equations for the variation of the osculating keplerian orbital elements@xcite , which are valid for any kind of disturbing acceleration @xmath31 , independently of its physical origin , can be safely used for working out the orbital effects of . in particular , the gauss equations for the semi - major axis @xmath0 and eccentricity @xmath1 of the orbit of a test particle moving around a central body of mass @xmath32 are @xmath33,\\ \\ % \dert e t & = & \rp{\sqrt{1-e^2}}{na}\left\{a_r\sin f + a_{t}\left[\cos f + \rp{1}{e}\left(1 - \rp{r}{a}\right)\right]\right\ } : \end{array}\lb{gauss}\ ] ] they allow one to work out the rates of changes of @xmath0 and @xmath1 averaged over one orbital period @xmath8 as t= ( ) _ 0^p_b(t)_kdt , = a , e.in @xmath34 are the right - hand - sides of evaluated onto the unperturbed keplerian ellipse . in @xmath35 are the radial and transverse components of a the generic disturbing acceleration @xmath31 , @xmath36 is the semilatus rectum , @xmath37 is the unperturbed keplerian mean motion related to the orbital period by @xmath38 , @xmath39 is the newtonian constant of gravitation , and @xmath40 is the true anomaly . since the new data analysis@xcite does not rule out the line joining the sun and the spacecrafts as a direction for the pa , we will assume that is entirely radial , so that @xmath41 . using the eccentric anomaly @xmath42 as a fast variable of integration turns out to be computationally more convenient . to this aim , useful relations are lll dt & = & ( ) de , + + t - t_0 & = & ( ) , + + f & = & . as a result , @xmath0 and @xmath1 experience non - vanishing secular variations lll a t & = & - , + + e t & = & - . notice that are exact in the sense that no approximations in @xmath1 were assumed . moreover , they do not depend on @xmath43 . in order to make a meaningful comparison of with the latest empirical results from planetary orbit determinations , we recall that modern data records cover at least one full orbital revolution for all the planets with the exception of neptune and pluto . the author of ref . , in producing the epm2006 ephemerides , made a global fit of a complete suite of standard dynamical force models acting on the solar system s major bodies to more than 400,000 observations of various kinds ranging over @xmath44 yr ( @xmath45 ) . among the about @xmath46 estimated parameters , there are the planetary orbital elements as well . according to table 3 of ref . , the formal , statistical errors in @xmath0 for saturn and uranus are lll _ a_^(epm2006 ) & = & 4,256 m + + _ a_^(epm2006 ) & = & 40,294 m , + + so that lll _ a_^(epm2006 ) & = & 46 m yr^-1 + + _ a_^(epm2006 ) & = & 433 m yr^-1 , + + can naively be inferred for their rates by simply dividing by @xmath47 . the pa was not modeled in the epm2006 . it is important to remark that the figure for @xmath48 quoted in was obtained without processing the radiotechnical observations of the cassini spacecraft . according to , the putative pa - induced secular changes of the semi - major axes of saturn and uranus are lll a_^(pio)&= & 42,505 m yr^-1 + + a_^(pio)&= & 290,581 m yr^-1 . + + these are about @xmath49 orders of magnitude larger than : even by re - scaling the formal uncertainties of by a factor of 10 , the pa - type anomalous rates of would still be about 2 orders of magnitude too large to have escaped from a detection . such conclusions are confirmed , and even enforced , by using the latest results published in ref . whose authors explicitly estimated secular changes of the semi - major axes of the first six planets with the epm2010 ephemerides based on more than 635,000 observations of different types over @xmath50 yr ( 1913 - 2010 ) . the authors of ref . , whose goal was a different one , did not model the pa . they obtain correlation between @xmath51 and @xmath52 was reported in ref . . ] a^(epm2010 ) _ = 13 m yr^-1,which is 4 orders of magnitude smaller than the predicted value of . incidentally , we note that if acted on , say , the earth , it would cause a variation in its semi - major axis as large as a^(pio)_=14.5 m yr^-1corresponding to a shift of a^(pio)_=1,348.5 mover @xmath44 yr . the formal , statistical uncertainty in the terrestrial semi - major axis is@xcite _ a_^(epm2006)=0.138 m , i.e . about 4 orders of magnitude smaller than . even by re - scaling by a factor of 10 , would still be 3 orders of magnitude too large . the epm2010 ephemerides correlation between @xmath53 and @xmath54 was reported in ref . . ] yield@xcite a_^(epm2010)=210 ^ -5 m yr^-1,i.e . 6 orders of magnitude smaller than the prediction of . in principle , it may be argued that was not included in the mathematical models fitted to the planetary data in the orbit determination process , so that its signature may have partly or totally been absorbed in the estimation of , say , the planetary state vectors : thus , the entire observational record should be re - processed by using ad - hoc modified dynamical force models explicitly including itself . however , this argument may have a validity especially when the magnitude of a putative anomalous effect of interest is close to the accuracy of the orbit determination process : it is not the case here . moreover , as far as the constant pa term is concerned , independent analyses@xcite in which it was explicitly modeled as a radial extra - acceleration acting on the outer planets substantially confirmed its neat incompatibility with the observations , as suggested in earlier studies@xcite . concerning the possibility that the pa is present at saturn , the authors of ref . yield an equivalent acceleration a^()_pio=(4.5811.80)10 ^ -10 m s^-2for its constant term . it is known that a radial constant extra - acceleration causes a secular precession of the perihelion@xcite given by ^(pio)=.the present - day accuracy in empirically constraining any anomalous precession @xmath55 of the perihelion of saturn from the inpop10a ephemerides , in which the pa was not modeled , is@xcite _ _ = 0.65 milliarcsec cty^-1.thus , and yield magnitude smaller than . if the time - dependent linear part of the pa were a genuine effect of gravitational origin causing an extra - acceleration , it would affect the orbital motions of the planets with secular variations of their semi - major axis and eccentricity if it were oriented towards the sun . the magnitude of such putative effects for saturn and uranus is up to @xmath2 orders of magnitude larger than the current accuracy in determining their orbital motions from the observations without modeling the pa itself . a partial removal of a pa - like signature may , in principle , have occurred in the standard parameter estimation procedure , so that the entire planetary data set should be re - processed with ad - hoc modified dynamical models explicitly including a hypothetical time - dependent pa - type acceleration as well . however , supported by the results of analogous tests for the constant form of the pa actually implemented by independent teams of astronomers with different purposely modified dynamical models , we feel that it is unlikely that such a potential removal of a pa - type signature can really justify the absence of such a huge effect in the currently available processed planetary data . concerning the existence of a constant pa - type acceleration at heliocentric distances of about @xmath3 au as large as @xmath56 m s@xmath5 , latest results about the maximum allowed size of anomalies in the perihelion precession of saturn yield a figure 5 orders of magnitude smaller . e. v. pitjeva , use of optical and radio astrometric observations of planets , satellites and spacecraft for ephemeris astronomy , in _ a giant step : from milli- to micro - arcsecond astrometry proceedings iau symposium no . 248 , _ eds . w. j. jin , i. platais and m. a. c. perryman ( cambridge university press , 2008 ) , pp . 20 - 22 . e. m. standish , planetary and lunar ephemerides : testing alternate gravitational theories , in _ aip conf . recent developments in gravitation and cosmology -3rd mexican meeting on mathematical and experimental physics , vol . 977 _ , eds . a. macias , c. lmmerzahl and a. camacho ( american institute of physics , 2008 ) , p. 254 . a. fienga , j. laskar , p. kuchynka , h. manche , m. gastineau and c. leponcin - lafitte , gravity tests with inpop planetary ephemerides , in _ sf2a-2009 : proceedings of the annual meeting of the french society of astronomy and astrophysics _ , eds . m. heydari - malayeri , c. reyl and r. samadi ( socit francaise d astronomie et d astrophysique , 2009 ) , p. 105 . e. m. standish , testing alternate gravitational theories , in _ relativity in fundamental astronomy : dynamics , reference frames , and data analysis , proceedings iau symposium no . 261 _ , eds . s. a. klioner , p. k. seidelmann and m. h. soffel ( cambridge university press , 2010 ) , p. 179 .	we work out the impact that the recently determined time - dependent component of the pioneer anomaly ( pa ) , if interpreted as an additional exotic acceleration of gravitational origin with respect to the well known pa - like constant one , may have on the orbital motions of some planets of the solar system . by assuming that it points towards the sun , it turns out that both the semi - major axis @xmath0 and the eccentricity @xmath1 of the orbit of a test particle would experience secular variations . for saturn and uranus , for which modern data records cover at least one full orbital revolution , such predicted anomalies are up to @xmath2 orders of magnitude larger than the present - day accuracies in empirical determinations of their orbital parameters from the usual orbit determination procedures in which the pa was not modeled . given the predicted huge sizes of such hypothetical signatures , it is unlikely that their absence from the presently available processed data can be attributable to an for them in the estimated parameters caused by the fact that they were not explicitly modeled . the magnitude of a constant pa - type acceleration at @xmath3 au can not be larger than @xmath4 m s@xmath5 according to the latest observational results for the perihelion precession of saturn . # 1eq . ( [ # 1 ] ) # 1#2eq . ( [ # 1])-eq . ( [ # 2 ] ) # 1eq . ( [ # 1 ] ) # 1#2eq . ( [ # 1])-eq . ( [ # 2 ] ) # 1#2 # 1#2 # 1#1 # 1``#1 '' # 1#2^#2_.#1 # 1#2#3 # 1#2#1#2 # 1[#1 ] # 1#2#3 # 1
galaxies that radiate most of their extremely large , quasar - like luminosities ( @xmath2 10@xmath3 ) as infrared dust emission the ultraluminous infrared galaxies ( ulirgs ; sanders & mirabel 1996 ) dominate the bright end of the galaxy luminosity function in the nearby universe @xcite . they have been used , extensively , to derive information on the dust - obscured star - formation rate , dust content , and metallicity in the early universe @xcite . understanding the nature of nearby ulirgs , and more particularly , determining whether they are powered by starbursts or active galactic nuclei ( agns ) , or both , is of great importance in modern extragalactic astronomy . spectroscopic observation of the thermal infrared wavelength range ( 320 @xmath0 m ) is currently one of the most powerful methods of determining the energy sources of ulirgs . at this wavelength range , dust extinction is lower than at shorter wavelengths ( @xmath42 @xmath0 m ) , so it becomes possible to detect and measure the emission from more highly obscured energy sources , with smaller uncertainty in dust extinction corrections . in addition , spectral features in this waveband can be used to distinguish between starburst and agn activity . polycyclic aromatic hydrocarbon ( pah ) emission is detected in starbursts but not in agns , making it a good indicator of starburst activity @xcite , while the presence of agns hidden behind dust can be recognized through absorption features . pah emission and absorption features , detected in ground - based slit spectra ( e.g. , dudley 1999 ; imanishi & dudley 2000 ; soifer et al . 2002 ) and in _ iso _ spectra taken with large apertures ( e.g. , tran et al . 2001 ) , have been utilized to investigate the energy sources of ulirgs . in ulirgs , there are two potentially energetically significant components . one is a weakly obscured ( a@xmath5 @xmath6 20 mag)m is insignificant , which is roughly the criterion that a@xmath7 @xmath6 1 mag , or a@xmath5 @xmath6 20 mag if the galactic extinction curve of @xmath8 @xcite is adopted . ] starburst in the extended ( kpc scale ) host galaxy and also at the nucleus . the other is a nuclear , compact ( less than a few hundred pc in size ) , highly obscured energy source , powered by an agn and/or a compact starburst . recent ground - based high spatial resolution imaging observations @xcite have shown that the weakly obscured ( @xmath9 mag ) starbursts are energetically insignificant , and that the nuclear , compact , highly obscured energy sources dominate in ulirgs @xcite . however , the nature of these nuclear sources is not easy to determine , because dust extinction can be significant even in the thermal infrared wavelength range . the observed pah - to - infrared luminosity ratios in ulirgs have been found to be smaller than in known starburst galaxies by a large factor @xcite . assuming that the luminosities of the pah emission are a good indicator of the absolute magnitudes of starbursts , the smaller ratios in ulirgs could be due either to an increase in the dust extinction of the pah emission , or a dominant contribution from agns to the infrared luminosities of ulirgs . determining which explanation is correct requires reliable estimates of the dust extinction toward the starbursts in ulirgs . however , these estimates become complicated if the assumption of a `` foreground screen '' dust geometry is not valid for the starbursts @xcite . to understand the nature of the compact , highly obscured energy sources in ulirgs , the spectral shapes of the nuclear compact emission can provide more insight than just the absolute luminosities of pah emission , because the equivalent widths of pah emission are , by definition , robust to the amount of dust extinction . if the nuclear compact emission of ulirgs displayed significantly lower equivalent widths of pah emission than typically seen in starburst galaxies , and , in addition , strong absorption features were detected , it would suggest that energetically important dust - obscured agns were present . since the attenuation of the extended , weakly obscured emission is much smaller than that of the emission from the compact , highly obscured energy sources , the observed 320 @xmath0 m spectra taken with large apertures can be strongly contaminated by the extended emission , even though the total infrared ( 81000 @xmath0 m ) luminosities of ulirgs are dominated by the compact energy sources . in this case , the signatures of absorption features toward the compact energy sources can be diluted . hence spectroscopy with a narrow ( less than a few arcsec ) slit comparable to the spatial extent of the compact component is better suited to probe the nature of the compact energy sources in ulirgs . ground - based slit spectroscopy in the @xmath10-band ( 2.84.1 @xmath0 m ) has several important advantages for investigating the nature of dust - enshrouded agns : 1 . the dust extinction curve is fairly grey ( wavelength - independent ) at 310 @xmath0 m @xcite . 2 . the dust around an agn has a strong temperature gradient , in that the inner ( outer ) dust has a higher ( lower ) temperature . the temperature of the innermost dust is determined by the dust sublimation temperature ( t @xmath11 1000 k ) . in the case of blackbody radiation ( @xmath12 [ @xmath0 m ] @xmath13 t [ k ] @xmath11 3000 ) , the 34 @xmath0 m continuum emission is dominated by dust with @xmath141000 k located in the innermost regions , while the @xmath1110 @xmath0 m continuum emission is dominated by dust with @xmath15 k located further out . thus , the dust extinction toward the @xmath1110 @xmath0 m continuum emission regions , as estimated using the optical depth of the 9.7 @xmath0 m silicate dust absorption feature , is much smaller than the dust extinction toward the agn ; the dereddened agn luminosity will be underestimated if the extinction correction is made based on @xmath1110 @xmath0 m data alone . by contrast , the dust extinction toward the innermost , 34 @xmath0 m continuum emission regions , as estimated using 34 @xmath0 m data , is a better indicator of the dust extinction toward the agn itself . in the case that an agn is the only energy source , the 34 @xmath0 m continuum emission predominantly originates in the innermost , 8001000 k dust . the dust extinction toward the 34 @xmath0 m continuum emission can therefore be reliably fitted with a foreground screen dust model . in a real galaxy , weakly obscured starburst emission may contribute to the observed nuclear 34 @xmath0 m flux . however , as long as the agn is sufficiently luminous that agn - powered continuum emission contributes significantly to the observed 34 @xmath0 m flux , evidence for an agn can be found , provided that the starburst contribution is subtracted properly , even if the absorption optical depth toward the agn is larger than unity at 34 @xmath0 m ( @xmath16 mag or @xmath17 mag ; rieke & lebofsky 1985 ) . we have previously demonstrated this in two ulirgs : iras 08572 + 3915 ( @xmath18 mag ; imanishi & dudley 2000 ) , and ugc 5101 ( @xmath19 mag ; imanishi et al . 3 . at 2.84.1 @xmath0 m , in addition to the 3.3 @xmath0 m pah emission , 3.4 @xmath0 m carbonaceous dust absorption and 3.1 @xmath0 m h@xmath20o ice absorption are present . as we will show later in this paper , all these features can be used in a complementary manner to constrain the properties of the compact energy sources . they are simultaneously observable with recently available spectrographs that cover the whole @xmath10-band atmospheric window ( 2.84.1 @xmath0 m ) , if the target sources are at @xmath21 . in this redshift range , a statistically significant number of ulirgs have been found with _ iras_. 4 . in the near future , slit spectra of ulirgs at 5.340 @xmath0 m will be obtainable using the _ sirtf _ irs . by combining 34 @xmath0 m and 5.340 @xmath0 m slit spectra , the presence of a strong temperature gradient in the dust ( a signature of an agn ) can be investigated through the comparison of the optical depths of absorption features at different wavelengths at 320 @xmath0 m @xcite . in the study of ulirgs , it is particularly important to detect and quantify elusive agns that are deeply buried in dust along all sightlines ( hereafter buried agns ; imanishi et al . ulirgs that are classified optically as liners and have cool far - infrared colors ( sanders et al . 1988 ) have generally been taken to be starburst - powered @xcite . however , @xcite argued that liner ulirgs may contain buried agns ( see also antonucci 2001 ) . in this paper 2.84.1 @xmath0 m ground - based slit spectra of liner ulirgs with cool far - infrared colors are reported , to search for observational evidence for energetically important , buried agns . throughout this paper , h@xmath22 @xmath23 75 km s@xmath24 mpc@xmath24 , @xmath25 = 0.3 , and @xmath26 = 0.7 are adopted . the seven liner ulirgs in table [ tbl-1 ] were observed . table [ tbl-1 ] summarizes their infrared emission properties . all have cool far - infrared colors . the optical classifications as liners are based on @xcite and @xcite . table [ tbl-2 ] summarizes the observing log . the 3.14.0 @xmath0 m slit spectrum of ugc 5101 has already been presented in @xcite . descriptions of the observations and data analysis were offered in that letter . to summarize , the observations were made at the irtf on mauna kea , hawaii , using the grism mode of nsfcam @xcite . although the wavelength coverage of the nsfcam grism mode is 2.84.1 @xmath0 m , results at @xmath27 3.1 @xmath0 m were not presented in @xcite because our main aim in that study was to investigate the 3.3 @xmath0 m pah emission and 3.4 @xmath0 m carbonaceous dust absorption features , and because the earth s atmospheric transmission at 2.83.1 @xmath0 m is poorer than at @xmath28 3.3 @xmath0 m . however , after the publication of @xcite , @xcite reported the detection of 6.0 @xmath0 m h@xmath1o ice absorption in ugc 5101 , and suggested that the 3.1 @xmath0 m h@xmath1o ice absorption feature might be present in our 3.14.0 @xmath0 m spectrum . therefore we carefully analyzed the spectrum at 2.83.1 @xmath0 m to investigate the 3.1 @xmath0 m h@xmath1o ice absorption feature , whose profile is very broad , extending over at least @xmath29 2.83.3 @xmath0 m in the rest - frame ( smith , sellgren , & tokunaga 1989 ) . the 2.84.1 @xmath0 m spectra of the other ulirgs were obtained with the infrared spectrograph and camera ( ircs ; kobayashi et al . 2000 ) at the subaru telescope on mauna kea , hawaii . the sky was clear during the observations of these objects and the seeing at k was measured to be 0@xmath3050@xmath308 in full - width at half maximum . a 0@xmath309-wide slit and the @xmath10-grism were used with a 58-mas pixel scale . the achievable spectral resolution is @xmath11140 at 3.5 @xmath0 m . a standard telescope nodding technique , with a throw of 7 arcsec along the slit , was employed to subtract background emission . ulirgs and corresponding standard stars were observed with an airmass difference of @xmath40.1 to correct for the transmission of the earth s atmosphere , and to provide flux calibration . standard data analysis procedures were employed , using iraf . initially , bad pixels were replaced with interpolated signals from the surrounding pixels . bias was subsequently subtracted from the obtained frames and the frames were divided by a spectroscopic flat frame . finally the spectra of ulirgs and standard stars were extracted . wavelength calibration was performed taking into account the wavelength - dependent transmission of the earth s atmosphere . the spectra of ulirgs were divided by the observed spectra of standard stars , multiplied by the spectra of blackbodies with temperatures appropriate to individual standard stars ( table [ tbl-2 ] ) , and then flux - calibrated . appropriate binning of spectral elements was performed , particularly at @xmath31 @xmath0 m in the observed frame , so as to give an adequate signal - to - noise ratio in each element . the resulting spectral resolution @xmath32 at @xmath31 @xmath0 m is @xmath6100 . although the earth s atmospheric transmission curve at 2.83.3 @xmath0 m is highly wavelength - dependent if observed at high spectral resolution of @xmath33 ( figure [ fig1]a ) , it is fairly smooth at lower spectral resolutions , with r @xmath6 100 ( figure [ fig1]b ) . thus , even if the net positions of the target object and standard star on the slit differ slightly ( on the sub - pixel scale ) along the wavelength direction , the standard data analysis described above is expected to produce no significant spurious features in spectra with r @xmath6 100 . figure [ fig2 ] shows flux - calibrated 2.84.1 @xmath0 m spectra of the seven ulirgs . since our slit spectra cover physical scales larger than a few hundred pc in all the ulirgs , emission from both agns and compact ( less than a few hundred pc ) starbursts should be fully detected . all the ulirgs clearly show the 3.3 @xmath0 m pah emission . to estimate its flux , luminosity , and rest - frame equivalent width , we make the reasonable assumption that the profiles of the 3.3 @xmath0 m pah emission in these ulirgs are similar to those of galactic star - forming regions and nearby starburst galaxies ; the main emission profile then extends between @xmath34 @xmath23 3.243.35 @xmath0 m @xcite . the flux excess above the solid lines in figure [ fig2 ] should thus be ascribed to the 3.3 @xmath0 m pah emission . table [ tbl-3 ] summarizes the results for the 3.3 @xmath0 m pah emission in these seven ulirgs . the spectrum of ugc 5101 shows clear 3.4 @xmath0 m carbonaceous dust absorption with an observed optical depth of @xmath35 = 0.65 @xcite . in other ulirgs , there is no clear indication of this feature , partly because this absorption feature is intrinsically not so strong ( @xmath35/a@xmath5 @xmath11 0.0040.007 ; pendleton et al . 1994 ) . in addition to the 3.3 @xmath0 m pah emission feature , there is another important common feature of the spectra of five ulirgs ( ugc 5101 , iras 00188@xmath360856 , iras 03250 + 1606 , iras 16487 + 5447 , and iras 21329@xmath362346 ) : the continuum emission is concave in these spectra . at the shorter wavelength side of the 3.3 @xmath0 m pah emission , the continuum flux level initially decreases with decreasing wavelength , but then begins to increase again at @xmath37 @xmath0 m . in iras 00188@xmath360856 and iras 03250 + 1606 , the spectra become flat at the shortest wavelength parts ( @xmath34 @xmath4 2.7 @xmath0 m ) . this behavior is quite similar to that of the infrared - luminous galaxy ngc 4945 , in which 3.1 @xmath0 m h@xmath1o ice absorption was clearly detected in an _ iso _ spectrum @xcite . earlier ground - based @xmath10-band spectroscopy has revealed that many bright galactic sources behind substantial columns of dust and molecular gas show broad 3.1 @xmath0 m h@xmath1o ice absorption , with a main profile peaking at @xmath34 @xmath11 3.05 @xmath0 m and extending at least to @xmath34 @xmath23 2.83.3 @xmath0 m ( smith et al . 1989 ; smith , sellgren , & tokunaga 1993 ; smith 1993 ) . a recent _ iso _ spectrum of ngc 4945 suggests that the absorption may extend to wavelengths as short as @xmath38 @xmath0 m @xcite . in addition to this main profile , a weak longer - wavelength wing at @xmath393.7 @xmath0 m can also be distinguished @xcite . the concave - shaped continuum emission found in the spectra of these five ulirgs is consistent with the presence of this broad 3.1 @xmath0 m h@xmath1o ice absorption . some seyfert 2 galaxies with lower infrared luminosities were observed on the same nights as these ulirgs , but no signatures of strong 3.1 @xmath0 m h@xmath1o ice absorption were found . the spectra of standard stars observed before these ulirgs on the same nights were divided by other standard stars observed after , but the resulting spectra also never showed such signatures . the strong 3.1 @xmath0 m h@xmath1o ice absorption signatures found only in the spectra of the ulirgs are thus believed to be genuine and intrinsic to the ulirgs , and not an artifact caused by insufficient cancellation of the earth s atmospheric transmission . in fact , for ugc 5101 and iras 00188@xmath360856 , the clear detection of h@xmath1o ice absorption at 6.0 @xmath0 m was also reported by @xcite . to estimate conservative lower limits for the optical depths of the broad 3.1 @xmath0 m h@xmath1o ice absorption feature , the concave quadratic dashed lines in figure [ fig2 ] are adopted as the lowest plausible continuum levels . in iras 23234 + 0946 and iras 23327 + 2913 , the presence of strong 3.1 @xmath0 m h@xmath1o ice absorption is not clear in the case of the adopted concave continuum . for the remaining five ulirgs , the observed optical depths of the 3.1 @xmath0 m h@xmath1o absorption ( @xmath40 ) as well as the 3.4 @xmath0 m carbonaceous dust absorption ( @xmath35 ) are summarized in table [ tbl-4 ] . these @xmath40 values should be regarded as strict lower limits because : ( 1 ) linear continuum levels that connect data points at the shortest and longest wavelength parts of the spectra would give rise to larger values of @xmath40 , and ( 2 ) the data points at the shortest wavelength parts , used to determine continuum levels , may still be affected by the 3.1 @xmath0 m h@xmath1o ice absorption feature in some cases . compared to ugc 5101 , iras 00188@xmath360856 and iras 21329@xmath362346 , the presence of the broad h@xmath1o ice absorption feature is apparently less clear in iras 03250 + 1606 and iras 16487 + 5447 . however , as long as the spectral shapes of the 3.3 @xmath0 m pah emission in these ulirgs do not differ significantly from those observed so far in starburst galaxies , the continuum flux levels on the shorter wavelength side of the 3.3 @xmath0 m pah emission are clearly lower than on the longer wavelength side , which is consistent with the presence of the broad 3.1 @xmath0 m h@xmath1o ice absorption . the very broad emission - like features that extend between @xmath41 3.63.9 @xmath0 m and 3.43.8 @xmath0 m in iras 03250 + 1606 and iras 16487 + 5447 , respectively , can not be ascribed solely to the 3.3 @xmath0 m pah emission ; instead , they are caused by a combination of 3.1 @xmath0 m h@xmath1o ice absorption , 3.3 @xmath0 m pah emission , and ( possibly ) 3.4 @xmath0 m carbonaceous dust absorption . the observed 3.3 @xmath0 m pah to infrared luminosity ratios for these seven ulirgs are l@xmath42/l@xmath43 @xmath11 13 @xmath4410@xmath45 ( table [ tbl-3 ] ) , which are 3 to 10 times smaller than the ratios for starburst galaxies ( l@xmath42/l@xmath43 @xmath11 10@xmath46 ; mouri et al . 1900 ; imanishi 2002 ) . it can therefore be concluded that _ weakly obscured ( a@xmath47 mag ) nuclear _ starbursts are not contributing significantly to the infrared luminosities of these ulirgs . since the l@xmath42 values are based on slit spectroscopy measurements , the luminosities of _ extended _ ( kpc scale ) weakly obscured starbursts are not taken into account . however , as mentioned in @xmath481 , such extended starbursts have been shown to be energetically insignificant in ulirgs @xcite . for ugc 5101 , @xcite estimated , based on the _ iso _ s large aperture measurement , a 6.2 @xmath0 m pah emission flux of @xmath49 w @xmath50 , so that its luminosity is @xmath51 ergs s@xmath24 in our adopted cosmology . the resulting @xmath52 , which is only @xmath53 of the ratio typically seen in starburst galaxies ( @xmath54 : fischer 2000 ) . thus , weakly obscured extended starbursts are unlikely to be the dominant energy source in ugc 5101 , either . the dominant energy sources should be compact and heavily obscured ; for the five ulirgs , the presence of such buried energy sources is supported by the detection of strong absorption features in our 2.84.1 @xmath0 m spectra . the detection of strong ( @xmath40 @xmath55 0.6 ) 3.1 @xmath0 m h@xmath1o ice absorption can , in principle , provide stringent constraints on the nature of the buried energy sources at the nuclei of the five ulirgs . the 3.4 @xmath0 m carbonaceous and 9.7 @xmath0 m silicate dust absorption features are detectable if an adequate amount of dust in the _ diffuse inter - stellar medium _ is present in front of the continuum - emitting energy sources . however , in order for h@xmath1o ice absorption to be detected , dust grains _ covered with an ice mantle _ must be present in the foreground . such a situation can arise only when the dust grains are located in molecular clouds , and are sufficiently shielded from the uv radiation from any energy sources either inside or outside the molecular clouds @xcite . in other words , the detection of strong h@xmath1o ice absorption requires that a large amount of dust in _ molecular clouds _ be present in front of the continuum - emitting energy sources . foreground molecular clouds in the host galaxies could be responsible for the strong 3.1 @xmath0 m h@xmath1o ice absorption detected toward the nuclei if the host galaxies were seen from nearly edge - on . however , there is no strong evidence that the host galaxies of these ulirgs are edge - on @xcite . @xcite reported that 6.0 @xmath0 m h@xmath1o ice absorption is systematically stronger in ulirgs than in less infrared - luminous seyfert 2 galaxies . these seyfert 2 galaxies also generally show no detectable or weak ( @xmath56 ) 3.1 @xmath0 m h@xmath1o ice absorption ( imanishi et al . 2002 , in preparation ) . the systematically larger optical depths of h@xmath1o ice absorption toward the nuclei of ulirgs can reasonably be explained by a larger amount of highly concentrated molecular gas ( and dust ) in their nuclei @xcite . we thus regard it as likely that the strong ( @xmath40 @xmath55 0.6 ) 3.1 @xmath0 m h@xmath1o ice absorption detected toward the nuclei of these five ulirgs are caused by nuclear concentrations of molecular gas and dust . note that , although the host galaxy of ngc 4945 is viewed edge - on @xcite , a significant fraction of the strong 3.1 @xmath0 m h@xmath1o ice absorption ( @xmath57 ; spoon et al . 2000 ) is likely to originate in a highly concentrated nuclear component of ice - covered dust grains , on a scale of less than a few hundred pc @xcite . in the case of _ normal _ starbursts , it seems reasonable to assume that the continuum - emitting sources and molecular gas and dust are spatially well mixed @xcite . a detailed infrared study of the nearby starburst galaxy m82 suggests that the mixed - dust geometry is a better fit to the observed data than a foreground screen dust distribution @xcite . since the oscillator strengths of species are fixed , there are upper limits on the absorption optical depths that can be produced by the mixed - dust geometry ( unless unusual abundances are considered ) . this is because the foreground , less - obscured , and therefore less - attenuated , emission ( which shows only weak absorption features ) will dominate the observed fluxes . in practice , nearby starburst galaxies such as m82 and ngc 253 show weak ( @xmath58 @xmath11 0.2 ) or undetectable absorption by h@xmath1o ice @xcite or dust grains @xcite . this issue will be discussed quantitatively below . in galactic molecular clouds , the optical depth of the h@xmath1o ice absorption and dust extinction ( @xmath59 ) are correlated , with a relation given by @xmath60 . the value of @xmath61 has been found to be almost constant , @xmath110.06 , in galactic molecular clouds , which exhibit various different levels of star - forming activity @xcite . this is because it reflects the abundance of h@xmath1o ice relative to dust grains in the regions where an ice mantle can survive ; it is therefore unlikely to differ substantially from the galactic value , even in starburst galaxies . the a@xmath62 value in the above relation is the threshold required for dust grains to be sufficiently shielded from the ambient uv radiation . this value can vary dramatically , from 26 mag in quiescent molecular clouds ( taurus and serpens ; whittet et al . 1988 ; eiroa & hodapp 1989 ; murakawa et al . 2000 ) to 15 mag in active star - forming molecular clouds ( @xmath63-oph ; tanaka et al . 1990 ) . in starburst galaxies with a more intense uv radiation field than the @xmath63-oph molecular clouds , @xmath64 is almost certainly larger than 15 mag , but its exact value is unknown . thus the relation @xmath65 is adopted in starburst galaxies , where @xmath66 mag , or @xmath67 where @xmath68 is the fraction of dust that is covered with an ice mantle . figures [ fig3]a and [ fig3]b show schematic diagrams of a mixed - dust geometry . in this geometry , the observed flux i(@xmath69 ) is given by @xmath70 where @xmath71 is an unattenuated intrinsic flux and @xmath69 is the optical depth at each wavelength , which takes different values inside and outside the absorption features . for 34 @xmath0 m continuum emission outside the 3.1 @xmath0 m h@xmath1o ice and 3.4 @xmath0 m carbonaceous dust absorption features , @xmath72 @xcite , where , by definition , @xmath73 . at the peak wavelength of the 3.1 @xmath0 m h@xmath1o ice absorption feature , @xmath74 where the first and second terms give the flux attenuation for 34 @xmath0 m continuum and the h@xmath1o ice absorption feature , respectively . thus , the optical depth of the h@xmath1o ice absorption feature in an observed 34 @xmath0 m flux is @xmath75 \\ & = & \ln [ ( 1+f ) \frac{1 - e^{-0.06 \times a_{\rm v}}}{1 - e^{-0.06 \times a_{\rm v } \times ( 1+f ) } } ] \end{aligned}\ ] ] this last formula implies that , for a fixed @xmath68-value , @xmath40 increases with increasing a@xmath5 , but saturates at a certain value . the @xmath68-value can be constrained based on observations of the nearby starburst galaxy m82 . based on a mixed - dust model , the dust extinction toward the farthest parts of the starburst region is estimated to be a@xmath5 @xmath11 55 mag @xcite . the observed optical depth of the 3.1 @xmath0 m h@xmath1o ice absorption is @xmath76 @xcite . based on these values , @xmath77 is obtained . in other words , in the starburst galaxy m82 , @xmath78 of dust grains are covered with an ice mantle . if the cores of ulirgs are simply a scaled - up version of normal starbursts with a mixed - dust geometry and the @xmath68-value is @xmath110.3 , then @xmath40 should be @xmath110.2 . however , the @xmath68-value is strongly dependent on the fraction of active star - forming regions relative to quiescent molecular gas in molecular clouds . if the cores of ulirgs are powered by starbursts with a mixed - dust geometry , the @xmath68-value in ulirgs should be even smaller than in less infrared luminous starburst galaxies @xcite . here , however , @xmath68 @xmath11 0.3 is adopted for ulirgs in order to determine a conservative upper limit for @xmath40 within the framework of a mixed - dust geometry . even if this conservative value is adopted , equation ( 7 ) shows that @xmath40 can not be larger than 0.3 . therefore , it can be concluded that _ the observed @xmath40 values of @xmath550.6 in these five ulirgs are clearly incompatible with mixed - dust geometry and f @xmath79 0.3_. when the contribution from any weakly obscured starburst emission to the observed 34 @xmath0 m fluxes is subtracted , the @xmath40 values toward the buried energy sources will be even larger , making this conclusion more robust . the same argument is applicable to the 3.4 @xmath0 m carbonaceous dust absorption in ugc 5101 @xcite . for the 3.4 @xmath0 m carbonaceous dust absorption feature , @xmath80 is obtained @xcite , where the term @xmath81 , the fraction of dust grains not covered by an ice mantle , is included because the 3.4 @xmath0 m carbonaceous dust absorption feature may be suppressed if the dust grains are covered with an ice mantle @xcite . based on the same arguments as above , it can be quantitatively demonstrated that @xmath35 can not be larger than 0.2 in the mixed - dust geometry . the observed large optical depth of @xmath82 in ugc 5101 @xcite is again incompatible with a mixed - dust source geometry . in order to explain the strong absorption features in these five ulirgs , dust in a foreground screen must be obscuring the 34 @xmath0 m continuum - emitting energy sources this configuration requires that the energy sources be centrally concentrated and deeply buried in nuclear dust and molecular gas , as shown in figure [ fig3]c . the centrally - concentrated nature of the buried energy source in ugc 5101 is naturally explained by a buried agn @xcite . furthermore , as mentioned in @xmath481 , the observed rest - frame equivalent widths of the 3.3 @xmath0 m pah emission in starburst galaxies should be little changed by dust extinction . the significantly lower equivalent width in ugc 5101 ( table [ tbl-3 ] ) than in starbursts ( @xmath11120 nm ; moorwood 1986 ; imanishi & dudley 2000 ) also supports the presence of agns that produce strong 34 @xmath0 m continuum emission , but essentially no pah emission . when our data are combined with ground - based high spatial resolution mid - infrared ( 825 @xmath0 m ) imaging data @xcite , we can give more insight into the nature of the buried energy source . @xcite estimated the size of the spatially - unresolved core of ugc 5101 at 12.5 @xmath0 m to be @xmath4200 pc . if the attenuation of the mid - infrared flux from the spatially - unresolved compact core in ugc 5101 is negligible , the surface brightness of the core emission becomes @xmath551.0 @xmath13 10@xmath83 kpc@xmath84 @xcite . however , the detection of strong 3.4 @xmath0 m carbonaceous dust and 3.1 @xmath0 m h@xmath1o ice absorption toward the nucleus of ugc 5101 requires that 1 . the actual size of the buried centrally - concentrated energy source is substantially smaller than that of the dust and molecular gas , and 2 . the dust extinction toward the buried energy source is @xmath85 mag @xcite . on point ( 1 ) , since dust around a centrally - concentrated energy source shows a temperature gradient ( see @xmath481 ) , the size of the cooler dust region , which is also responsible for the foreground dust extinction , is larger than that of the @xmath11250 k dust traced by the 12.5 @xmath0 m emission ( @xmath4200 pc ) . however , the size of the cool dust emission ( traced by the 1.3 mm emission ) at the core of the ulirg arp 220 has been shown to be compact ( @xmath4200 pc ; sakamoto et al . furthermore , the sizes of the high density molecular gas regions at the cores of ulirgs have been estimated to be no more than @xmath11500 pc in diameter @xcite . thus , it seems reasonable to assume that the actual size of the buried centrally - concentrated energy source in ugc 5101 is only a fraction of this , @xmath11100 pc or less . on point ( 2 ) , the dust extinction of @xmath19 mag means that the flux attenuation at 12.5 @xmath0 m is more than factor of 25 @xcite . when these corrections are taken into account , the actual surface brightness of the centrally - concentrated buried energy source in ugc 5101 is @xmath86 kpc@xmath84 . this exceeds the peak surface brightness of star clusters in starburst galaxies ( @xmath115 @xmath13 10@xmath87 kpc@xmath84 ; meurer et al . 1997 ; soifer et al . 2000 ) by a factor of more than two . the high surface brightness is best explained by the presence of an energetically important buried agn in ugc 5101 . if a buried agn is present at the nucleus of ugc 5101 , the dust is expected to show a strong temperature gradient ( see @xmath481 ) . are the available data consistent with its presence ? @xcite could not explicitly find the signature of a temperature gradient based on their imaging data of ugc 5101 . here , it is searched for using spectroscopic data . as mentioned in @xmath481 and @xcite , if a strong temperature gradient is present in the dust , the optical depth of the 3.4 @xmath0 m carbonaceous dust absorption feature probes deeper into the dust distribution than that of the 9.7 @xmath0 m silicate dust absorption . thus , the @xmath35/@xmath88 ratio toward the nucleus of ugc 5101 is expected to be significantly larger than the value seen when no significant temperature gradient is present ( @xmath35/@xmath890.07 ; roche & aitken 1984 , 1985 ; pendleton et al . 1994 ) . after subtracting the contribution from the weakly obscured starburst emission to the observed 34 @xmath0 m flux , the @xmath35 value toward the buried agn at the core of ugc 5101 was estimated to be @xmath110.8 @xcite , while only a lower limit for the observed @xmath88 value ( @xmath551.5 ) was given by @xcite based on a measurement with a large aperture using _ iso_. although the @xmath88 value is a lower limit , its actual value is unlikely to exceed 2.0 , judging from the actual spectrum of ugc 5101 ( spoon et al . 2002 ; their figure 8) . thus the @xmath35/@xmath88 ratio is @xmath90 . the actual @xmath88 value toward the buried agn in ugc 5101 could be larger than the observed value , because weakly obscured starburst emission is likely to contribute significantly to the observed @xmath1110 @xmath0 m _ iso _ flux . however , since the equivalent width of the 6.2 @xmath0 m pah emission in the _ iso _ spectrum of ugc 5101 is less than half that of starburst galaxies @xcite , at least half of the observed continuum flux at @xmath555 @xmath0 m should originate in the buried agn , implying that the @xmath35/@xmath88 ratio toward the buried agn is @xmath91 . thus , these spectroscopic data are consistent with the presence of a temperature gradient of dust at the core of ugc 5101 . if the dust around an agn has a torus - like geometry , low - density clumpy clouds which are directly illuminated by the central agn can exist at relatively large distances from the agn along the axis of the torus - geometry dust ( the so - called the narrow line regions or nlrs ) . nlrs produce the strong high - excitation forbidden emission lines seen in agn . however , in a buried agn , x - ray dissociation regions ( maloney et al . 1996 ) , rather than nlrs , are expected to be produced around the agn , so that high - excitation forbidden emission lines will be weak , consistent with the observed data of ugc 5101 @xcite . @xcite and @xcite found that the infrared to hcn ( j=1 - 0 ) luminosity ratios in ulirgs , including ugc 5101 @xcite , roughly follow the relation established in less infrared - luminous starburst galaxies . by regarding the hcn ( j=1 - 0 ) luminosities as a tracer of the mass of high density ( @xmath5510@xmath92 cm@xmath46 ) molecular gas where star - formation actually occurs , @xcite argued that the infrared luminosities of ulirgs are powered by starbursts . however , in the luminous inner regions ( central few hundred pc ) of ulirgs , the gas pressures - and therefore molecular gas densities - are inevitably extremely high , whether the energy sources are starbursts or agns @xcite . the relative constancy of the infrared to hcn luminosity ratio from less infrared luminous starburst galaxies to ulirgs simply indicates that much of the infrared emission arises from dust in dense molecular clouds , which is inevitable whatever the sources of the luminosities of ulirgs . for the two ulirgs which do not show strong 3.1 @xmath0 m h@xmath1o ice absorption ( iras 23234 + 0946 and iras 23327 + 2913 ) , there is no strong observational evidence for the presence of buried agns . however , for the other four ulirgs which show strong 3.1 @xmath0 m h@xmath1o absorption ( iras 00188@xmath360856 , iras 03250 + 1606 , iras 16487 + 5447 , and iras 21329@xmath362346 ) , the centrally - concentrated nature of their buried energy sources and systematically ( roughly a factor of two ) lower - equivalent - width pah emission than starbursts ( table 3 ) are also naturally explained by the presence of energetically important , buried agns . assuming that approximately half of the observed 34 @xmath0 m fluxes originate in buried agns , the @xmath40 and @xmath35 toward the buried agns are twice as large as the observed values . these @xmath40 and @xmath35 values reflect the column density of foreground dust , toward the buried agns , covered with and without an ice mantle , respectively ( @xmath484.2 ) . based on the relation of @xmath40/a@xmath5 @xmath11 0.06 , @xmath35/a@xmath5 @xmath11 0.0040.007 ( @xmath48 4.2 ) , and @xmath94 @xmath11 0.06 ( @xmath48 1 ) , and on a foreground screen dust model ( @xmath48 1 ) , it is found that the dereddened 34 @xmath0 m dust emission luminosities powered by the agns could be comparable to the infrared luminosities of these four ulirgs . it is sometimes argued that ulirgs with cool far - infrared colors can not be powered by agns because agns show warm far - infrared colors ( e.g. , downes & solomon 1998 ) . indeed , many known seyfert 2 galaxies show warm far - infrared colors @xcite . however , as mentioned in @xmath484.2 , the systematically higher detection rate of h@xmath1o ice absorption in ulirgs , as compared to less infrared luminous seyfert 2 galaxies , is naturally explained by a higher column density of nuclear dust in ulirgs . this explanation is supported by imaging data at @xmath1110 @xmath0 m : while emission from the nuclei of less infrared luminous seyfert 2 galaxies is argued to be optically thin at @xmath1110 @xmath0 m @xcite , that from the nuclei of ulirgs is estimated to be optically thick @xcite . the higher column densities of dust around buried agns in ulirgs can naturally produce infrared spectral energy distribution with cooler color than in less infrared luminous seyfert 2 galaxies . it is quite possible that many other ulirgs with cool far - infrared colors contain energetically important agns deeply buried in dust . @xcite have found that a 60 @xmath0 m luminosity function in the local universe , derived based on @xmath1115000 galaxies , shows a bump at @xmath95 , and suggested that agn activity begins to be energetically important at @xmath96 . this result also favors the buried agn scenario in the majority of nearby ulirgs . strong ( @xmath97 ) 3.1 @xmath0 m h@xmath1o ice absorption has been detected toward the nuclei of five ulirgs : ugc 5101 , iras 00188@xmath360856 , iras 03250 + 1606 , iras 16487 + 5447 , and iras 21329@xmath362346 . all these ulirgs have cool far - infrared colors and are classified optically as liners . this detection requires a foreground screen dust geometry , which is best explained if centrally - concentrated , compact energy sources are buried in nuclear dust and molecular gas . for all the ulirgs , the observed equivalent widths of the 3.3 @xmath0 m pah emission are significantly lower than those of starburst galaxies . these observational data are naturally explained by the presence of agns deeply buried in dust , and support our previous arguments that liner ulirgs with cool far - infrared colors may possess energetically important , buried agns . we are grateful to drs . j. rayner , b. golisch , h. terada , k. aoki , b. potter , d. scarla , and s. harasawa for their support during our irtf and subaru observing runs . we also thank dr . c. c. dudley for his useful comments , and dr . y. gao for personally showing his preprint prior to publication . prm is supported by the nsf under grant ast-9900871 . this research made use of the nasa / ipac extragalactic database ( ned ) , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . ables , j. g. et al . 1987 , mnras , 226 , 157 alonso - herrero , a. , quillen , a. c. , simpson , c. , efstathiou , a. & ward , m. j. 2001 , aj , 121 , 1369 antonucci , r. r. j. 2001 , tenerife lectures ( astro - ph/0103048 ) barger , a. j. , cowie , l. l. , & richards , e. a. 2000 , aj , 119 , 2092 barvainis , r. , maloney , p. , antonucci , r. , & alloin , d. 1997 , apj , 484 , 695 blain , a. w. , smail , i. , ivison , r. j. , & kneib , j. -p . 1999 , mnras , 302 , 632 brock , d. , joy , m. , lester , d. f. , harvey , p. m. , & ellis , h. b. jr . , 1988 , apj , 329 , 208 de grijp , m. h. k. , miley , g. k. , & lub , j. 1987 , a&as , 70 , 95 downes , d. , & solomon , p. m. 1998 , apj , 507 , 615 dudley , c. c. 1999 , mnras , 307 , 553 dudley , c. c. , & wynn - williams , c. g. 1997 , apj , 488 , 720 dudley , c. c. , & wynn - williams , c. g. 1999 , mnras , 304 , 549 eiroa , c. , & hodapp , k. -w . 1989 , a&a , 210 , 345 fischer , j. 2000 , in _ iso _ beyond the peaks , ed . a. salama , m. f. kessler , k. leech , & b. schulz ( esa sp-456 ; noordwijk : esa ) , 239 ( astro - ph/0009395 ) gao , y. , & solomon , m. 2002 , apj , submitted genzel , r. et al . 1998 , apj , 498 , 579 imanishi , m. 2000 , mnras , 319 , 331 imanishi , m. 2002 , apj , 569 , 44 imanishi , m. , & dudley , c. c. 2000 , apj , 545 , 701 imanishi , m. , dudley , c. c. , & maloney , p. r. 2001 , apj , 558 , l93 kim , d. -c . , veilleux , s. , & sanders , d. b. , apjs , 143 , in press ( astro - ph/0207373 ) kobayashi , n. , et al . 2000 , ircs : infrared camera and spectrograph for the subaru telescope , in proc . spie 4008 : optical and ir telescope instrumentation and detectors , eds m. iye & a. f. moorwood , 1056 lutz , d. et al . 1996 , a&a , 315 , l269 lutz , d. , veilleux , s. , & genzel , r. 1999 , apj , 517 , l13 maloney , p. r. 1999 , apss , 266 , 207 maloney , p. , hollenbach , d. , & tielens , a. g. g. m. 1996 , apj , 466 , 561 mcleod , k. k. , rieke , g. h. , rieke , m. j. , & kelly , d. m. 1993 , apj , 412 , 111 mennella , v. , munoz caro , g. m. , ruiterkamp , r. , schutte , w. a. , greenberg , j. m. , brucato , j. r. , & colangeli , l. 2001 , a&a , 367 , 355 meurer , g. r. , heckman , t. m. , lehnert , m. d. , leitherer , c. , & lowenthal , j. 1997 , aj , 114 , 54 moorwood , a. f. m. 1986 , a&a , 166 , 4 mouri , h. , kawara , k. , taniguchi , y. , & nishida , m. 1990 , apj , 356 , l39 murakawa , k. , tamura , m. , & nagata , t. 2000 , apjs , 128 , 603 murphy , jr , t. w. , armus , l. , matthews , k. , soifer , b. t. , mazzarella , j. m. , shupe , d. l. , strauss , m. a. , & neugebauer , g. 1996 , aj , 111 , 1025 murphy , jr , t. w. , soifer , b. t. , matthews , k. , armus , l. , & kiger , j. r. 2001 , aj , 121 , 97 pendleton , y. j. , sandford , s. a. , allamandola , l. j. , tielens , a. g. g. m. , & sellgren , k. 1994 , apj , 437 , 683 rieke , g. h. , & lebofsky , m. j. 1985 , apj , 288 , 618 roche p. f. , & aitken d. k. 1984 , mnras , 208 , 481 roche p. f. , & aitken d. k. 1985 , mnras , 215 , 425 roche , p. f. , aitken , d. k. , smith , c. h. , & ward , m. j. 1991 , mnras , 248 , 606 sakamoto , k. , scoville , n. z. , yun , m. s. , crosas , m. , genzel , r. , & tacconi , l. j. 1999 , apj , 514 , 68 sanders , d. b. , & mirabel , i. f. 1996 , ara&a , 34 , 749 sanders , d. b. , soifer , b. t. , elias , j. h. , neugebauer , g. , & matthews , k. 1988 , apj , 328 , l35 scoville , n. z. et al . 2000 , aj , 119 , 991 shure , m. a. , toomey , d. w. , rayner , j. t. , onaka , p. , & denault , a. j. 1994 , proc . spie , 2198 , 614 smith , r. g. 1993 , mnras , 264 , 587 smith , r. g. , sellgren , k. , & tokunaga , a. t. 1989 , apj , 344 , 413 smith , r. g. , sellgren , k. , & brooke , t. y. 1993 , mnras , 263 , 749 soifer , b. t. , sanders , d. b. , madore , b. f. , neugebauer , g. , danielson , g. e. , elias , j. h. , lonsdale , c. j. , & rice , w. l. 1987 , apj , 320 , 238 soifer , b. t. et al . 2000 , aj , 119 , 509 soifer , b. t. , neugebauer , g. , matthews , k. , egami , e. , weinberger , a. j. , 2002 , aj , in press ( astro - ph/0207444 ) solomon , p. m. 2001 , in starburst galaxies : near and far , proceedings of a workshop held at ringberg castle , ed . l. tacconi , & d. lutz ( heidelberg : springer - verlag ) , 173 ( astro - ph/0101482 ) solomon , p. m. , downes , d. , & radford , s. j. e. 1992 , apj , 387 , l55 spoon , h. w. w. , koornneef , j. , moorwood , a. f. m. , lutz , d. , & tielens , a. g. g. m. 2000 , a&a , 357 , 898 spoon , h. w. w. , keane , j. v. , tielens , a. g. g. m. , lutz , d. , moorwood , a. f. m. , & laurent , o. 2002 , a&a , 385 , 1022 sturm , e. , lutz , d. , tran , d. , feuchtgruber , h. , genzel , r. , kunze , d. , moorwood , a. f. m. , & thornley , m. d. 2000 , a&a , 358 , 481 surace , j. a. , & sanders , d. b. 1999 , apj , 512 , 162 surace , j. a. , sanders , d. b. , & evans , a. s. 2000 , apj , 529 , 170 takeuchi , t. t. , yoshikawa , k. , & ishii , t. t. 2002 , apj , submitted tanaka , m. , sato , s. , nagata , t. , & yamamoto , t. 1990 , apj , 352 , 724 taniguchi , y. , yoshino , a. , ohyama , y. , & nishiura , s. 1999 , apj , 514 , 660 tokunaga a. t. , sellgren k. , smith r. g. , nagata t. , sakata a. , nakada y. , 1991 , apj , 380 , 452 tran , q. d. et al . 2001 , apj , 552 , 527 trung , d- , v- , lo , k. y. , kim , d. -c . , gao , y. , & gruendl , r. a. 2001 , apj , 556 , 141 veilleux , s. , kim , d. -c . , sanders , d. b. , mazzarella , j. m. , & soifer , b. t. 1995 , apjs , 98 , 171 veilleux , s. , kim , d. -c . , & sanders , d. b. 1999 , apj , 522 , 113 whittet , d. c. b. , bode , m. f. , longmore , a. j. , adamson , a. j. , mcfadzean , a. d. , aitken , d. k. , & roche , p. f. 1988 , mnras , 233 , 321 lcrrrrccl ugc 5101 & 0.040 & 0.25 & 1.03 & 11.54 & 20.23 & 45.5 & 45.5 & 0.09 ( cool ) + iras 00188@xmath360856 & 0.128 & @xmath40.12 & 0.37 & 2.59 & 3.40 & 45.8 & 45.946.0 & 0.14 ( cool ) + iras 03250 + 1606 & 0.129 & @xmath40.10 & @xmath40.16 & 1.38 & 1.77 & 45.6 & 45.645.7 & @xmath40.12 ( cool ) + iras 16487 + 5447 & 0.104 & @xmath40.08 & 0.20 & 2.88 & 3.07 & 45.6 & 45.7 & 0.07 ( cool ) + iras 21329@xmath362346 & 0.125 & @xmath40.08 & @xmath40.16 & 1.65 & 2.22 & 45.6 & 45.645.7 & @xmath40.10 ( cool ) + iras 23234 + 0946 & 0.128 & @xmath40.06 & @xmath40.20 & 1.56 & 2.11 & 45.6 & 45.645.7 & @xmath40.13 ( cool ) + iras 23327 + 2913 & 0.107 & @xmath40.06 & 0.22 & 2.10 & 2.81 & 45.6 & 45.645.7 & 0.10 ( cool ) + llcclccc ugc 5101 & 2001 april 9 & 18 & 90 & hr 4112 & 3.4 & f8v & 6200 + iras 00188@xmath360856 & 2002 august 19 & 25 & 90 & hr 72 & 5.0 & g0v & 5930 + iras 03250 + 1606 & 2002 august 19 & 23.3 & 0 & hr 763 & 4.3 & f7v & 6240 + iras 16487 + 5447 & 2002 march 28 & 36 & 70 & hr 5949 & 6.3 & a0v & 9480 + iras 21329@xmath362346 & 2002 october 24 & 32 & 90 & hr 7898 & 4.5 & g8v & 5400 + iras 23234 + 0946 & 2002 october 24 & 32 & 120 & hr 8653 & 4.6 & g8iv & 5400 + iras 23327 + 2913 & 2002 october 24 & 24 & 90 & hr 8955 & 5.1 & f6v & 6400 + lcccc ugc 5101 & 16@xmath983 & [email protected] & @xmath111 & @xmath1135 + iras 00188@xmath360856 & [email protected] & 13@xmath981 & @xmath111 & @xmath1150 + iras 03250 + 1606 & [email protected] & 12@xmath981 & @xmath113 & @xmath1180 + iras 16487 + 5447 & [email protected] & [email protected] & @xmath111 & @xmath1170 + iras 21329@xmath362346 & [email protected] & [email protected] & @xmath111 & @xmath1150 + iras 23234 + 0946 & [email protected] & [email protected] & @xmath112 & @xmath1170 + iras 23327 + 2913 & [email protected] & [email protected] & @xmath111 & @xmath1150 + note . ( 1 ) : object name . ( 2 ) : observed flux of the 3.3 @xmath0 m pah emission . ( 3 ) : observed luminosity of the 3.3 @xmath0 m pah emission . col . ( 4 ) : observed 3.3 @xmath0 m pah - to - infrared luminosity ratio . ( 5 ) : rest - frame equivalent width of the 3.3 @xmath0 m pah emission . those for starbursts are @xmath11120 nm ( see text ) . lcc ugc 5101 & @xmath550.8 & 0.65 + iras 00188@xmath360856 & @xmath551.6 & @xmath40.2 + iras 03250 + 1606 & @xmath550.6 & @xmath40.1 + iras 16487 + 5447 & @xmath550.8 & @xmath40.35 + iras 21329@xmath362346 & @xmath550.9 & @xmath40.2 +	ground - based 2.84.1 @xmath0 m slit spectra of the nuclei of seven ultraluminous infrared galaxies ( ulirgs ) that are classified optically as liners and have cool far - infrared colors are presented . all the nuclei show 3.3 @xmath0 m polycyclic aromatic hydrocarbon ( pah ) emission , with equivalent widths that are systematically lower than those in starburst galaxies . strong 3.1 @xmath0 m h@xmath1o ice absorption , with optical depth greater than 0.6 , is also detected in five nuclei , and 3.4 @xmath0 m carbonaceous dust absorption is detected clearly in one of the five nuclei . it is quantitatively demonstrated that the large optical depths of the h@xmath1o ice absorption in the five sources , and the 3.4 @xmath0 m absorption in one source , are incompatible with a geometry in which the energy sources are spatially mixed with dust and molecular gas , as is expected for a typical starburst , but instead require that a large amount of nuclear dust ( including ice - covered grains ) and molecular gas be distributed in a screen in front of the 34 @xmath0 m continuum - emitting sources . this geometrical requirement can naturally be met if the energy sources are more centrally concentrated than the nuclear dust and molecular gas . the low equivalent widths of the pah emission compared to starbursts and the central concentration of the nuclear energy sources in these five ulirgs are best explained by the presence of energetically important active galactic nuclei deeply buried in dust and molecular gas .
the description of the motion of binary black holes within the einsteinian theory of gravity is quite a challenging problem and is likely to be finally solvable by numerical means only . however , it seems that the full success will occur in a far future . it has turned out in the past that analytical developments did always influence numerical investigations , hence further progress on the analytical side is certainly desirable . in this paper we shall present an analytical solution of truncated einstein equations for systems of non - spinning point - like objects , which shows many aspects of past numerical models . most of the numerical computations of quasi - stationary initial data for black - hole binaries are based on the assumption that the space metric is conformally flat @xcite . they often assume as well the existence of an approximate helical killing vector which permits to elaborate some manageable formulations ( also see @xcite ) . recently , the time - symmetric initial data of brill - lindquist @xcite were generalized by adding a non - conformally flat contribution that incorporates pieces of information provided by perturbative post - newtonian calculations @xcite . the present paper does not follow this line , but rather keeps the spirit of former numerical simulations . we essentially perform two simplifications : on the one hand , we adopt the often - used assumption of conformally flat space metric ; on the other hand , we attribute the non - linear gravitational field energy to point - like sources , which are modeled as dirac delta distributions . we solve the resulting truncated einstein equations analytically and get what we call , in reference to the point - like character of the source support , the _ skeleton _ solution . the calculations are carried out for @xmath0 bodies in a mathematically sound way by working from the beginning in @xmath1-dimensional space - time and taking the limit @xmath2 in the final expressions , so that the dirac delta distributions modeling the sources are consistently handled @xcite . our space metric and extrinsic curvature may be regarded as some generalization of the multiple black hole brill - lindquist initial value solution @xcite to arbitrary black - hole momenta preserving its original asymptotic structure . moreover , the energy function can be used as generator of the dynamical evolution of the system . we shall refer to it as the skeleton hamiltonian associated to the full arnowitt - deser - misner ( adm ) hamiltonian . it agrees with the exact hamiltonian _ ( i ) _ in the test body limit , _ ( ii ) _ at the first post - newtonian ( 1pn ) approximation , i.e. , at the first order in powers of @xmath3 ( @xmath4 denoting the speed of light ) beyond the newtonian dynamics , and _ ( iii ) _ in the limit of special relativity in absence of gravity . the skeleton gravitational field is given explicitly , as well as the equations of motion . as a first application , we shall investigate the location of the last stable ( circular ) orbit ( lso ) for binary systems . in order to reach a proper accuracy , the last - stable - orbit parameters are worked out up to the tenth post - newtonian ( 10pn ) approximation within the skeleton model . the comparison of second and third post - newtonian results derived from the skeleton model with the corresponding quantities computed in the full einstein theory delivers a quantitative measure for the missing terms in the skeleton approach . we are interested in a system of @xmath0 black holes interacting gravitationally . its description will be achieved within the adm hamiltonian framework using a @xmath1 splitting of space - time . the spatial indices are denoted by latin letters and vary from @xmath5 to @xmath6 , whereas the space - time indices are greek and vary from @xmath7 to @xmath6 . in a given chart , @xmath8 , the motion of the @xmath9th black hole is described by a curve @xmath10 , for @xmath11 . the conjugate momentum of the position @xmath12 is called @xmath13 . the constant mass parameter of the @xmath9th black hole is denoted by @xmath14 . the hamiltonian variables associated to the gravitational field are taken to be ( i ) the space metric @xmath15 induced by the full space - time metric @xmath16 on the hypersurface @xmath17 , and ( ii ) its conjugate momentum @xmath18 where @xmath19 is the newtonian gravitational constant . the other components of @xmath16 are parametrized by the lapse and shift functions : @xmath20 and @xmath21 , with @xmath22 standing for the inverse of @xmath15 . it is also useful to introduce the shift vector @xmath23 . the spatial indices will be raised and lowered with the help of @xmath22 and @xmath15 , respectively . we now assume that the `` center - of - field '' @xmath24 of object @xmath9 can be modeled in the fictitious space - time associated with the flat metric by means of the dirac distribution @xmath25 . this prescription is known to lead to the brill - lindquist hamiltonian when the black hole velocities are set to zero , provided the mass parameters @xmath14 , @xmath11 , are identified with the brill - lindquist masses @xcite . it also leads to the correct 2pn dynamics @xcite for compact binaries , which can itself be recovered by means of the extended - body approach @xcite . moreover , the singularities generated by the distributions @xmath26 have shown to be curable by dimensional regularization at the 3pn order @xcite . similarly , we shall see in the present work that , for an appropriate choice of the space dimension @xmath6 , the @xmath26-type sources do not generate any ambiguity , so that the 3-dimensional quantities can be obtained by dimensional regularization . finally , this point - particle approach allows to bypass delicate investigations about the behavior of the gravitational field near the objects . in an asymptotically flat space - time of dimension @xmath27 , the hamiltonian of @xmath0 point masses provided by the adm formalism reads @xcite @xmath28 where @xmath29 . the super - hamiltonian density @xmath30 appearing under the integral sign is defined by @xmath31 while the supermomentum density @xmath32 is given by @xmath33 in eq . @xmath34 denotes the space curvature of the hypersurface @xmath17 and @xmath35 is the determinant of the matrix @xmath15 . as the lapse and shift functions are lagrange multipliers , the variation of the hamiltonian with respect to them leads to the constraint equations @xmath36 before solving eqs . , we must first fix the coordinate system . it is convenient to work in the so - called adm transverse trace - free ( admtt ) gauge @xcite defined by the two conditions @xmath37 in this gauge , the metric decomposes into its trace @xmath38 , which may be parametrized as @xmath39 ( in order to get a simple expression for the curvature density ) , and a transverse trace - free part @xmath40 . we thus have @xmath41 \pi^{ii } = 0.\end{gathered}\ ] ] the momentum @xmath42 needs itself to be split into a transverse trace - free contribution @xmath43 , and a rest @xmath44 , @xmath45 these two terms are uniquely defined assuming they decay as @xmath46 at spatial infinity and are given as a combination of derivatives of the poisson integral @xmath47 , and the poisson integral of the latter quantity , namely @xmath48 . we can then try to compute @xmath49 and @xmath44 from the two constraint equations supplemented by the boundary conditions at spatial infinity resulting from the asymptotic flatness of space - time ; notably @xmath50 , @xmath51 . it is actually more convenient to solve the hamiltonian constraint for the quantity @xmath52 rather than @xmath49 to get rid off the spurious constant at spatial infinity . we thus have @xmath53 after solving the constraint equations , we insert the ensuing expressions for @xmath54 and @xmath44 into the right - hand side of eq . . we get a _ reduced _ hamiltonian depending on @xmath12 , @xmath13 , @xmath55 , and @xmath43 only : @xmath56 \nonumber\\ & = \frac{c^4}{16 \pi g } \int d^d\mathbf{x}~\partial_i ( \partial_j \gamma_{ij}-\partial_i \gamma_{jj } ) \nonumber\\ & = -\frac{c^4}{16 \pi g } \int d^d\mathbf { x}~\delta \phi.\end{aligned}\ ] ] this hamiltonian contains the full information about the evolution of the matter @xmath57 and field @xmath58 variables @xcite . as the change of variables @xmath59 is not canonical , the poisson brackets of @xmath55 and @xmath43 take a form involving a transverse trace - free projection operator @xcite . the reduced hamiltonian @xmath60 provides the equations of motion of the @xmath0 bodies together with the evolution equations of the gravitational field . as they are strongly coupled it is hopeless to solve them analytically , but they can be solved perturbatively using , notably , the post - newtonian iterative scheme @xcite . the resulting dynamics is highly reliable as long as the typical speeds remain smaller than , say , about @xmath61 . when binary black holes reach their last stable orbits , the reliability of the approximation breaks down since the system enters a strong field regime . the only way to understand precisely this crucial stage is to perform numerical simulations in full general relativity . our purpose here is to search for a simple non - perturbative model applicable to arbitrary strong fields , eventually leaving the einsteinian theory but always staying close enough to it , in order to give an element of comparison with numerical investigations . the space metric obtained in our approach should also be usable as initial condition . we shall adopt here a wilson - mathews - type prescription @xcite , setting the non - conformally flat part of @xmath15 , namely @xmath40 , equal to zero . as @xmath55 is of order @xmath62 , the difference between our @xmath15 and the einsteinian space metric shows up at the 2pn level . for spherically symmetric space - times the difference is zero . in the case where the present model is used to fix initial cauchy data , the condition of a conformally flat space metric needs to hold on only one hypersurface , say @xmath63 . choosing @xmath55 to be zero on @xmath64 is always possible , though it is not physically realistic in general for isolated self - gravitating systems . our attitude in this instance will be to accept the resulting error . in return , we shall be able to determine the field @xmath15 without assuming zero velocities of the bodies at initial time . we thus demand here that @xmath65 and the problem reduces to solve the constraints @xmath66 , @xmath67 under this condition . the space metric @xmath15 is then proportional to @xmath68 , which implies that in the supermomentum @xmath32 the term @xmath69 vanishes [ cf . eq . ] . the field momentum @xmath42 relates to @xmath70 through the identities @xmath71 in particular , @xmath70 is symmetric and trace - free ( stf ) as is @xmath42 . the determinant @xmath35 appearing in @xmath30 is merely @xmath72 finally , the 3-curvature density @xmath73 takes a remarkably simple form @xcite : @xmath74 collecting these relations together with the constraint equations , we find : [ eq : constraint_system ] @xmath75 & -\frac{16\pi g}{c^2 } \sum_{a=1}^\mathcal{n } m_a \delta_a \psi^{-1 } \left ( 1 + \psi^{-\frac{4}{d-2 } } \frac{p_a^2}{m_a^2 c^2 } \right)^{\frac{1}{2 } } \ , , \\ \label{eq : constraint_momentum } \partial_j\pi^j_{\,i } = & -\frac{8\pi g}{c^3 } \sum_{a=1}^\mathcal{n } p_{ai } \delta_a \ , , \end{aligned}\ ] ] with @xmath76 . in the system the second equation decouples from the first one and is solved by searching for a particular solution proportional to the symmetric trace - free part of the derivative of a certain vector potential @xmath77 ; namely , @xmath78 the equation for the divergence @xmath79 is obtained by applying the operator @xmath80 simultaneously on the two sides of , hence @xmath81 . this yields immediately @xmath82 , and from the constraint it follows that : @xmath83 we now notice that the @xmath6-dimensional laplacian of the @xmath84th power of @xmath85^{\frac{1}{2}}$ ] satisfies @xmath86 in the sense of functions , and that , in the sense of distributions , @xmath87 as a result we may choose : @xmath88 which is nothing but the 1pn vector potential @xmath89 of ref . @xcite verifying @xmath90 . our second simplification consists in stating that the homogeneous solution of eq . , @xmath91 , is precisely zero , which implies @xmath92 . when this prescription is imposed on one unique hypersurface , it merely represents a particular choice of initial data . the resulting field momentum @xmath93 is identical to the rescaled extrinsic curvature proposed by bowen and york in their time - asymmetric formulation of the cauchy problem @xcite and adopted e.g. in the puncture method @xcite or various calculations of the last stable circular orbit @xcite . although it is a mathematical solution of the constraint equations for a conformally flat metric , it is not realistic in the sense that it does not result from an evolution process in absence of incoming radiation . it does not coincide either with the wilson - mathews momentum @xmath94 derived from the condition @xmath95 : @xmath96 where the quantity @xmath97 , defined as the @xmath98 part of the solution [ matching the form ( 3.6 ) ] of the equation @xmath99 , is explicitly given in paper @xcite for @xmath100 . the difference @xmath101 is only of order @xmath98 , which corresponds to the 2pn approximation . on the hamiltonian level , it is of 3pn order only . one might then argue that the wilson - mathews condition @xmath102 , valid for all times , i.e. particularly @xmath103 , is in contradiction to the condition @xmath104 . however , it is not the case because the momentum constraint equation is linear in @xmath105 and decouples from any other field equation when the space metric is conformally flat , so that the quantity obtained by adding an arbitrary homogeneous solution to @xmath70 still satisfies the constraint . as our @xmath70 differs from @xmath106 by such a homogeneous solution ( i.e. a transverse - trace free object ) , there is no contradiction between the two approaches . the integration of the hamiltonian constraint is much more involved , principally because of the appearance of @xmath54 in the source and the presence of a non - compact support term . on the contrary , the compact support term is harmless . it is made of a sum of dirac distributions @xmath26 that formally determine the boundary conditions near the black holes . by subtracting from the conformal factor the @xmath107 terms entering @xmath54 , we get an object that satisfies an equation whose source is entirely free of @xmath26 s ( this is the essence of paper @xcite ) . by contrast , we shall keep here the constraint equation under its original form in order to determine the dependence of the coefficient of @xmath107 on the positions @xmath12 and momenta @xmath13 . to solve the hamiltonian constraint iteratively , we have to face the delicate computation of @xmath108 . the expansion of the prefactor @xmath109 in powers of @xmath54 is known to give rise to poles when @xmath2 at the level corresponding to the 3pn approximation @xcite . in full general relativity , these poles cancel in the hamiltonian with similar quantities coming from @xmath55 , but such a cancellation can not occur in our case . in particular , _ the 3pn approximation of the wilson - mathews model does not exist _ for binary black - holes . we must therefore treat the @xmath110 term differently . let us first note that this term enters the space metric at the 3pn order only , since @xmath111 , but contributes to the 1pn hamiltonian [ because of the global factor @xmath112 in eq . ] . to be consistent with the accuracy of the former assumption @xmath102 which amounts to neglect some 2pn corrections in the hamiltonian , we have to keep at least the leading term in the post - newtonian expansion of @xmath113 , modulo a possible total space derivative . now , the tensor density @xmath70 is itself proportional to the symmetric trace - free part of the space derivative @xmath114 , so that @xmath115 as a consequence , the latter expression can be split into a `` skeleton '' term @xmath116 with compact support , and a `` flesh '' term @xmath117 which takes into account the field between the point particles . the flesh term contributes to the hamiltonian through the quantity : @xmath118 we are allowed to regard it as negligible in our scheme and keep the skeleton term alone in the decomposition of @xmath119 . our third simplification consits thus in performing the substitution @xmath120 = \frac{16 \pi g}{c^3 } \psi^\frac{2 - 3d}{d-2 } \sum_{a=1}^\mathcal{n } p_{aj } v_j \delta_a \ , . \end{gathered}\ ] ] after the replacement is made , the right - hand side of eq . takes the form @xmath121 , where the @xmath122 s are some unknown functions . for @xmath6 belonging to an appropriate range of values , @xmath122 is regular at @xmath123 , and @xmath124 . we see that in our model the source of the hamiltonian constraint reduces to some `` skeleton '' made of a linear combination of dirac distributions . for this reason , all quantities computed in the present approximation will be referred to as skeleton quantities . defining the @xmath0 quantities @xmath125 by the relation @xmath126 the function @xmath54 can be written as @xmath127 and the conformal factor reads @xmath128 where we have posed , in order to lighten the notation , @xmath129 it is also useful to introduce the ( regularized ) value of @xmath49 at the particle position @xmath130 : @xmath131 where @xmath132^\frac{1}{2}$ ] . after making the replacement on the right - hand side of eq . , inserting there expressions and , we arrive at @xmath133^{\frac{1}{2 } } \nonumber\\ & + \bigg(1 + a_d \sum_{b \neq a } \frac{\alpha_b}{r_{ab}^{d-2 } } \bigg)^{\frac{2 - 3d}{d-2 } } \frac{p_{ai } v_{ai}}{c } \bigg\ } \delta_a,\end{aligned}\ ] ] with @xmath134 . making use of eq . , we see that @xmath135 because of the linear independence of the distributions @xmath26 , the hamiltonian constraint is finally equivalent to a system of @xmath0 algebraic equations which reads @xmath136^\frac{1}{2 } + \frac{p_{ai } v_{ai}/c}{\left(1 + a_d \sum_{b \neq a } \frac{\alpha_b}{r^{d-2}_{ab } } \right)^\frac{3d-2}{d-2 } } , \quad a=1,\,\ldots,\,\mathcal{n}.\ ] ] solving the above system provides in principle the values of all the @xmath125 s . these quantities represent some effective masses , as suggested by the approximate equality @xmath137 . they are functions of @xmath138 ( for @xmath139 ) and @xmath13 [ defined implicitly by eqs . ] . the skeleton hamiltonian reads @xmath140 which shows that the adm mass @xmath141 is just the sum of the @xmath0 effective masses @xmath125 . the hamiltonian depends only on the matter variables : @xmath142 , so that the system is automatically conservative . the space metric is deduced from eqs . , , and : @xmath143 by construction , @xmath144 and @xmath15 coincide with their counterparts in the einsteinian theory up to the 1pn order . the test - body limit is achieved by taking the mass of one black hole , say of label 1 , much larger than the masses of its companion : @xmath145 , where @xmath146 . to get the test - body hamiltonian exactly , one has to let formally @xmath147 go to infinity ( @xmath148 ) in the hamiltonian while keeping @xmath149 , @xmath146 , constant as well as all linear momenta @xmath13 . it follows that particularly the flesh term @xmath117 does not contribute to this approximation . in absence of gravity ( @xmath150 ) our hamitlonian has the form of the exact special - relativistic hamitlonian for the system of @xmath151 free point particles . when @xmath152 for all @xmath9 , we recover the brill - lindquist solution , for which eqs . reduce to @xmath153 . as a consequence our model also generalizes the brill - lindquist data . let us observe that the particle masses @xmath14 can not be equated to the adm masses computed at the spatial infinities of the @xmath0 black hole sheets when one or several momenta are non - zero . when such an identification is performed , it leads automatically to the same relation between @xmath125 and @xmath14 as in the zero momentum case , whereas the correct dynamics up to the 1pn order is known to be the one that derives from the hamiltonian determined by the set of eqs . with @xmath154 . the actual dependence of the effective masses on the @xmath14 s must take the kinetic energy of the system into account . this indicates a relative boost of the black hole sheets with respect to each others . the issue is complicated by the fact that the bowen - york data for one single black hole do not result from a boosted schwarzschild field @xcite . this point , being still rather unclear , needs to be investigated in detail . we do not intend to discuss it here in the adm formalism , the evolution of the space metric in @xmath1 dimensions is given by [ cf . ( 4.70 ) in ref . @xcite ] . ] @xmath155 where the operator @xmath156 denotes the space covariant derivative . its explicit action on the shift function @xmath157 after symmetrization is @xmath158 for the conformally flat 3-metric @xmath159 , we have @xmath160 , hence @xmath161 after replacing in the above equation @xmath162 and @xmath163 by the expressions following from eq . and dropping all contributions proportional to @xmath164 , we arrive at : @xmath165 the evolution equation for the field momenta @xmath42 is known in @xmath6 dimensions as well [ cf . ( 4.71 ) in ref . it reads @xmath166 + \frac{n}{\sqrt{\gamma } } \bigg [ \pi^{ij } \pi^k_{\ , k } - 2 \pi^i_{\ , k}\pi^{kj } + \frac{1}{2 } \gamma^{ij } \left(\pi^{kl } \pi_{kl } - \frac{1}{2 } ( \pi^k_{\,k})^2 \right ) \bigg ] \nonumber\\[1ex ] & - \left [ \pi^{kj } d_k n^i + \pi^{ki } d_k n^j - d_k ( \pi^{ij } n^k ) \right ] + \frac{8\pi g}{c^4 } n \sum_{a = 1}^\mathcal{n } \frac{p_{ak}p_{al}}{m_a } \gamma^{ik}\gamma^{jl } \left(1+\frac{p_{am}p_{an}}{m_a^2 c^2}\gamma^{mn}\right)^{-\frac{1}{2 } } \delta_a \,.\end{aligned}\ ] ] in the adm gauge , the trace ( with respect to flat metric @xmath68 ) of reduces to @xmath167 + \frac{8\pi g}{c^4 } n \psi^{-\frac{8}{d-2 } } \sum_{a = 1}^\mathcal{n } \frac{p_{a}^2}{m_a } \left(1+\frac{p_a^2}{m_a^2 c^2}\psi^{-\frac{4}{d-2}}\right)^{-\frac{1}{2 } } \delta_a = 0 \ , . \end{gathered}\ ] ] making now use of eqs . , , , and of the two useful relations @xmath168 and @xmath169 , eq . can be rewritten as @xmath170 + \frac{8\pi g}{c^4 } \psi^{-\frac{d+2}{d-2 } } n \sum_{a = 1}^\mathcal{n } \frac{p_a^2}{m_a } \left(1+\frac{p_a^2}{m_a^2 c^2}\psi^{-\frac{4}{d-2}}\right)^{-\frac{1}{2 } } \delta_a = 0 \,.\end{gathered}\ ] ] the form of the latter equation suggests to solve it for the auxiliary function @xmath171 rather than for the lapse @xmath172 @xcite . at this stage , we apply to the right - hand side of eq . the replacement which amounts to deleting a flesh term in the product @xmath173 . since the newtonian potential comes from the @xmath3 part of the lapse , the term @xmath174 is indeed a 2pn quantity and may be neglected consistently with our approximation . finally , we obtain a poisson equation for @xmath175 with the source term being some linear combination of the dirac deltas : @xmath176 + m_a \psi_a^{-2 } \bigg(1+\frac{p_a^2}{m_a^2c^2}\psi_a^{-\frac{4}{d-2}}\bigg)^{-\frac{1}{2 } } \left ( 1 + \frac{d}{d-2}\frac{p_a^2}{m_a^2c^2}\psi_a^{-\frac{4}{d-2 } } \right ) \bigg ) \delta_a \ , . \end{gathered}\ ] ] following the same argument as in the derivation of the conformal factor in sec . [ sec : skeleton ] , we may write @xmath177 so that @xmath178 . we next replace the left - hand side of eq . by virtue of the latter relation , and equate the coefficients of @xmath26 . we are led to @xmath179 & \quad + m_a \psi_a^{-2 } \left(1+\frac{p_a^2}{m_a^2c^2 } \psi_a^{-\frac{4}{d - 2}}\right)^{-\frac{1}{2 } } \nonumber\\[1ex ] & \qquad \times \left(1 + \frac{d}{d - 2 } \frac{p_a^2}{m_a^2c^2 } \psi_a^{-\frac{4}{d - 2 } } \right ) \bigg ) , \quad a=1,\,\ldots,\,\mathcal{n}.\end{aligned}\ ] ] this system of @xmath0 equations for @xmath0 unknowns provides in principle the values of the monopoles @xmath180 entering the expression of @xmath175 , eq . . the lapse function itself is then given by @xmath181 for @xmath182 and @xmath100 , we recover the result of ref . @xcite in the time - symmetric limit @xmath152 ( with @xmath183 ) . in order to solve eq . for the shift function @xmath184 , we take the first and second order space derivatives of both sides . we obtain : @xmath185 \label{eq : equation_divergence_ni_wilson } \frac{d-1}{d } \delta(\partial_i n^i ) & = - \partial_{ij } \left(\psi^{-\frac{2d}{d-2 } } n \pi^i_{\ , j } \right ) \ , . \end{aligned}\ ] ] the source term of the second equation is the sum of a rational function of @xmath186 , @xmath11 , namely @xmath187 , plus a distribution , @xmath188 . the poisson integral of the term with non - compact support is not explicitly known , but that of the distributional term is easy to compute . now , the new `` flesh '' term does not contribute at the newtonian level . it is therefore at least post - newtonian and will eventually play a role in the dynamics only at the 2pn order ; it will be consistently neglected in the present model . thus , the substitution will be supplemented by the replacement : @xmath189 = -\frac{8\pi g}{c^3 } \psi^{-\frac{2d}{d-2 } } n \sum_{a=1}^\mathcal{n } p_{ai } \delta_a \,.\end{gathered}\ ] ] taking eq . into account , eq . becomes @xmath190 the solution of can be obtained with the help of formula . it reads @xmath191 & \times \sum_{a=1}^\mathcal{n } \chi_a \psi_a^{-\frac{3d-2}{d-2 } } p_{ai } \partial_i \left ( \frac{1}{r_a^{d-2 } } \right ) \,.\end{aligned}\ ] ] the poisson equation after inserting the above expression for @xmath192 and performing the substitution takes the form @xmath193 & \quad \times \sum_{a=1}^\mathcal{n } \chi_a \psi_a^{-\frac{3d-2}{d-2 } } p_{aj}\partial_{ij}\left(\frac{1}{r_a^{d-2}}\right ) \nonumber\\[1ex ] & + \frac{16\pi g}{c^3 } \sum_{a=1}^\mathcal{n } \chi_a \psi_a^{-\frac{3d-2}{d-2 } } p_{ai } \delta_a \,.\end{aligned}\ ] ] by virtue of the formulas and , the solution of eq . is simply @xmath194 \times \bigg ( \frac{d-2}{(4-d)(d-1 ) } p_{aj } \partial_{ij } r_a^{4-d } - 4 \frac{p_{ai}}{r_a^{d-2 } } \bigg ) \,.\end{gathered}\ ] ] for a black - hole binary , the number of coupled equations composing the system that gives the effective masses reduces to two , and the sum @xmath195 contains only one term . it is now convenient to regard the two quantities @xmath196 as the unknowns : @xmath197 a = 1,2 , \quad b \neq a.\end{gathered}\ ] ] let us restrict ourselves to the space dimension @xmath198 . the constant @xmath199 reads then @xmath200 , and the system consists of two equations for the two unknowns @xmath201 : @xmath202 \quad b=1,2 , \quad a \neq b \ , .\end{gathered}\ ] ] it is clearly possible to decouple the two equations . in appendix [ app : polynomial ] we show that each @xmath201 is a root of a polynomial of order 200 . the equations of motion of a black - hole binary read [ eq : generic_equations_of_motion ] @xmath203 \dot{p}_{ai } & = -\frac{\partial h}{\partial x_{a}^i},\end{aligned}\ ] ] where the skeleton hamiltonian @xmath144 expressed in terms of the quantities @xmath196 equals @xmath204 using the theorem of implicit differentiation , we put eqs . in a form apparently more convenient for numerical integration . we start by differentiating eq . with respect to the positions . we find ( here @xmath205 and @xmath206 ) : @xmath207 \label{eq : psi_derivative_2_equation } \partial_{2i}\psi_b & \equiv \frac{\partial\psi_b}{\partial x_2^i } = \tilde{\eta}_{a2}^i - \zeta_a \partial_{2i } \psi_a,\end{aligned}\ ] ] where [ eq : tilde_variables ] @xmath208 & + \frac{g \partial_{1i } ( p_{aj } v_{aj}/r_{12})}{2 c^3 \psi_a^7 } , \\[1ex ] \label{eq : eta_a2 } \tilde{\eta}_{a2}^i \equiv & -\tilde{\eta}_{a1}^i , \\[1ex ] \label{eq : zeta_a } \zeta_a \equiv & -\frac{\partial \psi_b(x^i_c , p_{ci},\psi_c)}{\partial \psi_a } \nonumber\\[1ex ] = & \frac{g m_a}{2 c^2 r_{12 } \psi_a^2 } \frac{1 + 3 p_a^2/(m_a^2 c^2 \psi_a^4 ) } { [ 1 + p_a^2/(m_a^2 c^2 \psi_a^4)]^{\frac{1}{2 } } } \nonumber\\[1ex ] & + \frac{7 g \ , p_{aj } v_{aj}}{2 c^3 r_{12 } \psi_a^8 } \nonumber \\[1ex ] = & \frac{\chi_b - 1}{\chi_a}.\end{aligned}\ ] ] equality follows from the fact that the right - hand side of relation depends on @xmath209 and @xmath210 only through the black - hole separation @xmath211^\frac{1}{2}$ ] , and that we have @xmath212 . solving eqs . with respect to @xmath213 and eqs . with respect to @xmath214 , we find [ eq : psi_derivatives ] @xmath215 \partial_{2i}\psi_b & = -\partial_{1i}\psi_b \ , .\end{aligned}\ ] ] it remains to evaluate the momentum derivatives @xmath216 and @xmath217 . the system of equations for @xmath218 is given by @xmath219 where we have posed : @xmath220 & \quad + \frac{g\,\partial_{p_{1i}}(p_{aj}v_{aj})}{2 c^3 r_{12 } \psi_a^7}.\end{aligned}\ ] ] the coefficients @xmath221 are identical to those appearing in @xmath213 because they represent the result of the differentiation of the right - hand side of relation with respect to @xmath196 at @xmath12 and @xmath13 constant . the solution of eqs . with respect to @xmath218 reads @xmath222 from which we deduce @xmath223 by exchanging the particle labels . it is convenient to introduce the rescaled variables @xmath224 making use of eqs . , , , and , the equations of motion can be written in the form [ eq : equations_of_motion ] @xmath225 \dot{p}_{1i } & = -\frac{c^2}{1 - \zeta_1\zeta_2 } \big ( ( 1 - \zeta_2 ) \eta_1^i - ( 1 - \zeta_1 ) \eta_2^i \big ) \nonumber\\[1ex ] & \quad - \frac{2 c^4}{g } ( \psi_1 + \psi_2 - 2 ) n_{12}^i \nonumber \\[1ex ] & = \frac{c^2}{1 - \zeta_1 \zeta_2 } \big((1 - \zeta_1 ) \zeta_2 \eta_1^i - ( 1 - \zeta_2 ) \zeta_1 \eta_2^i \big ) \nonumber \\[1ex ] & \quad - c \left(\frac{1}{\psi_1 ^ 7 } + \frac{1}{\psi_2 ^ 7}\right ) \partial_{1i } ( p_{1j } v_{1j } ) \,,\end{aligned}\ ] ] where the last equality follows from relations and definitions . similar equations for @xmath226 and @xmath227 hold with the role of labels 1 and 2 exchanged . knowing the positions and momenta at a given time @xmath228 , we can solve the algebraic equations numerically , calculate @xmath229 , @xmath230 and @xmath221 by means of eqs . , , , and determine @xmath231 as well as @xmath232 from the evolution equations . as an example of application of our skeleton hamiltonian , we shall discuss in this section the parameters of the last stable ( circular ) orbit ( lso ) in the relative motion of two inspiraling compact objects . this class of systems is of primary importance , being one of the most promising sources of detectable gravitational waves for ground - interferometry experiments as ligo ( laser interferometer gravitational wave observatory ) , virgo , geo600 , and tama300 . because of gravitational - wave emission , the separation between the bodies decreases adiabatically in time while the frequency increases according to the relativistic version of the kepler law . the process ends as soon as the first instable orbit is reached before the coalescence phase . the point where this occurs changes according to the model ( see refs . @xcite and references therein for some comparisons ) . for binary black holes the skeleton hamiltonian reads @xmath233 , where @xmath234 and @xmath235 are solutions of the coupled system of equations . it is impossible to solve this system analytically , but it is not difficult to solve it perturbatively within the post - newtonian setting , treating @xmath236 as a small parameter . at the @xmath237pn approximation , we search the post - newtonian expansion of the coefficients @xmath125 in the form @xmath238 note that we need to carry the series up to the order @xmath239 in @xmath240 to get an @xmath237pn - accurate dynamics with respect to the exact skeleton evolution ( which first differs from the exact einstein s evolution at the 2pn order ) . we now substitute into eqs . and expand their right - hand sides up to the order @xmath239 . by identifying the coefficients of different powers @xmath241 , @xmath242 , we obtain a system of algebraic equations for @xmath243 . we display here the first three equations : @xmath244 \frac{\alpha_1^{(1)}}{m_1 } & = - \frac{(p_{1i}p_{1i})^2}{8m_1 ^ 4 } + \frac{g m_2}{2r_{12 } } \left ( - \frac{5 p_{1i } p_{1i}}{2 m_1 ^ 2 } + \frac{7 p_{1i } p_{2i}}{2 m_1 m_2 } \right . \\[2ex ] & \quad \left . + \frac{n_{12}^i n_{12}^j p_{1i } p_{2j}}{2 m_1 m_2 } - \frac{\alpha_2^{(0)}}{m_2 } \right ) + \frac{g^2 m_2 ^ 2}{4 r_{12}^2 } , \\[2ex ] \frac{\alpha_1^{(2)}}{m_1 } & = \frac{(p_{1i } p_{1i})^3}{16 m_1 ^ 6 } + \frac{g m_2}{2 r_{12 } } \left(\frac{9 ( p_{1i}p_{1i})^2}{8 m_1 ^ 4 } - \frac{5 p_{1i}p_{1i } \alpha_2^{(0)}}{2 m_1 ^ 2 m_2 } \right . \\[2ex ] & \quad \left . - \frac{\alpha_2^{(1)}}{m_2 } \right ) + \frac{g^2 m_2 ^ 2}{2 r_{12}^2 } \left ( \frac{15 p_{1i } p_{1i}}{4 m_1 ^ 2 } - \frac{49 p_{1i } p_{2i}}{4 m_1 m_2 } \right . \\[2ex ] & \quad \left . - \frac{7 n_{12}^i n_{12}^j p_{1i } p_{2j}}{4 m_1 m_2 } + \frac{\alpha_2^{(0)}}{m_2 } \right ) - \frac{g^3 m_2 ^ 3}{8 r_{12}^3 } \,;\end{aligned}\ ] ] similar equalities hold with labels 1 and 2 exchanged . it is straightforward to solve up this system to the required order . for convenience , we shall consider instead of @xmath144 the dimensionless hamiltonian ( after ruling out the constant rest - mass contribution ) @xmath245 @xmath246 denoting the reduced mass of the system . at the @xmath237pn level , we have @xmath247 we restrict our study to circular orbits in the center - of - mass reference frame . within the adm gauge , the frame shift is achieved by imposing the condition @xmath248 whereas circularity requires that @xmath249 the form of @xmath250 becomes particularly simple when use is made of the reduced relative position and momentum : @xmath251 when both relations and hold , the hamiltonian can be expressed with the help of @xmath252 and @xmath253 only . notice also that it depends on the masses @xmath147 and @xmath254 exclusively through the symmetric mass ratio @xmath255 . in the next step , we introduce the conserved ( reduced ) angular momentum @xmath256 along the cirular orbit of radius @xmath257 : @xmath258 after eliminating the linear momentum @xmath253 by means of eq . , the dimensionless energy @xmath259 of the binary system becomes a function of the variables @xmath257 and @xmath256 : @xmath260 the coordinate radius @xmath257 of the circular orbit and the angular momentum @xmath256 are related through one of the canonical equations of motion , namely @xmath261 a relation that may be regarded as expressing the quasi - stationarity of the configuration . remarkably , it can be shown in this case , i.e. , for circular motion , that the adm mass of the system equates the so - called komar mass @xmath262 , i.e. , the mass - like quantity proportional to the monopole part of @xmath263 at spatial infinity satisfying the relation @xmath264 @xcite . the proof of the equality @xmath265 is given in appendix [ app : komar ] for an arbitrary number of space dimensions . equation is solved perturbatively for @xmath266 . by plugging the result into eq . , we find the gauge invariant link between the energy @xmath267 and the angular momentum @xmath256 in the case of circular motion : @xmath268 = & -\frac{1}{2j^2 } - \frac{1}{8j^4 } ( 9+\nu ) \varepsilon \nonumber\\[1ex ] & - \frac{1}{16j^6 } \big(81 + 41\nu-5\nu^2\big ) \varepsilon^2 + { \cal o}(\varepsilon^3).\end{aligned}\ ] ] it is convenient to rewrite formula in terms of the dimensionless variable @xmath269 the parameter @xmath270 is now treated as a small quantity of the order @xmath240 [ cf . the definition establishes the link between @xmath270 and @xmath256 . we invert it order by order to obtain @xmath256 as a function of @xmath270 . inserting @xmath271 into eq . leads to a gauge invariant expression of the energy @xmath267 depending on the parameter @xmath270 . the calculation is performed at the 10pn approximation in order to come reasonably close to the exact result . we display here the explicit formula with the 3pn - accuracy . it reads @xmath272 the higher order post - newtonian coefficients @xmath273 ( for @xmath274 ) can be found in appendix [ app:10pn ] . we want to compare the result of eq . with the relation @xmath275 of the wilson - mathews model ( remember that in this model the post - newtonian expansion terminates at the 2pn order ) in eq . ( c4 ) of ref . @xcite and replacing there the angular momentum @xmath256 by @xmath276 according to eq . ( b1 ) of ref . @xcite . ] : @xmath277 the 2pn coefficient proportional to @xmath278 differs from that of the post - newtonian computations carried out in general relativity . indeed , the energy for circular orbits in einsteinian theory , up to the 3pn order , is @xcite @xmath279 where , according to ref . @xcite we set the parameter @xmath280 ( defined in paper @xcite ) equal to zero ( @xmath281 ) , getting the solution corresponding to the brill - lindquist initial data . by comparing eqs . and it is obvious that our energy @xmath267 starts to differ from the general - relativistic one , @xmath282 , at the 2pn order . cdddddddddd & & & & & & & & & & + @xmath283 & 0.5224 & 0.1077 & 0.0725 & 0.0624 & 0.0583 & 0.0563 & 0.0554 & 0.0548 & 0.0546 & 0.0544 + @xmath267 & -0.1622 & -0.0689 & -0.0566 & -0.0529 & -0.0514 & -0.0508 & -0.0505 & -0.0503 & -0.0503 & -0.0502 the position @xmath284 of the last stable orbit is determined from the requirement that @xmath285 . the corresponding energy @xmath286 [ computed by means of eq . ] is depicted in fig . [ fig : ex_plot ] for equal - mass binaries ( i.e. , for @xmath287 ) as a function of the dimensionless orbital frequency @xmath288 . numerical values are given in table [ tab : ex_table ] . the case of a test particle moving in a schwarzschild background is also plotted for comparison . similarly to what happens in einsteinian theory @xcite , the instabilities arise `` earlier '' ( i.e. , at lower frequency ) for higher post - newtonian orders in the skeleton model . the 2pn last stable orbit of our skeleton dynamics is quite far from that of general relativity in the sense that the absolute difference @xmath289 ( @xmath290 ) between the skeleton frequency @xmath291 ( the skeleton energy @xmath292 ) and the einsteinian one @xmath293 ( @xmath294 ) is not negligible with respect to the general - relativistic value : @xmath295% ( @xmath296% ) . the location of the wilson - mathews data suggests that modifications of the extrinsic curvature affect the parameters of the last stable orbit more than the choice of a conformally flat space metric @xcite . the 3pn skeleton values are even further away from the einsteinian 3pn ones , but they are closer to those of the effective one - body approach improved by pad approximants than the straight 3pn einsteinian values . the 10pn data are particularly interesting , considering that they may be regarded as good approximants to the full skeleton model . remarkably , their position in the @xmath297 plane relative to the 3pn data is nearly the same as that of the point referring to numerical simulations relative to the point referring to the 3pn data in general relativity . in fig . [ fig : ex_plot ] , the relative location of the last stable orbit indicated by the white circle obtained from numerical computations with respect to the 3pn result for corotating bodies ( black triangle ) is somehow plausible if we believe that it is essentially due to the wilson - mathews truncation . it is indeed comparable to the relative location of irrotational 2pn einsteinian and wilson - mathews data . we also want to point out that a realistic estimation of the last stable orbit does not necessarily follow from the standard 3pn calculation . it may well be close to the star - like point derived from the effective one - body model with pad approximant . there is currently no definite argument to decide what is the most likely statement . as a result , we regard the position of these points as defining the present uncertainty in the location of the last stable orbit derived analytically . p.j . thanks the theoretisch - physikalisches institut of the fsu jena for hospitality during the realization of the present work . is thankful to the spring-2003 numerical relativity group at caltech , particularly greg cook , lee lindblom , niall murchadha , and masaru shibata , for stimulating discussions . we also thank christian thierfelder for useful comments about the presentation of the evolution equations . this work is supported by the eu programme `` improving the human research potential and the socio - economic knowledge base '' ( research training network contract hprn - ct-2000 - 00137 ) , by the caltech visitors program , and by the deutsche forschungsgemeinschaft ( dfg ) through sfb / tr7 `` gravitationswellenastronomie '' . got an additional support from the polish kbn grant no . 5 p03b 034 20 . equation for @xmath298 , after multiplication by @xmath299 , can be written in the form @xmath300 , where @xmath301 and @xmath302 ^ 2 [ \psi_1 ^ 4 + p_1 ^ 2/(m_1 ^ 2 c^2)]$ ] are polynomials in @xmath303 with non - zero constant terms , of order 7 and 4 , respectively . a similar relation holds for @xmath304 . we now subtract @xmath305 from the left and right - hand side , and square the new identity . we obtain : @xmath306 = \left(\frac{g m_2}{2 r_{12 } c^2 } \right)^2 \psi_2 ^ 8 \left(\psi_2 ^ 4 + \frac{p_2 ^ 2}{m_2 ^ 2 c^2}\right ) \,.\end{gathered}\ ] ] next , we multiply the equality ( [ eq : equation_gamma1 ] ) by @xmath307 , make the substitution @xmath308 , and expand the terms @xmath309 , @xmath310 , in powers of @xmath311 with the help of the binomial formula . if @xmath237 is even , the contribution @xmath312 , with @xmath313 $ ] , is of polynomial kind , whereas that of @xmath314 has the form of a polynomial @xmath315 multiplied by the algebraic function @xmath316 ; @xmath317 and @xmath315 have degrees @xmath318= 7n$ ] and @xmath319 = 7n - 3 $ ] , respectively . the same decomposition holds for odd @xmath237 with different expressions for @xmath320 and @xmath321 . at last , we gather all polynomial quantities into the first member of the equality while the non polynomial contributions @xmath322 are recast to the second member . the final equation is @xmath323 \nonumber\\ & = - p_{1 \ , ( 4)}^{1/2 } \bigg[(\psi_1 - 1)^2 \sum_{k=0}^6 \genfrac{(}{)}{0pt}{}{14}{2k + 1 } p_{1\ , ( 7)}^{2k+1 } \psi_1^{4 ( 13 - 2k ) } p_{1\ , ( 4)}^{6-k } - \frac{g p_{2i } v_{2i}}{r_{12 } c^3 } ( \psi_1 - 1 ) \psi_1^{49 } \sum_{k=0}^3 \genfrac{(}{)}{0pt}{}{7}{2k } p_{1\ , ( 7)}^{2k } \psi_1^{4 ( 7 - 2k ) } p_{1\ , ( 4)}^{3-k } \nonumber\\ & \qquad \qquad - \left(\frac{g m_2 \psi_1 ^ 7}{2 r_{12 } c^2 } \right)^2 \bigg(\sum_{k=0}^5 \genfrac{(}{)}{0pt}{}{12}{2k+1 } p_{1 \ , ( 7)}^{2k+1 } \psi_1^{4 ( 11 - 2k ) } p_{1 \ , ( 4)}^{5-k } + \frac{p_2 ^ 2 \psi_1^{28}}{m_2 ^ 2 c^2 } \sum_{k=0}^3 \genfrac{(}{)}{0pt}{}{8}{2k+1 } p_{1\ , ( 7)}^{2k+1 } \psi_1^{4 ( 7 - 2k ) } p_{1\ , ( 4)}^{3-k } \bigg ) \bigg ] \,.\end{aligned}\ ] ] we see that the left - hand side of eq . is made of a sum of polynomials of degrees 100 , 99 , 98 , 98 , 98 , the first of them involving the only non - zero constant term . the right - hand side is the product of @xmath322 and polynomials of degrees 97 , 96 , 95 , 95 respectively that cancel when @xmath324 . as a consequence , @xmath303 is the root of a polynomial of degree 200 with a non - zero constant term . in this appendix , we restrict ourselves to a binary system . the black holes are thus labeled by @xmath325 . our starting point is the system of two sets of equations satisfied by the monopoles @xmath125 and @xmath180 entering the conformal factor @xmath49 and the auxiliary function @xmath171 : [ eq : alpha_beta_binary ] @xmath326 & \quad + m_a \psi_a^{-1 } \left(1+\frac{p_a^2}{m_a^2c^2}\psi_a^{-\frac{4}{d-2}}\right)^{\frac{1}{2 } } \ , , \\[2ex ] \label{eq : beta_a_binary } \beta_a & = \chi_a \bigg ( \frac{3d-2}{d-2 } \frac{p_{ai}v_{ai}}{c } \psi_a^{-\frac{4(d-1)}{d-2 } } \nonumber\\[2ex ] & \quad + m_a \psi_a^{-2 } \left(1+\frac{p_a^2}{m_a^2c^2}\psi_a^{-\frac{4}{d-2}}\right)^{-\frac{1}{2 } } \nonumber\\[2ex ] & \qquad \times \left(1 + \frac{d}{d-2 } \frac{p_a^2}{m_a^2c^2 } \psi_a^{-\frac{4}{d-2 } } \right ) \bigg ) \ , , \end{aligned}\ ] ] with ( here @xmath327 ) [ eq : psi_chi_binary ] @xmath328 \label{eq : chi_a_binary } \chi_a & = 1 - a_d \frac{\beta_b}{r_{12}^{d - 2 } } \,.\end{aligned}\ ] ] it is straightforward to show that for a binary system described by the functions @xmath49 and @xmath175 of the purely monopolar form given by eqs . and , the adm mass coincides with its komar mass if and only if the quantities @xmath125 and @xmath180 satisfy the simple relation : @xmath329 we shall assume that the relative orbits are circular . after introducing the reduced relative momentum @xmath253 and eliminating its square through the relation @xmath330 , all quantities become functions of @xmath331 and @xmath256 only . by virtue of eqs . , the effective mass @xmath125 can then be treated as a function of the _ three _ variables @xmath196 , @xmath331 , and @xmath256 : @xmath332 with the help of eqs . , it is not difficult to check that the following relations , crucial for our proof , are fulfilled by @xmath125 : [ eq : alpha_a_derivative ] @xmath333 \label{eq : alpha_a_derivative_r12 } \frac{\partial \alpha_a(\psi_a , r_{12 } , j)}{\partial r_{12 } } & = \frac{d - 2}{2r_{12 } } \bigg(\alpha_a - \beta_a \frac{\psi_a}{\chi_a } \bigg)\ , . \end{aligned}\ ] ] note that , writing down eqs . , we have already employed the relations . the adm hamiltonian , restricted to circular orbits , reads @xmath334 the circularity condition implies @xmath335 , which is equivalent to @xmath336 after substituting into we obtain the equality ( valid along circular orbits ) : @xmath337 according to eq . , it can be rewritten as @xmath338 on the other hand , the equations in the circular case take the form @xmath339 by differentiating both sides with respect to @xmath331 and resorting to relations , we obtain the system of two equations ( where @xmath340 ) @xmath341 which we solve with respect to the derivatives @xmath342 . after eliminating @xmath196 and @xmath343 by means of relations , the solution reduces to @xmath344 to complete the proof of eq . , it is enough to substitute eqs . into the left - hand side of equality and make use of eqs . . the coefficients @xmath273 ( for @xmath274 ) entering formula for energy @xmath267 along cirular orbits as a function of the parameter @xmath270 read @xmath345 e_6 & = & \frac{45927}{1024 } + \frac{662619}{1024 } \nu - \frac{233737}{128 } \nu^2 + \frac{1162513}{1024 } \nu^3 + \frac{56481}{1024 } \nu^4 - \frac{3777}{1024 } \nu^5 , \\[2ex ] e_7 & = & \frac{264627}{2048 } + \frac{16343855}{6144 } \nu - \frac{136513223}{18432 } \nu^2 + \frac{67182599}{27648 } \nu^3 + \frac{249565613}{31104 } \nu^4 - \frac{766814851}{497664 } \nu^5 - \frac{148888223 } { 13436928 } \nu^6 , \\[2ex ] e_8 & = & \frac{12196899}{32768 } + \frac{993480761}{98304 } \nu - \frac{403684801}{16384 } \nu^2 - \frac{19713422131}{884736 } \nu^3 + \frac{532659868169}{3981312 } \nu^4 - \frac{1096304004055}{11943936 } \nu^5 \nonumber\\[2ex ] & & + \frac{42034682027}{17915904 } \nu^6 + \frac{29197027769}{644972544 } \nu^7 , \\[2ex ] e_9 & = & \frac{70366725}{65536 } + \frac{2394150375}{65536 } \nu - \frac{13003419935}{196608 } \nu^2 - \frac{9560242585}{32768 } \nu^3 + \frac{248408863105}{196608 } \nu^4 - \frac{129128080475}{98304 } \nu^5 \nonumber\\[2ex ] & & + \frac{30481497595}{98304 } \nu^6 + \frac{189766295}{16384 } \nu^7 + \frac{1599565 } { 65536 } \nu^8 , \\[2ex ] e_{10 } & = & \frac{813439341}{262144 } + \frac{100457157571}{786432 } \nu - \frac{69694465361}{589824 } \nu^2 - \frac{44773413415657}{21233664 } \nu^3 + \frac{558727148465215}{63700992 } \nu^4 \nonumber\\[2ex ] & & - \frac{2097054471026245}{191102976 } \nu^5 + \frac{500159286898463}{143327232 } \nu^6 + \frac{858674754130903}{2579890176 } \nu^7 - \frac{233795515077109}{5159780352 } \nu^8 \nonumber\\[2ex ] & & - \frac{177721652120353}{417942208512 } \nu^9 , \\[2ex ] e_{11 } & = & \frac{4710988269}{524288 } + \frac{228788885795}{524288 } \nu + \frac{411952929451}{4718592 } \nu^2 - \frac{28395688669829}{2359296 } \nu^3 + \frac{3163918218071455}{63700992 } \nu^4 \nonumber\\[2ex ] & & - \frac{8137959523310875}{127401984}\nu^5 + \frac{6129421057080619}{1146617856 } \nu^6 + \frac{9328338011915891}{322486272 } \nu^7 - \frac{180325112745666065}{30958682112 } \nu^8 \nonumber\\[2ex ] & & - \frac{11728912802084585}{278628139008 } \nu^9 + \frac{1098645689138995}{2507653251072 } \nu^{10}.\end{aligned}\ ] ]	an approximate analytical and non - linear solution of the einstein field equations is derived for a system of multiple non - rotating black holes . the associated space - time has the same asymptotic structure as the brill - lindquist initial data solution for multiple black holes . the system admits an arnowitt - deser - misner ( adm ) hamiltonian that can particularly evolve the brill - lindquist solution over finite time intervals . the gravitational field of this model may properly be referred to as a _ skeleton _ approximate solution of the einstein field equations . the approximation is based on a conformally flat truncation , which excludes gravitational radiation , as well as a removal of some additional gravitational field energy . after these two simplifications , only source terms proportional to dirac delta distributions remain in the constraint equations . the skeleton hamiltonian is exact in the test - body limit , it leads to the einsteinian dynamics up to the first post - newtonian approximation , and in the time - symmetric limit it gives the energy of the brill - lindquist solution exactly . the skeleton model for binary systems may be regarded as a kind of analytical counterpart to the numerical treatment of orbiting misner - lindquist binary black holes proposed by gourgoulhon , grandclment , and bonazzola , even if they actually treat the corotating case . along circular orbits , the two - black - hole skeleton solution is quasi - stationary and it fulfills the important property of equality of komar and adm masses . explicit calculations for the determination of the last stable circular orbit of the binary system are performed up to the tenth post - newtonian order within the skeleton model .
understanding atomistic phenomena at contacting surfaces is fundamental to the improvement of many modern applications , ranging from experimental methods like atomic or friction force microscopy ( afm / ffm ) @xcite or nanoindentation @xcite to nanotechnologies employed , for example , in nano-/microelectromechanical - systems ( nems / mems ) @xcite . particularly , the lifetime , performance , and reliability of such systems depend sensitively on the interfacial processes between contacting materials . furthermore , detailed studies of such interfaces allow for gaining valuable insights necessary for the description of tribological processes , like nanoscale wear @xcite , which is not yet understood comprehensively because of its highly complex nature @xcite . interfaces between metals and counterparts like ceramics @xcite or diamond @xcite are of high technological as well as fundamental interest because they possess benefits of both types of the involved material classes , such as high thermal stability , degradation resistance , and interesting mechanical properties @xcite . such interfaces are used for various applications ranging from protective coatings to communication devices and nanoelectronics @xcite . in this paper , two distinct material pairings are investigated , namely the metal - ceramic al / tin interface and the metal - insulator cu / diamond ( c@xmath0 ) interface . these interfaces are conceived as contacts between hard and soft materials . in reality , however , surfaces are usually not pristine . for example , when aluminum is exposed to air a thin oxide film is formed at the al surface . this passivation prevents further oxidation and results in an excellent corrosion resistance of the material @xcite . the typical thickness of such a film is up to . as an initial step towards such exceedingly complex interfaces , the effect of a monolayer of oxygen and hydrogen covering al and c@xmath0 surfaces , respectively , will be discussed in this work . beginning in the 1980s , classical molecular dynamics ( md ) simulations have become a commonly applied tool to investigate nanotribological phenomena and they are a part of the standard computational methods in this field nowadays , see , e.g. , refs . . during the last decade , additionally , the use of density functional theory ( dft ) calculations has been introduced in nanotribology , see , e.g. , refs . , and should be seen as an extension to the more common computational tools in tribology . advantages of dft include that it is independent of empirical potentials , i.e. , dft allows for parameter - free calculations via an accurate quantum - mechanical description of systems . on the other hand , dft calculations are currently limited to relatively small systems of typically a few hundred atoms because of computational challenges . since dft calculations yield very reliable results for the investigated class of systems @xcite , this method is employed here to analyze the electronic and atomic structure of the investigated interfaces , e.g. , to determine adhesion energies . results obtained with dft , such as potential - energy curves , can be used as a starting point for a multi - scale approach in which data is handed over to , e.g. , large - scale classical md simulations @xcite . in the last years also quantum - classical embedding techniques have been developed and improved allowing to treat crucial parts of a system with high accuracy methods such as dft , while most of the system is evaluated using less expensive methods . @xcite such joint approaches combined with the continuously increasing available computer power and advances in software tools promise the possibility to study even larger and more realistic systems in the near future . investigations on al / tin interfaces have been conducted by various researches using experimental @xcite as well as theoretical @xcite methods . the role of interfacial species at al / tin interfaces is , however , less studied . liu et al . @xcite and zhang et al . @xcite investigated the effects of hydrogen and zn as well as mg , respectively , on al / tin interfaces . both computational studies concluded that the interfacial adhesion is reduced by the additional species at the interface . here , the emphasis lies on the role of oxygen , since aluminum is usually covered by an oxide layer under ambient conditions . @xcite further information on oxide layers on aluminum surfaces can be found , e.g. , in refs . . as a first step towards a more detailed description of al / tin interfaces , the al slab is terminated by one layer of oxygen in the current work , which focuses on the possibility of material transfer and its ab - initio simulation . insights on copper / diamond interfaces with and without interfacial species have been presented by various researchers @xcite . cu / c@xmath0 will be used as an abbreviation below . a better understanding necessarily includes detailed work on the structure of clean and hydrogen terminated diamond surfaces . for cu / c@xmath0 interfaces , guo et al . @xcite studied the fracture behaviour by comparing the work of separation and the work of decohesion . the structure and work of separation for such interfaces including h terminated diamond was examined computationally by wang and smith @xcite . they found a strong decrease in the adhesion energy and thus the work of separation of cu / c@xmath0 interfaces upon the introduction of interfacial hydrogen , but they did not investigate material transfer explicitly . their results are in agreement with experiments performed by pepper @xcite on the frictional properties of these interfaces . furthermore , in a recent study involving md simulations of nanoscratching zhu et al . @xcite observed that the minimum depth of material transfer at cu surfaces can be as thin as only one atomic layer of cu depending on the machining conditions . in this work , as for the al / tin interfaces , the emphasis is put on material transfer and its ab - initio simulation . to investigate the properties of al / tin interfaces and cu / c@xmath0 interfaces terminated with oxygen and hydrogen , respectively , we performed calculations of the approach and subsequent separation of two slabs forming one of the aforementioned interfaces . the focus of this paper is put on al / tin interfaces , while results on contacting cu and diamond slabs are given for comparison , validation of the method as well as a more general view of the subject . all results were calculated within the framework of dft using the vienna ab initio simulation package ( vasp ) @xcite . vasp uses periodic boundary conditions and a plane - wave basis set within the projector augmented - wave ( paw ) formalism @xcite . the interaction between the ionic core and the valence electrons is mediated by the corresponding paw pseudopotentials . the generalized gradient approximation ( gga ) in the perdew , burke , and ernzerhof ( pbe ) parametrization @xcite was used to describe the exchange and correlation ( xc ) functional , unless explicitly mentioned otherwise . while ggas often underestimate binding and adhesion energies @xcite , the local - density approximation ( lda ) @xcite usually overestimates these quantities @xcite . further examples but also deviations from this trend for the behaviour of different xc functionals are presented in ref . . here , calculations employing the lda were performed for comparison and validation . for the same reason the van der waals ( vdw ) density functional ( df ) optb86b @xcite was employed , which incorporates a nonlocal correlation term approximating vdw interactions . the `` opt '' vdw - dfs were applied to diverse materials , e.g. , see refs . , and proved to be of good accuracy . accurate total energies are guaranteed by a careful choice of the calculation parameters . k_-space integrations were performed using a @xmath1-centered @xmath2 monkhorst - pack mesh @xcite . the energy cutoff for the plane - wave basis was set to . both settings allow for total energies accurate to . while atomic relaxations were performed employing the method of methfessel and paxton @xcite to first - order with a smearing of , the tetrahedron method with blchl corrections @xcite was used for all static calculations . since the method of methfessel and paxton should only be used for metallic systems , gaussian smearing of was chosen for relaxations of the cu / c@xmath0 interfaces . a damped molecular dynamics ( md ) algorithm was selected in order to relax the atomic structures until an energy convergence criterion of was fulfilled . each converged relaxation was followed up by a static calculation to obtain more accurate total energies . for electronic self - consistency cycles a convergence criterion of was used . all simulations were performed at . the damped md relaxation scheme was chosen instead of the widely used conjugate - gradient or quasi - newton algorithms , because these two algorithms caused convergence problems and the tendency to remain stuck in local minima . the simulation model is described here for the case of an al / tin interface , using the methodology presented by the authors in an earlier publication @xcite . the same route is followed in the case of cu / c@xmath0 interfaces . to model our al / tin systems we constructed simulation cells from a fcc ( 111 ) al slab at the bottom and a rock salt ( 111 ) tin slab above ( see fig . [ fig : al - tin - side ] ) . the gap between the two slabs is called the `` interface distance '' which is measured from the top layer of the lower slab to the bottom layer of the upper slab , regardless of termination . in the following the term `` slab height '' refers to the vertical distance between the bottom al and top tin layers , which is the sum of the interface distance and the heights of the two slabs . to investigate the effect of oxygenation , an al slab was terminated by oxygen at the interface ( see fig . [ fig : al - o - tin - side ] ) . in this work 1@xmath31 surface cells were used , which represent infinitely extended surfaces because of the periodic boundary conditions . the tin slab consisted of a minimum of six ti and six n layers , while the al slab are built of seven layers . these geometric parameters are consistent with previous investigations . @xcite . 0.25 0.25 in the [ 111 ] direction tin is built from alternating layers of pure ti or n ; therefore , the ( 111 ) surface of a tin slab is either terminated by a layer of ti or n. the termination depends on the chemical potential of nitrogen . in an earlier publication @xcite it was shown that both cases are found in reasonable nitrogen concentration ranges by performing a thermodynamic analysis @xcite to calculate the surface gibbs free energy for the off - stoichiometric slabs @xcite . the presented results are in agreement with the refs . . in this work the main focus is put on n terminated slabs , since this termination is favorable at small deviations of the nitrogen chemical potential from its molecular reference , i.e. at ambient conditions . over a wide range of the nitrogen chemical potential the ( 001 ) tin surface is actually the most stable orientation ; however , al / tin interfaces between ( 001 ) slabs do not show any material transfer . @xcite thus , results for interfaces between ( 111 ) surfaces , which show a similar stability at certain values of the chemical potential and allow for material transfer at least between pristine surfaces , are presented in this work . dipole corrections @xcite perpendicular to the interface ( z direction ) were found to be negligible . to decouple the periodically repeated simulation cells in the z direction , a vacuum spacing of at least was included in the simulation cell . performing bulk calculations the lattice parameters of the single materials , and for al and tin , respectively , were obtained . the experimental lattice constants @xcite of al and tin are and , respectively , confirming our calculation to be accurate within 0.5% relative error . for the simulation cells combining al and tin slabs , an intermediate lattice parameter of was used for the lateral xy lattice vectors to equalize the relative error of about 2.6% for both materials . the consequences of such distortions were investigated in detail in an earlier publication @xcite , where no influence of the chosen lattice parameter on the occurrence of material transfer was observed . for the z direction the material - specific values for the layer distances were used assuming a pseudomorphic interface . it has to be noted that in reality such a simple mutual adjustment of thick slabs by a combination of stretching and compression does not usually occur , since dislocations at the interface or an incommensurate contact are more plausible . thus , some of the atoms would not be aligned perfectly according to the stacking of their respective original slabs as well as across the interface , but rather sample slightly different local environments . for computational reasons , these different local arrangements across the interface are assessed by considering various configurations at the interface which constitute limiting cases . the simulation of the approach and separation of the slabs follows the method described in ref . . for convenience the main points are repeated here . the approach and subsequent separation of the two slabs were simulated by moving the upper slab in discrete steps along the z direction and allowing for electronic and atomic relaxations after each step . for the atomic relaxations the bottom three al as well as the top tin ( three ti and three n ) layers were kept rigid at bulk - like distances , while the intermediate `` free '' ones were allowed to fully relax . this is depicted in fig . [ fig : al - tin - initial ] for the n - terminated case . for the approach and separation step sizes of and , respectively , were used . for the observation of material transfer the separation is more critical ; therefore , a smaller step size was used , which allows for higher accuracy but causes increased computational costs . the approach and separation of the slabs were initiated from equilibrium structures , i.e. , separately relaxed slabs and the structure with the lowest energy determined during the approach , respectively . to simulate a realistic separation process only the topmost , rigid tin layers were moved in discrete steps in the positive z direction , again allowing for electronic and atomic relaxations after each step , i.e. , a separation of the slabs at the initial interface was not promoted by the enforced step - wise motion . to study the effects of different local environments at the interface , several alignments of the slabs were investigated by placing the upper slab laterally on various sites with respect to the surface of the lower slab . as mentioned above these different configurations represent limiting scenarios for more complex realistic interfaces . the definitions of the alignments are shown in fig . [ fig : bond - sites ] by marking the high - symmetry points on the n - terminated ( 111 ) tin surface where the next al or o atom can be placed . the alignments are named according to the stacking of the bottom tin layer with respect to the al slab , the position of interfacial species is not taken into account in this naming convention . the different configurations are compared via the interaction energy @xmath4 , which is defined as the difference of the total energy of the interacting slabs @xmath5 at slab height @xmath6 and the reference energies of the two independent slabs , @xmath7 and @xmath8 , @xmath9 1 surface cell of a n - terminated ( 111 ) tin surface . filled circles indicate atoms in the top surface layer for each species ( ti and n are given by large purple and small cyan circles , respectively ) , while empty circles label atoms below the top surface layer . to obtain a ti - terminated ( 111 ) tin surface the ti and n atoms of the shown n - terminated surface have to be exchanged . high - symmetry points ( top , hcp , fcc ) are highlighted . ] the fcc adsorption site is favored by a single oxygen atom on a 1@xmath31 ( 111 ) al surface cell . the oxygen is strongly bound to the al surface exhibiting a rather small binding distance of about with an adsorption energy of about with respect to molecular , gas - phase oxygen . @xcite the hcp site is slightly less favorable by about , while the interaction on the top site is much weaker and not favorable considering molecular oxygen as the reference . additionally , the placement of two o atoms on each 1@xmath31 al surface cell was tested . the two atoms were located initially above the fcc and hcp sites of the al surface . during a subsequent relaxation the fcc o atom moved beneath the top al layer , while the hcp o atom stayed above the al surface . the distance between the top o - al compound and the remaining o terminated al slab grew to about . this large distance combined with a very weak adhesion ( ) of this compound on the surface shows that the termination with one o atom per surface cell is preferential on a 1@xmath31 al surface cell . the possibility of an al@xmath10o@xmath11-like surface oxide @xcite is not considered here . since material transfer at interfaces is of interest , the energy necessary to remove layers from a material is an important quantity . for this work , additionally the question arises whether this removal energy is affected by an oxygen layer . the removal of the layers was simulated by placing the layers at a large distance from the remaining slab , which does not allow for interactions between the slab and layers . results for pristine al slab including the effect of compressive and tensile stress were presented in an earlier publication @xcite . here , only the values necessary for comparison are repeated in table [ tab : removal ] . typically , it is energetically unfavorable to remove just one al layer compared to at least two layers . moreover , the removal energies are not changing significantly for two or more layers ; therefore , the removal of two al layers is used as a reference here . as mentioned above , oxygen is strongly bound to the al slab . thus , it is unlikely that only the oxygen is removed and so the more plausible case of and the removal of the oxygen layer together with al layers is investigated here . the data in table [ tab : removal ] shows that an oxygen layer hardly influences the removal energies which indicates that the bond breaking between the al layers is the determining factor . nevertheless , oxygen may still passivate a surface by suppressing interactions , i.e. weakening the interfacial adhesion , across an interface with another material such as tin . .[tab : removal ] energy costs to remove the top two al layers , with or without an additional oxygen layer , from a pristine and a oxygenated al slab , respectively , using pbe and lda . the removal energies are given in per 1@xmath31 surface cell . [ cols="<,^,^ " , ] 0.15 0.15 0.15 0.15 + 0.15 0.15 0.15 0.15 for ti - terminated tin slabs the adhesion energies at the al / tin interfaces are also reduced upon the introduction of oxygen at the interface . for different alignments at the interface the values of the adhesion energies range between to and to for the oxygenated and non - oxygenated @xcite contact , respectively . the decrease is less pronounced than for the n - terminated case because of more favorable ti - o interactions . nevertheless , material transfer becomes again unfavorable because of the interfacial oxygen , since the adhesion energies become smaller than the al - o removal energies of about . the relevant information is summarized in table [ tab : loop - data ] . in addition to the al / tin interfaces , cu / c@xmath0 interfaces were investigated to test the used method for the simulation of material transfer on a different system . analogous to the studies of al / tin , results on interfaces between the ( 111 ) surfaces of cu / c@xmath0 are discussed in this section . in general , the diamond ( 111 ) surface exhibits a ( 2 @xmath3 1 ) reconstruction ; however , in the case of hydrogen termination as well as at cu / c@xmath0 interfaces no reconstructions are observed and ( 1 @xmath3 1 ) configurations are found @xcite . thus , in this work only ( 1 @xmath3 1 ) structures are considered . at the diamond ( 111 ) surface two terminations are possible . here , the single dangling bond configuration is assumed , because it is more stable for reasonable hydrogen chemical potentials than the triple dangling bond terminations @xcite . particularly , the former configuration is more favourable at interfaces with copper @xcite . the energy necessary to remove material from an ( 111 ) cu slab is about , which will be compared to the adhesion energies for several alignments at cu / c@xmath0 interfaces in the following . of the examined high symmetry contact configurations , which are defined in the same way as for the al / tin interfaces , the top and fcc alignments result in the strongest and weakest interfaces , respectively . the adhesion energy for the top configuration is about , while for the fcc case it amounts to about . since these adhesion energies are larger than the cost to remove material from the cu slab , material transfer is expected to occur for all configurations . employing the method of approach and subsequent separation of the two slabs described above leads indeed to the material transfer of one layer of cu towards the c@xmath0 slab for all tested contact scenarios , which are considered to be the limiting cases . figures showing the corresponding interaction - energy curves are included in the supplementary material . @xcite these results support the dft - based work of guo et al . @xcite who predicted the occurrence of fracture at cu / c@xmath0 interface between the first two metallic layers or the 2nd and the 3rd layers , because of a bond weakening resulting from the contact with the c@xmath0 counterbody . as for the al / tin interfaces , the effect of additional interfacial species was studied at the contact of cu and c. since the dangling bonds at diamond surfaces are often saturated by hydrogen atoms , this species was chosen for our investigation . at the c@xmath0 ( 111 ) surface the h atoms prefer to sit at top sites in a distance of just above each c atom forming a p(1 @xmath3 1 ) pattern . this is in good agreement with literature , e.g. in ref . . the adhesion energies for all studied contact configurations of the cu / c@xmath0 interfaces are extremely reduced to about for hydrogen terminated c@xmath0 surfaces in a similar fashion as for the oxygenated al surfaces at the al / tin interfaces . since these adhesion energies are much lower than the cu removal energies of about , no material transfer is expected . this predicted behavior is verified by the method of approach and separation . figures of the resulting interaction - energy curves are shown in the supplementary material . @xcite the same results with respect to material transfer were obtained employing further xc functionals , namely lda and vdw - optb86b . as expected using these functionals the adhesion as well as removal energies for all investigated interfaces are increased similar to the al / tin interfaces . for the clean surfaces lda and vdw - optb86b yield very similar adhesion energies within about for each configuration . these energies are about and for the fcc and top alignment , respectively . as for pbe , the adhesion energies are drastically reduced for interfaces involving hydrogenated c@xmath0 surfaces . using lda the adhesion energies are between for the fcc configuration and for the top case , while for vdw - optb86b these energies range between about and for the corresponding alignments . the energies necessary to remove cu layers from a cu slab are also increased to and for lda and vdw - optb86b , respectively . these results show that the choice of the investigated xc functionals does not affect the results for material transfer . the findings presented above agree well with theoretical and experimental results presented in the introduction including the strong effect of interfacial hydrogen on the adhesion energy @xcite and the possibility of a transfer of only one atomic layer of cu @xcite . additionally , performing large - scale molecular dynamics ( md ) simulations on the rubbing contact of hydrogenated diamond bodies and tungsten surfaces , stoyanov et al . @xcite found that material transfer between c@xmath0 and w mainly occurred at regions where the hydrogen had been depleted from the diamond surface because of the rubbing process . this observation is in qualitative agreement with our study finding material transfer only possible for non - hydrogenated diamond surfaces . the adhesion energies at al / tin and cu / c@xmath0 interfaces are reduced significantly by the presence of a monolayer of oxygen and hydrogen at al / tin and cu / c@xmath0 interfaces , respectively . this reduction is investigated by analyzing the densities of states and charge density differences at the aforementioned interfaces which reveal distinct bonding situations . particularly , the possibility of material transfer at the interfaces is of interest . the occurrence of material transfer was studied by simulating the approach and subsequent separation of two slabs . for the clean surfaces material transfer of one or two atomic layers was found to be favorable for both investigated material combinations depending on the local interfacial configuration . the additional interfacial species o and h passivate the involved surfaces with respect to material transfer because of the strongly reduced adhesion energies for all investigated configurations . this is in agreement with the observation that metal - ceramic interfaces break at the interface or in bulk areas according to their interfacial adhesion @xcite . the results with respect to material transfer were not affected upon the use of different approximations for the exchange - correlation functional . the investigation of more complex interfacial species such as extended oxides instead of the studied monolayers promises to be very interesting in order to model many systems in a more realistic fashion . while , in principle , the method employed in this work can be applied to any pair of materials , in practice this can be computationally quite cumbersome . large simulation cells , which translate into high computational demands , become necessary for complex materials or materials with an unfavorable bulk lattice mismatch because the translational symmetry has to be preserved and the lattice distortions should be as low as possible . on the other hand , larger cells allow to account for additional features such the minimization of lattice mismatches and the possibility to model dislocations or quasi - incommensurate contacts at interfaces . furthermore , surface roughness can be treated to some extent by using a stepped surface or a regular arrangement of asperities . the authors thank g. vorlaufer for fruitful discussions . g.f . is grateful for the financial support from the german research foundation ( dfg ) via sfb 986 m@xmath12 . , p.o.b . , p.m. , and j.r . acknowledge the support by the austrian science fund ( fwf ) : f4109-n28 sfb vicom . part of this work was funded by the austrian comet - program ( project k2 xtribology , no . 849109 ) via the austrian research promotion agency ( ffg ) and the province of niedersterreich , vorarlberg , and wien . part of this work was supported by the european cooperation in science and technology ( cost ; action mp1303 ) . the authors also appreciate the ample support of computer resources by the vienna scientific cluster ( vsc ) . [ fig : al - tin - initial ] and [ fig : bond - sites ] were created employing vesta @xcite , fig . [ fig : chg - diff-111 ] utilizing visit @xcite , and fig . [ fig:111-nterm - sep ] using vmd @xcite . 100ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1103/physrevlett.56.930 [ * * , ( ) ] link:\doibase 10.1016/s1369 - 7021(05)00845-x [ * * , ( ) ] http://www.myilibrary.com?id=954[__ ] ( , , ) link:\doibase 10.1016/s1748 - 0132(07)70140 - 8 [ * * , ( ) ] link:\doibase 10.1098/rsta.2007.2170 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.88.215501 [ * * , ( ) ] http://link.aps.org/doi/10.1103/physrevlett.101.125501 [ * * , ( ) ] link:\doibase 10.1038/nnano.2010.3 [ * * , ( ) ] link:\doibase 10.1038/nnano.2012.255 [ * * , ( ) ] link:\doibase 10.1007/s12541 - 012 - 0224-y [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1063/1.3277013 [ * * , ( ) ] link:\doibase 10.1016/0167 - 5729(94)00005 - 0 [ * * , ( ) ] link:\doibase 10.1016/0956 - 7151(92)90256-e [ * * , ( ) ] http://www.books24x7.com/marc.asp?bookid=37295[__ ] ( , , ) link:\doibase 10.1103/physrevlett.63.766 [ * * , ( ) ] in http://www.crcnetbase.com/doi/abs/10.1201/9780849377877.ch20[__ ] , vol . ( , ) pp . link:\doibase 10.1098/rsta.2005.1621 [ * * , ( ) ] in http://linkinghub.elsevier.com/retrieve/pii/b9780444527721500366[__ ] ( , ) pp . link:\doibase 10.1088/0022 - 3727/41/12/123001 [ * * , ( ) ] link:\doibase 10.1039/c2fd00120a [ * * , ( ) ] link:\doibase 10.1103/physrevlett.115.025502 [ * * , ( ) ] link:\doibase 10.1088/0022 - 3727/48/46/465308 [ * * , ( ) ] link:\doibase 10.1016/j.cpc.2016.10.017 [ ( ) , 10.1016/j.cpc.2016.10.017 ] link:\doibase 10.1103/physrevlett.64.3054 [ * * , ( ) ] link:\doibase 10.1103/physrevb.70.241401 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/23/26/265003 [ * * , ( ) ] link:\doibase 10.1021/la200783a [ * * , ( ) ] link:\doibase 10.1103/physrevlett.108.126103 [ * * , ( ) ] link:\doibase 10.1021/nl204019k [ * * , ( ) ] link:\doibase 10.1103/physrevb.90.195418 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevb.91.195436 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/8/32/003 [ * * , ( ) ] link:\doibase 10.1016/s0039 - 6028(01)01225 - 0 [ * * , ( ) ] link:\doibase 10.1016/j.mser.2003.09.001 [ * * , ( ) ] link:\doibase 10.1209/0295 - 5075/26/8/005 [ * * , ( ) ] link:\doibase 10.1021/ct5001044 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.93.175503 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.105.075502 [ * * , ( ) ] link:\doibase 10.1038/ncomms3441 [ * * , ( ) ] link:\doibase 10.1016/s0040 - 6090(97)01067 - 5 [ * * , ( ) ] link:\doibase 10.1063/1.1372162 [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1016/0927 - 796x(95)80001 - 8 [ * * , ( ) ] link:\doibase 10.1016/j.actamat.2004.04.022 [ * * , ( ) ] link:\doibase 10.1016/j.actamat.2006.07.011 [ * * , ( ) ] link:\doibase 10.1016/j.commatsci.2006.05.017 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/19/22/226003 [ * * , ( ) ] link:\doibase 10.1063/1.3703663 [ * * , ( ) ] link:\doibase 10.1088/0965 - 0393/22/3/035020 [ * * , ( ) ] link:\doibase 10.1038/srep17380 [ * * , ( ) ] link:\doibase 10.1038/srep18554 [ * * , ( ) ] link:\doibase 10.1016/j.physe.2017.01.025 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/17/35/002 [ * * , ( ) ] link:\doibase 10.1103/physrevb.52.14954 [ * * , ( ) ] link:\doibase 10.1016/s0039 - 6028(00)00578 - 1 [ * * , ( ) ] link:\doibase 10.1016/s0039 - 6028(97)00348 - 8 [ * * , ( ) ] link:\doibase 10.1103/physrevb.53.13725 [ * * , ( ) ] link:\doibase 10.1016/0039 - 6028(96)00837 - 0 [ * * , ( ) ] link:\doibase 10.1016/0039 - 6028(96)00836 - 9 [ * * , ( ) ] link:\doibase 10.1016/s0039 - 6028(97)00168 - 4 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.87.186103 [ * * , ( ) ] link:\doibase 10.1116/1.571616 [ * * , ( ) ] link:\doibase 10.1016/j.commatsci.2016.03.023 [ * * , ( ) ] link:\doibase 10.1103/physrevb.47.558 [ * * , ( ) ] link:\doibase 10.1103/physrevb.49.14251 [ * * , ( ) ] link:\doibase 10.1103/physrevb.54.11169 [ * * , ( ) ] link:\doibase 10.1016/0927 - 0256(96)00008 - 0 [ * * , ( ) ] link:\doibase 10.1103/physrevb.50.17953 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.1758 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.77.3865 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.5521 [ * * , ( ) ] link:\doibase 10.1103/physrevb.23.5048 [ * * , ( ) ] link:\doibase 10.1103/physrevb.59.14992 [ * * , ( ) ] link:\doibase 10.1103/physrevb.76.024309 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/22/2/022201 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.195131 [ * * , ( ) ] link:\doibase 10.1088/0953 - 8984/24/42/424216 [ * * , ( ) ] link:\doibase 10.1021/jp504695 m [ * * , ( ) ] link:\doibase 10.1021/jp503829c [ * * , ( ) ] link:\doibase 10.1103/physrevb.86.035451 [ * * , ( ) ] link:\doibase 10.1063/1.4866175 [ * * , ( ) ] link:\doibase 10.1021/jp511268s [ * * , ( ) ] link:\doibase 10.1103/physrevb.13.5188 [ * * , ( ) ] link:\doibase 10.1103/physrevb.40.3616 [ * * , ( ) ] link:\doibase 10.1103/physrevb.49.16223 [ * * , ( ) ] link:\doibase 10.1103/physrevb.62.2899 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.81.1038 [ * * , ( ) ] link:\doibase 10.1103/physrevb.65.035406 [ * * , ( ) ] link:\doibase 10.1103/physrevb.83.115418 [ * * , ( ) ] link:\doibase 10.1016/j.ssc.2010.04.034 [ * * , ( ) ] link:\doibase 10.1103/physrevb.46.16067 [ * * , ( ) ] @noop _ _ , ed . ( , ) @noop `` '' link:\doibase 10.1016/j.actamat.2013.12.029 [ * * , ( ) ] link:\doibase 10.1107/s0021889811038970 [ * * , ( ) ] in @noop _ _ ( ) pp . link:\doibase 10.1016/0263 - 7855(96)00018 - 5 [ * * , ( ) ]	the effect of monolayers of oxygen ( o ) and hydrogen ( h ) on the possibility of material transfer at aluminum / titanium nitride ( al / tin ) and copper / diamond ( cu / c@xmath0 ) interfaces , respectively , were investigated within the framework of density functional theory ( dft ) . to this end the approach , contact , and subsequent separation of two atomically flat surfaces consisting of the aforementioned pairs of materials were simulated . these calculations were performed for the clean as well as oxygenated and hydrogenated al and c@xmath0 surfaces , respectively . various contact configurations were considered by studying several lateral arrangements of the involved surfaces at the interface . the consistency of the dft results was evaluated by using different approximations for the exchange - correlation functional . material transfer is typically possible at interfaces between the investigated clean surfaces ; however , the addition of o to the al and h to the c@xmath0 surfaces was found to hinder material transfer . this passivation occurs because of a significant reduction of the adhesion energy at the examined interfaces , which can be explained by the distinct bonding situations .
the term `` citizen science '' refers to the activities of people who are not paid to carry out scientific research ( `` citizens '' ) , but who make intellectual contributions to scientific research nonetheless . they come from all walks of life , and their contributions are diverse , both in type and research area . this review is about the astronomy projects they have participated in to date , the tasks they have performed , and how astronomy has benefited and could benefit further from their efforts . citizen involvement in science pre - dates the profession itself . the earliest example of collaboration between professional and amateur astronomers seems to have been edmund halley s call for observations of the 1715 total eclipse of the sun which crossed central england @xcite . missed both the time of the beginning of the eclipse and that of total darkness . '' ] since then there has been a long and honourable tradition of amateur observers making important discoveries and significant sustained contributions . however , the advent of the world wide web has changed the face of professional and amateur collaboration , providing new opportunities and accelerating the sharing of information . people are now connected to each other on a scale that has never happened before . professional scientists can interact with citizens via a range of web - based media , including purpose - built citizen science websites which increase the potential for shared data analysis and exploration , as well as for data collection . meanwhile , communities of citizens have sprung into existence as like - minded people have been able to find and talk to each other in a way that is almost independent of their geographical location . the result has been an exponential increase in citizen involvement in science . the field is evolving very quickly , with more and more professional scientists becoming aware of the possibilities offered by collaborating with , for example , specialists operating outside the usual parameters of professional astronomical observation , or tens of thousands of people eager to perform microtasks in their spare time . our aim in this work is to review the astronomical ( and occasionally wider ) literature for productive citizen science projects , and distill the characteristics that made these case studies successful . as our title states , this is a review of ideas for astronomy : we will look forward as well as back , and try to answer the following questions . what are the particular niches that citizen science fills , in our field ? what is the potential of citizen science in astronomy , and how can it be realized ? citizen science has a significant impact on its participants , whether they be sitting in a university office or in front of a home computer or mobile phone screen . this review is about the impact that citizen astronomy has had , and can have , on the progress of research in astronomy . this review is organised as follows . astronomy research typically starts with observations : so do we , in . we then proceed to consider visual classification , data modeling and finally citizen - led enquiry in sections [ sec : class][sec : explore ] . with this overview in place , we take a look in at the population of citizens who take part in astronomical research . we then turn to the future , and speculate on how citizens might contribute to astronomy there ( ) , and finish with some concluding remarks in . there is currently an active community of well - equipped amateur observers making astronomical observations of great utility . there are also many other citizens observing the night sky with less sophisticated equipment and as we shall see , there are even some examples of citizens making astronomical observations almost inadvertently . what astronomical data are the citizenry taking , and what is it being used for ? in this section , we review some of the citizen contributions to active observations of the night sky . `` passive '' contributions will be described in below . the steady improvements and increasing affordability of digital technology , in addition to the ease of data sharing and communications , have considerably expanded the realm of amateur astronomy in the past two decades . meanwhile , professional observatories are always over - subscribed , with resources necessarily being divided between particular areas of sky , or samples of objects , or on a few astronomical questions : tuning the parameters of professional observations to optimize all possible scientific enquiries would seem an impossible task . what types of niches does this leave for amateur observers to fill ? what are the strengths that amateur observers can play to ? the first key advantage that amateurs have is time availability . determinations of meteor frequencies ( for example ) require observations on short timescales ( minutes ) , whereas the slow evolution of giant planets ( for example ) occurs on longer timescales ( years and decades ) . amateur observations can be frequent and repetitive , but also long standing . the second , related , advantage is that of flexibility : whenever a new phenomenon is discovered ( e.g. , a new comet , or anything changing the appearance of the familiar planetary discs ) , observers will be keen to catch a glimpse irrespective of the scientific value of their observations . this reaction can be near instantaneous , compared to the need to allocate telescope resources among the professional community , and , when made by a networked community , provides naturally well - sampled coverage across the globe . the third benefit is contextual . professional observations are often taken in a very different wavelength range , focus on a narrower spatial region , or employ spectroscopic techniques that do nt yield images . in some situations , near - simultaneous wide field optical imaging by citizen scientists can provide useful additional constraints on the process of interest . the example case studies below serve to illustrate this synergy between amateur and professional observations , and also to highlight instances of professional - amateur ( `` pro - am '' ) collaboration . while the solar system provides some of the most amenable targets for amateur observation , `` deep sky '' observations by the non - professional community provide important further insight into the capabilities and potential of citizen astronomers . although survey telescopes provide the vast majority of modern solar system discoveries , citizen astronomers occupy some useful observational niches . small solar system objects moving against the fixed - star background can be detected in a set of ccd frames either by eye or by automated software . targets include near - earth asteroids ( neas , with orbits intersecting those of the terrestrial planets ) , main belt asteroids between mars and jupiter , and comets making their journey towards the sun from the outer solar system . the extreme familiarity of citizen astronomers with a particular region of sky , planet or nebula , allows them to immediately identify peculiarities or new features . a protocol for citizen discovery has been established : the position of any new object is compared to existing catalogues , and if no existing details are found then the new discovery and its ephemerides can be reported to the iau minor planet center . if observations are repeated for at least two nights by one or several observers , then a new denomination is provisionally assigned to the discovery . an electronic circular then reports the discovery to the wider world . for example , the nea 2012 da14 was initially reported to the minor planet center by a team of amateur observers affiliated with the la sagra sky survey at the astronomical observatory ( spain ) , and subsequently characterised by professional astronomers during its closest approach in february 2013 ( e.g. , * ? ? ? * ) . as with asteroids , the majority of new comet discoveries are made by automated surveys , but a small and stable number of discoveries come from amateurs with small telescopes , typically in regions poorly covered by survey telescopes ( e.g. , regions close to the sun ) . c/2011 w3 ( lovejoy ) , a kreutz sungrazer comet , is one such example , discovered by t. lovejoy and circulated via the central bureau for astronomical telegrams ( cbat ) ( e.g. , * ? ? ? the oort cloud comet c/2012 s1 ( ison ) was spotted by v. nevski and a. novichonok in images from the international scientific optical network , which spurred a major international effort to observe its perihelion passage as it disintegrated @xcite . amateurs are also contributing to the search for a sub - category of objects with a detectable cometary coma within the asteroid belt . recent discoveries of these main belt comets , which appear to be asteroids that are actively venting their volatiles at perihelion , are beginning to blur the distinction between asteroids and comets . the t3 project , a collaboration between the university of rome and several amateur observers , began in 2005 with the detection of a coma around asteroid 2005 sb216 @xcite , and has gone on to detect at least eight main belt comets @xcite . these early citizen science discoveries , followed up by professional astronomers , have generated new insights into the properties and variety of comets , and the dynamic and evolving nature of our solar system . the discovery of comet shoemaker - levy 9 ( co - discovered by amateur observer d. levy ) before its collision with jupiter @xcite is a classic example . in general , it is the global distribution of citizen observers and the long - baselines of their observations that enable new discoveries of minor bodies in our solar system . beyond first detections , citizen observers can aid in the detailed study of the physical and orbital characteristics of these newly discovered solar system bodies . these amateur - led contributions are typically published via the _ minor planet bulletin_. photometric monitoring of an asteroid as it rotates provides information on its physical parameters such as its shape , rotation rate and orientation ; monitoring of a comet s coma , dust and plasma tails can reveal dynamic structures , determine the locations of active venting regions and reveal outbursts and other events associated with the outgassing ( see * ? ? ? * for a comprehensive review ) . planetary atmospheres make tantalising targets for citizen observers , being large , bright , colourful and highly variable from night to night ( e.g. , . the long - term monitoring provided by the network of amateur astronomers provides valuable insights into the meteorology and climate of these worlds , tracking the motions of clouds , waves and storms as they are transported by atmospheric winds to probe the physical and chemical processes shaping their climates . for example , the global distribution of giant planet observers permits global monitoring of jupiter and saturn as they rotate over 10 hours . citizens upload raw filtered images and colour composites , organised by date and time , to online servers , such as the planetary virtual observatory and laboratory ( pvol ) maintained for the international outer planets watch ( iopw * ? ? ? those images can be used by amateurs and professionals alike to study quantitatively the visible activity , including measuring wind speeds from erupting plumes @xcite , investigating the strength and changes to the large vortices ( e.g. , the 2006 reddening of jupiter s oval ba , * ? ? ? * ) , and determining the life cycle of the belt / zone structure @xcite . for saturn , a close collaboration between citizen scientists and cassini spacecraft scientists ( known as saturn storm watch ) has allowed correlation of lightning - related radio emissions detected by the spacecraft with visible cloud structures on the disc ( e.g. , * ? ? ? * ) , which would not have been possible with the targeted regional views provided by cassini s cameras alone . furthermore , it was the amateur community that first spotted the eruption of saturn s enormous 2010 - 2011 storm system , which was monitored over several months @xcite . video monitoring has been used by citizen observers to enable high resolution `` lucky '' imaging of jupiter . the best images , at moments of clear seeing , from the high - resolution video frames are selected , extracted and stacked together , using custom software to correct for the distortions associated with the telescope optics and residual atmospheric seeing . software written by citizen scientists for free distribution to active observers , such as registax and autostakkert , allows them to process their own video files , thus avoiding the need for transfer of large datasets to some centralised server ( see * ? ? ? * for a thorough review ) . descriptive records of morphological changes are maintained and continuously updated by organisations of citizen scientists such as the british astronomical association ( baa ) and the association of lunar and planetary observers ( alpo ) . the baa s jupiter section is a team of amateurs with substantial expertise in jupiter s appearance @xcite ; their regular bulletins describe the changing appearance of the banded structure and the emergence of new turbulent structures and weather phenomena , and keep a record of the long - term atmospheric changes . active citizen observing also provides long - term monitoring in the inner solar system . venus photochemical smog shields the planet s surface from view , but discrete cloud features can be used to study the super - rotation of the venusian atmosphere , and the occurrence of a mysterious ultraviolet absorber at high altitudes . for example , the venus ground - based image active archive was created by esa to provide contextual observations supporting the venus express mission @xcite . the martian atmosphere , with its ephemeral clouds , seasonal co@xmath0 polar ice cycles , and dust storms , continues to prove popular among citizen observers , although these are a minor supplement to the wealth of high - resolution information being returned by orbital and surface missions to the red planet . as with other planetary targets , amateur observations provide the long temporal records of the evolution of atmospheric features . groups such as the international society of mars observers ( ismo ) , the british astronomical association ( baa ) and the international mars watch program quantitatively and qualitatively assess these amateur images . finally , although citizen observations of uranus and neptune are in their infancy and require telescopes with diameters exceeding 25 cm , there have been confirmed reports of atmospheric banding and discrete cloud features when near - infrared filters ( to maximise the contrast between the white clouds and the dark background ) and long exposure times of tens of minutes are used . citizen monitoring of all of these worlds ( summarised in ) provides the long - baselines , flexibility and high frequency of imaging needed to understand the forces shaping their evolving climates . the increasing adoption of video monitoring of planetary targets means that unexpected , short - lived events on the surfaces of those bodies are now more likely to be observed by citizen astronomers than by professional observatories . for example , an impact scar near jupiter s south polar region was first discovered in imaging by australian amateur anthony wesley on july 19th , 2009 . this led to an international campaign of professional observations to understand the asteroidal collision that had created the scar ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? although the 2009 impact was out of view from the earth , at least three flashes have been confirmed between 2010 and 2012 , and the light curves used to determine the sizes and frequency of objects colliding with jupiter ( e.g. , * ? ? ? citizen scientists have developed free software to allow observers to search for impact flashes in an automated way ( e.g. , jupiter impact detections and lunarscan from the alpo lunar meteoritic impact search for transient impact flashes recorded on the moon ) . amateur observers have contributed to several exoplanet investigations , responding to detections made by professional surveys and making important contributions to the light curves of the targets . in the case of exoplanet transits , the challenge is to measure the 1% diminution in starlight as a giant planet transits in front of its parent star . @xcite point out three methods whereby amateurs can contribute to the characterisation of exoplanetary systems : first , by frequent observations of known transits to refine ephemeris ; second , by searching for transit time variations that can reveal additional planets in a system ; and third , by searching for previously unidentified transits in known planetary systems ( e.g. , the discovery of the transit of hd 80606b from a 30 cm telescope near london , * ? ? ? in a planetary microlensing event , significant brightening of the background star is required to make a planet orbiting the microlens visible at all ; if additional caustic crossings are caused , the resulting exoplanetary microlensing feature is of just several days duration , calling for high frequency , on demand monitoring a situation well matched to the capability of a global network of small telescope observers ( see e.g. * ? ? ? high magnification events detected by the ogle and moa surveys have been broadcast by the microfun and planet networks ( now merged ) to globally - distributed professional and amateur observers to follow up . these collaborations have been very successful , helping enable characterisation of over a dozen exoplanet systems ( see e.g. * ? ? ? * ; * ? ? ? * and references therein ) . the american association of variable star observers ( aavso ) supports and coordinates the efforts of over 900 amateur observers who are interested in bright , nearby variable stars ; in 2013 , the community made over a million observations , either visually or with ccd or dslr cameras , and logged them into a shared database . the aavso provides a number of services to assist the volunteers , including training material , an online data entry tool that carries out basic error checking , several pieces of software to assist the observers in checking their own observations , and data reviews by aavso staff . despite its name , aavso observers are located all over the world , with two thirds of the membership working outside the us . some of the community s larger telescopes can be operated robotically , and have been linked together into a network , aavsonet . this network is engaged in an ongoing all sky survey , apass , which is carrying out a survey of over 50 million stars between 10th and 17th magnitude , in 5 optical filters ( @xmath1 ) . the data processing and calibration is being done as a pro - am collaboration , and the data is being released at approximately annual intervals . the distributed nature of the aavso community means that it can produce continuous light curves for stars at a wide range of declinations . the aavso data has been used extensively by professional astronomers needing the most up to date optical measurements of stellar variability in , for example , the ss cyg system @xcite , optical light curves taken simultaneously with monitoring being carried out by space telescopes and/or at different wavelengths ( see * ? ? ? * for a successful joint aavso hst program ) , or who are performing long baseline data mining analyses of variable star populations . the aavso , in partnership with several professional astronomers and education specialists , successfully coordinated the nsf - funded `` citizen sky '' project to monitor the 20092011 eclipse in the espilon aurigae binary star system . the results from this campaign @xcite were used by @xcite to help interpret their interferometric imaging of the obscuring disk in the system . while amateur astronomers have aquired a great deal of very useful data , the general population is better equipped than ever to image the sky and make that data available for scientific analysis . this has been demonstrated by two recent professionally - led studies , that made use of a largely passive observing community connected via online social networks not usually associated with astronomy . @xcite used more than 2000 images scraped from the photo sharing website flickr as inputs to a reconstruction of the orbit of comet holmes . this comet was bright enough to be visible with the naked eye during its 2007 apparition , and a large number of photographs were taken of it and uploaded to the flickr site . @xcite were able to astrometrically calibrate many of the images using their automatic image registration software , ` astrometry.net ` @xcite , which detects the stars in each image and matches them to those in professionally - assembled catalogs . this had been enabled as a flickr `` bot , '' crawling over all images submitted to the ` astrometry.net ` group and sending the photos owners messages showing them where on the sky their images were taken . the time of observation could in many cases be derived from metadata included in the image headers . the calibrated images trace out the trajectory of the comet , producing a result which is close to that obtained from the jpl horizons system @xcite . estimates of orbital parameters from flickr images alone are accurate , when compared to the horizons values , to within a few standard deviations . as the authors point out , while in this case the photographers did not realize they were participating in a scientific study , the potential of combining powerful calibration software with large amounts of citizen - supplied imaging data is made clear . this method of `` unconscious '' citizen science may prove to have significant value in fields beyond astronomy too , if models of the statistical sampling can be developed , with ecological surveys of images submitted to sites like flickr likely in the next few years . as well as observing other planets for impacts , citizen scientists have also played a crucial role in the recording of asteroid impacts on earth , although not always realizing the usefulness of their observations . video footage of the fireball and shockwave of the february 2013 chelyabinsk meteor @xcite were essential to scientifically characterise the impactor and its likely origins , despite the fact that these records were largely captured accidentally by autonomous security cameras . trajectories reconstructed from these records even permitted the recovery of meteorites from a debris field on the ground . such objects are fragments of comets and asteroids , the debris left over from the epoch of planetary formation : their numbers , sizes and composition provide a window into the earliest evolutionary stages of our solar system . the statistics of these impacts reveal the risk of threats in our local neighbourhood , and these statistics are currently actively provided via a global network of citizen scientists , sharing and publicising their observations of meteors via the international meteor organisation ( imo ) . however , visual observations of meteors can also be tracked with no such active participation . by searching the archive of short text messages submitted to the web service twitter , barentsen et al ( priv . comm . ) were able to detect several new meteor showers . naked - eye observers had spotted shooting stars and tweeted about them to their followers , giving rise to a detectable signal in the steam of tweets that night . observing the night sky with a telescope is perhaps the most familiar of the activities of amateur astronomers , but as the previous section showed , citizens are also actively involved in the processing and interpretation of the data they have taken . in this and the next section we look at projects where much larger archival astronomical datasets have been made available to crowds of citizens , who are asked to inspect images and light curves , and help describe and characterize the features in them . despite significant advances in machine learning and computer vision , the visual inspection of data remains an important part of astronomy , as it continues to take advantage of the amazing human capacity for visual pattern recognition . while many in the 1990s predicted that the increasing size of astronomical datasets would make such time - intensive inspection impossible , the extensive reach of the world wide web has enabled the involvement of hundreds of thousands of citizen scientists in this form of `` crowd - sourced '' data analysis . while significant preliminary work had been carried out by nasa s `` clickworkers '' ( see below ) , the project that first illustrated the potential of crowd - sourcing for astronomical purposes was stardust@home . the team asked volunteers to scan through images of samples returned from comet wild-2 by the _ stardust _ mission , attracted a large audience to the apparently unprepossessing task of looking for dust grains in an effort to identify samples of material from outside our solar system . the site was built on bossa , an early attempt to build a generalized platform for such crowd - sourcing projects , and featured a stringent test which volunteers had to pass before their classifications would be counted . despite this hurdle , more than 20,000 people took part , and a variety of dust grains were removed from the aerogel for further study , contributing two of the seven candidate interstellar grains presented in a recent science paper@xcite . perhaps the most significant long - term impact of stardust@home , though , was the demonstration that large amounts of volunteer effort were available even for such seemingly uninspiring tasks such as hunting dust grains in images unlikely to be described as intrinsically beautiful , and that , with a suitable website design and stringent testing , scientifically valuable results could be obtained . the stardust@home experience directly inspired the development of galaxy zoo , perhaps the most prominent scientific crowd - sourcing project to date . galaxy zoo was built on the continued importance of morphological classification of galaxies , first introduced in a systematic fashion by hubble , and later developed by , among others , de vaucoleurs . while the morphology of a galaxy is closely related to its other properties , such as colour , star formation history , dynamics , concentration and so on , it is not entirely defined by then : there is more information in resolved images of galaxies than is captured in these observables . one approach was to develop simple proxies ( e.g cas @xcite ) , but these are at best approximations for true morphology . in an effort to prepare for large surveys , such as the sloan digital sky survey ( sdss ) , @xcite ( @xcite,@xcite ) , and later , @xcite developed neural networks trained on small samples of expert classified images , in order to automate the process of classification arguing that the size of the then up - coming surveys left no place for visual classification . the performance of these automatic classifiers depended on the input parameters , including colour , magnitude and size . these variables correlate well with morphology , but are not themselves morphological , and when included they dominate the classification . in particular , for galaxies which do not fit the general trends , such as spirals with dominant bulges , or star - forming ellipticals , automated classifiers , whether using these simple measures or more complex proxies for morphology such as texture , fail to match the performance of expert classifiers@xcite . as a result , @xcite , @xcite , and others have spent substantial amounts of time visually classifying tens of thousands of galaxies . inspired by stardust@home , a small group led by one of the authors ( lintott ) created galaxy zoo in 2007 to provide basic classifications of sdss galaxies classifiers were presented with a coloured image centered on and scaled to one of more than 800,000 galaxies , and could select from one of six options to characterise that object s morphology : clockwise , anti clockwise and edge - on spirals , ellipticals , mergers and `` star / dont know . '' aside from an easily - passed initial test , little knowledge was required or indeed presented to classifiers , enabling them to proceed quickly to doing something real soon after arriving at the site ; this approach , in contrast to stardust@home , was successful in encouraging large numbers of visitors to participate . this tactic in which both passing and sustained engagement provide substantial contribution is illustrated in which shows results from galaxy zoo 2 . this later version of the project asked for more detailed classifications via a decision tree containing questions such as ` how prominent is the bulge ? ' , and later iterations of the project have applied a similar approach to galaxies drawn from _ hubble space telescope _ surveys including gems , goods , cosmos and candels . to date , several hundred thousand people have participated in the galaxy zoo project , but such figures would be meaningless if the classifications provided were not suitable for science . with sufficient effort to ensure each galaxy is classified multiple times ( as many as 80 for many galaxy zoo images ) , these independent classifications need to be combined into a consensus . as discussed in later sections , this can become complex , but for galaxy zoo a simple weighting which rewards consistency , first described in @xcite , was deemed sufficient . importantly , combining classifications provides not only the assignment of a label but , in the vote fraction in a particular category , an indication of the reliability of the classification . this allows more subtle biases , such as the propensity for small , faint or distant galaxies to appear as elliptical regardless of their true morphology , to be measured and accounted for ( see * ? ? ? the net result is that the galaxy zoo classifications are an excellent match for results from expert classification , and have produced science ranging from studies of red spirals @xcite to investigations of spiral spin @xcite . a full review of galaxy zoo science is beyond the scope of this review ; a recent summary is given in the galaxy zoo 2 data release paper by @xcite . it is worth noting that some of the project s most important results have been the result not of interaction with the main classification interface , but represent rather serendipitous discoveries made by participants . we return to these in below . if studying galaxies remains , at least in part , a visual pursuit , then the same is certainly true of planetary science . nasa s clickworkers , which asked volunteers to identify craters on the martian surface , lays claim to be the oldest astronomical crowd - sourcing project . the consensus results matched those available from experts at the time , but failed to go beyond this promising start to produce results of real scientific value . more recently , interfaces inviting classifiers to look at the moon , mercury , mars and vesta have been launched and attracted significant numbers of classifications ; however , although preliminary results have been promising@xcite these projects have yet to produce datasets that have been used by the planetary science community in the same way that galaxy zoo has by the astronomical community . the recent release of the first paper from the cosmoquest moon mappers project @xcite may indicate that this will change . not all astronomical crowd - sourced visual classification is led by professional scientists . jupos is an amateur astronomy project involving a global network of citizen observers to monitor the appearance of planetary atmospheres . recent software developments have provided a much more quantitative perspective on these citizen observations . the winjupos software was developed by a team of citizen scientists led by g. hahn ; it allows multiple images of a giant planet to be stacked with a correction for the rapid rotation of jupiter or saturn ( once every 10 hours ) , then re - projected onto a latitude - longitude coordinate system , so that the precise positional details of atmospheric features can be determined via `` point - and - click , '' relying on the citizen s ability to identify features on the planetary disc visually . by doing this over many nights surrounding jupiter s opposition , the community builds up enormous drift charts , comprising tens of thousands of positional measurements , for these features , ranging from the tiniest convective structure being moved by the jet streams , to the largest vortices ( e.g. * ? ? ? the charts reveal the dynamic interactions within the jovian weather layer , and the long - term stability of their zonal jets ( see e.g. , the regular bulletins provided by the jupiter section of the british astronomical association ) . the positions can be extrapolated forward in time , enabling targeted observations by professional observatories or even visiting spacecraft . the juno mission , scheduled to arrive at jupiter in 2016 , is reliant on the citizen observer community to provide this sort of contextual mapping for the close - in observations from the orbiter . this long - term record of jupiter s visible appearance by citizen scientists has proven to be an invaluable resource for jovian atmospheric scientists . [ snzoo ] the three defining characteristics of `` big data '' have come to be accepted as volume , velocity and variety . time - domain astronomy projects , that indeed require the immediate inspection of challenging volumes of live , high velocity , complex data , can benefit from citizen science , as shown by two recent projects , supernova zoo and planet hunters . while transients such as supernovae or asteroids can often be found through the use of automatic routines , visual inspection is still used by many professional science teams as part of their process of selecting candidates for follow - up . the most successful attempt to use crowd - sourcing to attack these problems to date has been the offshoot of galaxy zoo described in @xcite . data from the palomar transient factory @xcite was automatically processed and images of candidate supernovae uploaded on a nightly basis ; this triggered an email to volunteers who , upon responding , were shown the new image , a reference image and the difference between the two . by analyzing the answers given by the volunteers to a series of questions , candidates were sorted into three categories , roughly corresponding to `` probable supernova , '' `` likely astrophysical but non - supernova transient '' and `` artifact . '' the results were displayed on a webpage and used by the science team to select targets for follow - up . despite the supernova zoo site attracting many fewer classifiers than galaxy zoo , it was highly effective in sorting through data , with consensus typically reached on all images within 15 minutes of the initial email being sent . the large dataset generated by this project was used by @xcite to develop a supervised learning approach to automatic classification for ptf transients . the performance of this routine , which for a false - positive rate of 1% is more than 90% complete , depends on the kind of large training set that can be generated by crowds of inspectors ; this suggests a future path for large surveys in which citizen science provides initial , training data and is followed by machines taking on the remaining bulk of the work . encouragingly , @xcite s method , which makes use of a set of 42 features extracted from survey images , has performance which is insensitive to a small fraction of mislabeled training data , suggesting that the requirements for accuracy of citizen science projects which aim to calibrate later machine learning may be less stringent than otherwise thought . a different approach to crowd - sourced classification in time - domain astronomy is exemplified by the planet hunters project , which asks volunteers to examine light curves drawn from the dataset provided by the _ mission in order to identify interesting events in retrospect . while the task of identifying transits from extrasolar planets is , at first glance , one which seems more suited for automated than for human analysis , the success of planet hunters in identifying more than fifty planet candidates missed by the automatic routines suggests that there remains a role for inspection by eye in cases where the relevant science requires samples of high completeness . several of the planets found by planet hunters are unusual : ph1b , the project s first confirmed planet @xcite and a circumbinary , is the first planet known in a four - star system . others , including the more than forty candidates identified in the habitable zone of their parent star by @xcite , might have been expected to be recovered by more conventional searches . planet hunters , therefore , is acting as an independent test of the kepler pipeline s efficiency @xcite and has inspired improvements in subsequent analysis @xcite . in most visual classification projects , working on archived image data with little time pressure , the random assignment of task to classifier , followed by simple , democratic treatment of the classifications has been judged sufficient . however , the need for rapid processing of images in time domain astronomy projects has prompted the investigation of more efficient analyses of the classification data . using the supernova zoo project s archive as a test , @xcite developed a bayesian method , ibcc , for assessing classifier performance ; in this view , each classification provides information both about the subject of the classification and about the classifier themselves . classifier performance given subject properties can thus be predicted and an optimum set of task assignments calculated . moreover , work by simpson et al . , as well as @xcite and @xcite on galaxy zoo data , suggests that accuracy can be maintained with as few as 30% of classifications . this sort of optimization will be increasingly important for online citizen science , especially in projects that use a live stream of data , rather than an archive , since the classification analysis will need to be done in real time . steps towards real - time classification analysis have been taken in the space warps project . space warps is a rare object search : volunteers are shown deep sky survey images and asked to mark features that look as though they are gravitationally lensed galaxies or quasars ( marshall et al , more et al in prep ) . extensive training is provided via an ongoing tutorial that includes simulated lenses and known non - lenses , and immediate pop - up feedback as to whether these training images were correctly classified . because real lenses are rare ( appearing once every @xmath2 images , depending on the dataset ) , the primary goal is to reject the multitude of uninteresting images so that new ones can be inspected and this drives the need for efficiency . marshall et al ( in preparation ) derived a simplified version of the ibcc classification analysis that updates a probablistic model of both the subjects and the agents that represent the classifiers in a statistically online manner . this analysis was run daily during each of the space warps projects , and subjects retired from the stream as they crossed a low probability threshold . this algorithm is being implemented into the web application itself for future datasets . the increased efficiency of visual classification projects that will come with real - time analysis will enable feedback on the projects progress to be given much more promptly an important part of the collaboration between professionals and amateurs in crowd - sourcing projects . although , as described in the previous section , astronomical analysis led the development of citizen science as a data analysis tool , it has quickly been adopted by other fields . in some cases , this adoption has been explicit . the tools developed for stardust@home were developed into a general purpose library for citizen science , bossa . both this and the zooniverse platform ( which hosts many of the examples described above ) support projects from fields as diverse as ecology and papyrology . this diversity allows general lessons about project design to be drawn ; indeed , this is an active area of research for academic fields as diverse as computer science , economics and social science . a recent paper by @xcite , for example , compares planet hunters and seafloor explorer , a zooniverse project which explores the health of fisheries off the coast of north america . projects from other fields can also suggest strategies which could be adopted by future citizen astronomy projects . in particular , future projects involving analysis of survey data which has been collected for a multitude of purposes may require a more sophisticated model for data analysis than the simple decision tree presented by projects such as galaxy zoo . this zooniverse project invites the visual classification of animals in photographs from more than two hundred motion - sensitive `` camera traps '' installed in the serengeti national park , and enables a particularly sophisticated volunteer response . driven in part by the need for an interface which allows volunteers to state the obvious ( for example , identifying elephants , lion or zebra ) and also to provide more obscure classification ( for example , distinguishing between different species of gazelle ) , a variety of classification paths are presented . in addition to just clicking buttons identifying species , volunteers can opt for a decision tree - like approach , or choose from a variety of similar species ( `` looks like an antelope ... '' ) or search the descriptions provided in order to make an informed classification ( `` show me all animals whose descriptions involve ears ' ' ) . this hybrid model has proved successful not only in encouraging classification , but also in encouraging learning ; over a snapshot serengeti classifier s `` career '' they are increasingly likely to chose more direct routes . another aspect of project strategy , and design , relates to the engagement of the volunteers . the online citizen astronomy projects developed so far have tended to emphasise co - operation between volunteers , and the results being due to a team effort . elsewhere , experiments with a more competitive approach to citizen science , `` gamifying '' the activity , have been performed . the eyewire project , based at mit , seeks to supplement machine learning identification of neurons in three - dimensional scans . notably , this project incorporated some `` gamified '' elements into its design . participants in the project , who are asked to identify connected regions throughout a three - dimensional scan , earn points based on participation and also have a separate , publicly visible , accuracy score . in addition to overall leader boards , the project also runs short challenges including a regular friday `` happy hour '' in which participants compete on specific problems . eyewire is also notable for its other strong community elements , with a chat room open and available to all participants in the project ( supplemented , incidentally , by a `` bot '' built by a participant which answers frequently asked questions from new users ) . its first result , which drew on mapping of so - called ` starburst ' neurons , was published in mid-2014 @xcite . new understanding of the world comes from the interpretation fitting of data with a physical model . such `` data modelling '' often involves technical difficulties that computers may find hard to overcome , associated with complex and/or computationally expensive likelihood functions . humans , by applying their developed intuition , can contribute a great deal to the exploration of a model s parameter space by closing in quickly on those configurations that fit the data well . this process can be particularly satisfying , rather like solving a puzzle . meanwhile , many `` machine learning '' techniques effective in one field can often be adapted to astronomical problems : there are plenty of citizens with the skills to do this . how have citizen scientists been involved in model making and data fitting in astronomy , and other fields , to date ? @xcite provided volunteers with a fairly flexible set of annulus - drawing tools , for annotating circularly - symmetric `` bubble '' features in colour - composite ( 24.0 , 8.0 and 4.5@xmath3 m ) infrared images from surveys carried out by the spitzer space telescope . these bubbles are hypothesized to have been caused by recently - formed high mass stars at the centre each . the ( bubble ) model in this case is simple and recognizable , making both the interface construction and its operation relatively straightforward . the large sample of bubble models have been used to investigate the possibility of further star formation being triggered at the bubble surfaces @xcite . a subsequent effort @xcite used data provided by the project to train a machine learning algorithm , _ brut _ , in bubble finding . _ brut _ is able to identify a small number of sources which were not identified in the @xcite catalog . these bubbles were difficult for humans to identify , owing to their lying close to bright sources , and so having low contrast relative to their surroundings . in addition , _ brut _ has proved effective at identifying suspect bubbles included in the previous ( pre - citizen ) surveys . given the relatively small size of the mwp sample , the main use of machine learning here has been to provide an independent check on the citizen classification data ; for larger samples , as discussed below , an approach in which machine learning is trained on citizen science data , and gradually takes over the classification task could be considered . the space warps project ( ) has an informal data modeling element . the classification interface is restricted to enabling identification of candidate gravitationally - lensed features , but all the images are available via the project s discussion forum . a small team of volunteers ( including several citizens who helped design the project ) has engaged in modeling some of the identified lens candidates using web - based software developed and supported by the project science team . results from a small test program show that the model parameters derived by the ensemble of citizens are as accurate as those derived by experts ( kueng et al , in prep ) . a pilot collaborative modeling analysis was carried out and written up by a small group of space warps volunteers @xcite . this has been perhaps the most advanced attempt at data modeling in astronomical web - based citizen science @xcite . here , simple n - body simulations of galaxy mergers were performed in a java applet , and the results selected according to visual similarity to images of galaxy mergers ( previously identified in the galaxy zoo project ) . a key hypothesis here is that the inspectors of the simulation outputs would be able to find matches to the data more readily than a computer could , for two reasons . first is that humans are good at _ vague _ pattern matching : they do not get distracted by detailed pixel value comparisons but instead have an intuitive understanding of when one object is `` like '' another . the second is that initializing a galaxy merger simulation requires a large number of parameters to be set and its this high dimensionality that makes the space of possible models hard to explore for a machine . humans should be able to navigate the space using their intuition , which is partly physical and partly learned from experience gained from playing with the system . initial tests on the marging system arp 86 showed the crowd converging on a single location in parameter space , and that the simulated mergers at this location do indeed strongly resemble the arp 86 system . the authors have since collected thousands of citizen - generated models for a sample of a large number sdss merging systems ( holincheck et al , in preparation ) . one of the most successful examples of crowd - sourced , `` manual '' data modeling is the online multi - player 3-d protein modeling game , foldit @xcite in this pioneering project , players compete in teams to find the best lowest free energy molecular structures for particular protein `` puzzles . '' these puzzles are naturally visualizable in three dimensions , but they nevertheless involve thousands of degrees of freedom , in a parameter space that is notoriously hard to explore . under the hood is the professional rosetta structure prediction methodology ; the player s scores are simply the negative of the rosetta - computed energy . foldit provides an accessible interface to the rosetta toolkit , which provides multiple ways to interact with the protein structure as the global minimum energy solution is sought . the rosetta model parameter free energy hyper - surface is completely analogous to the complex likelihood surface of any non - linear model , the kind of model that is to be found in planetary system dynamics , gravitational lenses , merging galaxies , and many other astrophysical data analysis situations . results from foldit have been very encouraging , with the players discovering several new protein configurations , leading to improved enzyme performance @xcite and new understanding of retroviral drug design @xcite . the team have suggested several features of foldit that appear to them to have underpinned its success . recipes for manipulating the protein structures ( that codify strategies ) can be shared within teams , and later made available by the foldit team to the whole community these algorithms evolve rapidly as different players modify them , and can rival ( if not out - perform ) strategies developed by professional scientists @xcite . the game provides multiple sources of motivation ( competition between players , collaboration within a team , short term scores , long term status ) which appeal to a variety of players . we now turn to data modeling by citizens implementing machine learning techniques in astronomy , via analysis challenges organised by members of the professional astrophysics community . the measurement of weak gravitational lensing by large scale structure ( `` cosmic shear '' ) relies on the measurement of the shapes of distant , faint galaxies with extreme accuracy . the step @xcite and great @xcite blind galaxy shape estimation challenges have had an enormous impact on the field , revealing biases present in existing techniques , and providing a way for researchers outside the world of professional cosmology to participate . in particular , the great08 challenge saw very successful entries from two ( out of a total of 11 ) teams of researchers from outside of astronomy ( albeit still professional researchers ) , including the winner . a companion , somewhat streamlined galaxy shape measurement challenge , `` mapping dark matter , '' which was hosted at the kaggle website @xcite . the wider reach of this platform led to over 70 teams making over 700 entries to the competition ; many of the teams did not contain professional astronomers , although most were still from academia . in a comparison with the great challenges , the @xcite found a factor of several improvement in shear accuracy over comparable previous challenges , and suggested two interesting explanations for this success . first , the challenge was designed to be as accessible as possible , with an extensive training set of data that needed very little explanation ; in this way the challenge was geared towards _ idea generation_. second , they noted that the competitive nature of the challenge ( a webpage leaderboard was updated in real time as entries were submitted ) seemed to stimulate the analysts into improving their submissions . kaggle offers cash prizes , which will have had some effect as well ( the pot was $ 3000 for this challenge , even if indirectly ) . two more astronomical kaggle challenges have since been set . the first involved inferring the positions of dark matter halos based on their weak lensing effects @xcite this challenge attracted the attention of 357 teams , perhaps due to its larger prizes , and led to an improvement in halo position accuracy of 30% . it also sparked some debate in its forums as to the design of the challenge : the models used to generate the data , the size of the test datasets ( and consequent stability of the leaderboard ) , the choice of leaderboard metric and so on . these issues are also of generic importance for scientists looking to crowd - source algorithm development . it is interesting to note that the kaggle forums are a useful resource for the kaggle development team : the citizens who are active there do influence the design of the site infrastructure and challenge rules ( d. harvey , priv . the most recent kaggle astronomy challenge was to reproduce the galaxy zoo crowd - sourced galaxy morphologies , given the same color images . like foldit s `` recipes , '' the kaggle challenges are crowd - sourcing the development of new algorithms . as data science plays an increasingly important role in industry and commerce , we can expect the number of citizens interested in applying their skills to science problems in their spare time to grow . the challenge is to present those problems in meaningful ways , to enable high value contributions to be made . while members of this community may not identify as `` citizen astronomers , '' there is clearly an opportunity for citizen data scientists to play an important support role . the previous sections have focused on specific , and somewhat isolated activities in which citizens have participated . in most cases , the community s involvement has been a _ contribution _ to a scientific investigation defined by professionals . the most important part of any scientific investigation is the question at its heart : what is it we are trying to find out about the universe ? in this section we look at some cases where the process of enquiry , the science itself , has been instigated or led by citizens . in principle , this is an area of great potential . the constraints of funding proposals and management of research groups can often mean that professional scientists focus very narrowly on particular topics of research , specializing in particular techniques or datasets . steering away from this course implies taking risks with time management , and allocation of resources to an ultimately fruitless research area can be detrimental to careers . citizen scientists are largely free of these managerial and budgetary constraints , and are able to devote their attentions to whatever topics interest them . moreover , we might expect outsiders to ask some unusual questions , and make connections and suggestions that highly focused professionals may not have thought of . what are some enquiries that citizens have led in astronomy to date , and how have they been enabled and supported ? in this project , cassini s observations of lightning emissions were connected to active amateur observations of the convective cloud structures within the giant planet atmosphere @xcite , and the vertices of saturn s bizarre north polar hexagon @xcite ( a 6-sided planet - encircling wave that has persisted for at least 30 years but that has only recently been observed , by amateur astronomers ) . in the first case , citizen scientists wished to identify the source of the radio emission detected by cassini , after being alerted to them on the planetary virtual observatory and laboratory . in the latter case , the long - term evolution of the hexagon vertices was used to understand what sort of wave this is , and to identify its origins . the best known serendipitous discovery emerging from the galaxy zoo project is `` hanny s voorwerp '' @xcite , a galaxy - scale light echo which reveals a recent ( @xmath4 years ago ) shutdown of agn activity in ic 2497 , a neighboring spiral galaxy @xcite . the discovery of the voorwerp was first recorded in the galaxy zoo forum a few weeks after the project started , and inspired a more systematic search for similar phenomena in other galaxies . this project , made possible by the deep engagement in the forum community of galaxy zoo science team member bill keel , succeeded in finding more than forty instances of clouds which appear to have been ionized by agn activity , in systems a third of which show signs of significant drops in activity on a timescale of tens of thousands of years . the ability of the zoo volunteers to carry out their own research , moving far beyond the mere `` clockwork '' required by the main interface , is best illustrated by the discovery of the galaxy zoo green peas @xcite . these small , round and , in sdss imaging , green systems are dwarf galaxies with specific star formation rates ( sfr per unit mass ) which are unprecedented in the local universe , matched only by high - redshift lyman - break galaxies . volunteers not only identified these systems , but organized a systematic search and further review of them . this effort included the use of tools designed by sdss for professional astronomers to acquire and study spectroscopic data . while the discovery of the peas demonstrates the exploration ability of the galaxy zoo citizen community , it is important to note that the simpler , initial interaction provided by the main classification interface was necessary in order to develop that community in the first place . the participants in the citizen scientists investigation of the peas did not arrive on the site wanting to dig into spectra or confident of their ability to do so ; these were the results of their participation . the project acted as an `` engine of motivation '' in inspiring its participants to become more involved . the data modelling examples of all involved modeling infrastructure provided by either the project s developers or their science teams . planet hunters provides a case where citizens have carried out their own modeling analysis , using their own tools . critical to this endeavour was the ability of a small , and increasingly expert , group of volunteers to identify objects worthy of further analysis . for galaxy zoo , the forum had served this purpose but , as as the project matured , participation in discussions became restricted to a small and decreasing fraction of the community . planet hunters was the first zooniverse project to introduce an integrated discussion tool , known as _ talk_. classifiers were asked , after viewing each light curve , whether they wanted to discuss what they had seen ; more advanced users could then harvest interesting candidates from these posts . for example , the candidates presented in @xcite were initially collated by volunteers . their involvement was not limited to collecting planet hunters candidates . making use of the kepler archive , these advanced users were able to investigate the full set of data for candidate stars , producing periodograms and making fits to transits to derive planet candidate properties . some of this analysis , for example checking the kepler field for background sources , can be carried out online with tools originally intended for professional astronomers , but much was done off line using excel or other software . nor was this sort of work restricted to planet candidates ; interesting variable stars , including several new rr lyrae systems , and cataclysmic variables ( e.g. * ? ? ? * ) have been discovered and analysed by planet hunters volunteers . examples such as those above show that advanced work is possible within distributed citizen science projects , but that this requires volunteers to take on such tasks themselves . in order to increase the number , and perhaps the diversity , of volunteers moving beyond simple classification , experiments have been conducted to provide more scaffolded experiences . one of the most ambitious was the galaxy zoo : quench project ( trouille et al . in prep ) which offered volunteers the opportunity to `` experience science from beginning to end . '' in this project , classification of a sample of potential post - merger galaxies selected from the main galaxy zoo sample was followed by open exploration of both the classification data and the metadata for these galaxies ( available from the sloan digital sky survey ) by the volunteers , enabled by a `` dashboard . '' thousands of users ( around 20% of those who participated in the classification stage and discussion ) led to the formulation of a set of astrophysical interesting conclusions ; a small number of participants ( 10 ) collaborated on writing a paper ( in preparation ) . these later stages required intensive support from professional scientists ( and in fact it was constraints on their time that prevented earlier submission of the paper ) . quench demonstrated that a hierarchical approach , with simple tasks leading to more advanced analysis , can be successful in encouraging large numbers of volunteers to move beyond simple classification ; the number participating in exploring the data was much higher as a percentage of participants than in planet hunters . however , once engagement with the literature ( either by reading or writing ) is required there remains no substitute for significant involvement by professionals . having surveyed some of the activities involving citizen scientists , we can now consider some questions about this community itself . who participates in citizen science , and what motivates them ? who is participating in citizen astronomy ? we might expect the demographics to vary with activity , and with the level of commitment required . we have some understanding of at least the former division from two studies that were carried out approximately simultaneously , one of the community participating in galaxy zoo , and another of the american association of variable star observers ( aavso ) . @xcite surveyed the galaxy zoo volunteer community to investigate their motivations ( below ) , via a voluntary online questionnaire . the 11,000 self - selected galaxy zoo users identified as 80% male , with both genders having an approximately uniform distribution in age between their mid - twenties and late fifties . the authors point out that this is close to the us internet user age distribution , except for slight but significant excesses in numbers of post-50s males , post - retirement people of both genders , and a deficit in males under 30 . the survey respondents also tended to be more highly educated than average us internet users , with most holding at least an undergraduate degree , and around a quarter having a masters or doctorate . very similar findings were reported by @xcite from a survey of cosmoquest project participants . these findings can be compared with a survey of the members of aavso : @xcite received over 600 responses ( corresponding to about a quarter of the society s members ) . the education levels of the aavso repondents matches the galaxy zoo community very closely ; the aavso age distribution is more peaked ( in the mid fifties ) , with a similar post-60 decline but also a marked absence of younger people . the online nature of the galaxy zoo project seems to have increased the participation of younger ( pre middle - age ) people . likewise , the galaxy zoo gender bias , while itself extreme , is less so than at aavso , where some 92% of survey respondents were male . one additional piece of information provided by the aavso survey is the profession of the variable star observers : most ( nearly 60% ) of the survey respondents were found to be working in science , computer science , engineering and education . the galaxy zoo and aavso communities differ by more than just the nature of their activity . the smaller aavso community is arguably more engaged in its research , in the sense that a larger fraction of its membership is active in taking observations and contributing to analyses . it would be very interesting to know how citizen scientist motivation varied with the level of participation : dividing the galaxy zoo community into volunteers that contribute to the forum and those who do nt could be interesting ; perhaps more so would be to repeat the analysis of @xcite over a wide range of projects , and look for trends there . the emergent picture thus far , however , is of a well - educated ( and often scientifically trained ) but male - dominated citizen science community , whose female and younger membership is likely to have been , at least in part , enabled via projects being hosted online . continuing to lower the barriers to entry for currently under - represented demographic groups would seem both important , and within reach . what motivates citizen scientists ? the two demographic studies referred to above also covered this question ; having previously @xcite identified 12 categories of motivation in an earlier pilot study , @xcite asked the 170,000 galaxy zoon volunteers at the time to comment on how motivated they were by each of these categories , and which was their primary motivation . the 6% who responded gave consistent answers to those given by around 900 forum users who responded in a separate appeal , allowing conclusions about this presumably more engaged sub - population to be drawn . a desire to _ contribute _ to science was found to be the dominant primary motivation , being selected by 40% of respondents . _ astronomy _ , _ science _ , _ vastness _ , _ beauty _ and _ discovery _ were all motivation categories that were found to very important to the volunteers , while _ fun _ , _ learning _ and _ community _ were less important . the aavso demographic survey @xcite found similar results : over a third of variable star observers cited _ involvement in science and research _ as their primary source of motivation . however , a similar number gave an _ interest in variable stars _ as theirs , perhaps reflecting a stronger focus on the science questions involved than is present in the galaxy zoo community . both groups of citizen scientists are clearly quite serious in their reasons for taking part : their motivations are actually very close to those of professional scientists , as many readers of this review will recognize . perhaps surprisingly , the participants in online data analysis citizen science projects seem to a large extent to be a distinct community from those who participate in more traditional amateur astronomical activities . galaxy zoo classifiers , for example , are not , for the most part , regular amateur observers . while research on the skill , and conceptual understanding , that people aquire while participating in citzien science activities is still in its early stages , there are some hints that continued engagement is correlated with both performance in the task at hand , and understanding of the physics and astronomy underlying the task . @xcite offered galaxy zoo and moon zoo volunteers the opportunity to take questionnaires that tested their understanding of the astrophysics associated with each project , and found that performance on this questionnaire correlated with high levels of participation in the projects . in the space warps project , the probabilistic model for the crowd includes a measure of each classifier s skill ; a strong correlation is seen between a classifier s skill , and the number of images they have seen ( marshall et al , in prep . ) . it seems as though the skillful classifiers remain engaged in the project for a long time , while almost no long - term participants have low skill an observation consistent with the volunteers being motivated by contributing to science . as seen in and above , non - astronomical projects may have much to teach us about `` gamification '' as a motivator the inclusion , either explicitly or implicitly , of game - like mechanics such as scores , `` badges '' or other rewards , leader boards , and so on . the foldit team present a strong case for games as drivers of activity in citizen science , and the kaggle challenges depend on competition to stimulate engagement . what has been the experience in citizen astronomy so far ? an early experiment with galaxy zoo showed that the addition of a score de - incentivised poor classifiers , but also resulted in the best classifiers leaving , presumably having been satisfied once a top score was achieved . a recent study by @xcite of the zooniverse s old weather project , which included basic rankings for classifiers , highlighted these dangers , identifying volunteers who were alienated by the addition of this game - like score.they felt discouraged when top scores could not be matched , and worried about data quality if the scoring scheme rewarded quantity of classifications rather than accuracy . taking seriously the above finding that citizen scientists are motivated by a perception of authentic participation in research , it seems right to be cautious about introducing elements which are , or which are perceived to be , in tension with this primary motivation . furthermore , the introduction of a significant incentivizing scheme relies on an accurate model of what `` correct '' behaviour would look like . this may prove to be a significant barrier to accuracy if such a model is not available . for example , in planet hunters , such a model would not have included unusual systems such as ph1b . where a strong incentive scheme results in near - uniform classifier behaviour , a loss of flexibility in later data analysis could be incurred . a strong comparison of the type of reward structure utilized by eyewire and the approach used by projects such as galaxy zoo is needed , in order to inform future project design . the surveys described in the previous section reveal a community of people many of whom may have left academic science behind as soon as they finished their education , but who still maintain a passion for astronomy and the boundaries of knowledge . their thirst for new information , and the desire to be part of the scientific process drives them to actively observe the night sky or to participate in analysis of large datasets . for the more motivated people involved in citizen science , being part of a community , albeit a distributed one , brings great enjoyment and satisfaction . with the connectivity brought by the internet , there is a social aspect of citizen science that unites people with this shared interest , which may be far removed from their `` normal '' lives . however , _ community _ was not found to be a strong motivator for the galaxy zoo volunteers but it is nevertheless very important for the galaxy zoo forum users . more recent zooniverse projects have sought to widen participation in community discussion , hypothesizing not that it will more strongly motivate people , but because it will help them make better contributions . tests of hypotheses like this should be helpful in guiding citizen science project design . during this review a picture has emerged of two types of very active and engaged citizen astronomy community , which we might label observers and classifiers . although theses communities come together in differing ways ( by self - assembly through local groups linked by national and international networks , or by joining online projects built by professional organisations ) , they have reached a similar degree of internet - enabled connectedness , both with each other and with the groups of professional astronomers with whom they collaborate . they also share the common motivation of being involved in , and contributing to , science . in this section we look ahead , to the next decade or so , and discuss the likely paths that citizen astronomy will take , as the available technology advances and professional astronomy evolves . in it we try to identify the niches that citizens might best occupy in this changing environment , and also some key challenges that those who find themselves planning citizen science projects are likely to have to face . in professional astronomy , the wide field survey era is upon us : sdss provided the data for galaxy zoo , and other , larger surveys are planned or underway . key science drivers for projects such as lsst and the square kilometer array include mapping cosmological structure back into the reionisation era , and further opening the time domain ; these will yield datasets of significantly increased volume , throughput rates , and complexity . follow up observations of new discoveries made at greater depths will be made with giant facilities such as alma and the various planned extremely large telescopes , while distributed arrays of robotic telescopes , operating in remote regions with excellent atmospheric conditions , and trained to observe a target in a regular fashion over multiple nights will be able to take advantage of wealth of new transient phenomena . these future advances in technology may in one sense widen the gap between citizen scientists and professionals again . for example , networked telescopes capable of quasi - continuous observations over 24 hour periods could be used to develop a consistent high - quality dataset for cloud tracking on venus , mars or the giant planets ; as the images would be homogenous , we can envisage automated software identifying morphological peculiarities over time , replacing the crowd - sourced citizen analysis currently underway . however , such an investment would require both international funding and considerable time and effort : the availability of citizen observers will remain a factor . however , the advances in hardware becoming available to citizen observers suggest other roles they could play . larger optics , more sensitive cameras , and spectral coverage extending to longer wavelengths in the infrared could permit citizen investigations of uranus and neptune , the trans - neptunian and kuiper belt objects , and a wider variety of bright variable stars . transits of extrasolar planets in front of their parent stars would be permitted from modest observatories provided they had stable conditions . new platforms might also become available to the citizen scientist , including balloon - borne observatories that provide crisper and more detailed observations of astronomical targets . we may well see networks of citizen deep sky observers investigating new bright transients found in the wide field surveys , while continuing to expand their own surveys . aside from pushing the observational boundaries , one challenge that amateur astronomy may face is its own big data problem . for example , solar system video monitoring projects are likely to need automated feature detection of some kind ; other observing campaigns may also generate more data than is easily manipulated . will this community take to crowd - sourcing its visual inspection ? the zooniverse platform is currently being redeveloped to enable easy upload of images and launch of projects ; such a facility may be used by citizen scientists as well as by professionals . the point at which human review of data is no longer necessary has been forecast for decades , but as we have seen above , the number of problems for which manual review of images or data is still carried out is considerable . even if the proportion of data for which human inspection is necessary decreases dramatically over the next decade ( due to advances in automatic analyses ) , the continued growth in the size of astronomical datasets should ensure that there remains plenty for citizen scientists to do . consider the example of optical transients . the lsst system overview paper @xcite gives a conservative estimate of @xmath5 alerts per night . even if , after automated brokerage , only 1% of these require human classification , then that still might lead to 10,000 objects requiring inspection and interpretation every night roughly one every 100 seconds . given the increased reliability , and likelihood of serendipitous discovery , provided by citizen inspection , we should take seriously the incorporation of open inspection into plans for lsst transients . similar arguments ( with large error bars ) can be made for other surveys : inspection of transients for lofar already requires some human intervention @xcite . implicit in this way of thinking is the sharing of work between human and machine classifiers . a simple example of human - machine task allocation was mentioned in , where machine analysis of ptf images identified those that contained candidate supernovae needing inspection by volunteers . it is worth noting that the inclusion of human inspection changed the nature of the machine learning task ; instead of optimising for purity ( producing a small but accurately classified set of candidates ) , the task for machine learning became one of identifying a subset of the images which contained many false positives but also a complete set of all supernovae . in this example , human and machine classification proceeded in series rather than in parallel , but more complex interactions can be imagined . the accuracy of machine learning typically depends on the quality of the training or `` gold standard '' data which can be provided for the problem in question . citizen science projects can assist by providing training sets which are orders of magnitude larger than might otherwise have been available , while work by @xcite established that the confidence intervals provided by classifications from multiple volunteers can also improve machine learning accuracy . predicting human responses ( in the form of probabilities of classification ) is an easier task than straightforward sorting . we expect , therefore , intermediate - size surveys to benefit in the future from a `` citizen science phase , '' in which data is classified by volunteers prior to the automation of the task . this pattern has already been followed by the ptf supernova project discussed above , but perhaps it is more useful to think of the citizen scientists as providing training sets on demand , so that as conditions change from night to night , or the performance of the instrument evolves over time , a small percentage of the total data is always processed by humans in order to provide a constantly updated training set . if we are using classifications of gold standard data to assess the performance of human classifiers , it is straightforward to include machine classification in the same system . in this way , the task of classification could be shared dynamically and in real time between machine and human classifiers , improving the efficiency of the system . significant work has already been carried out for the nearly analogous problem of assigning tasks to an ensemble of imperfect machine classifiers whose characteristics are known and for mechanical turk - like systems where a fixed payment is provided for a task but the problem of adding in volunteers is significantly harder . for the machine - only case , each classification task can be treated as having a known cost ( perhaps the processing time necessary for a given routine ) , but when assigning tasks to volunteers , who are able to leave whenever they like , other costs must be taken into account . in order to create a viable system , it is , in fact , necessary to measure how _ interesting _ a task or set of tasks is , and this requirement may conflict with the need for efficiency . as an example , consider a galaxy zoo - like system which assigned the hardest galaxies to the best classifiers . this would result in a steady diet of faint fuzzy objects for the best classifiers ; if they are motivated in part by the variety of images seen , then such a system would tend to systematically drive away its best classifiers . a study of snapshot serengeti even revealed that seeing a more impressive image early in a classifier s career ( as measured by the number of volunteers who added it to their list of favourites ) tended to decrease the number of classifications received from that volunteer in the long run . considering individual classifications in isolation is clearly not sufficient ; the entirety of a volunteer s career must be considered when assigning tasks . we should be wary of over - specialisation even when efficiency is paramount . complexities like these indicate a clear need for research into novel systems for task assignment , in order to scale citizen science to the challenges of the next generation of surveys . as we have seen in previous sections , volunteers can and do move beyond simple classification problems , and such behaviour could become increasingly important as the volume and complexity of astronomical data continues to increase . we can imagine providing user - friendly , web - based tools enabling fairly sophisticated data analysis to be performed by anyone with a browser . the experience documented above invites us to consider the possibility of teams of citizens performing analyses that currently require a significant amount of research student time . checking survey images and catalogs for processing failures and fitting non - linear models to data are just two possibilities . just as research students adapt and develop the tools they are first presented with , the kaggle and foldit experiences point strongly towards a model where citizens are also enabled to evolve their tools . open source tool code is a minimal requirement in this model ; finding ways beyond this to support citizen algorithm development seems to be likely to pay off . in terms of supporting citizen - led enquiry , an example of best practice exists in the way that the sloan digital sky survey s _ sky server _ provided tools for both professional ( or advanced ) researchers alongside simplified versions aimed primarily at educational use . this structure has the twin benefits of providing near - seamless transitions from simple to more advanced interfaces , and of providing extra pressure to make the resulting interfaces easily usable ( something which benefits all users , not just citizen scientists ! ) . designers of science user interfaces for upcoming large projects would do well to bear these twin audiences in mind . indeed , the more citizen - accessible the interfaces to the upcoming public wide - field survey databases can be made , the better chance we will give ourselves of enabling and supporting `` bottom - up '' citizen science . this term , introduced by muki haklay and collaborators , represents an ambition to produce citizen science projects that are driven by the participants . moving beyond the top down structure of most astronomical citizen science projects is , as we have shown , a significant challenge - but one that is , perhaps , worth taking on . as well as enabling access to the data , citizen science projects looking to engage larger crowds of volunteers will likely face some other challenges . we might expect contributing to science via large international public datasets to appeal to citizens of many nations : while translation of project materials is simple , coordinating a scientific _ discussion _ across multiple language barriers could prove difficult . having a critical mass of professional scientists interacting in each language would seem the most important factor . even within a single language group , collaboration is difficult to achieve with very large numbers . the hierarchical system of citizen discussion moderators bridging the gap between science teams and the crowd has worked well in the zooniverse projects , although it requires significant commitment and effort from the volunteer moderators . access to professional scientists can be somewhat improved by regular webcasts ( as provided , for example , by the galaxy zoo science team ) ; this provides at least partially the scientific dialogue that is most valuable to the citizen collaborators . certainly these can supply much - needed feedback as to the utility of the citizens efforts , as the professionals report on how the citizen - provided data is being used . we might imagine regular broadcasts from the projects providing the data as playing a significant role in motivating and sustaining a crowd of volunteers . however , for the foreseeable future it remains the case that astronomical surveys and other organisations will seek to use citizen science as a way of expanding the amount of science that can be done ; systems which rely on significant intervention from paid professionals will likely fail ( or at least be a luxury available to few ) . a focus on systems which can maximise scientific return and volunteer participation without substantial intervention remains necessary . over the last two decades , citizen astronomy has undergone a period of rapid growth , primarily due to the sharp increase in the ease with which people can form communities and work together via the world - wide web . a number of very productive `` pro - am collaborations '' have formed , to observe a variety of bright astronomical objects in ways that capitalise on the flexibility , availability and skill of the amateur observing community . professional - led visual classification projects have appeared , attracting three orders of magnitude more citizens to the field than were previously engaged in amateur observational research . citizen - classified training sets have been used to improve the performance of machine learning approaches , suggesting that we should think in terms of `` human - machine partnerships . '' citizens have been challenged to take part in data analysis tasks of increasing sophistication and difficulty , and experiments in professionally - guided `` bottom up '' citizen research have begun . in this review , we have consistently seen that the best citizen science in astronomy has come from organised communities asked to play to their strengths , and which operate in niches insufficiently occupied by either professional observers or automated classification software . the citizen astronomers are passionate about the subject , and are encouragingly motivated by being of service to science . we must recognize that a critical feature of `` citizen science '' is the enabling of amateurs to make authentic contributions to the research topic in question : this in turn should drive us to seek out those tasks that can not be done by other means . the observational and classification citizen scientist communities are similar in their diversity regarding both their motivation and their ability to contribute ; this diversity means that good citizen science projects are ones that provide both a low barrier to entry , but that also provide ( or support the development of ) tools that enable their emergent experts to maximize their contributions to science . indeed , the most dedicated volunteers have proved capable of developing and using fairly advanced astronomical techniques , suggesting that we are likely to continue to see increasing numbers of citizens co - authoring papers in high impact research journals . while not everyone who takes part in a project wants to move to more advanced work , providing the opportunity to do so is important . each of the case studies presented in this review has been an experiment in citizen science : amateur and professional astronomers alike have had good ideas for ways to make use of the public s skills and abilities , tried them out , and made progress in astronomy and in doing so revealed something about how citizen science can work . human potential is vast : citizen astronomy seems to us to be an experiment well worth continuing . we are most grateful to the following people for their suggestions , comments , and clarifications : grischa hahn , ... we thank david w. hogg , arfon smith , laura whyte , ... for many useful discussions about the practice of citizen science in astronomy . pjm and lnf were supported by royal society research fellowships at the university of oxford . the work of pjm was also supported in part by the u.s . department of energy under contract number de - ac02 - 76sf00515 . cjl acknowledges ... natexlab#1#1 jd , yeomans dk , chamberlin ab , chodas pw , jacobson ra , et al . 1997 . in _ bulletin of the american astronomical society _ , eds . mf bietenholz , n bartel , mp rupen , aj beasley , da graham , vi altunin , t venturi , g umana , je conway , vol . 29 of _ bulletin of the american astronomical society _ j , de pater i , brecht sh , deming d , meadows v , et al . 2004 . _ lessons from shoemaker - levy 9 about jupiter and planetary impacts _ , chap . cambridge planetary science . cambridge univ . press , new york , 159184 a , wallin j , borne k , lintott c , smith a , et al . 2010 . in _ galaxy wars : stellar populations and star formation in interacting galaxies _ , eds . b smith , j higdon , s higdon , n bastian , vol . 423 of _ astronomical society of the pacific conference series _ j , holincheck a , borne k , lintott c , smith a , et al . 2010 . in _ galaxy wars : stellar populations and star formation in interacting galaxies _ , eds . b smith , j higdon , s higdon , n bastian , vol . 423 of _ astronomical society of the pacific conference series _	we review the relatively new , internet - enabled , and rapidly - evolving field of citizen science , focusing on research projects in stellar , extragalactic and solar system astronomy that have benefited from the participation of members of the public , often in large numbers . we find these volunteers making contributions to astronomy in a variety of ways : making and analyzing new observations , visually classifying features in images and light curves , exploring models constrained by astronomical datasets , and initiating new scientific enquiries . the most productive citizen astronomy projects involve close collaboration between the professionals and amateurs involved , and occupy scientific niches not easily filled by great observatories or machine learning methods : citizen astronomers are most strongly motivated by being of service to science . in the coming years we expect participation and productivity in citizen astronomy to increase , as survey datasets get larger and citizen science platforms become more efficient . opportunities include engaging the public in ever more advanced analyses , and facilitating citizen - led enquiry by designing professional user interfaces and analysis tools with citizens in mind . the making of this review is still in progress . the most up to date pdf file can be downloaded from github at http://tinyurl.com/citizenastronomyreview we invite feedback on the review via its repository s _ issues _ at http://tinyurl.com/citizenastronomyreviewfeedback and aim to submit to araa on september 26 .
the first evidence for the pentaquark hadron @xmath7 , discovered by the leps collaboration at spring-8 @xcite , was subsequently confirmed in some other experiments @xcite . however , many other experiments failed to find the @xmath7 signal ( for surveys see @xcite ) . most of them came from the data analysis of high - statistics high - energy experiments . these null results at high energies were not so much surprising because it is natural to expect a sizable suppression in the production of the more complicated five - quark system compared to the conventional three - quark hyperons @xcite . but the state of affairs became dramatic after the recent publication of the high statistics results of the clas collaboration @xcite . the first experiment is designed to search for the @xmath0 signal in @xmath9 in direct @xmath10 interactions at relatively low photon energy , @xmath11 gev . the second one aimed to search for the @xmath0 signal in @xmath12 and @xmath13 reactions . within the experimental significance , no @xmath0 signal was observed . note however , that recently the diana collaboration confirmed a former result for @xmath0 production in @xmath14 interaction with xe nuclei @xcite . another positive , but low statistics result on @xmath0 production in @xmath15 interaction was obtained in kek @xcite . therefore , the situation concerning the existence of the pentaquark state remains controversial . coming back to the high statistics clas experiments , one can conclude that if the @xmath0 exists , then the null result means that we do not understand the mechanism of @xmath0 photoproduction in elementary @xmath16 reactions . indeed , in all theoretical studies ( for references , see the recent review paper @xcite ) the cross section of this reaction is defined by the @xmath17 and @xmath18 exchange dynamics . in the first case , the amplitudes are proportional to the product of the @xmath0-nucleon - kaon coupling constant @xmath19 and the form factor @xmath20 , where @xmath21 are the four momenta of the baryon ( nucleon or @xmath0 ) and the kaon , respectively . one of the hadrons is far off - shell . if one uses the @xmath22 decay width ( @xmath23 ) as an input parameter , then the @xmath19 coupling is fixed , but unfortunately , there are no guiding rules for the off - shell form factors which bring some ambiguity into the theoretical predictions . for @xmath18 exchange processes the situation is even worse . in this case we do not know the @xmath24 coupling constant ( the ambiguity of its estimate is rather large @xcite ) and the `` off - shellness '' in the @xmath0- nucleon-@xmath18 vertex is much greater because of the large mass difference between @xmath18 and @xmath17 mesons . the clas null result for a finite @xmath0 decay width means large off - shell suppression of the corresponding amplitudes and small @xmath24 coupling constant . therefore , the best way to check whether the @xmath0 exists or not is to study the @xmath1 fusion reaction with a quasi - free kaon and a nucleon in the initial state . in this case the @xmath19 coupling is fixed ( for given @xmath23 ) , and there is no ambiguity with the off - shell form factor because all hadrons are on the mass shell . this situation may be realized in the reaction @xmath25 ( @xmath6 ) with the @xmath0 showing up as a peak in the @xmath26 invariant mass distribution as shown in ref . there are several conditions which can enhance this effect . first , the @xmath4 invariant mass must be close to the mass of @xmath5 . in this case , the total amplitude is the coherent sum of two amplitudes with charged and neutral kaon exchange shown in fig . [ fig:1 ] . tree level diagrams for the reaction @xmath27 . the exchange of charged and neutral kaons is shown in ( a ) and ( b ) , respectively . ] the dominance of the @xmath18 meson exchange in @xmath5 photoproduction @xcite results in a constructive interference between the two amplitudes which enhances the @xmath0 signal . second , the deuteron wave function greatly suppresses the processes with a fast moving recoil nucleon , therefore , the experiment must be able to measure an extremely slowly moving recoil ( spectator ) nucleon which participates in the @xmath28 reaction . and third , the @xmath4 pair must be knocked out in the forward direction . in this case , the momentum of the recoil kaon is small , and it can merge with the slowly moving spectator nucleon to produce a @xmath0 . the clas experiment @xcite to search for @xmath0 was designed to study the direct @xmath29 reaction and , in principle , it does not satisfy the above conditions . thus , the @xmath0 and the outgoing neutron have finite momenta , and , therefore , the experiment has a neutron momentum cut of @xmath30 gev / c . in order to reduce the contribution of @xmath31 mesons coming from @xmath5 excitation the data analysis makes a cut on the @xmath5 mass , i.e. the @xmath4 invariant mass is outside the @xmath5 mass . it has cuts for the kaon momenta , @xmath32 ( gev / c ) , and cuts for the angles for positive and negative particles , @xmath33 and @xmath34 degrees , respectively . all these experimental conditions ( the @xmath4invariant mass , momenta and the angle cuts ) while being quite acceptable for studying the @xmath35 reaction result in a large suppression of the @xmath36 formation process in the @xmath9 reaction and reduce the ratio of @xmath0 resonance contribution ( signal ) to background ( noise ) - s / n . in order to avoid the obvious difficulty in measuring the slowly moving recoil nucleon one has to analyze the @xmath8 $ ] missing mass distribution @xcite . in this case , all momenta , allowed by the conservation laws participate in the process and , of course , the dominant contribution would come from slowly moving nucleons . as a result , the total cross section strongly increases . unfortunately , in this case the background processes increase roughly by a factor of two compared to the exclusive @xmath9 reaction , because both the @xmath26 and @xmath37 final states now contribute . nevertheless , even under this circumstance such experimental conditions can give a better chance to see the @xmath0 signal , in case it exists . the aim of the present paper is to extend the results of ref . @xcite for the inclusive reaction @xmath38 , where @xmath3 , towards finding favorable kinematical conditions for a manifestation of the @xmath0 signal . we are going to show that this signal is independent of the mechanism of the elementary @xmath39 reaction if the @xmath4 pair is produced in the forward hemisphere . our paper is organized as follows . in sec . ii we consider the kinematics of a @xmath40 reaction and define the observables . in sec . iii we briefly discuss the elementary @xmath41 reaction which will be used later on for estimating the resonant effect and background . sec . iv is devoted to a description of the associated @xmath42 photoproduction in @xmath43 interactions , where we discuss the most favorable kinematics for the coherent effect and the dependence of the cross section on @xmath0 spin and parity . in sec . v we discuss two dominant components of the non - resonant background : spectator and rescattering channels . in sec . vi we present our main results and give a comparison of a possible @xmath0 signal for the inclusive reaction @xmath38 with favorable kinematics and the exclusive reaction @xmath9 under the clas conditions . we show that in latter case the @xmath0 signal is weak due to the experimental conditions . the summary is given in sec . vii . in appendices a and b we show some details for the kinematics considered and the amplitudes of the elementary rescattering processes . the differential cross section of the reaction @xmath44 , where @xmath45 or @xmath37 , shown in fig . [ fig:2 ] , reads reaction @xmath46 . ] @xmath47}\equiv \frac{d\sigma}{dm_x dm_y d\omega\,d\omega_{x}\,d\omega_{y } } = \frac{1}{64\pi^2s_d}\frac{p_f}{p_i } \frac{1}{6}\sum\limits_{nk,\lambda , m_d , m_p , m_n } \|t_{fi}^{nk}\|^2\ , \frac{{\tilde{q}}\ , } { 16\pi^3}\ , \frac{{\tilde{\bar q } } } { 16\pi^3}. \label{e1}\ ] ] hereafter , we use the following notations : @xmath48 is the @xmath49 pair with mass @xmath50 , @xmath51 is the @xmath4 pair with mass @xmath52 , @xmath53 is the absolute value of the three - momentum of @xmath51 in the @xmath43 center of mass system ( c.m.s . ) , @xmath54 is the absolute value of the photon momentum in the c.m.s . , @xmath55 denotes the square of the total energy in this system , @xmath56 and @xmath57 stand for the absolute values of the @xmath17 and @xmath31 mesons momenta in the rest frames of the @xmath48 and @xmath51 systems , respectively . the indices @xmath58 correspond to the spin projections of the deuteron , outgoing proton and nucleon , respectively , @xmath59 is the photon helicity ; @xmath60 and @xmath61 are the solid angles of the directions of flight of @xmath17 and @xmath31 mesons in the rest frames of the @xmath48 , and @xmath51 systems , respectively ; @xmath62 is the solid angle of the @xmath51 system in the c.m.s . the quantization axis @xmath63 is chosen along the photon momentum , and the @xmath64 axis is perpendicular to the production plane of @xmath48 and @xmath51 pairs : @xmath65 , where @xmath66 is the three momentum of the @xmath51 system in the c.m.s . @xmath67 represents sum of the amplitudes of the resonant ( @xmath68 ) , semi - resonant ( @xmath69 ) , and non - resonant ( @xmath38 ) processes . the invariant mass distribution @xmath70 is defined as a 6-dimensional integral @xmath71 } \,d m_y\,d\cos\theta\,d\omega_{x } d\omega_{y}. \label{e2}\ ] ] in order to define the four momenta of all particles involved in the process appearing as arguments of the corresponding elementary amplitudes , we use the following incoming kinematical variables : photon 4-momentum ( laboratory system ) : @xmath72 ; deuteron 4-momentum ( laboratory system ) : @xmath73 ; invariant masses @xmath50 and @xmath52 ; the polar angle of @xmath4 pair photoproduction in the c.m.s . @xmath74 ; and the solid angles @xmath60 and @xmath61 . using these variables we now calculate all momenta in the @xmath75 c.m.s . ( for details see appendix a ) and then , transform them to the laboratory system . that is because the deuteron wave function is only well defined in the laboratory system . in our study we analyze the missing mass distribution in the range @xmath76 , where @xmath77 and @xmath78 in several selected regions of the invariant mass @xmath79 mev . the @xmath80 transition leads to a @xmath0 signal in the missing mass distribution . associated @xmath81 photoproduction manifests itself most clearly for @xmath82 and @xmath83 . the coherent signal must be suppressed outside of the resonance position . to analyze this situation we choose @xmath84 at the resonance position with @xmath82 and at a larger value ( @xmath85 gev ) . we also analyze the sensitivity of the @xmath0 signal to the @xmath4 pair photoproduction angle to get a maximum value for the s / n ratio . this gives the conditions for the range of integration over @xmath74 . integration over @xmath86 in eq . ( [ e2 ] ) is performed in all regions . the mechanism of @xmath88 photoproduction in @xmath89 interaction is quite complicated because many processes can contribute . in our consideration we select the channel with an intermediate excitation of @xmath5 , @xmath90 , and denote it hereafter as the `` resonant '' channel . as we will demonstrate , this process is dominant in the associated @xmath91 photoproduction at @xmath92 gev . we denote all other channels as `` non - resonant '' background . of course , this notation is rather conventional , because the @xmath88 pairs can also be produced from the virtual vector mesons , hyperon resonances other than @xmath5 , and so on . in this case the notation `` resonant '' selects just the @xmath5 resonance excitation . in this work we do not put emphasis on @xmath0 photoproduction in @xmath89 interactions because , at the considered kinematics when @xmath4 is produced in the forward direction with a fast moving proton , this channel is strongly suppressed by the deuteron wave function . in this section all variables are given in the @xmath89 c.m.s . in this part we follow closely our previous paper @xcite and recall the main aspects of our considerations for the sake of completeness . we assume , that at low photon energies , close to threshold , the amplitude of @xmath5 excitation in the @xmath41 reaction may be described by the effective lagrangian formalism , whereas at high energies , the regge model with the @xmath18 exchange as a leading trajectory can be used . the value @xmath94 gev is chosen as the matching point between these two regimes . the tree level diagrams for @xmath95 reaction at low energies are shown in fig . [ fig:3 ] . the diagrams ( a ) , ( b ) , ( c ) and ( d ) correspond to the @xmath96 , @xmath97 , @xmath98 exchange amplitudes and the contact term , respectively , and are denoted as the born terms . the diagram ( e ) describes the @xmath96-channel @xmath18 exchange amplitude . we neglect the photon interaction within the decay vertex and restore the gauge invariance by the proper choice of the contact term . tree level diagrams for the reaction @xmath99 . ] the amplitudes of the @xmath100 and @xmath101 reactions at low energy read [ res_ampll ] @xmath102 u_p(p)\,\varepsilon^\mu~,\\ & & { a^{\lambda^}}_{fi}(\gamma n)=\bar u_{\lambda^}^\sigma(p_\lambda^ * ) \left [ { { { \cal m}^s}}_{\sigma\mu } + { { { \cal m}^t}}_{\sigma\mu}(k^ * ) \right ] u_n(p)\ , \varepsilon^\mu~,\end{aligned}\ ] ] the total cross section of the reaction @xmath103 as a function of the photon energy . the experimental data are taken from ref . the dot - dashed curve is a fit to this data by @xmath104b)@xmath105^{2.1}$ ] . the long - dashed curve represents the cross section for a constant amplitude . the solid curve corresponds to a solution in the low and high energy regimes . the dashed curve describes the extrapolation of the effective lagrangian model to the high energy region . see @xcite for more details . ] where @xmath106 , @xmath107 are the @xmath5 and nucleon spinors , respectively , and @xmath108 is the photon polarization vector . at high energy ( @xmath109 gev ) they are replaced by the @xmath96-channel @xmath18 meson exchange amplitude with reggeized @xmath18 meson propagator . the explicit form of the transition operators @xmath110 as well as the choice of parameters are given in @xcite . the total cross section of the reaction @xmath103 as a function of the photon energy from ref . @xcite together with available experimental data @xcite is exhibited in fig . [ fig:4 ] . similar results are obtained in refs . @xcite using slightly different approaches . the differential cross section of the reactions @xmath103 ( a ) and @xmath111 ( b ) as a function of the kaon photoproduction angle in @xmath89 c.m.s . at @xmath112 , and 2.3 gev . , title="fig : " ] the differential cross section of the reactions @xmath103 ( a ) and @xmath111 ( b ) as a function of the kaon photoproduction angle in @xmath89 c.m.s . at @xmath112 , and 2.3 gev . , title="fig : " ] the differential cross section for the @xmath103 @xmath113 and 3.8 gev . experimental data from ref . @xcite . ] figure [ fig:5 ] shows the differential cross sections for @xmath100 and @xmath101 as a function of the kaon production angle in the @xmath89 c.m.s . at different @xmath114 in the near - threshold region . the difference in shape for these two reactions at forward photoproduction angles is explained by the sizeable contribution of the born amplitudes in the @xmath115 reaction . in the @xmath10 reaction the born term ( @xmath97-channel ) is small , and the main contribution comes from the @xmath18 exchange process . at backward photoproduction the shapes and the absolute values of the cross sections for @xmath115 and @xmath10 are similar to each other , but the total cross sections for @xmath115 is larger . at @xmath116 gev it varies from 0.59 to 1.14 @xmath117b as compared with 0.27 to 1.08 @xmath117b for @xmath10 . as we will see in the next section , the dominant contribution to the associated @xmath91 photoproduction comes from the backward angle of the @xmath17 photoproduction in @xmath95 reaction . in fig . [ fig:6 ] we show the differential cross section at @xmath1182.1 and 3.8 gev together with available experimental data @xcite . one can see that for increasing initial photon energy the cross section decreases at backward angles for the @xmath17 photoproduction . therefore , we expect that the threshold region with @xmath119 gev is most favorable for studying associated @xmath91 photoproduction which reflects the @xmath0 formation . note that a similar approach for the @xmath120 photoproduction based on the effective lagrangian formalism was recently developed by nam , hosaka , and kim @xcite . the difference to our approach consists in ( i ) a different choice of the form factors and ( ii ) a different @xmath121 coupling constant , which results in an enhancement of the born terms . in this case one gets a large constructive interference in the @xmath115 reaction and essentially a destructive interference in the @xmath10 reaction which leads to a strong suppression of the latter one . the difference in the different parameter sets may be resolved experimentally . finally , let us mention that in our approach the total sign of the @xmath5 photoproduction amplitude follows the sign of the @xmath18 exchange amplitude . thus , in the @xmath115 reaction the interference between @xmath18 exchange and born terms is constructive , i.e. their total sign coincide with the sign of the @xmath18 exchange amplitude . in the @xmath10 reaction the @xmath18 exchange is the dominant channel . but su(3 ) symmetry predicts opposite signs for the @xmath122 and @xmath123 couplings which results in opposite signs of the total amplitudes in @xmath115 and @xmath10 reactions . in ref . @xcite we assumed that the dominant contribution to the non - resonant @xmath125 reaction comes from the virtual vector meson production ( @xmath126 ) and intermediate @xmath127 excitation ( @xmath128 ) . we believe that the vector meson contribution is under control because the mechanism of real vector meson photoproduction is well known . the differential cross section of @xmath129 meson photoproduction based on the model of ref . data from refs . @xcite . ] as an example , in fig . [ fig:7 ] we show the differential cross sections for @xmath129 meson photoproduction at @xmath130 gev calculated using the model of ref . @xcite together with the available experimental data . one can see that the description of this reaction is quite reasonable . next , the coupling constant of @xmath131 can be extracted from the @xmath132 decay , and the @xmath133 and @xmath134 couplings can be found from su(3 ) symmetry relations . then , the contribution to @xmath135 photoproduction from the virtual vector meson excitation may be easily evaluated . but at this moment , we have to make two comments . first , in the @xmath124 reaction the virtual vector mesons are off mass shell and , therefore , one has to introduce the corresponding form factors @xcite . the form factors , together with the vector meson propagators , strongly suppress contributions of the virtual @xmath136 and @xmath137 mesons leaving only a noticeable contribution from the @xmath129 meson which is almost on - shell because of the small decay width of the @xmath129 meson . second , in the data analysis for @xmath0 photoproduction the contribution of the @xmath129 meson can be excluded by making a corresponding `` @xmath129-meson cut '' @xcite . nevertheless , we discuss it here in order to fix other sources of @xmath135 photoproduction , having in hand only the total cross section @xmath138 of the @xmath139 reaction @xcite . the situation with the contribution from the @xmath127 is not so transparent . at @xmath140 gev , there is some difference ( @xmath141 ) on the level of @xmath142% between the total cross section @xmath138 and the total contribution from @xmath5 and vector meson excitations . this difference increases at higher energies , because @xmath138 increases with energy , whereas the contribution from @xmath5 decreases with energy and the contribution from the @xmath129 mesons stays constant . at low energies , @xmath141 may be identified with the virtual excitation and decay of hyperons other than @xmath5 . thus for example , oh , nakayama , and lee considered contributions from @xmath143 , and @xmath144 @xcite , roberts included additionally contributions from @xmath145 and @xmath146 @xcite . in principle , one can also add contributions from @xmath147 and @xmath148 hyperons , listed in pdg @xcite . another source of @xmath88 pair photoproduction in @xmath89 reaction is the so - called drell process @xcite where the incoming photon virtually decays into a @xmath135 pair with subsequent quasi - elastic or charge - exchange @xmath149 re - scattering . also a @xmath88 pair may be produced from the virtual decay @xmath150 with a subsequent inelastic @xmath151 transition . the total cross section of the @xmath152 reaction as a function of the photon energy . thin solid curve is the contribution of @xmath5 excitation , the dashed curve depicts the contribution of the vector mesons decay , all other sources denoted as @xmath153 are shown by the short - dashed curve . the solid curve is the sum of all processes . experimental data from ref . @xcite . ] it is quite clear that a consistent description of all the listed background sources is well beyond the present state - of - the - art because one needs a fairly large number of poorly known strength parameters , form factors , phases etc . moreover , we need a proper description of the high energy behavior of these processes . on the other hand , in case of a large number of background sources we can assume random relative phases between them which leads to cancellations of the interference terms . also , as a first approximation one can choose the incoherent sum of the squares of the amplitudes to be a constant . this means that the energy dependence at low energies of this source of @xmath135 pairs is defined essentially by the phase space factor . our analysis of the @xmath5 photoproduction shows that this approximation works well ( see the solid and long dashed curves in fig . [ fig:4 ] at low energy ) . the value of the constant matrix element can be obtained from a comparison with experimental data for the @xmath154 reaction . in our further analysis we parameterize the amplitude of the additional contribution @xmath141 ( for the sake of a concise notation , we denote it as @xmath153 ) by the constant matrix element with @xmath155 gev@xmath156 . this parameterization , being quite reasonable at low energy with @xmath157 gev , results in a somewhat larger rise of the cross section and overestimates the data by @xmath158 at @xmath159 gev . to fit the data , we multiply @xmath160 by a correction factor @xmath161 @xmath162 with the matching point @xmath163 gev . in fig . [ fig:8 ] we show the total cross section of the @xmath152 reaction together with available experimental data @xcite . to summarize this section we conclude , that for the elementary @xmath152 process which will be used in our analysis of the @xmath38 reaction we have selected and described explicitly the @xmath5 and vector meson ( @xmath129 meson ) excitation channels . the sum of all other possible processes is parameterized effectively by a constant matrix element . the energy dependence of this channel follows the phase space . at higher energies , @xmath164 gev , this dependence is slightly corrected . in case of @xmath165 we use the corresponding expressions for @xmath5 photoproduction in the @xmath10 reaction , keeping the @xmath166 contribution the same as in the @xmath115 reaction . now we turn to the associated @xmath81 photoproduction off the deuteron . basically , our consideration of @xmath168 is similar to that in ref . @xcite , however , we make several modifications . therefore , for completeness , we recall the main aspects of our model to fix the new points . we assume that main contribution comes from the charge and neutral @xmath17 meson exchange , shown in fig . [ fig:1 ] ( a ) and ( b ) , respectively , and we do not discuss the diagrams with direct @xmath0 photoproduction being important at backward angles of @xmath4 pair photoproduction . in calculating the @xmath36 vertices we consider the @xmath0 decay width as an input parameter , taking @xmath169 mev @xcite . the amplitudes of the associated @xmath170 photoproduction are expressed through the transition operators of the `` elementary '' process @xmath171 as @xmath172 where the transition operator @xmath173 defines the amplitude of @xmath5 photoproduction and uses the sum of transition operators in eqs . ( [ res_ampll]a ) and ( [ res_ampll]b ) ; @xmath174 and @xmath175 stand for the deuteron @xmath176 coupling vertex and the deuteron spinor , respectively , @xmath177 , and @xmath178 is the momentum of the exchanged kaon . we begin our consideration for the case of @xmath0 spin - parity @xmath179 . generalization and discussion of our results for another @xmath0 spin - parity is relegated to the end of this section . following refs . @xcite we assume that the dominant contribution to the loop integrals comes from their pole terms . the consideration of the regular parts with off - shell kaons needs incorporation of the corresponding off - shell form factors which brings an additional ambiguity into the model . thus , our estimate may be considered as a lower bound of the coherence effect . the pole part may be evaluated by summing all possible cuttings of the loops , as shown in fig . [ fig:9 ] . diagrammatic representation of cutting ( indicated by crosses ) the loop diagrams . ] calculating the imaginary parts we use the following substitutions for the propagators of the on - shell particles ( cutkovsky rules @xcite ) , shown by crosses in fig . [ fig:9 ] , @xmath180 and the identity @xmath181 with @xmath182 . we also use the standard representation of the product of the deuteron vertex function and the attached nucleon propagator through the non - relativistic deuteron wave function @xmath183 where @xmath177 , and @xmath184 is the deuteron wave function with the spin projection @xmath185 and the nucleon spin projections @xmath186 and @xmath187 . by using eqs . ( [ cut1 ] ) - ( [ cut3 ] ) , one can express the principal parts of the invariant amplitudes in eq . ( [ gamma - d1 ] ) as @xmath188 where @xmath189 is the spatial component of the corresponding 4-vectors , defined as @xmath190 and @xmath191 . indices @xmath192 and @xmath193 refer to the left and right cutting diagrams in fig . [ fig:9 ] , respectively . the function @xmath194 is the cosine of the polar angle of the internal nucleon momentum @xmath195 in a deuteron when the @xmath196 axis is along the momentum @xmath189 . @xmath197 with @xmath198 and @xmath199 . the @xmath5 photoproduction and @xmath0 decay amplitudes read @xmath200 now we have an additional cut @xmath201 , compared to ref . @xcite which suppresses the integrals in eq . ( [ gamma - d2 ] ) and reduces the values of the corresponding cross sections . the effective deuteron vertex reads @xmath202 where @xmath203 is the cosine of the polar angle of @xmath195 , @xmath204 denotes the deuteron wave function in momentum space @xmath205 normalized as @xmath206 in our calculation we use the deuteron wave function derived from the `` realistic '' paris potential . we checked that the final result does not depend on the fine details of the deuteron wave function and practically does not depend on the choice of the potential . calculating the loop integrals in eq . ( [ gamma - d2 ] ) , one has to be careful with the proper determination of the 3-momentum of @xmath195 which is the argument of the corresponding elementary amplitudes in the integrals . the azimuthal angle of @xmath195 is chosen to be zero because all momenta are in the production plane . in order to get the internal momentum @xmath195 in the laboratory system with the @xmath196 axis along the beam direction , we make the following transformation @xmath207 where @xmath208 is the polar angle of momentum @xmath209 . the differential cross section of the associated @xmath4 and @xmath49 photoproduction , integrated over the @xmath4 invariant mass in the range @xmath210 mev at @xmath211 is related to the differential cross section of the associated @xmath91 photoproduction as @xmath212 where @xmath213 is the integral over the breit - wigner @xmath214 decay distribution @xmath215 with @xmath216 mev and the branching ratio @xmath217 . the differential cross section of the coherent @xmath170 photoproduction reads @xmath218 where @xmath219 and @xmath220 are the amplitudes of the charge and neutral current exchange , respectively , depicted in fig . [ fig:1 ] . in this equation , averaging and summing over the spin projections in the initial and the final states are performed . the difference between @xmath219 and @xmath220 consists of different elementary amplitudes for the @xmath221 and @xmath222 reactions , and in an opposite sign of the @xmath223 and @xmath224 couplings which is a consequence of the zero isospin of @xmath0 . the relative sign of the amplitudes of @xmath221 and @xmath222 follows the relative sign of the @xmath225 and @xmath226 coupling constants and , according to su(3 ) predictions , is opposite . therefore , the sum of the charged and neutral @xmath17 meson exchange diagrams leads to a constructive interference between @xmath219 and @xmath220 , and an enhancement of the cross section of the associated @xmath91 photoproduction . the average momenta @xmath227 and @xmath228 in the loop diagrams shown in figs . [ fig:9 ] ( @xmath192 ) and ( @xmath193 ) , respectively , as a function of the @xmath4 photoproduction angle . ] in fig . [ fig:10 ] we show the average momenta @xmath229 and @xmath228 in the loop diagrams depicted in figs . [ fig:9 ] ( @xmath192 ) and ( @xmath193 ) , respectively , as a function of the @xmath4 photoproduction angle in the c.m.s . this example corresponds to @xmath113 gev and @xmath230 gev . the definition of this averaging is given as usual , @xmath231 with @xmath232 where @xmath233 is the integrand in the loop integrals . one can see that the average momenta have a minimum at @xmath234 . near this position the corresponding amplitudes have a maximum . at large angles , mean values of @xmath235 are large and as a result , the corresponding amplitudes are very small because of the exponentially small value of the deuteron wave function at large @xmath235 . in fig . [ fig:11 ] we show the angular distribution of the differential cross section @xmath236 of the @xmath237 reaction at @xmath113 gev , and the missing mass @xmath238 $ ] , @xmath239 gev , and at @xmath240 and 1.62 gev in ( a ) and ( b ) , respectively . the angular distribution of the differential cross section @xmath236 at @xmath239 gev , @xmath113 gev and @xmath240 and 1.62 ( gev ) , shown in ( a ) and ( b ) , respectively . the solid and dashed curves correspond to resonant and coherent background contributions.,title="fig : " ] the angular distribution of the differential cross section @xmath236 at @xmath239 gev , @xmath113 gev and @xmath240 and 1.62 ( gev ) , shown in ( a ) and ( b ) , respectively . the solid and dashed curves correspond to resonant and coherent background contributions.,title="fig : " ] the solid curves correspond to the resonance contribution , i.e. @xmath241 , while the dashed curves shows the contribution from the non - resonant @xmath242 processes depicted schematically in fig . [ fig:12 ] . + diagrammatic representation of the associated non - resonant @xmath243 photoproduction . ] the same as in fig . [ fig:11 ] , but for different @xmath114 . ( a ) and ( b ) correspond to the dependence on the @xmath5 photoproduction angle in the c.m.s . and the laboratory frame , respectively.,title="fig : " ] the same as in fig . [ fig:11 ] , but for different @xmath114 . ( a ) and ( b ) correspond to the dependence on the @xmath5 photoproduction angle in the c.m.s . and the laboratory frame , respectively.,title="fig : " ] one can see , that in the @xmath5 region , where @xmath244 , the resonant cross section is about one order of magnitude larger than the contribution of the non - resonant channels discussed in the previous section . outside the resonance position , say for @xmath245 mev , the result is opposite , namely , the resonant contribution is strongly suppressed because of a small @xmath5 total decay width , and the processes with non - resonant @xmath246 transitions become dominant . in fig . [ fig:13 ] ( a ) and ( b ) we show the resonant cross section as a function of the @xmath5 photoproduction angle in the c.m.s . and the laboratory system , respectively , for several values of the photon energy . the value of the cross section at maximum and the position of the maximum depends on the energy . one can see that the @xmath0 formation in associated @xmath91 photoproduction is hardly measurable if the detector acceptance does not allow measuring the @xmath4 pairs at small angles @xmath247 . finally , let us discuss the dependence of the associated @xmath91 photoproduction on the spin and parity of the @xmath0 . the case of @xmath248 is especially attractive because the small @xmath0 decay width @xcite has a natural explanation for this assignment of the @xmath0 spin and parity @xcite . the effective lagrangians of the @xmath249 interactions are expressed usually in the following form @xcite @xmath250 where @xmath251 and @xmath17 are the @xmath0 , nucleon and kaon fields , @xmath252 , and @xmath253 . for the fixed @xmath22 decay width the coupling constant @xmath19 depends on the spin and parity of @xmath0 as @xmath254}~,\end{aligned}\ ] ] where @xmath255 is the @xmath0 decay momentum . these equations result in the following relation @xmath256 where , for example for @xmath169 mev , @xmath257 . using this estimate one can expect naively that in case of @xmath258 ( @xmath259 ) the coherent @xmath91 photoproduction would be enhanced ( suppressed ) roughly by two orders of magnitude compared to the case of @xmath260 . but the real situation is far from this expectation . the matrix elements defining the @xmath0 formation are proportional to the products @xmath261 with @xmath262 where @xmath178 is the kaon momentum , @xmath263 and @xmath264 denote the spin projections of @xmath0 and nucleon , respectively . as a result , the large ( small ) value of @xmath265 is compensated by the corresponding small ( large ) value of @xmath266 . for a qualitative estimate of such a compensation let us consider the combination @xmath267 where the nucleon may be off - shell and express @xmath268 through the @xmath0 decay width @xmath269 where @xmath270 is the square of the nucleon momentum in the @xmath271 vertex . from these equations one can conclude that in case of an on - shell nucleon with @xmath272 : ( i ) @xmath268 does not depend on parity , and ( ii ) @xmath273 . the latter is the consequence of the spin factor @xmath274 in the expression for the decay width . the dependence on parity arises only for off - shell nucleons . @xmath268 increases ( decreases ) for @xmath260 and @xmath275 ( @xmath276 ) at the off - shell region with @xmath277 . the increase for @xmath278 and @xmath279 is different , because in the former case @xmath268 is defined by the interplay of suppression and enhancement factors . on average , the ratio of @xmath280 would be slightly larger than 2 . the same as in fig . [ fig:13 ] ( a ) , but for different @xmath0 spin and parities . @xmath113 gev and @xmath230 gev . ] in some sense , the dependence of the amplitude of the associated @xmath91 photoproduction on spin and parity of the @xmath0 is rather similar to that of @xmath281 in our example . the diagrams with the on - shell hadrons in the @xmath0 formation vertex ( see fig . [ fig:9 ] ( @xmath192 ) ) do not depend on @xmath0 parity , and for @xmath282 their contribution is roughly two times greater than for @xmath283 . the amplitudes with the off - shell nucleon in the @xmath0 formation vertex ( fig . [ fig:9 ] ( @xmath193 ) ) are enhanced for @xmath284 and suppressed for @xmath285 . but this off - shell modification is not so large , because the contribution of the nucleons which are far off - shell are suppressed by the deuteron wave function . in fig . [ fig:14 ] we show the angular distribution of the differential cross section @xmath236 at @xmath239 gev , and @xmath113 gev , @xmath240 and different @xmath0 spin and parities , @xmath286 and @xmath287 . the ratio of the cross section at their maximum position for different @xmath288 reads @xmath289 this result is in agreement with our qualitative analysis , namely , the cross section for @xmath282 on average is 2.4 times greater than for @xmath283 . for @xmath260 and @xmath279 the cross sections are enhanced compared to the cases of @xmath290 and @xmath291 , respectively . now two questions arise . first , whether the associated @xmath81 photoproduction may be seen against other non - resonant processes in the resonance region with @xmath244 , and second , whether this signal is suppressed outside the resonance region and why ? to answer these questions we have to analyze the background processes . the spectator channels are depicted in fig . [ fig:15 ] . diagrammatic representation of background spectator channels . ( a , b ) : quasi - free @xmath292 photoproduction , ( c , d ) : quasi - free non - resonant @xmath293 photoproduction . ] contributions of these channels to the invariant mass distribution are defined by eqs . ( [ e1 ] ) and ( [ e2 ] ) , where the amplitudes are expressed via products of the amplitude of the elementary @xmath294 reactions @xmath295 and the deuteron wave function @xmath129 as @xmath296 where the deuteron wave function is defined in eq . ( [ gamma - d4 ] ) , and the index @xmath297 corresponds to the different elementary sub - processes discussed in sec . background contributions to the missing - mass distribution in the @xmath38 reaction at @xmath113 gev for quasi - free @xmath5 photoproduction , vector meson and hyperon excitations ( spectator mechanism ) , and rescattering channels shown by solid , dot - dashed , long - dashed , and dashed curves , respectively . ( a ) and ( b ) correspond to @xmath230 and 1.62 gev , respectively.,title="fig : " ] background contributions to the missing - mass distribution in the @xmath38 reaction at @xmath113 gev for quasi - free @xmath5 photoproduction , vector meson and hyperon excitations ( spectator mechanism ) , and rescattering channels shown by solid , dot - dashed , long - dashed , and dashed curves , respectively . ( a ) and ( b ) correspond to @xmath230 and 1.62 gev , respectively.,title="fig : " ] the background contributions for the quasi - free @xmath5 photoproduction , vector meson and hyperon excitation are shown in fig . [ fig:16 ] by solid , dot - dashed , and long - dashed curves , respectively . if the @xmath4 invariant mass is close to the @xmath5 mass ( cf . [ fig:16](a ) ) the quasi - free @xmath5 photoproduction gives the dominant contribution to the background . the next important contribution comes from the @xmath153 channel , parameterized by the constant matrix element . the contribution of the vector mesons in the region where the @xmath298 $ ] missing mass is around the @xmath0 mass is rather small . indeed , it does not contribute to the @xmath10 reaction and , moreover , it is suppressed dynamically . thus , the kinematics of the associated @xmath91 photoproduction in the forward direction requires a fast @xmath31 and slow @xmath14 . in this case , the @xmath135 invariant mass is far from the @xmath129 meson mass . but the situation changes at large values of the @xmath298 $ ] missing mass . in this case the available values of @xmath135 invariant mass cover the @xmath129 meson mass region , and the contribution of the @xmath129 meson excitation becomes essential . the case when the @xmath4 invariant mass is far from the @xmath5 mass is shown in fig . [ fig:16 ] ( b ) . now , the quasi - free @xmath5 photoproduction is suppressed , whereas the other channels have the same order of magnitude . let us consider proton and @xmath31 meson rescattering when a @xmath4 pair is produced in the @xmath89 interaction as shown in fig . [ fig:17 ] ( a , b ) and ( c , d ) , respectively . in principle , the @xmath299 and @xmath300 rescattering in quasi - free @xmath242 photoproduction from a deuteron must be taken into account too , but we skip them at the present stage , assuming first that such processes give a small correction to the spectator ( quasi - free photoproduction ) channels considered in the previous section and second , the cross section for elastic @xmath301 scattering is much smaller than for @xmath302 scattering @xcite . moreover , as we will show , the dominant contribution here comes from proton rescattering and , therefore , kaon rescattering is a small part . rescattering of the proton ( a , b ) and @xmath31 meson ( c , d ) in the @xmath303 photoproduction from the deuteron . ] basically , the amplitudes of the rescattering processes are evaluated similarly to the amplitudes of the associated coherent @xmath42 or @xmath304 photoproduction considered in sect . iii , where we assumed the dominance of the imaginary part of the corresponding loop diagrams , calculated by cutting rules . but there are several new aspects : + first , the rescattered particles are outside of the production plane and , therefore now , in the loop integrals we have to integrate over the virtual nucleon momentum @xmath235 and the azimuthal angle @xmath305 . if the polar and azimuthal angles of the momentum @xmath189 are @xmath208 and @xmath306 , respectively , then the three dimensional vector of the virtual nucleon in the laboratory frame , with the @xmath196-axis along the beam direction , reads @xmath307 where @xmath308 the polar angle @xmath309 is fixed by the on - shell conditions ( cf . ( [ cosine ] ) ) @xmath310 where the four momenta @xmath311 are defined as follows . + @xmath235 rescattering : @xmath312 @xmath31 rescattering : @xmath313 were @xmath314 and @xmath53 refer to the momenta of the outgoing nucleon and the proton of the @xmath48 and @xmath51 systems , respectively . we remind the reader that the indices @xmath315 refer to the cut loops shown in fig . [ fig:9 ] ( @xmath192 ) and ( @xmath193 ) , respectively . the @xmath316 and @xmath302 rescattering channels in @xmath38 reaction at @xmath113 gev for @xmath230 and 1.62 gev shown at ( a ) and ( b ) , respectively . , title="fig : " ] the @xmath316 and @xmath302 rescattering channels in @xmath38 reaction at @xmath113 gev for @xmath230 and 1.62 gev shown at ( a ) and ( b ) , respectively . , title="fig : " ] to preserve the energy - momentum conservation in the loop vertices one has to be careful with the determination of the azimuthal angles @xmath306 . the corresponding expression reads @xmath317 next , one has to choose the effective amplitudes in the loop integrals . we take them as a product of the deuteron wave function , photoproduction of the @xmath318 pair in @xmath319 reaction and elastic scattering amplitudes . for the proton and @xmath31meson rescattering they read correspondingly @xmath320 where @xmath321 and @xmath264 are the spin projections of the outgoing proton and nucleon , respectively . in our calculations we use an on - shell approximation for the elastic scattering amplitudes @xmath322 and @xmath323 , taken from the experimentally measured cross sections of elastic scattering . details of the employed parameterizations are given in appendix b. an important point is related to the @xmath324 vertex . in our model it consists of three components : @xmath5 excitation , the vector meson contribution , and the remaining background contribution denoted above as @xmath153 . consider first the @xmath5 channel . due to the rescattering kinematics , the invariant mass of the virtual @xmath4 pair in the loop integrals covers the @xmath5 resonance region even when the invariant mass of the outgoing @xmath235 and @xmath31 meson is outside the resonance position . this results in increasing the background contribution at @xmath325 . but the situation is not so simple . since the @xmath5 has a small total decay width , @xmath326 mev , its decay length is large , @xmath327 at a @xmath5 velocity @xmath328 . in other words , the @xmath5 decays mostly outside of the deuteron . similar or even larger suppression is expected for the case of the @xmath129-meson contribution . therefore , we can simply neglect these two channels . in fig . [ fig:18 ] we show the relative contributions of @xmath235 and @xmath31 rescattering for different channels . enhancement of the @xmath235-rescattering is explained by the difference between the cross sections for @xmath316 and @xmath302 elastic scattering . the angular dependence of the missing mass distribution for the associated @xmath91 ( solid curve ) photoproduction and the background spectator processes ( dashed curve ) at @xmath329 gev , @xmath230 gev , and @xmath113 gev . ] to summarize this section we conclude that at low energy , say @xmath330 gev , the dominant component of the non - resonant background comes from the quasi - free @xmath5 spectator channel , the next important contribution is composed of the @xmath153 spectator channel and @xmath316 rescattering , if the invariant mass of the @xmath4 pair is inside the @xmath5 resonance position with @xmath230 gev . outside of the @xmath5 resonance position the quasi - free @xmath5 spectator channel is strongly suppressed , whereas other channels remain on the same level finally , let us examine the angular dependence of the spectator channel , similar to the associated @xmath91 photoproduction ( cf . figs . [ fig:13 ] and [ fig:14 ] ) . the corresponding result is shown in fig . [ fig:19 ] where we present simultaneously contributions from the associated @xmath91 photoproduction ( signal ) and from the background ( noise ) dominated by the spectator channels . the calculation is for the resonance region with @xmath329 gev , @xmath230 gev , and @xmath113 gev . we choose the case of @xmath331 . one can see that the spectator channel has a sharp peak caused by the deuteron wave function with a maximum close to the maximum for coherent @xmath91 photoproduction . at small angles the noise decreases much faster than the signal . therefore , we can conclude that the largest value for the s / n ratio is expected at extremely forward @xmath4 photoproduction angles , say @xmath332 . below , we discuss the prediction for the @xmath0 formation processes under two different conditions . the first is @xmath0 photoproduction at low energy ( @xmath330 gev ) in the inclusive @xmath38 reaction . it can be studied , for example , by leps at spring-8 and/or crystal - barrel at elsa ( bonn ) . the second is the formation @xmath0 photoproduction process in exclusive @xmath9 reaction in a wider energy interval ( @xmath11 gev ) with the experimental conditions of the clas collaboration measurement @xcite . all calculations are made for a total @xmath0 decay width of @xmath169 mev . we calculate the missing mass distribution in the inclusive @xmath38 reaction with two cuts . the first one is the @xmath129-meson cut . we exclude all events with a @xmath135 invariant mass close to the mass of the @xmath129 meson : @xmath333 mev @xcite . the second one is the angular cut : we keep only forward @xmath4 pair photoproduction with @xmath334 . this cut gives a maximum s / n ratio . hereafter , we define the corresponding missing mass distribution as @xmath335 , where the superscript @xmath336 indicates , conditionally , an alignment in the forward photoproduction of the @xmath4 pair . the @xmath8 $ ] missing mass distribution in the @xmath337 reaction together with the partial contributions of different background channels at @xmath230 gev and @xmath330 gev . the contributions from the @xmath0 signal , spectator and rescattering processes are shown by the thin , long dashed , and dashed curves , respectively . ( a ) and ( b ) correspond to calculations without and with the angular cut @xmath338 , respectively.,title="fig : " ] the @xmath8 $ ] missing mass distribution in the @xmath337 reaction together with the partial contributions of different background channels at @xmath230 gev and @xmath330 gev . the contributions from the @xmath0 signal , spectator and rescattering processes are shown by the thin , long dashed , and dashed curves , respectively . ( a ) and ( b ) correspond to calculations without and with the angular cut @xmath338 , respectively.,title="fig : " ] figure [ fig:20 ] illustrates the effect of the angular cut in the missing mass distribution @xmath337 reaction at the @xmath5 resonance position with @xmath339 ( @xmath230 gev ) , averaging over the energy interval @xmath330 gev . here , we choose the case of @xmath331 . first of all , in the @xmath5 resonance region one can see a distinct effect of the associated @xmath91 photoproduction as a sharp @xmath0 peak against the flat non - resonant background with and without the angular cut . the angular cut increases the s / n ratio . [ fig:20 ] ( a ) shows the missing mass distribution calculated for @xmath340 ( c.m.s . ) , whereas in fig . [ fig:20 ] ( b ) the angular interval is limited to @xmath341 in accordance with the results in fig . [ fig:19 ] . the angular cut suppresses the cross sections of both the @xmath0 signal and the background ; it further modifies the shape of the background in such a way as to get the maximum s / n ratio at the resonance position @xmath342 . the shape modification is explained by the suppression of the quasi - free spectator channel at large invariant mass @xmath50 . when @xmath50 increases the maximum in the quasi - free background distribution , shown in fig . [ fig:19 ] ( dashed curve ) , moves towards large angles being outside of the integration region and , therefore , the contribution of this channel decreases . as a result , the angular cut creates an enhancement of the s / n ratio by about @xmath343 . the same as in fig . [ fig:20 ] ( b ) , but for @xmath85 gev . ] figure [ fig:21 ] illustrates the case when the @xmath4 invariant mass is far from the @xmath5 resonance position , @xmath85 gev . now we also see some @xmath0 peak generated mainly by the associated non - resonant processes . the resonance channel is suppressed because of the small @xmath5 decay width . the background near the @xmath0 peak is dominated by the @xmath153 and rescattering channels . since now the charged and neutral @xmath17 meson exchange diagrams contribute incoherently , the ratio @xmath344 would be smaller compared to the ratio in the resonance region , @xmath345 . neglecting the rescattering channels and the background shape modification , one can get the following qualitative estimate @xmath346 the background shape modification and the rescattering channels result in decreasing @xmath344 . missing mass distribution in the @xmath337 reaction at @xmath330 gev and @xmath230 gev for different @xmath0 spin and parity : ( a ) , ( b ) , ( c ) and ( d ) correspond to the @xmath0 spin and parity @xmath347 , and @xmath348 , respectively . , title="fig : " ] missing mass distribution in the @xmath337 reaction at @xmath330 gev and @xmath230 gev for different @xmath0 spin and parity : ( a ) , ( b ) , ( c ) and ( d ) correspond to the @xmath0 spin and parity @xmath347 , and @xmath348 , respectively . , title="fig : " ] missing mass distribution in the @xmath337 reaction at @xmath330 gev and @xmath230 gev for different @xmath0 spin and parity : ( a ) , ( b ) , ( c ) and ( d ) correspond to the @xmath0 spin and parity @xmath347 , and @xmath348 , respectively . , title="fig : " ] missing mass distribution in the @xmath337 reaction at @xmath330 gev and @xmath230 gev for different @xmath0 spin and parity : ( a ) , ( b ) , ( c ) and ( d ) correspond to the @xmath0 spin and parity @xmath347 , and @xmath348 , respectively . , title="fig : " ] in fig . [ fig:22 ] we show the missing mass distribution for different @xmath0 spin and parity : @xmath349 . the signal - to - noise ratio for different @xmath288 reads @xmath350 this result is in qualitative agreement with the previous analysis of the differential invariant mass distributions ( cf . ( [ res11 ] ) ) . note that the value of the invariant mass distribution at the resonance is independent of the @xmath0 decay width . but since all experiments have a finite experimental resolution , the measured signal is smeared by the experimental resolution , and this smeared signal must depend on the value of @xmath23 . the missing mass distribution for @xmath351 at @xmath330 gev and @xmath230 gev and for different @xmath0 spin and parity , folded with a gaussian resolution function . ] in fig . [ fig:23 ] we show the missing mass distribution folded with a gaussian distribution function @xmath352~,\end{aligned}\ ] ] with @xmath353 mev , which imitates a finite experimental resolution . in this case , the height of the resonance peak ( s ) decreases proportional to the factor of @xmath354 therefore , in order to see this peak above the background , one needs rather good experimental resolution even assuming that the @xmath0 spin parity is @xmath279 . finally , let us estimate the total @xmath0 formation cross section in inclusive @xmath38 reactions with angle and @xmath129-meson cuts . for @xmath248 , the @xmath0 photoproduction cross section at maximum is equal to @xmath355 and this value is independent of the @xmath0 decay width . then , the total cross section for the @xmath0 signal reads @xmath356 results for other assignments of @xmath357 can be evaluated using eq . ( [ res111 ] ) . the reaction @xmath360 has been analyzed recently by the clas collaboration @xcite . we note that this experiment was designed for studying the @xmath0 photoproduction in direct @xmath35 elementary processes . in order to enforce the @xmath0 signal the data analysis was performed with some specific cuts . for convenience , we denote those cuts together with acceptance of the clas detector as the clas experimental conditions . no sizable @xmath0 signal in the @xmath26 invariant mass distribution was observed . therefore , it seems to be interesting and important to estimate the cross section for @xmath0 formation for the condition of this experiment . if we find that the formation cross section is greater than the experimental accuracy of the clas experiment , then it indicates problems for the @xmath0 : either @xmath0 does not exist , or it does exist but the @xmath0 width is much smaller than 1 mev . the non - resonant background structure of the @xmath26 invariant mass distribution in @xmath351 at @xmath11 gev and for clas conditions . ] the acceptance of the clas detector allows to detect ( i ) proton and kaons with momenta greater than 0.35 and 0.25 ( gev / c ) , respectively ; ( ii ) the angles of the direction of flight of the positively and negatively charged particles are greater than 9@xmath361 and 15@xmath361 ( laboratory system ) , respectively . the data analysis was performed with + ( iii ) @xmath129-meson cut @xmath362 gev , + ( iv ) @xmath5 cut @xmath363 , + ( v ) neutron momentum cut @xmath364 gev / c . the @xmath5 cut almost kills the associated @xmath91 photoproduction shown in fig . [ fig:1 ] . `` almost '' means that at @xmath365 mev the @xmath5 signal is rather weak , but finite . nevertheless , the main contribution to the @xmath0 formation comes from the non - resonant channels shown in fig . [ fig:12 ] . the neutron momentum cut strongly reduces the spectator processes , shown in fig . [ fig:15 ] , making rescattering channels dominant . in fig . [ fig:24 ] we show the non - resonant background for @xmath26 invariant mass distribution accounting for the clas experimental conditions . one can see the dominance of the rescattering ( mainly @xmath366 rescattering ) channel . two other cuts are dangerous for the formation processes . as we have shown in sec . iv , the dominant contribution to the @xmath0 formation comes from the forward photoproduction of a @xmath4 pair with @xmath367 in laboratory system . in this case , the @xmath14 meson is a slowly moving particle . therefore , the acceptance restriction for momentum @xmath368 gev / c together with the angle limitation for the proton and @xmath31 meson reduce the cross section of @xmath0 formation . note that for the inclusive photoproduction discussed above the @xmath4 pairs are detected in the forward direction , i.e. , there are no such restrictions ( cuts ) , and any value of @xmath369 consistent with conservation laws is allowed . the @xmath26 invariant mass distribution in @xmath351 at @xmath11 gev and for conditions ( i)-(v ) . ] [ fig:25 ] shows the @xmath26 invariant mass distribution in @xmath370 at @xmath11 gev and for clas conditions ( i ) -(v ) . for convenience , we denote this distribution as @xmath371 , where the superscript `` @xmath372 '' indicates the clas conditions . one can see some @xmath0 signal against the non - resonant background dominated by the rescattering channels . we have chosen the more favorable case of @xmath373 . however , even in this case the s / n ratio is about three times smaller compared to the case shown in fig . [ fig:20 ] ( b ) . left panel : missing mass distribution in inclusive @xmath337 at @xmath330 gev and the @xmath4 photoproduction angular cut ( @xmath374 ( c.m.s . ) ) and @xmath129-meson cut . right panel : @xmath26 invariant mass distribution in exclusive @xmath370 at @xmath11 gev and for clas experimental conditions(i)-(v ) . experimental data from ref . @xcite . in both cases , @xmath375 and the @xmath0 signal is folded with a gaussian resolution function with a width of 3 mev.,title="fig : " ] left panel : missing mass distribution in inclusive @xmath337 at @xmath330 gev and the @xmath4 photoproduction angular cut ( @xmath374 ( c.m.s . ) ) and @xmath129-meson cut . right panel : @xmath26 invariant mass distribution in exclusive @xmath370 at @xmath11 gev and for clas experimental conditions(i)-(v ) . experimental data from ref . @xcite . in both cases , @xmath375 and the @xmath0 signal is folded with a gaussian resolution function with a width of 3 mev.,title="fig : " ] the @xmath0 photoproduction cross section at maximum is equal to @xmath376 using this value we can evaluate the total @xmath0 formation cross section for conditions ( i)-(v ) @xmath377 in the latter case , the additional factor 2 means that only the @xmath378 decay channel is under consideration . our estimate of the @xmath0 formation cross section for the clas experiment is three times smaller than the upper bound for the @xmath0 signal ( 0.3 nb ) reported in @xcite . for illustration , in fig . [ fig:26 ] we exhibit simultaneously the missing mass distribution in the inclusive @xmath38 reaction , averaged over the interval @xmath330 gev with the @xmath4 photoproduction angular cut ( @xmath374 ( c.m.s . ) ) ( left panel ) and the @xmath26 invariant mass in the exclusive @xmath379 reaction for the clas experimental conditions ( i)-(v ) ( right panel ) together with the available experimental data @xcite . remember , that the @xmath26 invariant mass distribution shown in @xcite is obtained after removing certain processes : contributions from @xmath129 meson and @xmath5 excitations , and the neutron spectator channels , because this experiment intends to search for direct @xmath0 production in the @xmath35 reaction . in our analysis of the @xmath0 formation process in the clas experiment we include all experimental conditions ( i)-(v ) . in order to be close to the conditions of the data analysis in ref . @xcite we include in the consideration the acceptance correction factor which restores the full 4-body phase space broken by the cuts ( i)-(v ) . our results with acceptance corrections are shown in fig . 26 .. ) to 0.15 nb . ] the model satisfactory describes the data at low @xmath380 and slightly underestimates them at higher invariant mass with @xmath381 gev . it is not surprising because our simple model does not pretend for detailed description of all aspects of the @xmath43 interaction in a wide interval of the photon energy . our main purpose is the analysis of @xmath0 signal - to - noise ratio at @xmath382 , where the model looks quite reasonable . for the left and right panels of fig . 26 the @xmath0 signal is folded with a gaussian resolution function with @xmath353 mev . notice , that utilizing the gaussian resolution function reduces the @xmath0 signal at the maximum position by a factor @xmath383 ( cf . ( [ res13 ] ) ) mev is rather illustrative . using eq . ( [ res13 ] ) one can easily re - estimate the amplitude of the @xmath0 signal for any value of @xmath384 . ] . from fig . [ fig:26 ] ( right ) one can see that the effective @xmath0 signal predicted for the clas conditions ( i)-(v ) is comparable to the statistical fluctuations . therefore , absence of a bright @xmath0 signal in the clas data @xcite does not exclude its possible manifestation under more favorable experimental conditions . in summary , we analyzed the possible manifestation of the @xmath0 formation process in inclusive @xmath38 reactions . if the @xmath0 exists , then in the @xmath8 $ ] missing mass distribution there must be a distinct @xmath0 peak . its strength depends on the @xmath0 spin and parity , and has a maximum value for @xmath331 . we found that at forward angles of the @xmath4 pair photoproduction the signal - to - noise ratio is most favorable . we also analyzed the recent results of the clas collaboration and found that the present experimental conditions are not favorable for studying the @xmath0 processes . the corresponding signal to noise ratio is small , and the @xmath0 signal is comparable to the statistical fluctuations due to the experimental acceptance . in our model a distinct @xmath0 signal is caused by the constructive interference of the @xmath5 photoproduction from the proton and neutron in the associated @xmath91 photoproduction off the deuteron . in calculations we use relatively old data for @xmath5 photoproduction off the proton at photon energies greater than the most favorable ones for the associated @xmath91 photoproduction making a corresponding extrapolation . for a more detailed study of this effect , new high - statistics low - energy data both for @xmath385 and @xmath386 are greatly desired , especially for large kaon photoproduction angles . a similar problem concerns fixing the non - resonant background . the elementary @xmath152 cross section is very important here . in our analysis we used old data with low accuracy . it is clear that for understanding the @xmath0 formation processes one needs more accurate low - energy data for this elementary subprocess , too . however , our main results have a general character . thus , it seems more reliable to detect the @xmath0 signal in the @xmath387 fusion reaction realized in associated @xmath91 photoproduction , which may be seen in inclusive @xmath388 reaction for certain experimental conditions . finally , we note that the @xmath0 formation reaction together with other accompanying processes considered in the present paper may be studied experimentally at the electron and photon facilities at leps of spring-8 , jlab , crystal - barrel of elsa , and graal of esrf . we appreciate many fruitful discussions with t. nakano , and we thank d. diakonov , h. ejiri , m. fujiwara , k. hicks , a. hosaka , t. mibe , r. mutou , m. naruki , t. sato , and k. yazaki for useful comments . we appreciate b. mecking for careful reading our manuscript and valuable suggestions . one of the authors ( a.i.t . ) thanks h. enyo for offering the hospitality at riken . this work was supported by bmbf grant 06dr121 , gsi - fe . let us consider the determination of the momenta of all outgoing particles in the @xmath390 reaction at fixed input parameters defined in sec . ii in detail . the square of the total energy in the c.m.s . is @xmath391 where @xmath392 is the deuteron mass and @xmath114 is the photon energy in the laboratory system . the momenta @xmath393 and @xmath53 read @xmath394 where @xmath395 is the triangle kinematical function . the four momenta of proton and @xmath31 meson in the @xmath51 rest frame is defined by the mass @xmath52 and the solid angle @xmath61 . thus , the absolute value of the decay 3-momentum reads @xmath396 then the four momenta are defined as @xmath397 and @xmath398 respectively . note , that here the @xmath399 axis is taken along @xmath66 , and the @xmath400 axis coincides with @xmath401 . next , we boost these momenta to the c.m.s . along the @xmath402 axis as @xmath403 where @xmath404 and @xmath405 . then , we rotate the coordinate system around the @xmath64 axis by the angle @xmath74 to get the momenta in the c.m.s . with @xmath406 along the photon momentum @xmath407 @xmath408 similarly , we transform the momenta of the outgoing nucleon @xmath409 and @xmath17 meson of the @xmath48 system with obvious substitutions : @xmath410 , @xmath411 , @xmath412 , and @xmath413 . let us consider the elastic @xmath415 scattering . the amplitude is related to the differential cross section via @xmath416 where @xmath417 and @xmath418 are the three dimensional relative momenta and the proton spin projections in the initial and the final states , respectively , @xmath97 is the square of the total energy in the c.m.s . , and @xmath74 denotes the scattering angle . in our calculations we take the differential cross section from experiment @xcite , given as a function of the scattering angle in certain energy intervals . in rescattering processes , one of the incoming proton may be off - shell , and therefore , generally , @xmath419 . thus , the scattering angle is defined as @xmath420 with @xmath96 , @xmath421 and @xmath422 given as @xmath423 where @xmath178 and @xmath424 are the kaon four momenta in initial and final states , respectively , and @xmath425 is the square of the four momentum of the incoming off - shell nucleon . we use the following parameterization of the differential cross section ( in mb ) @xmath426 where @xmath427 is the kaon momentum in laboratory frame in gev / c and @xmath428 the spin dependence of the amplitude is chosen in the simplest form as @xmath429 for the differential cross section of the elastic @xmath433 scattering we use the parameterization of ref . @xcite : @xmath434 where @xmath435 , @xmath235 and @xmath436 are the proton momentum and energy in the c.m.s . , respectively . the parameters @xmath437 , and @xmath372 are listed in table . i. 99 t. nakano _ [ leps collaboration ] , phys . . lett . * 91 * , 012002 ( 2003 ) . barmin _ et al . _ [ diana collaboration ] , phys . atom . nucl . * 66 * , 1715 ( 2003 ) ; + s. stepanyan _ [ clas collaboration ] , phys . lett . * 91 * , 252001 ( 2003 ) ; + v. kubarovsky _ et al . _ [ clas collaboration ] , phys . lett . * 92 * , 032001 ( 2004 ) ; ibid . * 92 * , 049902(e ) ( 2004 ) ; + j. barth _ et al . _ [ saphir collaboration ] , phys . b * 572 * , 127 ( 2003 ) ; + a.e . asratyan , a.g . dolgolenko , and m.a . kubantsev , phys . atom . nucl . * 67 * , 682 ( 2004 ) ; + a. airapetian _ et al . _ [ hermes collaboration ] , phys . b * 585 * , 213 ( 2004 ) ; + s. chekanov _ et al . _ [ zeus collaboration ] , phys . b * 591 * , 7 ; + m. abdel - bary _ et al . _ [ cosy - tof collaboration ] , phys . b * 595 * , 127 ( 2004 ) ; + a. aleev _ et al . _ [ svd collaboration ] , phys . nucl . * 68 * , 974 ( 2005 ) . k. hicks , prog . part . phys * 55 * , 647 , ( 2005 ) . n. muramatsu , in proceedings of xiv international workshop on deep inelastic scattering(dis06 ) april 20 - 24 ( 2006 ) , tsukuba , japan , edited by m. kuze _ et al_. ( world scientific , singapore ) ( to be published ) ; r.a . arndt , i.i . strakovsky , and r.l . workman , phys . c * 68 * , 042201(r ) ( 2003 ) , erratum-_ibid . _ * 69 * , 019901(e ) ( 2004 ) ; + j. haidenbauer and g. krein , phys . rev . c * 68 * , 052201(r ) ( 2003 ) ; a. sibirtsev , j. haidenbauer , s. krewald , and u - g . meissner , phys . b * 599 * , 230 ( 2004 ) ; a. sibirtsev , j. haidenbauer , s. krewald , and u - g . meissner , eur . phys . j. a * 23 * , 491 ( 2005 ) ; a. casher and s. nussinov , phys . b * 578 * , 124 ( 2004 ) r. l. jaffe and a. jain , phys . d * 71 * , 034012 ( 2005 ) . t. hyodo and a. hosaka , phys . d * 71 * , 054017 ( 2005 ) . t. nishikawa , y. kanada - enyo , o. morimatsu , and y. kondo , phys . rev . d * 71 * , 076004 ( 2005 ) [ arxiv : hep - ph/0411224 ] .	we analyze the possibility to produce an intermediate @xmath0 via a @xmath1 formation process in @xmath2 ( @xmath3 ) reactions at some specific kinematical conditions , in which a @xmath4 pair is knocked out in the forward direction and its invariant mass is close to the mass of @xmath5 ( @xmath6 ) . the @xmath7 signal may appear in the @xmath8 $ ] missing mass distribution . the ratio of the signal ( cross section at the @xmath7 peak position ) to the smooth background processes varies from 0.7 to 2.5 depending on the spin and parity of @xmath0 , and it decreases correspondingly if the @xmath4 invariant mass is outside of the @xmath5-resonance region . we analyze the recent clas search for the @xmath0 in the @xmath9 reaction and show that the conditions of this experiment greatly reduce the @xmath0 formation process making it difficult to extract a @xmath0 peak from the data .
since the discovery of tev @xmath3-rays from the bl lac objects , mrk 421 ( punch et al . 1992 ) and mrk 501 ( quinn et al . 1996 ) , detailed very high energy observations of these nearby blazars ( z = 0.031 , z = 0.034 ) have been made . measurements of flux variation with time , particularly simultaneous measurements at several wavelengths , constrain models of particle acceleration and @xmath3-ray production in the jets . spectral energy density measurements constrain both the models of the jets and of the infra - red ( ir ) photon density in the intervening intergalactic medium . the possibility of absorption of @xmath3-rays by ir radiation has been predicted for some time ( see , e.g. , nikishov 1962 ; gould & schrder 1967 ; stecker , de jager & salamon 1992 ; biller et al . 1998 ; vassiliev 1999 ) . the general picture which has emerged for the spectral energy density of emitted radiation from bl lacs has two components , a lower one with energies extended up to about 100 kev attributed to synchrotron radiation from electrons , and a higher one with energies sometimes extending to the tev range , usually attributed to inverse compton scattering ( see , e.g. , maraschi , ghisellini & celotti 1992 , marscher & travis 1996 ) . there are also competing models ( mannheim & biermann 1992 ; mannheim 1993 ; mannheim 1998 ) which assume that the higher energy component arises from protons , either by proton - induced synchrotron cascades ( pic models ) or by decays and/or interactions of secondary particles such as neutral pions and neutrons , or synchrotron radiation from proton beams ( mcke & protheroe 2000 ; aharonian 2000 ) . see catanese and weekes ( 1999 ) and mukherjee ( 2001 ) for reviews of tev observations and an overview of relevant models . mrk 421 and mrk 501 are particularly useful in separating the spectral characteristics intrinsic to the object from absorption effects in the intervening medium because they have almost the same redshift . they also exhibit strong flares in the tev energy regime , well above typical quiescent levels , making detailed spectral measurements possible for both ( gaidos et al . 1996 ; catanese et al . 1997 ; protheroe et al . 1997 ; aharonian et al . . measurements by various tev astronomy groups have shown that the energy spectrum of mrk 501 is not a simple power law ( samuelson et al . 1998 ; aharonian et al . 1999a ; djannati - ata et al . 1999 ; aharonian et al . 2001 ) but has significant curvature . the two - component nature of multiwavelength blazar spectra implies that , over a sufficiently wide energy range , tev spectra must be intrinsically curved . the measured curvature however depends on the distance of the energy range of the data from the ic peak . during the strong flaring activity the synchrotron peak of mrk 501 appears to shift to above 100 kev ( catanese et al . 1997 ; pian et al . 1998 ) , with the ic peak shifting to several hundred gev ( samuelson et al . 1998 ) . measurements of the hegra collaboration have the highest energies extending to @xmath4 20 tev ; their spectrum is fit better with an exponential cutoff at @xmath5 tev ( aharonian et al . 1999a ; aharonian et al . 2001 ) , rather than a simple parabolic correction to the power law as used in samuelson et al . ( 1998 ) . several groups have determined energy spectra for mrk 421 , both at low average flux levels ( @xmath6 crab ) ( aharonian et al . 1999b ; krawczynski et al . 2001 ; bazer - bachi et al . 2001 ) and from intense flares ( 2.8 - 7.4 crab ) ( zweerink et al . 1997 ; krennrich et al . analysis of the intense flare data showed that mrk 421 had a spectral index different from mrk 501 . the data could be acceptably fit with a simple power law , although there was weak evidence for curvature ( krennrich et al . the shape of the spectral energy distribution for mrk 421 ( aharonian et al . 1999b ; krennrich at al 1999b ) and mrk 501 generally appears independent of the flux level ( aharonian et al . 1999a ) , although some evidence for spectral variability has been reported by djannati - ata et al . ( 1999 ) and krawczynski et al . ( 2001 ) for mrk 501 . in this letter , we present results from @xmath3-ray observations of mrk 421 taken during intense flares in january - march 2001 with the whipple observatory 10 m telescope yielding a spectrum spanning the energy range between 260 gev and 17 tev . the spectrum has high statistical precision and shows a distinct cutoff with a characteristic energy of about 3 - 6 tev . the observations were made with the whipple observatory 10 m @xmath3-ray telescope equipped with the granite - iii high resolution camera ( finley et al . the fine granularity ( @xmath7 ) of the 379 photomultiplier camera provides good sensitivity for point sources . the sensitive energy range of the instrument is @xmath8 200 gev to greater than 20 tev . based on finer sampling of @xmath3-ray images , the linear response of the telescope at the highest energies is improved in comparison with previous camera configurations . the use of a different type of photomultiplier ( hamamatsu r-960 , previously hamamatsu r1398 ) , a complete re - coating of the mirrors and the installation of a new set of light concentrators necessitated a comprehensive and detailed new calibration of the telescope . three methods were used : the first was based on laboratory measurements of the individual instrument components , the second utilized the calibrated cherenkov light signal from single secondary cosmic - ray muons , and the third used simulated cosmic - ray showers to match observed distributions of the parameters of the background images . all three methods provide an energy calibration consistent within 20% for the absolute energy of primary @xmath3-rays . the calibration can be evaluated by checking the energy spectrum of the crab nebula , which is a standard candle for tev @xmath3-ray astronomy . the measurements of the spectrum of the crab nebula with the whipple telescope over several years data using different camera configurations , are all statistically consistent , showing the same flux constant and energy spectral index . therefore , we also show the crab spectrum in this paper for comparison with the mrk 421 spectrum . the crab data consist of 15.4 hours of on - source observations ( zenith angle @xmath9 ) with the same amount of data for the background estimate . mrk 421 was more active in 2000 and 2001 than in previous years of observations , with the most intense flaring episodes in 2001 . the unusually high flaring states in 2001 provide remarkable statistics , and we have chosen the strongest flares for the analysis of the spectral energy distribution . observations of flaring states on january 21 , january 31 , february 1 - 3 , february 27 , march 19 , march 21 , march 26 - 27 , 2001 provide together @xmath10 photons above 260 gev . to test the validity of combining various flares we have also divided the data into two subsets and found no statistical inconsistency in the derived energy spectral indices . the statistical significance of the excess events corresponds to more than 86 standard deviations . the selected 2001 flaring data consist of 30.8 hours of on - source observations at zenith angles less than @xmath11 . for background comparison , the same amount of off - source data is used from observations made at similar zenith angles . the on - source data provide an average rate of 12.5 @xmath12/min , corresponding to 3.7 crab . the excellent signal to background ratio can be seen from figure 1 . this plot shows the alpha distribution ( for explanation of alpha parameter see caption of figure 1 or reynolds et al . 1993 ) of events with energies greater than 1 tev from the source ( solid line ) after applying @xmath3-ray selection criteria . the dotted line shows the corresponding distribution for the background measurements . the data analysis , @xmath3-ray selection and energy estimate use the methods developed by mohanty et al . these @xmath3-ray selection criteria are derived from parameter distributions of simulated @xmath3-ray showers as a function of their total light intensity ( size ) in the camera . the criteria vary with size and are set so that they keep 90% of the @xmath3-ray images whose centroids lie within @xmath13 of the center of the camera . to avoid the difficulties of modeling the trigger electronics we apply an additional cut , requiring that a signal of at least 15.1 photo electrons ( p.e . ) , 13.6 p.e . and 12.1 p.e . be present in the three highest image pixels , respectively . the differential energy flux values derived from the intense flaring states of mrk 421 in 2001 , and for comparison the crab nebula , are shown in figure 2 . the spectrum of the crab nebula can be well fit by a power law of the form : @xmath14 giving a @xmath15 of 6.9 for 10 degrees of freedom ( probability 73% ) . this result is consistent , within its limited statistics , with previous measurements made by the whipple collaboration ( hillas et al . 1998 ; mohanty et al . 1998 ; krennrich et al . 1999a ) , showing that the analysis methods , the calibration of the detector , and the reconstruction of energy spectra are all consistent at the 2 sigma level including systematic uncertainties . the first set of errors on the measured spectral index are statistical and the second are systematic . the systematic errors are determined by varying the gain of the system by 20% , and by varying the cut efficiencies ( see mohanty et al . 1998 ) . fitting a power law to the energy distribution of mrk 421 yields the following result : @xmath16 giving a @xmath15 of 410.7 for 10 degrees of freedom . the spectrum is clearly not compatible with a simple power law form . a curvature fit for the mrk 421 spectrum yields : @xmath17 @xmath18 . giving a @xmath19 for 9 degrees of freedom with a chance probability of @xmath20 . as shown in figure 2 , the mrk 421 spectrum exhibits clear curvature , but a parabola is not a suitable shape . assuming that part of the curvature in the spectrum is due to a cutoff , the data can be fit by : @xmath17 @xmath21 with @xmath22 . while the @xmath15 ( 25.2 for nine degrees of freedom or a probability of 0.3% ) is not particularly good , the fit using an exponential cutoff is much better than the fits using the two simple forms considered above . we should point out here that for the first time statistical errors of the flux measurements in tev astronomy reached the level of 2% , likely making systematic uncertainties dominant . because of this , indicated confidence levels should be accepted with a caution as they are based only on statistics of detected photons . we have also examined the flux values at 5.6 and 8.2 tev , which appear high to the eye , @xmath23 and @xmath24 , respectively above the exponential cutoff fit . considering the statistical uncertainties we do not regard this as a significant feature in the data . in future , a more complete analysis will be pursued to understand systematic errors at the percent level . in derivation of the cutoff energy we have included two most prominent sources of systematics : uncertainties in the reconstruction of the spectrum with a cutoff derived from simulation tests ( test spectra ) , and uncertainty in absolute energy calibration of the instrument ( 20% ) . we also tried a `` super - exponential '' form ( e.g. , see stecker , de jager & salamon 1992 ; aharonian et al . 2001 ) , but with no improvement in quality of fit . the result is : @xmath17 @xmath25 with @xmath26 . the @xmath15 of the fit is 24.0 for eight degrees of freedom ( probability of 0.2% ) . for the first time , the mrk 421 energy spectrum at very high energy has been determined over a sufficiently wide energy range , and with sufficient statistical precision , that an important feature ( a cutoff ) could be discerned . the spectrum is best described by a power law attenuated by an exponential cutoff at an energy of @xmath27 tev . the cutoff feature could have several origins , e.g. , the termination of the particle energy distribution in the primary beam , absorption near the @xmath3-ray source ( dermer & schlickeiser 1994 ) or absorption in intergalactic space by ir background radiation fields . if the cutoff energy varied in time , or differed for mrk 421 and 501 , it would not be due to extragalactic absorption . we examined our previous published flare spectra of mrk 421 ( krennrich et al . 1999a ; zweerink et al . 1997 ) and found them consistent with the present spectrum . fitting the previous mrk 421 spectrum ( krennrich et al . 1999a ) with the parametrization of @xmath28 @xmath29 results in an acceptable @xmath15 of 19.4 for 13 degrees of freedom ( p = 11% ) . we also examined our previously published spectrum for mrk 501 ( samuelson et al . , 1998 ) and find that its spectrum can be fit by @xmath30 with @xmath31 . however , fitting the mrk 501 spectrum with @xmath28 @xmath32 results in a poor fit with a @xmath15 of 37.6 for 13 degrees of freedom ( p = @xmath33 ) . the cutoff energy for mrk 501 is consistent with the cutoff energy for mrk 421 , but a different spectral index is required for an acceptable fit . thus , we find no inconsistency with the assumption that the cutoff might be due to the ir background . more detailed studies of spectral variability of mrk 421 are in progress and will be presented elsewhere . energy spectra for mrk 421 reported by the hegra collaboration ( aharonian et al . 1999b ) are based on data taken in 1997 and 1998 with the source at a lower average flux level ( @xmath34 crab ) . however , the primary sensitivity of the hegra array is at energies above 1 tev requiring a comparison above that energy . those results indicate a power law spectrum of @xmath35 . when the mrk 421 spectral data presented here are fit with a power law using only data points above 1 tev , a spectral index of @xmath36 with a @xmath15 of 106.8 for 6 degrees of freedom is measured ( @xmath37 if the 1.2 tev - 8.2 tev range is fitted , @xmath15 of 11.1 for 4 degrees of freedom ) which is consistent with the hegra result , considering statistical and systematic uncertainties between two different instruments . in summary , we have presented a mrk 421 spectrum over the energy range 260 gev to 17 tev which shows an exponential - like cutoff in the range 3 - 6 tev , similar to the cutoff found earlier for mrk 501 ( see figure 3 ) which has almost the same redshift . the new mrk 421 spectrum is consistent with those reported earlier both by our group and others . the occurrence of a cutoff at similar energy in the very high energy spectra of the two equally distant @xmath3-ray blazars mrk 421 and mrk 501 might indicate that tev @xmath3-rays are absorbed by ir background radiation fields . future investigations regarding the implications for the energy density of extragalactic ir background radiation are underway ( vassiliev 2001 ) . protheroe , r. j. , bhat , c. l. , fleury , p. , lorenz , e. , teshima , m. , weekes , t. c. 1998 , proc . of 25th int . cosmic ray conf . ( durban , south africa ) , eds . m.s . potgieter , b.c . raubenheimer , d.j . van der walt , 8 , 317	exceptionally strong and long lasting flaring activity of the blazar markarian 421 ( mrk 421 ) occurred between january and march 2001 . based on the excellent signal - to - noise ratio of the data we derive the energy spectrum between 260 gev - 17 tev with unprecedented statistical precision . the spectrum is not well described by a simple power law even with a curvature term . instead the data can be described by a power law with exponential cutoff : @xmath0 @xmath1 with @xmath2 . mrk 421 is the second @xmath3-ray blazar that unambiguously exhibits an absorption - like feature in its spectral energy distribution at 3 - 6 tev suggesting that this may be a universal phenomenon , possibly due to the extragalactic infra - red background radiation .
since its very origins , the exact renormalization group ( erg ) @xcite has proved to be a powerful tool for studies of non - perturbative effects in quantum field theory ( see recent reviews in @xcite ) . a particularly interesting case is that of an effective scalar field theory in two dimensions . as it was first conjectured by zamolodchikov @xcite , for a @xmath0 symmetric theory there should exist an infinite set of non - perturbative fixed points corresponding to the unitary minimal series of @xmath3 conformal field theories , where @xmath4 @xcite . morris @xcite showed numerically that such points do exist . the calculation was performed with a reparametrization invariant version of the legendre erg equation @xcite expanded in powers of derivatives . it was also pointed out there that to the level of the local potential approximation only the continuum limits described by periodic solutions and corresponding to critical sine - gordon models could be obtained . to find the expected set of fixed points the calculations had to be taken to the next order in the derivative expansion . this constituted a manifestation of the non - perturbative nature of the phenomena , and remarkably the legendre erg equation was powerful enough to locate and describe with good accuracy the expected set of 2d field theories . in this work we study the same @xmath0 symmetric scalar field theories in two dimensions but now with the polchinski erg equation @xcite . we present preliminary results which complement the results obtained with the legendre erg equation . in sect . 2 we follow the article by ball et al . @xcite to present the basic equations of the formalism this will allow us to set up notation for sect . 3 where we analyse the equations to second order in the derivative expansion . in sect . 4 we discuss the results and present our conclusions . the polchinski equation @xcite for a scalar theory can be written as follows @xcite & & = _ k(^2)+ds+ + & & _ - _ _ p [ cv1 ] . here @xmath5 is a general wilsonian action which can be written in terms of dimensionless variables as follows [ ; t ] & = & _ p _ p p^2 ( k(^2 ) ) ^-1 _ -p + _ int[;t ] , [ action1 ] + _ int[;t ] & = & dy . [ action2 ] in eq . ( [ cv1 ] ) the partial derivative on @xmath5 means it only acts on the explicit @xmath6 dependence of the couplings and the prime in the momentum derivative means it does not act on the delta function of the energy - momentum conservation , and @xmath7 . @xmath8 is a ( smooth ) regulating function which damps the high energy modes satisfying the normalization condition @xmath9 . the renormalized field @xmath10 changes with scale according to @xmath11 \hat{\varphi}_{p},\ ] ] where @xmath12 is the anomalous scaling dimension . to the second order in the derivative expansion we consider the two terms which are written explicitly in eq . ( [ action2 ] ) . within this approximation the polchinski erg equation reduces to the following system @xcite & = & f+2az-2ff+^+f+^-xf , [ eq1 ] + & = & z+bf^2 - 4zf-2zf+^-xz- z-/2 , [ eq2 ] where @xmath13 , @xmath14 , @xmath15 and the potentials @xmath16 and @xmath17 are defined in eq . ( [ action2 ] ) . the dots and primes denote the partial derivatives with respect to @xmath18 and @xmath19 respectively . the parameters @xmath20 and @xmath21 reflect the scheme dependence of the equations and are equal to @xmath22 , @xmath23 . here @xmath24 and @xmath25 , @xmath26 parametrize the regulating function in eq . ( [ action1 ] ) and are defined by & = & ( -1)^n+1k^(n+1)(0 ) , + i_n & = & -_(^2)^nk ( ^2)=-_d_0^dz z^d/2 - 1-nk(z ) , where @xmath27 stands for the @xmath28-th derivative of @xmath29 and @xmath30 . in the next section we search for fixed - point solutions , i.e. for functions @xmath31 and @xmath32 which are independent of @xmath18 and satisfy the system & & f+2az-2ff+^+f+^-xf=0 , [ eqfp1 ] + & & z+bf^2 - 4zf-2zf+^-xz- z-/2=0 . [ eqfp2 ] we will choose the initial conditions ( according to the terminology adopted in the literature on the erg equations ) set by the @xmath0 symmetry : @xmath33 and @xmath34 and by the normalization condition : @xmath35 . for the value of the first derivative of @xmath31 at the origin we will take the condition @xmath36 , where @xmath37 is a free parameter . the anomalous dimension @xmath12 at a fixed point becomes the critical exponent @xmath38 . to solve eqs . ( [ eqfp1 ] ) , ( [ eqfp2 ] ) for @xmath39 we consider the recursive numerical method already tested for @xmath40 @xcite . the physical fixed point solutions @xmath41 , @xmath42 at the fixed point value @xmath43 are regular for @xmath44 and have a certain asymptotic behavior as @xmath45 . thus the natural method for finding the correct numerical solution is to select those which can be integrated as far as possible in @xmath19 . a generic solution will end at a sharp singularity for a finite value of @xmath19 . the difficulty lies in the nonlinear and stiff nature of the equations and the need to fine tune @xmath1 and @xmath37 . this makes the direct integration of the system too hard . one way out is to solve it recursively . unlike the case @xmath40 studied in a number of articles @xcite , one faces an additional difficulty in two dimensions . it is not possible to start the iterative procedure by setting in eq . ( [ eqfp1 ] ) @xmath46 and @xmath47 as it is prescribed by the consistency of the leading approximation . for @xmath39 the polchinski equation in the leading order has only periodic or singular solutions for all values of @xmath37 . to overcome this difficulty one has to consider @xmath48 as the initial value to start the iterations consequently , an analysis of the leading order polchinski equation @xmath49 with the initial conditions @xmath33 , @xmath36 for non - zero @xmath1 is required . we studied eq . ( [ eqfp01 ] ) for @xmath39 numerically for a wide range of values of @xmath1 and @xmath37 . our results show that for each @xmath50 we can fine tune @xmath37 in such a way as to obtain a non - trivial regular fixed point solution . the set of such values @xmath51 form a discrete series of continuous lines @xmath52 ( see fig . 1 ) . in fact simple arguments can be presented which explain the appearance of the lines in the parameter space corresponding to regular fixed - point solutions @xmath41 . let @xmath53 denote position of the pole of a generic solution of eq . ( [ eqfp01 ] ) . suppose that for some values @xmath54 the solution is regular , i.e. @xmath55 . let us take another value @xmath56 sufficiently close to @xmath57 . assuming that the function @xmath53 is continuous , it is clear that there should exist the value @xmath58 such that again @xmath59 . hence there is a line of the `` constant value '' @xmath60 in the parameter space . when we move along a fixed line the solutions @xmath41 do not change their shape significantly . moreover , their shape follows a regular pattern when passing from one curve to the other similar to solutions obtained by morris @xcite . this can be considered as a sign for the existence of the infinite discrete set of fixed points corresponding to the minimal unitary series of conformal models . we also would like to note that for @xmath61 and @xmath62 there are no other non - trivial fixed - point solutions besides the ones corresponding to the lines discussed here . = 1.0 we would like to note that a similar picture takes place in other dimensions . for @xmath40 we found the same discrete set of lines in the @xmath63-plane corresponding to regular solutions of eq . ( [ eqfp01 ] ) , but in this case they are situated in the interval @xmath64 ( see fig . 1 ) . as one can see there is a line ( upper line in fig . 1 ) which crosses the @xmath37-axis at @xmath65 . this is the value of the parameter @xmath37 for which a non - trivial fixed point solution of the polchinski equation was found in the leading order ( local potential approximation ) @xcite . the important observation is that there is only one line in the @xmath63-plane with positive values of @xmath1 . since according to general arguments at physical fixed points @xmath66 , this suggests that for @xmath40 there is only one non - trivial fixed point . this is the wilson - fischer fixed point found in numerous previous studies @xcite . recall that for @xmath39 all the lines are situated in the @xmath67 half - plane , hence one can expect an infinite number of non - trivial fixed points . one more remark is relevant here . by a simple scaling analysis of eq . ( [ eqfp01 ] ) it can be shown that there is a certain mapping between the lines @xmath52 in different dimensions . when we pass from one dimension to another the line experiences a vertical shift and scaling transformation . more details about this mapping will be presented elsewhere . we now pass to the study of the system ( [ eqfp1 ] ) , ( [ eqfp2 ] ) . we have seen that there are families of solutions of eq . ( [ eqfp01 ] ) corresponding to a given fixed point . it turns out that when the second equation of the system is taken into account , this degeneracy disappears . we solved the system using the following iteration procedure developed by ball et al . first we set @xmath68 , choose some initial value @xmath69 and fine tune @xmath37 to the value @xmath70 corresponding to the regular solution @xmath71 of the first equation ( [ eqfp1 ] ) ( or ( [ eqfp01 ] ) ) . of course , the point @xmath72 lies on one of the lines described above ( see fig . 1 ) ) . as the next step we insert the function @xmath71 into the second equation ( [ eqfp2 ] ) for a fixed @xmath21 and fine tune @xmath1 to the value @xmath73 for which a regular solution @xmath74 exists . then we substitute @xmath74 and @xmath75 into the first equation of the system and find a regular solution for a fixed value of @xmath20 thus obtaining a new value @xmath76 and a new function @xmath77 . we repeat this process keeping @xmath20 and @xmath21 fixed . as a result a sequence of functions @xmath78 , @xmath79 , @xmath80 , @xmath81 , and a sequence of numbers @xmath82 , @xmath83 , @xmath84 , @xmath85 , are obtained , and we test them for convergence . for @xmath40 and for the relevant line , associated with the wilson - fischer fixed point , we confirmed the results by ball et al . @xcite . the new feature in our calculations is that we took @xmath86 as the initial value of the iterating procedure , whereas in @xcite only @xmath87 was considered . we conclude that the numerical method converges and that the rate of convergence is controlled by @xmath20 for fixed @xmath88 . the best @xmath20 was shown to correspond to the inflexion point where @xmath89 changes sign . the important observation is that the final values @xmath38 and @xmath90 to which the iterations converge ( i.e. the fixed - point values ) do not depend on the initial value @xmath91 . when @xmath88 is closer to the fixed - point value the rate of convergence is of course faster . the final value @xmath38 depends on @xmath21 linearly . for @xmath92 the two equations decouple and there is no need for iterations to find a solution of the system . we just have to adjust @xmath21 such that @xmath93 . for the wilson - fischer fixed point @xmath94 we found @xmath95 and @xmath96 . = 1.0 for @xmath39 the situation is totally different . for any line @xmath97 we start with and the initial value @xmath98 the iterative procedure turns out to be divergent if @xmath99 . only for @xmath92 we have been able to find a solution to the now decoupled system ( [ eqfp1 ] ) , ( [ eqfp2 ] ) by adjusting @xmath21 . similar to the case @xmath40 we have not found any natural criteria to select the value of @xmath21 since for each line @xmath21 depends monotonically on @xmath1 decreasing as @xmath100 . to determine the fixed - point solutions , corresponding to the minimal unitary series of conformal field theories , we have fixed the value of @xmath21 by a fit to the series of exact values for the anomalous scaling dimension @xmath38 . in this way we found @xmath101 . the fixed - point solutions for @xmath41 display regular behaviour and are reminiscent of those obtained by morris @xcite . in particular they have @xmath102 extrema , @xmath103 ( see fig . the fixed point solutions for @xmath42 have the same pattern of extrema , though their profiles are different from those of morris . next for @xmath104 and and corresponding values @xmath43 we have calculated the critical exponent @xmath2 . for this we considered perturbations of the functions @xmath31 and @xmath32 around the fixed - point solutions , @xmath105 and substituted them into eqs . ( [ eq1 ] ) , ( [ eq2 ] ) . after linearization we obtained the system -2g_nf_-2f_g_n+_^+g_n+ _ ^-xg_n&= & _ ng_n , + h_n-2f_h_n-2g_nz_-4f_h_n-4g_nz _ & & + + _ * ^-xh_n-_h_n+ 2bf_g_n&=&_nh_n , where @xmath106 and @xmath107 are calculated for @xmath43 . for @xmath0 perturbations the initial conditions are @xmath108 , @xmath109 . we also imposed the normalization condition @xmath110 . away from the fixed point ( but sufficiently close to it ) we relaxed @xmath111 to be different from zero , @xmath112 . then @xmath113 and @xmath114 were fine tuned so that polynomially growing eigenfunctions were obtained . for @xmath39 , @xmath92 and @xmath104 we have calculated the critical exponent @xmath115 . the results for @xmath38 and @xmath2 are given in table 1 . we conclude that the polchinski erg equation gives the values for the critical exponent which match quite well the exact conformal field theory values @xcite and the results by morris @xcite , also included in the table . we note that the best fit value for @xmath21 that we have found actually corresponds to a well defined subset of the regulating functions . an explicit example is @xmath116 , where @xmath117 to ensure @xmath118 and @xmath119 , and @xmath120 to give @xmath104 . values obtained by morris @xcite ( @xmath57 , @xmath121 ) and the exact results of conformal field theory ( @xmath122 , @xmath123 ) . in this work we have studied the solutions of the polchinski erg equation for an effective @xmath0-symmetric scalar field theory in the two - dimensional space @xmath124 . we have seen that this equation provides a reliable non - perturbative evidence for the existence of the fixed - point solutions corresponding to the minimal unitary series of conformal field theories and allows to calculate the anomalous dimension and the critical exponents with good accuracy . this constitutes another positive test of the power of the erg approach . at the same time our studies are complementary to similar calculations within the erg approach based on the equation for the legendre action @xcite . as mentioned above , in the leading order of the derivative expansion ( local potential approximation ) the consistent value for @xmath1 is 0 and only periodic sine - gordon type fixed - point solutions can be obtained . however , we have found that there are continuous families of fixed - point solutions corresponding to a series of lines @xmath52 in the @xmath63 plane which have a part with positive values @xmath125 . it was also argued that these lines correspond to the multicritical fixed points of the theory . it is by taking into account the second order in the derivative expansion that we find isolated fixed - point solutions . we have found the first 10 points out of the infinite series and calculated the critical exponent @xmath2 for them . the results depend on the choice of the regulating function . the value of @xmath21 , for which a regular solution exists , depends linearly on @xmath1 , so the criterium of minimal sensitivity can not be applied to fix @xmath21 . the best fit to the conformal field theory values of @xmath1 gives us @xmath92 , @xmath101 . no other regulating functions have been found to work . for @xmath39 whenever @xmath126 , the iterative procedure is not seen to converge . in fact , for @xmath92 there is no need for iterations since the two equations of the polchinski approach decouple . this is in sharp contrast to the case of @xmath40 where convergence was checked for @xmath126 @xcite . for the values @xmath92 , @xmath104 , giving the best fit , our results are comparable in accuracy and sometimes better ( the accuracy also increases with multicriticality ) than those of morris @xcite . we would like to note that fixing @xmath21 by the best fit to exact results for the anomalous dimensions @xmath1 is reminiscent of fixing the renormalization scheme dependence in the perturbative renormalization group . it is also similar to fixing the regulator by the condition of the reparametrization invariance for the erg equation for the legendre action ( that corresponds to the limiting case of @xmath92 , @xmath127 ) @xcite . another important point we would like to mention is that in our analysis we have not found non - trivial fixed points other than those corresponding to the minimal models . this is what one expects from zamolodchikov s @xmath128-theorem @xcite . the conclusion is already clear from the analysis of the leading order polchinski equation for @xmath48 when the lines in the @xmath129-plane corresponding to non - trivial fixed points are plotted . as final comments we would like to mention that to obtain an estimate of the error of our numerical results in table 1 one needs to carry out the calculations to the next order of the derivative expansion . it would be also interesting to expand the analysis for higher dimension operators . we would like to thank tim morris and jos latorre for some fruitful discussions during the workshop . we also acknowledge financial support by fundao para a cincia e a tecnologia under grant number cern / s / fae/ + 1177/97 . yu.k . acknowledges financial support from fellowship praxis xxi + /bcc/4802/95 . r.n . acknowledges financial support from fellowship praxis xxi / bpd/14137/97 . morris , int . j. mod . b12 ( 1998 ) 1343 ; in : `` zakopane 1997 . new developments in quantum field theory '' ( nato workshop on theoretical physics , zakopane , 1997 ) , p. 147 ( hep - th/9709100 ) ; preprint , hep - th/9802039 . a. hasenfratz and p. hasenfratz , nucl . b270 ( 1986 ) 687 ; t.r . morris , phys . b329 ( 1994 ) 241 ; p.e . haagensen , yu . kubyshin , j.i . latorre and e. moreno , phys . b323 ( 1994 ) 330 ; n. tetradis and c. wetterich , nucl . b422 ( 1994 ) 541 ; j. comellas , nucl . phys . b509 ( 1998 ) 662 .	we investigate a @xmath0-symmetric scalar field theory in two dimensions using the polchinski exact renormalization group equation expanded to second order in the derivative expansion . we find preliminary evidence that the polchinski equation is able to describe the non - perturbative infinite set of fixed points in the theory space , corresponding to the minimal unitary series of 2d conformal field theories . we compute the anomalous scaling dimension @xmath1 and the correlation length critical exponent @xmath2 showing that an accurate fit to conformal field theory selects particular regulating functions .
time - distance helioseismology @xcite constructs wave packets by cross - correlating time signals between any two points separated by a fixed horizontal angular distance on the solar surface . it then measures their travel time between the two points , by fitting a gabor wavelet to the observed temporal cross - correlation @xcite . this travel time is then inverted to study different properties @xcite ( e.g. sound speed perturbations , subsurface flows , meridional circulation ) in the solar interior that influence the wave packet , which can not otherwise be studied by the global oscillations . the success in applying the technique depends on being able to design a phase speed filter of a certain width , that selects acoustic waves within a certain range of horizontal phase speeds , and this improves the signal - to - noise ratio @xcite . waves with the same horizontal phase speed travel the same horizontal distance on the solar surface , and therefore can be collectively used to probe the sub - surface features in the sun , since they sample the same vertical depth inside the sun . the filtering operation is carried out in the frequency domain , and it improves the signal - to - noise ratio by removing unwanted signals and waves that deviate a lot from the chosen phase speed , as these do not contribute to the cross - correlation , but instead can degrade it , and make the estimation of travel times inaccurate . the phase speed filter has a gaussian shape centered at the desired phase speed , which is chosen to study waves traveling a particular horizontal distance . previous approaches to measuring the travel time between any two points was by fitting a gabor wavelet @xcite to the measured cross correlation . this wavelet was derived by assuming that the amplitude of the solar oscillations has a gaussian envelope in frequency of a certain width and peaked at a frequency where the power of the solar p - modes is concentrated , and moreover it considered only the radial component of the displacement . however , during data processing a phase speed filter is used . also , the observed displacement on the solar surface has both radial and horizontal components . the horizontal component usually is ignored in the travel - time measurements . we show that it has a significant effect , particularly , for measurements far from the disk center and for moderate horizontal distances , contributing to systematic errors . to remove the shortcomings , we derive a new analytical cross - correlation wavelet that incorporates the phase speed filter , and also includes both the radial and horizontal components of the doppler velocity oscillation signal . this wavelet retains the structure of the gabor wavelet , and is a function of the filter parameters : central phase speed and width , and also on the amplitude , phase and group travel times that depend on the oscillation properties and the dispersive nature of the solar medium . by including the horizontal component in the cross - correlation we also see a dependency on the location of the two points being cross - correlated , and moreover , the cross - correlation due to the horizontal component is a weighted sum of phase speed filtered gabor wavelets , with the weights depending on the location of the cross - correlated points , the horizontal travel distance and the angular degree . comparing this with the original gabor wavelet formula @xcite we can estimate how the phase - speed filter and the horizontal component shift the measured travel times . for the line depth and intensity observations , that only have a scalar ( radial ) component , in the weakly dispersive limit which is true for wave packets that probe the deeper layers in the sun and travel large horizontal distances on the solar surface , corresponding to small values of angular degree both formulae are similar . hence , using the old gabor wavelet in this regime should not effect the travel time measurements due to the phase speed filtering . on the other hand , wave packets constructed by cross - correlating time signals separated by a small horizontal distance , are distorted by the dispersive medium in the outer layers of the sun , and , hence , the new formula should be used to account for the travel time shifts due to the phase - speed filtering procedure . in this section , we generalize the fitting formula of @xcite by including the phase - speed filtering . we first consider acoustic waves , that are observed by measuring either intensity fluctuations or line depth observations on the solar surface . they are represented as a sum of standing waves or normal modes at a point @xmath0 in the solar interior and time @xmath1 , and can be written as @xmath2 where , @xmath3 , and each normal mode is specified by a 3-tuple @xmath4 of integer parameters , corresponding angular frequency @xmath5 , the mode amplitude @xmath6 , the phase @xmath7 and the spatial eigenfunction @xmath8 as a function of the radial variable @xmath9 and the angular variables @xmath10 . the integer @xmath11 denotes the degree and @xmath12 the azimuthal order , @xmath13 , of the spherical harmonic @xmath14 which is a function of the co - latitude @xmath15 and longitude @xmath16 . here , @xmath17 is a legendre function , and @xmath18 is a normalization constant . these describe the angular structure of the eigenfunctions . the third integer @xmath19 of the 3-tuple @xmath4 is called the radial order . for a spherically symmetric sun the eigenfunctions @xmath20 can be separated into a radial function @xmath21 and an angular component @xmath22 @xcite . this representation is valid for example for the scalar intensity observations , @xmath23 and also all modes with the same @xmath19 and @xmath11 have the same eigenfrequency @xmath24 , regardless of the value of @xmath12 . in reality the sun is not spherically symmetric , that causes this degeneracy in @xmath12 to be broken . in time - distance helioseismology we measure the travel time of wave packets by forming a temporal cross - correlation between the oscillation signals at two locations separated by an angular distance on the solar surface @xmath25 , where @xmath26 is the solar radius . to model this we represent the solar oscillations on the solar surface as a linear superposition of normal modes , that are band - limited in angular frequency @xmath27 . @xmath28 where @xmath29 is the temporal fourier transform @xcite of equation ( 1 ) . we ignore mode damping and assume spherical symmetry , hence we can replace @xmath5 by @xmath24 and invoke the band - limited nature of the solar spectrum , @xcite , by setting @xmath30 , and we obtain from equation ( [ eq4 ] ) the frequency band - limited signal @xmath31 @xmath32 where , the gaussian frequency function @xmath33 captures the band - limited nature of the amplitude of the solar modes , and is given by @xmath34 the band - limited nature of the modes is controlled by the width @xmath35 , and a central frequency @xmath36 , where the power of the modes is peaked in the @xmath37 diagram . these two parameters change for different data sets , that probe different regions of the solar interior . a phase speed filter is applied to the signal to select modes from the @xmath37 diagram for the purpose of constructing the cross - correlation wave packet . the phase speed filter is specified by a gaussian centered around a phase speed @xmath38 and a width @xmath39 as parameters , and is given by @xmath40 where , the phase speed @xmath41 , @xmath42 , @xmath43 is the horizontal wave number . the role of the phase speed filter is to select waves with a small range of phase speeds , the range is specified by the width @xmath39 . all these waves travel almost the same horizontal distance on the solar surface , and sample approximately the same vertical depth in the solar interior . hence , it is crucial to select @xmath39 appropriately so as to increase the signal - to - noise ratio in the cross - correlation function , and hence be able to better resolve the sub - surface structures in the sun . due to the band - limited nature of the oscillation amplitudes , only values of @xmath44 which are close to @xmath45 contribute to the sum in equation ( [ eq5 ] ) , and hence we taylor expand @xmath44 about the central frequency @xmath36 , upto first order : @xmath46 \approx l(\omega_o ) + \frac{dl}{d\omega}(\omega - \omega_o ) \label{eq8}\ ] ] the equation ( [ eq8 ] ) can be written in terms of the angular group velocity @xmath47 and angular phase velocity @xmath41 , evaluated at @xmath48 , and using the fact @xmath49 , we have @xmath50 likewise , the phase velocity @xmath51 can be expanded about the point @xmath52 in the @xmath37 diagram to yield @xmath53 where , @xmath54 , and the filter width @xmath39 is evaluated at @xmath55 , and is a constant . phase speed filtering takes place in the frequency domain and consists of multiplying the filter function @xmath56 with the band - limited fourier transformed signal @xmath57 from equation ( [ eq5 ] ) . @xmath58 equation ( [ eq11 ] ) can be inverse fourier transformed to yield the phase speed filtered temporal signal @xmath59 at a point @xmath60 on the solar surface . this signal now consists of a superposition of modes that lie in a region of the @xmath37 diagram , that is the intersection region of the frequency band - limited gaussian envelope of the solar modes and the phase speed filter . @xmath61 the temporal cross - correlation function @xmath62 of the phase speed filtered signal @xmath59 between two points a and b having coordinates @xmath63 and @xmath64 respectively on the solar surface with a fixed angular separation distance @xmath65 , as a function of the time lag @xmath66 is @xmath67 where , @xmath68 is the complex conjugate of the phase speed filtered signal @xmath69 , and @xmath70 is the length of the time series . substituting the expression for @xmath71 from equation ( [ eq12 ] ) into equation ( [ eq13 ] ) one gets after applying the orthonormality of @xmath72 in the temporal integral in equation ( [ eq13 ] ) @xmath73 where , @xmath74 is the complex conjugate of @xmath75 . since the phases are random , we assume that the term @xmath76 in the double sum for @xmath12 and @xmath77 is zero except when @xmath78 . in this case equation ( 14 ) becomes @xmath79 we can simplify equation ( 15 ) by applying the addition theorem of spherical harmonics @xcite @xmath80 where , @xmath81 is the legendre polynomial of order @xmath11 , @xmath82 for large @xmath11 @xcite , and @xmath65 is the angular distance on the solar surface between the two points a @xmath83 and b @xmath84 , and is ( see appendix a equations ( [ eq43 ] ) and ( [ eq44 ] ) , appendix d , figure [ fg6 ] and appendix e , figure [ fg8 ] ) @xmath85 we can approximate for large @xmath11 , small @xmath65 , such that @xmath86 is large @xcite @xmath87 \approx \sqrt{\frac{2}{\pi l\delta } } \cos \left(l\delta - \frac{\pi}{4}\right ) \label{eq18}\ ] ] where , @xmath88 is the bessel function of the first kind . with this approximation from equation ( [ eq18 ] ) , the addition theorem in equation ( [ eq16 ] ) becomes @xmath89 using equation ( [ eq18a ] ) , the equation ( [ eq15 ] ) for the phase speed filtered cross - correlation becomes after taking the real part @xmath90 the double sum in equation ( [ eq19 ] ) can be reduced to a convenient sum of integrals if we regroup the modes so that the outer sum is over the phase speed @xmath91 , and the inner sum is over @xmath24 @xcite . the expression for @xmath62 is an even function of @xmath66 , so we get an identical expression if we replace @xmath66 by @xmath92 ( negative time lag ) . the product of cosines in equation ( [ eq19 ] ) can be transformed by using the trigonometric identity , and this results in @xmath93 \label{eq20}\ ] ] in equation ( [ eq20 ] ) we let @xmath94 corresponding to the positive time lag @xmath66 and @xmath95 corresponding to the negative time lag @xmath92 . since cosine is an even function , it is seen that @xmath96 , hence we can drop @xmath97 from equation ( [ eq20 ] ) , while extending the sum to negative values of @xmath24 . we now substitute the expressions for the gaussian envelope frequency function and the phase speed filter from equations ( [ eq6 ] ) and ( [ eq10 ] ) respectively into equation ( [ eq20 ] ) and since @xmath24 is a dummy variable in the inner sum , it is replaced by @xmath27 for notational convenience . moreover , the inner sum over @xmath24 can be replaced by an integral over @xmath27 , hence the discrete inner sum in equation ( [ eq20 ] ) becomes @xmath98 \cos \left(\omega\tau - l\delta + \frac{\pi}{4 } \right ) \label{eq21}\ ] ] the integral in equation ( [ eq21 ] ) is negligible for large frequencies hence there is very little error made in extending the frequency limit to @xmath99 and @xmath100 . here , we choose the coefficient @xmath101 for the gaussian envelope function , such that @xmath102 in order to evaluate the integral in equation ( [ eq21 ] ) we taylor expand @xmath44 , and using the linear dispersion relation from equation ( [ eq9 ] ) we get @xmath103 @xmath104 where , the group travel time is defined as @xmath105 and phase travel time @xmath106 . substituting equations ( [ eq22 ] ) and ( [ eq23 ] ) into the exponent of the phase speed filter and the argument of the cosine term respectively in equation ( [ eq21 ] ) , and carrying out the integral @xcite , which is bounded since the exponentials decay at both the limits of integration , with some algebra yields @xmath107 where , @xmath108 and @xmath109 are dimensionless quantities representing the relative deviation of the group and phase travel times respectively from the filter phase travel time , @xmath110 and @xmath111 . @xmath112 the argument of the cosine term can be written as @xmath113 where , the shifted frequency , @xmath114 and the shifted phase travel time due to the phase speed filter is @xmath115 the dimensionless parameter @xmath116 is zero in the absence of phase speed filtering , which occurs when the filter width tends to infinity or in the non - dispersive limit when @xmath117 . also from equation ( [ eq24b ] ) we see a shift in the phase travel times due to the phase speed filtering , which tends to zero when @xmath118 tends to zero . also if we choose @xmath38 to be either @xmath119 or @xmath120 , then the effect of phase speed filtering can be removed since @xmath117 . summing equation ( [ eq24 ] ) over phase velocities we get the final cross - correlation . @xmath121 in the non - dispersive limit when @xmath117 , that is @xmath122 , or in the limit of no phase speed filtering , large @xmath123 , equations ( [ eq24 ] ) and ( [ eq25 ] ) reduce to the original gabor wavelet formula @xmath124 \cos \left [ \omega_o(\tau - \tau_p ) + \frac{\pi}{4 } \right ] \label{eq27}\ ] ] in previous papers @xcite , the factor of @xmath125 has been absorbed in the phase travel time . by defining the phase travel time @xmath126 , the gabor wavelet in equation ( [ eq27 ] ) reduces to the familiar form with the sinusoidal part being @xmath127 $ ] . we conclude that while the phase - speed filtering procedure does not change the functional form of the basic time - distance helioseismology fitting formula , it systematically shifts the travel times , if the filter parameter , @xmath38 , is different from the actual phase or group speeds for a given distance . we now consider the doppler shift observations which are commonly used in helioseismology . the doppler velocity depends on both the radial and horizontal components of the displacement . for simplicity , we take the axis of the spherical harmonics to be in the plane of the sky , orthogonal to the line of sight . the displacement vector @xmath128 at location @xmath0 and time @xmath1 is written for the spherically symmetric sun as @xcite @xmath129 \right \ } \exp \left \{i(\omega_{nl}t - \alpha_{nlm})\right \ } \label{eq28}\ ] ] where , @xmath130 , @xmath131 are the radial and horizontal components of the displacement eigenfunctions respectively , and @xmath132 , @xmath133 and @xmath134 are unit vectors in @xmath9 , @xmath15 and @xmath16 directions respectively ( see appendix a , figure [ fg5 ] ) . thus , the observed doppler signal on the solar surface is the projection of the displacement vector onto the line - of - sight direction . without loss of generality we take the x - axis as the line - of - sight direction , hence the projection is just @xmath135 , where @xmath136 is the unit vector along the x - axis and is given by ( see appendix a , equation ( [ eq40 ] ) ) @xmath137 the doppler signal on the solar surface is obtained by taking the dot product between @xmath138 and the line of sight direction @xmath136 , and is therefore @xmath139 . for a mode @xmath4 it is @xmath140 \right \ } \label{eq30}\ ] ] where , @xmath141 the ratio of horizontal to radial component @xmath142 shows dependence in frequency , and since the solar modes are band - limited in frequency , and peaked around @xmath36 , we replace the frequency @xmath24 by a constant @xmath143 . here , @xmath144 is the acceleration due to gravity at the solar surface , @xmath145 is the dimensionless frequency and @xmath146 is a factor that depends on the boundary condition used at the solar surface @xmath25 . for the solar 5-min oscillations @xmath147 , and for the boundary condition that the lagrangian pressure perturbation vanishes at the solar surface : at @xmath148 , leads to @xmath149 , and this shows that at the solar surface the radial component dominates . for other types of boundary conditions the factor @xmath146 may depend on @xmath44 . if one selects waves with other frequencies , which are peaked around @xmath150 instead of @xmath151 3.3 mhz for the 5 minute oscillations , then the value of @xmath152 at @xmath150 can be easily calculated from @xmath153 . with the time dependence the doppler signal is @xmath154 the cross - correlation @xmath155 can be computed for the phase speed filtered doppler signal @xmath156 in a similar manner as for equation ( [ eq12 ] ) and involves the product of the projected line of sight doppler signals at the two locations a @xmath83 and b @xmath84 on the solar surface ( see figures [ fg6 ] and [ fg8 ] ) . here we have replaced @xmath157 by the gaussian frequency function @xmath33 , which models the amplitude of the solar modes . @xmath158 in evaluating this cross - correlation we need the summation over @xmath12 of the spherical harmonics and its derivatives , which are given in appendix c. collecting all the sums over @xmath12 we define the total sum ( see appendix c equations ( [ eq89])-([eq92 ] ) ) @xmath159 substituting the expressions for @xmath160 and its derivatives from appendix b equations ( [ eq50 ] ) , ( [ eq51 ] ) and ( [ eq52 ] ) , we obtain for large @xmath11 , small @xmath65 , such that @xmath86 is large @xmath161 \label{eq35}\ ] ] where @xmath162 and @xmath163 . equation ( [ eq35 ] ) can be written as @xmath164 = c_l \sqrt{\frac{2}{\pi l\delta } } \cos \left(l \delta - \frac{\pi}{4 } + \zeta \right ) \label{eq36}\ ] ] where , the phase factor @xmath165 , @xmath166 , @xmath167 and @xmath168 . the value of @xmath44 outside the trigonometric functions is evaluated at @xmath36 and is @xmath169 . the phase factor @xmath170 in general is a function of frequency @xmath24 , but since the solar oscillations are band - limited and their power is peaked at the frequency @xmath36 , @xmath170 is evaluated at the central frequency @xmath36 . this also simplifies the evaluation of the integral , and a closed form expression can be derived . the phase shift @xmath170 is measured in radians but can be converted to units of time by dividing by @xmath36 . the cross - correlation is @xmath171 substituting the expression for @xmath172 from equation ( [ eq36 ] ) into equation ( [ eq37 ] ) and transforming the product of the cosines to a sum we obtain @xmath173 \label{eq38}\ ] ] where , @xmath174 and @xmath175 correspond to positive and negative time lags respectively . evaluating the inner sum in equation ( [ eq38 ] ) in a similar manner as in equation ( [ eq20 ] ) one obtains @xmath176 the shifted phase travel time @xmath177 due to the horizontal component is therefore , @xmath178 summing equation ( [ eq39 ] ) over phase velocities we get the final cross - correlation . @xmath179 in deriving this equation we used the asymptotic formula for the legendre functions , which allowed us to obtain the explicit expression for the phase shift . we verified by direct numerical calculations of the cross - covariance functions that this approximation is sufficiently accurate . a comparison of the analytical formula in equation ( [ eq36 ] ) for @xmath180 with the direct numerical mode summation is shown in figure [ fg1 ] . this approximation holds when the travel times are much greater than the oscillation period of five minutes , which is the case for typical time - distance measurements . from the equations ( [ eq35 ] ) and ( [ eq37 ] ) we observe that the cross - correlation of the doppler line of sight velocity involving the horizontal component of the displacement is a weighted sum of the phase speed filtered gabor wavelets , with the weights depending on the location @xmath181 of the two points being cross - correlated , and the angular degree @xmath11 of the modes . this weighted sum can be conveniently combined incorporating a phase factor into the cosine term . from equation ( [ eq39a ] ) we see a shift in phase travel times due to the horizontal component in addition to that introduced by the phase speed filtering . in figure [ fg2 ] , we see a variation in the travel time shift due to the horizontal component as we move point b ( see appendix e , figure [ fg8 ] ) around the annulus by changing the angle @xmath182 ( see appendix d , figure [ fg6 ] ) , cross correlating with the fixed point a. figure [ fg3 ] illustrates the phase shift for the projection on the solar disk . the greatest phase shift changes happen for the outmost points in the direction from the disk center . averaging the travel time shift over the annular angle @xmath182 , for different horizontal distances @xmath65 and distances from disk center @xmath183 , to simulate the observed mean travel times , we show in figure [ fg4 ] that this time shift increases ( becomes more negative ) as we go away from disk center and also for increasing horizontal distances . the shifts are quite appreciable , on the order of a few seconds , and have been observed in the real data from soho / mdi @xcite . at disk center c , ( see appendix e , figure [ fg7 ] ) which according to our convention is at @xmath184 , which is the origin of the yz plane perpendicular to the line of sight direction @xmath136 . in polar coordinates the disk center is specified by @xmath185 , at which the horizontal component is zero in equation ( [ eq30 ] ) , and the radial component is maximum . we see from figure [ fg2 ] , when point a is at disk center , the time shift is independent of @xmath182 , moreover there will always be a small travel time shift for this case since point b is at a distance @xmath65 from the disk center , and it has a non - zero horizontal component that when cross - correlated with the radial component of the signal at point a , which is at disk center , leads to a travel time shift . a dispersive medium is where the phase velocity @xmath119 is quite different from the group velocity @xmath120 @xcite . due to this for a fixed distance @xmath65 , the group and phase travel times are different . in the solar case the wave packets that travel short distances , probe the outer layers of the sun , which are quite dispersive due to the presence of large gradients . hence the wave packets got by the cross - correlation in equations ( [ eq24 ] ) and ( [ eq25 ] ) are distorted as they travel through the medium . moreover , the approximations made by retaining only the first order terms in the taylor expansion of the dispersion relation in equation ( [ eq8 ] ) may be inadequate to completely model the physical effects associated with small distances , hence care must be exercised in interpreting the results got by using these wavelets for small distances . on the other hand , wave packets that travel large angular distances @xmath65 , probe the deeper layers of the sun , that are weakly dispersive , and hence undergo less distortion . it is quite interesting to note that the gabor wavelet in equation ( [ eq27 ] ) , in the absence of the phase speed filter is got in the non - dispersive limit of equations ( [ eq24 ] ) and ( [ eq25 ] ) when @xmath117 , that is @xmath122 , for the intensity or line depth signal which has only the radial component . this is quite intuitive since in this limit there is no phase speed filtering , and @xmath186 , hence the phase speed filtered signal in equation ( [ eq11 ] ) reduces to the frequency band - limited signal in equation ( [ eq5 ] ) . this is consisted both physically and mathematically . to form wave packets the medium need not be dispersive . waves in a range of frequencies and wave lengths can interfere to form a wave packet . the wave packets that travel through a dispersive medium are distorted , where as those through a non - dispersive medium are not distorted , but retain their original shape . the dimensionless quantities @xmath187 and @xmath188 measure the amount of deviation from non - dispersiveness . hence to fit the observed temporal cross - correlation for the intensity or line depth signal where , @xmath188 and @xmath187 are small , there will be little error in using the gabor wavelet of equation ( [ eq27 ] ) , since the new gabor wavelet of equations ( [ eq24 ] ) and ( [ eq25 ] ) is very close to the old gabor wavelet @xcite in this non - dispersive or weakly dispersive limit for the line depth or scalar intensity signal that have only the radial component . in this paper we introduce a mathematical formalism to deal with the horizontal component of the displacement , which is used to derive a new formula to estimate travel times in time - distance helioseismology . it models the observed cross - correlation , by considering both the radial and horizontal components of the displacement , and also includes the phase speed filter on the frequency band - limited solar oscillations . the form of the gabor wavelet is retained by the present formula , and in the non - dispersive limit for the line depth or intensity observations it reduces to the previously used gabor wavelet that was derived without a phase speed filter , considering only the radial component . including the horizontal component , the cross - correlation is a weighted sum of phase speed filtered gabor wavelets , with the weights depending on the angular position of the two points being cross - correlated , and the angular degree of the modes . due to the phase speed filtering and inclusion of the horizontal component of the displacement , shifts in phase travel times are observed . in particular , the horizontal component induces a systematic shift in travel times , whose absolute value increases as we go away from disk center and also for large horizontal distances . this will in turn have implications in the inversions to study flows and the sub - surface properties in the sun and also to formulate other local helioseismic approaches that involve the horizontal component . this project was supported by the nasa mdi grant to stanford university . bracewell , r. , 2000 , the fourier transform and its applications , 3rd edition ( new york : mcgraw - hill ) christensen - dalsgaard , j. 2002 , _ rev . _ , 74 , 1073 christensen - dalsgaard , j. 2003 , lecture notes on stellar oscillations , 5th edition . duvall , t. l. , jr . , jefferies , s. m. , harvey , j. w. & pomerantz , m. a. 1993 , , 362 , 430 duvall , t. l. , jr . 1997 , sol . phys . , 170 , 63 duvall , t. l. , jr . 2003 , esa sp-517 : gong+ 2002 . local and global helioseismology : the present and future , 12 , 259 giles , p. m. 1999 , ph.d . thesis , stanford university gradshteyn , i. s. and ryzhik , i. m. 1994 , page 531 , table of integrals , series , and products ( 5th edition ) , academic press , san diego jackson , j. d. , 1999 , classical electrodynamics , 3rd edition ( new york : john wiley & sons ) jefferies , s.m . , osaki , y. , shibahashi , h. , duvall , t.l . harvey , j.w . , & pomerantz , m.a . 1994 , , 434 , 795 kosovichev , a. g. 1996 , , 461 , l55 kosovichev , a.g . , duvall , t.l . , jr . in score-96 : solar convection and oscillations and their relationship , ed . pijpers et al . , 1997 ( kluwer ) lighthill , m.j . , 1978 , waves in fluids ( cambridge : cambridge university press ) smart , w. m. , 1977 , textbook on spherical astronomy , 6th edition ( cambridge : cambridge university press ) unno , w. , osaki , y. , ando , h. , & shibahashi , h. , 1989 , nonradial oscillations of stars ( university of tokyo press , tokyo ) whitham , g.b . , 1974 , linear and nonlinear waves ( new york : john wiley & sons ) winch , d. e. and roberts , p. h. 1995 , j. austral math soc ser b 37 , 212 zhao , j. 2004 ph.d . thesis , stanford university we define the spherical coordinate system used in this paper ( see figure [ fg5 ] ) . consider fixed cartesian unit vectors @xmath189 and the corresponding spherical unit vectors @xmath190 that vary with the angles @xmath10 . the equations relating these sets of unit vectors are @xmath191 @xmath192 @xmath193 any point @xmath194 in spherical coordinates @xmath195 is @xmath196 , @xmath197 and @xmath198 . consider two points a @xmath199 and b @xmath200 on the solar surface @xmath25 ( see figure [ fg6 ] , appendix d ) , with coordinates @xmath201 and @xmath202 , the angular distance @xmath65 between them can be calculated by taking the dot product of the unit vectors at these two locations and is given by @xmath203 where , @xmath204 and + @xmath205 ( see figure [ fg6 ] , appendix d ) . substituting this into equation ( [ eq43 ] ) we obtain @xmath206 we derive various approximations for @xmath207 and its derivatives for large @xmath11 , small @xmath65 and large @xmath86 following the approach of @xcite . @xmath87 \approx \sqrt{\frac{2}{\pi l\delta } } \cos \left(l\delta - \frac{\pi}{4}\right ) \label{eq45}\ ] ] from the recursion relation @xcite , @xmath208 for small @xmath65 equation ( [ eq46 ] ) becomes , @xmath209 differentiating equation ( [ eq46 ] ) with respect to @xmath65 gives @xmath210 for small @xmath65 and large @xmath11 , equation ( [ eq48 ] ) becomes , @xmath211 now for small @xmath65 , large @xmath11 and large @xmath86 the various approximations are after using the recursion relations from equations ( [ eq47 ] ) and ( [ eq49 ] ) @xmath212 @xmath213 @xmath214 using the addition theorem of the derivatives of spherical harmonics @xcite , we derive the expressions for the terms in equation ( [ eq34 ] ) . the angular distance @xmath65 is , @xmath215 differentiating equation ( [ eq53 ] ) with respect to @xmath216 , @xmath217 , @xmath218 and @xmath219 we obtain @xmath220 @xmath221 @xmath222 @xmath223 the expressions in equations ( [ eq54])-([eq57 ] ) can be equated to sine and cosine functions of @xmath182 and @xmath224 @xcite , which are shown in figure [ fg6 ] , appendix d using the results of spherical trigonometry @xcite . according to the addition theorem of spherical harmonics @xcite @xmath225 for notational convenience we denote @xmath226 . differentiating equation ( [ eq58 ] ) with respect to @xmath217 and applying the chain rule leads to @xmath227 differentiating equation ( [ eq58 ] ) with respect to @xmath219 and applying the chain rule leads to @xmath228 differentiating equation ( [ eq58 ] ) with respect to @xmath216 and applying the chain rule leads to @xmath229 differentiating equation ( [ eq61 ] ) with respect to @xmath217 and applying the product and chain rule gives @xmath230 \label{eq62}\ ] ] differentiating equation ( [ eq61 ] ) with respect to @xmath219 and applying the product and chain rule gives @xmath231 \label{eq63}\ ] ] differentiating equation ( [ eq58 ] ) with respect to @xmath218 and applying the chain rule leads to @xmath232 differentiating equation ( [ eq64 ] ) with respect to @xmath217 and applying the product and chain rule gives @xmath233 \label{eq65}\ ] ] differentiating equation ( [ eq64 ] ) with respect to @xmath219 and applying the product and chain rule gives @xmath234 \label{eq66}\ ] ] now using equations ( [ eq30 ] ) , ( [ eq58])-([eq66 ] ) we obtain for the various terms in equation ( [ eq34 ] ) @xmath235 for notational convenience we denote the angular coordinates of points @xmath236 and @xmath237 by @xmath238 @xmath239 @xmath240 @xmath241 @xmath242 ^ 2 \cos\theta_1 \cos\phi_1 \cos\theta_2 \cos\phi_2 s_{nl}^{\theta \theta } = f^{\theta \theta}_{52}(\vec a_{12 } ) \frac { d^2p_l(\cos\delta ) } { d\delta^2 } + f^{\theta \theta}_{51}(\vec a_{12 } ) \frac { dp_l(\cos\delta ) } { d\delta } \label{eq71}\ ] ] @xmath243 ^ 2 \cos\theta_1 \cos\phi_1 \frac { \sin\phi_2 } { \sin\theta_2 } s_{nl}^{\theta \phi } = f^{\theta \phi}_{62}(\vec a_{12 } ) \frac { d^2p_l(\cos\delta ) } { d\delta^2 } + f^{\theta \phi}_{61}(\vec a_{12 } ) \frac { dp_l(\cos\delta ) } { d\delta } \label{eq72}\ ] ] @xmath244 @xmath245 ^ 2 \frac { \sin\phi_1 } { \sin\theta_1 } \cos\theta_2 \cos\phi_2 s_{nl}^{\phi \theta } = f^{\phi \theta}_{82}(\vec a_{12 } ) \frac { d^2p_l(\cos\delta ) } { d\delta^2 } + f^{\phi \theta}_{81}(\vec a_{12 } ) \frac { dp_l(\cos\delta ) } { d\delta } \label{eq74}\ ] ] @xmath246 ^ 2 \frac { \sin\phi_1 } { \sin\theta_1 } \frac { \sin\phi_2 } { \sin\theta_2 } s_{nl}^{\phi \phi } = f^{\phi \phi}_{92}(\vec a_{12 } ) \frac { d^2p_l(\cos\delta ) } { d\delta^2 } + f^{\phi \phi}_{91}(\vec a_{12 } ) \frac { dp_l(\cos\delta ) } { d\delta } \label{eq75}\ ] ] where we evaluate the derivatives of @xmath65 with respect to @xmath216 , @xmath217 , @xmath218 and @xmath219 using equations ( [ eq54])-([eq57 ] ) and the equations from @xcite , we obtain the following expressions @xmath247 @xmath248 @xmath249 @xmath250 @xmath251 ^ 2 \cos \theta_1 \cos \phi_1 \cos \theta_2 \cos \phi_2 \frac{\sin \gamma_1 \sin \gamma_2 } { \sin \delta } \label{eq80}\ ] ] @xmath252 ^ 2 \cos \theta_1 \cos \phi_1 \cos \theta_2 \cos \phi_2 \cos \gamma_1 \cos \gamma_2 \label{eq81}\ ] ] @xmath253 ^ 2 \cos \theta_1 \cos \phi_1 \sin \phi_2 \frac{\sin \gamma_1 \cos \gamma_2}{\sin \delta } \label{eq82}\ ] ] @xmath254 ^ 2 \cos \theta_1 \cos \phi_1 \sin \phi_2 \cos \gamma_1 \sin \gamma_2 \label{eq83}\ ] ] @xmath255 @xmath256 ^ 2 \sin \phi_1 \cos \theta_2 \cos \phi_2 \frac{\cos \gamma_1 \sin \gamma_2}{\sin \delta } \label{eq85}\ ] ] @xmath257 ^ 2 \sin \phi_1 \cos \theta_2 \cos \phi_2 \sin \gamma_1 \cos \gamma_2 \label{eq86}\ ] ] @xmath258 ^ 2 \sin \phi_1 \sin \phi_2 \frac{\cos \gamma_1 \cos \gamma_2}{\sin \delta } \label{eq87}\ ] ] @xmath259 ^ 2 \sin \phi_1 \sin \phi_2 \sin \gamma_1 \sin \gamma_2 \label{eq88}\ ] ] in computing the cross correlation in equation ( [ eq37 ] ) , we need to evaluate @xmath172 in equation ( [ eq34 ] ) as @xmath260 where , @xmath261 @xmath262 @xmath263 in this appendix we derive the angular coordinates @xmath200 of point b on the annulus , which is at a fixed annular angle @xmath182 and a given angular distance @xmath65 from the fixed point a with angular coordinates @xmath199 at the center of the annulus . we use equations ( [ eq53 ] ) and ( [ eq54 ] ) from appendix c to express the angular coordinates of b in terms of the fixed angles @xmath264 . the geometry for the calculation is shown in figure 6 . @xmath265 @xmath266 we now consider the case when the two points are close together , which is true for small distances @xmath65 . they are related as @xmath267 and @xmath268 , for small increments @xmath269 and @xmath270 . the equations ( [ eq93 ] ) and ( [ eq94 ] ) can be expanded in terms of the increments , and using the standard trigonometric approximations for small angles @xmath271 and @xmath272 we obtain , @xmath273^{\frac{1}{2 } } \label{eq95}\ ] ] @xmath274^{\frac{1}{2 } } } \label{eq96}\ ] ] for fixed @xmath65 and @xmath182 we use equations ( [ eq95 ] ) and ( [ eq96 ] ) to calculate @xmath200 for small angular separation . for larger separations we can solve the linear equations ( [ eq93 ] ) and ( [ eq94 ] ) to calculate @xmath217 , and equation ( [ eq56 ] ) to compute @xmath219 , and equations ( [ eq55 ] ) and ( [ eq57 ] ) to calculate @xmath275 and @xmath276 respectively . in this appendix we consider the orthographic projection shown in figure [ fg7 ] , where the points on the surface of the sun are projected onto the @xmath277 plane p , which is perpendicular to the line of sight direction @xmath136 . the origin o is at the center of the sun of radius @xmath26 . we denote a point a@xmath278 on the solar surface as @xmath194 with axis @xmath279 and as @xmath280 the coordinate of the same point projected as a on the projection plane p with axis @xmath277 . the @xmath281 axis is perpendicular the the projection plane . the projection is oriented with north along @xmath282 axis and east along the @xmath283 axis . the projection is centered at co - latitude of @xmath284 and longitude of @xmath285 . with this convention , the north pole n has coordinates @xmath286 , the south pole s is @xmath287 , and the point of tangency c between the sphere and the plane is @xmath288 . the point c is also called the disk center . the projection plane is @xmath289 . the coordinates of any point a@xmath278 with co - latitude @xmath216 and longitude @xmath218 on the solar surface are @xmath290 the equation of the sphere is @xmath291 . the coordinates of the same point on the plane are @xmath292 . the point of tangency c between the sphere and the plane is the disk center and has coordinates @xmath293 on the plane . hence , the distance from the disk center c @xmath294 to any point a @xmath280 on the plane p is denoted by @xmath183 , which is fixed for each computation of the cross correlation . any point a@xmath278 on the solar surface and hence the corresponding projection a on the plane p can be specified by @xmath199 or alternatively by the two parameters @xmath183 and the angle @xmath218 . in figure [ fg8 ] , the center to annulus geometry is shown with the fixed center point a @xmath199 , which is at a distance @xmath183 from disk center c @xmath293 , and a point b @xmath200 is on the annulus @xmath295 of radius @xmath65 . @xmath296 the angular coordinates of the disk center c , are @xmath297 . when the point a is at the disk center @xmath298 , and hence the distance @xmath299 . for fixed values of @xmath300 , together with the coordinates @xmath199 of the annulus center a , we can compute the coordinates @xmath200 of point b on the annulus by solving the linear system of equations ( [ eq93 ] ) and ( [ eq94 ] ) to get @xmath217 and solving equation ( [ eq56 ] ) to compute @xmath219 .	in time - distance helioseismology , the time signals ( doppler shifts ) at two points on the solar surface , separated by a fixed angular distance are cross - correlated , and this leads to a wave packet signal . accurately measuring the travel times of these wave packets is crucial for inferring the sub - surface properties in the sun . the observed signal is quite noisy , and to improve the signal - to - noise ratio and make the cross - correlation more robust , the temporal oscillation signal is phase - speed filtered at the two points in order to select waves that travel a fixed horizontal distance . hence a new formula to estimate the travel times is derived , in the presence of a phase speed filter , and it includes both the radial and horizontal component of the oscillation displacement signal . it generalizes the previously used gabor wavelet that was derived without a phase speed filter and included only the radial component of the displacement . this is important since it will be consistent with the observed cross - correlation that is computed using a phase speed filter , and also it accounts for both the components of the displacement . the new formula depends on the location of the two points on the solar surface that are being cross correlated and accounts for the travel time shifts at different locations on the solar surface .
quantum spin hall effect ( qshe ) , which is proposed as a new class of topological state of matter in two dimensions , has generated a great deal of interest @xcite . contrary to the integer or the fractional quantum hall state , which is induced by a magnetic field that breaks the time reversal symmetry , qshe is caused by a strong spin - orbit interaction that maintains the time reversal symmetry . kane and mele proposed a @xmath0 classification for this kind of the new topological state@xcite . for materials with spatial inversion symmetry , the index could be easily computed as the product of parities of the wave function at several high symmetry points in the brillouin zone . the @xmath0 classification can be generalized to three - dimensional systems as well @xcite . recently , topological insulators suitable for room temperature applications are also predicted@xcite . the qsh state has the helical edge states , namely , having two counter - propagating edge states for the two opposite spin polarizations . the helical edge states are stable against time - reversal conserving perturbations , since backscattering processes need to connect the upper and lower edges of the sample . the probability of backscattering is exponentially suppressed as the sample width is increased . recent experiment @xcite provides evidences of the qshe in hgte / cdte quantum well ( qw ) structures , as predicted theoretically@xcite . the de - coherence effect in qsh samples is also investigated.@xcite some interesting questions emerge such as the response of a qsh state to the disorder , and the interplay of the helical edge states and the bulk states . according to the celebrated scaling argument , in two dimensions the bulk electron wave functions are localized in the presence of any weak disorder @xcite . but there are two known exceptions , one is the two dimensional systems with a strong spin - orbital coupling , and the other is the quantum hall transition between different plateaus@xcite . with the discover of qshe state in two dimensions , study of the localization is in demand . sheng et al investigated the disorder effect in honeycomb lattice qshe sample , they found the qshe phase is robust against weak disorder@xcite . ono and obuse _ et al _ studied the critical properties of the transition from a metal to a qsh state , and they found the results to be somewhat controversial@xcite . recently , li@xcite _ et al _ studied the transport properties of the hgte / cdte qws in the presence of disorder , they found as the increasing of the disorder strength , the initially un - quantized conductance became quantized , _ i.e. _ the sample enters into the qsh state because of the disorder , so they named the state as `` topological anderson insulator''(tai ) . the newly anomalous quantized conductance plateau is caused by the edge transport , which is indirectly revealed by the unchanged plateau value along with width variation for two - terminal calculation and quantized transmission coefficient for four - terminal calculation . however , the detailed mechanics of the edge transport is less clear . in the present paper , we study the effect of disorder on the electronic state of hgte / cdte qws . we carry out keldysh s nonequilibrium - green - function ( negf ) calculations based on a four - band tight binding model . first , we perform the conductance calculations for two different geometries . in the case of a stripe geometry(see fig .1(a ) ) , the presence of the helical edge states are evident from the band spectrum , the tai phase described in reference 12 appears . while for the cylindrical geometry , _ i.e. _ periodical boundary condition along @xmath1 direction(see fig . 1(b ) ) , there is no edge state , the bulk state is localized by disorder and there is no quantized conductance . these results strongly support the thesis that anomalous conductance plateau is due to the edge transport and give a better understanding of the topological feature of the tai phase . second , we obtain the distributions of the local currents for the two - terminal strip samples with different chemical potentials and disorder strengths . the evolvement of the local current vector configurations gives rise to a direct demonstration of the impurity influence in the hgte / cdte qws . by analyzing these local current configurations , the transport phenomena in both normal qsh region and tai region are clearly explained . moreover , the detailed results , such as the coexistence of the bulk and edge states at the dip point ( see fig.6 ) and the bulk states assisted backscattering obtained from the local current vector configurations shed further light on the mechanism of the disorder induced edge states for the tai . the rest of this paper is organized as follows . in section ii , we introduce the effective tight - binding model . the formulas and calculation method are also described . in section iii , the numerical results and their discussions are presented . finally , a conclusion is given in section iv . as a starting point , we introduce the effective hamiltonian for the hgte / cdte qws with anderson impurity in the tight - binding representation . we consider a square lattice with four special orbit states @xmath2,@xmath3 , @xmath4 , @xmath5,@xmath6 denotes the electron spin . through symmetry consideration , the effective hamiltonian can be written as@xcite @xmath7 here @xmath8 is the site index , and @xmath9 and @xmath10 are unit vectors along the @xmath11 and @xmath1 directions . @xmath12 represents the four annihilation operators of electron on the site @xmath13 with the state indexes @xmath14,@xmath15 , @xmath4 , @xmath16 . @xmath17 , and @xmath18 are on - site disorder energies uniformly distributed in the range @xmath19 $ ] with the disorder strength @xmath20 @xmath21 and @xmath22 are the five independent parameters that characterize the clean hgte / cdte samples . it is clear that near the @xmath23 point the lattice hamiltonian ( eq . ( 1 ) ) in k - representation can be reduced to the continuous hamiltonian in ref.6 when we take @xmath24 , @xmath25 , @xmath26 , @xmath27 , and @xmath28 . here @xmath29 is the lattice constant and all the parameters @xmath30 can be controlled experimentally@xcite . moreover , the eq.(1 ) can be directly obtained by discretizing spatial coordinates of the continuous hamiltonian using the substitution @xmath31.@xcite in this paper , we apply the model to two geometric devices . the device a ( see fig . 1(a ) ) is of stripe geometry , while the device b ( see fig . 1(b ) ) is of cylindrical geometry which can be obtained by rolling the device @xmath29 into a tube . without disorder(w=0 ) , the energy spectrum of such two geometric devices can be calculated by diagonalizing eq.(1 ) using the periodic boundary condition in x direction @xcite . next we investigate how the disorder affects the transport properties of such systems . for both devices a and b , the size of the central region is @xmath32 . for convenience , we assume that the anderson impurities only exist in the ( red ) filled region and the temperature is zero . in our simulations , a small external bias @xmath33 is applied between the two terminals . with the help of the negf method , the local current between neighboring sites @xmath34 and @xmath35 is calculated from the formula@xcite @xmath36(v_{l}-v_{r})\ ] ] where @xmath37 are the voltages at the @xmath38-l , r . @xmath39 denote the state indices . @xmath40 is electron correlation function with line width function @xmath41 $ ] , the green function @xmath42^{-1}$ ] and the hamiltonian in the central region denoted as @xmath43 . the retarded self - energy @xmath44 due to the coupling to the lead-@xmath45 can be calculated numerically @xcite . note that for eq . ( 1 ) , the spin up(@xmath46 ) subsystem and spin down ( @xmath47 ) subsystem are decoupled . the local current between neighboring sites @xmath34 and @xmath35 with spin index @xmath48 @xmath49 can also be calculated from eq.(2 ) by summing over only the state index with the corresponding unitary subsystem . the current @xmath50 flowing through the device is calculated by summing all the local current @xmath51 for an arbitrary column . after obtaining the current @xmath50 , the linear conductance @xmath52 is given by @xmath53 . in addition , the linear conductance can be directly obtained by @xmath54 $ ] . the agreement between the two methods gives strong confirmation of our analytical derivations and numerical calculations . , and gap parameter @xmath55 ( c ) , @xmath56 ( e ) . ( d ) and ( f ) demonstrate the one - dimensional energy bands for device b with the same parameters as ( c ) and ( e ) , respectively.,width=340 ] in the following numerical calculations , we choose the realistic material parameters @xmath30 that arrived from the hgte / cdte qws.@xcite the sample width ( or circumference ) is fixed to @xmath57 with the lattice constant @xmath58 . since the model is only valid in small @xmath59 , we confine the fermi energy within a small region near the @xmath60 point . in the presence of disorder , the conductance @xmath61 , conductance fluctuation @xmath62 , the local current @xmath49 etc . , are all averaged over up to 500 random configurations . ( a)(b ) and conductance fluctuation @xmath62 ( c)(d ) vs disorder strength w for different fermi energy @xmath63 and gap parameter m. ( a)(c ) for device a and ( b)(d ) for device b. the central region length @xmath64 . other parameters ( a , b , c , and b ) are the same as fig . 1(c)-1(f ) . , let us first compare the two devices in geometry and topology . in device @xmath29 with open boundary ( fig.1(a ) ) , there are two edges in the y direction . thus , the edge states can exist in such device . since the fig . 1(b ) is a cylinder without edge , the edge states are prohibited in the device b. in figures 1(c ) and 1(e ) , the band structures of clean hgte / cdte qws of are plotted . there exist a bulk energy gap approximately of @xmath65 in both figures . moreover , there are two degenerate bands ( edge states ) cross inside the gap for gap parameter @xmath66 ( fig . 1(c ) ) . in contrast , the crossing bands vanished when @xmath67 was tuned up to above zero ( fig . 1(d ) ) . these results are in agreement with those of previous studies@xcite . in figures 1(d ) and 1(f ) , we plot the corresponding band structures in the cylindrical geometry . for both gap parameter @xmath68 and @xmath69 , though the degeneracy is higher due to the enhanced symmetry , the bulk energy gap is nearly unchanged and there is only tiny shift in the energy bands . however , there is a big difference between the two samples , the edge states which cross inside the gap vanish in the later one . the phenomenon originates from the topology of the device , when the sample is sufficient large , the bulk states are hardly affected by the topology change at the edge while the edge states are totally destroyed . in conclusion , rolling the sample from a strip to a cylinder destroys the edge channels but maintain the bulk state properties . next , we investigate how the transport properties are affected by disorder . fig .2(a ) and 2(c ) plot the conductance @xmath61 and @xmath62 versus disorder strength for device a. when the system is in the inverted regime ( @xmath66 ) with the fermi energy inside the bulk gap ( @xmath70 ) , for a range of disorder strength @xmath71 , the two terminal conductance is quantized without much fluctuations . such observation agrees with the previous result that the qshe is robust against weak disorder @xcite . however , when the chemical potential is tuned up into the bulk band region near the @xmath60 point , no matter @xmath72 or @xmath66 , the new intriguing phenomena emerge . the conductance @xmath61 decreases while the fluctuation @xmath62 increases when the disorder is first applied . when the disorder strength continues to increase , instead of localization , the conductance begins to increase to a quantized value ( @xmath73 ) and maintains at this value for a certain range before eventually decreases . meanwhile , the conductance fluctuation @xmath62 decreases to zero and vanishes for the corresponding w. the anomalous conductance plateau indicates that the sample becomes a topological insulator . more importantly , different from the traditional topological insulator ( qhe etc ) , the quantized value is induced by the impurities . this disorder induced anomalous conductance plateau was discovered in a very recent work.@xcite naturally , there exists a question : what is the mechanism that causes such an anomalous plateau ? , m=-10mev , central region size @xmath74 under disorder strength w=0 ( a ) w=50mev(b ) w=110mev ( c ) w=220mev ( d ) . the inset in fig . 3(a ) is the schematic of local current flow vector . the vector direction represent the local current flow direction and the vector length is proportional to the logarithm of local current value.,width=340 ] to answer this question , we first examine @xmath61 and @xmath62 as a function of disorder strength for device b using the same parameters as device a [ see figs . 2(b ) and 2(d ) ] . unlike the device a , the transport properties follow the traditional anderson metal - insulator transition feature,@xcite and the anomalous conductance plateau is absent . for instance , the conductance monotonously decreases to zero with increase of the disorder strength and the critic disorder strength @xmath75 increases with raising of the fermi energy . significantly , the metal - insulator transition point for device b is roughly at the starting point of the anomalous plateau for device a. take @xmath76 for example [ see black solid line in fig . 2(a ) and 2(b ) ] , the anomalous plateau sets up at the disorder strength @xmath77 for device a and for that threshold value the device b becomes an insulator(@xmath78 ) . as shown in fig .1 , the bulk states in the device a and the device b are the same . only the edge states are completely destroyed by rolling the device a into the device b. this gives a direct evidence that the anomalous conductance plateau originates from the edge states . , fermi energy @xmath79 , and disorder strength w=0 ( a ) w=100mev ( b ) w=150mev ( c ) w=250mev ( d).,width=355 ] , fermi energy @xmath79 , and disorder strength w=0 ( a ) w=65mev ( b ) w=100mev ( c ) w=250mev ( d).,width=385 ] to get a better insight into the microscopic origin of the conductance variations , we examine the disorder effect through the local - current - vector - flow configurations . due to the time reversal symmetry , we only consider the spin up subsystem , the influence of spin down subsystem can be directly obtained by time reversal symmetry . here , the local current flow vector on site * i * is defined as @xmath80 . in fig.3 , the typical distributions of local currents for device a in traditional qshe region are plotted . for a clean sample ( see fig.3(a ) ) , the local currents locate mainly on the upper edge and their values decay exponentially towards the bulk . surprisingly , the small disorder initially makes the edge channel narrower ( see fig . though the mechanism is unclear , we note such phenomenon was also observed recently by chu _ et al_.@xcite in their paper , this narrowing effect was indirectly observed by the decreasing of the oscillation period of the a - b ring , while in this paper such effect is directly shown by the spatial distributions of the local currents . when the anderson disorder strength is getting larger , the local currents spread to the bulk and broaden the edge channels again(see as fig . however , only when the disorder strength exceeds the critical value @xmath75 , the spread local - current flow can reach the lower edge channels with different chirality , the effective backscattering ( as shown in the local current flow vector located near the lower edge in the region @xmath81 in fig .3(d ) ) can take place , leading to the reduction of the conductance between the two terminals . these pictures explain why the traditional quantized plateau is robust under weak disorder and how it is destroyed in the strong disorder limit . and @xmath57 at the dip point . ( b ) the position - related current @xmath82 vs longitude axis x. @xmath82 is defined as the summation of @xmath83 for four layers in the corresponding region . the other parameters are the same as ( a ) . ( c ) the two - terminal conductance g vs. disorder strength w with different sample lengths @xmath84 . , width=328 ] next , the fermi energy is tuned to @xmath79 , sitting slightly above the bulk gap . a positive gap parameter @xmath56 , for which there is no helical edge states inside the bulk gap for the clean sample , is chosen in the following simulations . the conductance @xmath61 vs. disorder strength @xmath71 shown in fig . 2(a ) can be classified to four regions ( i ) without disorder , ( ii ) before the anomalous plateau , ( iii ) on the anomalous plateau , and ( iv ) after the anomalous plateau . the typical configurations of local - current - flow vector in such four regions are plotted in fig.4 . for the clean sample ( see fig . 4(a ) ) , the local current not only flow forward along the upper edge , but also uniformly flow forward in the bulk . while @xmath71 is increased from zero into the region ( ii ) , the local currents in the bulk of the disorder region become smaller and more irregular in directions ( see fig . 4(b ) ) , which directly shows the decline of the bulk transport and the localization of the bulk states . fig .4(c ) demonstrates the distribution of the local currents when the device is in the region ( iii ) . the most interesting phenomena exist in the disorder region @xmath85 $ ] . the local currents in the bulk decline to zero while the residual currents flow in the upper edge and flow without much scattering . in addition , throughout the region ( iii ) , the bulk transport vanishes due to the localization of all the bulk states . in contrast , the edge transport shows the same behavior as in the traditional qshe region : the local currents deeply spread into the bulk without any effective backscattering with @xmath71 increasing until it connected the opposite edge channels(see as fig . 4(d ) ) . obviously , such edge transport will lead to the quantized plateau . in the negative gap parameter situation , for example @xmath76,(see fig . 5 ) the evolvement of local current configurations resembles that of @xmath72 situation . so far , the fig . 4(c ) and fig . 5(c ) are the most strong evidence directly showing that the tai is caused by the edge transport . in addition , these plots provide a vivid microscopic picture demonstrating the influence of disorder on tai . up to now , we have explained the majority phenomena emerged in the tai . nevertheless , the transport properties in the region ( ii ) of @xmath86 figure demands a detailed study for the following two reasons . ( i ) the conductance is not monotonously decreasing with increasing disorder strength but shows a dip feature prior to the anomalous plateau . obviously , at the dip point , the system is neither in tai phase nor in normal anderson insulator phase due to its nonzero and non - quantized conductance value . ( ii ) for both types of gap parameter @xmath72 and @xmath66 , the dip does exist prior to the anomalous plateau and the conductance @xmath61 behaves similarly after the dip . thus , revealing the cause for the dip feature may help us to understand the mechanism of the formation of the edge states . in fig.6(a ) , the distribution of the local currents with the disorder strength @xmath71 being fixed at the dip is plotted . the local currents flow are larger close to the edges than in the bulk . the predominant edge transport is clearly seen . in fig . 6(b ) we plot the position - related current @xmath82 versus longitude axis @xmath11 . where @xmath82 is defined as the summation of @xmath87 for four layers in the corresponding region . for example , given sample width @xmath57 , @xmath82 for the upper edge is defined as @xmath88 . with the help of @xmath82 , one can quantitatively analyze the local currents . on the whole , the behavior of the position - related current @xmath82 is similar to the local current configurations , but it is smoother with the position @xmath11 . more significantly , one can observe from fig . 6(b ) , the local currents flow from left to right with rapidly decreasing magnitude for upper edge , vice versa for the lower edge . this phenomenon can be attributed to the bulk state assisted backscattering between two edges . the fact that the position - related current @xmath82 in the bulk is small but nonzero indicates that the bulk states are not fully localized at the dip . the scattered carriers in the upper edge can hop through such delocalized bulk states to the lower edge which leads to the backscattering processes . in other words , disorder not only destroys the bulk transport but also quickly destroys the edge transport . because of this , the conductance @xmath61 is lower than plateau value at the dip point . for the region after dip ( see fig . 2(a ) ) , increasing of @xmath71 slowly destroys the bulk transport , but it also suppresses the bulk - assisted backscattering mechanism at the same time . it leads to an overall increase of the conductance @xmath61 . when the bulk states become fully localized , the anomalous plateau shows up . in fig.6(c ) , the conductance @xmath61 versus @xmath71 with different sample lengths @xmath84 are plotted . the dip feature is clearer for a large @xmath84 because of the increasing probability of the backscattering between the two edges . however , for all @xmath84 , the anomalous plateaus appear with the same disorder strength due to the fully localized bulk states . in summary , the disorder effect in hgte / cdte quantum wells is studied . we confirm the existence of the topological anderson insulator(tai ) phase . conductances calculated for the stripe and cylinder samples reveal the topological feature of tai and support the idea that helical edge states cause the anomalous quantized plateau . with the help of local - current - vector configurations for different chemical potentials and disorder strengths , the basic physical phenomena emerged in the normal qshe region and in the tai region are clearly understood . in particular , the analysis of the local current configurations provides us with the importance of the bulk - states - assisted backscattering in tai that in turn help us to understand the mechanism of the formation of the disorder induced edge states . the work is supported by nsfc under grant nos . 10525418 , 10734110 , and 10821403 , and by 973 program project xcx is supported by us - doe and oklahoma c - spin center . we thank j. k. jain , s.q . shen , xuele liu and waigen zhang for helpful discussions . part of the calculations was performed on the hpcc clusters at osu . haijun zhang , chao - xing liu , xiao - liang qi , xi dai , zhong fang , s. c. zhang , arxiv : 0812.1622v1 ; accept by nature physics ; y. xia , l. wray , d. qian , d. hsieh , a. pal , h lin , a. bansil , d. grauer , y. s. hor , r. j. cave , and m. z. hasan , arxiv:0812.2078v1 . m. k@xmath89nig , steffen wiedmann , christoph bruene , andreas roth , hartmut buhmann , laurens w. molenkamp , xiao - liang qi and s. c. zhang science * 318 * , 766 ( 2007);m . k@xmath89nig , hartmut buhmann , laurens w. molenkamp , t. l. hughes , chao - xing liu , xiao - liang qi and s. c. zhang j. phys . . jpn . * 77 * , 031007 ( 2008 ) .	we study the disorder effect on the transport properties in the hgte / cdte semiconductor quantum wells . we confirm that at a moderate disorder strength , the initially un - quantized two terminal conductance becomes quantized , and the system makes a transition to the novel topological anderson insulator ( tai ) . conductances calculated for the stripe and cylinder samples reveal the topological feature of tai and supports the idea that the helical edge states may cause the anomalous quantized plateaus . the influence of disorder is studied by calculating the distributions of local currents . base on the above - mentioned picture , the phenomena induced by disorder in the quantum spin hall region and tai region are directly explained . our study of the local current configurations shed further light on the mechanism of the anomalous plateau .
the numerical simulation of the schrdinger equation and the visualization of its solutions has become an important part of quantum - mechanical education on all levels @xcite . but , it is almost unavoidable to stumble accross strange phenomena , when one attempts a numerical solution of the dirac equation . in this article we show visualizations of some of these peculiarities and try to give an informal explanation . the dirac equation is the fundamental equation for relativistic quantum mechanics . hence , it belongs to the most important equations in modern physics . among its big successes is the very accurate description of the energy levels of the hydrogen atom . on the other hand , the occurrence of several unexplained paradoxes casts doubts on its status and interpretation . the unexpected behavior occurs even for innocent looking initial conditions . a canonical set of initial conditions for the time - dependent schrdinger or dirac equations is given by the set of gaussian wave packets . they describe more or less localized quantum states for which the product of the uncertainties in position and momentum is minimal . on the other hand , the set of gaussian initial conditions covers most cases of practical interest , because any wave packet can be approximated by a superposition of a finite number of gaussian states . the motion of gaussian wave packets according to the one - dimensional free schrdinger equation shows little surprises . the example in fig . [ fig : g ] shows a nonrelativistic gaussian wave packet with average momentum zero . initially , the wave packet is well localized , but it spreads during the time evolution . as the wave packet gets smeared out , its height decreases , because its norm as a square - integrable function must remain constant . the wave packet remains a gaussian function for all times . spreading of a gaussian wave packet according to the schrdinger equation.,width=230 ] time evolution of a gaussian initial wave packet according to the dirac equation.,width=230 ] according to the free schrdinger equation , the average position @xmath0 and the average momentum @xmath1 of the wave packet obey the rules of classical mechanics . moreover , the spreading of the wave packet is independent of the average velocity of the wave packet . this spreading of the position distribution would also be observed for a cloud of classical particles whose density in position space is a gaussian function , provided that the momenta of the particles also have a gaussian distribution . [ fig : a ] shows a numerical solution of the one - dimensional free dirac equation . it has the same gaussian initial distribution as the nonrelativistic wave packet in fig . [ fig : g ] , yet its behavior is quite different . the relativistic wave packet wiggles back and forth , becomes non - gaussian for @xmath2 , and soon develops characteristic ripples . this result is so strange that anybody with some experience with quantum mechanics ( but not with the dirac equation ) would first assume that the numerical method is at fault . this is a good example supporting the argument that the numerical solution of an equation is rarely sufficient to understand a phenomenon . this motivates a more careful theoretical analysis in order to understand the origin of this strange behavior . this article contains several black - and - white images of solutions of the dirac equation . computer - generated animations showing the whole time evolution can be found on the internet @xcite . these movies show both components of the solution with a color code for the phase of complex numbers . thus they reveal more information than the simple black - and - white reproductions of the position probability density in this article . the forthcoming book @xcite will be accompanied by a cd - rom containing a large collection of similar movies , animations , and simulations . they illustrate and supplement the theoretical exposition in order to provide students with an intuitive understanding that is hard to achieve by studying mathematical formulas . see the web site @xcite for more details and the ideas behind the project visual quantum mechanics . in this article , we discuss the time - dependent free dirac equation in one space dimension . we write it as an evolution equation in the familiar `` schrdinger form '' @xmath3 the free dirac hamiltonian @xmath4 is the matrix - differential operator @xmath5 where @xmath6 and @xmath7 are the famous pauli matrices , and @xmath8 . the expression for @xmath4 can be interpreted as a linearization of the relativistic energy - momentum relation @xmath9 the square of the dirac operator is just given by @xmath10 for numerical computations and for the visualizations it is advantageous to use units where @xmath11 . these can be obtained from the si - units by a simple scaling transformation . hence , in the following we use dirac equation in the form @xmath12 instead of @xmath6 and @xmath7 , we could use any other pair of pauli matrices . this would give a unitarily equivalent formulation . all images in this article would remain unchanged . the phenomena explained here also occur in higher dimensions , but the one - dimensional situation is much easier to visualize . for the dirac hamiltonian in three space - dimensions , pauli matrices are not sufficient ; @xmath13-dirac matrices are needed instead . the book @xcite contains more information about the dirac equation in three dimensions . a suitable state - space for the solutions of the dirac equation must consist of vector - valued functions @xmath14 because the operator @xmath4 is a two - by - two matrix . these two - component wave functions are usually called dirac spinors . the word `` spinor '' might appear inappropriate . in one dimension , all magnetic fields that could possibly affect the spin are pure gauge fields . hence , the two components of do not describe the spin . rather , they reflect the appearance of negative energies . ( however , the doubling of the spinor components in three dimensions is caused by the spin . ) the free dirac hamiltonian in momentum space is the matrix @xmath15 which for each @xmath16 has the eigenvalues @xmath17 and @xmath18 . as the hamiltonian represents the energy of the quantum system , one is led to the conclusion that the possible values of the energy of a free particle can be positive or negative . correspondingly , the dirac equation has two types of plane - wave solutions , which we denote by @xmath19 and @xmath20 . for each @xmath16 , @xmath21 where @xmath22 and @xmath23 are eigenvectors of the matrix @xmath24 belonging to the eigenvalues @xmath25 and @xmath18 , respectively . hence , we have @xmath26 and therefore @xmath19 ( @xmath20 ) is called a plane wave with positive ( negative ) energy . in order to be in line with the formalism of quantum mechanics , one usually requires that ( for each @xmath27 ) the components of a dirac - spinor be square - integrable , @xmath28 one should be aware that this mathematical requirement is closely related to the interpretation of the solutions . our choice makes perfect sense if we agree with the following tentative interpretation of relativistic wave functions : suppose @xmath29 is a normalized dirac spinor , then @xmath30 is the probability to find the particle in the interval @xmath31 . here `` normalized '' means that the above integral from @xmath32 to @xmath33 is equal to @xmath34 . similarly , and in complete analogy to nonrelativistic quantum mechanics , @xmath35 is interpreted as the probability to find the momentum of the particle in the interval from @xmath36 to @xmath37 . ( the hat denotes the fourier transform with respect to @xmath38 . ) this interpretation is consistent with the choice of @xmath39 ( the generator of spatial translations ) as the momentum operator . the unitarity of the time - evolution generated by the dirac equation guarantees that the normalization of dirac spinors is time - independent . hence , if there is one particle at the beginning , there will be one particle at the end . this remains true for arbitrary external fields . contrary to a common belief , the dirac equation is not able to describe pair production within the framework of quantum mechanics . for this one needs the formalism of quantum field theory where many - particle wave functions are built from one - particle solutions ( and from one - antiparticle solutions , which are related to negative - energy solutions ) . square - integrable wave packets can be obtained from the plane waves by superposition . this is the same procedure that works for the schrdinger equation . with suitable coefficient functions , any square - integrable solution of the dirac equation can be written in the form @xmath40 the coefficient functions @xmath41 and @xmath42 can be determined from the fourier transform @xmath43 of the initial function @xmath44 by a projection onto the positive or negative energy subspace : @xmath45 numerical integration of provides a method to compute the free time evolution of an arbitrary initial function that can be a useful alternative to a finite difference scheme . more about the mathematics of the one - dimensional dirac equation can be found in @xcite . for many reasons , including some of the observations in this article , the interpretation given above may be critisized . therefore , one should consider this interpretation merely as a convenient working hypothesis that allows us to apply the well - established mathematical formalism of quantum mechanics . in this section , we describe and visualize three strange phenomena shown by solutions of the free dirac equation . some explanations will be given in the following sections . as a first example , we compute the free time evolution of the gaussian dirac spinor @xmath46 a few snapshots of the solution are shown in fig . [ fig : a ] . this image shows the position probability density @xmath47 according to the interpretation given above . compared to the simple spreading of the nonrelativistic gaussian , this wave packet shows a very complicated motion . space - time diagram and worldline of the position mean value for the wave packet in fig . [ fig : a].,width=230 ] the choice of the numerical constants in is motivated by the following consideration . the fourier transform of the wave packet is @xmath48 it describes a momentum distribution that is well localized in the interval @xmath49 $ ] . this corresponds to a maximal speed less than @xmath50 , far below the speed of light @xmath51 . hence , we can be sure that all observations have little to do with relativistic effects due to velocities approaching the speed of light . [ fig : aworld ] shows a space - time diagram of this solution . here , the density plot visualizes the position probability density as a function of the space coordinate @xmath38 and the time coordinate @xmath27 in dimensionless units . the white curve shows the world line of the average position , which obviously does not obey classical ( relativistic ) kinematics . instead , the expectation value of the position operator performs a wiggling motion , commonly known as `` zitterbewegung '' @xcite . apart from the rapid oscillation , the wave packet drifts slowly to the right , although its average momentum is zero . nevertheless , it turns out that the momentum distribution ( as determined from the fourier transform of the wave packet ) is still a conserved quantity with average momentum zero . the second example , fig . [ fig : b ] shows the free time evolution of the initial spinor @xmath52 time evolution of a gaussian initial wave packet with positive momentum.,width=230 ] this spinor is very similar to . in position space , it is multiplied by a phase factor @xmath53 . in momentum space , this means a translation by @xmath54 . the momentum distribution belonging to is therefore a gaussian distribution centered around the average momentum @xmath54 . moreover , the momentum distribution is so narrow that all momenta contributing significantly to the wave packet are positive . nevertheless , the solution splits into two parts , and the smaller part moves to the left ( that is , with a negative velocity ) . space - time diagram and worldline of the position mean value for the wave packet in fig . [ fig : b].,width=230 ] the space - time diagram of this solution , shown in fig . [ fig : bworld ] , again shows the zitterbewegung of the position s mean value . but this time , the oscillation quickly fades away . we see that zitterbewegung is sustained only as long as the left - moving part and the right - moving part of the wave packet have some overlap in position space . the third example is shown in fig . [ fig : c ] . it realizes a wave packet with positive velocities . the initial wave packet was obtained as a superposition of a positive- and a negative - energy part : @xmath55 the positive - energy part has positive momentum . in momentum space , it is defined as @xmath56 the negative - energy part has negative momentum , @xmath57 in position space , both parts obviously move in the same direction . they interfere with each other , thereby causing the ripples in the position distribution . time evolution of a wave packet with positive velocity.,width=230 ] space - time diagram and worldline of the position mean value for the wave packet in fig . [ fig : c].,width=230 ] fig . [ fig : cworld ] shows a space - time diagram of this motion . there is no zitterbewegung at all ; the world line of the average position is a straight line . interestingly , the peaks of the wave packet move with superluminal speed . this is strange because , at least in principle , local variations of the position probability density are observable . let us now proceed with a theoretical analysis . the average velocity of the wave packet shows a slow drift to the right , which is clearly visible in fig . [ fig : aworld ] . it is remarkable that the solution is not symmetric with respect to reflections at the origin , although the initial condition is . indeed , the dirac equation is not invariant under the replacement @xmath58 but the replacement @xmath59 in the wave function does not describe the physical space reflection . like any lorentz transformation , the space reflection has a part that acts on the components of the wave function . the correct way to describe the space reflection in the hilbert space of the dirac equation is the parity transform @xmath60 hence , the wave packet is not invariant under a parity transform , @xmath61 is not a scalar multiple of @xmath29 . an example of a parity invariant solution is provided by the gaussian initial wave packet @xmath62 which has only an upper component and satisfies @xmath63 for all times . [ fig : dworld ] shows a space - time diagram of the position probability density for this solution . the average velocity is zero , and the wave packet shows no zitterbewegung in the expectation value although the position probability density shows ripples similar to the first solution . space - time diagram and worldline of the position mean value for the wave packet .,width=230 ] the average velocity of a wave packet is described by the classical velocity operator @xmath64 it corresponds to the definition of the velocity in terms of momentum and energy , which is familiar from classical relativistic mechanics . note that this relation between velocity and momentum depends , in particular , on the sign of the energy . for a wave packet with negative energy , a positive momentum @xmath65 thus corresponds to a negative average velocity @xmath66 . this is also reflected by the phase velocity of the plane waves . the phase velocity of the plane waves is @xmath67 note that the phase velocity is always faster than the velocity of light , and even tends to infinity in the limit of small momenta . this does not matter , because no information can be transmitted with phase velocity . actually , a plane wave is spread over all of space - time and all information that is carried by a plane wave is already everywhere . the group velocity of relativistic wave packets is always slower than or equal to the velocity of light . but the sign of the phase velocity carries over to the sign of the group velocity . momentum distributions of the positive and negative energy parts of the wave packet in fig . [ fig : amom].,width=172 ] for the wave packet , fig . [ fig : amom ] shows the momentum distributions of the parts with positive and negative energy , that is , the functions @xmath68 and @xmath69 , according to . we see that the positive - energy part has its momentum distribution slightly shifted towards positive momenta , whereas the negative - energy part has a negative average momentum . for a positive - energy wave packet , a positive average momentum means a positive average velocity , as usual . but , for a negative - energy wave packet , the negative average momentum corresponds to a positive average velocity . hence , the whole wave packet has a positive average velocity , which can be seen clearly in fig . [ fig : aworld ] . the opposite direction of velocity and momentum in the negative - energy part of a wave packet immediately explains the behavior of the wave packet in fig . [ fig : b ] . for this wave packet , the momentum distributions of the parts with positive and negative energy are shown in fig . [ fig : bmom ] . we see that both parts consist of positive momenta . for the smaller negative - energy part , this corresponds to a negative velocity . hence , the wave packet in position space is a superposition of an approximately gaussian wave packet with positive energy moving to the right , and a smaller part with negative energy moving to the left ( also with a positive average momentum ) . momentum distributions of the positive and negative energy parts of the wave packet in fig . [ fig : b].,width=172 ] in case of the third example , the negative - energy part has negative momenta , as shown in fig . [ fig : cmom ] . here the two parts are of the same size and they have opposite momenta , they have the same average velocity . hence , the parts with positive and negative energy move together in position space . the opposite phase - directions of the two parts cause the interference effects . at the average momentum @xmath70 , the initial state contains the superposition of the plane waves @xmath71 . in both components of the spinor , this is proportional to a superposition of @xmath72 and @xmath73 . this , precisely , causes the sine - like distortions of the gaussian position probability density visible in fig . [ fig : c ] . momentum distributions of the positive and negative energy parts of the wave packet in fig . [ fig : c].,width=201 ] hence , the interference ripples have their origin in the phase interference of the parts with positive and negative energy . as the phase moves with a superluminal average velocity @xmath74 , also the interference pattern moves with superluminal speed . because it is just an interference pattern , it carries no information from one point in space to another . hence , despite the fact that local variations in the position probability density are , in principle , observable , the fact that these variations move at superluminal speed is no contradiction to the special theory of relativity . a classical example for this effect is the apparent motion of moir patterns . if two periodic patterns with slightly different periodicity lengths are slowly shifted with respect to each other , the pattern of the superposition ( the moir pattern , which is locally observable ) can move with almost arbitrary speed . in fact the recently discussed phenomenon of superluminal tunneling also consists in the motion of an interference pattern of plane waves ( thereby the transmitted wave packet changes its shape , but no information is being transmitted with superluminal speed ) . a local perturbation of the interference pattern can only propagate with a velocity less than @xmath75 . the zitterbewegung ( that is , the shape of the white curve in the space - time diagram ) is described by the expectation value ( in the given initial state ) of the operator ( see , for example , @xcite ) @xmath76 here @xmath66 is the classical velocity operator , and @xmath77 describes the oscillation , @xmath78 note that @xmath79 is the instantaneous velocity according to the dirac equation , because the time derivative of the position in the heisenberg picture is just @xmath80 = c\,\sigma_1 . \label{e : standard velocity operator}\ ] ] the initial wave packets and are eigenvectors of @xmath6 , belonging to the eigenvalue @xmath81 . hence the initial velocity of these wave packets is @xmath82 , hence in fig . [ fig : aworld ] and fig . [ fig : bworld ] , the slope of the white curve at @xmath83 is @xmath84 degrees . in momentum space , @xmath77 is just a multiplication by a matrix - valued function of @xmath65 . the operator @xmath77 anticommutes with @xmath4 . this means that @xmath77 maps a state with positive energy onto a state with negative energy . suppose that @xmath85 is , for example , a positive - energy state . then @xmath86 has negative - energy and is thus orthogonal to @xmath85 . hence , the scalar product @xmath87 must be zero . the expectation value @xmath87 can only be nonzero for wave packets having both positive- and negative - energy parts . wave packets located in different regions of momentum or position space are orthogonal . the operator @xmath77 is a multiplication operator in momentum space which means that it does not change the localization properties of a wave packet in momentum space . hence , zitterbewegung can only be significant , if the function @xmath88 has a significant overlap with the function @xmath89 in momentum space . in position space , the operator @xmath77 is nonlocal , but it turns out , that it does not change the approximate localization of a wave packet all too much . hence , if a wave packet @xmath90 is ( approximately ) located in some region @xmath91 of position space , then the function @xmath92 is approximately located in a neighborhood of that region . as a consequence , zitterbewegung is only significant as long as the parts with positive and negative energy are close to each other in position space . now , let us consider our examples . [ fig : amom ] shows that the momentum distributions of the positive and negative energy parts of the initial wave packet have a significant overlap in momentum space . this fact does not change with time , because the momentum ( and hence the momentum distribution of the initial wave packet ) is a conserved quantity according to the free dirac equation . both parts of the initial wave packet have average momenta close to zero , corresponding to average ( classical ) velocities close to zero . hence , the corresponding parts will remain close together also in position space . hence , for this wave packet , zitterbewegung is a sustained phenomenon . ( actually , the amplitude of the zitterbewegung vanishes very slowly , as @xmath93 . a mathematical argument for the asymptotic decay of zitterbewegung is given in @xcite . ) for the wave packet in fig . [ fig : b ] , the momentum distributions of the positive and negative energy parts overlap completely ( see fig . [ fig : bmom ] ) . hence , we can indeed observe zitterbewegung , but only as long as the two parts occupy approximately the same region in position space . because the two parts have opposite velocities , they quickly separate , and the amplitude of the zitterbewegung decreases rapidly . for the wave packet in fig . [ fig : c ] , the parts with positive and negative energy move together in position space , but there is no zitterbewegung , because these parts are located in different regions of momentum space . the interference effects occur only for those solutions of the dirac equation that are composed of parts with positive and parts with negative energy . the origin for the ripples in the position distribution lies in the fact that the parts with positive and negative energy have opposite phase directions . because any square - integrable spinor - valued function ( in particular , a gaussian spinor ) is , in general , a superposition of positive and negative energies , one is likely to run into these phenomena when one computes a numerical solution of the dirac equation . one may argue that superpositions of positive and negative energy states are not physically observable . a wave packet with only one sign of energies shows no zitterbewegung and behaves reasonably . indeed , in quantum field theory , one constructs a hilbert space of many - particle electron states from the positive - energy solutions of the one - particle dirac equation , and the positronic states are built from charge - conjugated negative - energy solutions . a problem with this approach is that the electronic and the positronic states can not always be separated in the presence of external fields .	this article investigates some solutions of the time - dependent free dirac equation . visualizations of these solutions immediately reveal strange phenomena which are caused by the interference of positive- and negative - energy waves . the effects discussed here include the _ zitterbewegung _ , the opposite direction of momentum and velocity in negative - energy wave packets , and the superluminal propagation of the wave packet s local maxima .
the successful comparison of hydrodynamical models to a wide range of data from heavy ion collisions suggests that the produced quark - gluon fluid is a viscous fluid with perhaps the lowest known viscosity to entropy density ratio ( @xmath0 ) in nature . the low values of @xmath0 , coupled with the fact that these hydrodynamical models employ an equation of state , appear to indicate that the matter is thermal , or at least nearly isotropic , with the ratio of the longitudinal pressure to the transverse pressure close to unity . further , to reproduce key features of the data , it appears important that hydrodynamics be applicable at very early times of less than a fermi after the collision . there is some elasticity to the above conclusions , and it is conceivable that the hydrodynamic paradigm may be modified . nevertheless , the phenomenology is sufficiently robust to approach seriously . from a theoretical perspective , at first glance , it seems astonishing that hydrodynamics is applicable at all to such small systems , and at such early times . hydrodynamics is an excellent effective field theory of qcd , but for long wavelength modes and at late times @xcite . in kinetic theory frameworks , the scattering rates of quarks and gluons have to be sufficiently strong to counter the rapid longitudinal expansion of the system . this appears challenging . from these elementary , and perhaps naive considerations , to paraphrase a quote by wigner in another context , hydrodynamics is unreasonably effective " . a weak link in hydrodynamic models is the statement that the system isotropizes / thermalizes at very early times . most hydrodynamic models choose an initial time @xmath1 fm . nearly all these models ignore the pre - equilibrium dynamics prior to this time . the one model that does incorporate pre - equilibrium dynamics , the ip - glasma model , does so imperfectly @xcite . there is therefore a practical problem of how and when pre - equilibrium dynamics can be matched on to a hydrodynamic description . this of course is tied to resolving the more general conceptual problem of how thermalization occurs in qcd . the latter will be one of the two subjects of the discussion here . even if thermalization does occur in heavy ion collisions , as the hydrodynamic models suggest , there is the interesting question of whether this framework is applicable to smaller size systems . how and where does hydrodynamics break down ? does it apply to p+a and p+p collisions , as some interpretations of data ( on long range rapidity correlations ) in these collisions suggest ? if it works for high multiplicities , at what multiplicities do we see an onset of hydrodynamic behavior ? are there alternative explanations for what is seen in the data ? the interpretation of long range rapidity correlations in p+p and p+a collisions will be the other topic discussed here . both topics will be discussed within weak coupling frameworks here . it is a common misunderstanding that weak coupling implies weakly interacting . that is not the case for systems with high occupancy . it is a legitimate question to ask whether weak coupling is the right framework for heavy ion collisions at rhic and lhc , the coupling is not particularly weak . at one level , an answer is that this is the only framework we know how to compute in systematically and reliably . but this answer is also profoundly unsatisfying . a better answer is that weak coupling frameworks describe many non - trivial features of heavy ion collisions . it is however not a universal panacea , which disappoints some people , but that ca nt be helped until some smart person solves qcd . the lack of a satisfactory framework to address dynamical aspects of qcd in strong coupling is a powerful motivation for ads / cft duality inspired frameworks . the questions regarding the applicability of these methods to heavy ion collisions are well known , and i will not revisit them here . the next section will discuss the problem of thermalization in weak coupling . we will then discuss the recent results from p+p and p+a collisions on collimated long range rapidity correlations ( the ridge " ) . since many of the issues discussed are open , and are the subject of much debate , conclusions may be premature . i will conclude instead with some personal reminiscences of gerry brown , whose early mentorship made it possible for me , however imperfectly , to tackle these issues . multiparticle production at central rapidities is dominated by gluon configurations carrying small fractions @xmath2 of the momenta of the colliding nuclei . perturbative qcd ( pqcd ) predicts , and data from hera confirm , that the occupancy of small @xmath2 modes in a proton is large for fixed momentum transfer @xmath3 . the occupancy saturates at @xmath4 for any given @xmath5 , for sufficiently small @xmath2 . this phenomenon is called gluon saturation , and the saturation scale @xmath6 is the dynamical ( semi- ) hard scale at which maximal occupancy is attained for a given @xmath2 @xcite . a small @xmath2 probe with @xmath7 , will see a nearly perfectly absorptive black disk of gluons . the color glass condensate ( cgc ) is a classical effective field theory of such highly occupied gluon configurations @xcite . systematic computations ( lo , nlo , ... ) are feasible in this framework . an ever present issue is factorization : what gluon modes are universal resolved when the nuclei are coarse grained at different resolutions and what modes participate in the scattering process under consideration . factorization in nucleus - nucleus collisions has been proven to leading logarithmic accuracy ; in plainspeak , the largest logarithmically divergent terms in @xmath2 , at each order in perturbation theory , can be resummed into nuclear wavefunctionals that can be probed through various final states . in nucleus - nucleus collisions , the leading order cgc result for components of the stress energy tensor @xmath8 is given by solutions of yang - mills equations for two lightlike classical sources ( representing the large @xmath2 static color charges of the two nuclei ) with null initial conditions for fields at negative infinity @xcite . the numerical solution to this problem is known @xcite . on the ( proper ) time scale @xmath9 , the system is highly anisotropic : @xmath10 and @xmath11 , where @xmath12 is the energy density , @xmath13 the transverse pressures in the @xmath2 and @xmath14 directions orthogonal to the beam , and @xmath15 is the longitudinal pressure of the expanding nuclei after the collision . this is straightforwardly understood to be a consequence of the classical field solutions @xmath16 being independent of the space - time rapidity @xmath17 . this non - equilibrium matter , often called the glasma @xcite , is far from isotropy by @xmath18 ; the former is the saturation scale in the nuclear wavefunctions , the latter is the hard scale in the glasma . while the latter derives from the former , they do not have to be identical . ] at @xmath19 . the large @xmath20 modes are frozen out , with distributions given by leading order pqcd expressions . interestingly , the softer @xmath21 modes are described by a 2-d classical bose - distribution @xmath22 @xcite . what happens at nlo ? parametrically small quantum fluctuations ( o(1 ) relative to the leading @xmath23 classical field ) can grow exponentially as @xmath24 @xcite this is due to an instability of the small fluctuation propagator in the classical field , an instability that has been interpreted previously in the language of weibel @xcite and nielsen - olesen @xcite instabilities . so , if we were to try and do a systematic computation , @xmath25 , we would find terms at nlo that would be as large as the lo contribution at times @xmath26 , indicating a breakdown of the perturbative expansion at these times . note that this power counting makes sense only in very weak coupling ; for realistic values of @xmath27 relevant for experiments , the @xmath28 is not appreciably different from @xmath29 . just as in the case of small @xmath2 resummations , the leading @xmath30 at each nth sub - leading order in perturbation theory can be resummed and written as @xcite @xmath31\;f_0 [ \alpha ] t_{\rm lo}^{\mu\nu } [ a_{\rm class . } + \alpha](x)\ , , \label{eq : em - tensor}\ ] ] where @xmath32 = \exp\left ( - \frac{1}{2 } \int_{\tau = 0^+ } d^3 u d^3 v \alpha(u ) \gamma^{-1}(u , v ) \alpha ( v)\right ) \ , , \ ] ] denotes the spectrum of fluctuations , and @xmath33 , the small fluctuations propagator in the classical background field at proper time @xmath34 after the collision . to compute @xmath35 , it is sufficient to compute solutions of the small fluctuation equations of motion ( initialized as plane waves at negative infinity ) through the field strengths of one of the nuclei and into the forward light cone at @xmath34 after the collision . this is quite challenging but precisely such a computation was performed recently @xcite . computing eq . ( [ eq : em - tensor ] ) is equivalent to solving the 3 + 1-d classical yang - mills equations with the initial condition @xmath36 \ , . \label{eq : quantum}\ ] ] the coefficients @xmath37 , where @xmath38 collectively denote quantum numbers labeling the basis of solutions , are random gaussian - distributed complex numbers . the explicit expressions for the small fluctuations and their conjugate momenta , denoted here by @xmath39 , are given in @xcite . while the computations in @xcite may be important for a variety of final states sensitive to very early times after the collision , their relevance for the problem of thermalization in heavy ion collisions is unclear . firstly , the resummation in eq . ( [ eq : em - tensor ] ) will receive sub - leading corrections @xmath40 with @xmath41 . these all become of order unity on time scales that are parametrically @xmath42 ; this suggests that the resummation procedure may be fragile when extended to later times . secondly , it has been argued that an effective memory loss " of initial conditions may occur when oscillating functions , each corresponding to a gaussian distributed initial condition of varying amplitude , are summed over @xcite . more generally , we argued that the ` late - time ' dynamics of the glasma is described by a universal nonthermal fixed point , the properties of which are independent of the initial conditions @xcite . the picture we have developed is as follows . by timescales @xmath43 , phase decoherence sets in and the classical fields , in @xmath44 gauge , can be decomposed into quasi - particle modes described by single particle distributions , as @xmath45\ ; , \end{aligned}\ ] ] and a corresponding expression for the canonically conjugate momenta . here @xmath46 denote the ( time dependent ) transverse polarization vectors of modes with transverse momentum @xmath47 , rapidity wave number @xmath48 and polarization index @xmath49 in the non - interacting theory and @xmath50 denotes complex conjugation . the classical - statistical ensemble is defined by the distribution of the coefficients @xmath51 , which are the complex gaussian random numbers introduced previously . the single particle distributions are parametrized as @xmath52 . if the dynamics is faithfully captured by quasi - particle excitations for @xmath53 , details of the evolution from earlier times @xmath54 up to @xmath55 are subsumed in the initial occupancy @xmath56 , and in the anisotropy of the initial momentum distribution @xmath57 . with the initial conditions of eq . ( [ eq : quantum2 ] ) for the gauge fields , and likewise , an initial condition for canonical conjugate momenta , the 3 + 1-d yang - mills equations can be solved for longitudinally expanding non - abelian plasmas . to ensure the dynamics is classical for the entire lifetime of the evolution , the amplitude of the gauge fields has to be chosen to be much larger than those obtained by replacing @xmath58 by the quantum ` @xmath59 ' . for the results of the simulations discussed here , we have chosen @xmath60 corresponding to @xmath61 . we will comment shortly on the extrapolation of these results to more realistic values of @xmath27 . we find strikingly that the classical - statistical evolution of the system at late times demonstrates self - similar behavior independent of the initial conditions . specifically , the single particle distributions extracted from the numerical simulations at later times by inverting eq . ( [ eq : quantum2 ] ) , and the corresponding expression for canonically conjuage momenta , satisfy @xmath62 where @xmath63 denotes a _ stationary _ distribution in time , that describes the spectral properties of the non - thermal fixed point . this is seen for a wide range of @xmath56 and @xmath57 values , and has been checked to hold for initialization times ranging from @xmath64 and @xmath65 . in each case , regardless of the initial condition , the system flows to the same attractor solution by @xmath66 . the scaling exponents @xmath67 and @xmath68 describe the temporal evolution of the characteristic hard longitudinal and transverse momentum scales respectively : @xmath69 one can therefore equivalently write @xmath70 . the scaling exponent @xmath71 describes the overall decrease of the distribution amplitude in time . the hard scales in eq . ( [ eq : scales ] ) are gauge invariant quantities @xcite that can be measured on the lattice ; @xmath68 , @xmath67 and the combination @xmath72 can be extracted from the logarithmic derivatives of @xmath73 , @xmath74 , and the energy density respectively . very careful measurements , on the largest lattices to date , have been performed and one extracts @xcite @xmath75 , @xmath76 and @xmath77 . the scaling exponents can also be extracted from moments of the single particle distribution in eq . ( [ eq : scalingf ] ) ; within systematic uncertainties , they are consistent with the values quoted here . the decrease in the occupancy as @xmath78 implies that the system that began with an occupancy of @xmath79 will reach an occupancy of unity at time scales @xmath80 . the classical simulations will no longer be valid after this time scale . however the parametrically large window of @xmath81 , where these simulations are valid , provides insight into the rich dynamics of the thermalization process . the transverse momentum of hard gluons does not change during this time ; their occupancy only decreases due to the redshift in the longitudinal momentum ; they are distributed as @xmath82 , with @xmath83 , very early in the evolution . if the system were to free - stream , the ratio @xmath84 would decrease as @xmath85 ; instead , the strong scattering of gluons prevents this , and @xmath86 instead . while the anisotropy does increase , the lowest value of the ratio @xmath84 is @xmath87 at the end of the classical regime . when occupancies are less than @xmath79 , but greater than unity , there exists a dual description of the system in terms of either classical fields or kinetic theory @xcite . the latter framework however is sensitive to dynamics at the debye screening scale that is non - perturbative ; this sensitivity allows for a number of kinetic scenarios , including the possible late time role of plasma instabilities @xcite or transient bose - einstein condensates @xcite . instead , we find that our numerical simulations are consistent with the late stage classical dynamics of the bottom - up " thermalization scenario @xcite . in this kinetic scenario , the dynamics of overoccupied gluons is dominated by small angle elastic scattering . the self - similar scaling behavior extracted from both gauge invariant and gauge fixed observables in the classical - statistical field simulations of the expanding non - abelian plasma finds a simple a posteriori explanation in the context of wave turbulence . such analyses were performed previously in scalar field theories that are used to describe the thermalization process in the post - inflationary early universe @xcite . in close analogy to this earlier work , the self - similar behavior characteristic of wave turbulence in the gauge theory can be interpreted in the framework of a kinetic equation @xmath88f(p_t , p_z,\tau)=c[p_t , p_z,\tau;f ] \;,\end{aligned}\ ] ] for the single particle distribution @xmath89 with a generic collision term @xmath90 $ ] for @xmath91 scattering processes . for the self - similar distribution in eq . ( [ eq : scalingf ] ) , the scaling behavior of the collision integral @xmath90 = ( q \tau)^{\mu}\ , c[(q \tau)^{\beta}p_t,(q \tau)^{\gamma}p_z;f_s]$ ] , is described in terms of the exponent @xmath92 , the precise form of which depends on the underlying interaction . substituting this scaling function into eq . ( [ eq : boltzmann ] ) leads to the _ time - independent condition _ @xmath93\;. \end{aligned}\ ] ] the non - thermal attractor solution observed in our lattice simulations is a nontrivial solution of this kinetic equation . such scaling analyses of kinetic equations , demonstrate that the scaling exponents characterizing the attractor can be classified on very general grounds of dimensionality , conservation laws and boundary conditions for the evolution @xcite . the observation of such a self - similar scaling solution of the expanding non - abelian plasma is a powerful indication for universal behavior far from equilibrium , as seen in dynamical systems ranging from the inflaton dynamics of relativistic scalar fields to table - top cold atom experiments @xcite . thus while the bottom - up scenario correctly captures the universal properties of the turbulent thermalization process , the phenomenon may be more general than indicated by the specifics of the bottom - up scenario . preliminary results from simulations of longitudinally expanding _ scalar _ fields appear to bolster these conclusions @xcite . what then of isotropization and thermalization ? though the system does not free - stream and is strongly interacting , it does not isotropize either in the classical regime where the simulations are valid . in the bottom - up kinetic scenario , which matches our results in the classical regime , the system can be followed all the way to thermalization . in the quantum regime ( where @xmath94 ) , @xmath95 inelastic interactions become important . for @xmath96 , soft gluons produced through inelastic @xmath95 interactions contribute very little to the total number but begin to dominate debye screening . the latter begins to influence dynamics . in this regime , one finds that @xmath97 , independent of time , and @xmath98 . the transverse pressure is still dominated by the hard gluons , and continues to have the parametric form @xmath99 ; the longitudinal pressure , on the other hand , is given by @xmath100 . the ratio of the two then gives @xmath101it thus begins to rise again in the quantum regime . by @xmath102 , the number of soft gluons exceeds the number of hard gluons . the soft gluons collide frequently the rate of expansion for @xmath103 is slower than the relaxation time of soft gluons . the soft sector has therefore thermalized by @xmath104 . the system is however anistropic as a whole because of the remaining gluons in the hard sector . it has recently been conjectured in the context of jet quenching that hard gluons in the quark - gluon plasma lose their energy via a medium induced self - similar turbulent cascade of soft gluons @xcite a quantum turbulent attractor . this framework should also be applicable to describe how the remaining hard gluons at the scale @xmath18 are finally quenched and become part of the thermalized plasma . in the bmss scenario , this occurs at the time @xmath105 and at a temperature of @xmath106 . subsequently , the temperature of the system decreases as @xmath107 as for a hydrodynamic system undergoing one dimensional expansion . ( [ fig : summary ] ) illustrates , via the temporal evolution of @xmath84 , one scenario , based on classical - statistical field theory and kinetic theory , of the rich non - equilibrium structure that precedes the formation of a thermal quark - gluon plasma . while all the estimates here were obtained in a weak coupling framework , there is no show stopper for extending these ideas to larger , more realistic , couplings . if one puts in realistic values of @xmath27 , one gets reasonable ball park numbers for the saturation scales estimated at rhic and lhc . more quantitative phenomenology will depend on computing the pre - factors in these estimates ; such computations are in progress @xcite . even though the system may not be describable by hydrodynamics at early times , it is still strongly interacting and generating flow . hence it is plausible that when such effects are included , the start time for hydrodynamics could be moved to later times , and still give good agreement with flow data from rhic and lhc . it is also interesting to ask if there are direct signatures in the data of anisotropic early time dynamics . recent efforts have focused on electromagnetic signatures , in particular the photon spectrum @xcite . schematic illustration of the thermalization process for the example of the bulk anisotropy . the time evolution in the shaded yellow regime pictures the results of this work , with the red ellipse symbolizing the observed non - thermal fixed point . the curves beyond the classical regime reflect the evolution in the bottom - up scenario @xcite , with the possibility of a second turbulent attractor in the quantum regime @xcite . from ref . , scaledwidth=55.0% ] an alternative scenario is implied by the numerical simulations in @xcite based on eq . ( [ eq : quantum ] ) . recall this expression is derived at @xmath108 , in the theoretical formulation presented in @xcite . it is found that for @xmath109 ( @xmath110 ) , the ratio @xmath84 rises rapidly and appears to reach a constant value that is approximately 0.6 by @xmath111 and remains constant up to the maximum time presented in the simulation of @xmath112 . on the surface , this appears to contradict the results of @xcite since in the latter @xmath84 decreases as @xmath113 . assuming both numerical simulations are accurate , the difference could be attributed to a ) the different couplings chosen in the two simulations , and b ) the different initial conditions corresponding to eq . ( [ eq : quantum ] ) for @xcite and eq . ( [ eq : quantum2 ] ) for @xcite . the first point can be addressed straightforwardly . simulations performed at @xmath114 ( corresponding to @xmath115 ) , show the same late time behavior as for the larger values of @xmath116 shown in @xcite . a more contentious point could be the differing initial conditions . our claim however is that they will give the same result at late times for both sets of initial conditions . to appreciate this , consider what the lifetime of the classical regime is for @xmath117 , the value of the coupling in the simulations of @xcite . in the bottom - up scenario , this lifetime is @xmath118 . other kinetic scenarios will give comparable times , if not longer , for the applicability of classical - statistical dynamics . the simulations presented in @xcite end at @xmath119 , about @xmath120 of the lifetime of the classical regime . our prediction is that if the simulations are followed for longer , within the time when classical dynamics is valid , the ratio of @xmath84 will turn over and decrease as a function of @xmath121 . another hint that this may be the right interpretation of the results of @xcite , is the fact that the end of the instability regime is at @xmath122 , which is roughly the time when @xmath84 becomes large in the simulations of @xcite . finally , simulations of a scalar field theory at weak coupling with i ) the ` field driven ' initial conditions of eq . ( [ eq : quantum ] ) and ii ) the ` fluctuation driven ' initial conditions of eq . ( [ eq : quantum2 ] ) , give the same late time attractor behavior , well within the domain of validity of classical statistical simulations @xcite . the situation can be settled concretely by performing the simulations of @xcite on larger lattices for later times . it will also be important to extract the quasi - particle spectrum of @xcite , as for instance , done in @xcite . it is of crucial importance to settle this issue one way or the other . if the claim of @xcite is borne out , classical field dynamics can be matched to viscous hydrodynamics , in a more sophisticated version of the ip - glasma model @xcite . if not , one has to work harder , and match classical dynamics to kinetic theory first , before thermalization is attained . it is at one level more cumbersome , but on another , provides further insight into the rich dynamics that becomes possible in strongly correlated non - abelian plasmas . interesting ideas to merge classical field theory and kinetic theory can be found in @xcite . of course , weak coupling approaches could be wrong altogether , and nature , at realistic coupling , may have different plans in mind . the data from heavy ion collisions however is so plentiful , and of such high quality , that one can expect significant progress in the near future . one line of practical inquiry is to see how tweaks of the early time dynamics in the very successful ip - glasma model influence comparisons with experiment . data on two particle correlations in high multiplicity proton - proton and proton - nucleus collisions at the lhc have a component that is only weakly dependent on the rapidity separation ( @xmath123 ) between the pairs and is strongly collimated in the azimuthal angular separation ( @xmath124 ) between the pairs . such ridge "- like structures were first seen in a+a collisions at rhic , at the lhc , and more recently , in very central deuteron - gold collisions at rhic . for recent reviews , see @xcite . in perturbative qcd the only two parton correlation that one obtains is the back - to - back @xmath125 correlation from the di - jet graph ; this explains why none of the event generators saw the ridge collimation at @xmath124 . in the cgc effective theory , the nearside @xmath124 collimation is obtained from connected two gluon production qcd graphs called glasma graphs " . they are qcd interference graphs . in conventional perturbative qcd computations , these graphs are ignored for good reason . their contribution at high @xmath21 and in peripheral collisions is negligibly small . however , most remarkably , the high occupancy of gluons ( for transverse momenta @xmath126 , where @xmath127 is the saturation scale ) in rare high multiplicity proton - proton events enhances such graphs by @xmath128 . this corresponds to a strikingly large enhancement of @xmath129 for typical values of the probed qcd fine structure constant @xmath27 ! thus in the cgc power counting , gluon saturation ensures that glasma graphs provide a significant additional contribution in high multiplicity events to di - jet " qcd graphs . the importance of glasma graphs was first discussed in @xcite and the formalism developed in @xcite . it was first postulated as an explanation of the high multiplicity cms proton - proton ridge in @xcite , and a quantitative description of the nearside collimated yield obtained in @xcite . the di - jet contribution that is long range in rapidity is described by bfkl dynamics @xcite . we showed in @xcite that bfkl dynamics does well in describing the awayside spectra in high multiplicity proton - proton collisions . the description is significantly better than pythia-8 @xcite , and @xmath130 qcd graphs in the quasi multi regge kinematics ( qmrk ) @xcite . both of these approaches overestimate the awayside yield especially at larger momenta . the glasma+bfkl cgc framework @xcite describes reasonably well the associated yield per trigger obtained in p+pb data at @xmath131 tev / nucleon by the cms collaboration @xcite . subsequently , both the alice @xcite and atlas @xcite collaborations published their di - hadron correlation results from the first lhc p+pb run . the alice experiment has an acceptance in @xmath132 of @xmath133 , while the atlas experiment has an acceptance of @xmath134 , close to the cms acceptance of @xmath135 . in addition to the lhc results , the phenix collaboration at rhic reanalyzed their deuteron - gold data at 200 gev / nucleon and extracted a ridge signal in very central events @xcite . all three of the experiments show that when the two particle yield in peripheral collisions is subtracted from the central events , a dipole structure remains that is long range in rapidity . this is precisely what we anticipated in our glasma+bfkl graph scenario . specifically , this is because the bfkl di - jet contribution has a weak dependence on centrality and the net glasma graph contribution is symmetric around @xmath136 . the lhc experiments have analyzed @xcite the @xmath137 fourier moments of the jet - subtracted di - hadron yields in proton - nucleus collisions . ( note that this has not been done yet in proton - proton high multiplicity collisions because the jet yield is very large , potentially generating large systematic uncertainties in any such subtraction . ) they obtained a significant value of @xmath138 which approaches that observed in peripheral a+a collisions . because of the @xmath139 dependence of our glasma graph contribution , we too expect a significant @xmath138 . however , the experiments observed a distinct @xmath140 contribution . in addition , @xmath141 , @xmath138 extracted from four particle cumulants , is significant . both of these effects were widely interpreted as problematic for the initial state cgc picture . however , there is also an interference contribution between the two contributions that must be there . previously , we had not included it because it is small . our preliminary conclusions are that this contribution gives a triangular azimuthal anisotropy and that it has the right centrality and @xmath21 dependence seen in the data @xcite . the observation of large ridges in proton - nucleus collisions that are comparable in size to those in peripheral nucleus - nucleus collisions have reinvigorated interpretations in terms of hydrodynamic flow . additional force for the flow argument is adduced from the mass ordering of @xmath142 for different hadron species @xcite . in p+a collisions , it looks remarkably similar to that in peripheral a+a collisions , both in shape and magnitude . computations in a hydrodynamical model reproduced a number of features of the proton - lead data @xcite . other recent hydrodynamics based discussions of the ridge in p+a collisions can be found in refs . however , hydrodynamics when applied to p+a collisions is very sensitive to assumptions about the dynamics of the initial state and values of @xmath0 . the hydrodynamic description of small size systems has recently been studied in the framework of the ip - glasma+music model @xcite . though the model gives excellent fits to data at both rhic and lhc in a+a collisions , the values of @xmath138 and @xmath140 in @xmath143 collisions are factors of 2 - 3 smaller than the data for the same values of @xmath0 that describe the a+a data @xcite . to help distinguish between final state models alone , it is essential to perform ` apples - to - apples ' comparisons of data in very different centrality classes in a+a and p+a that correspond to the same @xmath144 . very recently , we performed a comprehensive study of single inclusive multiplicities and multiplicity distributions in p+p , p / d+a and a+a collisions within the ip - glasma framework @xcite . with the results of this study , flow generated from the initial state configurations that generate the @xmath145% central a+a collisions can be compared directly to the flow generated from initial state configurations corresponding to @xmath146% central events in p+pb collisions . the preliminary results are consistent with the studies in @xcite that indicate that @xmath138 generated is too small to describe the data . we note however that if the spatial distribution of gluons is less uniform than that given by the ip - glasma framework , larger flow moments may be generated . this further emphasizes the sensitivity of results to assumptions about the spatial distribution of glue in rare configurations of the colliding protons and nuclei . a detailed discussion of our take on the initial state versus final state issues on the ridge in p+p , p+a and a+a collisions can be found in @xcite . here we will provide a rough sketch of our perspective on these issues . in high multiplicity events , the power counting in the cgc tells us that the underlying glasma background is azimuthally isotropic . this contribution is the leading contribution in powers of the coupling and the number of colors @xmath147 . jetty contributions are sub - leading in powers of the coupling but leading in @xmath147 . the interplay between the two , as a function of centrality and transverse momenta , determines the broad features of two particle correlations in high multiplicity events . similar conclusions , couched in the language of sphericity " observables , have been advanced in the pythia event generator framework @xcite . however , in addition to the leading azimuthally isotropic and jetty contributions in the glasma , there is an intrinsic collimation in delta @xmath148 that is long range in rapidity . these arise from the so - called glasma graphs we described previously . they are leading in powers of the coupling in high multiplicity events but are sub - leading ( @xmath149 suppressed ) in the number of colors . the effect is purely a quantum interference effect , because the color structure of the graphs in the amplitude and in the complex conjugate amplitude are different . some of the glasma contributions are hanbury - brown twiss ( hbt ) like interference contributions , but others are not the collimation is sensitive to the detailed shape of the unintegrated gluon distributions in the protons and nuclei . in a+a collisions , the azimuthally isotropic leading contributions to multiparticle production develop a sensitivity to the geometry of the overlap region due to final state interactions and generate the long range ridge like structure due to flow . the glasma graphs are tiny in comparison . however , when the multiplicity of the isotropic contribution is small , the intrinsic glasma contribution has the chance to peek its head out of the underlying background , a colorful diamond shining through the muck . this long range quantum correlation provides an intriguing alternative to the ubiquitous flow explanation that is applied to the small sized p+a and p+p systems . when we put together the following * our recent results on the triangular anisotropy in this framework , * the disagreement in the ip - glasma+music model with @xmath138 and @xmath140 measured in p+a collisions , especially when contrasted to the remarkable success of the ip - glasma+music model in describing a+a collisions , * the recent hbt results from the alice collaboration showing that hbt radii in p+a collisions have the same slope in p+p and p+a collisions , in contrast to a+a collisions @xcite , there is some cause for optimism that this manifestation of long range gluon ` entanglement ' will survive . in addition to further developments in theory , a wealth of data coming out in the near future will further test this picture relative to other , mostly final state , explanations . i first met gerry brown when he invited me to visit stony brook , whilst i was still an undergraduate at the university of chicago . he asked me if i had the proverbial fire in my belly . i mumbled something , and that was that i was hired . it being 1987 , stony brook was the tower of bbal @xcite , with gerry conducting that noisy and unruly orchestra with aplomb . the wild excitement notwithstanding , supernova physics was at that time not quite to my taste . but gerry did nt hold it against me , and instead paired me with prakash , who proved to be a model advisor . gerry persisted in being very kind to me and sent me off to spend summers in berkeley , copenhagen and minneapolis . i admired greatly gerry s intuitive approach and passion for physics that he communicated effortlessly , as well as his wide erudition and dry wit . i was fortunate though not to catch the sharp edge of the last , which did not endear him to everyone . when i was about to graduate , and job prospects looked dim , gerry came to me and said i should stay an extra year and work with him if things did nt work out . they did work out but part of me regrets that missed opportunity . r.vs research was supported by doe contract no . de - ac02 - 98ch10886 . he thanks his collaborators juergen berges , kirill boguslavski , adam bzdak , kevin dusling , bjoern schenke , soeren schlichting and prithwish tribedy for generously contributing their insight into the topics discussed here . he is grateful to bjoern schenke and soeren schlichting for a close reading of the manuscript , and to edward shuryak for valuable editorial comments . b. schenke , p. tribedy , r. venugopalan , phys . lett . * 108 * , 252301 ( 2012 ) ; phys . c * 86 * , 034908 ( 2012 ) ; c. gale , s. jeon , b. schenke , p. tribedy , r. venugopalan , phys . rev . lett . * 110 * , 012302 ( 2013 ) .	the late gerry brown was not shy to tackle complex scientific problems that took time to play out but yielded in the end a deeper understanding of many - body phenomena . in this note , prepared for a memorial volume in his honor , we provide a perspective on a couple of outstanding scientific puzzles that have their origin in our understanding of the thermalization of matter in ultrarelativistic heavy ion collisions , and possibly , in high multiplicity proton - proton and proton - nucleus collisions .
a small fraction of massive stars end their lives with spectacular explosions one or two orders of magnitude more luminous than normal supernovae ( sne ) . after the initial puzzling discoveries of the luminous sne 2005ap @xcite and 2006gy @xcite , modern wide - field surveys over the past decade began to uncover these superluminous sne ( slsne ) in greater numbers . the energy scales involved in these explosions challenge our understanding of conventional sn explosions . normal sne resulting from iron core collapse have characteristic energy scales of @xmath310@xmath15 erg of kinetic energy and @xmath310@xmath16 erg emitted as optical radiation ( @xmath310@xmath17 erg s@xmath6 for @xmath310@xmath18 s ) . the slsne are far off this scale they peak at optical luminosities of up to @xmath34@xmath19 erg s@xmath6 @xcite and emit a total of up to 4@xmath410@xmath15 erg optically @xcite . this large energy scale motivates the question of what physics powers these sne , and how to accommodate these objects within the standard understanding of massive star evolution . theorists have proposed a number of exotic power sources , including the pair instability mechanism ( e.g. , @xcite ) and reprocessed spindown energy released by a newly formed magnetar @xcite . another possibility is interaction with a dense circumstellar medium ( csm ) @xcite , requiring extreme csm masses and densities whose origin remains unexplained ( see @xcite for one possibility ) . all of these models require additional ingredients beyond the normal stellar evolutionary processes . @xcite has attempted to impose order on the menagerie of objects achieving sufficient peak luminosities to be classified as slsne ( @xmath20 mag is a typical requirement ) by sorting them into three categories . all of the hydrogen - rich objects were classified as slsne - ii and all exhibit signs of being powered by dense csm interaction , with the possible exception of sn 2008es @xcite . he split the objects lacking hydrogen into two classes , the rare slsne - r that have slow photometric decline rates consistent with being powered by the radioactive decay of a very large synthesized mass of @xmath21ni , and the relatively homogeneous class of slsne - i , whose power source is still mysterious . a few caveats have been raised . the slsne - r are interpreted to be the results of pair - instability sne . however , existing models for the pair instability process prefer extremely low metallicity , and may be in conflict with the observed spectrum and spectral energy distribution ( sed ) of slsne - r ( e.g. , @xcite ) . also , it is not clear how homogeneous the slsne - i class really is . although the spectra of most appear to be similar to those of sn 2005ap and scp06f6 @xcite , the rise times and peak luminosities of published objects vary by factors of @xmath35 @xcite . all slsne - i to date have had hot spectra and been bright in the rest - frame near - ultraviolet ( nuv ) relative to normal sn seds . in this paper , we present the discovery of ps1 - 10afx , an extreme slsn at redshift @xmath0 that does not fit into this classification scheme and is distinct from all previous slsne . the peak luminosity is comparable to the highest known and the rise time is the fastest measured . the spectra show no evidence for hydrogen and lack any analog in the existing sample of slsne . instead , they most closely resemble those of line - blanketed normal sne ic . in section 2 , we present the suite of optical and near - infrared ( nir ) observations . the host galaxy is described in section 3 . we compare our observations of ps1 - 10afx to known sne in section 4 . in section 5 , we construct the sed and bolometric light curve . we then compare ps1 - 10afx to existing slsn models in section 6 . all calculations in this paper assume a flat @xmath22cdm cosmology with @xmath23=74 km s@xmath6 mpc@xmath6 , @xmath24=0.27 , and @xmath25=0.73 @xcite . [ [ section ] ] the pan - starrs1 ( ps1 ) telescope has a 1.8 m diameter primary mirror that images a field with a diameter of 3.3 @xcite onto a total of sixty @xmath26 pixel detectors , with a pixel scale of 0.258@xcite . a more complete description of the ps1 system , hardware and software , is provided by @xcite . the ps1 observations are obtained through a set of five broadband filters , designated as @xmath27 , @xmath28 , @xmath29 , @xmath30 , and @xmath31 . although the filter system for ps1 has much in common with that used in previous surveys , such as the sloan digital sky survey ( sdss ; @xcite ) , there are differences . most important for this work , the @xmath30 filter is cut off at 9300 , giving it a different response than the detector response defined @xmath32 , and sdss has no corresponding @xmath31 filter . further information on the passband shapes is described by @xcite . photometry is in the `` natural '' ps1 system , @xmath33 , with a single zeropoint adjustment @xmath34 made in each band to conform to the ab magnitude scale @xcite . photometry from all other sources presented in this paper is also on the ab scale . ps1 magnitudes are interpreted as being at the top of the atmosphere , with 1.2 airmasses of atmospheric attenuation being included in the system response function . the ps1 medium deep survey ( mds ) consists of 10 fields across the sky that are observed nightly when in season ( @xmath35 months per year ) with a typical cadence of 3 d between observations in @xmath27@xmath28@xmath29@xmath30 in dark and gray time , while @xmath31 is used near full moon . ps1 data are processed through the image processing pipeline ( ipp ; @xcite ) on a computer cluster at the maui high performance computer center . the pipeline runs the images through a succession of stages , including flat - fielding ( `` de - trending '' ) , a flux - conserving warping to a sky - based image plane , masking and artifact removal , and object detection and photometry . transient detection using ipp photometry is carried out at queen s university belfast . independently , difference images are produced from the stacked nightly mds images by the ` photpipe ` pipeline @xcite running on the odyssey computer cluster at harvard university . the discovery and data presented here are from the ` photpipe ` analysis . ps1 - 10afx was first detected in mds imaging on 2010 august 31.35 ( ut dates are used throughout this paper ) at a position of @xmath35 = 22@xmath3611@xmath3724162 , @xmath38 = @xmath390009@xmath404349 ( j2000 ) , with an uncertainty of 0@xmath411 in each coordinate . the strong detections in @xmath29 and @xmath30 combined with non - detections in @xmath27 and @xmath28 over the next few nights immediately garnered our attention . the unusual colors are evident in the color images of the field presented in figure [ hostfig ] . we constructed deep template stacks from all pre - explosion images and subtracted them from the ps1 observations using ` photpipe ` @xcite . details of the photometry and generation of ps1 sn light curves will be given by rest et al . ( 2013 , in prep . ) and scolnic et al . ( 2013 , in prep . ) . the typical spacing between mds observations in the same filter corresponds to only 1.3 d in the rest frame of ps1 - 10afx , so in some cases we have co - added the photometry from adjacent observations to increase the significance of marginal detections or the depth of the non - detections . the final ps1 - 10afx photometry , after correction for @xmath42 mag of galactic extinction @xcite , is given in table [ phottab ] and shown in figure [ multiplot ] . we fit polynomials to the @xmath30 data points within 15 rest - frame days of peak to determine the time of maximum light , for which we adopt a modified julian date ( mjd ) of 55457.0 ( = 2010 september 18.0 ) . all phases referred to subsequently are in rest - frame days referenced from this date . in addition to the ps1 observations , we obtained two epochs of multicolor photometry using the gemini multi - object spectrographs ( gmos ; @xcite ) on the 8-m gemini - north and south telescopes and one epoch of imaging using the inamori - magellan areal camera and spectrograph ( imacs ; @xcite ) on the 6.5-m magellan baade telescope . the images were processed using standard tasks and then archival fringe frames were subtracted from the gmos images using the ` gemini ` iraf package . we used several sdss stars in the field to calibrate the gemini images , while the imacs zeropoints were checked with observations of standard star fields obtained the same night . we subtracted the deep ps1 templates in the corresponding filter from each image using the isis software package @xcite to correct for host contamination . the final photometry is listed in table [ phottab ] . the @xmath43 and @xmath44 magnitudes agree well with the ps1 observations at similar epochs . however , the @xmath45 observations exhibit an offset from the @xmath30 light curve due to the filter response differences noted above ( gmos and imacs are closer to sdss ) . we therefore add 0.2 mag to the @xmath45 photometry points for consistency with @xmath30 whenever we refer to the combined @xmath46-band light curve . finally , we obtained late - time observations of the host galaxy in @xmath47 and @xmath43 on 2011 october 21.1 ( phase @xmath39167 d ) using the low dispersion survey spectrograph-3 ( ldss3 ) on the 6.5-m magellan clay telescope ( table [ hosttab ] ) . lrcccc 55440.5 & @xmath486.9 & @xmath27 & @xmath4923.50 & & ps1 + 55445.0 & @xmath485.0 & @xmath27 & @xmath4923.91 & & ps1 + 55450.9 & @xmath482.6 & @xmath27 & @xmath4923.77 & & ps1 + 55455.4 & @xmath480.7 & @xmath27 & @xmath4923.22 & & ps1 + 55470.3 & 5.6 & @xmath27 & @xmath4923.68 & & ps1 + 55480.7 & 9.9 & @xmath27 & @xmath4923.85 & & ps1 + 55439.0 & @xmath487.5 & @xmath28 & @xmath4923.13 & & ps1 + 55445.0 & @xmath485.0 & @xmath28 & @xmath4923.47 & & ps1 + 55450.2 & @xmath482.8 & @xmath43 & 24.18 & 0.30 & gmos - n + 55450.9 & @xmath482.5 & @xmath28 & 23.87 & 0.41 & ps1 + 55455.4 & @xmath480.7 & @xmath28 & @xmath4923.42 & & ps1 + 55465.0 & 3.4 & @xmath43 & 23.73 & 0.20 & gmos - s + 55470.3 & 5.6 & @xmath28 & @xmath4923.63 & & ps1 + 55480.8 & 9.9 & @xmath28 & @xmath4923.55 & & ps1 + 55425.0 & @xmath4813.4 & @xmath29 & @xmath4924.21 & & ps1 + 55438.5 & @xmath487.7 & @xmath29 & @xmath4923.84 & & ps1 + 55441.5 & @xmath486.5 & @xmath29 & 23.10 & 0.16 & ps1 + 55444.4 & @xmath485.3 & @xmath29 & 22.57 & 0.12 & ps1 + 55447.4 & @xmath484.0 & @xmath29 & 22.32 & 0.06 & ps1 + 55450.4 & @xmath482.8 & @xmath29 & 22.15 & 0.11 & ps1 + 55450.2 & @xmath482.8 & @xmath44 & 22.18 & 0.11 & gmos - n + 55453.4 & @xmath481.5 & @xmath29 & 22.00 & 0.05 & ps1 + 55456.4 & @xmath480.3 & @xmath29 & 21.99 & 0.05 & ps1 + 55465.0 & 3.4 & @xmath44 & 22.25 & 0.16 & gmos - s + 55465.5 & 3.5 & @xmath29 & 22.28 & 0.20 & ps1 + 55468.3 & 4.7 & @xmath29 & 22.39 & 0.12 & ps1 + 55477.2 & 8.5 & @xmath29 & 23.20 & 0.19 & ps1 + 55480.2 & 9.7 & @xmath29 & 23.28 & 0.42 & ps1 + 55483.2 & 11.0 & @xmath29 & 23.45 & 0.22 & ps1 + 55486.3 & 12.2 & @xmath29 & @xmath4922.90 & & ps1 + 55499.8 & 17.9 & @xmath29 & @xmath4923.87 & & ps1 + 55508.8 & 21.7 & @xmath29 & @xmath4924.08 & & ps1 + 55513.1 & 23.5 & @xmath44 & @xmath4925.01 & & imacs + 55426.0 & @xmath4813.0 & @xmath30 & @xmath4923.47 & & ps1 + 55436.4 & @xmath488.6 & @xmath30 & @xmath4922.79 & & ps1 + 55439.3 & @xmath487.4 & @xmath30 & 23.06 & 0.37 & ps1 + 55442.5 & @xmath486.1 & @xmath30 & 22.83 & 0.37 & ps1 + 55445.4 & @xmath484.8 & @xmath30 & 22.24 & 0.15 & ps1 + 55448.4 & @xmath483.6 & @xmath30 & 21.88 & 0.12 & ps1 + 55450.2 & @xmath482.8 & @xmath45 & 21.59 & 0.13 & gmos - n + 55454.4 & @xmath481.1 & @xmath30 & 21.69 & 0.10 & ps1 + 55457.3 & 0.1 & @xmath30 & 21.76 & 0.11 & ps1 + 55465.0 & 3.4 & @xmath45 & 21.44 & 0.19 & gmos - s + 55466.4 & 3.9 & @xmath30 & 21.82 & 0.10 & ps1 + 55475.2 & 7.6 & @xmath30 & 22.11 & 0.17 & ps1 + 55478.2 & 8.9 & @xmath30 & 22.13 & 0.17 & ps1 + 55481.2 & 10.1 & @xmath30 & 21.98 & 0.19 & ps1 + 55484.2 & 11.4 & @xmath30 & 22.77 & 0.30 & ps1 + 55499.2 & 17.7 & @xmath30 & @xmath4923.04 & & ps1 + 55513.1 & 23.5 & @xmath45 & 23.83 & 0.30 & imacs + 55463.4 & 2.7 & @xmath31 & 21.09 & 0.16 & ps1 + 55492.8 & 15.0 & @xmath31 & @xmath4921.95 & & ps1 + 55523.3 & 27.8 & @xmath31 & @xmath4921.82 & & ps1 + 55463.4 & 2.7 & @xmath50 & 20.96 & 0.10 & niri + 55463.3 & 2.6 & @xmath51 & 20.99 & 0.17 & mmirs + 55463.4 & 2.7 & @xmath51 & 21.19 & 0.10 & niri + 55481.1 & 10.1 & @xmath51 & 22.22 & 0.26 & mmirs + 55485.0 & 11.7 & @xmath51 & 22.28 & 0.30 & mmirs + 55496.0 & 16.3 & @xmath51 & 22.19 & 0.10 & hawk - i + 55515.1 & 24.3 & @xmath51 & 22.84 & 0.14 & hawk - i + 55552.0 & 39.8 & @xmath51 & @xmath4921.90 & & hawk - i + 55561.0 & 43.6 & @xmath51 & @xmath4921.90 & & hawk - i + 55463.4 & 2.7 & @xmath52 & 21.20 & 0.19 & niri + 55485.0 & 11.7 & @xmath52 & 21.05 & 0.44 & mmirs + 55490.1 & 13.9 & @xmath52 & 21.94 & 0.24 & hawk - i + 55515.1 & 24.3 & @xmath52 & 22.28 & 0.39 & hawk - i + 55463.4 & 2.7 & @xmath53 & 21.46 & 0.26 & niri + 55485.0 & 11.7 & @xmath54 & @xmath4921.35 & & mmirs [ phottab ] the high redshift and red optical colors of ps1 - 10afx indicate that nir photometry is required to constrain the sed and bolometric luminosity . we received director s discretionary ( dd ) time at gemini - north to obtain @xmath55 photometry near maximum light using the near infrared imager and spectrometer ( niri ; @xcite ) . the observations were acquired on 2010 september 24.4 , at a phase of @xmath392.7 d. additional observations were obtained over the subsequent month in @xmath56 with the mmt & magellan infrared spectrograph ( mmirs ; @xcite ) on the magellan clay telescope . to follow the light curve in the rest - frame optical to late times , we obtained a series of @xmath57 observations using dd time on the 8-m very large telescope ( yepun ) with the high - acuity wide - field k - band imager ( hawk - i ; @xcite ) . finally , we used the fourstar infrared camera @xcite on the magellan baade telescope to obtain late - time @xmath56 observations of the host galaxy of ps1 - 10afx on 2011 september 18 , 2011 september 21 , and 2011 december 7 ( phases @xmath39153 d , @xmath39154 d , and @xmath39186 d ) , respectively ( table [ hosttab ] ) . the images were flat fielded , sky subtracted , and stacked using standard tasks in iraf , including the ` gemini ` package for the niri data , except for the hawk - i data , which were processed using the instrument pipeline . the images from each instrument were then calibrated using the same 2mass stars in each field , except for the @xmath50 image , which was calibrated using archival niri zeropoints . the @xmath52 and @xmath54 host galaxy fluxes were then subtracted numerically from each datapoint in those filters . for the @xmath50 data point , we interpolated the host galaxy flux between the measured values in @xmath31 and @xmath51 to subtract off the host galaxy contribution ( only a small correction ) . we were able to perform image subtraction on most of the @xmath51 images using the isis software package with the late - time fourstar image as a template , but the others had the host @xmath51 flux subtracted numerically . the final host - corrected photometry points are shown in figure [ nirplot ] and listed in table [ phottab ] , including the calibration uncertainty in the errors . after noticing the unusually red @xmath58 colors of ps1 - 10afx , we immediately triggered optical spectroscopy using gmos - n . a pair of dithered 1800 s observations were taken on 2010 september 11.26 ( phase @xmath59 d ) with the r400 grating and og515 order - blocking filter to cover the wavelength range of 5400@xmath489650 . we used standard tasks in iraf to perform basic two - dimensional image processing and spectral extraction . we used our own idl procedures to apply a flux calibration and correct for telluric absorption based on archival observations of spectrophotometric standard stars . the observations were performed at a mean airmass of 1.5 , but the 1-wide long slit was oriented within 20 of the parallactic angle @xcite , so the overall spectral shape is reliable at these red wavelengths . a second epoch of spectroscopy was obtained on 2010 september 26.04 ( phase @xmath393.4 d ) using gmos - s . the observations and instrumental setup were similar to the first epoch , except that a redder grating tilt was used to cover the range 5880@xmath4810100 . this spectrum has a lower signal - to - noise ratio ( s / n ) than the first one , but the main sn features are still present . we attempted a final epoch of spectroscopy using gmos - s in nod - and - shuffle mode on 2010 november 5 , but ps1 - 10afx had faded significantly , so we only detected a faint continuum with a hint of sn features . we do not consider this spectrum further . we observed ps1 - 10afx with the folded - port infrared echellette ( fire ; @xcite ) spectrograph on the magellan baade telescope on 2010 september 18.17 ( phase @xmath390.1 d ) , using the high throughput , low - resolution prism mode to cover the range of [email protected] @xmath60 m . the spectral resolution is a strongly decreasing function of wavelength , but is @xmath61 in @xmath51 . six 150 s dithered exposures were obtained for a total of 900 s of on - source time . despite the short exposure time , the continuum of ps1 - 10afx is well detected blueward of @xmath31.6 @xmath60 m . we reduced and combined the spectra using the firehose package , including a correction for telluric absorption obtained from observations of the a0v star hd 208368 using the algorithm of @xcite . we obtained a mmirs spectrum on 2010 september 28 ( phase @xmath394.2 d ) , using the @xmath62 filter and @xmath51 grism to cover the range [email protected] @xmath60 m . a total of 90 minutes were spent on source in a series of dithered 300 s exposures . we used standard tasks in iraf to combine and extract the spectra . the final optical and nir spectra are shown in figure [ specplot ] . all of our optical spectra exhibit an emission line near 8903 , which we identify as [ ] @xmath633727 emission from the host galaxy , as well as absorption from the @xmath632800 doublet at the same redshift of @xmath0 , which we adopt as the sn redshift . popular models for slsne do not predict detectable radio emission at high redshift , as discussed by @xcite . however , given its unusual properties , we observed ps1 - 10afx on 2010 september 12.2 ( phase @xmath482.4 d ) with the karl g. jansky very large array @xcite as part of our nrao key science project `` exotic explosions , eruptions , and disruptions : a new transient phase - space , '' with 256 mhz of bandwidth centered at 4.96 ghz . the data were reduced with standard tasks in the astronomical image processing system ( aips ; @xcite ) , using j2212 + 0152 as the gain calibrator and 3c48 as the absolute flux density calibrator . these observations yielded a non - detection of 19@xmath6418 @xmath60jy . we measure a host redshift of @xmath65 from fits to the [ ] line in our highest s / n spectrum ( phase @xmath482.8 d ) . a double - gaussian fit to the @xmath632800 doublet absorption in the same spectrum ( inset in figure [ specplot ] ) is blueshifted by 120@xmath6412 km s@xmath6 relative to the [ ] rest frame . such an offset is typical of rest - frame uv selected star - forming galaxies at this redshift @xcite and was also seen in observations of the slsn ps1 - 11bam @xcite . this effect has been interpreted to be caused by absorption occurring in galactic - scale outflows driven by star formation . the rest - frame equivalent widths ( @xmath66 ) of the two lines are @xmath66(@xmath632796)[email protected] and @xmath66(@xmath632803)[email protected] . these values are larger than for ps1 - 11bam @xcite , but slightly lower than the median of the intrinsic absorbers from gamma - ray burst ( grb ) host galaxies @xcite . in addition , @xmath632852 absorption is present , but the uv spectral slope of ps1 - 10afx is so red that the s / n rapidly decreases to the blue , making it hard to confirm the presence of other expected strong interstellar absorption lines , such as @xmath632600 . we obtained photometry of the host galaxy of ps1 - 10afx in eight filters from @xmath47 to @xmath54 ( table [ hosttab ] ) to probe the host stellar population from the rest - frame nuv ( @xmath32000 ) to the nir ( @xmath30.9 @xmath60 m ) . the host is well detected in our deep @xmath29@xmath30@xmath31 ps1 template images , which we supplemented with the observations from other facilities described above . host photometry was performed with a consistent aperture of radius 1.7 in all filters . we fit a suite of single stellar population age models from @xcite to the data , assuming a red horizontal branch morphology , a salpeter initial mass function , and a metallicity of @xmath67 . the best - fit model is shown in figure [ hostplot ] and has an age of 10@xmath14 yr and requires a small amount of internal extinction ( a@xmath68 mag ) . the derived stellar mass is @xmath31.8 @xmath4 10@xmath13 m@xmath69 . the @xmath47 magnitude of the host galaxy corresponds to a nuv continuum luminosity of 1.3@xmath410@xmath70 erg s@xmath6 ( after correction for internal extinction ) , which implies a star formation rate ( sfr ) of @xmath318 m@xmath69 yr@xmath6 using the calibration of @xcite . we also estimate a consistent value of @xmath313 m@xmath12 yr@xmath6 from the observed [ ] flux of @xmath7110@xmath72 erg @xmath73 s@xmath6 @xcite . these sfrs would be a factor of @xmath32 lower without any correction for internal extinction . ccccc @xmath47 & 4825/2020 & 24.01 & 0.10 & ldss3 + @xmath43 & 6170/2580 & 23.69 & 0.10 & ldss3 + @xmath29 & 7520/3150 & 23.43 & 0.13 & ps1 + @xmath30 & 8660/3625 & 22.75 & 0.10 & ps1 + @xmath31 & 9620/4030 & 22.29 & 0.28 & ps1 + @xmath51 & 12500/5230 & 22.02 & 0.14 & fourstar + @xmath52 & 16500/6910 & 21.93 & 0.22 & fourstar + @xmath54 & 21490/9000 & 21.53 & 0.25 & fourstar [ hosttab ] this galaxy is significantly more massive than the previous hosts of slsne . @xcite examined the host galaxies of a sample of luminous sne and found them to have generally low stellar masses and high specific sfrs . in particular , all of the slsne - i in their sample had dwarf hosts , with a median stellar mass of @xmath32@xmath410@xmath74 m@xmath12 . by comparison , ps1 - 10afx has a significantly more massive host with a lower specific sfr ( @xmath38@xmath410@xmath75 yr@xmath6 ) than their sample ( with a median of 3@xmath410@xmath76 yr@xmath6 ) . only the luminous host galaxy of the slsn - ii 2006gy exceeded our estimate for the stellar mass of the host of ps1 - 10afx . objects discovered subsequently have continued this trend , having either dwarf host galaxies or no host at all detected to date despite deep observations @xcite . while it has been argued that the association of slsne with dwarf galaxy hosts is driven by their low metallicity ( e.g. , @xcite ) , the massive host of ps1 - 10afx appears to be inconsistent with a low metallicity . for example , the relationships of @xcite imply a super - solar metallicity of @xmath77(o / h ) = 8.85 for our derived stellar mass and sfr , although we caution that we can not measure the metallicity directly with the available data . an additional point of contrast is provided by the sample of long - duration grbs , whose occurrence is also widely believed to be associated with low - metallicity environments ( e.g. , @xcite ) . the host galaxy of ps1 - 10afx is brighter in @xmath54 than any of the grb host galaxies at similar redshifts in the optically - unbiased sample of @xcite , although a few of the `` dark '' grb hosts studied by @xcite are similarly luminous . the absolute magnitudes of ps1 - 10afx s host prior to any internal extinction correction , @xmath78 mag and @xmath79 mag , are @xmath30.4 mag brighter than @xmath80 , the characteristic magnitudes in @xcite function fits to the field galaxy luminosity functions at this redshift @xcite . the @xmath54 magnitude of the host is also near the median of the distribution of field galaxies at this redshift weighted by star formation rate @xcite . combined , these points suggest that the environment of ps1 - 10afx is representative of typical star - forming galaxies at this redshift , while being distinct from the other hydrogen - poor slsne . the host is marginally resolved in our template images , with a full width at half - maximum ( fwhm ) of @xmath31@xmath412 , or 10 kpc at this redshift . astrometric alignment of difference images taken near maximum light with the templates shows that ps1 - 10afx is aligned with the centroid of the host to within 0@xmath411 ( 0.8 kpc ) . the other slsn with a massive host galaxy , sn 2006gy @xcite , is also close to the nucleus , perhaps implying an unusual star formation environment . despite its extraordinary luminosity ( see section 5 ) , ps1 - 10afx has spectra that more closely resemble those of a normal sn ic than any known slsn . we show our highest s / n spectrum ( phase @xmath482.8 d ) in figures [ uvcomp ] and [ uvcomplog ] along with several comparisons drawn from the literature . relatively few uv spectra of core - collapse sne exist , but the available data allow us to sample a wide variety of phenomena . the most prominent feature in the ps1 - 10afx spectrum is a broad p - cygni feature with an absorption minimum near 3730 , which we identify as the typical h&k absorption seen in most types of sne , blueshifted by about @xmath81 km s@xmath6 . there are several other weaker features in the spectrum in the range 3000@xmath483500 . the @xmath393.4 d gmos spectrum is noisier , but similar , with some evidence that the broad absorption minimum decreased in velocity , although the overlapping [ ] emission and strong night sky residuals make this uncertain . these spectra are in strong contrast to the sn 2005ap - like class of slsne @xcite . ps1 - 10ky is a high - redshift example of this class @xcite and it has a much bluer nuv continuum with no feature ( figure [ uvcomp ] ) . a reasonable concern is that ps1 - 10afx only appears to have redder colors due to host - galaxy extinction . however , matching the uv spectral slope of ps1 - 10afx to ps1 - 10ky requires @xmath82 mag of extinction ( gray line in figure [ uvcomp ] ) , assuming a galactic reddening curve @xcite . this is unsatisfactory for two reasons . first , ps1 - 10afx is already more luminous in the nuv than the sn 2005ap - like objects ( see below ) , so @xmath83 mag would imply an upward correction of two orders of magnitude to an already extreme peak luminosity . second , the strongest uv spectral features of the sn 2005ap - like objects are the trio of features blueward of 3000 first seen in scp06f6 @xcite , with the continuum redward of that being largely smooth except for features @xcite . ps1 - 10afx has fundamentally different spectral features with no correspondence in the ps1 - 10ky spectrum , so the simplest explanation is that it has an intrinsically cooler photosphere . the sn iin 1998s is shown in figure [ uvcomp ] as a well - studied example of a sn undergoing strong csm interaction . sn 1998s at this phase showed no broad p - cygni features in the uv , but had a blue continuum with narrow absorption lines from interstellar and circumstellar material @xcite . the slsne 2006gy and 2006tf are higher - luminosity versions of sne dominated by hydrogen - rich csm interaction @xcite . these objects have redder spectra than sn 1998s , but still exhibit balmer lines and deep narrow absorptions that have no analog in the ps1 - 10afx spectrum . the slsn 2008es had almost featureless blue spectra near maximum light @xcite , but the plotted spectrum is from a later epoch when balmer lines were beginning to develop strength in the rest - frame optical spectrum . the nuv spectrum of the broad - lined sn ic 2002ap is included because the engine - driven sne associated with grbs have similar spectra , even if 2002ap itself was not unusually luminous @xcite . sn 2002ap has much broader and more blended features than ps1 - 10afx . thus , none of these objects in figure [ uvcomp ] resembles ps1 - 10afx . the slsn 2007bi has been identified as a potential pair - instability sn @xcite . it is more similar to ps1 - 10afx , with a definite feature , but is still significantly bluer and has several additional spectral features that do not match . instead , the set of spectra from normal sne plotted in figure [ uvcomplog ] provide a much better basis for comparison . the sne ia and sn 1993j ( type iib ) spectra have been blueshifted by the indicated amounts to approximately match the absorption minima to ps1 - 10afx . over the range 2800@xmath484000 , the spectra in figure [ uvcomplog ] are all very similar , with the dominant features being due to in all objects . a notch near 3950 is from @xmath634130 . an emission peak at 2900 also appears to be common to these objects , along with a bluer peak near 3150 . shortward of @xmath32800 , the sne ia diverge from the core - collapse objects . this dropoff in flux , along with a pair of deep absorption features near 2450 and 2600 , is caused by iron - group elements , primarily @xcite . the lack of these features in ps1 - 10afx is evidence that the abundance of newly - synthesized @xmath21ni ( which decays to @xmath21co ) near the photosphere is much lower than in a sn ia . one purpose of these comparisons is to demonstrate the apparent broad similarity of the nuv spectra of ps1 - 10afx to those of sne coming from three very different progenitors : a bare core of a massive star ( sn 1994i : @xcite ) , a partially - stripped massive star in a binary system that retained part of its hydrogen envelope ( sn 1993j : @xcite ) , and the thermonuclear explosions of white dwarfs ( sne 2011fe and 2011iv : @xcite ) . sne 2011fe and 2011iv have some spectral differences from each other near 3000 , exhibiting the potential for spectroscopic variation in objects with similar progenitors . other than the lack of cobalt near the photosphere , the remaining differences between ps1 - 10afx and these other objects are relatively minor . in addition , the overall uv spectral slope of ps1 - 10afx is very similar to those of sne 1993j and 1994i , after correction of both by @xmath84 mag of reddening @xcite . the fact that the only sne in figures [ uvcomp ] and [ uvcomplog ] that have similar spectral features to ps1 - 10afx also have similar sed shapes provides further evidence that the extinction of ps1 - 10afx is not large . if the ps1 - 10afx spectra were corrected for a large amount of reddening , they would be bluer and we would expect different spectral features to be present at the higher implied temperatures . we further explore the spectra using the sn spectrum synthesis code ` syn++ ` with ` synapps ` @xcite to fit the near - maximum - light gmos / fire spectra . these codes are based on the same assumptions as the original ` synow ` ( e.g. , @xcite ) treatment of resonant - scattering lines in the sobolev approximation above a sharp photosphere that emits as a blackbody ( bb ) . the atomic and ionic species are assumed to have levels populated in local thermodynamic equilibrium at some excitation temperature . ` synow ` is a tool designed for line identifications and is not a full self - consistent spectral model . our best ` synow ` model is shown in figure [ synplot ] and contains seven ions . we assumed equal bb continuum and excitation temperatures of @xmath85 k and a photospheric velocity of @xmath8 km s@xmath6 . the deep absorption trough near 3700 is dominated by h&k . and combine to produce several of the wiggles observed in the range @xmath86 . naturally explains the observed absorption near 4250 , while also producing absorption shortward of @xmath32700 . contributes to both the blue wing of the deep feature and the notch near 3950 . the feature near 3950 is present in the overlap between the gmos and fire data and has a consistent strength that is significantly larger than in the synthetic spectrum . with ` synow ` , we have to be careful not to overproduce absorption by the @xmath636355 doublet because no strong feature is seen in the data , although the s / n and the spectral resolution become very poor in the observed - frame @xmath52 . we introduce to help reproduce the wiggle near 5000 , but the identification is not unique and its presence is not required . in addition , we include to help suppress the uv flux . that ion does not contribute sufficiently strongly to any specific feature for us to make a positive identification , but it has been included in previous ` synow ` models of sne ic @xcite because of its expected strength in either helium or carbon / oxygen - rich sn ejecta at these temperatures @xcite . we also consider several other species that were not included in the final fit . as discussed above , produces strong nuv absorptions in sne ia that are not present in ps1 - 10afx . d is commonly seen in all types of sne , but at this redshift it would fall in the strong telluric absorption between @xmath51 and @xmath52 , so we have no constraint on its presence from the observations . we also tested because of its use in prior ` synow ` analyses of sne ib @xcite . scandium results in an additional absorption minimum near @xmath33500 similar to one seen in our data , but does not produce a sufficiently strong improvement in the model to justify its inclusion , particularly given the expectation that it should only become important at lower ejecta temperatures @xcite . there is also an apparent absorption feature near 5300 that is not fit by the model . can produce a feature near that wavelength , but the identification is not certain . lines of hydrogen or helium would be very important if present , but there is no evidence for either , although the low s / n of our nir data precludes strong statements about helium . the sn 2005ap - like slsne have lines of singly - ionized carbon and oxygen in their spectra at early times ( e.g. , @xcite ) , with possible contributions from doubly - ionized cno elements @xcite , but those lines do not appear to be present in our ps1 - 10afx spectra . overall , we can reproduce the shape of the spectrum in the rest - frame optical reasonably well , but the nuv is more problematic . our synthetic spectrum has features at many of the same wavelengths as the real data , but the overall shape in the nuv is less well fit . adding additional ions to the fit might potentially help the uv shape of the synthetic spectrum , but would not be well motivated or add any additional insight . alternatively , the radiative transfer in the uv is known to be complex and we could be encountering the limitations of ` synow ` . the most important point is that the ions we have used are typical for generic fits to sne ia and ic spectra ( e.g. , @xcite ) , but differ from those identified in most slsne - i spectra near maximum light @xcite . existing light curves of slsne are heterogeneous , and the objects have been found at a wide range of redshifts , making direct comparisons difficult . following previous work @xcite , in figure [ uplot ] we compare the rest - frame @xmath87 light curve of ps1 - 10afx ( corresponding to observed - frame @xmath30 ) to several of the most luminous sn 2005ap - like sne in the literature . in addition , we include the hydrogen - rich sn 2008es because its light curve is similar to the others and it did not develop strong balmer lines until well after maximum light @xcite . for comparison , we also plot the @xmath88 light curves of the broad - lined sne ic 1998bw and 2002ap , two objects of normal luminosities @xcite . given the uncertain seds of many of these objects , we do not perform a full @xmath89-correction , but instead correct the observed ab magnitudes in filters whose effective wavelengths are close to @xmath87 in the observed frame by @xmath90d@xmath91(@xmath46)/10 pc ) @xmath48 2.5 @xmath92(1+@xmath46 ) , where d@xmath91 is the luminosity distance . to be conservative , we assume that ps1 - 10afx is unaffected by reddening here and in all following discussion , despite the strong absorption along the line of sight and the inference that the host galaxy sed requires some degree of internal extinction . we presented some evidence above that the reddening to the sn is not large , but it is possible that we have underestimated the peak luminosity of this object . it is apparent from figure [ uplot ] that at peak ps1 - 10afx was slightly more luminous in @xmath87 ( @xmath93 mag ) than any previous slsne . however , by @xmath310 d on either side of maximum light , ps1 - 10afx falls below all the others . this emphasizes the high peak luminosity and fast evolution of the light curve . these objects have rather different seds , so comparisons of the energetics require constructing a bolometric light curve . to proceed further , we first construct seds at three epochs near the times of our nir observations , at phases of @xmath94 , @xmath95 , and @xmath96 d. we fit third - order polynomials to the observed photometry in each @xmath28@xmath29@xmath30 filter around each sed epoch to estimate the flux . we then estimate the error bars on these fitted fluxes by generating monte carlo realizations of the data where we repeatedly randomly adjust the observed fluxes by drawing from a normal distribution having a width of the measured errors , refitting , and determining the variance of the interpolated flux at the desired epoch . we perform a similar procedure whenever we need to interpolate fluxes to a common epoch . for the seds at phases of @xmath94 and @xmath96 d , we use the actual measured @xmath31 and nir fluxes . for the @xmath95 d sed , the observed nir fluxes were quite noisy , so we again interpolate the fluxes in @xmath51 and @xmath52 by fitting second - order polynomials to the observed light curves to generate the points on the sed . the resulting sed evolution is shown in figure [ sedplot ] . we then fit bb spectra to the seds at each epoch . our @xmath94 d sed has the most complete information . the dashed line in figure [ sedplot ] shows the result of a bb fit to the full sed . the derived bb temperature is @xmath97 k , but clearly the curve is a poor fit to the data at rest wavelengths longward of @xmath35000 . if instead we restrict the fit to @xmath31 and the nir bands , the bb temperature is not well constrained ( @xmath98 k ) due to a lack of points blueward of the peak . the formal best fit is plotted as the solid blue line in figure [ sedplot ] and provides a much better fit to the nir data points ( by construction ) , but significantly overestimates the rest - frame uv flux . this is to be expected if line blanketing in the nuv provided by iron - peak elements supresses the uv flux , as is typical for sne ic . for example , the shape of the sed of the normal sn ic 1994i near maximum light ( figure [ sedplot ] ) is similar to that of ps1 - 10afx , but the detailed model fits by @xcite find an underlying bb temperature of 9230 k at the same epoch . the data at phases @xmath95 and @xmath96 d are less complete , so we fit a simple bb to all of the points on the sed and do not correct for line blanketing , which would only be expected to increase as the ejecta cool . the derived bbs show the ejecta clearly cooling from @xmath3@xmath99 k at a phase of @xmath94 d to @xmath3@xmath100 k at @xmath96 d , although @xmath101 should only be regarded as lower limits to the effective temperature ( @xmath102 ) on the later two dates . independent of the details of line blanketing , we can directly see the sed becoming redder , and hence implying cooler temperatures , by looking at the color evolution . in figure [ colorplot ] , we construct @xmath29@xmath48@xmath30 and @xmath30@xmath103 color curves . in each case , we interpolate the light curve in the bluer band to the epochs of observation for the noisier redder band . the @xmath29@xmath48@xmath30 color is consistent with a constant ( [email protected] mag ) before maximum light , but both colors became redder with time after maximum . we use the bb fits to construct a bolometric light curve , starting with the near - maximum - light sed ( phase @xmath393 d ) shown in figure [ sedplot ] . a trapezoidal integration of the observed fluxes from @xmath28 to @xmath53 gives a minimum bolometric luminosity ( @xmath104 ) near peak of ( [email protected])@xmath105 erg s@xmath6 . to account for flux emitted redward of our observations , we add the tail of the bb and extend the trapezoidal integration to long wavelengths , which provides only a 14% upward correction to the total flux . we repeat the process for the other two sed epochs . after including the bb tails , the derived bolometric luminosities are ( [email protected] ) , ( [email protected] ) , and ( [email protected])@xmath105 erg s@xmath6 at phases of @xmath393 , @xmath3912 , and @xmath3924 d , respectively . although we lack nir data at other epochs , we can use these fits to estimate the bolometric light curve in combination with the well - sampled ps1 light curves . for each of our three sed epochs , we define a multiplicative bolometric correction factor to obtain @xmath104 from @xmath106@xmath107 in the @xmath30 band . motivated by the lack of strong color evolution before maximum light , we apply the bolometric correction factor from @xmath393 d to all pre - maximum @xmath30 data points . for the post - maximum data , we use a smoothly varying time - dependent correction factor to evolve from the @xmath393 d value to the @xmath3924 d value . we repeat the process for the @xmath29 band , except that we use a smaller ( but constant ) bolometric correction factor before maximum light than the derived @xmath393 d value to better match the @xmath30 results and account for the small amount of @xmath29@xmath48@xmath30 color evolution between @xmath393 d and the pre - maximum data . the combined bolometric light curve is plotted in figure [ boloplot ] . a third - order polynomial fit to the peak of the bolometric luminosity curve gives a maximum value of ( [email protected])@xmath19 erg s@xmath6 , equivalent to 1.1@xmath410@xmath108 l@xmath12 , not including the uncertainty in the bolometric correction factors ( or potential extinction ) . to check that our derived maximum bolometric luminosity of ps1 - 10afxis realistic and that our procedure for extrapolating the sed to the nir has not introduced any significant error , we can take advantage of the fact that the shapes of the spectra we have of ps1 - 10afx and the derived seds are very similar to those of a sn ia near maximum light ( cf . figure [ uvcomplog ] ) . one of the most distant spectroscopically confirmed sne ia known , hst04sas , is at @xmath109 @xcite , consistent with the redshift of ps1 - 10afx . interpolating the f850lp ( @xmath3@xmath46 band ) light curve of hst04sas to a phase of zero days produces an estimate of the observed magnitude of @xmath110 mag ( ab ) , while fits to ps1 - 10afx show that @xmath30@xmath111 mag . assuming that the similarity of the seds requires no further correction factors and a typical sn ia peak bolometric luminosity of @xmath31.2@xmath410@xmath112 erg s@xmath6 ( e.g. , @xcite ) , the 3.8 mag difference between the objects corresponds to a peak bolometric luminosity for ps1 - 10afx of @xmath11310@xmath5 erg s@xmath6 , in excellent agreement with our value derived above . in fact , the peak of @xmath114 mag is more luminous than any of the hydrogen - poor objects collected by @xcite mag . ] . this statement is somewhat uncertain due to the heterogeneous and fragmentary nature of the available data for seds and bolometric correction factors for the objects in the literature . sne 2008es @xcite , 2005ap @xcite , and scp06f6 @xcite all have similar peak bolometric luminosities to ps1 - 10afx , but the exact comparison depends on the treatment of the bolometric correction for these uv - luminous sources . the only slsn of any type with a clearly higher peak luminosity is the peculiar hydrogen - rich transient css100217 @xcite . however , the sn nature of that object is still questionable due to possible confusion with the active galactic nucleus of its host galaxy . in addition to the high peak luminosity of ps1 - 10afx , another striking fact about the light curve is the rapid time evolution . despite the rolling survey nature of ps1 , our first detection is at a phase of only @xmath115 d , with non - detections prior to that ( figure [ boloplot ] ) . it is traditional in sn studies to find the rise time by fitting a fireball model of the form @xmath116 for @xmath117 at early times , where @xmath118 is the explosion time . for ps1 - 10afx , this gives a @xmath119 d ( figure [ boloplot ] ) . allowing the power - law index to vary gives a best fit with @xmath120 and @xmath121 d. regardless of the exact parameterization , the rise in @xmath30 is a factor of 3.5 in flux in the 7.4 d before peak . this apparent rise time is much faster than the fastest known slsne ( figure [ uplot ] ; @xcite ) . the decay timescale is also unusually fast . the fwhm of the bolometric light curve is only 18 d. the usual parameterization of sn light curve shapes , @xmath122@xmath123 , is the decline in magnitudes in 15 d after maximum light . the bolometric light curve of ps1 - 10afx has @xmath122@xmath123=0.95 mag . by comparison , only sn 1994i has a faster @xmath122@xmath123(@xmath124 ) in the collection of sne ib / c light curves of @xcite , although several additional objects had @xmath122@xmath123(@xmath125 ) that are comparable . figure [ uplot ] demonstrates that in @xmath87 , ps1 - 10afx also has a faster light curve decay than slsne of similar peak luminosity . integrating the observed bolometric light curve from the first detection at a phase of @xmath487.4 d to the last detection at @xmath3924 d provides a lower limit on the emitted energy of @xmath37@xmath410@xmath126 erg ( over only 31 d ) . reasonable extrapolations of the light curve to later times will increase the total to @xmath310@xmath15 erg , comparable to most other hydrogen - deficient slsne , but not notably high ( e.g. , @xcite ) . the fast timescale of ps1 - 10afx compensates for its high peak luminosity . the difficult constraints placed by the rise time on any model can be seen from the inferred emitting radius . assuming @xmath127 k as derived above , the bb radius ( @xmath128 ) of the photosphere at maximum light is 5.5@xmath129 cm . this is larger than for other hydrogen - poor slsne , which typically have inferred radii of @xmath310@xmath10 cm @xcite , as expected from the higher luminosity and lower @xmath101 of ps1 - 10afx . if @xmath102 is actually as low as the color temperature of @xmath3@xmath130 k , then the inferred radius at maximum light increases to 1.6@xmath131 cm . the most conservative set of numbers at maximum light ( @xmath128=5.5@xmath129 cm , @[email protected] d ) requires an expansion velocity of @xmath350,000 km s@xmath6 if @xmath118 is the true explosion time , in obvious conflict with the observed photospheric velocities seen in the spectrum . material moving at the ` synow`-derived velocity of @xmath8 km s@xmath6 will take 58 d to reach that radius , implying an explosion @xmath350 d before our first detection . there is an observed example of such an effect with the slsn 2006 oz . it exhibited a precursor plateau with a luminosity of @xmath32@xmath410@xmath112 erg s@xmath6 prior to rising to the main peak of the light curve , which @xcite interpreted as being due to a recombination wave in an oxygen - rich csm . a weighted average of the ps1 - 10afx non - detections between phases @xmath132 and @xmath133 d produces a mean luminosity of ( [email protected])@xmath410@xmath112 erg s@xmath6 . at the 2@xmath134 level , this non - detection is consistent with a hypothetical plateau being present at the level seen in sn 2006 oz , but unobserved due to the limiting magnitudes of our survey . at phases @xmath3912 and @xmath3924 d , the @xmath104 quoted above and @xmath101 shown in figure [ sedplot ] combine to produce bb radii of 1.4 and 1.3@xmath131 cm , respectively , although these are likely overestimated due to the underestimate of @xmath102 caused by line blanketing . notably , these values are consistent with the photosphere begining to recede through the ejecta ( in both the physical and comoving senses ) as they become optically thin . by contrast , the bb radii inferred for the interacting slsne 2006tf and 2006gy increased linearly with time at the same rate as the velocities measured from the spectra , consistent with an expanding optically - thick shell @xcite . although the measurement errors make it less certain , this also appears to be true for at least some sn 2005ap - like sne ( e.g. , @xcite ) . the kinetic energy of a powerful sn explosion provides an attractive reservoir of energy to power slsn light curves , but only if that energy can be efficiently converted to radiation . for ps1 - 10afx , this already requires an initial sn energy of at least a @xmath13510@xmath15 erg in order for the object to radiate 10@xmath15 erg and still have ejecta velocities as high as those observed . this generally requires a csm with a mass comparable to that of the ejecta and a radius comparable to the photospheric radius at maximum light in order to maximize efficiency and not lose energy to adiabatic expansion ( e.g. , @xcite ) . another attraction of interaction models is that the sn can explode well before the start of the optical rise of the sn if the csm is sufficiently dense to trap radiation from the shock @xcite . this could greatly alleviate the conflict described above between the rise time , ejecta velocity , and photospheric radius . the required csm densities in these models are much larger than for well - studied examples at lower luminosity . @xcite have identified two regimes of interest when the csm is this dense . if the radiation from the shock can escape while the shock is still interacting with the csm , then the light curve will be broad and exhibit a slow decay after maximum light . their model works well for the light curves of slsne like 2006gy . an additional consequence of this model is that there should be spectral signatures of the ongoing csm interaction , such as emission lines from the unshocked csm ( e.g. , @xcite ) . other luminous sne iin in the literature , such as sne 2008fz @xcite and 2008am @xcite , with their blue continua and strong balmer emission lines , can also be fit by similar models . the much faster light curve of ps1 - 10afx and lack of strong emission lines imply a different regime , one where the outer radius of the csm is comparable to or smaller than the diffusion length of photons from the shock . in this framework , the optical light we detect is then associated with the breakout of the shock from the outer radius of the csm . in the context of a dense wind with a density profile that scales with radius as @xmath136 out to a maximum radius of @xmath137 , where @xmath138 is a density normalization parameter , the total mass of the wind is @xmath139 . in the models of @xcite , @xmath138 can be estimated from the diffusion timescale as @xmath140 ( @xmath141/days ) g cm@xmath6 , if we assume an ionized he - rich composition . we can approximate the diffusion timescale @xmath141 by the rise time and set @xmath137 to the bb radius at maximum light to produce an estimate of the wind mass of @xmath143 m@xmath12 . these values are only approximate , as @xcite have shown that in the regime of interest , the analytic approximations are outside the range of validity and require true hydrodynamic calculations . models with shock breakout occurring in a dense , truncated wind can produce bolometric light curves with the desired peak luminosities and time scales , but it is unclear if this can be done consistently with the other information about the object . @xcite were able to approximately match the light curve shapes of the slsne 2010gx and 2005ap , but the derived @xmath101 were hard to match to the observations . a theoretical caveat is that the observed color temperature could deviate from @xmath102 if the shock is not in equilibrium when it breaks out @xcite . this deviation is caused by the nature of the opacity , and usually implies that the color temperature is hotter than the true @xmath102 ( e.g. , see @xcite for models of 2006gy ) , which would in turn require that the photosphere be at even larger radii than we have found . this effect might then conflict with the fast timescale of the light curve , which otherwise is indicative of a more compact csm @xcite , but detailed modeling with radiative transfer is needed to correctly evaluate these concerns . further , a key component of the rising sn luminosity in the shock breakout models of both @xcite and @xcite is that the temperature rises to a maximum near the peak of the light curve and declines thereafter . although we lack full sed information on the rise of the light curve for ps1 - 10afx , the observed @xmath144 color shows no evolution prior to the peak of the light curve ( figure [ colorplot ] ) . despite the interpretation uncertainties introduced by the possibilities of line blanketing or deviations from equilibrium , the evolution of @xmath101after maximum light results in a noticeable reddening of the @xmath145 color . if @xmath101 were rising by a similar amount while the light curve rose to its peak , we would have been able to detect it . a separate uncertainty is the origin of this csm , which could not be produced by steady stellar winds due to the high required mass - loss rates and the unexplained truncation at @xmath137 . most theoretical work has focused on winds , but the material could also be in a shell @xcite . @xcite invoked a pulsational pair instability to produce shells of matter at the appropriate distances , but these require very massive stars . other possibilities include common envelope ejection @xcite and instabilities in the late stages of stellar evolution @xcite . another challenge for csm interaction being the dominant contributor to the luminosity of ps1 - 10afx is the appearance of the spectra . interaction - powered sne convert the kinetic energy of the explosion into uv / optical continuum radiation , frequently reprocessing some of the continuum into strong emission lines from the dominant constituents of the csm external to the expanding shock . the spectral comparisons in figure [ uvcomp ] show several examples of objects consistent with this picture . however , ps1 - 10afx is even more luminous than those objects and yet shows p - cygni features originating in the sn ejecta , including a deep absorption comparable to that seen in sn 1994i . the lack of strong emission lines from the csm can be explained if the material has a sharp outer radius , but interaction sufficiently luminous to dominate the light curve should dilute and weaken spectral features from the ejecta due to the `` top - lighting '' effect @xcite , if not completely hide emission from the unshocked ejecta . for example , ptf 09uj is a candidate for being a lower - luminosity version of a shock breakout through a dense ( hydrogen - rich ) wind and , as expected , its spectra are very blue and lack strong p - cygni features @xcite . it is not clear how strong csm interaction could dominate the luminosity of ps1 - 10afx without leaving spectral signatures . in summary , existing csm interaction models can match the gross photometric properties of ps1 - 10afx ( peak luminosity , timescales ) , but only if the models have the freedom to assume the necessary csm structure . the origin of such unusual csm is not explained _ a priori _ in these models . in addition , the details of the sed , color evolution , and emitted spectra appear to be in conflict with the observations , although more detailed radiative transfer calculations are necessary to be confident in the model predictions , which are largely based on bb assumptions . normal hydrogen - deficient sne have light curves powered by the diffusion of energy deposited by the radioactive decay of @xmath21ni and @xmath21co . the light curves of most slsne - i appear to fade too rapidly relative to their peak luminosities to have radioactively - powered ejecta @xcite , and ps1 - 10afx is no exception . with a rise time of 12 d , the peak bolometric luminosity requires a nickel mass of @xmath314 m@xmath12 . however , making standard assumptions about the structure of the ejecta @xcite allows us to use the photospheric velocity of 11,000 km s@xmath6 and light curve timescale of 12 d ( assuming that the diffusion timescale is comparable to the rise time ) to estimate the ejecta mass as @xmath32 m@xmath12 , clearly in contradiction of the high nickel mass . another potential power source suggested for slsne is the spindown energy from a newly - born magnetar @xcite . the high luminosity and fast rise of ps1 - 10afx stretch the existing models to the bounds of plausibility . inspection of figures 4 and 5 of @xcite shows that the light curve of ps1 - 10afx requires a magnetar solution to have a very fast initial spin period ( @xmath146 ms , near the breakup speed of a neutron star ) , a low ejecta mass ( @xmath31 m@xmath12 ) , and a high magnetic field ( @xmath14710@xmath148 g ) . this is because the high luminosity requires a high spindown power , while the fast light curve requires both fast spindown times and short diffusion timescales . we examined magnetar - powered light curves using the formalism of @xcite and found parameters that produced light curves that approximately matched the shape of the bolometric light curve of ps1 - 10afx . one such model is plotted in figure [ boloplot ] . however , it is not clear whether this fit should be regarded as more than a numerical curiosity because the physical assumptions underlying the model break down . in the parameter range appropriate for ps1 - 10afx , the magnetar dipole spindown energy ( 5@xmath410@xmath15 ( @xmath149/2 ms)@xmath150 erg ) dominates over the initial sn kinetic energy ( assumed to be 10@xmath15 erg ) . the @xcite models make an ansatz that the entire spindown energy of the magnetar is simply thermalized spherically at the base of the sn ejecta and deposited into the internal energy . because this process accelerates and heats the ejecta , the final kinetic energy , the peak luminosity , and the temperature are coupled . the model plotted in figure [ boloplot ] predicts expansion velocities of @xmath925,000 km s@xmath6 , in contradiction of the observed spectrum . in addition , the model predicts a radius at maximum light ( from the ejecta velocity and rise time ) that when combined with the luminosity implies a bb temperature in excess of @xmath151 k ( with strong evolution at other times ) . this contradicts the observed @xmath101 near maximum light and the constant red @xmath29@xmath48@xmath30 color during the rise of the light curve . we find that increasing the assumed sn kinetic energy relative to the spindown energy does not alleviate these issues because of the difficulty in simultaneously satisfying the constraints from the expansion velocity seen at maximum light ( which implies a normal energy - to - mass ratio , and hence high mass if the initial sn energy is high ) with the fast diffusion timescale ( which requires a low mass ) . the models of @xcite start from massive star progenitors and proceed through the explosions , but they only probe slower rotating , and hence less energetic , magnetars than those required for ps1 - 10afx . @xcite also examined explosion models for type ib / c sne with the addition of a central energy source that mimicked the effect of a magnetar . they followed the effects of a central energy source on the ejecta produced by the explosion models and generated light curves and spectra . if the central energy source deposited @xmath310@xmath15 erg of energy , the resulting sne had high velocities and were blue at maximum light , which appears promising for explaining the sn 2005ap - like class of slsne , but is very different from ps1 - 10afx . as noted above , the light curve of ps1 - 10afx requires even more energy input from the magnetar . therefore , we disfavor magnetar models at this time . @xcite have proposed a scenario that shares some similarities with the magnetar model in that their model has a post - explosion internal energy source , in this case provided by fallback accretion onto a newly formed compact remnant , likely a black hole . their suite of models can only match the fast timescale and high peak luminosity of ps1 - 10afx if the progenitor is a blue supergiant and the initial sn explosion energy is extremely low with an ejecta mass of @xmath31 m@xmath12 . their hydrogen - poor models do not reach the same peak luminosities , which is a problem for ps1 - 10afx . in addition , these models have the same problem as the magnetar ones in simultaneously satisfying the high peak luminosity and relatively cool photospheric temperature . in all of the preceding discussion , we have necessarily assumed spherical symmetry in order to convert our observations into intrinsic quantities . however , core - collapse sne are all known to be aspherical at some level ( e.g. , @xcite ) . the most basic effect of asphericity is to increase the uncertainty in @xmath104 and dilute the connection between observables such as the rise time and physical quantities such as the diffusion timescale . the broad - lined sne ic associated with grbs have been extensively studied in the context of aspherical models , with different studies coming to conflicting conclusions for the same object ( e.g. , sn 1998bw : @xcite ) . the models of @xcite for sn 1998bw predict that lines of sight along the major axis of a bipolar outflow will see light curves with faster rise times and higher luminosities than average , perhaps indicating a general trend with relevance for ps1 - 10afx . however , despite the high explosion energies ( @xmath4910@xmath152 erg ) and fairly high nickel masses ( @xmath30.4 m@xmath12 ) assumed in their models , the maximum peak luminosity was still an order of magnitude lower than for ps1 - 10afx . in addition , the spectral features of ps1 - 10afx are not as blended as those of broad - lined sne ic ( figure [ uvcomp ] ) . we conclude that an aspherical sn 1998bw - like model is unlikely to produce an object that resembles ps1 - 10afx unless some other ingredient is added to the models . another possibility in a csm interaction scenario is that the material with which the sn is interacting is distributed in an aspherical fashion , perhaps allowing a distant observer to see both the luminous continuum emission from the interaction region and the sn ejecta . however , such models still require the combination of the covering fraction of the csm , the radiative efficiency factor , and the sn kinetic energy to be sufficiently high to produce the @xmath310@xmath15 erg of optical emission that we measure . the slsn models of @xcite invoke remnant protostellar disks around massive stars , but do not simultaneously fit the high peak luminosity and the fast timescale of ps1 - 10afx . in addition , such a model has similar difficulties as spherical csm interaction models in explaining the strength of the p - cygni features in the observed spectra if there is continuum emission from an interaction region that dilutes the strength of the spectral features from the ejecta . we now consider whether gravitational lensing could result in an object with the spectrum and light curve shape of a normal sn , but artificially boosted in luminosity . however , the peak luminosity of ps1 - 10afx is a factor of @xmath950 higher than normal sne ic , requiring an extremely high magnification factor despite the lack of an obvious lens . while this work was being refereed , a preprint by @xcite expanded on this hypothesis further and proposed that ps1 - 10afx was instead a normal sn ia and therefore had to be lensed by a magnification factor of @xmath320 . there are several constraints against the lensing hypothesis , starting with the lack of an apparent lens . in standard @xmath22cdm cosmologies , the optical depth to lensing with amplification factors @xmath4910 is very small , 10@xmath153 @xmath48 10@xmath154 for sources at the redshift of ps1 - 10afx ( e.g. , @xcite ) , and the mass distribution of the lenses for those extreme magnifications is peaked at cluster - scale haloes of @xmath310@xmath148 m@xmath12 @xcite . however , there is no foreground galaxy cluster visible in the ps1 images , as @xcite also concluded on the basis of a clustering analysis of independent canada - france hawaii telescope images of the field . the source present along the line of sight to ps1 - 10afx exhibits [ ] emission and the sed ( including a balmer break ) of a galaxy at the same redshift as the sn . the [ ] and rest - frame uv continuum luminosities produce similar estimates of the sfr . these observations are all consistent with the continuum flux from the source along the line of sight being dominated by , if not entirely due to , emission from the host galaxy of ps1 - 10afx , and leave little room for a contribution from a hypothetical lens galaxy at lower redshift . these constraints led @xcite to propose the existence of a dark lens from a previously unknown population , such as a free - floating black hole or a dark matter halo with few baryons . this provocative conclusion is based on the premise that ps1 - 10afx has to be classified as a normal sn ia and therefore the intrinsic luminosity is known . we demonstrated above that the spectra of ps1 - 10afx are similar to those of both sne ia and ic near maximum light ( section 4 ) . we also note that the presence of an absorption feature from @xmath636355 does not uniquely classify an object as a sn ia because that line is also present in some sne ic . as an experiment , we took near - maximum light spectra of an unambiguous nearby sne ic that exhibits features , sn 2004aw @xcite , resampled the data , and added noise to match the s / n of our ps1 - 10afx fire data that show the possible absorption feature near 6100 . these noisy spectra were then run through the supernova identification ( snid ) code of @xcite , using the updated templates of @xcite , with care taken to exclude templates from the same object . we found that typically all of the top 20 best matching templates are sne ia . this is in part due to the biased sample of spectral templates used by sn classification codes , where sne ia are largely overrepresented . @xcite also claim that the light curve of ps1 - 10afx is a good match for sne ia . however , we note that in order to get a good reduced-@xmath155 for their fit , they had to artificially increase the error bars in the nir and exclude the @xmath51-band data . furthermore , all of the ps1 - 10afx observations prior to a phase of @xmath156 d fall below the sn ia prediction , including the @xmath30 non - detection at @xmath115 d that is @xmath30.7 mag below their model . this is a consequence of the fast rise time of ps1 - 10afx , which we measure above to be @xmath157 d if we assume that the rise scales as @xmath116 , and @xmath158 d if we relax that constraint . although spectroscopically normal sne ia show a small dispersion in their rise times , objects with normal light curve shapes have rise times of 16@xmath4818 d in @xmath159 ( e.g. , @xcite ) . @xcite demonstrated that the rise in @xmath88 ( closer in rest wavelength to our @xmath30 observations ) was 2.3 d shorter , but still significantly longer than observed for ps1 - 10afx . at phases of @xmath160 to @xmath161 d , normal sne ia emit fluxes about 20@xmath4840% of the peak @xcite , but that is about 4@xmath134 above our non - detections ( figure [ boloplot ] ) . in summary , although sne ia have reasonably similar spectra and light curve shapes to ps1 - 10afx , the details of the rise time are not a good fit for sne ia and the spectra of the unambiguous sn ic 2004aw are also similar to sne ia when degraded to our low s / n ratio . therefore , an identification of ps1 - 10afx with normal sne ia is not required by the data , and there is no need to invoke a new population of unseen dark gravitational lenses . instead , we accept the lack of an apparent lens at face value and conclude that ps1 - 10afx was actually an unusually luminous hydrogen - deficient sn . we present multiwavelength observations of ps1 - 10afx at redshift @xmath0 , perhaps the most luminous sn yet discovered . the combination of observables presents strict constraints on any theoretical interpretation . these are : * a peak bolometric luminosity of 4.1@xmath410@xmath5 erg s@xmath6 * an observed rise time of @xmath312 d * a fast light curve decay , with @xmath122@xmath123= @xmath162 mag * at least 7@xmath410@xmath126 erg of optical radiation emitted , with a total likely closer to 10@xmath15 erg * red color near maximum light ( @xmath163 k ) , although @xmath102 may be closer to @xmath99 k * constant uv color before maximum light * photospheric velocities near maximum light of @xmath3@xmath8 km s@xmath6 * spectra that most closely resemble normal sne ic , with deep p - cygni absorption * photometric and spectroscopic evidence for some line blanketing in the nuv from iron - peak elements , although not as much as for sne ia * a massive ( @xmath32@xmath410@xmath13 m@xmath12 ) host galaxy that is unlikely to have an extremely low metallicity we surveyed existing models for slsne and found that none were acceptable . in particular , the large inferred @xmath128 near maximum light is very hard to reconcile with the fast observed rise time and the measured photospheric velocities because the sn ejecta need too much time to reach such large radii . the magnetar models that match the peak luminosity and rise time do so by producing a much higher temperature at a smaller radius . these flaws are generic to models powered by internal energy sources , although if the onset of the internal energy source can be sufficiently delayed after the initial explosion , then maybe the conflict between the velocities and radii can be avoided . shock breakout scenarios invoking dense csm provide a promising solution by allowing the sn to explode well before the optical emission is detectable . if the csm is sufficiently dense out to @xmath35@xmath410@xmath10 cm , then the light curve will only rise after the ejecta have had time to reach that radius . however , these models leave the normal sn ic - like spectrum unexplained and are in apparent conflict with the observed color evolution before maximum light . more detailed radiative transfer calculations of the models are necessary to know whether these flaws can be avoided . in addition , it is not clear how to produce the special csm structure that has to be assumed ( a truncated wind or a shell ) . ps1 - 10afx was initially identified as an object of interest due to its unusual colors ( @xmath164 2 mag at peak ) and would not have been found without the ps1 observation strategy , which includes regular @xmath29 and @xmath30 observations as part of the search . at redshifts below @xmath31 , an object like ps1 - 10afx would have a strong detection in @xmath28 and would not stand out without knowing the redshift . unlike many of the sn 2005ap - like sne , the time - dilated light curve of a ps1 - 10afx - like sn at lower redshift would not be unusually long and would not be associated with the attention - grabbing lack of a visible host galaxy . however , ps1 - 10afx was more than a magnitude brighter than its host galaxy , which potentially provides a selection criterion to increase the odds of finding similar objects in the future . in the sample of slsne found by ps1 , sn 2005ap - like objects @xcite outnumber ps1 - 10afx by at least an order of magnitude . our poor understanding of spectroscopic incompleteness and the relevant selection effects precludes a more precise estimate of rates at this time . an additional complication to be considered is the role of metallicity . the massive host of ps1 - 10afx is in contrast with the low - mass @xcite and low - metallicity @xcite hosts of known hydrogen - poor slsne . the natal composition of the progenitor could have an indirect effect on the explosion through stellar evolutionary processes or it could directly affect the appearance through the opacity in the outer ejecta ( could sn 2005ap - like events with higher metal abundances exhibit line - blanketed spectra like ps1 - 10afx ? ) . most of the published slsne that lack hydrogen in their spectra are similar to sn 2005ap and scp06f6 @xcite . sn 2007bi @xcite and a couple of as - yet unpublished objects @xcite were the only known exceptions . they exhibited rather different spectra and had very slow light curve decays , which were interpreted by @xcite to be the result of decay of a large amount of radioactive @xmath21ni produced in a pair - instability explosion . this result has been challenged on theoretical grounds by @xcite , who show that the observations do not match theoretical expectations for pair - instability sne . with the discovery of ps1 - 10afx , we now have further evidence that the most luminous sne are a heterogeneous lot , even when just considering the objects lacking hydrogen . a key question for future investigations is whether these diverse outcomes of stellar evolution can be produced by variations on a single underlying physical model or whether multiple pathways exist to produce sne of these extraordinary luminosities . ps1 - 10afx differs from existing slsne - i in almost every observable except for the peak luminosity . it is hard to understand how all of these differences could be produced with only small modifications to a single model , and may indicate that ps1 - 10afx is the first example of a different channel for producing slsne . this paper includes data gathered with the 6.5-m magellan telescopes located at las campanas observatory , chile . we thank the staffs at ps1 , gemini , magellan , and the vlt for their assistance with performing these observations . we are grateful for the grants of dd time at gemini and the vlt . we acknowledge j. strader for obtaining some fourstar observations , as well as the mmirs observers who helped obtained some of the data here during the fall 2010 queue observations : m. kriek , i. labbe , j. roll , d. sand , and s. wuyts . sjs acknowledges funding from the european research council under the european union s seventh framework programme ( fp7/2007 - 2013)/erc grant agreement n@xmath165 291222 . the pan - starrs1 surveys ( ps1 ) have been made possible through contributions of the institute for astronomy , the university of hawaii , the pan - starrs project office , the max - planck society and its participating institutes , the max planck institute for astronomy , heidelberg and the max planck institute for extraterrestrial physics , garching , the johns hopkins university , durham university , the university of edinburgh , queen s university belfast , the harvard - smithsonian center for astrophysics , the las cumbres observatory global telescope network incorporated , the national central university of taiwan , the space telescope science institute , and the national aeronautics and space administration under grant no . nnx08ar22 g issued through the planetary science division of the nasa science mission directorate . some observations were obtained under program ids gn-2010b - q-5 ( pi : berger ) , gs-2010b - q-4 ( pi : berger ) , and gn-2010b - dd-2 ( pi : chornock ) at the gemini observatory , which is operated by the association of universities for research in astronomy , inc . , under a cooperative agreement with the nsf on behalf of the gemini partnership : the national science foundation ( united states ) , the science and technology facilities council ( united kingdom ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , ministrio da cincia , tecnologia e inovao ( brazil ) and ministerio de ciencia , tecnologa e innovacin productiva ( argentina ) . some observations were collected at the european organisation for astronomical research in the southern hemisphere , chile under ddt programme 286.d-5005 ( pi : smartt ) . the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc . some of the archival data presented in this paper were taken under programs go-5623 and go-9114 ( pi : kirshner ) and were obtained from the mikulski archive for space telescopes ( mast ) . stsci is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . some of the computations in this paper were run on the odyssey cluster supported by the fas science division research computing group at harvard university . tonry , j. , & onaka , p. 2009 , advanced maui optical and space surveillance technologies conference , proceedings of the advanced maui optical and space surveillance technologies conference , ed . : s. ryan , p.e40 .	we present the discovery of ps1 - 10afx , a unique hydrogen - deficient superluminous supernova ( slsn ) at redshift @xmath0 . the light curve peaked at @xmath1 mag , making ps1 - 10afx comparable to the most luminous known sne , with @xmath2 mag . our extensive optical and near - infrared observations indicate that the bolometric light curve of ps1 - 10afx rose on the unusually fast timescale of @xmath312 d to the extraordinary peak luminosity of 4.1@xmath410@xmath5 erg s@xmath6 ( @xmath7 mag ) and subsequently faded rapidly . equally important , the sed is unusually red for a slsn , with a color temperature of @xmath36800 k near maximum light , in contrast to previous hydrogen - poor slsne , which are bright in the ultraviolet ( uv ) . the spectra more closely resemble those of a normal sn ic than any known slsn , with a photospheric velocity of @xmath3@xmath8 km s@xmath6 and evidence for line blanketing in the rest - frame uv . despite the fast rise , these parameters imply a very large emitting radius ( @xmath95@xmath410@xmath10 cm ) . we demonstrate that no existing theoretical model can satisfactorily explain this combination of properties : ( i ) a nickel - powered light curve can not match the combination of high peak luminosity with the fast timescale ; ( ii ) models powered by the spindown energy of a rapidly - rotating magnetar predict significantly hotter and faster ejecta ; and ( iii ) models invoking shock breakout through a dense circumstellar medium can not explain the observed spectra or color evolution . the host galaxy is well detected in pre - explosion imaging with a luminosity near @xmath11 , a star formation rate of @xmath315 m@xmath12yr@xmath6 , and is fairly massive ( @xmath32@xmath410@xmath13 m@xmath12 ) , with a stellar population age of @xmath310@xmath14 yr , also in contrast to the young dwarf hosts of known hydrogen - poor slsne . ps1 - 10afx is distinct from known examples of slsne in its spectra , colors , light - curve shape , and host galaxy properties , suggesting that it resulted from a different channel than other hydrogen - poor slsne .
one of eliashberg s foundational achievements is his pioneering work on the existence theory of stein manifolds . these manifolds have been studied by complex analysts for most of the past century , and so have many equivalent definitions . the definition that most immediately indicates their fundamental nature is that they are complex affine analytic varieties , i.e. , complex manifolds that embed holomorphically as closed subsets of @xmath4 . the importance of stein manifolds mandates the study of two basic existence questions . most obviously , we should ask which abstract smooth manifolds admit stein structures . however , we can also work inside a fixed ambient complex manifold @xmath5 . every open subset of @xmath5 is itself a complex manifold , so we can ask which open subsets of @xmath5 are stein . the study of such stein open subsets has had a long and continuing history , already in the special case @xmath6 . for example , if @xmath5 is @xmath7 ( or any other stein manifold ) , then its stein open subsets @xmath8 have been characterized as being _ domains of holomorphy _ , meaning that they satisfy a certain maximality condition for extending holomorphic functions . ( specifically , for every connected , open @xmath9 extending outside of @xmath8 , and every component @xmath10 of @xmath11 , there are holomorphic functions on @xmath8 whose restriction to @xmath10 does not extend to @xmath12 . ) since the stein condition can be destroyed by a tiny perturbation of @xmath8 , we reformulate the ambient existence question to put it within reach of a topological answer : which open subsets of @xmath5 can be made stein by an _ isotopy _ , i.e. , a homotopy of the inclusion map through embeddings ? ( we define embeddings to be 1 - 1 immersions , but do not require them to be proper , so we are asking when an open set can be deformed into a new open set whose complex structure inherited from @xmath5 is stein . ) building on eliashberg s work of the late 1980s , we can now completely answer both the abstract and ambient existence questions . in complex dimension @xmath13 , eliashberg completely characterized those manifolds admitting stein structures by a simple differential topological condition . a similar statement can be obtained in the ambient setting . when @xmath3 , one obtains the corresponding characterizations by imposing additional delicate conditions . alternatively , one can eliminate these extra conditions by passing to the topological category , i.e. , working up to homeomorphism or up to isotopy through topological ( @xmath14 rather than @xmath15 ) embeddings , and invoking freedman theory . as an application , one obtains domains of holomorphy in @xmath16 realizing uncountably many diffeomorphism types of exotic @xmath17s . more generally , every domain of holomorphy in @xmath1 is topologically isotopic to uncountably many other diffeomorphism types of such domains ( all of which are homeomorphic ) . this paper surveys published and pending literature to give a unified view across all dimensions of these consequences of eliashberg s work . the abstract and ambient results are presented in sections [ abstract ] and [ ambient ] , respectively , and their proofs are sketched in section [ proofs ] . section [ onion ] reexamines the ambient theory from several other viewpoints , replacing isotopy by the sharper notion of ambient isotopy and discussing when embedded cw - complexes can be described ( after isotopy ) as nested intersections of stein open subsets . in the 4-dimensional topological setting , this frequently yields an uncountable neighborhood system realizing uncountably many exotic smooth structures . section [ appendix ] is an appendix proving several lemmas that are useful in earlier sections . to motivate eliashberg s work , we first consider necessary conditions for the existence of a stein structure . since a stein manifold @xmath8 is complex , it comes endowed with an almost - complex structure , that is , its tangent bundle @xmath18 is a complex vector bundle . we capture this structure with the bundle automorphism @xmath19 of @xmath18 given by fiberwise multiplication by @xmath20 . if such an almost - complex structure @xmath19 comes from a complex structure , then it uniquely determines the latter . however , we usually consider almost - complex structures only up to homotopy . the homotopy classification of complex bundle structures on @xmath18 ( or on any real vector bundle ) is a standard problem in algebraic topology . thus , it is reasonable to refine our original abstract existence question by asking which almost - complex structures on a given manifold ( if any exist ) are homotopic to stein structures . in addition to having an almost - complex structure , a stein manifold @xmath8 satisfies another classical condition : it admits an exhausting morse function ( i.e. a smooth , proper map @xmath21 whose critical points are nondegenerate ) such that each critical point has index @xmath22 . ( thus , @xmath8 has real dimension @xmath23 but the homotopy type of an @xmath2-complex . ) surprisingly , eliashberg s theorem asserts that these necessary conditions are also sufficient . [ absn ] @xcite . for @xmath13 , let @xmath8 be a @xmath23-manifold with an almost - complex structure @xmath19 . then @xmath19 is homotopic to a stein structure if and only if @xmath8 admits an exhausting morse function whose critical points all have index @xmath24 . in particular , @xmath8 admits a stein structure if and only if it admits an almost - complex structure and suitable morse function . the original proof appeared in @xcite , but a simpler and more expository version will be given in the book @xcite . we will discuss the latter version in section [ proofs ] . it is sometimes useful to replace morse theory by the language of handlebodies ( e.g. @xcite ) . recall that a _ handle of index @xmath25 _ , or _ @xmath25-handle _ , is a copy of @xmath26 attached to the boundary of an @xmath27-manifold along its attaching region @xmath28 . a ( topological or smooth ) _ handlebody _ @xmath29 is a manifold built from the empty set by successively attaching handles ( necessarily beginning with a 0-handle ) . in the smooth case , we glue each handle by a smooth embedding of its attaching region , then smooth the resulting corners . ( these corners occur along @xmath30 . by uniqueness of tubular neighborhoods , we can canonically identify a neighborhood of the corner locus with the product of @xmath30 with @xmath31 minus the open first quadrant , then smooth using the essentially unique smoothing of the last factor . ) handlebodies are necessarily _ locally finite _ in the sense that only finitely many handles attach to a given handle . ( attaching infinitely many handles to a compact boundary region would cause clustering destroying the manifold structure . ) we also require handlebodies to be _ morse ordered _ , so each @xmath25-handle attaches to a subhandlebody consisting of handles with index @xmath32 . for compact handlebodies ( i.e. those with finitely many handles ) , this costs no generality by a general position argument , but infinite handlebodies are restricted , e.g. requiring infinitely many 0-handles . we impose this constraint since the cores @xmath33 of the handles will then fit together inside the handlebody to give an embedded cw - complex @xmath34 . ( note that a cw - complex is morse - ordered by definition and locally finite if it embeds in a manifold . ) if there are no @xmath27-handles , then @xmath35 is homeomorphic ( diffeomorphic in the smooth case ) to @xmath36 $ ] , realizing @xmath29 as the mapping cylinder of a map @xmath37 . to relate morse functions with handlebodies , suppose @xmath38 is a critical value of an exhausting morse function @xmath39 , corresponding to a unique critical point @xmath40 in @xmath41\subset u$ ] . then @xmath42 $ ] is obtained from @xmath43 $ ] by adding a collar @xmath44 to the boundary and then attaching a handle with the same index as @xmath40 ( e.g. @xcite ) . if there are only finitely many critical points , we may then identify @xmath8 with the interior of a handlebody by collapsing out the intervening collars . in our situation , though , there may be infinitely many critical points ( although only finitely many in each @xmath45 $ ] since @xmath46 is proper ) . collapsing the collars may no longer yield a morse - ordered handlebody . one could remedy this by simply adding collars into the definition of a handlebody , but only at the expense of losing morse - ordering or local finiteness . this would complicate the statements of some of our theorems , and lose the natural cw cores and mapping cylinder structure . instead , we show in the appendix ( lemma [ decomp ] ) that suitable subdivision allows us to identify @xmath8 with the interior of a handlebody , even if @xmath46 has infinitely many critical points , and the maximal index of the handles is that of the critical points of @xmath46 . applying this to theorem [ absn ] , we conclude : [ absn1 ] under the hypotheses of eliashberg s theorem , @xmath19 is homotopic to a stein structure if and only if @xmath8 is diffeomorphic to the interior of a handlebody whose handles all have index @xmath24 . as we will see in section [ proofs ] , eliashberg constructs his stein structure from @xmath19 by induction on increasing critical values , or from the other viewpoint , by induction on handles . the induction step works in all cases except for 4-dimensional 2-handles , resulting in a proof for all @xmath47 . in the problematic case , the induction step may or may not work , depending on the framing of the 2-handle , so the following theorem results for stein _ surfaces _ ( i.e. the @xmath3 case ) . [ abs2 ] ( eliashberg . ) an oriented 4-manifold admits a stein structure if and only if it is diffeomorphic to the interior of a handlebody whose handles all have index @xmath48 , and for which each 2-handle is attached along a legendrian knot ( in the standard contact structure on the relevant boundary 3-manifold ) with framing obtained from the contact framing by adding one left twist . for terminology and additional explanation , see section [ proofs ] . the main observation there is that each 2-handle is attached to a finite union of 0- and 1-handles , whose boundary @xmath49 admits essentially a unique tight contact structure , with respect to which we can define legendrian knots and contact framings . theorem [ abs2 ] was essentially known to eliashberg when he wrote @xcite , but it was not contemporaneously published . we have sharpened it slightly , using the appendix ( lemma [ steindec ] ) to write it in the context of morse - ordered handlebodies without collars . ( this sharpening is what makes a concise statement possible when there are infinitely many handles . from the morse viewpoint , one must build the stein structure on the possibly complicated manifold below the level of a new 2-handle before the framing condition makes sense . ) for further discussion and applications of the theorem , see e.g. @xcite , @xcite , @xcite or @xcite . note that a handlebody with all indices @xmath50 has the homotopy type of a 2-complex , so if it is oriented , it automatically admits almost - complex structures ( respecting the given orientation ) . thus it suffices to hypothesize an orientation rather than an almost - complex structure . on the other hand , the almost - complex structure is completely determined by the legendrian link along which the 2-handles are attached ( see section [ proofs ] ) . unlike in other dimensions , there are manifolds admitting stein structures realizing some , but not all , homotopy classes of almost - complex structures . ( in fact , this occurs for every stein surface with @xmath51 , by the adjunction inequality from seiberg - witten theory @xcite . ) we now have a complete characterization , in all dimensions , of which manifolds admit stein structures , and which homotopy classes of almost - complex structures are so realized . however , the criterion for 4-manifolds is much harder to apply than in other dimensions . furthermore , if we ignore the delicate framing condition , we immediately run into trouble . the simplest handlebody interior requiring a 2-handle is @xmath52 . while this admits a complex structure ( @xmath53 ) , it admits no stein structure , because it contains a homologically essential 2-sphere with trivial normal bundle . ( by seiberg - witten theory , every smoothly embedded essential sphere in a stein surface has normal euler number @xmath54 @xcite . ) to simplify the 4-dimensional characterization , we apply the fundamental principle of freedman @xcite , @xcite , that high - dimensional differential topology frequently works in dimension 4 after we give up smoothness and enter the purely topological world . we replace the problematic 2-handles by more flexible _ casson handles _ , which are homeomorphic to the open 2-handle @xmath55 but have exotic smooth structures . we then recover a precise analog of eliashberg s theorem up to homeomorphism . [ abstop ] @xcite . an oriented 4-manifold is homeomorphic to a stein surface if and only if it is homeomorphic to the interior of a handlebody whose handles all have index @xmath48 . every homotopy class of almost - complex structures on such a handlebody interior is realized by an orientation - preserving homeomorphism to a stein surface . note that 4-dimensional topological handlebodies are uniquely smoothable , since they are made by gluing along 3-manifolds and the gluing maps are uniquely smoothable . thus , it does not matter whether the handlebody ( or 4-manifold ) in the theorem is topological or smooth . to see why a homeomorphism determines a bijection of homotopy classes of almost - complex structures ( which can be reinterpreted as spin@xmath56-structures since the base is essentially a 2-complex ) , see @xcite . one way to interpret this result is that eliashberg s theorem holds in all dimensions , provided that when @xmath3 we are allowed to change the smooth structure on the manifold . the resulting smooth structures have peculiarities typical to dimension 4 . such stein surfaces are typically not diffeomorphic to the interior of any compact 4-manifold , even if the original handlebody is finite . furthermore , a single handlebody often corresponds to uncountably many diffeomorphism types of stein surfaces . we now return to the question of which open subsets @xmath8 of a fixed complex surface @xmath5 are isotopic to stein open subsets . we wish to find characterizations analogous to the theorems of the previous section . note that @xmath8 inherits a complex structure from @xmath5 , and any isotopy of @xmath8 induces a homotopy of complex structures ( which we now view as almost - complex structures ) . thus , we should ignore our previous hypotheses regarding almost - complex structures , since they are now built into the topological setup . we are left with hypotheses regarding morse functions or handlebodies , and these turn out also to suffice in the ambient setting . we first state the analog of theorem [ absn ] and corollary [ absn1 ] . [ ambin ] ( eliashberg . ) for @xmath47 , an open subset @xmath8 of a complex @xmath2-manifold @xmath5 is smoothly isotopic to a stein open subset if and only if it admits an exhausting morse function whose critical points all have index @xmath24 , or equivalently , it is diffeomorphic to the interior of a smooth handlebody with all indices @xmath24 . our requirement that handlebodies be morse - ordered and without collars is again convenient here , for we can think of embedded handlebodies as tubular neighborhoods of smoothly embedded cw - complexes . we investigate this correspondence more deeply in section [ onion ] . theorem [ ambin ] is implicit in @xcite in the sense that it follows by the method of proof of theorem 1.3.6 of that paper . however , it does not seem to have previously appeared in the above form . there is essentially a complete proof in a paper of forstneri and slapar @xcite . their main objective was to extend eliashberg s work to make a manifold stein while a preassigned continuous map to some complex manifold becomes homotopic to a holomorphic map . if one erases the parts concerning the auxiliary map , however , one essentially obtains the proof of theorem [ ambin ] implicit in @xcite . ( more precisely , one immediately obtains theorem [ ambin ] by applying theorem 1.2 of @xcite to the complex manifold @xmath8 , setting the auxiliary complex manifold equal to a point . ) next we consider the ambient analog of theorem [ abs2 ] . we have already seen that theorem [ ambin ] fails when @xmath3 . ( consider the obvious embedding @xmath59 . ) the problem is that the manifold @xmath8 may not even abstractly admit a stein structure . the following theorem bypasses this difficulty . [ ambi2 ] @xcite . an open subset @xmath8 of a complex surface is smoothly isotopic to a stein open subset if and only if the induced complex structure on @xmath8 is homotopic ( through almost - complex structures on @xmath8 ) to a stein structure on @xmath8 . thus , the difficulty of making open subsets stein by smooth isotopy in this dimension is already contained in the problem of abstractly creating a stein structure , and is addressed by theorem [ abs2 ] . [ r4 ] let @xmath60 be one of the exotic @xmath0s of example [ abstract](b ) . we have already seen that @xmath8 admits a stein structure . but @xmath8 is contractible , so it has a unique homotopy class of almost - complex structures compatible with the given orientation . by arranging the identification of @xmath0 with @xmath1 to respect this orientation , we guarantee that the induced complex structure on @xmath8 is homotopic to a stein structure . since a smooth isotopy preserves the diffeomorphism type of the domain , theorem [ ambi2 ] exhibits a domain of holomorphy in @xmath1 diffeomorphic to @xmath8 , an exotic @xmath0 . according to example [ abstract](b ) , we obtain uncountably many diffeomorphism types of such exotic @xmath0 domains of holomorphy in @xmath1 . more generally , we can replace @xmath1 in this argument by an @xmath61-ball , obtaining a similar uncountable family inside any complex surface . the same method can be used to find many contractible domains of holomorphy in @xmath1 that ( unlike exotic @xmath0s ) are bounded by smooth , compact , pseudoconvex homology 3-spheres @xcite . finally , we turn to the analog of theorem [ abstop ] . as in that theorem , the extra hypothesis disappears when we replace 2-handles by casson handles . we obtain a precise analog of theorem [ ambin ] for @xmath3 , up to topological isotopy . [ ambitop ] @xcite , @xcite . an open subset @xmath8 of a complex surface @xmath5 is topologically isotopic to a stein open subset if and only if it is homeomorphic to the interior of a handlebody with all indices @xmath48 . this is essentially proven , in an expository manner , in @xcite . it is also explicitly given in @xcite as a corollary of a stronger theorem ( essentially theorem [ handle](c ) below ) . it can be derived from theorem [ ambi2 ] and a strengthened form of theorem [ abstop ] : the stein surface associated by the latter theorem to the above @xmath8 can actually be assumed to smoothly embed in @xmath8 , so that the embedding is topologically isotopic to the identity map on @xmath8 , and theorem [ ambi2 ] immediately gives the required stein open subset . alternatively , theorem [ ambitop ] can be used to prove theorem [ abstop ] : the handlebody interior of the latter theorem can be easily given a complex structure homotopic to the given almost - complex structure . ( for example , construct a suitable continuous map from the core 2-complex to a rational surface , then homotope it to an immersion and pull back the complex structure . ) theorem [ ambitop ] with @xmath62 completes the proof . \a ) any given smooth embedding @xmath63 has a tubular neighborhood diffeomorphic to @xmath64 . this is topologically ( but obviously not smoothly ) isotopic to a stein open subset . as in example [ abstract](a ) , we can realize any positive integer @xmath65 as the minimal genus of the generator of such a domain of holomorphy ( for any fixed isotopy class of embeddings of @xmath66 ) . for more on stein neighborhoods of isotoped embedded surfaces , see @xcite , @xcite . \b ) more generally , let @xmath67 be any open subset homeomorphic to the interior of a handlebody with all indices @xmath48 . then @xmath8 is topologically isotopic to a stein open subset @xmath12 . by example [ r4 ] , there are uncountably many diffeomorphism types of stein exotic @xmath0s in @xmath68 . we can connect any one of these to @xmath12 by a 1-handle in @xmath1 , preserving the stein condition . ( this requires a small amount of control of the boundaries of these subsets , but is not a problem here . ) by the method of @xcite , we can then distinguish uncountably many diffeomorphism types of these domains of holomorphy , all topologically isotopic ( hence homeomorphic ) to @xmath8 @xcite . thus , if @xmath67 is topologically isotopic to any domain of holomorphy , then there are uncountably many diffeomorphism types of such . note that this construction preserves the minimal genus of every homology class of @xmath12 , so in ( a ) above we have uncountably many diffeomorphism types for each embedding of @xmath66 and each @xmath69 . one would expect this uncountability to be typical , with any complex surface @xmath5 in place of @xmath1 . this can be proven in many special cases ( @xcite , see also theorem [ onion ] below ) , but the general case seems out of range of current technology . the theorems of the previous two sections can be thought of as manifestations of gromov s _ h - principle _ @xcite , @xcite . for many problems such as constructing complex or symplectic structures or smooth immersions , one first has to solve a topological problem at the level of bundle theory . the h - principle for such a problem asserts that a solution at the level of bundle theory guarantees a solution to the complete problem . this clearly fails for putting complex structures on closed manifolds : a solution at the level of bundles is an almost - complex structure , but many closed , almost - complex manifolds do not admit complex structures , and there seems to be no hope of reducing the gap to a topological criterion . in contrast , we have seen that the existence of stein structures and domains of holomorphy are determined by topological criteria . we will now sketch the proofs of the theorems of the two previous sections , emphasizing the underlying topology for the h - principle , which is the source of the difficulties surrounding 4-dimensional 2-handles . we sketch a proof from a preliminary version of @xcite . we are given an almost - complex manifold @xmath70 of complex dimension @xmath2 , and an exhausting morse function @xmath46 on @xmath8 whose critical points all have index @xmath24 . we inductively assume that for a certain regular value @xmath71 , the almost - complex structure @xmath19 restricts to an actual complex structure on the compact submanifold @xmath72 $ ] . we also assume that on @xmath73 , the function @xmath46 is suitably compatible with @xmath19 , i.e. , it is a _ ( strictly ) plurisubharmonic _ ( or _ j - convex _ ) function . ( the significance of this will be indicated later . ) we wish to homotope @xmath19 relative to @xmath73 to replicate the induction hypotheses for a new regular value @xmath74 , where for simplicity we assume @xmath75 $ ] contains a unique critical point , whose index is @xmath25 . induction then homotopes @xmath19 to an honest complex structure on @xmath8 for which @xmath46 is an exhausting plurisubharmonic function , and the latter condition guarantees the complex structure is stein by a theorem of grauert @xcite . we now extract the bundle - theoretic data relevant for the induction step . we ascend to the level of the new regular value @xmath76 by attaching a handle @xmath77 to @xmath73 ( along with a collar that does not affect the homotopy - theoretic data ) . at the level of homotopy theory , @xmath78 is given by its core @xmath79 , which is attached along its attaching sphere @xmath80 . the complete attaching map @xmath81 is then determined up to isotopy by by its 1-jet along @xmath82 . on the real hypersurface @xmath83 , the almost - complex structure @xmath19 induces a unique field @xmath84 of complex hyperplanes . since the complex line bundle @xmath85 is canonically trivialized by a normal vector field to @xmath86 , the complex @xmath87-plane field @xmath88 carries all of the bundle - theoretic information about @xmath19 along @xmath86 . the handle @xmath78 also inherits an almost - complex structure @xmath89 from its inclusion in @xmath8 , and along the attaching region @xmath90 , this induces a complex hyperplane field @xmath91 ( mapping to @xmath92 complex linearly under @xmath93 ) . the structure @xmath89 is determined up to homotopy rel @xmath82 by its restriction to the bundle @xmath94 . thus , the essential data associated to gluing the almost - complex handle are @xmath89 on @xmath94 and the embedding @xmath95 , with the latter covered by a complex bundle map @xmath96 . ( note that this bundle map also carries the normal data determining @xmath93 . ) to complete the proof , we must adjust our setup so that @xmath89 takes a standard form . the descending disk of any critical point of a plurisubharmonic morse function is _ totally real _ , that is , its tangent bundle contains no complex lines . thus , it is natural to define the standard complex @xmath25-handle to be a tubular neighborhood in @xmath7 of @xmath97 , with the standard complex structure @xmath20 . ( this also clarifies why the indices of such critical points never exceed @xmath2 . ) the space of positively oriented complex vector space structures on @xmath98 is connected ( since a complex basis for any such @xmath19 , completed to a real basis using its @xmath19-image , is homotopic through real bases to the standard basis ) , so we may assume @xmath99 near the center point of @xmath100 . since @xmath78 is contractible , we can then find a homotopy @xmath101 on @xmath78 from @xmath89 to @xmath20 . [ hprin ] unless @xmath102 , there is an isotopy @xmath103 of the attaching map @xmath93 , and a choice of homotopy @xmath101 on @xmath78 as above , such that for each @xmath104 , the hyperplane field @xmath105 induced by @xmath101 on @xmath90 is sent @xmath101-complex linearly onto @xmath92 by @xmath106 . we can use @xmath107 to adjust the handle @xmath78 in @xmath8 , and extend @xmath101 to a homotopy on @xmath8 rel @xmath73 , deforming our entire setup rel @xmath73 so that @xmath99 along @xmath100 ( for @xmath47 ) . by a further adjustment , we can then arrange @xmath78 to be the standard complex @xmath25-handle , glued to @xmath73 holomorphically on a neighborhood of @xmath82 in @xmath78 . ( here we do not require that @xmath90 maps into @xmath86 , only that its 1-jet along @xmath82 maps to that of @xmath86 . this is because @xmath86 is @xmath19-convex while @xmath90 is @xmath19-flat . holomorphicity is then achieved by taking @xmath93 to be real - analytic on @xmath108 near @xmath82 , then complexifying . ) now @xmath109 is a complex manifold , and @xmath100 is a totally real and real - analytic submanifold intersecting @xmath86 standardly ( along @xmath82 we have @xmath110 ) . with some additional hard work , eliashberg produces a regular neighborhood @xmath111 of such a subset @xmath112 , for which the boundary is @xmath19-convex . this means that after an isotopy of @xmath8 fixing @xmath73 and sending @xmath113 to @xmath114 , we can assume that @xmath115 is plurisubharmonic . the induction step and proof are now completed by a proof of the lemma , which we discuss below . now we are given @xmath116 , with a suitable morse function on @xmath8 , and @xmath13 . we apply the previous method . for each handle @xmath117 , we obtain an isotopy @xmath118 of the inclusion map ( extending the isotopy of @xmath90 given by lemma [ hprin ] ) , and a homotopy @xmath101 of @xmath19 rel @xmath73 on @xmath8 , so that @xmath119 on @xmath100 . this time , however , we are not allowed to homotope @xmath19 to construct our stein open subset of @xmath5 . instead , we observe that the homotopy @xmath120 agrees with @xmath121 on @xmath122 and for @xmath123 . we infer that on @xmath94 , the structure @xmath124 is homotopic rel @xmath125 to @xmath119 . in particular , there is no homotopy - theoretic obstruction to making @xmath100 totally real by an isotopy of @xmath126 rel @xmath82 . we can now invoke an h - principle of gromov to conclude that such an isotopy exists , so we can assume @xmath100 is totally real and intersects @xmath86 standardly . after another ( @xmath127-small ) isotopy , we can assume @xmath100 is also real - analytic , so eliashberg s work extends the plurisubharmonic function as before . we again sketch the proof from @xcite ; details appear in @xcite lemma 3.1 . we would like to derive the lemma as a sort of h - principle . since @xmath86 is a level set of a plurisubharmonic function , it follows that @xmath92 is a _ contact structure _ on @xmath86 . since the tangent spaces of @xmath82 lie in @xmath128 , the fibers of @xmath129 are parallel to @xmath130 , and so lie tangent to @xmath90 . thus , @xmath131 lies in the @xmath20-complex hyperplane field @xmath132 , and so the embedding @xmath126 we are trying to construct must send @xmath82 to a manifold tangent to @xmath92 everywhere . such a manifold is called _ legendrian _ if @xmath133 and _ isotropic _ in general . the lemma now amounts to isotoping @xmath95 to an isotropic embedding while suitably controlling data about the almost - complex structures . by combining h - principles of gromov ( making a submanifold of a contact manifold isotropic ) and hirsch ( for the homotopy and remaining data ) , one can at least find a regular _ homotopy _ @xmath107 with the right properties ( for all @xmath134 ) . if @xmath135 , this will generically be an isotopy , and the lemma follows . when @xmath133 , we can still assume @xmath126 is an embedding , but there will generically be finitely many values of @xmath104 for which @xmath107 is only an immersion . it now suffices to assume @xmath133 and to turn the given regular homotopy into an isotopy for @xmath13 . since the lemma is trivial for @xmath136 , we turn to the next simplest case @xmath3 , where the method is only partially successful . there is an obstruction in @xmath137 to turning the regular homotopy @xmath107 into an isotopy rel @xmath138 , namely the signed number of double points in the surface @xmath139 , where @xmath140 . ( we must change @xmath141 by a regular homotopy in order to preserve the bundle data . ) we wish to kill the obstruction by appending an additional regular homotopy of the legendrian embedding @xmath126 . in order not to lose the legendrian condition , we homotope through legendrian immersions . since every contact structure can be locally identified with the standard model in @xmath142 , we can identify a neighborhood of some point of @xmath143 in @xmath86 with a standard neighborhood in @xmath144 , so that the standard `` front '' projection to @xmath31 appears as a cusp , as in figure [ yasha](a ) . ( see e.g. @xcite , @xcite for more on the standard contact structure and front projections of legendrian curves . ) a regular homotopy of the front projection produces ( b ) of the figure . this homotopy lifts uniquely to a regular homotopy of legendrian curves in @xmath144 . ( since negative slopes are always in front of positive slopes , ( b ) uniquely represents an embedded legendrian curve in @xmath144 obtained from ( a ) by a regular homotopy involving a legendrian curve with a generic double point . ) this local model shows how to change the obstruction by one unit , but unfortunately we can not control the sign of this change ( since the sign of the double point is predetermined through the underlying contact geometry ) . in fact , it is not possible to arbitrarily change the obstruction without losing control of the topological knot type of @xmath145 , and this is why eliashberg s method fails for 4-dimensional 2-handles . we return to this phenomenon below . for @xmath146 , the argument works much better . we have the analogous obstruction ( in @xmath137 or @xmath147 , depending on the parity of @xmath2 ) , but this time we can change it arbitrarily . the analogous front projection is obtained from figure [ yasha](a ) by crossing with extra dimensions ( to obtain an @xmath87-manifold in @xmath108 ) . instead of pushing an arc of the manifold through the adjacent sheet , we push through a compact @xmath87-manifold @xmath148 with boundary . one can compute that the resulting change in the obstruction is just the euler characteristic of @xmath148 , which can be any integer ( whereas a compact 1-manifold must have nonnegative euler characteristic ) . thus , we can modify the regular homotopy to make the obstruction vanish when @xmath146 . now we can perturb @xmath141 to make its image embedded , by a version of the whitney trick . ( if @xmath86 is not simply connected , we use the fact that the homotopy is @xmath14-small , so we can assume the whitney circles are sufficiently small that they are nullhomotopic . ) the resulting isotopy satisfies the lemma ( and exists unless @xmath102 ) . before extending the method to the 4-dimensional case , theorem [ abs2 ] , we clarify terminology . a 4-dimensional compact handlebody with all indices @xmath149 admits a unique ( tight ) contact structure on its boundary @xcite . a legendrian knot in a contact 3-manifold has an induced _ contact framing _ , determined by a vector field in @xmath92 transverse to the knot . in the special case of the attaching circle @xmath82 in the standard complex 2-handle in @xmath1 , this vector field is just @xmath150 for a tangent vector field @xmath151 to @xmath82 in @xmath31 . since @xmath151 has winding number 1 in @xmath31 , it follows that the contact framing of @xmath82 differs by one twist from the normal framing to @xmath100 in the handle . we conclude that for @xmath3 , any holomorphic 2-handle must be attached along a legendrian knot with framing obtained from its contact framing by adding a ( left ) twist . this is precisely the condition given in theorem [ abs2 ] . when @xmath3 , the proof of theorem [ absn ] only fails for 2-handles . the given condition on the framings allows us to bypass lemma [ hprin ] , to attach standard complex 2-handles with @xmath93 preserving @xmath92 along @xmath82 . these can be arranged to preserve the stein condition as before . in the ambient case , the main idea for the higher - dimensional theorem [ ambin ] still works , although the details are more delicate . we must incorporate the extra data about the stein structure homotopic to @xmath152 , in order to bypass lemma [ hprin ] for the 2-handles . however , it is too much to expect that the resulting stein open subset will be holomorphically equivalent to the hypothesized stein surface , so we must give up some information , relying on the contact structures to carry enough data . for a complete proof , see @xcite . to understand the context into which theorem [ abs2 ] fits , it is helpful to analyze lemma [ hprin ] when @xmath3 . in this case , the obstruction can be interpreted as a mismatch between framings . note that in figure [ yasha ] , for example , we can return from ( b ) to ( a ) by an isotopy ( although not through legendrian knots ) . the resulting regular homotopy from ( a ) to ( b ) and back carries along a normal framing , but returns it with two extra twists . ( draw a curve parallel to the one in ( a ) , carry the pair through the regular homotopy , and observe directly that two left twists appear . ) the resulting framing change can be interpreted as a normal euler class , for example on the immersed torus @xmath153 in @xmath154 that is homologous to @xmath155 . the phenomenon now reduces to the observation that for a closed , orientable , generically immersed surface , its normal euler number differs from the self - intersection number of its homology class by twice the signed number of double points ( which follows easily from the definitions , e.g. @xcite exercise 6.1.1(a ) and solution ) . thus , the framing condition in theorem [ abs2 ] corresponds to suitably killing the obstruction in lemma [ hprin ] . while we can not apply the whitney trick as in that lemma , it remains true that attaching circles can be @xmath14-small isotoped to become legendrian ( although we lose control of the almost - complex structures on the corresponding handles ) . we can then change the obstruction arbitrarily in one direction but not in the other . ( figure [ yasha ] changes the obstruction by 2 , but it can also be changed by 1 by adding a zig - zag . ) to apply theorem [ abs2 ] , the problem then becomes to draw a handle diagram for @xmath8 , where the attaching circles are legendrian with sufficiently large contact framings ( or _ thurston - bennequin invariants _ , in the language of legendrian knot theory ) . this is much harder than applying the higher - dimensional theory , but sometimes still tractable . see e.g. @xcite , @xcite or @xcite for examples and applications . once a legendrian handle diagram is specified for @xmath8 , the homotopy class of the corresponding stein structure is determined by the _ rotation numbers _ of the legendrian knots . these are defined as relative chern classes but can be computed combinatorially in a diagram . a knot can be made legendrian for only finitely many choices of the rotation number , once the thurston - bennequin invariant is specified . this reflects the fact that a 4-manifold may be stein realizing some , but not all , choices of the homotopy class of the almost - complex structure . in either case , the difficulty is that while we can attach the 2-handles to legendrian knots with the correct rotation numbers , we can not do so with the framings required for a stein structure . ( that is , the thurston - bennequin invariants will be too negative , so that the contact framings of the legendrian knots will be far from the framings of the 2-handles . ) as we have seen , however , these obstructions can be interpreted as normal euler numbers , which suitably count double points . we can kill the obstructions by replacing the core disks of the 2-handles by immersed disks . ( in the case of theorem [ ambitop ] , the cores may only be topologically embedded , in which case it takes work of freedman and quinn to find suitable smoothly immersed disks . ) we now obtain a stein structure , but on the wrong manifold each double point of an immersed disk contributes a new generator to the fundamental group . we can kill these generators by adding 2-handles , recovering @xmath8 but losing the stein structure . if we instead allow double points in the new 2-handles , we recover the stein structure but again have extra fundamental group . we inductively kill the extra generators at each stage by adding immersed disks at the next stage , obtaining an infinite union that inherits a stein structure . this procedure replaces each 2-handle by an infinite tower called a _ casson handle_. these casson handles are simply connected ( since any loop lies in a finite subtower , so it is nullhomotopic at the next stage ) . in fact , according to freedman @xcite , every casson handle is homeomorphic to an open 2-handle @xmath156 , so our final stein surface is homeomorphic to @xmath8 . for further details , see @xcite and @xcite , respectively . we have seen that for some of our purposes , the morse and handlebody viewpoints are interchangeable , but for other purposes , one viewpoint is more natural than the other . for 4-manifolds with infinite topology , for example , theorem [ abs2 ] is most easily stated using handlebodies . in this section , we adapt the theorems of section [ ambient ] to the setting of embedded handlebodies , which allows us to use a sharper notion of isotopy ( namely _ ambient _ isotopy ) . we also introduce a third viewpoint , focusing on the totally real cw - complex forming the core of the handlebody . we show that suitably embedded cw - complexes can be isotoped to become totally real and pseudoconvex in the sense of being nested intersections of stein open subsets . this is particularly striking in the 4-dimensional topological setting , where the resulting uncountable stein neighborhood systems frequently realize uncountably many diffeomorphism types , even though the neighborhoods are all homeomorphic relative to the embedded complex . throughout this section , we rely on informal terminology , hypothesizing that stein manifolds are made by a particular method . when we construct a stein manifold , it is clear what method we have used , but if the manifold is given to us in some other way , it is not clear whether the hypothesis applies . in @xcite and @xcite we use more precise and general terminology , with definitions that are currently too cumbersome to state here . instead , we try to indicate the most important points . for example , a stein manifold constructed by eliashberg s method is assumed to be exhibited as the interior of a handlebody with pseudoconvex boundary , and in the ambient setting , this handlebody is embedded . thus , we are given a partial collaring at infinity by pseudoconvex hypersurfaces , extending to a smooth mapping cylinder structure surrounding a cw - complex core that we can assume is smooth and totally real . the subhandlebodies inherit similar structure . a smooth ( or topological ) _ ambient isotopy _ of a manifold @xmath5 is an isotopy of the identity on @xmath5 through diffeomorphisms ( or homeomorphisms ) of @xmath5 . an isotopy of an open subset @xmath8 of @xmath5 need not extend to an ambient isotopy of @xmath5 , because the _ end _ of @xmath8 ( the part outside large compact subsets of @xmath8 ) can be troublesome . for example , @xmath157 is isotopic , but not ambiently isotopic , to the complement of a disk , and an arbitrary @xmath5 is not ambiently isotopic to any proper subset of itself . thus , theorem [ ambitop ] implies @xmath53 is topologically isotopic inside itself to a stein surface , but not ambiently . we fix this problem by replacing open subsets by properly embedded handlebodies , which are closed subsets . we obtain the following analog of the main theorems of section [ ambient ] . [ handle ] let @xmath29 be a smooth , @xmath23-dimensional handlebody with all handles of index @xmath24 . * if @xmath13 , then every smooth embedding of @xmath29 into a complex @xmath2-manifold @xmath5 is smoothly isotopic ( ambiently if the embedding is proper ) to one for which the image has stein interior and a totally real core . this image is made by eliashberg s method from the given handle decomposition . * if @xmath3 , the same holds for embeddings for which the pulled - back almost - complex structure is homotopic to a stein structure on @xmath158 made from the given handle decomposition . * if @xmath3 , then every topological embedding is topologically isotopic to one whose image has stein interior and a core that , except for one point on each 2-handle core , is smooth and totally real . if the embedding is proper and the image of @xmath159 is flat ( as defined below ) , then the isotopy is ambient . for ( a ) and ( b ) , the discussion of sections [ ambient ] and [ proofs ] applies with little change , so in particular , ( b ) is proved in @xcite and ( a ) follows from @xcite or @xcite . one merely needs an additional glance at the proof of theorem [ ambin ] in section [ proofs ] to see that at each stage of the induction , the constructed isotopy can be assumed to be ambient . ( since each handle of @xmath29 has a smooth , compact outer boundary , the isotopy extension theorem applies . ) these isotopies have a well - defined limit on @xmath29 , as well as on @xmath5 if the embedding is proper , since each point in the domain has a neighborhood on which the isotopies all agree after finitely many stages . ( each map of the limiting isotopy on @xmath5 is surjective since it is proper . this follows since @xmath29 is morse - ordered , so only finitely many handles are stacked above a given handle , and since each map sends each subhandlebody of @xmath29 into itself . ) the proof of ( c ) is given in @xcite . the main new difficulty is that the end of a casson handle is pathological the point - set boundary of a typically embedded casson handle can not be a manifold . to address this , we interleave many layers of embedded surfaces between the layers of immersed disks , to obtain the more subtle infinite towers of @xcite , whose point - set boundaries are 3-manifolds . ( increasing the genus of a surface has the same effect as introducing positive double points when we construct stein surfaces . ) core disks that are smooth except at one point are a byproduct of freedman s methods , and in our setting , the smooth parts are also totally real . the flatness hypothesis means that the embedding of @xmath159 can be extended to an embedding of @xmath160 . this rules out wild behavior analogous to alexander s horned sphere , and for proper embeddings , it is sufficient to guarantee that the isotopy is ambient . ( in @xcite , properness is replaced by a weaker condition , allowing us to also consider embeddings of infinite handlebodies into regions with compact closure . ) since the stein manifolds of ( a ) and ( b ) above are constructed by eliashberg s method , the resulting boundaries are pseudoconvex hypersurfaces in @xmath5 . ( here we use boundary " in the sense of manifolds , which only corresponds to the point - set sense if the embedding is proper . ) in contrast , the boundaries obtained in ( c ) are only topological ( flat ) 3-manifolds . these will be compact if @xmath29 is , but may not be smoothable in @xmath5 by any topological isotopy . it seems reasonable to introduce a notion of `` topological pseudoconvexity '' to describe such 3-manifolds bounding stein open subsets . theorem [ handle ] shows that being made by eliashberg s method is a weak hypothesis in fact , every stein manifold satisfies this hypothesis after a smooth isotopy within itself : [ structure ] let @xmath8 be a stein manifold . then @xmath8 is smoothly isotopic within itself to a stein open subset that is the interior of a smoothly and properly embedded handlebody @xmath161 constructed by eliashberg s method . in particular , the core of @xmath29 is a smooth , totally real @xmath2-complex ( @xmath162 ) . this can be assumed to be fixed by the given isotopy . furthermore , @xmath163 is diffeomorphic to @xmath164 . by lemma [ decomp ] ( or [ steindec ] if @xmath3 ) , there is a diffeomorphism @xmath165 from a handlebody interior . shrink @xmath166 into its own interior by the obvious proper isotopy , and apply theorem [ handle ] to the image in @xmath8 of the resulting @xmath167 . the consequent ambient isotopy changes @xmath46 into a new diffeomorphism for which the image @xmath29 of @xmath167 has the required properties . we can reinterpret theorem [ handle ] to focus on the cores of the handlebodies . call an embedded cw - complex _ tame _ ( in either the topological or smooth setting ) if we can thicken it to a handlebody , i.e. , it is the core of some embedded handlebody whose handles correspond to cells in the obvious way . ( for example , the fox - artin arc in @xmath144 is not a tame cw - complex , although it could be considered a smooth 1-complex with two 0-cells and an embedded open 1-cell . ) it is natural to ask which embedded complexes ( up to isotopy ) can be thickened to stein handlebodies as above . we answer the question and observe that the resulting complexes are pseudoconvex in the sense of being nested intersections of stein neighborhoods . [ ncplx ] let @xmath34 be a tamely embedded @xmath2-complex in a complex @xmath2-manifold @xmath5 . * in the smooth setting , if @xmath13 , then after a @xmath14-small smooth isotopy ( smoothly ambient if @xmath34 is a closed subset , and topologically ambient in general ) , @xmath34 is totally real , and is the core of a smoothly embedded handlebody constructed by eliashberg s method . thus , @xmath34 ( after isotopy ) can be described as a nested intersection of stein neighborhoods , smoothly isotopic to each other @xmath168 ( ambiently as before ) , forming a neighborhood system of @xmath34 if the latter is compact . * if @xmath3 , the same holds provided that @xmath34 thickens to a smoothly embedded handlebody satisfying the hypotheses of theorem [ abs2 ] , for which the abstract stein structure of that theorem is homotopic to the almost - complex structure inherited from the embedding in @xmath5 . * if @xmath3 and @xmath34 is tamely topologically embedded , then there is a @xmath14-small topological ambient isotopy after which , except on a finite subset of each 2-cell , @xmath34 is smooth and totally real , and @xmath34 is the core of a topologically embedded handlebody with stein interior . in fact , @xmath34 is a nested intersection of such stein neighborhoods , topologically ambiently isotopic to each other @xmath168 , forming a neighborhood system of @xmath34 if the latter is compact . note that in ( c ) , @xmath34 may not be topologically ambiently isotopic to any smoothly tame complex , and even if it is , there will typically be obstructions to making it totally real . while a smooth 2-cell can always be made totally real except on a finite set , the remaining complex points may obstruct the existence of stein neighborhoods . for example , @xmath169 is not smoothly isotopic to any complex with a stein neighborhood , but the theorem gives a topological ambient isotopy to a topological sphere with a stein neighborhood system . thicken @xmath34 to a handlebody and apply theorem [ handle ] . in the smooth setting , eliashberg s method allows the isotopy to be @xmath14-small ( once the handles are chosen sufficiently thin ) . to arrange this in ( c ) , first subdivide @xmath34 sufficiently , and use the fact that the isotopy of theorem [ handle ] sends each subhandlebody into itself . in either case , we can take the isotopy to be @xmath61-small , where @xmath61 varies continuously . if @xmath170 outside a neighborhood within which @xmath34 is closed , the isotopy will be topologically ambient . for ( a ) and ( b ) , the neighborhood system is obtained by repeatedly applying eliashberg s construction to thicken the given totally real @xmath34 . the neighborhood system in ( c ) follows from theorem [ onion ] below . eliashberg s method should actually provide more than just a totally real core with a stein neighborhood system . every smooth handlebody @xmath29 whose handles all have index @xmath171 admits a smooth mapping cylinder structure , so a map @xmath172\times \partial h\to h$ ] that is the identity on @xmath173 , sends @xmath174 onto the core @xmath34 of @xmath29 , and restricts to a diffeomorphism @xmath175\times \partial h\to h - k$ ] . for each @xmath176 $ ] , let @xmath177 . when @xmath29 is compact , these form a neighborhood system of @xmath34 i.e. , every neighborhood of @xmath34 contains some @xmath178 . by choosing eliashberg s handles to lie in 1-parameter nested families intersecting only in the core , it should be possible to control the construction to inductively arrange the resulting stein manifold @xmath179 so that every @xmath178 is also stein . we would then have a nested family of stein neighborhoods of the totally real @xmath34 , parametrized by @xmath175 $ ] , with any two neighborhoods smoothly isotopic rel @xmath34 ( ambiently in @xmath29 for @xmath180 ) . while this has not yet been been worked out in detail , it may be discussed further in @xcite or @xcite . in the 4-dimensional topological setting , we can no longer expect such nice structure , since @xmath181 will no longer inherit a smooth product structure . ( even when @xmath29 is compact , the stein surface homeomorphic to @xmath158 will not usually be diffeomorphic to the interior of a compact manifold . ) surprisingly , however , we can still use freedman theory to construct a topological mapping cylinder structure for which @xmath178 is stein whenever @xmath182 lies in the standard cantor set @xmath183 . while these stein neighborhoods are still topologically isotopic rel @xmath34 , they will typically not be diffeomorphic to each other . [ onion ] @xcite . let @xmath184 be a homeomorphism from a handlebody interior to a stein surface . suppose either that @xmath8 was constructed by eliashberg s method ( abstractly or ambiently ) as a handlebody interior , exhibited by the diffeomorphism @xmath78 , or that @xmath8 was obtained by theorem [ abstop ] , [ ambitop ] or [ handle](c ) , with @xmath78 the given homeomorphism . then after a topological ambient isotopy , the open subsets @xmath185 coming from the mapping cylinder structure on @xmath29 are stein whenever @xmath186 , and except for one point on each 2-cell , the core @xmath34 is smooth and totally real . if @xmath187 is nonzero but finitely generated ( for example ) , then we can assume the stein surfaces @xmath178 for @xmath186 realize infinitely many diffeomorphism types , and if @xmath8 topologically embeds in a ( possibly infinite ) blowup of @xmath1 ( for example ) , they realize uncountably many diffeomorphism types . a stronger and more detailed version of this is given in @xcite , with a more general and precise hypothesis replacing that of the method of construction . ( basically , @xmath8 should be constructed by eliashberg s method , allowing some 2-handles of @xmath29 to be replaced by generalized casson handles , so that the resulting subhandlebodies with finite towers attached have pseudoconvex boundaries and totally real cores . ) recall that by corollary [ structure ] , every stein surface is smoothly isotopic within itself to one admitting a diffeomorphism @xmath78 as above . in fact , the method of that corollary shows that the conclusion of theorem [ onion ] applies to _ any _ stein surface , once we restrict @xmath182 to @xmath188 $ ] . however , the theorem is most striking in the context of theorems [ abstop ] , [ ambitop ] and [ handle](c ) . in those cases , when @xmath29 is a finite handlebody , the resulting @xmath34 is compact with neighborhood system @xmath189 , even though @xmath8 and @xmath178 frequently can not be diffeomorphic to the interior of a finite handlebody . ( consider @xmath190 , for example . ) one would then expect the subsets @xmath178 to nearly always be pairwise nondiffeomorphic , and this can be proven in various cases @xcite . our definition of handlebodies in section [ abstract ] required them to be morse - ordered and locally finite , without collars between layers of handles . we saw that this was the correct language for expressing eliashberg s work in dimension 4 ( theorem [ abs2 ] ) . passing to the cores of the handles gave a correspondence between handlebodies and tamely embedded cw - complexes , suggesting corollary [ ncplx ] . theorem [ onion ] required our narrow definition of handlebodies in order to have the underlying mapping cylinder structure . to extend this theorem to arbitrary stein surfaces we then had to find stein handlebodies ( narrowly defined ) inside them via corollary [ structure ] . we now prove the necessary lemmas showing that morse functions can be refined ( introducing critical points ) to yield handle decompositions in our narrow sense , even if we start with infinitely many critical points . [ decomp ] let @xmath8 be a smooth @xmath27-manifold with an exhausting morse function @xmath46 . then @xmath8 is diffeomorphic to the interior of a handlebody @xmath29 with the same maximal index @xmath25 as @xmath46 . assume @xmath46 has infinitely many critical points , since the finite case is well - known . after a slight perturbation , @xmath46 determines a decomposition @xmath191 , where @xmath192 is empty and @xmath193 for a collar @xmath194 \times \partial p_i$ ] and handle @xmath195 of index @xmath196 . any handle decomposition of @xmath197 can be extended over @xmath198 using an arbitrary handle decomposition of @xmath199 : each @xmath200-handle @xmath78 of @xmath199 contributes an @xmath200-handle @xmath201 \times h$ ] and an @xmath202-handle @xmath203 \times h$ ] . a perturbation of the product structure on @xmath198 guarantees that each of its handles attaches to handles of lower index as required , as does the additional handle @xmath195 of @xmath204 . thus , we inductively obtain a handle decomposition @xmath205 of @xmath8 ( containing many @xmath27-handles ) that is compatible with the collars . let @xmath206 be the subhandlebody consisting of all handles of index @xmath196 except for @xmath25-handles of the form @xmath201 \times h$ ] along each @xmath199 . then for each @xmath20 , @xmath207\times q_i$ ] , where @xmath208 is the subhandlebody of handles with index @xmath32 . it is now routine to construct a diffeomorphism from @xmath158 to @xmath8 : first inductively construct , for each @xmath20 , a diffeomorphism from @xmath158 to @xmath209 , by pushing the unwanted handles of @xmath198 out of @xmath204 using the collar structure . then check that these diffeomorphisms can be assumed to converge to the required diffeomorphism @xmath210 . [ steindec ] every stein surface is diffeomorphic to the interior of a handlebody whose handles all have index @xmath48 , and for which each 2-handle is attached along a legendrian knot in the relevant contact 3-manifold , with framing obtained from the contact framing by adding one left twist . the stein structure induced by this handle decomposition via theorem [ abs2 ] is homotopic ( through almost - complex structures ) to the original . we wish to apply the method of the previous proof , but extra care is required for controlling 2-handle framings . we begin with an observation of goodman @xcite , that every ( closed ) contact 3-manifold @xmath86 admits a _ contact cell decomposition_. using convex surface theory ( @xcite , @xcite , see also @xcite ) , we construct a contact cell decomposition whose 1-skeleton contains any preassigned legendrian circle @xmath211 in @xmath86 . we then show that the 1-skeleton @xmath34 of such a decomposition is _ sufficiently fine _ , by which we mean that every 1-complex in @xmath86 can be contact isotoped , fixing @xmath34 , into an arbitrarily small neighborhood of @xmath34 . for simplicity , we assume @xmath86 is tight . then a tame cell decomposition of @xmath86 is _ contact _ if its 1-cells are legendrian and its 2-cells are convex with twisting @xmath212 . such a decomposition can be constructed , for example , from a triangulation , whose 1-skeleton we can assume contains @xmath211 , by isotoping the 1-cells rel @xmath211 to be legendrian with negative twisting . ( we measure the twisting using adjacent 2-cells , and arrange negative twisting on the 1-cells of @xmath211 using a suitably twisted model triangulation near @xmath211 . ) the 2-cells can then be assumed to be convex triangles with legendrian boundaries and negative twisting on each edge , so by the legendrian realization principle ( @xcite , @xcite , see also @xcite ) , we can subdivide to obtain the required structure . to see that such a 1-skeleton is sufficiently fine , we first contact isotope a given 1-complex @xmath213 into a preassigned neighborhood @xmath214 of the 2-skeleton . choose a 3-ball in the interior of each 3-cell , with convex boundary contained in @xmath214 . by convex surface theory , the boundary is essentially uniquely determined , so by @xcite , each ball can be contactomorphically identified with the standard ball in @xmath144 . using gray s theorem @xcite , it is now easy to push @xmath213 off of each ball into @xmath214 by a contact isotopy fixing @xmath34 . for any neighborhood @xmath12 of the 1-skeleton , we can choose @xmath214 sufficiently thin that @xmath213 can then be pushed into @xmath12 as required . this follows from examining a standard neighborhood of each 2-cell , where , for example , we push first toward the two elliptic boundary points , then transversely , preserving the characteristic foliation away from the elliptic points in each case . now let @xmath70 be an arbitrary stein surface . the basic theory of stein manifolds @xcite guarantees an exhausting plurisubharmonic morse function @xmath46 on @xmath8 , that we can assume is 1 - 1 on its critical set . its critical points all have index @xmath215 , and its regular level sets inherit contact structures . each descending disk of an index-2 critical point intersects nearby level sets below it in a legendrian circle , and when we thicken the disk to obtain a 2-handle , it is attached with framing differing from the contact framing by a left twist , as desired for the lemma . as before , the lemma is easy and well - known if @xmath46 has only finitely many critical points , so we again restrict to the infinite case and construct the required handle structure from @xmath46 using induction . for our induction hypotheses , we assume there is a handlebody @xmath29 , whose handles all have index @xmath48 and satisfy the required framing condition , with a diffeomorphism @xmath216 $ ] for a preassigned regular value @xmath71 . by theorem [ abs2 ] and its proof ( in the direction independent of lemma [ steindec ] ) , there is an induced stein structure @xmath217 on @xmath29 and contact structure on @xmath159 . we inductively assume we are given a neighborhood @xmath12 in @xmath159 of a sufficiently fine 1-complex @xmath218 , whose @xmath200-cells ( @xmath219 ) are disjoint from the handles of @xmath29 with index @xmath220 . ( in comparison with the previous proof , @xmath12 corresponds to the intersection of @xmath199 with the handles of @xmath205 of index @xmath32 , containing the foundation @xmath221 for the next layer of handles . ) we also assume that @xmath222 is a contactomorphism to @xmath223 with contact structure @xmath92 induced by @xmath19 , and that on @xmath73 minus finitely many interior points , @xmath224 agrees with @xmath19 up to homotopy through almost - complex structures inducing the given @xmath92 . ( it seems unlikely that deleting points is actually necessary , but it is easier to avoid the issue . ) note that the space of complex structures on @xmath225 inducing a given plane field on @xmath226 is contractible , so @xmath19 is essentially determined on @xmath86 by @xmath92 . the induction hypotheses are trivially satisfied for @xmath227 . for a regular value @xmath74 such that @xmath228 $ ] contains a unique critical value , @xmath45 $ ] is obtained from @xmath73 by attaching a collar and a handle ; we wish to extend over these to reproduce the induction hypotheses with @xmath76 in place of @xmath71 . we begin the induction step by adding handles to @xmath29 along @xmath12 , to form part of a collar of @xmath159 as in the proof of lemma [ decomp ] . let @xmath100 denote the descending disk of the unique critical point in @xmath75 $ ] , with @xmath229 in @xmath230 a legendrian circle or two points or empty , depending on the index of the critical point . let @xmath231 be the 1-skeleton of a contact cell decomposition of @xmath86 . since the previously given complex @xmath232 is sufficiently fine , there is a contact isotopy of @xmath86 sending @xmath34 into @xmath233 . by gray s theorem , we can assume the isotopy pushes the 0-cells of @xmath34 sufficiently close to those of @xmath234 that they avoid the 1- and 2-handles of @xmath29 . by reversing the isotopy and extending it to an isotopy of @xmath93 supported near @xmath159 , we can adjust @xmath93 so that @xmath233 contains @xmath34 ( as originally located , so in particular @xmath235 ) , without disturbing the induction hypotheses . by construction , the 1-complex @xmath236 thickens to a handlebody @xmath237 whose @xmath200-handles ( @xmath238 avoid the handles of @xmath29 with index @xmath220 . ( compare with @xmath221 in the previous proof . ) to extend @xmath29 over @xmath239\times ( \mbox{\rm 0-handles})\subset [ 0,1]\times q$ ] , let @xmath166 be the result of attaching a canceling 0-handle and 1-handle to @xmath29 at each vertex of @xmath240 , by eliashberg s method . then @xmath166 is a ( morse - ordered ) handlebody , with a stein structure @xmath241 extending @xmath217 , and a contact boundary . the latter agrees with @xmath242 outside a collection of 3-balls . since a tight contact 3-ball is determined by its boundary @xcite , there is a contactomorphism @xmath243 . this extends to a diffeomorphism @xmath244 that is supported on a disjoint union of 4-balls , so after we remove the center points of the 4-balls , @xmath245 is homotopic to @xmath217 preserving @xmath246 on @xmath159 . we replace @xmath29 by @xmath166 and @xmath93 by @xmath247 , preserving the induction hypotheses ( ignoring @xmath12 ) . we next construct the rest of @xmath239\times q$ ] . for each 1-cell of @xmath34 , the preimage in @xmath248 is a legendrian arc with endpoints in @xmath249 . after trimming the ends of this arc by removing its intersection with @xmath250\times q$ ] , connect its new endpoints by a 1-handle , then add a canceling 2-handle along the resulting legendrian circle . if we add the handles by eliashberg s method , the new 2-handles will satisfy the required framing condition . the resulting handlebody @xmath167 differs from @xmath166 only on a collection of 4-balls , so the previous argument allows us to replace @xmath29 by @xmath167 without disturbing the induction hypotheses ( again ignoring @xmath12 ) . the preimage @xmath251 of @xmath211 in @xmath252 sits in @xmath249 on top of the partial collar , and is a legendrian circle if @xmath253 . to complete the induction step , we must attach a new handle along @xmath251 , suitably map the result onto @xmath45 $ ] , and reconstruct @xmath254 . let @xmath255 be obtained from @xmath73 by attaching a handle with core @xmath100 using eliashberg s method . let @xmath256 be obtained from @xmath167 by attaching a handle @xmath78 with the corresponding attaching data , so we get a diffeomorphism @xmath257 . ( for a 2-handle , the attaching data consist of the legendrian circle @xmath251 and a normal framing , which by construction satisfies the condition required for the lemma . ) the two new handles intersect @xmath258 and @xmath113 , respectively , in regions @xmath259 that are either a solid torus or a punctured 3-sphere , depending on the index of @xmath78 . in each case , the ( tight ) contact structure is determined on @xmath259 by its restriction to @xmath260 . ( for punctured 3-spheres , this follows from @xcite . in the remaining case , the 2-handle attaches to a neighborhood of a legendrian circle , whose boundary can be assumed convex with a 2-component dividing set @xmath261 . the framing condition guarantees that each component of @xmath261 intersects the meridian of @xmath259 in a single point , so the classification of contact structures on a solid torus uniquely determines the contact structure @xcite , @xcite , cf . also theorem 5.1.30 of @xcite . ) thus , we can assume @xmath262 restricts to a contactomorphism @xmath263 . ( any self - diffeomorphism of @xmath264 extends over @xmath265 . ) since the new handles are 4-balls , we recover a homotopy from @xmath266 to @xmath267 in the complement of a finite set , preserving the contact boundary as usual . by a result in @xcite , we can assume @xmath111 was constructed so that there is a new exhausting plurisubharmonic function @xmath268 on @xmath8 for which @xmath269 and @xmath114 are level sets , with no critical points in between . we can then construct an isotopy @xmath107 of @xmath262 such that @xmath270 and each @xmath271 is a level set of @xmath272 , hence a contact manifold . by gray s theorem , we can assume each @xmath273 is a contactomorphism . then @xmath274 $ ] satisfies the induction hypotheses at the new level , once we construct a new @xmath218 and @xmath12 . for this last step , recall that @xmath34 is the 1-skeleton of a contact cell decomposition of @xmath86 . we pull this contactomorphically back to @xmath252 , and modify it inside a neighborhood of @xmath251 in @xmath249 to obtain a cell decomposition of @xmath258 . this only requires care when @xmath78 has index 2 . in that case , we require a pushoff of @xmath275 in @xmath252 to be a subcomplex of the new structure , and fill in @xmath276 with a single 2-cell and 3-cell . since the 2-cell intersects the dividing set of @xmath275 in two points , it has twisting @xmath212 . now for any index , we make the decomposition contact as in the first paragraph , noting that no change occurs outside a neighborhood of @xmath251 , and that no subdivision is necessary in @xmath277 in the index-2 case . the new sufficiently fine 1-skeleton @xmath278 lies in @xmath279 , it avoids the 2-handles of @xmath256 , and after a contact isotopy supported in @xmath280 , its 0-cells avoid the 1-handles of @xmath256 . thus , @xmath278 and @xmath281 complete the induction hypotheses . we have now inductively constructed a nested sequence of handlebodies , each with an embedding into @xmath8 . since the handles added at the @xmath282 stage are attached along a neighborhood @xmath12 contained in a subhandlebody added at the @xmath283 stage , the union @xmath284 is locally finite , and hence an infinite handlebody . each embedding into @xmath8 agrees with the previous one on the domain of the latter , except on a neighborhood of its boundary . it is routine to arrange these embeddings so that each point of @xmath285 has a neighborhood on which all but finitely many embeddings agree . we then have a limiting map @xmath286 that is easily seen to be a diffeomorphism . ( compare with the diffeomorphism constructed in the previous proof . ) similarly , the homotopies of almost - complex structures yield a limiting homotopy from @xmath287 to @xmath19 on the complement of a discrete and closed subset . ( here @xmath288 denotes the stein structure inherited by @xmath285 since its 2-handles are suitably framed . ) since the ( 2-dimensional ) core of @xmath284 can be chosen to miss the singular set , it is easy to construct a new homotopy that is defined everywhere on @xmath8 . a. casson , _ three lectures on new infinite constructions in 4-dimensional manifolds _ , ( notes prepared by l. guillou ) , a la recherche de la topologie perdue , progress in mathematics vol . 62 , birkhuser , 1986 , pp .	a unified summary is given of the existence theory of stein manifolds in all dimensions , based on published and pending literature . eliashberg s characterization of manifolds admitting stein structures requires an extra delicate hypothesis in complex dimension 2 , which can be eliminated by passing to the topological setting and invoking freedman theory . the situation is quite similar if one asks which open subsets of a fixed complex manifold can be made stein by an isotopy . as an application of these theorems , one can construct uncountably many diffeomorphism types of exotic @xmath0s realized as stein open subsets of @xmath1 ( i.e. domains of holomorphy ) . more generally , every domain of holomorphy in @xmath1 is topologically isotopic to other such domains realizing uncountably many diffeomorphism types . any tame @xmath2-complex in a complex @xmath2-manifold can be isotoped to become a nested intersection of stein open subsets , provided the isotopy is topological when @xmath3 . in the latter case , the stein neighborhoods are homeomorphic , but frequently realize uncountably many diffeomorphism types . it is also proved that every exhausting morse function can be subdivided to yield a locally finite handlebody of the same maximal index , both in the context of smooth @xmath2-manifolds and for stein surfaces . \b ) one of the most peculiar phenomena unique to 4-dimensional topology is the existence of exotic @xmath0s , that is , manifolds homeomorphic to euclidean space @xmath0 but not diffeomorphic to it . these have a complicated history , dating back to ideas of casson in the 1970s @xcite . most directly relevant to this paper is @xcite , which exhibits uncountably many diffeomorphism types of exotic @xmath0s , each occurring as an open subset of @xmath0 ( with its usual smooth structure ) . after simplifying ( @xcite , see also @xcite ) , one can explicitly build each member of such a family with a 0-handle , two 1-handles , a 2-handle and a casson handle ( taking the resulting interior ) . by a careful application of theorem [ abs2 ] , using the main idea behind theorem [ abstop ] to deal with the casson handle , one can endow such an uncountable family of exotic @xmath0s with stein structures @xcite . in contrast , there is an uncountable family of exotic @xmath0s that can not be smoothly embedded in @xmath0 @xcite , and these can not be constructed as handlebody interiors without using infinitely many 3-handles @xcite , so they can not admit stein structures .
one may wonder whether the two - time correlations @xmath158 reflect time - dependent measurements after the preparation of some initial state . we show that this is the case for the simple , but important example of a spin correlation for @xmath18 , i.e. , for @xmath73 . then we write @xmath159 where @xmath160 projects onto the states with @xmath161 . if @xmath7 denotes the density matrix of the total system before any state preparation we calculate where we assumed that the hamiltonian @xmath4 and the density matrix @xmath7 are invariant under total inversion @xmath163 so that the second term in equals the first one . finally , in we define the initial density matrix @xmath164 which results from @xmath7 by projecting it to the states with @xmath165 and its proper normalization . this clearly shows that in the studied case @xmath14 equals the time - dependent expectation value for a suitably prepared initial state . the above procedure can be modified to other observables . generally , we can consider @xmath166 to focus on the time - dependent expectation value @xmath167 starting from the initial density matrix @xmath168 . however , do not claim that a suitable operator @xmath169 is easy to find . this route remains to be explored in future work . with these matrix and vector elements we can compute @xmath66 in for various sets of conserved quantities . note that @xmath185 is linearly dependent on the @xmath102 quantities @xmath157 with @xmath186 due to @xmath187 similarly , @xmath188 depends linearly on them due to @xmath189 hence , one may either consider @xmath87 together with the @xmath102 quantities @xmath157 with @xmath186 _ or _ the three quantities @xmath190 . the first choice exploits all the known conserved quantities on the considered level of at most trilinear spin combinations . this is what is called ` all quantities ' in fig . 1 in the letter . no explicit formula can be given , but the required matrix inversion is easily performed for up to @xmath191 spins with any computer algebra program and up to @xmath192 spins by any subroutine package for linear algebra . the second choice of @xmath190 yields @xmath193 matrices and can be analysed analytically . inserting the elements in and in and those in into yields @xmath194 furthermore , these three quantities are conserved for any isotropic spin model so that we may also consider the system with the additional bond @xmath195 , see fig . 1 . thus we extend the above formulae by passing from @xmath88 to @xmath4 and hence from @xmath185 to @xmath196 . the modified scalar products are they lead to a bound @xmath198 as depicted in fig . 1 . the explicit formula is similar to the one in , but lengthy so that we do not present it here . it can be easily computed by computer algebra programs . . relates the non - decaying fraction @xmath26 to the relative bound for the overhauser field @xmath199 where @xmath144 is arbitrary if the central spin has @xmath18 . we stress , however , that the derivation yielding in ref . only holds for the csm so that we do not consider extensions to finite @xmath109 in this case . we use the freedom to choose @xmath144 to maximize the resulting lower bound for @xmath200 . we reuse all matrix elements of the norm matrix @xmath56 in and in . since uses the relative correlation we have to compute @xmath201 as well . furthermore , the vector elements of @xmath202 must be determined anew these elements allow us to determine the ratio @xmath204 for the three quantities @xmath190 or for all quantities , i.e. , @xmath87 and @xmath157 with @xmath205 . the ensuing lower bounds can be optimized by varying @xmath144 in such a way that the ratios become maximum yielding the best bounds . the latter step is easy to perform since the non - linear equation in @xmath144 to be solved to determine the maximum is just a quadratic one . in this way , the triangle and square symbols in fig . 1 are computed . the comparison to the bethe ansatz data for up to @xmath206 spins in ref . @xcite yields an excellent agreement within the accuracy with which we can read off @xmath26 from the numerically evaluated bethe ansatz correlation @xmath14 . this concludes the section on the required input of matrix and vector elements .	mazur s inequality renders statements about persistent correlations possible . we generalize it in a convenient form applicable to any set of linearly independent constants of motion . this approach is used to show rigorously that a fraction of the initial spin correlations persists indefinitely in the isotropic central spin model unless the average coupling vanishes . the central spin model describes a major mechanism of decoherence in a large class of potential realizations of quantum bits . thus the derived results contribute significantly to the understanding of the preservation of coherence . we will show that persisting quantum correlations are not linked to the integrability of the model , but caused by a finite operator overlap with a finite set of constants of motion . [ [ introduction . ] ] introduction . + + + + + + + + + + + + + the two - time correlation function of two observables reveals important information about the dynamics of a system in and out of equilibrium : the noise spectra are obtained from symmetric combinations of correlation functions , while the causal , antisymmetric combination determines the susceptibilities required for the theory of linear response . the two - time correlation function only depends on the time difference if at @xmath0 the system of interest is prepared in a stationary state whose density operator commutes with the time - independent hamiltonian . this is what will be considered in this work . since correlations generically decay for @xmath1 , important information about the system dynamics is gained if a non - decaying fraction of correlations prevails at infinite times . such non - decaying correlations are clearly connected to a limited dynamics in certain subspaces of the hilbert space . the question arises if such a restricted dynamics is always linked to the integrability of the hamiltonian . here integrability means that the hamiltonian can be diagonalized by bethe ansatz which implies that there is an extensive number of constants of motion . identifying and understanding those non - decaying correlations can be potentially exploited in applications for persistent storage of ( quantum ) information . in this letter we first prove that persisting correlations are not restricted to integrable systems by using a generalized form of mazur s inequality @xcite . this is in contrast to the behavior of the drude weight in the frequency - dependent conductivity of one - dimensional systems which appears to vanish abruptly once the integrability is lost , even if only by including an arbitrarily small perturbation . so far , the drude weight has been the most common application of mazur s inequality , see for instance refs . and references therein . second , we apply this approach to the central spin model ( csm ) @xcite describing the interaction of a single spin , e.g. , an electronic spin in a quantum dot @xcite , an effective two - level model in a nv center in diamond @xcite , or a @xmath2c nuclear spin @xcite , coupled to a bath of surrounding nuclear spins inducing decoherence . persisting spin correlations have been found in the csm by averaging the central spin dynamics over a bath of random classical spins @xcite or in markov approximation @xcite . finite - size calculations @xcite of the full quantum problem and stochastic evaluation @xcite of the exact bethe ansatz equations @xcite for small system sizes @xmath3 have also provided evidence for a non - decaying fraction of the central spin correlation , predicting a non - universal , system dependent value . its origin has remained obscure , and it has been speculated that the lack of spin decay might be linked to bose - einstein condensate - like physics @xcite . while it is fascinating to identify such non - decaying correlations , it is technically very difficult to rigorously establish them . approximate methods often miss precisely those intricate aspects allowing correlations to persist , especially when they explicitly exploit the assumption that the system relaxes towards a statistical mixture . numerical approaches are either restricted in system size @xcite , or they are limited in the maximum time which can reliably be captured @xcite . even analytical solutions @xcite can often only be evaluated in small systems @xcite . thus , a rigorous result establishing the existence of non - decaying correlations is highly desirable and we resort to mazur s inequality for this purpose . [ [ general - derivation . ] ] general derivation . + + + + + + + + + + + + + + + + + + + to establish the key idea and to fix the notation we present the following modified derivation related to suzuki s derivation in ref . . we consider the time - independent hamiltonian @xmath4 and the operator @xmath5 with a vanishing expectation value @xmath6 with respect to a stationary density operator @xmath7 , i.e. @xmath8=0 $ ] so that two - time correlation functions only depend on the time difference . note that @xmath7 does not need to be the equilibrium density operator . then , @xmath7 and @xmath4 have a complete common eigenbasis @xmath9 in a finite - dimensional hilbert space , and their spectra are @xmath10 and @xmath11 , respectively . we define the correlation function of @xmath5 as @xmath12 \\ \label{eq : lehmann } & = & \sum_{j , m } \rho_j \|a_{jm}\|^2 \exp(i(e_j - e_m)t),\end{aligned}\ ] ] so that eq . is its lehmann representation , and @xmath13 denotes the matrix element of @xmath5 . physically , @xmath14 stands for a measurement of @xmath15 at time @xmath16 after the evolution from the initial state prepared by applying @xmath5 at @xmath0 . especially , for @xmath17 of a spin @xmath18 in a disordered environment , @xmath14 is proportional to @xmath19 if @xmath20 , see supplement a for details . if @xmath21 exists , it is given by @xmath22 if @xmath23 does not exist , and @xmath24 , the long - time average @xmath25 is projecting out the time - independent part @xmath26 and uniquely defines the non - decaying fraction of the correlation . in practice , the lehmann representation requires the complete diagonalization of @xmath4 which is not feasible for large systems . hence one resorts to constants of motion . to this end , we define the scalar product for two operators @xmath27 and @xmath28 as @xmath29\ ] ] in the super - hilbert space of the operators . if a set of @xmath30 conserved linearly independent operators @xmath31 with @xmath32=0 $ ] is known , one may assume their orthonormality @xmath33 provided by a gram - schmidt process . then , we expand the operator of interest @xmath5 @xmath34 in this incomplete operator basis where @xmath35 and @xmath36 is the remaining rest with @xmath37 @xmath38 . substituting into the definition yields @xmath39 with @xmath40 . this relies on the constancy of ( i ) @xmath41 , of ( ii ) @xmath42 , and of ( iii ) @xmath43 all stemming from @xmath44=0 $ ] . for the last relation we have used the cyclic invariance of the trace and @xmath45=0 $ ] . if we knew @xmath46 , we would deduce @xmath47 . but in general this does not hold because @xmath36 may still contain a non - decaying part . but implies mazur s inequality @xmath48 for a given @xmath4 , the complete set of conserved operators @xmath49 is spanned by all pairs of energy - degenerate eigenstates @xmath50 the elements of @xmath49 are orthonormal with respect to the scalar product . the coefficent @xmath51 of @xmath52 takes the value @xmath53 so that the right hand side of equals @xmath26 as given by the lehmann representation . thus , the inequality is tight because it becomes exact for the _ complete _ set @xmath49 of conserved operators . the physical interpretation of eq . is straightforward in the heisenberg picture if we view the time - dependent observable @xmath54 as super vector . its components parallel to conserved quantities ( super vector directions ) are constant in time because these quantities commute with the hamiltonian . but all other components , which are perpendicular to the conserved super subspace , finally decay . if not all conserved operators are considered , the r.h.s . of decreases and only the inequality holds . generally , if _ any _ subspace of the space spanned by @xmath49 is considered mazur s inequality holds . one does not need to know the complete set of eigenstates of @xmath4 in order to calculate a lower bound : any finite ( sub)set of conserved operators is sufficient . now we proceed to generalize mazur s inequality for easy - to - use application . usually , some conserved operators @xmath55 are known but they are not necessarily orthonormal in general . rather their overlaps yield a hermitian , positive norm matrix @xmath56 with matrix elements @xmath57 . each operator @xmath55 can be represented as a linear superposition of the complete set of orthonormal @xmath31 . these superpositions can be summarized in a matrix @xmath58 so that @xmath59 where the vectors @xmath60 and @xmath61 contain the operators @xmath31 and @xmath55 as coefficients ; @xmath62 is the complex ( not hermitian ! ) conjugate of @xmath58 . a short calculation shows that @xmath63 . if we define the vector @xmath64 with complex components @xmath65 , the bound @xmath66 can be expressed by @xmath67 . in analogy , we compute @xmath68 with complex components @xmath69 . obviously , @xmath70 holds and the lower bound is computed by @xmath71 without resorting to orthonormalized operators , relying only on the scalar products of @xmath55 and @xmath5 . we have successfully eliminated the construction of a subset of orthogonal operators @xmath31 and related the lower bound to some known set of linear independent unnormalized conserved operators @xmath55 . the general lower bound is our first key result . a possible route to generalizations to various initial states is sketched in the supplement . [ [ central - spin - model . ] ] central spin model . + + + + + + + + + + + + + + + + + + + the hamiltonian of the csm reads @xmath72 where we assume all spins to be @xmath18 for simplicity . it is a generic model to study the interaction between a two - level system and a bath of spins or more generally a set of subsystems with finite number of levels . currently , it is intensively investigated for understanding the decoherence and dephasing in possible realizations of quantum bits @xcite . theoretical tools comprise chebyshev polynomial technique @xcite , perturbative approaches @xcite , generalized master equations @xcite , equations of motion @xcite various cluster expansions @xcite , bethe ansatz @xcite , density - matrix renormalization @xcite , and studies of the classical analogue @xcite . by focusing on @xmath73 , the correlation function defined in reveals important information on the decay of the central spin . due to isotropy no other components of the central spin need to be considered . given the smallness of the hyperfine couplings ( @xmath74 is in the range of @xmath75ev corresponding to percents of a kelvin @xcite ) the experimentally relevant temperature can be considered as infinite , and we take the spin system to be completely disordered , i.e. , @xmath76 , prior to the preparation of an initial state of the central spin , cf . supplement . for classical spins @xmath77 , there are strong analytical arguments that a fraction of central spin correlations persists unless there is a diverging number of arbitrarily weakly coupled spins in the bath @xcite . in the quantum case smaller systems have been studied and evidence for a non - decaying fraction of spin polarization @xcite has only be compiled in fairly small ( @xmath78 ) systems or up to fairly short times @xcite . based on the generalized mazur s inequality , we are able to address the nature and the lower bound of these non - decaying correlations for arbitrary system sizes . the total spin @xmath79 could serve as a first guess for a useful conserved quantity . only the @xmath80-component @xmath81 has an overlap @xmath82 ( we omit the subscript @xmath83 for brevity ) . the norm @xmath84 takes the value @xmath85 so that provides @xmath86 . irrespective of the considered distribution of the couplings @xmath74 , using only @xmath87 as single conserved operator does not provide a meaningful lower bound for thermodynamically large , or infinite baths . the next important conserved quantity is the energy @xmath88 itself . but , of course , @xmath89 because @xmath88 is a scalar and @xmath90 a vector component . the @xmath80-component of the product @xmath91 , @xmath92 , clearly fulfills @xmath93=0 $ ] and defines a conserved composite vector operator . we find @xmath94 where @xmath95 and @xmath96 . with this input eq . yields @xmath97 this bound remains finite for @xmath98 if the @xmath74 are drawn from a probability distribution @xmath99 with average @xmath100 and variance @xmath101 . for large @xmath102 one has @xmath103 and @xmath104 so that @xmath105 $ ] ensues for @xmath98 . this is a finite lower bound unless the average values @xmath100 vanishes . this rigorous bound is our second key result . for @xmath106 bath spins with @xmath107 , but normalized such that @xmath108 is the unit of energy , and various @xmath109 defined in . the inset compares @xmath26 from the average of the numerical data with @xmath110 $ ] to @xmath66 obtained from for the 3 quantities ( @xmath111 ) or for all quantities ( @xmath112 with @xmath113 ) . the estimates from the overhauser correlations @xmath114 are also shown . [ fig : ss ] ] for any finite system with non - vanishing sum @xmath115 , eq . provides a rigorous finite lower bound which is very easy to compute for any given set of couplings . it can serve to check the validity of numerical results such as provided in refs . @xcite . generically , distributions of the @xmath74 have finite values @xmath100 and @xmath116 . this is the case for nuclear spins in molecules @xcite or nv centers in diamond @xcite because the spin baths are finite . in quantum dots , the convergence and existence @xmath115 and @xmath117 is ensured even for arbitrary number of spins because the couplings are bounded from above , but become arbitrarily small due to exponential tails of the electron wave function @xcite . this leads to vanishing @xmath100 implying complete decay for infinite times . for large , but finite times , however , our results include the possibility of slow decays @xmath118 previously advocated for infinitely large spin baths @xcite . assuming exponential scaling for the couplings @xmath119 , where @xmath120 is inversely proportional to the number of relevant bath spins , to @xmath121 . it is not the total number of nuclear spins which is of the order of avogadro s constant . ] it is clear that @xmath115 and @xmath122 converge quickly for @xmath98 so that eq . implies @xmath123 . chen et al . @xcite have argued that at any given _ finite _ time @xmath16 , only those spins @xmath124 with couplings @xmath125 significantly influence the real - time dynamics of the central spin . hence , only an effective number @xmath126 of spins contribute to the correlation function implying @xmath127 for such a distribution function . the lower bound can be improved by considering the three conserved observables @xmath87 , @xmath128 , and @xmath129 . the required vector and matrix elements are given in the supplemental material . still the bound does not exhaust the numerically found value as depicted in the inset of fig . [ fig : ss ] for @xmath130 ( @xmath109 makes the system non - integrable , it will be defined in ) . even resorting to the integrability of the csm @xcite which implies @xmath131 $ ] with @xmath132 and @xmath133 does not account for the full non - decaying fraction obtained in finite size calculations @xcite , see circle in the inset of fig.[fig : ss ] . the bound has been computed considering @xmath87 and @xmath134 for @xmath113 ( for matrix elements see supplement ) . the above results suggest that the integrability is not the key ingredient for a finite non - decaying fraction . to support this claim we extend the hamiltonian by adding one extra coupling @xmath135 @xmath136 between the most weakly and the most strongly coupled bath spin , defined to be at @xmath137 and @xmath102 , respectively . its value @xmath109 is chosen to be @xmath138 so that it constitutes a sizable perturbation even for large spin baths . the modified time - dependence of @xmath14 is depicted for various @xmath109 in fig . [ fig : ss ] . a finite @xmath109 spoils the integrability completely @xcite , but leaves the quantities @xmath139 conserved . these three constants of motion generic for isotropic spin models are used to obtain the lower bound ( red curve ) in the inset of fig . [ fig : ss ] . obviously , @xmath66 is decreased smoothly and only moderately upon increasing @xmath109 in line with the numerically determined @xmath26 . there is no abrupt jump to zero , in contrast to what is known for the drude weight . the conclusion that integrability is only secondary for the non - decaying spin correlation is our third key result . at present it remains an open question which conserved quantities one has to include to yield a tight lower bound . we presume that higher powers of @xmath4 , for instance @xmath140 , have to be considered . such studies are more tedious and left for future research . instead , we take a mathematically less rigorous route based on the estimate by merkulov et al.@xcite @xmath141 where @xmath142 is the correlation of the overhauser field operator @xmath143 . note that an arbitrary @xmath144 can be included because @xmath145 differs from @xmath88 in only by an irrelevant constant for spin @xmath146 . this estimate was derived for a classical , large overhauser field @xcite and prevails in the thermodynamic limit of the quantum case : the overhauser field becomes a classical variable upon @xmath98 as shown in ref . @xcite . thus we now apply the general approach to @xmath147 . considering only @xmath148 as conserved operator already yields a meaningful lower bound for the overhauser field correlation function for @xmath98 @xmath149 recall @xmath150 and @xmath151 if the couplings are drawn from a normalized distribution function @xmath99 . this lower bound can be optimized by choosing the arbitrary value @xmath144 such that the bound becomes maximal . with the matrix elements given in the supplement @xmath152 can be improved considering the three constants @xmath139 or all integrals @xmath87 and @xmath153 . the results are also included in fig . [ fig : ss ] ( triangle and square symbols ) . they hold only for @xmath130 because the estimate applies only in this case . remarkably , the resulting estimates for @xmath26 seem to be tight . in particular , the easily evaluated estimate based on all integrals reproduces the numerically found @xmath26 to its accuracy . we applied the same estimate to the case @xmath119 studied by stochastically evaluating the bethe ansatz equations and found excellent agreement with the published data with @xmath154 in ref.@xcite as well . thus we conjecture that the non - decaying fraction @xmath26 in the central spin model is quantitatively described by @xmath155 if @xmath152 is determined from the @xmath156 integrals @xmath87 and @xmath157 . this constitutes our fourth key result . the small difference , however , between triangle ( from three constants of motion ) and square ( from @xmath156 constants of motion ) in fig . [ fig : ss ] indicates again that the significance of the integrability is limited . in summary , four key results are obtained : ( i ) an easy - to - use version of mazur s inequality to prove persisting correlations ; ( ii ) a rigorous finite lower bound for the infinite - time spin correlation in the csm , valid for the infinite system if the average coupling is finite ; ( iii ) only a small part of the persisting correlation is due to the integrability ; ( iv ) a quantitative estimate for the persisting correlation is conjectured , based on the overhauser field . clearly , the generalized inequality calls for application to other problems @xcite . the approach is easy to evaluate and can be used for very large systems and large numbers of constants of motion . thus it can prove fruitful in the intensely studied field of integrable systems , for instance in estimating drude weights . in the context of coherence in particular , various extensions of the csm , e.g. , by magnetic fields , anisotropies , or more intra - bath couplings suggest themselves to be investigated in the presented manner . we gratefully acknowledge helpful discussions with m. brockmann , a. faribault , a. greilich , and d. schuricht . financial support was given by the mercator stiftung under pr-2011 - 0003 ( gsu ) , by the studienstiftung des deutschen volkes ( ds ) and by the deutsche forschungsgemeinschaft under an 275/7 - 1 ( fba ) and uh 90/9 - 1 ( gsu ) . 43ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrevb.89.045317 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )
it has been generally accepted for decades that disk galaxies are ubiquitously embedded in massive dark matter halos . during the past few years , a consensus has been developing that elliptical galaxies also contain dark halos . their existence is predicted by hierarchical theories of galaxy formation , and has recently been supported by various observational evidence , such as gravitational lensing ( griffiths et al . 1996 ; keeton , kochanek , & falco 1998 ) and x - ray measurements of their hot gas atmospheres ( matsushita et al . 1998 ; loewenstein & white 1999 ) . although useful to infer the large - scale mass distribution of elliptical galaxies , these observations do not sufficiently constrain the detailed structure of the mass distribution at a few effective radii . this region is particularly important for understanding the coupling of the dark and luminous matter . in order to constrain the gravitational potential at these radii , other , kinematical , tracers are used . for disk galaxies , the neutral hydrogen gas , which radiates at 21 cm , forms an excellent kinematical tracer ; elliptical galaxies , however , usually lack the necessary amounts of interstellar gas . discrete tracers such as planetary nebulae , globular clusters and dwarf satellite galaxies can be used ( zepf et al . 2000 ; romanowski & kochanek 2001 ; kronawitter et al . 2000 ) , but due to their small numbers and larger distances to the center , they do not sufficiently constrain the gravitational potential at a few @xmath0 . this leaves stars as the main tracer for the mass distribution in elliptical galaxies in this region . the first stellar kinematical evidence for dark matter halos around elliptical galaxies came in the early 1990s . at that time , the available kinematical data consisted of the mean projected velocity @xmath1 and the projected velocity dispersion @xmath2 , at projected radii rarely larger than @xmath3 . for a number of elliptical galaxies , the projected velocity dispersion profile was found to decrease only slowly with projected radius , a behavior that was interpreted as a signature for the presence of a dark matter halo ( saglia , bertin , & stiavelli 1992 ; saglia et al . however , such @xmath2 profiles can also be generated by intrinsically tangentially anisotropic galaxy models , without the need for dark matter halos . the velocity dispersion profile alone does not contain sufficient kinematic information to constrain both the mass and the orbital structure of elliptical galaxies , a problem usually referred to as the mass - anisotropy degeneracy ( gerhard 1993 ) . this degeneracy can be broken by considering the higher order kinematical information contained in the line - of - sight velocity distributions ( losvds ) , usually parameterized by means of the gauss - hermite shape parameters @xmath4 ( where @xmath5 ; gerhard 1993 ; van der marel & franx 1993 ) . in particular , the additional information contained in the @xmath6 profile provides the key to breaking the mass - anisotropy degeneracy : the combination of a slowly decreasing velocity dispersion profile , together with a relatively large @xmath6 profile , is generally interpreted as evidence for a dark matter halo . profiles , we refer to section 3 of gerhard et al . ( 1998 ) . ] thanks to improved instrumentation and data reduction techniques , it is nowadays possible to determine @xmath1 , @xmath2 , @xmath7 and @xmath6 profiles with a reasonable accuracy , out to several @xmath0 . several authors have recently adopted such kinematical information to constrain the dark matter distribution in a number of elliptical galaxies ( rix et al . 1997 ; gerhard et al . 1998 ; saglia et al . 2000 ; kronawitter et al . 2000 ) . in this entire discussion , it has ( implicitly ) been assumed that elliptical galaxies consist of two dynamically important components : the stars moving as test particles in a gravitational potential , generated by both stellar and dark mass . during the past decade , it has become well established that elliptical galaxies also contain a surprisingly large amount of interstellar dust ( up to several million solar masses ) , most of it believed to be distributed diffusely over the galaxy ( roberts et al . 1991 ; goudfrooij & de jong 1995 ; bregman et al . this number must be revised an order of magnitude upward , if more detailed dust mass estimators ( merluzzi 1998 ) or additional submillimeter measurements ( fich & hodge 1993 ; wiklind & henkel 1995 ) are taken into account . this is still a negligible fraction of the total mass of the galaxy , such that the dust will hardly influence the gravitational potential . nevertheless , it has a significant role in galaxy dynamics : dust grains efficiently absorb and scatter optical photons . interstellar dust will therefore affect all observable quantities , including the observed kinematics . we are undertaking an effort to understand the impact of interstellar dust on the observed kinematics in elliptical galaxies . previously , we investigated how absorption by dust grains affects the light profile and the observed kinematics ( baes & dejonghe 2000 ; baes , dejonghe , & de rijcke 2001 ) . we found that the observed kinematics are affected only in the most central regions , the magnitude of these effects being of the order of a few percents . here we have extended our modeling to incorporate the process of scattering off dust grains . we will show in this letter that this has a considerable effect on the observed kinematics , in particular concerning the stellar kinematical evidence for dark matter halos . to investigate the effects of attenuation on the observed kinematics , we constructed a spherically symmetric elliptical galaxy model , consisting of a stellar and a dust component . for the stellar distribution , we adopted an isotropic hernquist model . it is a reasonable approximation for the observed surface brightness distribution of elliptical galaxies , and has the advantage that most of the internal stellar kinematics can be calculated analytically ( hernquist 1990 ) . the distribution of the ( diffuse ) dust in elliptical galaxies is not well constrained ; its presence has been demonstrated only indirectly . we adopted a dust distribution that is shallower than the stellar density , with a total visual optical depth of order unity . such a model seems to be indicated by the observed color gradients in elliptical galaxies ( witt et al.1992 ; goudfrooij & de jong 1995 ; wise & silva 1996 ) . for the optical properties of the dust grains ( the relative opacity , the scattering albedo , and the phase function ) , we adopted the values calculated by maccioni & perinotto ( 1994 ) and displayed in di bartolomeo , barbaro & perinotto ( 1995 ) . a more extended set of models , with a larger variety in optical depth and star - dust geometry , will be presented in a forthcoming paper ( m. baes & h. dejonghe , 2001 , in preparation ) . the inclusion of dust absorption in the calculation of the losvds was rather straightforward : it required an extra factor in the integration along the line of sight ( baes & dejonghe 2000 ) . if scattering is included , however , such an approach is not possible anymore . indeed , photons can now leave their initial path , and reach the observer through a completely different path . therefore , we adopted a monte carlo technique to include this process . basically , the method consists of following the individual path of a very large number of photons ( typically about @xmath8 ) , selected randomly from phase space , according to the phase space distribution function of the galaxy . knowledge of the path along which each photon leaves the galaxy , together with the kinematical signature of the star that emitted it , enables us to construct the surface brightness profiles and losvds of the galaxy , both with and without dust attenuation taken into account . the @xmath2 and @xmath6 profiles are extracted from the losvds ( the @xmath9 and @xmath7 profiles are identically zero , because we consider a spherically symmetric nonrotating galaxy model ) . a more detailed description of the method will be given in m. baes & h. dejonghe ( 2001 , in preparation ) . a check on the method was provided by the case where only absorption is taken into account : we found the results of the monte carlo simulation to be in perfect agreement with our previously obtained semianalytical results ( baes & dejonghe 2000 ) . in figure 1 , the effect of dust attenuation on the observed kinematics is illustrated . for the lines of sight that pass near the galaxy center , the kinematics are nearly unaffected . because of the modest ( and conservative ) dust masses we used , the high - velocity stars in the galaxy center are only slightly obscured . as a result , the velocity dispersion only decreases marginally . at large radii , however , the kinematics are seriously affected . the @xmath2 profile drops less steeply , and the @xmath6 parameter is significantly larger compared to the dust - free case . in figure 2 , we show the losvd at @xmath10 of our galaxy model , with and without dust attenuation . this figure clearly indicates that dust attenuation brings on significant high - velocity wings in the outer losvds . in particular , it should be noted that these losvds do not vanish at the local escape velocity : hence , stars are observed that can physically not be present on these lines of sight . these wings are a scattering effect , as demonstrated in figure 3 , and they are responsible for the increase of @xmath2 and @xmath6 . attenuation by interstellar dust apparently has a kinematical signature that is strikingly similar to the presence of a dark matter halo : a velocity dispersion profile that decreases more slowly than expected , and a relatively large @xmath6 profile . hence , dust - affected kinematics mimic the presence of a dark matter halo . to check this into more detail , we considered our dust - affected kinematics as an observational data set , and modeled it as any dynamical modeler would do , i.e. , without taking dust attenuation into account . we concentrated on the model with @xmath11 , and considered a data set consisting of the @xmath12-band photometry and @xmath13-band @xmath2 and @xmath6 profiles , out to @xmath14 . the modeling was performed with a powerful non - parametric modeling technique based on quadratic programming ( dejonghe 1989 ) . first , we tried to find out whether the data were consistent with a model with a constant @xmath15 . we constructed a set of dynamical models , with @xmath15 as a free parameter . the best - fitting model is represented by the dashed line in figure 4 . obviously , this fit is not satisfactory : it can fit the @xmath2 profile , but only through a strong tangential anisotropy , reflected in a strongly negative @xmath6 profile . it is impossible to fit both the @xmath2 and @xmath6 profiles with a constant @xmath15 model . in order to construct models with a rising @xmath15 , i.e. with a dark halo , we considered a set of hernquist potentials , each with a different scale length . we constructed a set of dynamical models , with both the potential scale length and the mass - to - light ratio as free parameters . the best fit to the photometry and the @xmath2 and @xmath6 profiles is plotted in solid lines in figure 4 . its potential has a scale length 45% larger than the original hernquist potential , and in the maximum stellar mass hypothesis ( the analogue for the maximum disk hypothesis in disk galaxies ) , the dark matter contributes roughly a third of the total mass within 1@xmath0 , and half of the total mass within the last data point . this clearly demonstrates that the effects of dust attenuation can mimic the presence of a dark matter halo . taking dust attenuation into account in dynamical modeling procedures will hence reduce or may even eliminate the need for a dark matter halo at a few @xmath0 . in view of these results , we may have to reconsider the stellar kinematical evidence for dark matter halos . in analogy with the mass - anisotropy degeneracy , which could be lifted by considering higher - order shape parameters for the losvds , we are now faced with a new degeneracy , which could be called the mass - dust degeneracy . indeed , the attenuation ( in particular the scattering ) by dust grains has the same effect on the stellar kinematics as a dark matter halo . at least , these results apply for the model we have explored in this letter , with a dust distribution shallower than the stars and with an optical depth @xmath11 . we are well aware that these assumptions are quite uncertain , and an extended set of models , with a large variety in optical depth and star - dust geometry , is being investigated . the new mass - dust degeneracy strongly complicates the use of stellar kinematics as a tracer for the mass distribution in elliptical galaxies . although the presence of dark matter halos nowadays seems firmly established at very large scales , this leaves us with a major problem concerning the determination of the dark halo properties at a few effective radii . there are two possible ways to break this new degeneracy . the first option is to observe the kinematics at near - infrared wavelengths , where the effects of dust are negligible . this , however , poses a serious observational challenge in the off - center regions . the second option is to include radiative transfer calculations in dynamical modeling techniques . besides a computational challenge , this requires a better knowledge of the spatial distribution and of the optical properties of dust grains in elliptical galaxies than we have today . the new generation of far - infrared and submillimeter instruments , such as alma and _ sirtf _ , will help to solve this problem . the authors would like to thank t. van albada and a. n. witt for stimulating discussions , and the referee , n. cretton , for his constructive remarks . m. b. acknowledges the financial support of fwo - vlaanderen . baes m. , dejonghe h. , 2000 , mnras , 313 , 153 baes m. , dejonghe h. , de rijcke s. , 2000 , mnras , 318 , 798 bregman j. n. , snider b. a. , grego l. , cox c. v. , 1998 , apj , 499 , 670 dejonghe h. , 1989 , apj , 343 , 113 di bartolomeo a. , barbaro g. , perinotto m. , 1995 , mnras , 277 , 1279 fich m. , hodge p. , 1993 , apj , 415 , 75 gerhard o. e. , 1993 , mnras , 265 , 213 gerhard o. , jeske g. , saglia r. p. , bender r. , 1998 , mnras , 295 , 197 goudfrooij p. , de jong t. , 1995 , a&a , 298 , 784 griffiths r. e. , casertano s. , i m m. , ratnatunga k. u. , 1996 , mnras , 282 , 1159 hernquist l. , 1990 , apj , 356 , 359 keeton c. r. , kochanek c. s. , falco e. e. , 1998 , apj , 509 , 561 kronawitter a. , saglia r. p. , gerhard o. , bender r. , 2000 , a&as , 144 , 53 loewenstein m. , white r. e. iii , 1999 , apj , 518 , 50 maccioni a. , perinotto m. , 1994 , a&a , 284 , 241 matsushita k. , makishima k. , ikebe y. , rokutanda e. , yamasaki n. , ohashi t. , 1998 , apj , 499 , l13 merluzzi p. , 1998 , a&a , 338 , 807 rix h .- w . de zeeuw p. t. , cretton n. , van der marel r. p. , carollo c. m. , 1997 , apj , 488 , 702 roberts m. s. , hogg d. e. , bregman j. n. , forman w. r. , jones c. , 1991 , apjs , 75 , 751 romanowsky a. j. , kochanek c. s. , 2001 , apj , 553 , 722 saglia r. p. , bertin g. , stiavelli m. , 1992 , apj , 384 , 433 saglia r. p. , bertin g. , bertola f. , danziger j. , dejonghe h. , sadler e. m. , stiavelli m. , de zeeuw p. t. , zeilinger w. w. , 1993 , apj , 403 , 567 saglia r. p. , kronawitter a. , gerhard o. , bender r. , 2000 , aj , 119 , 153 van der marel r. p. , franx m. , 1993 , apj , 407 , 525 wiklind t. , henkel c. , 1995 , a&a , 297 , l71 wise m. s. , silva d. r. , 1996 , apj , 461 , 155 witt a. n. , thronson j. a. jr . , capuano j. m. jr . , 1992 , apj , 393 , 611 zepf s. e. , beasley m. a. , bridges t. j. , hanes d. a. , sharples r. m. , ashman k. m. , geisler d. , 2000 , aj , 120 , 2928 ; the different curves correspond to those in fig . 1 . the losvds are normalized to unity at @xmath17 . the escape velocity @xmath18 at this line of sight is indicated by an arrow . if dust attenuation is not taken into account , no stars are observed with line - of - sight velocities larger than @xmath18 . if dust attenuation is considered , however , the losvd does not vanish at @xmath18 . ] . if there were no dust attenuation , the photon would travel along a straight line , and contribute a line - of - sight velocity @xmath19 to the losvd at the projected radius @xmath20 . in a dusty galaxy , however , photons can leave their initial path and reach the observer through a completely different path . in the figure , the photon is scattered twice , at @xmath21 and @xmath22 , and it leaves the galaxy along a line of sight with projected radius @xmath23 . during its journey through the galaxy , the photon carries along the kinematical signature of the star that emitted it . hence , it will contribute the line - of - sight velocity @xmath24 to the losvd at the line of sight @xmath23 . in particular , it is possible that the observed line - of - sight velocity @xmath24 exceeds the local escape velocity @xmath25 . this is indeed observed in the outer losvds : they do not vanish at the local escape velocity ( fig . 2).,width=245 ] . shown are the surface brightness @xmath26 , velocity dispersion @xmath2 , and @xmath6 profiles . the data are represented by the gray circles . the dashed line is the best fitting model with constant @xmath15 ; the solid line is the best fitting model with a dark matter halo ( see text ) . the projected kinematics of the input model , without dust attenuation taken into account , are represented by a dotted line . ]	hierarchical models of galaxy formation and various observational evidence suggest that elliptical galaxies are , like disk galaxies , embedded in massive dark matter halos . stellar kinematics are considered the most important tracer for this dark halo at a few effective radii . using detailed modeling techniques , several authors have recently presented stellar kinematical evidence of a dark halo for a number of elliptical galaxies . in these modeling techniques , dust attenuation ( absorption and scattering of starlight by dust grains ) has not been taken into account . nevertheless , elliptical galaxies contain a significant amount of interstellar dust , which affects all observable quantities , including the observed kinematics . we constructed a set of dynamical models for elliptical galaxies , in which dust attenuation is included through a monte carlo technique . we find that a dust component , shallower than the stellar distribution and with an optical depth of order unity , affects the observed kinematics significantly , in the way that it mimics the presence of a dark halo . if such dust distributions are realistic in elliptical galaxies , we are faced with a new mass - dust degeneracy . taking dust attenuation into account in dynamical modeling procedures will hence reduce or may even eliminate the need for a dark matter halo at a few effective radii .
in the following , we formulate a one - electron microcanonical distribution for triatomic molecules . we denote the positions of the nuclei by @xmath1 , @xmath2 and @xmath3 and the inter - nuclear distances by @xmath4 , @xmath5 and @xmath6 , see fig . one can show that the coordinates of the nucleus c are expressed in terms of the inter - nuclear distances as follows : we denote the position vector of the electron by @xmath8 and the distances of the electron from nuclei a , b and c by @xmath9 , @xmath10 and @xmath11 , respectively . we then define the confocal elliptical coordinates @xmath12 and @xmath13 using the nuclei a and b as the foci of the ellipse , that is , where @xmath15 and @xmath16}$ ] . the third coordinate @xmath17}$ ] is the angle between the projection of the position vector @xmath8 on the xy plane and the positive x axis ; it thus defines the rotation angle around the axis that passes through nuclei a and b. the potential of the electron in the presence of the nuclei a , b and c , which have charges @xmath18 , @xmath19 and @xmath20 , respectively , is given by @xmath21 this potential is expressed in terms of the confocal elliptical coordinates as follows the one - electron microcanonical distribution is given by @xmath23 where @xmath24 is the ionization energy of the one - electron triatomic molecule . note that the energy is given by @xmath25 . the electron momentum in terms of the confocal elliptical coordinates is expressed as follows the @xmath36 distribution goes to zero and is thus well - behaved when the electron is placed on top of either nucleus a or b. however , when @xmath37 , i.e. , the electron is placed on top of nucleus c , @xmath38 . we eliminate this singularity by introducing an additional transformation . setting @xmath39 , @xmath40 and expanding @xmath41 around @xmath42 we find @xmath43 where @xmath44 and @xmath45 are the values of @xmath12 and @xmath13 , respectively , when the electron is placed on top of the nucleus c. to eliminate the singularity in eq . ( [ eq : sing ] ) , we introduce a new variable @xmath46 , such that @xmath47 . the new distribution takes the form since @xmath16}$ ] , @xmath49 and @xmath46 take both negative and positive values and therefore , if we choose one @xmath50 for all values of @xmath13 , @xmath50 must be odd . moreover , to avoid the singularity when the electron is placed on top of nucleus c , @xmath50 must be such that @xmath51 , i.e. , @xmath52 . combining the above two conditions , yields @xmath53 . the new distribution @xmath54 goes to zero when the electron is placed on top of the nucleus c , i.e. , when @xmath55 , @xmath56 and @xmath57 . to set up the initial conditions we find @xmath58 so that @xmath59 and equivalently @xmath60 we then find the maximum value @xmath61 of the distribution @xmath54 , for the allowed values of the parameters @xmath12 , @xmath46 and @xmath62 . we next generate the uniform random numbers @xmath63}$ ] , @xmath64}$ ] , @xmath65}$ ] and @xmath66}$ ] , with @xmath67 and @xmath68 . if @xmath69 then the generated values of @xmath12 , @xmath46 and @xmath62 are accepted as initial conditions , otherwise , they are rejected and the sampling process starts again . following the above described formulation , we obtain the initial conditions of the electron with respect to the origin of the coordinate system . to obtain the initial conditions for the position of the electron with respect to the center of mass of the triatomic molecule , @xmath70 , in terms of the ones with respect to the origin , @xmath8 , we shift the coordinates by @xmath71 , where @xmath72 is given by @xmath73 with @xmath74 with @xmath75 , @xmath76 and @xmath77 the masses of the nuclei . as an example , we next obtain the probability densities of the position and the momentum of the electron that is initially bound in @xmath78 when the molecule is driven by an intense infrared laser field . we assume the other electron tunnel - ionizes in the initial state . we consider the @xmath78 triatomic molecule in its ground state , where the distance of the nuclei in the equilateral triangle arrangement is 1.65 a.u . and the first and second ionization energies are @xmath79 a.u . and @xmath80 a.u . , respectively . we find the ionization potentials and equilibrium distances of the initial state using molpro , which is a quantum chemistry package @xcite . for the microcanonical distribution the relevant ionization energy is @xmath81 , since @xmath82 is associated with the electron that tunnel - ionizes in the initial state . in fig . [ position ] ( b ) we plot the probability density of the position of the electron on the x - z plane for @xmath83 using the above described microcanonical distribution . to compare , in fig . [ position ] ( a ) we plot the quantum mechanical probability density of the position of the electron on the x - z plane . that is , we plot @xmath84 , where @xmath85 is the quantum mechanical wavefunction for the @xmath86 molecule , which we obtain using molpro . the two plots , fig . [ position ] ( a ) and ( b ) , show that the two probability densities for the electron position compare well . however , the microcanonical probability density underestimates the electron probability density between the nuclei and overestimates the electron probability density around the nuclei . [ ht ! ] plane . the middle panel shows the microcanonical probability density of the electron momentum plotted on the @xmath87 plane for all values of @xmath88 . the right panel shows the projections on the @xmath89 axis of the probability densities plotted in fig . [ momentum ] ( a ) and ( b ) . , title="fig:",scaledwidth=50.0% ] in addition , in fig . [ momentum ] ( b ) for all values of the electron momentum component along the y - axis , @xmath88 , we plot the probability density of the electron momentum on the @xmath90 plane using the microcanonical distribution . to compare , in fig . [ momentum ] ( a ) we plot @xmath91 , which we obtain by first computing the quantum mechanical wavefunction in momentum space using the quantum mechanical wavefunction @xmath85 computed from molpro and then by integrating over @xmath88 @xmath93 the two plots , fig . [ momentum ] ( a ) and ( b ) , show that the two probability densities for the electron momentum compare well . however , the microcanonical probability density overestimates the higher values of the electron momentum . this can be seen more clearly in fig . [ momentum ] ( c ) where we plot the probability density of the electron momentum along the @xmath89 axis both quantum mechanically and using our microcanonical distribution . to obtain the plots in fig . [ momentum ] ( c ) we project the probability densities of the electron momentum in fig . [ momentum ] ( a ) and ( b ) on the @xmath89 axis . [ momentum ] ( c ) clearly shows that the probability density of the electron momentum obtained from the microcanonical distribution overestimates the higher values of the momentum component @xmath89 . this is consistent with our previous finding that the microcanonical distribution overestimates values of the electron position around the nuclei resulting to higher values of the momentum . in conclusion , in the current work we have formulated a microcanonical distribution for a general one - electron triatomic molecule . this distribution can be used to describe the initial state of the bound electron in semiclassical models of strongly - driven two - electron triatomic molecules . 10 meckel m , comtois d , zeidler d , staudte a , pavii d , bandulet h c , ppin h , kieffer j c , drner r , villeneuve d m and corkum p b 2008 _ science _ * 320 * 1478 lefebvre c , lu h z , chelkowski s and bandrauk a d 2014 _ phys . rev . a _ * 89 * 023403 liu j , ye d f , chen j and liu x 2007 _ phys . _ * 99 * 013003 emmanouilidou a and staudte a 2009 _ phys . a _ * 80 * 053415 emmanouilidou a , lazarou c , staudte a and eichmann u 2012 phys . rev . a * 85 * 011402 ( r ) hugh h , lazarou c and emmanouilidou a 2014 _ phys . rev . a _ * 90 * 053419 landau l d and lifshitz e m 1977 _ quantum mechanics _ ( pergamon press , new york ) ; delone n b and krainov v p 1991 _ j. opt . b _ * 8 * 1207 murray r , spanner m , patchkovskii s and ivanov m y 2011 _ phys . lett . _ * 106 * 173001 meng l , reinhold c o and olson r e 1989 _ phys . a _ * 40 * , 3637 werner h j. knowles p j , knizia g , manby f , schtz r m , et al . 2012 _ molpro , version 2012.1 , a package of ab initio programs _ , see http://www.molpro.net	we formulate a microcanonical distribution for an arbitrary one - electron triatomic molecule . this distribution can be used to describe the initial state in strongly - driven two - electron triatomic molecules . namely , in many semiclassical models that describe ionization of two - electron molecules driven by intense infrared laser fields in the tunneling regime initially one electron tunnels while the other electron is bound . the microcanonical distribution presented in this work can be used to describe the initial state of this bound electron . the nonlinear response of multi - center molecules to intense laser fields is a fundamental problem . for instance , understanding the break - up dynamics of strongly - driven molecules paves the way for controlling and imaging molecular processes @xcite . semi - classical models are essential in understanding the dynamics during the fragmentation of multi - center molecules driven by intense infrared laser pulses . one reason is that treating the dynamics of electrons and nuclei at the same time poses an immense challenge for fully ab - initio quantum mechanical calculations . currently quantum mechanical techniques can only address one electron in triatomic molecules in two - dimensions @xcite . semi - classical models have provided significant insights , for example , in double ionization of strongly - driven @xmath0 with fixed nuclei " @xcite and in frustrated " double ionization during the fragmentation of strongly - driven @xmath0 @xcite , where one electron eventually stays bound in a highly excited state of the h atom . the initial state that is commonly employed by semi - classical models for strongly - driven two - electron atoms and molecules , for intensities in the tunneling regime , involves one electron that tunnel - ionizes in the field - lowered coulomb potential and another electron that remains bound . the electron that tunnel - ionizes emerges from the barrier with a zero velocity along the direction of the laser field , while its velocity perpendicular to the laser field is given by a gaussian distribution @xcite . the tunneling rate can be obtained using semi - classical treatments , for instance see @xcite for atoms and @xcite for molecules . the electron that is initially bound is commonly described in semi - classical models by a microcanonical distribution . to our knowledge , in the literature , a microcanonical distribution is available only for diatomic molecules @xcite . in this work , we formulate a microcanonical distribution for any one - electron triatomic molecule which can also describe the initial state of the bound electron in the above described semi - classical models .
the ultracold bosonic atoms in optical lattices have opened a new window to study many - particle quantum physics in a uniquely controlled manner @xcite . various schemes have been proposed to realize a wide range of many - body hamiltonians by loading ultracold atoms into a certain optical lattice @xcite . the advances in the manipulation of atom - atom interactions by feshbach resonances @xcite allowed experimental study of many - body systems accessible to the full regime of interactions , even to the very strongly interacting tonks - gas limit @xcite . recently a lot of exciting experiments in optical lattices have been implemented , including the superfluid - mott - insulator transition @xcite , non - linear self - trapping in a periodical optical lattice @xcite and repulsively bound atom pairs in an optical lattice @xcite . the basic physics of the ultracold atomic systems in optical lattice is captured by the bose - hubbard model ( bhm ) @xcite , which incorporates the competition between the interaction and hopping energy and has been successfully applied to interpret the superfluid - mott - insulator transition . as a fundamental model , the bhm has been widely applied to study the quantum phase transitions and dynamic problems in the optical lattices . very recently , winkler _ et al . _ @xcite have studied the dynamical evolution of the initially prepared bosonic atom pairs in an optical lattice . their experimental results indicate that the atom pairs with strongly repulsive on - site interaction @xmath0 exhibits a remarkably longer lifetime than the system with weakly repulsive interaction @xcite . at first glance , this result is counter - intuition because one may expect the repulsive interaction to separate the particles , instead to bind them together . the experimental result has stimulated theoretical investigations on the dynamics of repulsively bound pairs @xcite . in ref.@xcite , the theoretical understanding of the stable pair relies on the analytical solution of a two - particle problem by solving two particle lippmann - schwinger scattering equation on the lattice corresponding to the bose - hubbard hamiltonian@xcite . obviously , this method is only limited to a single repulsively bound pair and is not capable to extend to deal with many - particle dynamic problem . motivated by the experimental progress @xcite , in this paper we study the quantum dynamics of the repulsively pair states in the bhm both analytically and numerically . in the strongly repulsive limit , we develop an analytical method to deal with the dynamical problem based on the perturbation expansion of the hopping terms , whereas we can numerically solve the dynamic problem in the whole regime of interaction for a finite - size system which could be diagonalized exactly by the exact diagonalization method . the bose - hubbard hamiltonian ( bhh ) reads @xcite @xmath1 where @xmath2 is the creation ( annihilation ) operator of bosons on site @xmath3 @xmath4 , @xmath5 counts the number of bosons on the @xmath3th site , and @xmath6 denotes summation over nearest neighbors . the parameter @xmath7 denotes the hopping matrix element to neighboring sites , and @xmath8 represents the on - site interaction due to s - wave scattering . for an actual optical lattice , @xmath9 and @xmath0 are related to the depth of the optical lattice @xmath10 which is determined by the intensity of the laser . the lattice constant @xmath11 is half of the wave length of the laser @xmath12 @xcite . in this article , we focus on the dynamical evolution of the repulsively bound atom pairs in the periodic bose - hubbard model with @xmath13 , i.e. repulsive on - site interaction . in the following calculation , we will ignore all possible dissipations in the system , such as the loss of atoms by three - body collision . the paper is organized as follows . in sec . ii , we first review a general scheme to deal with dynamical evolution and present the exact result for the two - site problem . in sec . iii , we develop a perturbation method to study the dynamic evolution of the initially prepared state of atom pairs which works in the large @xmath0 limit . in sec . iv , we study the dynamical problem for a finite - size system by using the exact diagonalization method and compare the analytical results with the exact numerical results . the bose - hubbard model has been investigated by a variety of theoretical and numerical methods under different cases @xcite . most of the theoretical investigations concern the ground state properties , whereas the quantum dynamic problems are hardly dealt with and most of works are limited to the double - well ( two - site ) problems @xcite . given an initial state @xmath14 the evolution state at time @xmath15 can be formally represented as @xmath16 if we know all the eigenstates of the bhh which are the solutions of schrdinger equation @xmath17 we can get @xmath18 now it is straightforward to obtain the probability of finding the given initial state at time @xmath15 @xmath19 where @xmath20 and @xmath21 $ ] is the basis dimension of the @xmath22-particle system in a lattice with size @xmath23 . if the energy spectrum and its corresponding eigenstates @xmath24 are known , then the dynamical problem is exactly solved in principle . it is obvious that the dimension @xmath25 increases very quickly as the particle number @xmath22 and lattice site @xmath23 increase . therefore it is not practical to investigate the dynamical problem of a large system in this way , though one can solve it numerically for a finite - size system by exact diagonalization method . versus @xmath15 in units of @xmath9 for the two - site - two - boson case . a comparison between the exact and the approximate solution is made the solid lines denote the exact solutions given in sec . the dotted lines denote the approximate solutions by perturbation theory . from top to bottom the lines correspond to @xmath26 , @xmath27 and @xmath28 , respectively . , width=340 ] it is instructive to study the two - site system before presenting the analytical results based on perturbation theory and the numerical results for the systems with larger size . the dynamical evolution of an atom pair in the toy two - site model can be completely solved analytically . the results of this toy model will shed light on studies of larger systems . according to the scheme discussed above , it is quite easy to get the analytical expression of @xmath29 for a pair of atoms initially in a two - site lattice : @xmath30 where @xmath31 , @xmath32 and @xmath33 with the parameters @xmath34 and @xmath35 . the eigenenergies are given by @xmath36 , @xmath37 , and @xmath38 . it is quite easy to show the time evolution of the initial atom pair by the formula ( [ 2exact ] ) . in fig . 1 , we plot the time evolution of @xmath29 for @xmath26 , @xmath27 and @xmath28 . the results show that the larger the repulsive @xmath0 is , the probability of finding the initial atom pair is larger . when @xmath39 , we observe that @xmath40 , @xmath41 , @xmath42 and @xmath43 . in this limit , the terms including of @xmath44 in eq . ( [ 2exact ] ) , which is proportional to @xmath45 , merely give a very tiny contribution to @xmath29 accompanying with a very high - frequency oscillation , whereas the dominant part is given by @xmath46 it is obvious that @xmath47 if @xmath48 . therefore at the large @xmath0 limit , the atom pair is dynamically stable within a very large time scale . in the strongly repulsive interaction limit of @xmath49 , it is convenient to represent the bhh as @xmath50 where @xmath51 and @xmath52 are the on - site interaction part and tunnelling part respectively . we may solve the following dynamical evolution equation @xmath53 approximately by treating @xmath52 as a perturbation . in order to do that , we formally represent @xmath54 as @xmath55 with @xmath56 defined as @xmath57 . it is easy to check that @xmath56 fulfills the following motion of equation @xmath58 with @xmath59 therefore the formal solution of @xmath56 is given by @xmath60 \\ & = & \sum_{n=0}^\infty \frac{\left ( -i\right ) ^n}{n!}\int_0^tdt_1\cdots \int_0^tdt_nt\left [ \widehat{h}_j(t_1)\cdots \widehat{h}_j(t_n)\right ] , \end{aligned}\ ] ] where @xmath61 is the time - ordering operator . by using the above formula , now we can calculate @xmath62 perturbatively . up to the second order , we get @xmath63 from the definition of @xmath64 we see that it should be normalized . however , in the scheme of perturbation calculation , the normalized condition is not automatically fulfilled . in order to overcome this problem and get a well - defined @xmath29 , we need use a normalized @xmath65 which is defined as @xmath66 therefore in the scheme of perturbation expansion , the probability @xmath29 is defined by @xmath67 for the six - site - six - boson case . the comparison between the solution given by perturbation theory and the exact solution from numerical computation is shown . the analytical results are denoted by the dotted lines and the numerical result by solid lines . the top two lines correspond to @xmath26 and the other two lines correspond to @xmath68 . the time is in units of @xmath9.,width=340 ] for a given initial state @xmath69 represented in a basis of fock states @xmath70 with @xmath71 or @xmath72 it is easy to get @xmath51 @xmath73 @xmath74 for convenience , we first calculate the case with @xmath75 . by using the formula ( [ pb ] ) , we get @xmath76 . \label{psi0t}\end{aligned}\ ] ] therefore we have @xmath77 \label{p0t2}\end{aligned}\ ] ] where @xmath78 is the normalized constant . we note that , in the scheme of the second - order perturbation theory , formula ( [ psi0 t ] ) has the same form for different initial configurations such as @xmath79 and @xmath80 . actually , the formula ( [ psi0 t ] ) is valid for all the initial configurations where the atoms pairs are not neighboring to each others . for the other kind of initial configurations , it is straightforward to calculate @xmath29 by formula ( [ pt ] ) and ( [ pb ] ) . in order to check the validity of the perturbation theory , we plot the @xmath29 obtained by the perturbation theory for the two - site problem in fig . [ figure1 ] and make a rough comparison between the perturbation result and the exact result given in eq . ( [ 2exact ] ) . specially , for the simplest two - site case with @xmath81 or @xmath82 , we have @xmath83,\end{aligned}\ ] ] where the normalization constant @xmath84 is given by @xmath85.\end{aligned}\ ] ] as is shown in fig . [ figure1 ] , the larger the @xmath0 , the better the exact and the approximate solution fit . it is natural to use the formula ( [ pt ] ) to deal with the many - particle system . from ( [ psi0 t ] ) , we see that the perturbation theory should work well as long as @xmath86 . in order to compare with the exact result by numerical exact diagonalization method in next section , we apply ( [ p0t2 ] ) to calculate the time evolution of @xmath29 for a finite - size system with an initial state latexmath:[$\left\|\psi_0\right\rangle=\left can see that at large @xmath0 limit , the perturbation theory works quite well . our results show that @xmath88 decays slower for a larger @xmath0 . for a larger system with @xmath89 and the initial state given by @xmath90 , the time evolution of @xmath29 shows similar behaviors as that of the 2- and 6-particle systems . as shown in fig . [ figure3 ] , @xmath29 decays slowly accompanying with a high - frequency but very narrow oscillation . these results show that , when @xmath0 is large enough , @xmath88 mainly stays in the initial state within a large time scale . the larger the u is , the more obvious the case . for the 50-particle system with initial state @xmath90 versus @xmath15 in units of @xmath9 with the corresponding on - site interaction @xmath91.,width=340 ] the exact diagonalization ( ed ) method is very powerful from the viewpoint that it can give out all information of the system exactly . to solve the matrix eigenvalue problem of the hamiltonian , it is convenient to work in a basis of fock states @xmath92 with @xmath93 and the index @xmath94 labels the different compositions of occupation numbers for a fixed total particle number @xmath95 . through the exact diagonalization algorithm we compute exactly all the eigenvalues @xmath96 and the corresponding eigenstates @xmath97 of the hamiltonian ( [ bhh ] ) . then we can calculate the dynamic evolution of @xmath29 by using eq . ( [ p0 t ] ) . to give a concrete example and compare with the result obtained by the perturbation theory in sec . iii , we first consider an initial state @xmath98 . our result is shown in fig . [ figure2 ] . we can see that at large @xmath0 limit , the perturbation theory works quite well . we note that what was measured in the experiment @xcite is the number of remaining pairs after a variable hold time . to understand the experimental result qualitatively , we calculate the time - dependent normalized number of the atom pairs which can be represented as @xmath99 where @xmath100 is given by @xmath101 @xmath102 denotes the number of atom pairs in @xmath103 and @xmath104 is the total number of initial atom pairs . using the formula ( [ npt ] ) , we can calculate the normalized atom pair number for different on - site interaction @xmath0 and different initial state @xmath69 . , @xmath105 and @xmath106 . the time is in units of @xmath9 and the initial state is chosen as @xmath107 . , width=340 ] for simplicity , let us consider a one - dimensional ( 1d ) optical lattice with @xmath108 and @xmath109 , i.e. three boson atom pairs loaded into an optical lattice of nine sites . the time dependent normalized pair numbers for various interaction strengths are plotted in fig . [ figure4 ] with the initial state given by @xmath110 . evidently , the atom pairs in the case of @xmath111 stay longer than that in the case of @xmath112 . it is clear that the large on - site interaction @xmath0 effectively suppresses the dynamic dissociation of the initial pairs , which is in qualitative agreement with the experiment reported in ref . @xcite . and different initial states @xmath69 . from top to bottom the on - site interaction @xmath113= 1.0 , 10.0 , and 100.0 , respectively . in each graph , there are seven lines corresponding to seven different initial states . the initial states are denoted by different lines shown in the table at the bottom . the time is in units of @xmath9 . , width=340 ] next we study how the dynamics of the repulsively bound pairs is affected by the choice of the initial state . for the case with three bound pairs loaded into an optical lattice with nine sites , there are @xmath114 different configurations . taken the translational symmetry of the system into account , there are in fact only seven different configurations . for different initial state @xmath69 , we calculate the normalized atom pair number and plot them in fig . [ figure5 ] . according to fig . [ figure5 ] , although for different initial state @xmath115 the corresponding dynamics of the system is similar to each other , there are still some minor differences depending on whether the initial pairs are close together or not . for initial states @xmath116 , @xmath117 , and @xmath118 in which each pair is apart from the others , the time evolutions are almost the same . this is consistent with our result by perturbation theory . according to the second - order perturbation theory , the time evolution of @xmath29 for these initial states are same . similarly , the time evolution for initial states @xmath119 , @xmath120 , and @xmath121 are almost the same for different on - site interaction @xmath0 as is shown in fig . [ figure5 ] . we notice that the initial states with bound pairs repelling each other are dynamically stabler than the states with all atom pairs initially occupying neighboring sites . we also investigate the case in which there are three boson atom pairs loaded into a two - dimensional @xmath122 optical lattice . the results are similar to the one - dimensional case as shown in fig . 4 and fig . 5 . as @xmath123 . the graph on the top corresponds to @xmath113=1.0 and the graph on the bottom to @xmath113=10.0 . the time is in units of @xmath9 . , title="fig:",width=302 ] as @xmath123 . the graph on the top corresponds to @xmath113=1.0 and the graph on the bottom to @xmath113=10.0 . the time is in units of @xmath9 . , title="fig:",width=302 ] from above discussions , we could understand the stable atom - pair state due to the strongly repulsive interaction as a dynamically stable state . the physical origin is essentially that the energy conservation does not allow the particle to tunnel into the neighboring sites to save the repulsive interaction unless the obtained tunneling energy is comparable to the loss of repulsive energy @xcite . from theoretical point of view , a similar dynamically stable state is expected to appear for the case in which the initial state composed of three or more atoms . in order to clarify this point , we investigate an example case in which six atoms were initially loaded into only one site of a 1d optical lattice with nine sites . the initial state is chosen as @xmath124 . given the initial state , we then determine the dynamic evolution of the initial state on the whole lattice , _ i.e. _ , the spatial distribution of the normalized atom number on each site at different time point for different on - site interaction . the time - dependent normalized atom number on the @xmath125-th site can be calculated by @xmath126 where @xmath127 . the spatial distributions provide us useful information for the expansion dynamics of a condensate loaded in the optical lattice . at an incremental sequence of time points , we calculate the normalized atom number at each site . the results are shown in fig . [ figure6 ] . for a small @xmath0 @xmath128 , the atoms initially trapped in a site tend to tunnel to the neighboring sites and thus lead to the decrease of the normalized atom number in the initial site . however , when the on - site @xmath0 is large enough , the initially trapped atoms are dynamically frozen and the expansion to the neighboring sites is almost completely suppressed . in this paper , we theoretically investigate the dynamical stability of atom pairs induced by the strongly repulsive interaction in an optical lattice both analytically and numerically . firstly , in the large @xmath0 limit , we calculate the time evolution of an initial system composed of atom pairs in a scheme of the perturbation theory . the analytical results show that the initial state of atom pairs is dynamically stabler for larger repulsive interaction . then by the exact diagonaliztion method , we numerically study the stability of atom pairs in a finite - size system with few particles by calculating the time evolution of the normalized atom pair number . our results show that the initial state of atom pairs are dynamically stable and the dissociation of atom pairs is greatly suppressed when the strength of the on - site interaction is large . we also compare our numerical result with the result based on the perturbation theory and find out that they fit quite well in the large @xmath0 limit . our results imply that the experimentally observed repulsively bound pairs can be understood as a dynamically stable state . t. kinoshita , t. wenger and d. s. weiss , science * 305 * , 1125 ( 2004 ) ; b. paredes , a. widera , v. murg , o. mandel , s. flling , i. cirac , g. v. shlyapnikov , t. w. hnsch , and i. bloch , nature ( london ) * 429 * , 277 ( 2004 ) . m. d. girardeau , j. math . * 1 * , 516 ( 1960 ) ; l. tonks , phys . rev . * 50 * , 955 ( 1936 ) ; v. dunjko , v. lorent and m. olshanii , phys . lett . * 86 * , 5413 ( 2001 ) ; s. chen and r. egger , phys . a. * 68 * , 063605 ( 2003 ) .	we investigate the quantum dynamics of repulsively bound atom pairs in an optical lattice described by the periodic bose - hubbard model both analytically and numerically . in the strongly repulsive limit , we analytically study the dynamical problem by the perturbation method with the hopping terms treated as a perturbation . for a finite - size system , we numerically solve the dynamic problem in the whole regime of interaction by the exact diagonalization method . our results show that the initially prepared atom pairs are dynamically stable and the dissociation of atom pairs is greatly suppressed when the strength of the on - site interaction is much greater than the tunneling amplitude , i.e. , the strongly repulsive interaction induces a self - localization phenomenon of the atom pairs .
the process of star formation is currently not well understood . however , the main protostellar collapse phases of low - mass stars ( @xmath10.52m@xmath2 ) have been identified observationally and labelled as class 0 ( andr , ward - thompson & barsony 1993 ) , and class i ( lada & wilking 1984 ; lada 1987 ) protostars . these are believed to represent the phases during which the circumstellar envelope accretes onto the central protostar and disk ( e.g. andr 1994 ; ward - thompson 1996 ) . the final pre - main - sequence stages of classes ii & iii ( lada & wilking 1984 ; lada 1987 ) correspond to the classical t tauri ( ctt ) and weak - line t tauri ( wtt ) stages respectively ( andr & montmerle 1994 ) . these protostellar stages are beginning to be understood at least in outline ( for a review , see : andr , ward - thompson & barsony 2000 ) . protostellar infall has been reported in class 0 sources by a number of authors ( e.g. zhou et al . 1993 ; ward - thompson et al . however , the manner of the collapse remains a matter for debate . the ideas of static initial conditions for collapse ( e.g. shu 1977 ) have been disputed by many authors ( e.g. foster & chevalier 1993 ; whitworth et al . there is now a growing body of evidence which suggests that the collapse occurs from non - static initial conditions , and at a non - constant accretion rate which decreases with time ( e.g. kenyon & hartmann 1995 ; henriksen , andr & bontemps 1997 ; safier , mckee & stahler 1997 ; whitworth & ward - thompson 2001 ) . infall has even been detected in some starless cores , such as l1544 ( tafalla et al . 1999 ) . the densities and temperatures in the gas envelopes surrounding the youngest protostars are favourable for exciting a number of rotational molecular transitions , observable in the submillimetre waveband . the line profiles of these transitions contain information about both the physical state and dynamics of the envelope gas , and may potentially be used to test theoretical models of star formation , as many workers have shown ( e.g. bernes 1979 ; rybicki & hummer 1991 ; choi et al . 1995 ; juvela 1997 ; park & hong 1998 ; hogerheijde & van der tak 2000 ) . in this paper observations are presented of the protostellar candidates ngc1333iras2 and serpens smm4 , in transitions of hco@xmath3 , h@xmath4co@xmath3 , cs , co , @xmath4co and c@xmath5o . the hco@xmath3 and cs transitions preferentially trace high density gas , whereas co traces a much wider range of gas densities . the paper is laid out as follows : section 2 introduces the @xmath0-iteration method of numerical radiative transfer ; section 3 describes our approach to the modelling ; section 4 explores the sensitivity of the model to the various free parameters ; section 5 describes our observations and data reduction ; sections 6 & 7 present the results of our observations for ngc1333iras2 and serpens smm4 respectively ; section 8 compares the observations with the model predictions and finds the best fits to the data ; section 9 presents a brief summary of our main findings . the non - expert reader may prefer to read the second half of the paper first , starting from section 5 . consider a cloud of gas with a specified distribution of density , kinetic temperature and composition , which may have internal turbulent and systematic motions . let any radiation sources not forming part of the cloud itself also be specified . for each molecular species , there exists a steady state solution for the distribution of rotational energy level populations as a function of position in the cloud . in the modelling of this paper we numerically calculate this distribution for the idealised case of a spherically symmetric model cloud , and predict the observed line profiles . this is a complex problem , due to the fact that well separated parts of the cloud can interact radiatively with each other . the @xmath0-iteration method , described below , is conceptually one of the simplest techniques for solving this kind of problem . ( 80,90 ) the method is begun by choosing an initial radiation field in a more or less arbitrary manner . from this a ` false ' run of level populations may be calculated , by assuming the validity of the steady state rate equations . radiative transitions between these level populations will generally produce a radiation field which departs from the one originally assumed . if this radiation field is used to calculate a new set of level populations , and the procedure is repeated a sufficiently large number of times , the radiation field and level populations should eventually converge on a mutually consistent solution . @xmath0-iteration is a kind of diffusion process , where imbalances in the radiation field are smoothed out over a length - scale corresponding to approximately one optical depth at each iteration step . a number of radiative transfer models have been published ( e.g. rawlings et al . 1992 ; zhou 1992 ; walker et al . however , the modelling in this paper was carried out using a modified version of the stenholm @xmath0-iteration code developed by stenholm and subsequently expanded by little and co - workers at the university of kent ( stenholm 1977 ; matthews 1986 ; heaton et al . the code uses the above method to solve the spectral line radiative transfer problem for the rotational transitions of linear molecules in a spherically symmetric model cloud . radial profiles of systematic velocity , temperature , density , tracer molecule abundance and micro - turbulent velocity dispersion may be specified . the model cloud is discretised using a number of spherical shells , and the level populations in each shell are determined in the iteration process from the calculated mean radiation intensity in the co - moving shell frame . once the level populations have converged , a calculation is made of a simulated spectral line observation on the cloud . this part of the code carries out straightforward numerical integrations of the equation of transfer along parallel lines of sight through the cloud to find the emergent intensities for a grid of impact parameters , which are then weighted according to a two - dimensional gaussian beam profile function . it is important to find a reliable test for the convergence of the model . one possible condition is that the maximum fractional change @xmath6 , of the level population @xmath7 between iterations @xmath8 and @xmath9 , is less than some specified value @xmath10 : @xmath11 for all @xmath12 and j , where the subscripts specify the iteration numbers . in many cases , the value of @xmath13 is a good estimate of the absolute fractional error in the corresponding level population , in which case convergence is ensured by setting a suitably low value of @xmath10 ( a value of @xmath14 or less is usually found to be sufficient ) . however , when the rate of convergence is particularly slow , this test may give a misleading result , since the fractional change between two successive iterations may then significantly underestimate the total change produced over a large number of subsequent iterations . the rate of decrease of @xmath15 should somehow be taken into account . dickel & auer ( 1994 ) use the following procedure to estimate the total fractional error @xmath16 of the level population @xmath7 after the @xmath9th iteration : @xmath17 - 1}\right\|.\ ] ] the convergence criterion is then @xmath18 for all @xmath12 and j , for some specified @xmath10 much less than 1 . note that if @xmath19 then @xmath16 is much larger than @xmath20 i.e. the total error is much larger than the fractional change between successive iterations . the rate of convergence is fastest when the optical depth of the cloud is small , or when the hydrogen number density is much greater than the critical thermalisation density . when the optical depth is small , radiation can travel across the whole cloud in one iteration , allowing the level populations in different parts of the cloud to adjust to each other quickly . when the density is close to , or above , the critical density for the principal thermally accessible transitions of the molecule , the level populations are determined mainly by molecular collisions , and are insensitive to changes in the radiation field . for optically thick regions where the density is below the critical density , the convergence rate is generally much slower . the low density results in low collision rates , and the level populations therefore depend more sensitively on the radiation field , while the high optical depth means that the radiation field will take many iterations to adjust to the conditions in distant parts of the cloud . the code converges most rapidly near to lte and when optically thin . we have made several alterations to the code since it was made available to us . these have included changes to the internal structure of the code by , for example , improving the shell subdivision algorithm and incorporating rotation into the line profile calculation . however , the former did not affect the conclusions , it merely gave the code more resolution . some of the more significant changes made are now discussed in detail . an important modification made to the radiative transfer calculation was the removal of the approximation of core saturation used in the original code , known as the rybicki approximation ( rybicki 1972 ; stenholm 1977 ; rybicki 1984 ) , which may be summarised as follows : suppose that at some point in the cloud the following relation for the optical depth @xmath21 at frequency @xmath22 is satisfied for all directions * r*@xmath23 and solid angles @xmath24 : @xmath25 where @xmath26 is the distance to the boundary of the model and @xmath27 is some prescribed constant . the intensity i in all directions at that point is defined in the rybicki approximation to be equal to the local source function s , i.e : @xmath28 this equation is only applied for the particular values of @xmath22 for which @xmath29 . for increasing values of @xmath27 , the volume of the cloud to which the approximation applies becomes smaller , and disappears altogether when @xmath27 increases beyond the maximum optical depth of the cloud . for the problems we consider , the source function can not be assumed to be uniform over distances corresponding to a few optical depths , and the rybicki approximation can not be used . we therefore do not use the rybicki approximation in the modelling in this paper . ( 80,180 ) we have incorporated rotation into the code , in the form of ` solid body ' rotation i.e. the angular velocity of all points in the cloud about some fixed axis is a constant , but systematic radial motions in the rotating frame are still allowed . the radiative transfer effects of solid body rotation were fully accounted for by changes to the section of the code which calculates the emergent line profile from the cloud , after the @xmath0-iteration solution has been found . no change needs to be made to the @xmath0-iteration code itself , since the solution for the level populations in the cloud does not change when solid body rotation is introduced , as is now explained . the radiation field in a frame co - moving with the gas at a given point in the model can only be affected by a superimposed velocity field if the latter introduces changes in the relative line of sight velocities of other parts of the cloud ( ignoring relativistic effects of order @xmath30 or less ) . solid body rotation does not introduce changes in relative line of sight velocities from one part of the cloud to another . for example , consider a non - rotating cloud which has systematic radial velocities @xmath31 and @xmath32 at the positions @xmath33 and @xmath34 respectively . writing @xmath35 for the distance separating the two points , the component of relative velocity , @xmath36 , along the line of sight between @xmath33 and @xmath34 is given by : @xmath37/r_{12}. \label{vlos1}\ ] ] now assume the cloud is made to rotate in solid - body fashion with angular velocity @xmath38 . let the new systematic velocities at positions @xmath33 and @xmath34 be @xmath39 and @xmath40 , where : @xmath41 @xmath42 substituting into equation 5 , the terms in @xmath38 disappear , and the final expression is unchanged . this implies that the solid body rotation has no effect on the relative line of sight velocities between any two points in the cloud , and hence the radiation field in the co - moving frame at each point in the cloud is not affected by solid body rotation . this result only applies to the radiation emitted by the cloud itself . radiation entering the model from outside will be doppler shifted in the gas frame by the rotation , but the effect of this is completely negligible for realistic external radiation fields . however , solid body rotation does make a difference to the prediction for the observed line profile , since the rotation does introduce changes in the line of sight velocities of different parts of the cloud relative to an observer outside the cloud . if @xmath43 is the unit vector parallel to the line of sight from the observer to the cloud , and @xmath44 is the line of sight velocity ( relative to the observer ) of the gas at the position vector @xmath45 ( relative to the centre of the cloud ) , then : @xmath46 where @xmath47 is the component along the line of sight of the radial component of the gas velocity at @xmath45 , and @xmath48 and @xmath49 are , as before , spatial co - ordinates in the plane perpendicular to the line of sight at the distance of the cloud . @xmath50 and @xmath51 are the components of the angular velocity vector in this plane . the component of angular velocity parallel to the line of sight , @xmath52 , does not appear in equation 8 and has no effect on the observed spectrum from the cloud . analytical solutions of the equation of transfer are in general unobtainable for all but the simplest of problems . however , in the limit that the level populations are completely thermalised by collisions i.e. in local thermodynamic equilibrium ( lte ) the spectral line calculations from the numerical model may be compared directly with analytical predictions . in lte , the excitation temperature for each transition approaches the kinetic temperature @xmath53 of the gas , and the source function for the transition j@xmath54 becomes @xmath55 $ ] . if @xmath56 is constant over the cloud , then the solution to the equation of transfer is : @xmath57 where @xmath58 is the optical depth of the cloud at the frequency corresponding to the velocity bin @xmath59 . for simplicity , we consider the intensity emerging from a uniform density isothermal cloud layer of thickness @xmath60 , with no systematic motions , and with a constant velocity dispersion ( @xmath61 ) and total tracer molecule number density ( @xmath62 ) . the optical depth through the cloud layer may then be written as : @xmath63 @xmath64\phi({\rm v}_z)\ ] ] @xmath65\phi({\rm v}_z),\ ] ] where @xmath66 and @xmath67 are the number densities of tracer molecules in the upper and lower level of the transition , and other symbols take their usual meanings . in the last equation we have used the fact that in lte : @xmath68 and also : @xmath69 the explicit equation for @xmath70 is given by : @xmath71 and the number density @xmath66 of tracer molecules in the level j may be evaluated by : @xmath72 the sum in the denominator is the partition function , and usually converges rapidly after about ten or so terms ( depending on @xmath56 ) , and can be easily evaluated to the required accuracy . substituting these expressions into equation 9 , with @xmath73 , gives an expression for the predicted intensity as a function of @xmath59 . to compare these predictions with the stenholm code , we used a static model cloud with a uniform kinetic temperature , hydrogen number density , velocity dispersion , and abundance . the hydrogen number density was set several orders of magnitude above the critical density to enforce lte . figure 1 shows comparisons of the analytical and numerical predictions for the cs(j@xmath74 ) line , from lte models with progressively increasing cs abundance , covering a wide range of cloud optical depths . the excellent agreement between the predicted spectra in all parts of the line , and across the large range of optical depths , is a good indication that the integration of the equation of transfer along the line of sight is being carried out correctly . the lte comparison shown above , while instructive , does not test the ability of the code to carry out non - lte radiative transfer calculations reliably . we have not found any exact analytical solutions for non - lte radiative transfer problems in spherical geometry , to compare with the code in a similar manner as shown above . instead , we attempt to reproduce the line profiles obtained in a previous non - lte radiative transfer study ( krgel & chini 1994 , hereafter kc ) , which used a somewhat different numerical method . figure 2 shows a comparison of profiles taken from kc with the corresponding profiles calculated by our stenholm code . the figure caption gives details of the model cloud used , which has strong velocity and density gradients towards the centre , and is far from lte . both calculations use the cs collision rates of green & chapman ( 1978 ) . overall , the agreement between the two sets of line profiles is very good . the intensities of the line profiles predicted by stenholm are up to 10% weaker than the kc profiles , and the disagreement appears to be greatest for the smallest beam sizes . we checked our calculations , and found that this was not the result of inadequate sampling of the lines of sight used in the beam convolution , by re - calculating the profiles for a range of different beam sampling densities . the small disagreement in the line profile strengths is not serious , and may be caused by differences in the spatial , angular or frequency discretisation schemes used in the models , or small differences in the precise values of some of the parameters used . particularly encouraging is the very good agreement in the shapes of the corresponding line profiles . in this section we discuss our approach to the modelling of infall line profiles , and investigate the qualitative dependence of the predicted line profiles on a number of different model parameters . the approach to the modelling is necessarily a compromise between the desire to represent realistically the object being modelled , and the practical need to limit the parameter space and running time of the numerical simulations . the observations are modelled using the spherically symmetric micro - turbulent @xmath0-iteration radiative transfer code described above . even with the considerable simplification of spherical symmetry the number of model parameters is formidable . radial profiles of density , kinetic temperature , molecule abundance , systematic velocity , and micro - turbulent velocity width must be specified . additional parameters include the choice of inner and outer boundaries of the model , the incident radiation fields at the inner and outer boundaries , and the distance to the object . some assumptions must be made about the forms taken by these profiles , so that they may be described by specifying only a few parameters . some of the profiles are assumed to have a power law dependence with radius , which requires only the power law exponent and a normalisation factor to be specified . wherever possible , the choice of parametric forms used to describe the radial profiles is guided by theoretical considerations . we now discuss the profile of each physical parameter in turn . following the onset of fully dynamical collapse , the radial velocity profile should tend towards a freefall profile ( e.g. larson 1969 ; shu 1977 ; hunter 1977 ; zhou 1992 ; foster & chevalier 1993 ) . the pure free - fall velocity profile may be written as : @xmath75 the infall velocity profiles in non - rotating clouds which collapse from different initial conditions should approximately converge , even when the mass accretion rates are very different , as long as the comparison is made at times when the central protostellar masses are equal ( hunter 1977 ) . hence we use the singular isothermal sphere ( sis ) model velocity profiles to parameterise the infall velocity field in the radiative transfer modelling , which allows the velocity field to be described in terms of only two quantities : the effective sound speed @xmath76 and the infall radius @xmath77 . we stress that these are merely used as parameters characterising the velocity field , and their actual physical significance is model dependent . we have found the following empirical fit to the exact velocity profile predicted by the standard sis model , for @xmath78 : @xmath79 this equation fits the sis collapse field to within a few per cent . if the collapsing cloud has some initial rotation , then centrifugal support may become important at small radii , leading to a decrease in the radial velocity close to the centre of collapse . with an initial angular velocity @xmath80 , then the inner radius , @xmath81 , at which centrifugal forces begin to dominate depends on the mass already accreted , @xmath82 , and the radius @xmath83 enclosing this mass ( terebey , shu & cassen 1984 ) through the following relation : @xmath84 as discussed above , the radiative transfer code used to model the data in this paper makes use of the micro - turbulent approximation . this assumes that random ` turbulent ' motions can be treated in the same way as thermal molecular motions , by incorporating them into the local line profile function . comparisons between radiative transfer codes which model macroscopic and microscopic turbulence suggest that the main effect of relaxing the micro - turbulent approximation is to reduce the strength of self - absorption features ( e.g. park & hong 1995 ) . although observations appear to indicate random motions of some kind must be present in protostellar envelopes , very little theoretical work has been done on the role of turbulence and turbulent support in protostellar collapse . lizano & shu ( 1989 ) derived a ` turbulent equation of state ' using the well known ` larson s laws ' ( larson 1981 ) , which relate the observed velocity dispersion ( @xmath85 ) , size ( r ) , and average density ( @xmath86 ) in molecular clouds and cores : @xmath87 these correlations apply to ensembles of clouds and also to observations of individual clouds in different molecular tracers , and are associated with the turbulent motions in the clouds ( e.g. myers 1983 ) . combining the two relations above gives : @xmath88 zhou ( 1992 ) used this relation to infer the variation of turbulent velocity with density in his radiative transfer calculations of infall line profiles . vzquez - semadeni , cant & lizano ( 1998 ) have recently argued that larson s laws may be inappropriate to dynamically collapsing clouds , because they were established from observations of clouds which were in equilibrium configurations . these authors carried out numerical hydrodynamic and magneto - hydrodynamic simulations to test the behaviour of the turbulence during the collapse , and found that , in contrast to equation [ sigrho ] , the turbulent velocity dispersion actually increases with density during the collapse . for purely hydrodynamic and weakly magnetic collapse models the relation between the turbulent velocity dispersion and the density was seen to approach a power law of the form @xmath89 ( i.e. @xmath90 ) . in the case of strongly magnetically inhibited collapse , the power law index was observed to change to @xmath91 ( @xmath92 ) , consistent with the predicted behaviour for the slow compression of alfvn waves ( mckee & zweibel 1995 ) . the behaviour of the turbulence in infalling protostellar envelopes is thus theoretically rather uncertain . turbulence nevertheless plays a very important part in the formation of line profiles , and is one of the principal sources of uncertainty in the infall line profile modelling . in a region where the infall velocity has a free - fall profile there is often a strong tendency for the density profile to adopt an @xmath93 distribution , since if the rate of injection of material into the accretion stream is not changing rapidly with time , the quantity @xmath94 should be approximately independent of radius . the prediction for the density profile deep inside the infalling region may be written as : @xmath95 where , as before , @xmath96 is the infall radius , and @xmath76 is the effective sound speed . the exact model density profile inside the whole infalling region is very well approximated ( within @xmath11 per cent ) by the following empirical formula : @xmath97 . \label{shurho}\ ] ] outside the infall radius , the density profile is given by : @xmath98 there are several competing theoretical models which predict different forms of the density profiles at various stages in the contraction and collapse of a dense cloud core ( e.g. foster & chevalier 1993 ; basu & mouschovias 1994 ; whitworth & ward - thompson 2001 ) . however , once a central protostar has formed there is a tendency in the theoretical models for the forms of the density and velocity profiles to converge towards the sis model forms , in particular towards the centre of the collapse ( e.g. foster & chevalier 1993 ) . we therefore initially search for fits to our observations using density profiles of the above form , although we do not require that the adopted density profile uses the same values for the parameters @xmath99 and @xmath76 as used in the velocity profile . the temperature of the gas in an infalling protostellar envelope is determined by the balance of heating and cooling processes . ceccarelli , hollenbach & tielens ( 1996 ) modelled these processes to derive the temperature profile in protostellar envelopes with central embedded heating sources . they found that in all the cases studied , the gas temperature profile stays within @xmath130% of the dust temperature , usually lying slightly below the dust temperature , @xmath100 . in the following it is assumed that the gas temperature is well coupled to the dust temperature , although this assumption may be significantly in error for hydrogen number densities below @xmath101@xmath102 . if the emitted radiation reaching the optically thin part of the envelope peaks at a wavelength @xmath103 , then the temperature of the dust at a radius @xmath104 will be given by : @xmath105 the value of @xmath103 depends on the details of the dust radiative transfer through the optically thick inner region ( i.e. how the material in this region is distributed ) . kenyon et al . ( 1993 ) used the diffusion approximation in the optically thick region to estimate the value of @xmath103 . for a density profile of the form @xmath106 this gives : ( 80,90 ) @xmath107^{1/(4 + 4\beta)}.\ ] ] substituting this expression into the above equation and setting @xmath108 and @xmath109 ( appropriate for a hydrodynamically collapsing region ) , we find : @xmath110 the temperature profile in the optically thin part of the envelope is therefore only weakly dependent on the density normalisation and the luminosity ( see kenyon et al . 1993 for further discussion ) . for the objects modelled in this paper , the optically thick region is expected to be very much smaller than the beam size of the observations and is comparable to the inner shell radius used in the radiative transfer models . at sufficiently large radii , the gas temperature will approach the ambient temperature of the larger scale cloud . we therefore adopt the temperature profile appropriate to the optically thin limit ( @xmath111 where @xmath112 ) in the inner region , and switch to a flat profile at the radius where the power law inner profile falls below the ambient cloud temperature . ( 80,40 ) ( 80,40 ) the profiles of cs and relative abundance in an infalling protostellar envelope are governed by time dependent chemistry and the transfer of molecules between the gas phase and the ice mantles surrounding dust grains . this problem has been studied by rawlings et al . ( 1992 ) and willacy , rawlings & williams ( 1994 ) in the case of isothermal collapsing clouds . the objects studied in this paper contain internal stellar heating sources , and are unlikely to be well described by an isothermal model . the chemistry of internally heated infalling protostellar envelopes has been studied by ceccarelli , hollenbach & tielens ( 1996 ) . we refer the reader to all of these references for further discussion of the way in which chemistry can affect observed abundance profiles . given these uncertainties and the requirement to limit the parameter space , we used flat abundance profiles in our modelling . the fractional abundances of cs and hco@xmath113 measured in dense molecular gas cores occupy similar ranges , typically between @xmath114 and @xmath115 ( e.g. irvine et al 1987 ; mcmullin , mundy & blake 1994 ; blake et al . 1995 ) . the measured values of [ @xmath116c/@xmath4c ] at the galacto - centric radius of the sun ( which applies to all of our observed objects ) lie in the range 50100 ( wilson & rood 1994 ) . we therefore restrict the abundances used in the modelling to these ranges of values . ( 80,40 ) ( 80,40 ) it is useful to define a specific infalling envelope model , with plausible physical parameters , which may be used to investigate some general properties of infall spectral line profiles . we use the above density and velocity profiles at a stage when a mass m@xmath117/2 has already accreted onto the central protostar , and a further @xmath118m@xmath117/2 of envelope gas is infalling towards it . if we choose an effective sound speed of @xmath119 , then an infall radius of @xmath1183700au is implied . we therefore use the following model relations for the systemic velocity and hydrogen number density profiles : @xmath120 @xmath121 and outside the infall radius the systematic velocity is set to zero . the density profile is given by : @xmath122 we truncate the density profile at an outer radius of 10000au , which encloses a total mass of 2.75m@xmath117 . this corresponds to the typical radius at which dense cores merge with the more diffuse gas in the ambient molecular cloud . the adopted micro - turbulent velocity dispersion is @xmath123kms@xmath124 , which is assumed to be spatially constant . although there is little empirical or physical justification for a flat turbulent velocity profile ( see the above discussion ) , we make this assumption as a zeroth order approximation , in the absence of any clear consensus as to the correct description of turbulence in infalling envelopes . as discussed above , the kinetic temperature profile ( assumed to be equal to the dust temperature profile ) in the optically thin part of the envelope is expected to be fairly insensitive to the luminosity of the central source . we choose a canonical temperature profile of : @xmath125 the normalisation of the temperature profile was fixed using the dust temperature profile calculated by kenyon et al . ( 1993 ) , for a source luminosity of @xmath126 . the varying kinetic temperature in the envelope is not strictly consistent with the assumption of constant @xmath76 implied by the sis model , where @xmath127 . however , our primary concern is not to investigate the sis model solution in detail , but rather to investigate the essential features of line formation in infalling envelopes , whether the infall is described by the sis model or not . we set both the cs and relative abundances to @xmath128 , and assume a [ h@xmath116co@xmath3]/[h@xmath4co@xmath113 ] isotopic ratio of 60 . the set of molecular constants used in the radiative transfer modelling are listed in table 1 , where @xmath131 denotes the permanent molecular dipole moment and other symbols take their usual meanings . the data are taken from the jpl spectral line catalogue ( poynter & pickett 1985 ) . [ molecules ] .assumed physical constants of the molecules modelled . [ cols="^,^,^,^,^ " , ] as with ngc1333iras2 , the infall velocities in the model fit are much greater than would be predicted by the sis model given the density profile . from the infall velocity profile we estimate the mass of the central protostellar system to be @xmath132m@xmath117 . the temperature profile is normalised at a considerably higher level than in the iras2 fit , which would suggest that smm4 is considerably more luminous . ( 60,190 ) a high accretion luminosity for this object is also suggested by the large central protostellar mass and mass accretion rate ( @xmath133m@xmath117yr@xmath124 ) . hurt & barsony ( 1996 ) estimated the luminosity of smm4 as 9l@xmath117 , after deconvolving the hires iras images of the serpens region to obtain the far - infrared fluxes . jennings et al . ( 1987 ) estimated the luminosity of iras2 as 17l@xmath117 , also using iras fluxes , so there are clearly large uncertainties on these estimates . the parameters used in our model fits are inconsistent with these luminosities , and for our model to be correct , then either the smm4 luminosity calculated from the iras data , or our assumed protostellar radius , would have to be a significant underestimate . the predicted linewidths are considerably broader than the observations , as we found with ngc1333iras2 , although the peak intensities are in very good agreement . again , we suggest this may be indicating deviations of the infall velocity profile below the @xmath134 free - fall profile in the centre of the cloud . we found above that the smm4 envelope shows strong evidence for rotation ( @xmath135kms@xmath124pc@xmath124 ) , and suggested that the flattening seen in the spatial distribution of integrated hco@xmath136 emission may be caused by centrifugal support of the outer envelope . thus centrifugal braking may retard the infall velocity near the centre of the cloud . the calculated cs@xmath137 and cs@xmath138 line profiles match the observed line intensities reasonably well , but the peak velocities of the calculated profiles are less red - shifted than the observations . the enhanced blue - shifted wings seen in both of the observed cs spectra strongly suggest that outflow contributes to the emission in these lines . as discussed above , cs is believed to be enhanced relative to in bipolar outflows ( blake et al . 1995 ) . figure 33 shows the fits to the off - centre hco@xmath136 line profiles towards smm4 from our best - fit model . we included a solid body rotation of 28kms@xmath124pc@xmath124 in the model ( assuming rotation axis in plane of sky ) , consistent with the above centroid velocity gradient across smm4 . as the hco@xmath136 spectra at the ( 0,0 ) and ( 0,+7 ) positions have approximately equal intensities , we assumed that the centre of the cloud lies at the offset ( 0,+3.5 ) . although the model profiles somewhat underestimate the observed lines , we are encouraged by the overall agreement of the trends in the observed lineshapes across the envelope . it is likely that a better fit to the line intensities requires a shallower density profile than predicted by the sis model . in contrast to iras2 we find little evidence for outflow emission in the off - centre line profiles , and essentially all of the high velocity emission can be accounted for by the infall . we consider these results to provide convincing evidence that both infall and rotation are taking place in the the serpens smm4 envelope . the magnitude of the rotational velocity is large enough for centrifugal effects to have a significant influence on the gas dynamics . in this paper we have discussed the physical constraints which can be used to limit the number and range of parameters needed to define the spherically symmetric models in the stenholm radiative transfer code . we have studied the dependence of the predicted line profiles on a number of model parameters . increasing the infall velocities in the envelope tends to produce stronger asymmetry between the blue and red - shifted peaks of double - peaked line profiles , without strongly affecting the separation between the peaks . _ the value of the turbulent velocity dispersion is the most significant factor in determining the velocity separation between the peaks_. we carried out a preliminary investigation into the effect of adopting different power law dependences of the turbulent velocity width on the gas density and found the predicted line profiles to be extremely sensitive to the value of the power law exponent . we have studied the effect on the predicted line profiles of super - imposing a solid - body rotation onto the infall velocity field . the effect of the solid - body rotation on the on - source double - peaked line profiles was to ` smooth out ' the double - peaked structure . the centroid velocities and the line shapes of the off - centre profiles were found to be significantly altered by the solid - body rotation . on the red - shifted side of the rotational velocity gradient , a reversal of the asymmetry in the line profiles was seen . we measured the centroid velocity gradient predicted by our model for a solid - body rotating envelope , which was found to slightly underestimate the actual rotational velocity gradient used in the model . we selected the objects ngc1333iras2 and serpens smm4 for detailed modelling using the stenholm radiative transfer code , confining our search to the sis model forms for the density and velocity profiles in the modelling , although we did not require these profiles to use the same sis model parameters in the models . our best model fits to both the ngc1333iras2 and serpens smm4 data used velocity profiles which were more strongly developed than the sis model would predict , given the fit to the density profile . this is consistent with theoretical predictions that at the instant a central protostar is formed , the gas in the envelope has already accelerated to substantial velocities ( e.g. foster & chevalier 1993 ; jch ) , which contrasts with the static initial state assumed by the sis model , and is one of the principal distinguishing features between the sis model and other collapse models . our fits to the line profile observations of both objects slightly over - estimated the profile widths . we could produce better fits to these lines by lowering the infall velocity in the centre of the cloud , although this affected the fits to the main line profile slightly . the infall velocity may therefore be retarded from the @xmath134 free - fall relation towards the centre of the cloud , possibly by centrifugal braking . in the case of iras2 , the discrepancies between the model predictions and the observations for the off - source profiles could not possibly be accommodated within a plausible model of infall and rotation , and we attribute them to the outflow . the model fits to the off - centre hco@xmath136 line profiles of serpens smm4 included the solid - body rotation inferred from the hco@xmath136 centroid velocity gradient in the calculation . the predicted and observed profiles showed good agreement in the overall line shape although the model slightly underestimated the line intensities . we thus conclude that serpens smm4 is an infalling and rotating protostar . the authors would like to thank the staff of the joint astronomy centre , hawaii ( jach ) , and in particular the jcmt telescope support staff , for assistance during the taking of the observations presented in this paper . we would also like to thank les little for giving us a copy of the kent stenholm code . the jach is operated jointly by the uk pparc , canadian nsf and netherlands nwo . adams f. c. , shu f. h. , 1985 , apj , 296 , 655 andr p. , 1994 , in : montmerle th . , lada c. j. , mirabel i. f. , trn thanh vn j. , eds . , ` the cold universe ' , 179 , editions frontieres , paris andr p. , montmerle t. , 1994 , apj , 420 , 837 andr p. , ward - thompson d. , barsony m. , 1993 , 406 , 122 andr p. , ward - thompson d. , barsony m. , 2000 , in : mannings v. , boss a. p. , russell s. s. , eds . , ` protostars & planets iv ' , 59 , university of arizona press , tucson aspin c. , sandell g. , russell a. p. g. , 1994 , a&a , 106 , 165 bally j. , devine d. , reipurth b. , 1996 , apj , 473 , l49 barsony m. , 1997 , communications of the konkoly observatory , 100 , 189 basu s. , mouschovias t. , 1994 , apj , 265 , 824 bernes c. , 1979 , a&a , 73 , 67 bevington p. r. , robinson d. k. , 1992 , ` data reduction and error analysis for the physical sciences ' , mcgraw - hill , new york blake g. a. , sandell g. , van dishoeck e. f. , groesbeck t. d. , mundy l. g. , aspin c. , 1995 , apj , 441 , 689 butner h. m. , evans n. j. ii , harvey p. m. , mundy l. g. , natta a. , randich m. s. , 1990 , apj , 364 , 164 casali m. m. , eiroa c. , duncan w. d. , 1993 , a&a , 275 , 195 ( ced ) ceccarelli c. , hollenbach d. j. , tielens a. g. g. m. , 1996 , apj , 471 , 400 cernis k. , 1990 , ap&ss , 166 , 315 chavarria k. c. , de lara e. , finkenzeller u. , mendoza e. e. , ocegueda j. , 1988 , a&a , 197 , 151 choi m. , evans n. j. , gregersen e. m. , wang y. , 1995 , apj , 448 , 742 ciolek g. e. , mouschovias t. ch . , 1994 , apj , 425 , 142 cunningham c. t. , hayward r. h. , wade j. d. , davies s. r. , matheson d. n. , 1992 , int . j. infrared millimetre waves , 13 , 1827 curiel s. , rodriguez l. f. , moran j. m. , canto j. , 1993 , apj , 415 , 191 curiel s. , rodriguez l. f. , gomez j. f. , torrelles j. m. , ho p. t. p. , eiroa c. , 1996 , apj , 456 , 677 davies s. r. , cunningham c. t. , little l. t. , matheson d. n. , 1992 , int . j. infrared millimetre waves , 13 , 647 davis c. j. , chrysostomou a. , matthews h. e. , jenness t. , ray t. p. , 2000 , apj , 530 , 115 de lara e. , chavarria k. c. , lopez - molina g. , 1991 , a&a , 243 , 139 dent w. r. f. , cunningham c. , hayward r. , davies s. r. , wade d. , avery l. w. , mayer c. j. , masuda n. t. , 1993 , mnras , 262 , l13 dickel h. r. , auer l. h. , 1994 , apj , 437 , 222 eiroa c. , 1991 , in : reipurth b. , ed . , ` low mass star formation in southern molecular clouds ' , eso scientific reports , 11 , 197 , eso , garching eiroa c. , casali m. m. , 1989 , a&a , 223 , l17 eiroa c. , casali m. m. , 1992 , a&a , 262 , 468 foster p. n. , chevalier r. a. , 1993 , apj , 416 , 303 gmez de castro a. i. , eiroa c. , lenzen r. , 1988 , a&a , 201 , 299 green s. , chapman s. , 1978 , apjs , 37 , 169 gregersen e. m. , evans n. j. ii , zhou s. , choi m. , 1997 , apj , 484 , 256 harvey p. m. , wilking b. a. , joy m. , 1984 , apj , 278 , 156 heaton b. d. , little l. t. , yamashita t. , davies s. r. , cunningham c. t. , monteiro t. s. , 1993 , a&a , 278 , 238 henriksen r. , andr p. , bontemps s. , 1997 , a&a , 323 549 hodapp k .- w . , ladd e. f. , 1995 , apj , 453 , 715 hogerheijde m. r. , van dishoeck e. f. , blake g. a. , van langevelde h. j. , 1998 , apj , 502 , 315 hogerheijde m. r. , van der tak f. f. s. , 2000 , a&a , 362 , 697 hunter c. , 1977 , apj , 218 , 834 hurt r. l. , barsony m. , 1996 , apj , 460 , l45 hurt r. l. , barsony m. , wootten a. , 1996 , apj , 456 , 686 irvine w. m. , goldsmith p. f. , hjalmarson a. , 1987 , in : hollenbach d. j. , thronson h. a. , eds . , ` interstellar processes ' , 561 , dordrecht , reidel jayawardhana r. , hartmann l. , calvet n. , 2001 , apj , in press jhc jennings r. e. , cameron d. h. m. , cudlip w. , hirst c. j. , 1987 , mnras , 226 , 461 juvela m. , 1997 , a&a , 322 , 943 kenyon s. j. , calvet n. , hartmann l. , 1993 , apj , 414 , 676 kenyon s. j. , hartmann l. w. , 1995 , apjss , 101 , 117 krgel e. , chini r. , 1994 , a&a , 287 , 947 ( kc ) lada c. j. , 1987 , in : peimbert m. , jugaku j. , eds . , ` star forming regions ' , iau syposium 115 , 1 , reidel , dordrecht lada c. j. , 1991 , in : lada c. j. , kylafis n. d. , eds . , ` the physics of star formation and early stellar evolution ' , 329 , kluwer , dordrecht lada c. j. , alves j. , lada e. a. , 1996 , aj , 111 , 1964 lada c. j. , gottlieb c. a. , litvak m. m. , lilley a. e. , 1974 , apj , 194 , 609 lada c. j. , wilking b. , 1984 , apj , 287 , 610 langer w. d. , castets a. , lefloch b. , 1996 , apj , 471 , l111 larson r. b. , 1969 , mnras , 145 , 271 larson r. b. , 1981 , mnras , 194 , 809 liseau r. , sandell g. , knee l. b. g. , 1988 , a&a , 192 , 153 lizano s. , shu f. , 1989 , apj , 342 , 834 loren r. b. , 1976 , apj , 209 , 466 lovas f. j. , 1992 , j. phys . chem . , 21 , 181 mardones d. , myers p. c. , tafalla m. , wilner d. j. , bachiller r. , garay g. , 1997 , apj , 489 , 719 matthews h. , 1996 , jcmt guide for prospective users matthews n. , 1986 , phd thesis , university of kent mckee c. f. , zweibel e. g. , 1995 , apj , 440 , 686 mcmullin j. p. , mundy l. g. , blake g. a. , 1994a , apj , 437 , 305 mcmullin j. p. , mundy l. g. , wilking b. a. , hezel t. , blake g. a. , 1994b , apj , 424 222 mouschovias t. ch . , 1994 , in : clemens d. p. , barvainis r. , eds . , ` clouds , cores and low mass stars ' , asp conference series , 65 , 134 myers p. c. , 1983 , apj , 270 , 105 myers p. c. , bachiller r. , caselli p. , fuller g. a. , mardones d. , tafalla m. , wilner d. j. , 1995 , apj , 449 , l65 myers p. c. , mardones d. , tafalla m. , williams j. p. , wilner d. j. , 1996 , apj , 465 , l133 narayanan g. , walker c. k. , buckley h. d. , 1998 , apj , 496 , 292 padman r. , 1990 , ` specx users manual ' , mrao , university of cambridge park y .- s . , hong s. s. , 1995 , a&a , 300 , 890 park y .- s . , hong s. s. , 1998 , mnras , 300 , 528 poynter r. l. pickett h. m. , 1985 , applied optics , 24 , 2235 prasad s. s. , huntress w. w. jr . , 1982 , apj , 260 , 590 pudritz r. , norman c. a. , 1986 , apj , 301 , 571 rawlings j. m. c. , hartquist t. w. , menten k. m. , williams d. a. , mnras , 1992 , 255 , 471 rybicki g. b. , 1972 , ` line formation in magnetic fields ' , national centre for atmospheric research , boulder , colorado rybicki g. b. , 1984 , in : kalkofen w. , ed . , ` methods in radiative transfer ' , cambridge university press rybicki g. b. , hummer d. g. , 1991 , a&a , 245 , 171 safier p. n. , mckee c. f. , stahler s. w. , 1997 , apj , 485 , 660 sandell g. , knee l. b. g. , aspin c. , robson e. i. , russell a. p. g. , 1994 , a&a , 285 , l1 sargent a. i. , 1979 , apj , 223 , 884 shu f. h. , 1977 , apj , 214 , 488 stenholm l. g. , 1977 , a&a , 54 , 577 straizys v. , cernis k. , bartasit s. , 1996 , baltic astronomy , 5 , 125 strom s. e. , vrba f. j. , strom k. m. , 1976 , aj , 81 , 638 swade d. a. , 1989 , apj , 345 , 828 terebey s. , shu f. h. , cassen p. , 1984 , apj , 286 , 529 testi l. , sargent a. i. , 1998 , apj , 508 , 91 testi l. , sargent a. i. , olmi l. , onello j. s. , 2000 , apj , 540 , 53 ungerechts h. , gsten r. , 1984 , a&a , 131 , 177 vzquez - semadeni e. , cant j. , lizano s. , 1998 , apj , 492 , 596 walker c. k. , narayanan g. , boss a. p. , 1994 , apj , 431 , 767 ward - thompson d. , 1996 , ap&ss , 239 , 151 ward - thompson d. , buckley h. d. , greaves j. s. , holland w. s. , andr p. , 1996 , mnras , 281 , l53 warin s. , castets a. , langer w. d. , wilson r. w. , pagani l. , 1996 , a&a , 306 , 935 white g. , casali m. m. , eiroa c. , 1995 , a&a , 298 , 594 ( wce ) whitworth a. p. , bhattal a. s. , francis n. , watkins s. j. , 1996 , mnras , 283 , 1061 whitworth a. p. , ward - thompson d. , 2001 , apj , 547 , 317 willacy k. , rawlings j. m. c. , williams d. a. , 1994 , mnras , 269 , 921 williams j. p. , myers p. c. , 2000 , apj , 537 , 891 wilson t. l. , rood r. t. , 1994 , ara&a , 32 , 191 zhang c. y. , laureijs r. j. , clark f. o. , wesselius p. r. , 1988 , a&a , 196 , 236 zhou s. , 1992 , apj , 394 , 204 zhou s. , 1995 , apj , 442 , 685 zhou s. , evans n. j. , kompe c. , walmsley c. m. , 1993 , apj , 404 , 232	we present a radiative transfer model , which is applicable to the study of submillimetre spectral line observations of protostellar envelopes . the model uses an exact , non - lte , spherically symmetric radiative transfer ` stenholm ' method , which numerically solves the radiative transfer problem by the process of ` @xmath0-iteration ' . we also present submillimetre spectral line data of the class 0 protostars ngc1333iras2 and serpens smm4 . we model the data using the stenholm code . we examine the physical constraints which can be used to limit the number and range of parameters used in protostellar envelope models , and identify the turbulent velocity and tracer molecule abundance as the principle sources of uncertainty in the radiative transfer modelling . we explore the trends in the appearance of the predicted line profiles as key parameters in the models are varied , such as infall velocity profile , turbulence and rotation . the formation of the characteristic asymmetric double - peaked line profile in infalling envelopes is discussed . _ we find that the separation of the two peaks of a typical infall profile is dependent not on the evolutionary status of the collapsing protostar , but on the turbulent velocity dispersion in the envelope_. we also find that the line shapes can be significantly altered by rotation . fits are found for the observed line profiles of iras2 and smm4 using plausible infall model parameters . the density and velocity profiles in our best fit models are inconsistent with a singular isothermal sphere model ( sis ) , since for both objects modelled , the infall velocities appear further advanced than a sis model would predict , given the density profile . we find better agreement with a form of collapse which assumes non - static initial conditions in agreement with other recent findings . we also find some evidence that the infall velocities are retarded from free - fall towards the centre of the cloud , probably by rotation , and that the envelope of smm4 is rotationally flattened . latexl-.36em.3ex-.15em t-.1667em.7ex-.125emx = [ section ] [ firstpage ] stars : formation ism : clouds radio continuum : ism
one of the most important questions to be clarified in the hydrodynamical approach to the relativistic heavy ion physics is the effect of dissipative processes . the second order theory , first proposed by muller and developed by israel and stewart , has been considered standard approach for this problem @xcite . but it is quite complex and involves many unknown parameters from the point of qcd dynamics so that its complete application to practical problems such as relativistic heavy ion reactions has not been done yet @xcite . in this work , we propose an alternative approach to this question @xcite . we show that the physical origin of the second order theories can easily be understood in terms of memory effects . the irreversible current modified by the memory effects becomes consistently with causality and sum rules @xcite . based on this idea , we introduced the memory effect to the relativistic dissipative hydrodynamic of landau @xcite , where we introduce only one extra parameter , the relaxation time @xmath0 in addition to the usual viscosity coefficients of the navier - stokes equation . the resulting equation becomes hyperbolic @xcite . the effect of viscosity is also important when we discuss the possible generation and propagation of shock waves in the qcd medium created in the process of relativistic heavy ion collisions . as discussed extensively in this conference@xcite , it has been suggested that a high energy jet propagating in the qgp may generate a mach cone and observables associated with such phenomena may bring important information of the genuine hydrodynamical properties of the matter @xcite . the dynamical simulation of shock wave generation is very difficult even for the non - relativistic regime . a full 3d simulation of shock wave dynamics has never been done for the heavy ion collisions . in this work , we apply our formulation to the calculation of full 3d relativistic ( causal ) shock wave problem . the implementation of our method to the existing ideal hydro - codes is straightforward , particularly to those based on the local lagrangian coordinate system such as spherio @xcite . we organize the present work as follows . in the next section , we briefly introduce our formalism and discuss its application to the generation of shock waves . in section 3 , we present some results of 3d calculation of shock wave propagation within the causal formulation of the dissipative hydrodynamics . in section 4 , we discuss the result and perspectives . the fundamental problem of the first order theory like the navier - stokes theory comes from the fact that the diffusion equation is parabolic . the physical origin of this problem can be followed up to the fact that the irreversible current @xmath1 is assumed to be proportional to a thermodynamic force @xmath2 as @xmath3where the onsager coefficient @xmath4 is , in general , a function of thermodynamic quantities . usually @xmath2 is related to the inhomogeneity in the density . when the microscopic rearrangement time scale is not negligible compared to the time scale of the change in the irreversible current , then the above should be replaced by the equation of motion for the current , @xmath5where @xmath6 is the relaxation time . for very small @xmath7 we recover eq.([curr ] ) . thus eq.([curr ] ) can be understood as the large viscous limit of the damped motion , where the velocity ( current ) is proportional to the force ( aristotelian vision ) . it can be shown that the above modification is enough to convert the parabolic nature of a diffusion equation to hyperbolic one @xcite . for the relativistic hydrodynamics , we have to consider several different kind of thermodynamical forces related to the velocity and density inhomogeneity . they are @xmath8 and@xmath9where @xmath10 is the four - velocity of the fluid , @xmath11 with @xmath12 the chemical potential . these inhomgeneities generate the corresponding irreversible currents and the analogous equations to eq.(irrev ) for them should be @xmath13where @xmath14 and @xmath15 @xmath16 and @xmath17 are bulk viscosity , shear viscosity and thermal conductivity coefficients , respectively . the energy - momentum tensor is expressed with these currents as @xmath18 where @xmath19 and @xmath20 is the double symmetric traceless projection , @xmath21 whereas the conserved baryon number current is given by @xmath22 the hydrodynamic equations are @xmath23 although eqs.([heat],[hydro ] ) together with the equation of state give the complete description of the hydrodynamical motion of the system , in practice , some additional care to be taken , especially for the simulation of the shock wave dynamics . whenever there exists a shock wave , always occurs an entropy production through the shock front . in an idealized hydrodynamical approach , the shock front is a discontinuity in thermodynamical quantities in a hydrodynamic solution . mathematically speaking , it should be treated as the boundary condition to connect two distinct hydrodynamic solutions . physically , it is not a real discontinuity , but a quick change of the density in the region where the local equilibrium is not satisfied . thus it has a finite thickness at least a few times of the mean - free path ( typical microscopic scale of distance ) for a stationary shock . under a dynamical condition such as relativistic heavy ion collisions , the compression shock may have much more larger thickness due to the many complicated local transient properties . to reproduce true shock wave phenomena , the full degrees of freedom of the hydrodynamics , together with a proper boundary condition correctly connecting to regions through the non - equilibrated domain of the shock , are required . the usual numerical approach of hydrodynamics excludes such a possibility from the beginning . since there exist no short wavelength excitation modes due to the finite discretization , the energy and momentum conservation required by the hydrodynamics result in a very rapidly oscillations of the variables near the shock region . these quick oscillations are somewhat compensating the thermal energy associated to the entropy production throughout the shock front . in order to avoid these oscillations , von neumann and richtmeyer@xcite introduced the method of pseudo - viscosity . the idea is to put the bulk viscosity in the shock region to replace the entropy production through the shock front . the form of the pseudo - viscosity is chosen so that asymptotic values of thermodynamic quantities connects smoothly from one side to the other through the shock , satisfying the huguniot - rankine boundary condition . for the relativistic extension , we have to transform this formalism in the causal form . thus , we introduce the causal pseudo - viscosity by using the bulk viscosity coefficient @xmath24 in eq.([heat ] ) as @xmath25 $ ] for @xmath26 and @xmath27 otherwise where @xmath28 and @xmath29 and @xmath30 are constants proportional to the space resolution scale of the numerical solution . from the condition of causality , the relaxation time associated to this pseudo - viscosity should satisfy @xmath31this means , the larger the viscosity , the larger the relaxation time . in fig.1 , we show the example of creation and propagation of shock wave induced by a small qgp drop injected into the homogeneous qgp fluid at a relativistic energy ( @xmath32 ) . here , we take a massless ideal gas equation of state , and the incident drop has half the temperature of the background in the local rest frame . the relaxation time necessary to maintain the causality in this calculation is found to be @xmath33 . this seems to be large , but the effect of shock wave requires a large viscosity so that a large relaxation time is necessary . we can see the generation of mach cone and also the turbulent structure after the shock cone . we found that the cone angle depends on the viscosity and also the related relaxation time @xmath34 for a finite @xmath6 the mach cone opens more than the usual estimate for ideal fluid case . we presented the first full 3d relativistic viscous hydrodynamic calculation which preserve the causality correctly . in the case of shock wave simulation , it is fundamental to include the bulk viscosity to take account for the entropy generation through the shock front . to satisfy the causal propagation of the signal , the presence of viscosity requires a finite relaxation time . it is found that the mach cone angle depends on the viscosity and the relaxation time . it is also observed that a very turbulent structure after the shock . more quantitative and systematic study of the shock wave generation and its propagation in an expanding qgp is in progress . chaudhuri , nucl - th/0604014 , u. heinz , h. song and a.k . chaudhuri , phys.rev.c73:034904 ( 2006 ) , a. muronga and d.h . rischke , nucl - th/0407114 , a. muronga , phys.rev.c69:034903,2004 , phys.rev.lett.88:062302,2002 , r. baier , p. romatschke , u. a. wiedemann , hep - ph/0602249 , nucl - th/0604006 .	the first 3d calculation of shock wave propagation in a homogeneous qgp has been performed within the new formulation of relativistic dissipative hydrodynamics which preserves the causality . we found that the relaxation time plays an important role and also affects the angle of mach cone .
robustness is one of the most important properties of complex networks and has been widely explored on single networks in the last decade@xcite . however , complex systems are rarely isolated . the more casual situation is that networks usually interact with other networks such as transportation networks and financial systems@xcite . in the case of interdependent networks , conclusions are often far different from single networks . in particular , a removal of a very small fraction of nodes can lead to catastrophic failures on the whole network@xcite . a theoretical framework based on percolation theory has been established to analyze the resilience of interdependent systems@xcite , and much details have been explored@xcite . the fraction of interdependent nodes is one important factor that will influence the phase transition of the networks@xcite . also , the overlap of links can significantly change the properties of the percolation , and there is a critical point above which the emergence of the mutually connected component is continuous@xcite . the presence of degree correlations in multiplex networks can modify drastically the percolation threshold@xcite . most previous models have focused on interdependent random and scale - free networks in which space restrictions are not considered . however , many real - world systems such as power grid networks and computer networks are embedded in two - dimensional space@xcite . in interdependent random and scale - free networks , the overlap of links and degree correlations will change the properties of phase transition . nevertheless for spatially embedded interdependent networks which are modeled as square lattices , the overlap of links or the degree correlations of nodes lose their significance , since their network topologies are identical . the spatially interdependent networks are extremely vulnerable . any fraction of interdependent nodes will lead to first order transition@xcite . from an identical dependency map to totally random dependency map , the randomness of the dependency map may be one of the most important factors leading to the emergence of discontinuous percolation . in most real interdependent systems , dependencies are neither totally random nor identical . research on the resilience of intermediate systems that lie somewhere in between these two extremes is of high practical significance and needs further exploration . from this perspective , we study the relationship between the dependency s randomness and stability of the system of two interdependent spatially embedded networks . we use approximate entropy(@xmath0 ) as the measure of randomness . one of the big challenges here is how to introduce controlled degree of randomness into the system . therefore , we propose an intermediate model which describes the system with dependency map between identical map and totally random map . inspired by the constructing procedure of the watt - strogatz small - world model@xcite , starting from an identical dependency map , we rewire each dependency link at random with probability @xmath4 . by increasing @xmath4 from @xmath5 to @xmath6 , the @xmath0 increases monotonically . therefore , the traverse of randomness can be generally represented by @xmath4 . we reveal that there is a critical value @xmath7 , for which the percolation transition becomes continuous , whereas for any @xmath8 the collapse is discontinuous . changing the topologies on a single layer , we discover that @xmath7 is different for interdependent scale - free networks , watts - strogatz networks , and erds - rnyi networks . the critical threshold increases with @xmath4 when @xmath9 and remains approximately constant when @xmath10 . additionally , we present an analytical method for time scale of cascade failures based on critical p and find that the four topologies display rich transient properties when @xmath4 changes from @xmath5 to @xmath6 . finally , we analyze the influence of limited dependency length on spatial networks . we show that with the same dependency length , a linearly dependent system is always continuous , but not continuous for some locally randomly dependent system . our results show that the randomness of dependency may be one of important factors for extreme vulnerability of spatially interdependent systems . our model of interdependent networks is realized via two networks(@xmath11 ) a and b under full dependency . here one network is the copy of the opposite network and their average degree @xmath12(the same as a square lattice ) . the degree distribution of the scale - free network is @xmath13 where @xmath14 . in each network , each node has two types of links : connectivity link and dependency link . also every node in network a is connected with one and only one node in network b. for a square lattice , each node is connected to its four nearest neighbors within the same lattice via connectivity links . all dependencies in our model are mutual and bidirectional . dependency is taken to mean that if a node in network a is removed from the system and a node in b that depends on it will be removed from b as well . thus failures of nodes iterate until mutually connected giant component of both networks emerges . this process is called cascade failures and see methods for details of cascade process of the system . there are two extreme situations . i ) node @xmath15 in a depends on node @xmath16 in the b such that @xmath17 . we call it identity dependency map(fig.[fig1].a ) . ii ) the random dependency map as most papers considered(fig.1.b ) . like the constructing procedure of the watt - strogatz small - world model , starting from the identity dependency map , we rewire each dependency link at random with probability @xmath4 , while guaranteeing that each node in a depends on one and only one node in b(@xmath18 ) . we sample @xmath19 and plot them in fig.1 . note : we must figure out that our model is different from partially interdependent lattices proposed by amir bashan et al@xcite . in partially interdependent lattices , there are interdependent lattices with a fraction @xmath4 of interdependent nodes and the remaining @xmath20 of nodes autonomous . in our model , however , the remaining @xmath20 nodes are connected with the identical nodes in opposite network . the subtle difference between these two models will lead to huge difference in cascade failure process . we will illustrate it in fig.[fig2 ] . in fig.[fig2 ] , we can see that the cascade failures process differs much between these two models : with the same @xmath21 and @xmath22 , the size of the giant component in our model is @xmath23 , while the size of giant component in partially interdependent networks is @xmath24 . this apparently show that the our model is different partially interdependent lattices . entropy can be used to measure the randomness effectively@xcite . in fact , approximate entropy(@xmath0 ) is adopted in this paper for computation convenience . when @xmath25 , @xmath0 is nearly 0 , and when @xmath26 , it reaches its maximum . the @xmath0 is a continuous and monotonically increasing function of @xmath4(fig.3 ) . however the randomness is not fully represented by rewiring dependency links , since the locally randomly interdependent lattice@xcite in which @xmath26 is not totally randomly interdependent but with length constraints . then considering a more casual situation , the permutation of @xmath27 can not be exhausted by rewiring the dependency links of identical mapping at probability @xmath4 . but as the approximate entropy changes continuously with @xmath4 , we can change @xmath4 to generally represent all approximate entropies . through simulation , we find that there is a critical @xmath28 for an interdependent lattice network below which the percolation is second - order but first - order above . in fig.4 , we can see that for @xmath29 , the decrease of giant component occurs in multiple steps , characteristic of a second - order transition . for @xmath30 and @xmath31 , the giant component may completely collapse by removal of a small fraction of nodes , characteristic of a first - order transition(fig.4 ) . @xmath2 increases with @xmath4 . when @xmath4 is relatively small , @xmath2 increases approximately linearly . and when @xmath32 , @xmath2 remains almost constant , i.e. as more random the dependency is , as more fragile is the system . analogously , for interdependent scale - free networks , watts - strogatz networks and erds - rnyi networks , there is also a critical @xmath7 . we find that @xmath33 . additionally @xmath2 of lattice networks is generally greater than that of other networks . smaller fraction @xmath34 will lead to cascade failures . this means that an interdependent scale - free network is most robust under random attacks , while the square lattice is most vulnerable . a random network is more stable than a small - world one . the time scale of cascade failures , i.e. the time that the interdependent networks need to collapse to the stationary state is one evidently important merit for system s resilience . the number of iterations ( noi ) increases and reaches its peak at the critical @xmath7 and goes quickly down to a small value with @xmath35 ( fig.5 ) . so the noi at @xmath2 is an effective measure for time scale of the system . the noi at @xmath2 ( i.e.@xmath36 ) is a function of @xmath4 . @xmath36 increases quickly with @xmath4 when @xmath37 and declines very gently above @xmath7 for interdependent lattice networks . for a scale - free network , it increases until @xmath7 and then starts to decline . for random networks and small - world networks , it increases monotonously with @xmath4 but the variation tendency becomes gentle above @xmath7(fig.6 ) . all four interdependent systems have variation tendency s changes around @xmath7 . @xmath36 of interdependent lattice networks is greater than those of scale - free , small - world and random networks when @xmath38 . @xmath36 of interdependent square lattice is smaller than those of all other three network types when @xmath39 . the noi reflects the time scale of system collapse . our results show that the transient characteristics of the four systems go through rich changes with the variation of @xmath4 . finally we check locally interdependent networks in which the distance between two interdependent nodes is limited(@xmath40 i.e. @xmath41 and @xmath42 in reference@xcite ) . for simplicity , we consider one more special condition . here we split the whole network into @xmath43 small blocks and dependency links are randomized within each block . we find that there is critical distance @xmath44 under which the percolation is continuous but discontinuous above @xmath45(fig.7 ) . the @xmath45 is greater than @xmath46 in @xcite because the randomness(approximate entropy ) here is lower than that in @xcitewith the same distance . the corresponding approximate entropy @xmath47 of @xmath45 is approximately equal to @xmath1 . compared with locally random dependency , the linear dependency map is more robust . for the linear dependency map , the percolation is always continuous(fig.7 ) . although the dependency distance changes strongly , the approximate entropies of those dependency maps are almost equal to @xmath5 . so their percolation properties are nearly the same as that of a single lattice . it is thus clear that the randomness may be a more important factor leading to cascade failures than dependency distance . furthermore , it is possible that the randomness of dependency is related to other metrics of interdependent networks such as dimension . the dimension of networks is a function of the distribution of link lengths@xcite . for spatially embedded networks , the dimension is one of the most fundamental quantities to characterize its structure and very likely will influence its percolation property@xcite . however , how interdependency relationships between networks change the dimension of the system has not been figured out . in reference @xcite , the authors discover that the dependency length play an critical role in the percolation transition . however , we found that under linear dependency map , the change of dependency length influence the percolation property little . from the discrepancy of those two situations , it can be inferred that local property of dependency relationship make a big difference . and the local property of dependency will directly influence the local topological inter - similarity between networks . randomness happen to reflects the local property of dependency ( we can see this from the computation steps of approximate entropy in methods ) . in spatially interdependent embedded networks , the local characteristics of dependency can be more intuitively characterized as the relative length of dependency links . under linear dependency map , the relative length of dependency links and the approximate entropy of dependency map are nearly @xmath5 . no matter how large @xmath48 is , they changes little and are still nearly @xmath5 , so the percolation changes little . on the other hand , the smaller the relative length is , the less dimension is changed from that of single network . there should exit some relations between dimension and the randomness of the dependency map . in many real interdependent systems such as coupled power grid and communication network , the dependency relationship is neither completely regular nor completely random , but lies somewhere between these two extremes . the transition from regular to random dependency is one of the keys to extreme vulnerability of spatially interdependent systems . from the proposed intermediate cascade failure model ( from regular to random dependency ) , we find there is a transform from continuous percolation to discontinuous percolation with the randomness variation of the dependency map between two interdependent networks . we emphasize the generic character of our model because the dependency map could influence not only the resilience but also synchronization , disease spreading and other dynamic process in interdependent networks . with suitable modification , our model could be applied to understand the dynamical process in most real interdependent systems since the dependency maps between networks are more various and complicated instead of totally random dependency or regular dependency . the time scale of cascade failures is evidently important for system s resilience , but it has received little attentions in the analysis of resilience so far . in different dynamic processes , the characteristic time scales of systems vary greatly . for instance biological systems , social systems and financial market dynamics have time scale much longer than that of cascade failures of power grid . our analytic method based on critical p is simple and effective for characterizing the time scale of different systems . generally , the system which has a shorter time scale demands higher requirements for our responding speed to catastrophe and brings much bigger challenges for us to take mitigation actions than those with longer time scale . therefore , our method may provide a clue for research on revealing the transient mechanism and mitigation of cascade failures in interdependent networks . the randomness of the dependency maps the interdependent square lattice is measured by approximate entropy . entropy can effectively reflects the randomness of a sequence . however , for computation convenience , we choose the approximate entropy as the measure of the randomness of the system . the approximate entropy is denoted by @xmath0 and is computed by following steps@xcite : the percolation transition is studied by randomly removing a fraction @xmath61 of nodes and the links attached to them from both networks simultaneously . then , on each network , clusters which are detached from the largest connected component are removed . after that , the nodes in each network which lost their supporting nodes in the opposite network are removed . this in turn causes more clusters to break off from the giant component and this process is continued until no more clusters break off . first , we analyze the situation with totally random dependency maps . after the initial attack , only @xmath62 remains functional . each node in a that is removed causes the removal of its interdependent node in b. then only @xmath63 nodes in b remain alive . this produces further damage in a and causes cascading failures . the cascade failures can be represented by the recursive equation for the survived fraction @xmath64@xcite . @xmath65 in the limit @xmath66 , eq.(1 ) yields the equation for the mutual giant component at the steady state , @xmath67 equation ( 2 ) can be solved graphically by the intersection between the curve @xmath68 and the straight line @xmath69 . next , we consider the mutual percolation for more casual situations where the dependency is not totally random . for every dependency link , there is a probability @xmath4 to rewire it at random . this is equivalent to the situation with a fraction @xmath20 nodes mapping to itself and the remaining @xmath4 nodes have a random dependency map . the case of @xmath26 corresponds to the scenario of a random dependency map and @xmath25 is identical to the conventional percolation on a single lattice . for the initial attack which destroys a fraction @xmath61 of nodes , @xmath70 nodes are removed . we compute the probability @xmath71 that one node in a depends on the same node in b. for @xmath72 nodes in totally random dependency networks , the number of nodes @xmath73 in the same location of both networks is @xcite : @xmath74 @xmath75 when @xmath72 is very large , the computation of @xmath76 is very inconvenient . for computation simplication , when @xmath77 , we have : @xmath78 where @xmath79 is the euler s number and @xmath80 is the integer part of @xmath81 . then , @xmath82 so for each node , the probability that it is in the same location of a and b is : @xmath83 when @xmath84 , @xmath85 . the initial attack causes some number of nodes to be disconnected from the giant component in both networks a and b. furthermore , because of the dependency relationship , the nodes disconnected from a will lead to further damages . @xmath86 increases with @xmath71 . the greater @xmath71 is , the more nodes disconnected from the giant component of a overlap the nodes in b. so further damage decreases and cascade failures are weakened ( or prevented ) from the beginning . for @xmath25 , the cascade failures are prevented from the start and the percolation is continuous . when @xmath87 , the totally random dependency map will lead to a first - order transition . as @xmath4 increases , the percolation transition changes from a continuous transition to a discontinuous one . there must be one critical @xmath7 beyond which the percolation transition become discontinuous . this paper is supported by the national natural science foundation of china ( grant nos . 61170269 , 61472045 ) , the beijing higher education young elite teacher project ( grant no.yetp0449 ) , the asia foresight program under nsfc grant ( grant no . 61411146001 ) and the beijing natural science foundation ( grant no . 4142016 ) . 31ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ] + 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) in @noop _ _ ( , ) pp . @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( )	for interdependent networks with identity dependency map , percolation is exactly the same with that on a single network and follows a second - order phase transition , while for random dependency , percolation follows a first - order phase transition . in real networks , the dependency relations between networks are neither identical nor completely random . thus in this paper , we study the influence of randomness for dependency maps on the robustness of interdependent lattice networks . we introduce approximate entropy(@xmath0 ) as the measure of randomness of the dependency maps . we find that there is critical @xmath1 below which the percolation is continuous , but for larger @xmath0 , it is a first - order transition . with the increment of @xmath0 , the @xmath2 increases until @xmath0 reaching @xmath3 and then remains almost constant . the time scale of the system shows rich properties as @xmath0 increases . our results uncover that randomness is one of the important factors that lead to cascading failures of spatially interdependent networks . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the interdependent networks which fully consider the interactions between networks have been used to model real complex systems better . robustness is one of most important properties for interdependent networks especially spatially interdependent networks , since most of the infrastructure networks are spatially networks . in real interdependent networks , the dependency relationship are not usually random . thus we analyze how the randomness of dependency map affects the robustness of interdependent lattices which are used to model the spatially interdependent networks . we find that the randomness of dependency map between networks is quite critical for the robustness of interdependent lattices . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
much attention has been devoted to the electron spin dynamics in semiconductors for the past three decades.@xcite especially , recent experiments have shown extremely long spin lifetime ( up to hundreds of nanoseconds ) in @xmath0-type bulk zinc - blende semiconductors ( such as gaas).@xcite moreover , a lot more investigations have been performed on various low dimensional systems,@xcite and spin lifetime as long as tens of nanoseconds has been reported in ( 110)-oriented gaas quantum wells ( qws)@xcite at room temperature and in @xmath7-type gaas : be / al@xmath8ga@xmath9as double hetero - structures@xcite at low temperature . in these studies , understanding the spin relaxation / dephasing ( r / d ) mechanism is one of the most important problems as it is the prerequisite for the application of the spintronic devices . it is understood that the dayakonov - perel ( dp ) mechanism is the leading spin r / d mechanism in @xmath0-type zinc - blende semiconductors.@xcite this mechanism is composed of the contribution from the dresselhaus term,@xcite which is due to the lack of inversion symmetry in the zinc - blende crystal brillouin zone ( sometimes referred to as the bulk inversion asymmetry ) , and that from the rashba term,@xcite which originates from the asymmetric potential within a qw along the growth direction ( sometimes referred to as the structure inversion asymmetry ) . both appear as effective magnetic fields . for narrow ( 001 ) gaas qw without the additional large bias voltage , the dresselhaus term is the leading term:@xcite @xmath10 in which @xmath11 represents the average of the operator @xmath12 over the electronic state of the lowest subband . under the finite square well assumption , @xmath13 where @xmath14 and @xmath15 are the lowest energy solutions of the equations @xmath16 and @xmath17 with @xmath18 , @xmath19 and @xmath20 denoting the well depth , well width and the effective mass respectively , in the limiting case , @xmath21 . @xmath22 is the spin splitting parameter.@xcite there are a lot of theoretical investigations on the spin r / d due to the dp mechanism lately.@xcite most of them are within the framework of single - particle approximation and the coulomb scattering is thought to be irrelevant in the spin r / d . recently wu _ et al . _ performed a full microscopic investigation on the spin r / d and showed that the single - particle approach is inadequate in accounting for the spin r / d.@xcite in this approach , the momentum dependence of the effective magnetic field ( the dp term ) and the momentum dependence of the spin diffusion rate in the direction of the spacial gradient@xcite or even the random spin - orbit interaction@xcite serve as inhomogeneous broadening.@xcite in the presence of the inhomogeneous broadening , any scattering ( even the spin - conserving scattering ) , including the coulomb scattering,@xcite can cause irreversible dephasing . moreover , this approach also includes the counter effect of the scattering to the inhomogeneous broadening , _ i.e. _ , the suppression of the inhomogeneous broadening by the scattering . finally , this approach is valid not only near the equilibrium , but also far away from the equilibrium,@xcite and is applicable to both the strong ( @xmath23 ) and the weak ( @xmath24 ) scattering limits,@xcite with @xmath25 representing the momentum relaxation time . in the weak scattering limit , the counter effect of the scattering is less important and adding additional scattering ( including the coulomb scattering ) causes stronger spin r / d . whereas in the strong scattering limit , adding additional scattering always increases the spin r / d time . the feature is more complicated when @xmath26.@xcite in above studies,@xcite we have been focusing on the high temperature regime ( @xmath27 k ) where the electron - acoustic ( ac ) phonon scattering , which is more complicated in numerical calculation than the electron - longitudinal optical ( lo ) phonon scattering , is negligible . in this paper , we extend the scope of our approach to study the spin kinetics at low temperature regime by including the electron - ac phonon scattering . moreover , we compare the spin relaxation time ( srt ) obtained from our theory with the experimental data over a wide temperature regime and show the excellent agreement of our theory with the experiment . we further show that the coulomb scattering is important to the spin r / d not only at high temperatures,@xcite but also at low temperatures . the electron density , impurity density , well width , temperature and electric field dependences of the srt are studied in detail . this paper is organized as follows : in sec . ii we set up the model and give the kinetic spin bloch equations . in sec . iii we compare our results with the experimental data . then , we investigate the temperature dependence of the spin relaxation under different conditions such as electron densities , impurity densities and well widths in sec . the effect of coulomb scattering is also addressed . the hot - electron effect in spin relaxation is investigated in sec . v. we summarize in sec . we start our investigation from an @xmath0-type gaas ( 001 ) qw with the growth direction along the @xmath28-axis . a moderate magnetic field @xmath29 is applied along the @xmath30-axis ( in the voigt configuration ) . the kinetic spin bloch equations can be constructed by using the nonequilibrium green function method:@xcite @xmath31 with @xmath32 representing the single particle density matrix elements . the diagonal and off - diagonal elements give the electron distribution functions @xmath33 and the spin coherence @xmath34 . the second term in eq . ( [ bloch ] ) describes the energy input from the external electric field @xmath35 . the coherent terms @xmath36 describe the precession of the electron spin due to the applied magnetic field @xmath29 and the effective magnetic field @xmath37 [ eqs . ( [ omegax]-[omegaz ] ) ] as well as the effective magnetic field from the hartree - fock coulomb interaction.@xcite @xmath38 in eq . ( [ bloch ] ) denote the electron - lo - phonon , the electron - ac - phonon , the electron - nonmagnetic impurity and the electron - electron coulomb scattering . their expressions are given in detail in ref . , except the additional matrix elements of the electron - ac - phonon scattering . for the electron - ac - phonon scattering due to the deformation potential , the matrix elements are given by @xmath39,@xcite and for the scattering due to the piezoelectric coupling , the matrix elements read @xmath40 for the longitudinal phonon and @xmath41 for the transverse phonon.@xcite here @xmath42 ; @xmath43 ; @xmath44 ev is the deformation potential ; @xmath45 g/@xmath46 is the mass density of the crystal ; @xmath47 m / s ( @xmath48 m / s ) is the velocity of the longitudinal ( transverse ) sound wave ; @xmath49 denotes the static dielectric constant ; and @xmath50 v / m represents the piezoelectric constant.@xcite the ac phonon spectra @xmath51 are given by @xmath52 for the longitudinal mode and @xmath53 for the transverse mode . the form factor is @xmath54\big\}^2\end{aligned}\ ] ] with @xmath55 . the numerical schemes of the electron - electron coulomb , the electron - impurity as well as the electron - lo phonon scattering have been given in detail in ref . , whereas the numerical scheme for the electron - ac - phonon scattering is presented in appendix a. the electron interface - phonon scattering is negligible due to the thick gaalas barrier . in addition , as we are going to explore the spin r / d over a wide range of electron densities , in the present paper we use the screening under the random phase approximation @xcite rather than the one in the limiting ( degenerate or nondegenerate ) cases for the screened coulomb potential:@xcite @xmath56 where @xmath57 is the bare coulomb potential and @xmath58 in this way , we also take into account the hot - electron effect on the screening . by numerically solving the kinetic spin bloch equations with all these scattering explicitly included , one is able to obtain the spin dephasing and relaxation times from the temporal evolutions of the spin coherence @xmath34 and the electron distribution functions @xmath59 . the irreversible spin dephasing time can be obtained by the slope of the envelope of the incoherently summed spin coherence@xcite @xmath60 , and the srt can be defined by the slope of the envelope of the difference between @xmath61 and @xmath62 , with @xmath63 . first , we compare the srt obtained from our microscopic approach with the experimental data by ohno _ et al . _ in 60 periods of gaas qws separated by @xmath64 nm thick al@xmath1ga@xmath2as barriers . the well width of each qw @xmath65 nm and the electron density @xmath66 @xmath5.@xcite the well depth of the gaas well confined by al@xmath67ga@xmath9as is roughly estimated to be @xmath68 % of @xmath69 ev , for @xmath70 and is therefore @xmath71 mev.@xcite differently from our previous fit@xcite with the experiment data at high temperatures by malinowski _ et al._@xcite where we had two fitting parameters , _ i.e. _ , the spin splitting parameter @xmath22 and the impurity density @xmath72 due to the absence of mobility data , here we have only one fitting parameter @xmath22 and the temperature sweeps from the very low temperature to the room temperature . the corresponding hall mobilities @xmath73 in the experiment@xcite can be found in ref . , also plotted in fig . 1(b ) . from the hall mobility , one can deduce the impurity density by calculating the transport mobilities@xcite @xmath74 with @xmath75 for the electron - ac phonon scattering due to the deformation potential ; @xmath76 for the electron - ac phonon scattering due to the piezoelectric coupling and the electron - lo phonon scattering ; and @xmath75 for the electron - ionized impurity scattering.@xcite the only fitting parameter @xmath22 is around @xmath77 in which @xmath78 ; @xmath79 denotes the band gap ; @xmath80 represents the spin - orbit splitting of the valence band ; @xmath81 stands for the electron mass in gaas ; and @xmath82 is a constant close in magnitude to the free electron mass @xmath83.@xcite for gaas when @xmath84 , @xmath85 ev@xmath86@xmath87 . the initial spin polarization @xmath88 is assumed to be @xmath89 % for weak polarization throughout the paper . in fig . 1(a ) , the srts @xmath90 obtained from our approach are plotted against the temperature with all the scattering included . @xmath29 and @xmath91 are taken to be zero , as in the experiment.@xcite @xmath92 , 1.0 and @xmath93 correspond to @xmath94 , 1.0 and @xmath95 with @xmath82 being the only not - fully - determined parameter in eq . ( [ gamma0 ] ) . one finds _ good _ agreement between our theory and the experiment data almost over the _ whole _ temperature regime . when @xmath96 is below @xmath97 k , there is no theoretical data due to the lack of experimental data for the hall mobility . et al . _ also fitted the same experiment data by using the single - particle theory without the coulomb scattering.@xcite they used a fourteen - band model to calculate the spin - orbit coupling . unlike our theory , their results can only give the boundary values of the srt in several cases rather than the exact data.@xcite this is because they did not take the full microscopic calculation , and the single - particle theory is inadequate in accounting for the spin r / d . the best - fitted value of the spin - orbit coupling parameter , _ i.e. _ , @xmath98 ev@xmath86@xmath87 , is close to the value calculated by kainz _ et al . _ @xmath99 ev@xmath86@xmath100 using the multiband envelope - function approximation.@xcite it is noted that the value for @xmath22 in gaas is still in debate . usually reported experimental values for @xmath22 ( @xmath101 ev@xmath86@xmath87 ) in bulk material are deduced from the dp spin relaxation mechanism within the framework of the single - particle approximation , where the coulomb scattering is not included.@xcite furthermore , the raman scattering experiment showed that @xmath102 ev@xmath86@xmath87 in asymmetric qw;@xcite and the hanle effect experiment showed that @xmath103 ev@xmath86@xmath87.@xcite theoretically , semi - empirical parameterized @xmath104 * k@xmath86p * calculations show that @xmath105 ev@xmath86@xmath87;@xcite and the self - consistent _ ab initio _ calculations predict @xmath106 and @xmath107 ev@xmath86@xmath87.@xcite our fitting result supports the last four experimental and theoretical results . it is also noted that our result further confirms the analytical result eq . ( [ gamma0 ] ) obtained from the perturbation,@xcite with @xmath84 . we now study the temperature dependence of the spin relaxation in detail . in the calculation , the electric field @xmath108 , the magnetic field @xmath109 t and the spin splitting parameter @xmath110 . we plot in fig . 2 the temperature dependence of the srt of gaas / al@xmath1ga@xmath2as qws with @xmath65 nm at different impurity densities when the electron densities are low ( @xmath111 @xmath5 ) [ fig . 2(a ) ] , medium ( @xmath112 @xmath5 ) [ fig . 2(b ) ] and high ( @xmath113 @xmath5 ) [ fig . 2(c ) ] respectively as solid curves . for the well width and the electron densities here , the linear terms in the dp terms [ eqs . ( [ omegax]-[omegaz ] ) ] are dominant , and only the lowest subband is relevant when @xmath114 k. it is seen from the figure that adding impurities always increases the srt . this is understood that the criterion of strong scattering @xmath23 is satisfied here at all temperatures , and therefore adding additional scattering always increases the srt.@xcite it is noted that @xmath25 here has been extended to include @xmath115 , _ i.e. _ , the contribution from the coulomb scattering . @xcite it is interesting to note that unlike our previous works focusing on high temperatures ( @xmath116k),@xcite the situation is more complicated at low temperatures . at low / medium electron densities [ fig . 2(a)/(b ) ] , the srt presents a peak at very low temperature ( near 20@xmath11730 k)/low temperature ( around 41 k ) and a valley around @xmath118 k ; whereas at high electron density [ fig . 2(c ) ] , the srt increases monotonically with @xmath96 . it is noted that at very low temperature ( around 20 k ) the electron - ac phonon scattering is negligible.@xcite @xmath119 from the electron - ac phonon scattering is around @xmath120 ps , two orders of magnitude larger than @xmath121 from the electron - electron coulomb scattering . in addition , @xmath122 from the impurity scattering is around 2 ps , one order of magnitude larger than @xmath121 , and has a very weak temperature dependence . therefore , the appearance of the peaks in fig . 2(a ) originates from the electron - electron coulomb scattering which dominates the scattering process . moreover , @xmath121 is a nonmonotonic function of temperature : @xmath123 at low temperature ( degenerate limit ) and @xmath124 at high temperature ( nondegenerate limit).@xcite the minimum of @xmath121 corresponds to the crossover from the degenerate limit to the nondegenerate one at @xmath125 . @xmath126 k when @xmath111 @xmath5 , in good agreement with the peaks obtained from our calculation with the exact coulomb scattering . therefore the srt increases / decreases with the temperature ( and the coulomb scattering ) in the degenerate / nondegenerate regime . once @xmath127 k , the electron - lo phonon scattering becomes comparable with the coulomb scattering and strengthens so rapidly with temperature that it completely surpasses the weak temperature dependence of the coulomb scattering : @xmath128 from the electron - lo phonon scattering varies from several picoseconds at @xmath118 k to several tenths of picosecond at @xmath129 k , and @xmath121 varies from @xmath130 ps to several picoseconds . therefore the srt increases with @xmath96 . when the electron density is @xmath131 @xmath5 , @xmath132 is nearly @xmath133 k. around this temperature , the electron - ac phonon scattering can not be overlooked although @xmath119 is still roughly one order of magnitude larger than @xmath121 . therefore , the reduction of the coulomb scattering after @xmath132 can be partly compensated by the increase of the electron - phonon scattering . as a result , one can see that the decrease of the srt after @xmath132 in fig . 2(b ) is much slower than that in fig . however , when the electron density is high enough , say @xmath134 @xmath5 in fig . 2(c ) , @xmath132 is nearly @xmath135 k , much larger than the case of low electron density . at this temperature , the electron - phonon scattering becomes comparable to the coulomb scattering and the strengthening rate of phonon scattering around this temperature is large enough to completely compensate , and even surpass , the weakening rate of the coulomb scattering . consequently the total scattering increases monotonously with @xmath96 . therefore the srt increases monotonically with @xmath96 . we further show the effect of the coulomb scattering on the spin relaxation . this was first proposed by wu and ning based on the inhomogeneous broadening induced by the energy dependence of the @xmath136-factor.@xcite then , we used our full microscopic approach and showed that the coulomb scattering makes marked contribution to the spin r / d when @xmath116 k when the inhomogeneous broadening is induced by the dp term.@xcite at low temperature ( @xmath137 k ) , glazov and ivchenko used perturbation method to show that the second - order coulomb scattering causes the srt.@xcite in the perturbation approach , the coulomb scattering contributes marginally to the spin r / d at high temperature . in our calculation , we include the coulomb scattering to all orders of the bubble diagrams as well as the counter effect of the coulomb scattering to the inhomogeneous broadening . in fig . 2 by plotting the srt for the case of @xmath138 , but without the coulomb scattering , as dashed curves , we show that the coulomb scattering makes marked contribution to the spin r / d over the whole temperature regime by increasing the spin r / d time.@xcite it is further seen from figs . 2(a ) and ( b ) that the peak disappears without the coulomb scattering . this is consistent with the previous discussion . it is interesting to see that in the absence of the coulomb scattering , the criterion for strong scattering regime @xmath23 is satisfied only when @xmath139 k. therefore the srt increases with @xmath96 when @xmath116 k. when @xmath140 k , @xmath141 is slightly smaller than @xmath130 , which is the intermediate scattering regime . the variation of the srt depends on the competition between the increase of the inhomogeneous broadening and the increase of the scattering with the temperature.@xcite for low / high electron density case , the temperature dependence of the electron - ac phonon scattering is less / more effective and the srt decreases / increases with @xmath96 . it is noted that in order to see the peaks at low electron density , it is important to have a high mobility sample ( low impurity density ) . this is because the ascendancy of the coulomb scattering can be impaired when the impurity scattering gets large enough and the total scattering is mainly determined by the impurity scattering . as the electron - impurity scattering depends weakly on the temperature , the temperature dependence of the inhomogeneous broadening from the dp term becomes the only variable element . therefore the srt decreases monotonically with @xmath96 as the solid curves with triangles in fig . 2 for the case of @xmath142 . this condition is not satisfied in the experiment by ohno _ et al._,@xcite and this is the reason why there is no peak in fig . 1 . however , apart from the peaks which are not observed yet , both the experiment and calculation show that the srt decreases with temperature at low electron densities when @xmath137 k. the srt at high electron density increases monotonically with temperature when the impurity density is low , which is also in agreement with the latest experiment by harley _ et al._@xcite finally we investigate the well width dependence of the srt . in fig . 3 we plot the srt versus temperature at well widths @xmath65 nm ( solid curves ) and 15 nm ( dashed curves ) respectively . we choose low and high impurity densities @xmath143 ( curves with dots ) and @xmath144 ( curves with triangles ) as well as low and high electron densities @xmath111 @xmath5 ( a ) and @xmath145 @xmath5 ( b ) . it is noted that the srt is enhanced by increasing the well width as @xmath146 in the dp term decreases with the increase of @xmath19 . moreover , as impurities further enhance the srt , it reaches to several nanoseconds at very low temperatures at high impurity density . we now turn to investigating the hot - electron effect on spin relaxation at low temperature . an electric field is applied parallel to the qw . similar to our previous investigation,@xcite electrons obtain a center - of - mass drift velocity ( and consequently an effective magnetic field proportional to the electric field ) and are heated to a temperature @xmath147 higher than @xmath96 . the numerical schemes of solving the hot electron problem has been laid out in detail in ref . .@xcite we plot the electric field dependence of the srt with @xmath148 nm and @xmath149 k for different impurity densities at low / high electron density ( @xmath111 @xmath5/@xmath145 @xmath5 ) in fig . 4(a)/(b ) . in the calculation , the magnetic field @xmath150 t and the spin splitting parameter @xmath110 as in the previous section . it is seen from the figure that unlike the high temperature case investigated before@xcite ( and also see fig . 4(c ) for @xmath151 k ) where the electric field can be applied easily to around 1 kv / cm , at low temperatures it can be applied only to a very small value due to the `` runaway '' effect.@xcite this is because at low temperature , the efficient electron - lo phonon scattering is missing and electrons are therefore very easily driven to very high momentum states by a very small electric field . it is interesting to see from the figure that differing from the high temperature case where the srt _ increases _ with the electric field ( see fig . 4 ( c ) and also ref . ) , here , for the case of low electron densities , the srt _ decreases _ with the field and for the case of high electron densities , the srt decreases / increases with the field at high / low impurity densities . these features at low temperature @xmath96 can be understood from the joint effects of the electric field @xmath91 to the scattering strength and the inhomogeneous broadening due to the dp term . on one hand , when the electric field is small , the ionized - impurity scattering , whose strength decreases slightly with the electron temperature @xmath147 , is dominant . when the electric field is further increased , @xmath147 and therefore the electron - ac phonon scattering is raised . if the impurity scattering is not too high , the electron - ac phonon scattering can then be dominant as discussed decades ago in ref . . these can be seen from the mobilities @xmath152 obtained from our calculation , which are plotted in the same figure for all the corresponding cases . one can see that @xmath153 increases slightly and monotonically with @xmath91 for the impurity - scattering - dominant case such as @xmath142 ; it decreases monotonically with @xmath91 for the case of very low / no impurity scattering such as @xmath154 , as the electron - ac phonon scattering always increases with the electron temperature @xmath147 ; for the case of low impurity scattering such as @xmath155 , @xmath153 first increases slightly then decreases with @xmath91 which shows the transition from the impurity - scattering - dominant regime to the electron - ac - phonon - scattering - dominant regime@xcite ( unless the runaway effect blocks the system to the later regime as shown in fig . 4(a ) for the case of low electron densities ) . on the other hand , electrons are driven to the higher momentum states by the electric field and experience a larger effective magnetic field [ eqs . ( [ omegax]-[omegaz ] ) ] . therefore the inhomogeneous broadening is increased . this tends to decrease the srt . in order to show the electric field dependence of the inhomogeneous broadening , we plot in fig . 5 the electron temperature @xmath147 as a function of electric field @xmath91 for all the corresponding cases in figs . 4(a ) and ( b ) . it is obvious from fig . 5 that the increase of @xmath147 for the low electron density case is much faster than that for the high one . therefore , the increase of the inhomogeneous broadening is the leading contribution in comparison with the electric field effect on the scattering . consequently , the srt decreases with @xmath91 for the case of low electron densities as shown in fig . for the case of high electron densities , when the impurity density is high such as @xmath144 in fig . 4(b ) , both the slight decrease of the scattering and the increase of the inhomogeneous broadening tend to suppress the srt . when the impurity density is low / zero , the strengthening of the scattering is dominant ( as shown in fig . 4(b ) the decrease of mobility with @xmath91 ) in comparison with the increase of the inhomogeneous broadening . this makes the srt decrease with @xmath91 . finally we point out that the coulomb scattering plays an essential role in the spin r / d in the presence of the electric field . it determines the hot - electron temperature @xmath147 which controls the inhomogeneous broadening and the scattering strengths . moreover , the coulomb scattering itself also contributes to the spin r / d . for comparison , we also plot the srt versus electric field at high temperature @xmath156 k for both low and high electron densities with @xmath143 in fig . at this temperature , the electron - lo phonon scattering is dominant . therefore @xmath153 always decrease with @xmath91 . for the well width and electron density we study , the linear dp term is dominant and the increase of the scattering is more important . therefore the srt increases with @xmath91 . for qws with larger well width so that the cubic term is dominant , the srt can decrease with @xmath91 as reported in our previous work at high temperatures.@xcite in summary , we have investigated the temperature dependence of the srt for @xmath0-type gaas ( 001 ) qws with small well widths from a full microscopic approach by constructing and numerically solving the kinetic spin bloch equations with all the relevant scattering explicitly included . in contrast to our previous studies at high temperatures ( @xmath27 k ) , we include the electron - ac phonon scattering which is absent in our previous studies so that we may extend the scope of our approach to the low temperature regime ( @xmath140 k ) . good agreement with experiment data@xcite is obtained from our theory over almost the _ whole _ temperature regime by using only one fitting parameter @xmath22 whose value agrees with many experimental and theoretical results . we show that the coulomb scattering plays an essential role in spin r / d over all the temperature regime . for qws with low electron densities but high mobility ( _ i.e. _ , low impurity density ) , the spin r / d is mainly controlled by the electron - electron coulomb scattering when @xmath157 k. we predict a peak in the @xmath90-@xmath96 curve . the closer the peak approaches the high temperature limit , the smoother the peak appears . after the peak , the srt increases with temperature . finally , such a peak disappears at sufficient high electron density where the srt increases monotonically with temperature . we point out that the peak origins from the coulomb scattering . specifically , it origins from the different temperature dependences of the coulomb scattering at the degenerate and the nondegenerate limits with the transition temperature @xmath158 . for low electron densities , @xmath159 k where the electron - phonon scattering is negligible . then one may observe an abrupt peak around @xmath132 . for medium electron densities , @xmath160 k@xmath161 k where the increase of the electron - ac phonon scattering partially compensates the decrease of the coulomb scattering when @xmath96 increases , one may observe a smooth peak around @xmath132 . nevertheless , for high electron densities , @xmath162 k , the increase of the electron - phonon scattering completely compensates the decrease of the coulomb scattering when @xmath96 rises . consequently the peak disappears . at high temperature ( @xmath116 k ) and low impurity density , when the well width is small so that the cubic terms in the dp terms are unimportant , the increase of electron - lo phonon scattering surpasses the increase of inhomogeneous broadening with temperature , so that the srt increases with temperature . however , when the impurity density is so high that electron - impurity scattering is the dominant scattering mechanism , the srt decreases monotonically with temperature for any electron density . this is because the temperature dependence of the electron - impurity scattering is very weak and the increase of the inhomogeneous broadening with temperature dominates the temperature dependence of the srt . we also show that larger well width leads to a slower spin relaxation . moreover , in the strong scattering limit , higher impurity density also leads to a slower spin relaxation . both effects can make the srt as long as nanoseconds at very low temperatures . the effect of electric field ( _ i.e. _ , the hot electron effect ) on the spin relaxation is also investigated . we show that the electric field dependence of the srt at low temperature appears again quite differently from that at high temperature due to the absence of electron - lo phonon scattering . moreover , we further show different electric field dependences of the srt at low and high electron densities . at low electron densities , the srt _ decreases _ with the electric field . when the electron density is high , it _ decreases / increases _ with the electric field for the case of high / low impurity densities . these features are in giant difference from the high temperature case where the srt _ increases _ monotonically with electric field for the same qws . more experiments are needed to explore the predictions presented in this manuscript . this work was supported by the natural science foundation of china under grant nos . 90303012 and 10574120 , the natural science foundation of anhui province under grant no . 050460203 , the knowledge innovation project of chinese academy of sciences and srfdp . the authors would like to thank i. c. da cunha lima for his critical reading of this manuscript to improve the english . @xmath163\bigr\ } -\bigl\{{\bf k } \leftrightarrow { \bf k - q}\bigr\}\ , \\ \left . \frac{\partial \rho_{\mathbf{k}}}{\partial t}\right \|_{\mbox{\small ac}}&=&\bigl\ { \pi\sum_{\mathbf{q}q_z\lambda}g^2_{\mathbf{q}q_z\lambda } \delta(\varepsilon_{\mathbf{k}}-\varepsilon_{\mathbf{k}-\mathbf{q } } -\omega_{\mathbf{q}q_z\lambda } ) [ \rho_{\mathbf{k}-\mathbf{q } } ( f_{\mathbf{k},\uparrow}+f_{\mathbf{k},\downarrow } ) + ( f_{\mathbf{k}-\mathbf{q},\uparrow}+f_{\mathbf{k}-\mathbf{q},\downarrow}-2 ) \rho_{\mathbf{k}}\nonumber\\ & & -2n(\epsilon_{\bf k}-\epsilon_{{\bf k}-{\bf q } } ) ( \rho_{\mathbf{k}}-\rho_{\mathbf{k}-\mathbf{q } } ) ] \bigr\ } -\bigl\{\mathbf{k}\leftrightarrow \mathbf{k}-\mathbf{q}\bigr\}\ , \end{aligned}\ ] ] with @xmath164 and @xmath165 standing for the same terms as the previous @xmath166 but with the interchange @xmath167 . @xmath168^{-1}$ ] represents the bose distribution . the division of the truncated two - dimensional momentum space is all the same as our previous work ( see fig . 8 in ref . ) . the two dimensional momentum space is thus divided into @xmath169 control regions , each with the same energy and angle intervals . the @xmath170-grid point of each control region is chosen to be the center of the region : @xmath171 with @xmath172 and @xmath173 . @xmath174 and @xmath175 . with the truncation energy @xmath176 and @xmath177 . unlike the electron - lo phonon scattering where the @xmath178-function in the scattering is used to carry out the integral of @xmath179 , more specifically @xmath180 , with @xmath181 , here the @xmath178-function is used to perform the integral of @xmath182 with @xmath183 , eds . d. d. awschalom , d. loss , and n. samarth ( springer , berlin , 2002 ) ; i. uti , j. fabian , and s. das sarma , rev . mod . phys . * 76 * , 323 ( 2004 ) . j. m. kikkawa and d. d. awschalom , phys . lett . * 80 * , 4313 ( 1998 ) . r. i. dzhioev , b. p. zakharchenya , v. l. korenev , d. gammon , and d. s. katzer , pisma zh . fiz . * 74 * , 200 ( 2001 ) [ jetp lett . * 74 * , 182 ( 2001 ) ] ; r. i. dzhioev , k. v. kavokin , m. v. lazarev , b. ya . meltser , m. n. stepanova , b. p. zakharchenya , d. gammon , and d. s. katzer , phys . b * 66 * , 245204 ( 2002 ) . t. c. damen , l. via , j. e. cunningham , j. shah , and l. j. sham , phys . . lett . * 67 * , 3432 ( 1991 ) . j. wagner , h. schneider , d. richards , a. fischer , and k. ploog , phys . b * 47 * , 4786 ( 1993 ) . a. p. heberle , w. w. rhle , and k. ploog , phys . lett . * 72 * , 3887 ( 1994 ) . s. a. crooker , j. j. baumberg , f. flack , n. samarth , and d. d. awschalom , phys . lett . * 77 * , 2814 ( 1996 ) ; s. a. crooker , d. d. awschalom , j. j. baumberg , f. flack , and n. samarth , phys . b * 56 * , 7574 ( 1997 ) . j. m. kikkawa , i. p. smorchkova , n. samarth , and d. d. awschalom , science * 277 * , 1284 ( 1997 ) . y. ohno , r. terauchi , t. adachi , f. matsukura , and h. ohno , phys . rev * 83 * , 4196 ( 1999 ) . y. ohno , r. terauchi , t. adachi , f. matsukura , and h. ohno , physica e ( amsterdam ) * 6 * , 817 ( 2000 ) . a. malinowski , r. s. britton , t. grevatt , r. t. harley , d. a. ritchie , and m. y. simmons , phys . b * 62 * , 13034 ( 2000 ) ; m. a. brand , a. malinowski , o. z. karimov , p. a. marsden , r. t. harley , a. j. shields , d. sanvitto , d. a. ritchie , and m. y. simmons , phys . rev . lett . * 89 * , 236601 ( 2002 ) . t. adachi , y. ohno , f. matsukura , and h. ohno , physica e ( amsterdam ) * 10 * , 36 ( 2001 ) . o. z. karimov , g. h. john , r. t. harley , w. h. lau , m. e. flatt , m. henini , and r. airey , phys . rev . lett . * 91 * , 246601 ( 2003 ) . s. dhrmann , d. hgele , j. rudolph , m. bichler , d. schuh , and m. oestreich , phys . * 93 * , 147405 ( 2004 ) . l. lombez , p. -f . braun , h. carrre , b. urbaszek , p. renucci . t. amand , x. marie , j. c. harmand , and v. k. kalevich , appl . lett . * 87 * , 252115 ( 2005 ) . j. strand , x. lou , c. adelmann , b. d. schultz , a. f. isakovic , c. j. palmstr@xmath184 m , and p. a. crowell , phys . b * 72 * , 155308 ( 2005 ) . a. m. tyryshkin , s. a. lyon , w. jantsch , and f. schffler , phys . lett . * 94 * , 126802 ( 2005 ) . dyakonov and kachorovskii , fiz . tekh . poluprovodn . v 20 , p 178 ( 1986 ) [ sov . semicond . * 20 * , 110 ( 1986 ) ] . x. cartorix , l. -w . wang , d. z. -y . ting , y. -c . chang , phys . b * 73 * , 205341 ( 2006 ) . w. h. lau and m. e. flatt , phys . b * 72 * , 161311(r ) ( 2005 ) . w. h. lau , j. t. olesberg , and m. e. flatt , phys . b * 64 * , 161301(r ) ( 2001 ) ; p. h. song and k. w. kim , phys . rev . b * 66 * , 035207 ( 2002 ) ; f. x. bronold , i. martin , a. saxena , and d. l. smith , phys . rev . b * 66 * , 233206 ( 2002 ) ; n. s. averkiev , l. e. golub , and m. willander , j. phys . : condens . matter * 14 * , r271 ( 2002 ) ; s. krishnamurthy , m. van schilfgaarde , and n. newman , appl . lett . * 83 * , 1761 ( 2003 ) ; f. x. bronold , a. saxena , and d. l. smith , phys . rev . b * 70 * , 245210 ( 2004 ) ; s. w. chang and s. l. chuang , phys . b * 72 * , 115429 ( 2005 ) ; z. g. yu , s. krishnamurthy , m. van schilfgaarde , and n. newman , phys . rev . b * 71 * , 245312 ( 2005 ) ; o. bleibaum , phys . rev . b * 69 * , 205202 ( 2004 ) ; * 71 * , 235318 ( 2005 ) ; x. cartoix , d. z. y. ting , and y. c. chang , phys . b * 71 * , 045313 ( 2005 ) . m. w. wu and h. metiu , phys . rev . b * 61 * , 2945 ( 2000 ) ; m. w. wu , j. supercond . * 14 * , 245 ( 2001 ) . m. w. wu and c. z. ning , eur . j. b * 18 * , 373 ( 2000 ) . m. w. wu , j. phys . . jpn . * 70 * , 2195 ( 2001 ) . m. w. wu and c. z. ning , phys . sol.(b ) * 222 * , 523 ( 2000 ) ; m. w. wu and m. kuwata - gonokami , solid state commun . * 121 * , 509 ( 2002 ) . j. l. cheng , m. q. weng , and m. w. wu , solid state commun . * 128 * , 365 ( 2003 ) . m. q. weng and m. w. wu , phys . b * 66 * , 235109 ( 2002 ) ; j. appl . phys . * 93 * , 410 ( 2003 ) ; m. q. weng , m. w. wu , and q. w. shi , phys . b * 69 * , 125310 ( 2004 ) ; l. jiang , m. q. weng , m. w. wu , and j. l. cheng , j. appl . phys . * 98 * , 113702 ( 2005 ) . m. q. weng and m. w. wu , phys . b * 68 * , 075312 ( 2003 ) ; * 71 * , 199902(e ) ( 2005 ) ; chin . phys * 22 * , 671 ( 2005 ) . m. q. weng , m. w. wu , and l. jiang , phys . rev . b * 69 * , 245320 ( 2004 ) . m. q. weng and m. w. wu , phys . rev . b * 70 * , 195318 ( 2004 ) . l. jiang and m. w. wu , phys . b * 72 * , 033311 ( 2005 ) . c. l , j. l. cheng , and m. w. wu , phys . rev . b * 73 * , 125314 ( 2006 ) . j. l. cheng and m. w. wu , j. appl . phys . * 99 * , 083704 ( 2006 ) . h. haug and a. p. jauho , _ quantum kinetics in transport and optics of semiconductor _ ( spinger - verlag , berlin , 1996 ) . p. vogl , in _ physics of nonlinear transport in semiconductor _ , edited by k. ferry , j. r. barker , and c. jacoboni ( plenum , new york , 1980 ) . g. d. mahan , in _ polarons in ionic crystals and polar semiconductor _ , edited by j. t. devreese ( north - holland , amsterdam , 1972 ) . , landolt - brnstein , new serious , vol . 17a , ed . by o. madelung ( springer , berlin , 1987 ) . it is noted that the random phase approximation is valid for all the electron densities we discuss in the present paper according to the criterion given by m. jonson , j. phys . c * 9 * , 3055 ( 1976 ) . h. haug and s. w. koch , _ quantum theory of the optical and electronic properties of semiconductors _ ( world scientific , singapore , 2004 ) . , edited by otfried madelung ( springer , berlin , 1996 ) , p.p . 151 ; e. t. yu , j. o. mccaldin , and t. c. mcgill , solid state phys . * 46 * , 1 ( 1992 ) . x. l. lei , j. l. birman , and c. s. ting , j. appl . phys . * 58 * , 2270 ( 1985 ) . m. lundustrom , _ fundamentals of the carrier transport _ ( cambridge university press , cambridge , england , 2000 ) ; j. singh , _ physics of the semiconductors and their heterostructures _ ( mcgraw - hill , new york , 1993 ) . a. g. aronov , g. e. pikus , and a. n. titkov , zh . fiz . * 84 * , 1170 ( 1983 ) [ sov . jetp * 57 * , 680 ( 1983 ) ] . v. a. marushchak , m. n. stepanova , and a. n. tikov , fiz . tverd . ( donetsk ) * 25 * , 3537 ( 1983 ) . b. jusserand , d. richards , g. allen , c. priester , and b. etienne , phys . b * 51 * , 4707 ( 1994 ) . a. t. gorelenko , b. a. marushchak , and a. n. tikov . nauk sssr , ser . fiz . * 50 * , 290 ( 1986 ) . m. cardona , n. e. christensen , and g. fasol , phys . b * 38 * , 1806 ( 1987 ) . a. n. chantis , m. van schilfgaarde , and t. kotani , phys . lett . * 96 * , 086405 ( 2006 ) . it is difficult to give an exact analytical expression of @xmath115 as we take into account of the coulomb scattering to all orders as well as the counter effect of the coulomb scattering to the inhomogeneous broadening . however , one can roughly estimate @xmath115 by considering the coulomb scattering to the lowest order ( 2nd order ) and further ignoring the counter effect in the degenerate and non - degenerate limits following the approach in ref . . p. y. yu and m. cardona , _ fundamentals of semiconductors _ ( springer , berlin , 2003 ) , 3rd ed . , p. 222 . m. m. glazov and e. l. ivchenko , zh . ksp . . fiz . * 126 * 1465 ( 2004 ) [ jetp * 99 * , 1279 ( 2004 ) ] . m. m. glazov and e. l. ivchenko , pisma zh . teor . fiz . * 75 * , 476 ( 2002 ) [ jetp lett . * 75 * , 403 ( 2002 ) ] . in the weak scattering regime , the coulomb scattering can also reduce the spin r / d time.@xcite private communication with r. t. harley . it is stressed here that unlike the high temperature case where the preparation of the steady - state initial electron distributions under strong electric field proposed in ref . is not essential as the time scale to reach the steady state after turning on the electric field is negligible in comparison with the time scale of the spin r / d , at low temperature case due to the weak electron - ac phonon scattering , the preparation is essential as both time scales can be comparable to each other . a. p. dmitriev , v. yu kachorovskii , m. s. shur , and m. stroscio , solid state commun . * 113 * , 565 ( 2000 ) ; a. p. dmitriev , v. yu kachorovskii , and m. s. shur , j. appl . phys . * 89 * , 3793 ( 2001 ) . e. m. conwell , _ high field transport in semiconductors _ ( academic press , new york and london , 1967 ) , p. 21 .	we perform a full microscopic investigation on the spin relaxation in @xmath0-type ( 001 ) gaas quantum wells with al@xmath1ga@xmath2as barrier due to the dyakonov - perel mechanism from nearly 20 k to the room temperature by constructing and numerically solving the kinetic spin bloch equations . we consider all the relevant scattering such as the electron acoustic - phonon , the electron longitudinal - optical - phonon , the electron nonmagnetic - impurity and the electron - electron coulomb scattering to the spin relaxation . the spin relaxation times calculated from our theory with a fitting spin splitting parameter are in good agreement with the experimental data by ohno _ et al . _ [ physica e * 6 * , 817 ( 2000 ) ] over the whole temperature regime ( from 20 k to 300 k ) . the value of the fitted spin splitting parameter agrees with many experiments and theoretical calculations . we further show the temperature dependence of the spin relaxation time under various conditions such as electron density , impurity density and well width . we predict a peak solely due to the coulomb scattering in the spin relaxation time at low temperature ( @xmath3 k ) in samples with low electron density ( _ e.g. _ , density less than @xmath4 @xmath5 ) but high mobility . this peak disappears in samples with high electron density ( _ e.g. _ @xmath6 @xmath5 ) and/or low mobility . the hot - electron spin kinetics at low temperature is also addressed with many features quite different from the high temperature case predicted .
the most important physical processes determining star formation are turbulence , gravity , and magnetic fields ( see the reviews by * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . on one hand , molecular cloud turbulence because it is supersonic compresses interstellar gas in shocks and filaments . if a critical amount of mass is swept up in a local compression , that gas can become gravitationally unstable and collapse , giving birth to new stars . on the other hand , turbulence and magnetic fields together carry about the same amount of energy as there is in gravitational binding energy in a whole molecular cloud ( e.g. , * ? ? ? thus , both turbulence and magnetic fields provide large - scale support against global cloud collapse . that support is needed , because otherwise molecular clouds would simply collapse as a whole , which is not observed @xcite . the role of gravity and magnetic pressure is unidirectional : gravity always acts to promote collapse , while magnetic pressure always counteracts collapse . the role of turbulence on the other hand is dual : it provides global support , but at the same time produces local compression the seeds for stellar birth . hence , turbulence , gravity , and magnetic fields are crucial for our understanding of star formation . these three physical effects are likely the major players in controlling the star formation rate ( sfr ) in the milky way and potentially in other galaxies , as discussed in ( * ? ? ? * hereafter paper i ) . there , we compared theoretical and numerical models of the sfr with observations of galactic clouds , finding very good agreement for a wide range of physical parameters , including extreme cases of turbulent forcing ( solenoidal versus compressive forcing of the turbulence ) , sonic mach numbers between 3 and 50 , and magnetic fields ranging from the super - alfvnic to the trans - alfvnic regime , all in the range of parameters observed in real molecular clouds . despite this large variety of physical parameters , the conversion of gas into stars is always quite inefficient in molecular clouds , with sfes of typically only a few percent @xcite . however , it is still poorly understood why this is the case . it is the aim of this study to shed light on this subject by providing a detailed analysis of the sfe in the simulations of paper i , and comparing them to observations of galactic clouds . the sfe of clouds is often uncertain or unknown . the main difficulty in estimating the sfe lies in obtaining a clear census of ysos , in particular in dense and confused regions where many protostellar disks are embedded and closely packed , such that the yso population is often underestimated in observations . there is some indication that the sfe ranges from less than 1% up to 50% in most clouds , depending on the particular cloud studied , its physical parameters , the actual subregion considered inside that cloud , and the evolutionary stage of that region . this is a very wide range of sfes . the aim of this paper is to advance our understanding of the physical processes that determine this range and provide measures to constrain the sfe for a particular cloud region observed . for instance , considered as a whole , giant molecular clouds typically only convert at most a few percent of their gas into stars ( e.g. , * ? ? ? * ; * ? ? ? in contrast , when looking at regions of dense core and cluster formation , observers find higher sfes ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , eventually approaching the local efficiency @xmath13 on scales of individual protostellar cores . this local efficiency @xmath13 determines the gas fraction of a single protostellar accretion envelope that actually ends up on the protostar(s ) forming in a dense core although gas is efficiently falling onto the protostar , jets , winds and outflows launched from the protostellar disk counteract the inflow , thereby limiting accretion such that @xmath14 . we estimated this local efficiency with by comparing our simulations with galactic observations by @xcite in paper i. it is thus plausible that the sfe naturally approaches the local efficiency @xmath13 on sufficiently small scales , when a single , dense , star - forming core is considered . here , we investigate the dependence of the sfe on molecular cloud scales and on physical parameters , such as the driving of turbulence , the sonic and alfvn mach numbers , and the virial parameter of a cloud . first , we study the structure of the column density and how it changes with @xmath12 in section [ sec : morphology ] . we then discuss the dependences of the probability density functions ( pdfs ) of the volumetric and column densities in section [ sec : pdfs ] . we find that the pdfs depend on the forcing of the turbulence and on the sonic and alfvn mach numbers . the pdfs develop power - law tails of flatting slope in all our numerical models when star formation proceeds , as seen in observations of clouds @xcite , suggesting that the pdf can be used to infer the sfe . in section [ sec : scaling ] , however , we find that a more reliable and independent distinction between star - forming and quiescent clouds is possible by studying the spatial scaling of column density with fourier or @xmath4-variance analyses @xcite , i.e. , by studying the density size scaling in molecular clouds . we extend this qualitative discriminator of star formation to a quantitative method for estimating the sfe in section [ sec : estimatingsfe ] , by measuring how the slope @xmath2 of the density power spectrum changes with @xmath12 , depending on the forcing of turbulence , the sonic mach number , and the magnetic field . inverting this relation yields a function @xmath15 , which can be used to estimate @xmath12 from a measurement of @xmath2 in a dust or integrated molecular line map . the advantage of this new method is that no a priori knowledge of star formation activity or yso counts is required to estimate sfe . in section [ sec : obs ] we apply this method to observations and compare inferred sfes with independent estimates , indicating good agreement . in section [ sec : uncertainties ] , we discuss uncertainties and limitations of the new method . our conclusions are summarized in section [ sec : conclusions ] . our numerical simulation techniques are explained in detail in paper i. here we only give a brief overview of the most important aspects of the simulation methodology and provide a complete list of simulation parameters as in paper i. we use the adaptive mesh refinement ( amr , * ? ? ? * ) code flash v2.5 @xcite to model isothermal , self - gravitating , magnetohydrodynamic ( mhd ) turbulence on three - dimensional ( 3d ) , periodic grids with resolutions of @xmath16@xmath17 grid points . these are all uniform - grid simulations , except for one @xmath18 simulation , where we use a root grid with @xmath19 cells and one level of amr with a refinement criterion to ensure that the local jeans lengths is covered with at least 32 grid cells , in order to resolve turbulent vorticity and magnetic - field amplification on the jeans scale @xcite . for solving the mhd equations , we use the hll3r positive - definite riemann solver @xcite . the mhd equations are closed with an isothermal equation of state , which is a reasonable approximation for dense , molecular gas of solar metallicity , over a wide range of densities @xcite . the self - gravity of the gas is computed with a multi - grid poisson solver ( see * ? ? ? to drive turbulence in the simulations , we apply a stochastic acceleration field @xmath20 as a momentum and energy source term . @xmath20 only contains large - scale modes , @xmath21 , where most of the power is injected at the @xmath22 mode in fourier space , which corresponds to half of the box size @xmath23 in physical space . we thus model turbulent forcing on large scales , as favored by molecular cloud observations ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? smaller scales , @xmath24 , are not affected directly by the forcing , such that turbulence can develop self - consistently there . we use the ornstein - uhlenbeck ( ou ) process to model @xmath20 , which is a well - defined stochastic process with a finite autocorrelation timescale @xcite . we set the autocorrelation time equal to the turbulent crossing time on the largest scales of the system , @xmath25 , with the rms sonic mach number @xmath26 and the constant sound speed @xmath27 , leading to a smoothly varying stochastic forcing in space and time ( for details , see * ? ? ? * ; * ? ? ? * ; * ? ? ? * paper i ) . we can adjust the mixture of solenoidal and compressive modes of our turbulent forcing , by applying a projection in fourier space . here we compare simulations with solenoidal forcing ( @xmath28 ) and compressive forcing ( @xmath29 ) , as well as an intermediate mixture of both . solenoidal and compressive forcing are extreme cases , while in real molecular clouds we expect some mixture with a range of possible ratios between solenoidal and compressive modes ( see the discussion of physical drivers of turbulence and their mode characteristics in paper i ) . for instance , @xcite describe a scenario in which the observed alignment of young protostars in the @xmath30 ophiuchi central region could have been triggered by a shock compression , induced by a supernova / wind shell and/or expanding @xmath31 regions from nearby ob associations . such expanding shells are compressive forcing mechanisms , because they primarily excite compressible velocity modes . in order to model collapse and accretion , we use an advanced amr - based approach for sink particles , in which only bound and collapsing gas is accreted ( for a detailed analysis , see * ? ? ? the key feature of this approach is to define a control volume centered on grid cells exceeding a density threshold . @xcite found that the jeans length must be resolved with at least 4 grid cells to avoid artificial fragmentation , leading to a resolution - dependent density threshold criterion for triggering sink particle creation checks , @xmath32 where the sink particle accretion radius @xmath33 is set to 2.5 grid - cell lengths at the maximum level of refinement , corresponding to half a jeans length at @xmath34 . this guarantees that the jeans length is still resolved with 5 grid cells prior to potential sink particle creation to avoid artificial fragmentation . grid cells exceeding the density threshold given by equation ( [ eq : rhosink ] ) , however , do not form sink particles right away . first , a spherical control volume with radius @xmath33 is defined around the cell exceeding @xmath34 , in which a series of checks for gravitational instability and collapse are performed . if all checks are passed , a sink particle is created in the center of the control volume . this procedure avoids spurious sink particle formation , and allows us to trace only truly collapsing and star - forming gas . once a sink particle is created , it can gain mass by accreting gas from the amr grid , but only if this gas exceeds the threshold density , is inside the sink particle accretion radius , is bound to the particle , and is collapsing toward it . if all these criteria are fulfilled , the excess mass above the density threshold defined by equation ( [ eq : rhosink ] ) is removed from the mhd system and added to the sink particle , such that mass , momentum and angular momentum are conserved by construction ( see * ? ? ? * ; * ? ? ? * for details ) . sink particle gravity is computed by direct @xmath35-body summation over all sink particles and grid cells . we use a second - order accurate leapfrog integrator to advance the sink particles on a timestep that allows us to resolve close and highly eccentric orbits of sink particles without introducing significant errors on super - resolution grid scales , as tested in @xcite . [ cols="<,>,^,^,^,^,^,^,^,^,^,^,^,^,^ " , ] * notes . * the upper and lower limits of the sfe quoted in the last column were estimated based on comparing the range of spectral slopes @xmath2 ( including the maximum uncertainties in equation [ eq : fitinverse ] combined with the additional fit uncertainties for @xmath2 from the observations ; see discussion in section [ sec : uncertainties ] ) with the numerical simulations in figure [ fig : slopes ] , covering a wide range of physical conditions ( with solenoidal and compressive forcing , and different magnetic field strengths ) . equation ( [ eq : fitinverse ] ) together with table [ tab : fits ] provide direct relations @xmath15 fitted to the simulation data in figure [ fig : slopes ] . references : ( 1 ) fourier analysis by @xcite ; ( 2 ) reanalysis of @xmath4-variance estimates in @xcite ; ( 3 ) reanalysis of @xmath4-variance estimates in @xcite ; ( 4 ) @xmath4-variance by @xcite ; ( 5 ) @xmath4-variance by @xcite . the first entry in table [ tab : obs ] is from a fourier analysis of the @xmath36 line of atomic hydrogen of the high galactic latitude ursa major ` cirrus ' by @xcite . they find a spectral slope @xmath37 . comparing this value , including the uncertainties in both simulations ( figure [ fig : slopes ] and equation [ eq : fitinverse ] ) and observations ( table [ tab : obs ] ) , we estimate that the ursa major cirrus has an @xmath7 . this is because even the maximum slope , @xmath38 is significantly negative and any of our numerical simulations did not produce stars for @xmath39 . although molecular condensations have been detected in the ursa major cirrus , it appears to have essentially no star formation activity , likely because of the high virial parameters estimated for this cloud @xcite . this is in good agreement with our findings of @xmath7 , based on the comparison of the observed slope @xmath2 with our numerical simulations . entries 25 , 7 , and 12 in table [ tab : obs ] are for observations of the polaris flare by @xcite and @xcite with different instruments ( iram , kosma , fcrao , cfa ) and different molecular transitions of @xmath40 and @xmath41 . we see some variation of the derived slope @xmath42 to @xmath43 , depending on the telescope and tracer used , as well as on the spatial scales over which the slope was determined @xcite . thus , we reanalyzed the @xmath4-variance data in @xcite , fitting average power laws over a wider range of scales than considered in @xcite , in order to obtain a rough estimate of the slope @xmath2 , less affected by the particular choice of scales for the fit . we find that our reanalysis mostly agrees with the estimates of @xmath2 in @xcite . we see that the overall slope @xmath2 in the polaris flare is of order @xmath44 , from which we estimate @xmath7 based on the model fits of figure [ fig : slopes ] and equation ( [ eq : fitinverse ] ) in the polaris flare , with a formal upper limit of @xmath45 . this is in good agreement with observations of the polaris flare , showing no sign of star formation activity , although the saxophone region does seem to indicate some trace of gravitational contraction ( n. schneider 2012 , private communication ) , which , however , has not led to any star formation in the cloud yet . entries 6 , 8 , and 9 are for dust extinction and co observations of vela , ngc 1333 in perseus , and the coalsack by @xcite and @xcite . the values of @xmath46 to @xmath47 were estimated based on average fits to the @xmath4-variance plots in @xcite and @xcite , taking into account a broad range of scales and all the error bars of those analyses . the purpose of our reanalysis is to obtain general trends of the slope @xmath2 , without constraining the fits to relatively small scales as done in @xcite and @xcite . this is why we find slightly different slopes here compared to their small - scale fits . these values of @xmath2 indicate @xmath7 with an upper limit of @xmath48 in vela and ngc 1333 , and @xmath45 in coalsack . the latter is in agreement with an independent estimate of star formation activity in coalsack , suggesting relatively low sfe @xcite . estimates of @xmath2 for the following clouds and cloud regions ( entries 10 , 11 , and 1324 ) range from @xmath49 to @xmath50 , suggesting sfes in the range @xmath7 to a maximum of 6% , when comparing to the range of sfes obtained in the simulations for such @xmath2 ( see figure [ fig : slopes ] , equation [ eq : fitinverse ] , and table [ tab : fits ] ) . those clouds include lupus , rosette , pipe , corona australis , and perseus , which indeed seem to be consistent with low to moderate star formation activity ( compare , e.g. , table 1 in * ? ? ? a notable exception may seem orion a+b , which contain a large number of ysos , indicating high star formation activity . in a recent survey of orion a+b , @xcite indeed detected 3479 ysos . however , orion a+b are also quite massive with @xmath51 @xcite . assuming that ysos produce stars with a mean mass of about @xmath52 ( which is in between the mean masses of the present - day stellar mass function , @xmath53 , and the initial mass function , @xmath54 , for disk galaxies ; see * ? ? ? * ) , we can estimate the sfe of orion a+b with @xmath55 . a similar estimate based on yso counts and assuming a median mass for each yso of @xmath56 is provided in @xcite . from their table 2 , we can compute @xmath57 with an assumed star formation timescale @xmath58 @xcite . using the values quoted for the low extinction threshold , @xmath59 , covering almost the whole cloud , we obtain @xmath60 for orion a and @xmath61 for orion b , consistent with our own estimate above . those sfes obtained from yso counts , however , are likely lower limits , because of the limited detectability of ysos on small spatial scales in crowed , dense regions . the yso completeness achieved with _ spitzer _ may only be 50% averaged over the inner @xmath62 of the trapezium region in orion , and only 30% within a radius of @xmath63 @xcite . thus , sfe estimates based on yso counts can be quite uncertain . however , they do fall in the range that we obtain for orion a+b by using the slope @xmath2 listed in table [ tab : obs ] . thus , even if no information on ysos is available for a given cloud , we can still get a lower and upper limit of the sfe from the slope @xmath2 , which is directly measurable from the column density map of a cloud alone . taurus , w3 , and chamaeleon ( entries 2527 ) all have @xmath64 , indicating . for instance , taurus has a total molecular mass of about @xmath65 @xcite , and more than 300 ysos were detected ( see table 1 in * ? ? ? * ; * ? ? ? similar to our estimate for orion a+b above , we find @xmath66 . another estimate based on yso counts in @xcite yields @xmath67 ( again using their low extinction threshold value to define @xmath68 ) . both are consistent with our formal upper limit of @xmath69 from @xmath2 , but significantly smaller than the mean , @xmath70 . taurus is indeed somewhat special , in that it may be more dominated by magnetic fields than other clouds . recent observations suggest that turbulence in taurus is trans - alfvnic @xcite , leading to lower sfrs than in super - alfvnic clouds ( see paper i ) . the rather low sfe in taurus may thus be a result of the relatively strong magnetic fields there . our estimate of from the slope @xmath2 is slightly biased toward larger values of @xmath12 , because stronger magnetic fields also produce slightly larger @xmath2 ( see figure [ fig : slopes ] , bottom panel ) . cloud region 28 in table [ tab : obs ] is a @xmath4-variance measurement by @xcite of the cygnus x giant molecular cloud from a high - resolution @xmath41 integrated intensity map . cygnus x is one of the most massive and most active star - forming regions in the galaxy , including both old and young populations of stars @xcite . on well - resolved scales , between @xmath71 and @xmath72 , we find a power - law slope @xmath73 from figure 2 in @xcite . comparing this slope to our simulation models in figure [ fig : slopes ] , we estimate that cygnus x has an sfe in the range . the lower limit of the @xmath12 is formally obtained from the lower limit of @xmath2 , while we actually know that cygnus x is forming stars , so @xmath9 . in particular , the massive dr21 filament in cygnus x shows signs of infall @xcite , which is likely the reason for cygnus x exhibiting a slightly positive @xmath2 on length scales representative of that filament ( length @xmath74 , width @xmath75 ) and sub - filaments collapsing toward the main filament @xcite . the range found for cygnus x is relatively large , but to the best of our knowledge the only currently available estimate of the @xmath12 for the whole cygnus x giant molecular cloud complex . finally , @xcite studied the @xmath4-variance spectrum of the serpens core observed by @xcite . @xcite estimated a slope @xmath76 , which is clearly positive and larger than any of the previous cloud regions studied , indicating @xmath77 ( last entry in table [ tab : obs ] ) . based on the total available gas mass and the mass spectrum of condensations reported in @xcite , we estimate an independent lower limit of the sfe in the serpens core of @xmath78 . recently , @xcite reported in serpens south and w40 . another independent estimate by @xcite suggests an in the region where most protostellar cores and protostars are located . all independent estimates agree with the lower limit , @xmath77 , that we find here based on @xmath2 . it is indeed plausible that the serpens core region has a relatively high @xmath12 , because unlike the large - scale cloud regions studied above the serpens core is a small - scale , high - density region , where eventually we expect a relatively large fraction of the gas to be accreted by stars . the serpens core region is close to the limit where the @xmath12 approaches the local efficiency @xmath79 , i.e. , the mass fraction of gas of an individual dense core that will actually fall onto the protostar . theoretical , numerical , and observational estimates suggest that @xcite . for instance , @xcite find and @xcite find . comparing the sfr and gas column densities in our numerical simulations with galactic observations by @xcite in paper i , we found best - fit values , which seem to suggest typical local efficiencies of @xmath80 , in good agreement with the aforementioned independent estimates . thus , about half of the gas in a dense core is accreted by the protostar , while the other half remains in the envelope or is driven out of the core by winds , jets , and outflows . in section [ sec : estimatingsfe ] we calibrated a simple , empirical relation given by equation ( [ eq : fitinverse ] ) between the spectral slope @xmath2 of the column density and the @xmath12 in the numerical simulations of section [ sec : sims ] . this relation can be applied to observational maps in which @xmath2 can be measured with fourier techniques such as the @xmath4-variance method , to obtain an estimate of the @xmath12 in the observed region . equation ( [ eq : fitinverse ] ) only includes an uncertainty in @xmath81 as suggested by our calibration with numerical simulations . the practical uncertainties in @xmath2 when measured in an observational map , however , can be higher than that . such uncertainties include finite telescope resolution , signal - to - noise , and sensitivity limits for a given observational tracer of column density . for instance , dust emission , extinction , or molecular line emission all suffer from sensitivity limits in both the high and low column - density regimes ( e.g. , molecular lines become optically thick at high column density or detectors are simply not sensitive enough to trace very low column density ) . such statistical and systematic uncertainties are not included in equation ( [ eq : fitinverse ] ) and must be added accordingly . a primary goal of this paper is to understand the transition from the global @xmath12 on large scales ( e.g. , a whole molecular cloud ) to the local core formation efficiency @xmath13 . the latter denotes the fraction of gas in a dense core that eventually goes into the star . this is not 100% , because some fraction , @xmath82 , of the gas in the core does not fall onto the protostar , but is driven out of the core by outflows , winds , and/or jets during the formation of the protostar in the center of the accretion disk . without specifying what mechanisms ( e.g. , radiative and/or magnetic ) cause the gas flow out of the core or how it actually happens , we simply note that we expect a local efficiency ( * ? ? ? * paper i , and references therein ) . however , even the best high - resolution observations of individual cores are just beginning to investigate this local regime of star formation and constrain @xmath13 . in surveys of structure and star formation activity of molecular clouds on intermediate and large scales , such cores appear as sites of high column density , but their internal structure is often not resolved . thus , the actual amount of gas contributing to star formation is uncertain and so is the actual total @xmath12 . we note that we did not take into account the local efficiency @xmath13 in equation ( [ eq : fitinverse ] ) , because the simulations did not include any form of feedback such as winds , jets , or outflows from the protostellar objects . such feedback effects are not accounted for and hence add to the uncertainty in the true @xmath12 when equation ( [ eq : fitinverse ] ) is applied to observational data . given the local efficiencies , it would seem reasonable to adjust the inferred sfe via equation ( [ eq : fitinverse ] ) down by the factor @xmath13 to estimate the actual total sfe of a cloud . in the application of equation ( [ eq : fitinverse ] ) to the observations presented in section [ sec : obs ] , however , we did not make such an adjustment , because we think that the same adjustment would be necessary in the case of yso counts to infer the sfe . this is because the effect of the local efficiency @xmath13 is typically also not accounted for when estimating the typical mass of a yso , since those observations are limited in resolution and likely include a significant fraction of gas / dust in the core that may not all be accreted by the star . thus , the local efficiency @xmath13 due to feedback introduces uncertainties as discussed above , which , however , similarly apply to sfe estimates obtained with yso counts . we furthermore note that the simulations studied here are highly idealized and can suffer from resolution issues . however , the effects of limited resolution are relatively small for our measurement of @xmath2 and we do understand them quite well ( see appendix [ app : spectres ] ) . a more severe issue is the choice of boundary conditions in the simulations . they are periodic in all directions . we discuss the limitations of this and related , different approaches to treat boundaries in paper i. for the present study of the sfe , the main limitation lies in the fact that our computational boxes are closed , so there is no gas inflow or outflow . real molecular clouds are not isolated , closed objects , but exchange mass with the medium in which they are embedded . this poses a general problem and uncertainty in defining the @xmath12 . imagine , for instance , a cloud accreting more gas from the outside than is turned into stars inside in the same amount of time . in that case , the @xmath12 would go down , even though stars form continuously , simply because the total gas content in the region of interest increases due to accretion from the outside . an even more remarkable thought experiment is one in which star formation has stopped at a given @xmath12 in a given volume , and suppose the stars formed in that region now blow away all the remaining gas in their surrounding and drive it out of that volume ( e.g. , by driving winds ) . according to the standard definition of the @xmath12 in equation ( [ eq : sfe ] ) , @xmath83 would remain the same in this process , but @xmath84 would go down , leading to an increasing @xmath12 , even though no more stars form . this is an example emphasizing that one must be quite careful with interpreting sfes in open systems . such definitions of the @xmath12 in open systems are discussed in @xcite and @xcite . we studied the star formation efficiency ( sfe ) in simulations of supersonic , self - gravitating , magnetized turbulence . in paper i , we found that the simulations yield star formation rates ( sfrs ) varying by orders of magnitude , depending on the virial parameter , on the forcing of the turbulence , and on the sonic and alfvn mach numbers , consistent with observed sfr column densities . here , we focused on the sfe and find that the density and column density distributions change significantly when star formation occurs and when the sfe increases . we find that the column density scaling is particularly sensitive to the sfe , enabling us to estimate the sfe from column density observations , without requiring a priori information on star formation activity or yso counts . more detailed results are listed in the following . 1 . we find that star formation affects the column density structure of clouds ( see figures [ fig : imagessol ] and [ fig : imagescomp ] ) . the effect of star formation on the global structure is stronger in low mach number clouds and for clouds with solenoidal forcing than for clouds with high mach number and/or compressive forcing , because strong turbulence alone produces high - density filaments , without requiring gravitational contraction . 2 . the volumetric and column density pdfs become wider with increasing mach number and more compressive forcing ( see figures [ fig : pdfs ] and [ fig : coldenspdfs ] ) . in contrast , increasing the magnetic field strength yields smaller standard deviations of the pdf ( see figure [ fig : magpdfs ] ) , as expected from equation ( [ eq : sigs ] ) . 3 . the pdfs of our simulations with extremely different forcing , sonic mach number , and magnetic field strengths , all develop power - law tails of flattening slope with increasing sfe . we find that the high - density tails of the pdfs are consistent with equivalent radial density profiles , @xmath0 with depending on @xmath12 , consistent with column density observations @xcite . 4 . the velocity spectra from our simulations with different forcing , sonic mach number , and magnetic field are all consistent with the scaling inferred in molecular cloud observations , @xcite , suggesting highly compressible , supersonic turbulence of burgers type on large scales . we find evidence for a transition to subsonic , kolmogorov turbulence with on scales smaller than the sonic scale ( see figure [ fig : velspectra ] ) . the velocity spectra are largely insensitive to star formation . 5 . unlike the velocity spectra , we find a strong change of the density and column density spectra with the onset of star formation . the slope @xmath2 of @xmath3 or equivalently of the @xmath4-variance spectrum of column density , @xmath5 , which are both related to the column density spectrum @xmath85 via equations ( [ eq : ptwodtothreedrelation ] ) and ( [ eq : dvarspectrelation ] ) , switches sign from @xmath86 for @xmath7 to @xmath87 when star formation proceeds , @xmath88 ( see figures [ fig : coldensspectra ] and [ fig : bspectra ] ) . 6 . the change in slope @xmath2 occurs in all our models , despite the wide range of physical cloud conditions probed in the simulations ( with solenoidal , mixed , and compressive forcing of the turbulence , and magnetic fields ranging from the super - alfvnic to the trans - alfvnic regime ; see table [ tab : sims ] ) . we confirmed this result for the slope @xmath2 with an independent measurement based on the @xmath4-variance technique , suitable for molecular cloud observations ( see figures [ fig : deltavar ] and [ fig : bdeltavar ] ) . 7 . studying the detailed dependence of the slope @xmath2 on @xmath12 in all our numerical models , we find that the @xmath12 can be estimated by measuring @xmath2 ( see figure [ fig : slopes ] ) . we provide a fitted relation , @xmath15 in equation ( [ eq : fitinverse ] ) for each numerical model set with best - fit parameters listed in table [ tab : fits ] . we compared observational measurements of @xmath2 in an @xmath89 cloud , in several giant molecular clouds and sub - regions , and in a dense star - forming core , with values ranging from @xmath10 to @xmath11 , all listed in table [ tab : obs ] . we infer the @xmath12 for each cloud or region studied and show that negative @xmath2 typically indicates @xmath7 , consistent with observations . clouds with @xmath90 and larger do form stars and exhibit a range of sfes , which is monotonically increasing with @xmath2 . estimates of @xmath12 from @xmath2 are consistent with independent measurements based on yso counts , where available . however , the @xmath15 relations found here provide an independent estimate of @xmath12 based on a column density map alone , without requiring a priori knowledge of star - formation activity or yso counts for a given cloud . the overall agreement between our simulations and observations of galactic clouds concerning the sfe studied here and the sfr in paper i is promising . we conclude that supersonic , magnetized turbulence is a key process , likely controlling the sfrs and sfes of molecular clouds in the milky way and potentially in other galaxies . we thank nicola schneider , timea csengeri , jrg fischera , jouni kainulainen , and volker ossenkopf for useful discussions on column density pdfs and @xmath4-variance measurements in galactic clouds . the work on pdf tails was partly inspired by discussions with philipp girichidis and lukas konstandin . we thank the anonymous referee for suggesting a critical discussion of the uncertainties in estimating the sfe in observations . c. f. acknowledges the australian research council for a discovery projects fellowship ( grant no . dp110102191 ) . r. s. k. acknowledges subsidies from the baden - wrttemberg - stiftung via contract research grant p - ls - spii/18 . this work was supported by the deutsche forschungsgemeinschaft , priority program 1573 ( `` physics of the interstellar medium '' ) and the collaborative research project sfb 881 ( `` the milky way system '' ) . the simulations presented in this work were run on supercomputers at the leibniz rechenzentrum ( project pr32lo ) and at the forschungszentrum jlich ( project hhd20 ) . the software used here was in part developed by the doe - supported asc / alliance center for astrophysical thermonuclear flashes at the university of chicago . figure [ fig : pdfres ] shows a resolution study of the density pdf @xmath91 for our two mach 10 models with solenoidal and compressive forcing ( left and middle panel ) , and for our mach 50 model with compressive forcing ( right panel ) at @xmath92 and @xmath93 . we compare resolutions @xmath94 , @xmath95 , and @xmath96 for the mach 10 models . we see no significant resolution dependence for @xmath97 in the log - normal part of the pdf , but a clear resolution dependence in the power - law tail , stretching the pdf to higher @xmath98 when the resolution is increased . in the solenoidal forcing case ( left panel ) , we find a break in the power law at @xmath99 , 5.2 , and 6.5 for @xmath94 , 256 , and 512 , respectively . these breaks seem to correlate with the density threshold for sink particle creation , equation ( [ eq : rhosink ] ) , @xmath100 , 4.5 , and 5.9 , shown as vertical short lines on top of the abscissa ( @xmath94 : dotted ; 256 : dashed , and 512 : solid ) , but are slightly higher than that . the pdf for densities @xmath101 seems converged , while the flatter part for which @xmath102 shifts to higher densities when the resolution is increased . the compressive forcing model with mach 10 ( middle panel ) does not show such a break , but a cutoff at large @xmath98 , which shifts to higher @xmath98 with increasing resolution , similar to the cutoff seen in the solenoidal forcing model at large @xmath98 . the pdf of the mach 50 , compressive forcing case ( right panel ) does not show a significant resolution dependence for the range of densities and @xmath103 analyzed here ( @xmath104 : dotted ; 512 : dashed ; and 1024 : solid ) , even though the sink particle creation densities are quite low ( see the vertical lines on top of the abscissa ) . figure [ fig : coldensspectra_res ] shows the same resolution study as in figure [ fig : pdfres ] , but for the density and column density power spectra as in figure [ fig : coldensspectra ] . this resolution study demonstrates that density and column density spectra are converged for @xmath105 . in contrast , velocity spectra are only converged for @xmath106 @xcite . this is because resolving the kinetic energy content of a vortex ( which is a vector quantity ) requires at least 30 grid cells . here we find that about 15 grid cells are sufficient to resolve turbulent and collapsing density structures . this in turn suggests that resolving the jeans length with 15 grid cells would be sufficient to capture the density structures in collapse simulations , but is insufficient to resolve kinetic vorticity and potential small - scale dynamo amplification of magnetic fields ( both of which are vector quantities ) , which still requires at least 30 grid cells per jeans length @xcite . the slopes @xmath2 that we estimate in are obtained in a converged range of scales , and hence these slopes are not affected by numerical resolution . figure [ fig : pdfspectseeds ] ( left ) indicates statistical fluctuations mostly in the low- and high - density tails of the pdfs when different random seeds of the turbulence ( seed1 , seed2 , seed3 ) are considered . this gives us an impression of the statistical uncertainties of the pdfs . the right panel shows the density and column density spectra of the same models , also indicating statistical fluctuations , which manifest in different amplitudes of the spectra , but the power - law slopes do not vary significantly with random seed . , a. , fisher , r. , graziani , c. , et al . 2008 , in astronomical society of the pacific conference series , vol . 385 , numerical modeling of space plasma flows , ed . n. v. pogorelov , e. audit , & g. p. zank , 145	we study the star formation efficiency ( sfe ) in simulations and observations of turbulent , magnetized , molecular clouds . we find that the probability density functions ( pdfs ) of the density and the column density in our simulations with solenoidal , mixed , and compressive forcing of the turbulence , sonic mach numbers of , and magnetic fields in the super- to the trans - alfvnic regime , all develop power - law tails of flattening slope with increasing sfe . the high - density tails of the pdfs are consistent with equivalent radial density profiles , @xmath0 with , in agreement with observations . studying velocity size scalings , we find that all the simulations are consistent with the observed scaling of supersonic turbulence , and seem to approach kolmogorov turbulence with @xmath1 below the sonic scale . the velocity size scaling is , however , largely independent of the sfe . in contrast , the density size and column density size scalings are highly sensitive to star formation . we find that the power - law slope @xmath2 of the density power spectrum , @xmath3 , or equivalently the @xmath4-variance spectrum of the column density , @xmath5 , switches sign from @xmath6 for @xmath7 to @xmath8 when star formation proceeds ( @xmath9 ) . we provide a relation to compute the sfe from a measurement of @xmath2 . studying the literature , we find values ranging from @xmath10 to @xmath11 in observations covering scales from the large - scale atomic medium , over cold molecular clouds , down to dense star - forming cores . from those @xmath2 values , we infer sfes and find good agreement with independent measurements based on young stellar object ( yso ) counts , where available . our relation provides an independent estimate of the @xmath12 based on the column density map of a cloud alone , without requiring a priori knowledge of star - formation activity or yso counts .
despite more than four decades of research , the emission from pulsars and their surrounding nebulae is still poorly understood . two important problems are the origin of pulsar @xmath0-ray emission and , for pulsar wind nebulae ( pwne ) , the so - called sigma - problem ( see the reviews by * ? ? ? * ; * ? ? ? * ; * ? ? ? the latter problem derives its name from the sigma - parameter , @xmath5 , which is the ratio of the energy density due to poynting flux over the particle energy density , with @xmath6 the local magnetic field strength , @xmath7 the lorentz factor of the pulsar wind , and @xmath8 the particle number density . according to theoretical models , most of the pulsar s rotational energy loss rate ( or spin - down power ) @xmath9 is due to poynting flux , i.e. the pulsar wind should have @xmath10 , but observations of pwne indicate that most of the energy that has been lost is actually contained by the relativistic electrons / positrons . somehow , the high @xmath11 flow is converted into kinetic energy somewhere between the pulsar s magnetosphere and the wind termination shock , which converts the radial pulsar wind into an isotropic , relativistic particle distribution @xcite . a third , and perhaps related , problem is the high wind multiplicity factor . the combination of a high magnetic and rapid rotation results in a strong electric potential in the magnetosphere . this potential will be neutralized by charged particles that are stripped from the surface of the neutron star @xcite . the associated charged particle density is @xmath12 with @xmath13 the pulsar s period and @xmath6 the local magnetic field . a fraction of these particles will escape through open field lines , resulting in a particle flux @xmath14 with @xmath15 the dipole surface magnetic field in units of @xmath16 g. however , x - ray ( e.g. * ? ? ? * ) and tev @xcite observations indicate that the number of relativistic electrons contained by pwne turns out to be orders of magnitude larger than @xmath17 , i.e. @xmath18 , with the multiplicity factor being @xmath19 for a young pulsar like b1509 - 58 @xcite . the origin of the additional plasma is likely electron / positron pair production in the magnetosphere . the pair production occurs in the presence of the high magnetic fields inside the magnetosphere , and requires the presence of high energy photons that are either the result of curvature radiation or inverse compton scattering . the electrons that cause the emission are accelerated due to the extremely large voltage drop across the open field lines @xcite . for the inverse compton scattering seed photons are necessary that are emitted by the hot polar caps of the pulsar , heated due to the bombardment by accelerated particles , or due to the cooling of the young neutron star . despite the many unsolved problems , pulsar research has thrived over the last decade thanks to many advances in observational techniques and numerical simulations . in particular high energy observations have contributed to a wealth of new information on pulsars and pwne ; from high spatial resolution x - ray images with _ chandra _ , revealing torii and jets ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , to a rapid progress in the field of tev astronomy ( * ? ? ? * for a review ) , which have revealed an unexpectedly large population of very extended pwne ( e.g. * ? ? ? * ; * ? ? ? this rapid growth in observational data has recently been augmented by the gev @xmath0-ray observatory _ fermi _ , which has greatly increased the sample of @xmath0-ray - emitting pulsars @xcite . here we present a statistical analysis of two samples of rotation powered pulsars . one is those of x - ray pulsars compiled by @xcite , the other the aforementioned sample of _ fermi_-detected pulsars . our analysis concentrates on what determines the radiative efficiency of pulsars and their pwne . we report a surprisingly strong correlation between the x - ray luminosity of pulsars and their pwne , which inversely correlates with characteristic age , at least for young pulsars . in contrast , the @xmath0-ray emission correlates positively with characteristic pulsar age , as already noted by @xcite . it is well known that the non - thermal x - ray luminosities of pulsars and pwne are strongly correlated with the spin - down luminosity @xmath9 of the pulsar @xcite . the efficiency with which the spin - down luminosity is converted into x - ray emission is usually indicated by the symbol @xmath20 with @xmath21 in the range of @xmath22 . an important question is what determines this efficiency for both @xmath0-ray and x - ray emission . for the pwne @xmath21 may provide information on how well the spin - down luminosity is converted into relativistic particles , i.e. it is related to the sigma - problem . for the statistical analysis described here we use the x - ray properties of pulsars and their nebulae as determined by @xcite . this sample is based on _ chandra _ observations . chandra_-acis instrument that was used has a poor timing resolution , so all pulsar luminosities are a combination of pulsed and unpulsed emission . however , given the high spatial resolution of _ chandra _ , the x - ray flux from the pulsar could be accurately separated from the x - ray flux from the pwn . the x - ray luminosities were derived from the 0.5 - 8 kev fluxes , corrected for interstellar absorption and using the distance estimated listed in the paper . @xcite list only the non - thermal x - ray luminosities , so the contribution of thermal x - ray emission , if present , was ignored . the uncertainties in the distance are the largest source of error in the luminosities . in general , the distance estimates may have errors of order @xmath23 , resulting in luminosity errors of order 4 . of course distance errors affect the pulsar and pwne luminosities in the same way . a source of error for the pwne luminosities may be underestimated flux contributions from low surface brightness emission at large radii . indeed , @xcite recently reported the detection of low x - ray surface brightness structures around several pwne by _ suzaku_. the fraction of the luminosity in these low surface brightness is , however , not more than a factor of 2 in luminosity . so all together , the logarithm of the luminosities may contain errors of the order of one decade . we included in our study a statistical analysis of the @xmath0-ray luminosities , based on the pulsed @xmath0-ray emission of pulsar reported by @xcite . we made sure that the @xmath0-ray luminosities of pulsars that were common to both the x - ray and @xmath0-ray samples were based on the same distance estimates , namely those adopted by @xcite . for both the x - ray and @xmath0-ray samples we excluded millisecond pulsars , and omitted pulsars without good distance estimates . in figure [ fig : eta ] we show the correlation of the x - ray luminosity efficiency @xmath21 of pulsars and pwne versus the characteristic spin - down age @xmath24 . we also included data points for two recently discovered pwne ; those surrounding the high energy pulsar j14003 - 6326 @xcite and axp 1e1547.0- 5408 @xcite . we did not include these in our statistical analysis , although their properties are consistent with the general trends we report below . figure [ fig : eta ] also shows the @xmath0-ray efficiency , as obtained from the pulsed @xmath0-ray luminosities determined by @xcite . this figure reveals the trend that young pulsars appear to have higher x - ray efficiencies than old pulsars . moreover , the behavior is similar for the luminosities of the pulsars and the pwne . figure [ fig : ppdot ] shows the timing properties of the pulsars in the two samples used for our statistical analysis in the @xmath25 diagram . it shows that on average the detected @xmath0-ray pulsars seem to be somewhat older than the x - ray detected pulsars / pwne , consistent with the trend that older pulsars are less efficient in emitting x - rays and more efficient in emitting @xmath0-rays . note , however , that an interpretation of this diagram is far from straightforward , because the detectability of a pulsar also depends on its distance and , for x - rays , on the interstellar absorption column . a trend of decreasing x - ray luminosity with @xmath26 was reported before by @xcite , but figure [ fig : eta ] reveals it to be due to a trend in the efficiency , not in the overall spin - down power . for pulsars with @xmath27 yr the efficiency appears to be more or less constant . in contrast , young pulsars are not so efficient in producing @xmath0-ray emission , as already noted by @xcite . the spread in the data points for a given characteristic is of order of 1 - 2 decades . as discussed above , the uncertainties in the luminosity can explain about 1 decade of this spread , but intrinsic variations , and for the @xmath0-ray emission , beaming properties are expected to contribute to the spread as well . in order to investigate the correlations between @xmath21 and @xmath26 further , one has to avoid @xmath21 as an independent variable , since both @xmath21 and @xmath26 are derived from combinations of the pulsar s period @xmath13 and period derivative @xmath28 ; the spin - down luminosity is given by @xmath29 with @xmath30 g@xmath31 the neutron star s moment of inertia . for that reason we base our regression analysis on the logarithm to denote the 10-based logarithm . ] of the x - ray/@xmath0-ray luminosity , as the quantity to be explained , and the independent variables @xmath32 , @xmath33 as the principle input variables for the model ( c.f . as explained above the errors are dominated by systematic errors , mostly due to difficulties in estimating distances . for that reason we used the unweighted least square method , which means that the errors in the best - fit parameters are based on the variance of the residuals . the disadvantage is that we do not have an intrinsic goodness of fit statistic . however , we can compare two hypotheses using the f - test statistic , defined as : @xmath34 here @xmath35 denotes the observed ( @xmath0-ray / x - ray ) luminosity and we use @xmath36 to indicate the expected value of @xmath35 based on the best - fit parameters , and the subscript @xmath37 and @xmath38 to indicate two different best - fit models , @xmath39 is the number of degrees of freedom and @xmath8 is the sample size . the probability that the improvement in the sum of squared residuals is by pure chance is given by the @xmath40-distribution . from the above it is clear that we fitted the linear relation : @xmath41 we compare this relation , using the f - test , with a functional form @xmath42 . this is essentially assuming that @xmath21 ( equation [ eq : eta ] ) is constant . in addition we fit the relation @xmath43 which corresponds to @xmath44 . this is a functional form that has been used before ( e.g. * ? ? ? * ) , and that seems to work particularly well for the @xmath0-ray emission from pulsars @xcite . since figure [ fig : eta ] suggests that young pulsars are more efficient x - ray emitters than old pulsars , we divided the x - ray sample in an old and a young population . as the precise characteristic age that should be chosen is a bit arbitrary , but lies somewhere between @xmath45 yr and @xmath46 yr , we decided to divide the sample in two more or less equally sized samples . this put the cut at @xmath47 yr . the result of the regression analysis is listed in table [ tab : fits ] , whereas figure [ fig : bestfit ] shows the results of applying the best - fit solutions to the different samples . the best - fit relations for young pulsars and pwne indicate that the x - ray luminosity does not scale with @xmath9 , since in that case we would expect for equation ( [ eq : fullfit ] ) @xmath48 and @xmath49 , whereas we find @xmath50 and @xmath51 and @xmath52 and @xmath53 for pwne and pulsars , respectively . this conclusion is based on both the best - fit parameters and their inferred errors , as on the f - test . the f - test , in fact , indicates that equation ( [ eq : fullfit ] ) provides a better fit than equation ( [ eq : eta ] ) at the 99.99% confidence level . interestingly , the values for @xmath54 and @xmath55 for pulsars and pwne are consistent with each other . for older pulsars the f - values indicate that the x - ray emission efficiency is just as well described by equation ( [ eq : eta ] ) with @xmath56 , as by a linear dependency on @xmath32 and @xmath33 . also this result is true for both the pulsar x - ray luminosity as for the pwn luminosity . the least square fits for the _ fermi_-sample shows that the relation @xmath4 is indeed a good description of the data . for equation ( [ eq : edotfit ] ) we find @xmath57 , close to the expected 1/2 , and for equation ( [ eq : fullfit ] ) we find @xmath58 and @xmath59 , whereas a scaling with @xmath60 implies @xmath61 and @xmath62 , in excellent agreement . the f - tests confirm this , and indicate that a constant value for @xmath63 can be rejected at the 99.9996% confidence level . finally , one may wonder whether the combined @xmath0-ray and x - ray luminosity is more closely correlated with @xmath9 , than the @xmath0-ray and x - ray luminosities individually . unfortunately , this is hard to determine from the present data sets as there are only 15 pulsars that the two samples have in common , with most of them being older pulsars . in fact , the combined x - ray/@xmath0-ray luminosity of all pulsars is dominated by the @xmath0-ray emission , with the exception of the crab pulsar . it is , therefore , not surprising that the best - fit relation between @xmath64 indicates @xmath65 , just like for @xmath0-ray luminosity . however , as the number of @xmath0-ray - detected pulsars will grow in the near future , it may be good to investigate the total radiative output from rotation powered pulsars in more detail , especially around @xmath66 yr , where the radiative output changes from x - ray to @xmath0-ray domination . we presented a statistical study of the x - ray and @xmath0-ray properties of rotation powered pulsars , with the aim of finding what trends underly the efficiency with which spin - down luminosity is converted into high energy radiation . we started our analysis by showing that young pulsars are efficient x - ray emitters , but poor @xmath0-ray emitters , whereas it is the other way around for old pulsars . a trend of low x - ray luminosity as a function of characteristic age was reported by @xcite , but it is shown here that it is related to the efficiency with which spin - down luminosity is converted to x - ray emission , and not just due to an overall decline in spin - down luminosity . in order to be as general as possible in finding trends in x - ray/@xmath0-rayluminosities our main results are based on a regression analysis of x - ray luminosity of the pulsars / pwne @xcite and @xmath0-ray luminosity of pulsars @xcite versus the pulse period and its derivative . this is a very generic method , which includes the possibility that the luminosity depends solely on spin - down luminosity . thus encompassing other possible dependencies , such as on spin - down luminosity . given the trend noted in figure [ fig : eta ] we divided the x - ray sample in two equally sized samples of young ( @xmath1 yr ) and old pulsars . our statistical analysis produced two new findings : 1 ) the x - ray luminosity of pulsars and their surrounding pwne appear closely correlated , with @xmath67 for young pulsars , whereas for old pulsars the best fit gives @xmath68 , the latter being close to , and statistically indistinguishable from , a constant x - ray emission efficiency @xmath21 ; 2 ) young pulsars are more efficient x - ray emitters than old pulsars , and have a different dependency for @xmath69 on @xmath13 and @xmath28 . we also confirm the findings by @xcite that 1 ) the @xmath0-ray luminosity is well described by @xmath4 ; 2 ) younger pulsars are less efficient in producing @xmath0-rays . the latter trend is , therefore , the reverse of the x - ray luminosity ( figure [ fig : eta ] ) . the question is what these findings reveal about pulsar x - ray emission mechanisms . first of all , the connection between the x - ray luminosities of pulsars and their pwne may be surprising , given that the luminosities of the pwne are , in general , affected by both the pulsar wind properties and the environment of the pulsar . in particular for young pwne the luminosity is possibly affected by its interaction with a surrounding supernova remnant ( snr ) . the reverse shock of the snr will during a certain phase of the snr evolution compress the pwn , which naively may be assumed to result in a brightnening of the pwn . however , a recent study by @xcite showed the behavior of the pwn luminosity to be more complex . their study indicates that the compression by the reverse shock leads to a brightening of the radio luminosity , but at the same time to an almost total quenching of the x - ray luminosity for a brief period ( for the specific model they calculated this happened between 18 - 30 kyr ) . apart from this brief phase , the x - ray luminosity traces the spin - down luminosity surprisingly well , with an expected fluctuation of about 0.3 decades in @xmath21 . given that apparently the x - ray luminosities of the pulsars and their pwne seem to trace each other , and have similar dependencies on the period and period derivative , one can not easily use the best - fit functions for young pulsars to derive a correlation of x - ray luminosity with some well known physical pulsar property , such as @xmath70 , @xmath71 , @xmath72 ( the magnetic field at the light cylinder ) , @xmath73 , or @xmath74 . for young pulsars the x - ray emission is poorly fit with a dependency on @xmath75 ( table [ tab : fits ] ) . this suggests that the x - ray luminosity from pulsars and pwne may require a more complex model than the @xmath0-ray luminosity , with its @xmath4 scaling . the best - fit formula does , however , explain the strong dependency on @xmath26 for young pulsars , since @xmath76 . secondly , the similar behavior of the pulsars and pwn x - ray luminosities suggest there is a physical connection between the two . it is usually assumed that the pulsar x - ray emission originates in the magnetosphere , whereas the x - ray emission from the nebulae comes from outside the termination shock . these are two distinct regions , which are separated by a region that encompasses the pulsar light cylinder and the so - called wind - zone @xcite , the region in which the pulsar wind is formed . it is tempting to speculate that the connection between the x - ray emission from the pulsars and pwne may have something to do with the pair multiplicity . models of pair creation in the magnetosphere indicate that the pair multiplicity is a function of the characteristic age ( e.g. * ? ? ? this would mean that both the x - ray emission from the pulsar and from the pwne are somehow proportional to the multiplicity . this is not a completely satisfactory explanation , because the x - ray synchrotron depends on the total energy contained by the pairs , not just by the total number of particles . another explanation for the similar behavior of pulsar and pwn luminosities is offered by the striped wind theory @xcite . according to this theory the alternating magnetic fields generated by obliquely rotating pulsars leads to reconnecting magnetic fields in the wind zone . this transforms magnetic energy into kinetic energy , thereby changing the pulsar wind from a high to a low @xmath11 outflow . according to @xcite the zone in which this heating and acceleration occurs could be the location of pulsed x - ray emission . such a model makes it easier to explain why there is a connection between the pulsar and pwn luminosities , as in both cases the energy is generated in the wind zone . the higher pair multiplicity of young pulsars may be of additional importance as there are simply more particles available to be accelerated . finally , there is one other issue to consider , namely , why young pulsars do not seem to be efficient in generating @xmath0-rays . the fact that it is the other way around for the x - ray emission , perhaps indicates that the pair multiplicity in young pulsars is so high that the electric fields in the magnetosphere are shorted out @xcite . it may also be worthwhile to investigate in more detail what the role of the seed photons for inverse compton upscattering in young pulsars is . the compton scattered photons ( and in some cases curvature radiation generated photons ) are converted in pairs , and form the basis of the pair multiplicity @xcite . the soft seed photons originate from the neutron star surface , and may the result of polar cap heating due to accelerated electron / positron beams , or they are the result of the cooling of the neutron star . the latter would induce a dependency on neutron star age , which could explain why young pulsars behave differently than old pulsars . we have only briefly mentioned the tev detected pwne . the tev luminosities of pwne do not seem to correlate with spin - down luminosity @xcite . the probable reason is that the tev emission from pwne is not so much determined by the current energy production of the pulsar , but reflects the time integrated energy input from the pulsar wind , which in itself scales with the initial spin - period @xmath77 . in x - rays the lifetime of the x - ray synchrotron emitting electrons is short compared to the pulsar lifetime , whereas in tev ( and in the radio ) this is not the case . we can therefore conclude that different parts of the electromagnetic spectrum inform us about different aspects of pulsars . the @xmath0-ray emission and pulsed radio emission is likely related to what happens in the magnetosphere , whereas tev and radio emission from the pwne inform us about the time integrated properties of pulsars . based on our statistical study we have added here the suggestion that the x - ray emission from the pulsar , like the x - ray emission from the pwne , may inform us about what happens in one of the least understood regions surrounding the pulsar : the wind zone . jv is supported by a vidi grant from the netherlands science foundation ( nwo ) . ry is supported by grant - in - aid from the ministry of education , culture , sports , science , and technology ( mext ) of japan , no . 19047004 , no . 21740184 , no . we have made use of the atnf online pulsar catalogue : http://www.atnf.csiro.au/research/pulsar/psrcat . \(2 ) @xmath85 & @xmath79 & @xmath86 & @xmath87 & @xmath88 & @xmath89 & @xmath90 + & @xmath54 & @xmath91 & @xmath92 & @xmath93 & @xmath94 & @xmath95 + & @xmath96-value and associated probability based on a comparison with formula 1 in this table . ] & @xmath97 & @xmath98 & @xmath99 & @xmath100 & @xmath101 + \(3 ) @xmath102 & a & @xmath103 & @xmath104 & @xmath105 & @xmath106 & @xmath107 + & b & @xmath108 & @xmath109 & @xmath110 & @xmath111 & @xmath112 + & c & @xmath113 & @xmath114 & @xmath115 & @xmath116 & @xmath117 + & @xmath96 & @xmath118 & @xmath119 & @xmath120 & @xmath121 & @xmath122 + & @xmath96 & @xmath123 & @xmath124 & @xmath125 & @xmath126 & @xmath127 + sample size & & & & 28 +	we present a statistical analysis of the x - ray luminosity of rotation powered pulsars and their surrounding nebulae using the sample of @xcite and we complement this with an analysis of the @xmath0-ray - emission of _ fermi _ detected pulsars . we report a strong trend in the efficiency with which spin - down power is converted to x - ray and @xmath0-ray emission with characteristic age : young pulsars and their surrounding nebulae are efficient x - ray emitters , whereas in contrast old pulsars are efficient @xmath0-ray emitters . we divided the x - ray sample in a young ( @xmath1 yr ) and old sample and used linear regression to search for correlations between the logarithm of the x - ray and @xmath0-ray luminosities and the logarithms of the periods and period derivatives . the x - ray emission from young pulsars and their nebulae are both consistent with @xmath2 . for old pulsars and their nebulae the x - ray luminosity is consistent with a more or less constant efficiency @xmath3 . for the @xmath0-ray luminosity we confirm that @xmath4 . we discuss these findings in the context of pair production inside pulsar magnetospheres and the striped wind model . we suggest that the striped wind model may explain the similarity between the x - ray properties of the pulsar wind nebulae and the pulsars themselves , which according to the striped wind model may both find their origin outside the light cylinder , in the pulsar wind zone .
let @xmath6 be a group generated by a finite set @xmath7 , where @xmath8 denotes the free group on @xmath7 . let @xmath9 be the number of elements of the normal subgroup @xmath10 of @xmath8 whose minimal representation as words in @xmath11 has length @xmath12 ; let @xmath13 be the number of ( not necessarily reduced ) words of length @xmath12 in @xmath11 that evaluate to @xmath14 in @xmath15 ; and let @xmath16 . the numbers @xmath17{f_n},\qquad\nu=\frac1d\limsup_{n\to\infty}\sqrt[n]{g_n}\ ] ] are called the _ cogrowth _ and _ spectral radius _ of @xmath18 . the grigorchuk formula @xcite states that @xmath19 we generalize this result to a somewhat more general setting : we replace the group @xmath15 by a regular graph @xmath20 , i.e. a graph with the same number of edges at each vertex . fix a vertex @xmath21 of @xmath20 ; let @xmath13 be the number of circuits ( closed sequences of edges ) of length @xmath12 at @xmath21 and let @xmath9 be the number of circuits of length @xmath12 at @xmath21 with no backtracking ( no edge followed twice consecutively ) . then the same equation holds between the growth rates of @xmath9 and @xmath13 . to a group @xmath15 with fixed generating set one associates its cayley graph @xmath20 ( see subsection [ subs : gptheory ] ) . @xmath20 is a @xmath0-regular graph with distinguished vertex @xmath22 ; paths starting at @xmath21 in @xmath20 are in one - to - one correspondence with words in @xmath11 , and paths starting at @xmath21 with no backtracking are in one - to - one correspondence with elements of @xmath8 . a circuit at @xmath21 in @xmath20 is then precisely a word evaluating to @xmath14 in @xmath15 , and a circuit without backtracking represents precisely one element of @xmath10 . in this sense results on graphs generalize results on groups . the converse would not be true : there are even graphs with a vertex - transitive automorphism group that are not the cayley graph of a group @xcite . even more generally , we will show that , rather than counting circuits and proper circuits ( those without backtracking ) at a fixed vertex , we can count paths and proper paths between two fixed vertices and obtain the same formula relating their growth rates . these relations between growth rates are consequences of a stronger result , expressed in terms of generating functions . define the formal power series @xmath23 then assuming @xmath20 is @xmath0-regular we have @xmath24 this equation relates @xmath25 and @xmath26 , and so relates _ a fortiori _ their radii of convergence , which are @xmath27 and @xmath28 . we re - obtain thus the grigorchuk formula . finally , rather than counting paths and proper paths between two fixed vertices , we can count , for each @xmath29 , the number of paths with @xmath30 backtrackings , i.e. with @xmath30 occurrences of an edge followed twice in a row . letting @xmath31 be the number of paths of length @xmath12 with @xmath30 backtrackings , consider the two - variable formal power series @xmath32 note that @xmath33 and @xmath34 . the following equation now holds : @xmath5 letting @xmath35 in this equation reduces it to the previous one . a generalization of the grigorchuk formula in a completely different direction can be attempted : consider again a finitely generated group @xmath15 , and an exact sequence @xmath36 where this time @xmath37 is not necessarily free . assume @xmath37 is generated as a monoid by a finite set @xmath7 . let again @xmath13 be the number of words of length @xmath12 in @xmath7 evaluating to @xmath14 in @xmath15 , and let @xmath9 be the number of elements of @xmath38 whose minimal - length representation as a word in @xmath7 has length @xmath12 . is there again a relation between the @xmath9 and the @xmath13 ? in section [ sec : nonfree ] we derive such a relation for @xmath37 the modular group @xmath39 . again there is a combinatorial counterpart ; rather than considering graphs one considers a locally finite cellular complex @xmath40 such that all vertices have isomorphic neighbourhoods . as before , @xmath13 counts the number of paths of length @xmath12 in the @xmath14-skeleton of @xmath40 between two fixed vertices ; and @xmath9 counts elements of the fundamental groupoid , i.e. homotopy classes of paths , between two fixed vertices whose minimal - length representation as a path in the @xmath14-skeleton of @xmath40 has length @xmath12 . we obtain a relation between these numbers when @xmath40 consists solely of triangles and arcs , with no two triangles nor two arcs meeting ; these are precisely the complexes associated with quotients of the modular group . let @xmath20 be a graph , that may have multiple edges and loops . we make the following typographical convention for the power series that will appear : a series in the formal variable @xmath41 is written @xmath1 , or @xmath26 for short , and @xmath42 refers to the series @xmath26 with @xmath43 substituted for @xmath41 . functions are written on the right , with @xmath44 or @xmath45 denoting @xmath46 evaluated at @xmath43 . we start by the precise definition of graph we will use : a _ graph _ @xmath20 is a pair of sets @xmath47 and maps @xmath48 satisfying @xmath49 the graph @xmath20 is said to be _ finite _ if both @xmath50 and @xmath51 are finite sets . a _ graph morphism _ @xmath52 is a pair of maps @xmath53 with @xmath54 and @xmath55 and @xmath56 given an edge @xmath57 , we call @xmath58 and @xmath59 @xmath60 s _ source _ and _ destination _ , respectively . we say two vertices @xmath61 are _ adjacent _ , and write @xmath62 , if they are connected by an edge , i.e. if there exists an @xmath57 with @xmath63 and @xmath64 . we say two edges @xmath65 are _ consecutive _ if @xmath66 . a _ loop _ is an edge @xmath60 with @xmath67 . the _ degree _ @xmath68 of a vertex @xmath43 is the number of incident edges : @xmath69 if @xmath68 is finite for all @xmath43 , we say @xmath20 is _ locally finite_. if @xmath70 for all vertices @xmath43 , we say @xmath20 is _ @xmath0-regular_. note that the involution @xmath71 may have fixed points . even though the edges of @xmath20 are individually oriented , the graph @xmath20 itself should be viewed as an non - oriented graph . in case @xmath72 has no fixed point , @xmath20 can be viewed as a geometric graph . a _ path _ in @xmath20 is a sequence @xmath73 , @xmath74 of edges and vertices of @xmath20 , with @xmath75 and @xmath76 for all @xmath77 and @xmath78 . the _ length _ of the path @xmath73 is the number @xmath12 of edges in @xmath73 . start _ of the path @xmath73 is @xmath79 , and its _ end _ is @xmath80 . if @xmath81 , the path @xmath73 is called a _ circuit _ at @xmath82 . in most cases , we will omit the @xmath83 from the description of paths ; they are necessary only if @xmath84 , in which case a starting vertex must be specified . we extend the involution @xmath72 from edges to paths by setting @xmath85 ( note that @xmath86 is a path from @xmath87 to @xmath82 ) . we denote by @xmath88 the set of paths , with a partially defined multiplication given by concatenation : if @xmath73 and @xmath89 be two paths with @xmath90 , their _ product _ is defined as @xmath91 . for two vertices @xmath92 we denote by @xmath93 $ ] the set of paths from @xmath43 to @xmath94 . we turn @xmath51 into a metric space by defining for vertices @xmath92 their _ distance _ @xmath95}\|\pi\|.\ ] ] the _ ball _ of radius @xmath12 at a vertex @xmath96 is the subgraph @xmath97 of @xmath20 with vertex set @xmath98 and edge set @xmath99 we define @xmath100 , @xmath101 and @xmath102 . this definition amounts to `` wrapping around disconnected edges '' . it has the advantage of preserving the degrees of vertices . we say a path @xmath73 has a _ bump at @xmath103 _ if @xmath104 ; if the location of the bump is unimportant we will just say @xmath73 has a bump . bump count _ @xmath105 of a path @xmath73 is the number of bumps in @xmath73 . a _ proper path _ in @xmath20 is a path @xmath73 with no bumps . let @xmath106 be a ring . a @xmath106-_labelling _ of the graph @xmath20 is a map @xmath107 the simplest examples of labellings are : * the _ trivial labelling _ , given by @xmath108 and @xmath109 for all @xmath57 ; * the _ length labelling _ , given by @xmath110 $ ] and @xmath111 for all @xmath57 . a @xmath106-labelling @xmath112 of @xmath20 induces a map , still written @xmath113 , by setting @xmath114 the labelling @xmath113 is _ complete _ @xcite if for any vertex @xmath43 of @xmath20 and any set @xmath115 of paths in @xmath20 starting at @xmath43 there is an element @xmath116 of @xmath106 , and this function @xmath117 satisfies @xmath118 for all paths @xmath73 and disjoint sets @xmath115 and @xmath119 ( @xmath120 denotes disjoint union ) . if @xmath115 is infinite , it is customary , though abusive , to write @xmath116 as @xmath121 . if @xmath106 is a topological ring ( @xmath122 , @xmath123 , the formal power series ring @xmath124 $ ] , etc . ) , completion of @xmath112 implies that @xmath125 when @xmath126 , but the converse does not hold . the completeness condition becomes that @xmath127 be a well - defined element of @xmath106 for all @xmath115 ; i.e. , the limit exists . generally , we define the following topology on @xmath106 : a sequence @xmath128 converges to @xmath129 if and only if @xmath130 tends to infinity . in the sequel of this paper all labellings will be assumed to be complete . the length labelling is complete for locally finite graphs ; more generally , @xmath112 is complete when @xmath106 is a discretely valued ring , @xmath131 has a positive valuation for all edges @xmath60 , and @xmath20 is locally finite . an arbitrary ring @xmath106 may be embedded in @xmath132 $ ] , where @xmath41 has valuation @xmath14 and @xmath106 has valuation @xmath129 ; if @xmath133 is a labelling , we define @xmath134 by @xmath135 ; and @xmath136 will be complete as soon as @xmath20 is locally finite . in particular the length labelling is obtained from the trivial labelling through this construction . in all the examples we consider the labelling is defined in this manner . throughout the paper we will assume a graph @xmath20 and two vertices @xmath137 have been fixed . we wish to enumerate the paths , counting their number of bumps , from @xmath21 to @xmath138 in @xmath20 . for a given complete edge - labelling @xmath112 , consider the series @xmath139}\pi^\ell\in\bbbk,\qquad { \mathfrak f}(\ell)=\sum_{\pi\in[{\star},{\dagger}]}u^{{\operatorname{bc}}(\pi)}\pi^\ell\in\bbbk[[u]].\ ] ] beware that in general @xmath140 and @xmath141 also depend on the choice of @xmath21 and @xmath138 . for all vertices @xmath96 let @xmath142.\ ] ] ( a combinatorial interpretation of these @xmath143 will be given in section [ sec : proof ] . ) let @xmath112 be a complete @xmath106-labelling of @xmath20 , and define a new labelling @xmath144 $ ] by @xmath145 then our main result is the following : [ thm : main ] with the definitions of @xmath136 and @xmath143 given above , @xmath136 is a complete labelling and we have in @xmath146 $ ] the equality @xmath147 we now explicit the definitions and main result for the length labelling on a locally finite graph . the integer - valued series @xmath148}t^{\|\pi\|}\in{{\mathbb n}}[[t]]\ ] ] is called the _ path series _ of @xmath149 . the series @xmath150}u^{{\operatorname{bc}}(\pi)}t^{\|\pi\|}\in{{\mathbb n}}[u][[t]]\subset{{\mathbb n}}[[u , t]]\ ] ] is called the _ enriched path series _ of @xmath149 . its specialization @xmath151 is called the _ proper path series _ of @xmath149 . in case @xmath152 , we will call @xmath26 the _ circuit series _ of @xmath153 and @xmath25 the _ enriched circuit series _ of @xmath153 . the circuit series is often called the _ green function _ of the graph @xmath20 . note that @xmath3 lies in @xmath154[[t]]$ ] because the number of bumps on a path is smaller than its length , so all monomials in the sum have a @xmath155-degree smaller than their @xmath41-degree ; hence for any fixed @xmath41-degree there are only finitely many monomials with same @xmath41-degree , because @xmath20 is locally finite . expressed in terms of length labellings , our main theorem then gives the following result : [ cor : main ] suppose @xmath20 is a @xmath0-regular graph . then we have @xmath156 because @xmath20 is regular , the @xmath143 defined above do not depend on @xmath43 and are all equal to @xmath157 thus theorem [ thm : main ] reads @xmath158 now writing @xmath159 and @xmath160 completes the proof . the special case @xmath35 of this formula appears as an exercise in @xcite . the meaning of the corollary is that , for regular graphs , the richer two - variable generating series @xmath3 can be recovered from the simpler @xmath1 . conversely , @xmath26 can be recovered from @xmath25 , for instance because @xmath161 . remember it is valid to substitute @xmath14 for @xmath155 in @xmath25 , because for any fixed @xmath41-degree only finitely many monomials with that @xmath41-degree occur in @xmath25 . in fact , much more is true , as we have the equality @xmath162 or after simplification @xmath163 where both sides are to be understood as power series in @xmath164 $ ] that actually reside in @xmath165 $ ] . then for any value ( say , in @xmath123 ) of @xmath155 we obtain an expression of @xmath26 in terms of @xmath166 . of particular interest is the case @xmath167 , where ( [ eq : c ] ) specializes to @xmath168 this equation appears in a slightly different form in @xcite . similarly , we have in @xmath169 $ ] the equality @xmath170 beware though that ( [ eq : d ] ) holds for formal variables @xmath171 and @xmath41 ; if we were to substitute a real number for @xmath41 , then the resulting series @xmath172 would converge absolutely for @xmath173 , where @xmath89 is the radius of convergence of @xmath174 , and in particular not in a neighbourhood of @xmath129 . the equalities ( [ eq : c ] ) and ( [ eq : d ] ) are easily derived from ( [ eq : b ] ) by letting @xmath175 and solving for @xmath41 and @xmath155 . [ cor : ratalg ] in the setting described above : 1 . if @xmath20 is finite , then both @xmath25 and @xmath26 are rational series . if @xmath26 is rational , then @xmath25 is rational too . @xmath25 is algebraic if and only if @xmath26 is algebraic . the proofs are immediate and follow from the explicit form of ( [ eq : b ] ) . the converse of the first statement of the preceding corollary will be proved in section [ subs : convc ] . the last statement appears in @xcite and @xcite . in the following section we draw some applications to other fields : group theory , language theory . we give applications of theorem [ thm : main ] and corollary [ cor : main ] to some examples of graphs in section [ sec : examples ] , and a derivation of a `` cogrowth formula '' ( as that of subsection [ subs : gptheory ] ) for a non - free presentation in section [ sec : nonfree ] . we give two proofs of the main result in sections [ sec : newpf ] and [ sec : proof ] . the first one , shorter , uses a reduction to finite graphs and their adjacency matrices . the second one is combinatorial and uses the inclusion - exclusion principle . using the first proof , we obtain in section [ sec : matrices ] an extension of a result by yasutaka ihara finally in section [ sec : freeamalg ] we show how to compute the circuit series of a free product of graphs ( an analogue of the free products of groups , _ via _ their cayley graph ) , and in section [ sec : direct ] do the same for direct products of graphs . the original motivation for formula [ eq : b ] was its implication of a well - known result in the theory of random walks on discrete groups . in this section we will show how @xmath26 is related to random walks and @xmath25 to cogrowth . this will give a generalization of the main formula ( [ eq : grigformula ] ) to homogeneous spaces @xmath176 , where @xmath38 does not have to be normal and @xmath37 is a free product of infinite - cyclic and order - two groups . for a survey on the topic of random walks see @xcite . throughout this subsection we will have @xmath177 . we recall the notion of growth of groups : let @xmath15 be a group generated by a finite symmetric set @xmath7 . for a @xmath178 define its _ length _ @xmath179 the _ growth series _ of @xmath18 is the formal power series @xmath180.\ ] ] expanding @xmath181 , the _ growth _ of @xmath18 is @xmath182{f_n}\ ] ] ( this supremum - limit is actually a limit and is smaller than @xmath183 ) . let @xmath184 be a subset of @xmath15 . growth series _ of @xmath184 relative to @xmath18 is the formal power series @xmath185.\ ] ] expanding @xmath186 , define the _ growth _ of @xmath184 relative to @xmath18 as @xmath187{f_n}.\ ] ] if @xmath188 is a transitive right @xmath15-set , the _ simple random walk _ on @xmath189 is the random walk of a point on @xmath188 , having probability @xmath190 of moving from its current position @xmath43 to a neighbour @xmath191 , for all @xmath192 . fix a point @xmath193 , and let @xmath194 be the probability that a walk starting at @xmath21 finish at @xmath21 after @xmath12 moves ( it does not depend on the choice of @xmath21 ) . we define the _ spectral radius _ of the random walk as @xmath195{p_n}.\ ] ] a group @xmath37 is _ quasi - free _ if it is free product of cyclic groups of order @xmath196 and @xmath197 . equivalently , there exists a finite set @xmath7 and an involution @xmath198 such that , as a monoid , @xmath199 @xmath37 is then said to be _ quasi - free on @xmath7_. all quasi - free groups on @xmath7 have the same cayley graph , which is a regular tree of degree @xmath200 . every group @xmath15 generated by a symmetric set @xmath7 is a quotient of a quasi - free group in the following way : let @xmath72 be an involution on @xmath7 such that for all @xmath192 we have the equality @xmath201 in @xmath15 . then @xmath15 is a quotient of the quasi - free group @xmath202 . the _ cogrowth series _ ( respectively _ cogrowth _ ) of @xmath18 is defined as the growth series ( respectively growth ) of @xmath203 relative to @xmath204 , where @xmath37 is a quasi - free group on @xmath7 . associated with a group @xmath37 generated by a set @xmath7 and a subgroup @xmath38 of @xmath37 , there is a @xmath200-regular graph @xmath20 on which @xmath37 acts , called the _ schreier graph _ of @xmath204 relative to @xmath38 . it is given by @xmath205 , with @xmath206 and @xmath207 i.e. two cosets @xmath208 are joined by at least one edge if and only if @xmath209 . ( this is the cayley graph of @xmath204 if @xmath210 . ) there is a circuit in @xmath20 at every vertex @xmath211 such that with @xmath212 for some @xmath192 ; and there is a multiple edge from @xmath213 to @xmath214 in @xmath20 if there are @xmath215 with @xmath216 . [ cor : homo ] let @xmath37 be a quasi - free group , presented as a monoid as @xmath199 let @xmath217 be a subgroup of @xmath37 . let @xmath218 denote the spectral radius of the simple random walk on @xmath219 generated by @xmath7 ; and @xmath220 denote the relative growth of @xmath38 in @xmath37 . then we have @xmath221 let @xmath20 be the schreier graph of @xmath204 relative to @xmath38 defined above . fix the endpoints @xmath222 , the coset of @xmath14 , and give @xmath20 the length labelling . let @xmath26 and @xmath25 be the circuit and proper circuit series of @xmath20 . in this setting , expressing @xmath223 and @xmath224 , we see that @xmath225 is the growth rate @xmath226{g_n}$ ] of circuits in @xmath20 , and @xmath227 the growth rate @xmath226{f_n}$ ] of proper circuits in @xmath20 . as both @xmath25 and @xmath26 are power series with non - negative coefficients , @xmath228 is the radius of convergence of @xmath26 and @xmath27 the radius of convergence of @xmath25 . let @xmath229 and consider the function @xmath230 this function is strictly increasing for @xmath231 , has a maximum at @xmath232 with @xmath233 , and is strictly decreasing for @xmath234 . first we suppose that @xmath235 , so @xmath236 is monotonously increasing on @xmath237 $ ] . we set @xmath35 in ( [ eq : b ] ) and note that , for @xmath238 , it says that @xmath25 has a singularity at @xmath41 if and only if @xmath26 has a singularity at @xmath239 . now as @xmath27 is the singularity of @xmath25 closest to @xmath129 , we conclude by monotonicity of @xmath236 that the singularity of @xmath26 closest to @xmath129 is at @xmath240 ; thus @xmath241 suppose now that @xmath242 . if @xmath243 , the right - hand side of ( [ eq : b ] ) would be bounded for all @xmath244 while the left - hand side diverges at @xmath245 . if @xmath246 , there would be a @xmath247 with @xmath248 ; and @xmath25 would have a singularity at @xmath249 . the only case left is @xmath250 . ( 300,300 ) ( 30,295)(0,0)[lb ] ( 275,50)(0,0)[lb ] ( -30,-10 ) [ cor : cogrowth ] let @xmath15 be a group generated by a symmetric finite set @xmath7 , let @xmath251 denote the spectral radius of the simple random walk on @xmath15 , and let @xmath227 denote the cogrowth of @xmath18 . then @xmath252 a variety of proofs exist for this result : the original @xcite by grigorchuk , one by cohen @xcite , an extension by northshield to regular graphs @xcite , a short proof by szwarc @xcite using operator theory , one by woess @xcite , etc . present @xmath15 as @xmath176 , with @xmath37 a quasi - free group and @xmath38 the normal subgroup of @xmath37 generated by the relators in @xmath15 , and apply corollary [ cor : homo ] . we note in passing that if @xmath253 , then necessarily @xmath254 . equivalently , we will show that if @xmath253 , then @xmath210 , so the cayley graph @xmath20 is a tree . indeed , suppose @xmath20 is not a tree , so it contains a circuit @xmath255 at @xmath21 . as @xmath20 is transitive , there is a translate of @xmath255 at every vertex , which we will still write @xmath255 . there are at least @xmath256 paths @xmath257 of length @xmath41 in @xmath20 starting at @xmath21 such that the circuit @xmath258 is proper ; thus @xmath259{\|s\|(\|s\|-1)^{t-2}(\|s\|-2 ) } = \sqrt{\|s\|-1}.\ ] ] in fact it is known that @xmath260 ; see @xcite . in this subsection we study the singularities the series @xmath25 and @xmath26 may have on their circle of convergence . the smallest positive real singularity has a special importance : for a series @xmath261 with positive coefficients , let @xmath262 denote its radius of convergence . then @xmath46 is _ @xmath262-recurrent _ if @xmath263 otherwise , it is _ @xmath262-transient_. as typical examples , @xmath264 is @xmath89-recurrent , as are all rational series ; @xmath265 is not @xmath89-transient . to study the singularities of @xmath25 or @xmath26 , we may suppose that @xmath152 ; indeed in was shown in @xcite and ( * ? ? ? * lemma 1 ) that the singularities of @xmath25 and @xmath26 do not depend on the choice of @xmath21 and @xmath138 . we make that assumption for the remainder of the subsection . we will also suppose throughout that @xmath20 is @xmath0-regular , that the radius of convergence of @xmath25 is @xmath27 and the radius of convergence of @xmath26 is @xmath266 . let @xmath20 be a connected graph . a _ proper cycle _ in @xmath20 is a proper circuit @xmath267 such that @xmath268 . the _ proper period _ @xmath257 and _ strong proper period _ @xmath269 are defined as follows : @xmath270 where by convention the @xmath271 of the empty set is @xmath197 . the graph @xmath20 is _ strongly properly periodic _ if @xmath272 . the _ period _ @xmath273 and _ strong period _ @xmath274 of @xmath20 are defined analogously with ` proper cycle ' replaced by ` circuit ' . @xmath20 is _ strongly periodic _ if @xmath275 . [ thm : cart ] let @xmath20 have proper period @xmath257 and strong proper period @xmath269 . then the singularities of @xmath25 on its circle of convergence are among the @xmath276 if moreover @xmath20 is strongly properly periodic , the singularities of @xmath25 on its circle of convergence are precisely these numbers . let @xmath20 have period @xmath273 and strong period @xmath274 . then the singularities of @xmath26 on its circle of convergence are among the @xmath277 if moreover @xmath20 is strongly periodic , the singularities of @xmath26 on its circle of convergence are precisely these numbers . if @xmath20 is connected and non - trivial , there is a path of even length at every vertex ( a sequence of bumps , for instance ) . all graphs are then either @xmath196-periodic ( if they are bipartite ) or @xmath14-periodic . if there is a constant @xmath10 such that for all @xmath96 there is at @xmath43 a circuit of odd length bounded by @xmath10 , then @xmath20 is strongly @xmath14-periodic ; otherwise @xmath20 is strongly @xmath196-periodic . the singularities of @xmath26 on its circle of convergence are then at @xmath278 , and also at @xmath279 if @xmath20 is strongly periodic with period @xmath196 . if @xmath20 is not strongly periodic , there may be one or two singularities on @xmath26 s circle of convergence ; consider for instance the @xmath280-regular tree , and at a vertex @xmath21 delete two or three edges replacing them by loops . the resulting graphs @xmath281 and @xmath282 are still @xmath280-regular and their circuit series , as computed in ( [ eq : treeloop ] ) , are respectively @xmath283 @xmath284 has singularities at @xmath285 on its circle of convergence , while @xmath286 has only @xmath287 as singularity on its circle of convergence . following the proof of corollary [ cor : homo ] above , we see that if @xmath288 the singularities of @xmath25 on its circle of convergence are in bijection with those of @xmath26 , so are at @xmath27 and possibly @xmath289 , if @xmath20 is strongly two - periodic . if @xmath290 , though , @xmath20 can have any strong proper period ; consider for example the cycles on length @xmath291 studied in section [ subs : cycles ] : they are strongly properly @xmath291-periodic . this simple result shows how @xmath20 can be approximated by finite subgraphs . [ lem : limit ] let @xmath20 be a graph and @xmath43 , @xmath94 two vertices in @xmath20 . let @xmath292 and @xmath293 be the path series and enriched path series respectively from @xmath43 to @xmath94 in @xmath20 , and let @xmath294 and @xmath295 be the path series and enriched path series respectively from @xmath43 to @xmath94 in the ball @xmath97 ( they are @xmath129 if @xmath296 ) . then @xmath297 recall that @xmath298 means that both terms are sums of paths , say @xmath299 and @xmath115 , such that the minimal length of paths in the symmetric difference @xmath300 tends to infinity . now the difference between @xmath294 and @xmath292 consists only of paths in @xmath20 that exit @xmath97 , and thus have length at least @xmath301 . the same argument holds for @xmath141 . the graph @xmath20 is _ quasi - transitive _ if @xmath302 acts with finitely many orbits . [ lem : gfcard ] let @xmath20 be a regular quasi - transitive connected graph with distinguished vertex @xmath303 , and let @xmath9 and @xmath13 denote respectively the number of proper circuits and circuits at @xmath303 of length @xmath12 . then @xmath304 if @xmath20 is finite , then @xmath290 , the degree of @xmath20 ; after a large even number of steps , a random walk starting at @xmath21 will be uniformly distributed over @xmath20 ( or over the vertices at even distance of @xmath21 , in case all circuits have even length ) . a long enough walk then has probability @xmath305 ( or @xmath306 if all circuits have even length ) of being a circuit . if @xmath20 is infinite , we consider two cases . if @xmath307 , i.e. @xmath26 is @xmath278-transient , the general term @xmath308 of the series @xmath309 tends to @xmath129 . if @xmath26 is @xmath278-recurrent , then , as @xmath20 is quasi - transitive , @xmath290 by ( * ? ? ? * theorem 7.7 ) . we then approximate @xmath20 by the sequence of its balls of radius @xmath184 : @xmath310 where we expand the circuit series of @xmath311 as @xmath312 . the same proof holds for the @xmath9 . its particular case where @xmath20 is a cayley graph appears in @xcite . note that if @xmath20 is not quasi - transitive , a somewhat weaker result holds @xcite : if @xmath20 is transient or null - recurrent then the common limsup is @xmath129 . if @xmath20 is positive - recurrent then the limsups are normalized coefficients of @xmath20 s perron - frobenius eigenvector . lemma [ lem : gfcard ] is not true for arbitrary @xmath0-regular graphs : consider for instance the graph @xmath282 described above . its circuit series @xmath286 , given in ( [ ex:4reg3 ] ) , has radius of convergence @xmath313 , and one easily checks that all its coefficients @xmath13 satisfy @xmath314 . we obtain the following characterization of rational series : for regular quasi - transitive connected graphs , the following are equivalent : 1 . @xmath20 is finite;[thm : rat : a ] 2 . @xmath1 is a rational function of @xmath41;[thm : rat : b ] 3 . @xmath315 is a rational function of @xmath41 , and @xmath20 is not an infinite tree.[thm : rat : c ] by corollary [ cor : ratalg ] , statement [ thm : rat : a ] implies the other two . by corollary [ cor : main ] , and a computation on trees done in section [ subs : trees ] to deal with the case @xmath316 , statement [ thm : rat : b ] implies [ thm : rat : c ] . it remains to show that statement [ thm : rat : c ] implies [ thm : rat : a ] . assume that @xmath317 is rational , not equal to @xmath14 . as the @xmath9 are positive , @xmath25 has a pole , of multiplicity @xmath30 , at @xmath27 . there is then a constant @xmath318 such that @xmath319 for infinitely many values of @xmath12 @xcite . it follows by lemma [ lem : gfcard ] that @xmath320 and the graph @xmath20 is finite , of cardinality at most @xmath321 . it is not known whether the same holds for regular , or even arbitrary connected graphs . certainly an altogether different proof would be needed . let @xmath7 be a finite set of cardinality @xmath0 and let @xmath72 be an involution on @xmath7 . a _ word _ is an element @xmath322 of the free monoid @xmath323 . a _ language _ is a set @xmath324 of words . the language @xmath324 is called _ saturated _ if for any @xmath325 and @xmath192 we have @xmath326 that is to say , @xmath324 is stable under insertion and deletion of subwords of the form @xmath327 . the language @xmath324 is called _ desiccated _ if no word in @xmath324 contains a subword of the form @xmath327 . given a language @xmath324 we may naturally construct its _ saturation _ @xmath328 , the smallest saturated language containing @xmath324 , and its _ desiccation _ @xmath329 , the largest desiccated language contained in @xmath324 . let @xmath117 be the monoid defined by generators @xmath7 and relations @xmath330 for all @xmath192 : @xmath331 this is a free product of free groups and order - two groups ; if @xmath72 is fixed - point - free , @xmath117 is a free group . write @xmath236 for the canonical projection from @xmath323 to @xmath117 . let @xmath332 $ ] be its monoid ring . then given a language @xmath333 we may define its _ growth series _ @xmath334 as @xmath335.\ ] ] this notion of growth series with coefficients was introduced by fabrice liardet in his doctoral thesis @xcite , where he studied _ complete growth functions _ of groups . for any language @xmath324 there holds @xmath336 where @xmath337 . for any language there exists a unique minimal ( possibly infinite ) automaton recognising it ( @xcite is a good reference ) . let @xmath20 be the minimal automaton recognising @xmath338 . recall that this is a graph with an initial vertex @xmath21 , a set of terminal vertices @xmath339 and a labelling @xmath340 of the graph s edges such that the number of paths labelled @xmath322 , starting at @xmath21 and ending at a @xmath341 is @xmath14 if @xmath342 and @xmath129 otherwise . extend the labelling @xmath136 to a labelling @xmath343 $ ] by @xmath344 because @xmath345 is saturated , and @xmath20 is minimal , @xmath346 ; then @xmath329 is the set of labels on proper paths from @xmath21 to some @xmath341 . choosing in turn all @xmath341 as @xmath138 , we obtain growth series @xmath347 counting the formal sum of paths and proper paths from @xmath21 to @xmath348 . it then suffices to write @xmath349 the following result is well - known : let @xmath15 be a finitely generated group , presented as a quotient @xmath350 with @xmath117 as in ( [ eq : defsigma ] ) . then @xmath351 is an algebraic series ( i.e. satisfies a polynomial equation ) if and only if @xmath350 is virtually free ( i.e. has a normal subgroup of finite index that is free ) . it is not known whether there exists a non - virtually - free graph whose circuit series ( as defined in corollary [ cor : main ] ) is algebraic . we will now prove theorem [ thm : main ] using linear algebra . we first assume the graph has a finite number of vertices ; for the computations refer to @xmath106-matrices and @xmath146$]-matrices indexed by the graph s vertices . this proof is hinted at in godsil s book as an exercise @xcite ; it was suggested to the author by gilles robert . for all pairs of vertices @xmath92 let @xmath352}\pi^\ell,\qquad{\mathfrak f}_{x , y}(\ell)=\sum_{\pi\in[x , y]}u^{{\operatorname{bc}}(\pi)}\pi^\ell\ ] ] be the path and enriched path series from @xmath43 to @xmath94 ; for ease of notation we will leave out the labelling @xmath112 if it is obvious from context . let @xmath353 denote the kronecker delta , equal to @xmath14 if @xmath354 and @xmath129 otherwise . for any @xmath355 , let @xmath356_x^y$ ] denote the @xmath357 matrix with zeroes everywhere except at @xmath358 , where it has value @xmath359 . then @xmath360 so that if @xmath361_{e^\alpha}^{e^\omega}\ ] ] be the adjacency matrix of @xmath20 , with labellings , then we have @xmath362 an equation holding between @xmath357 matrices over @xmath106 . similarly , letting @xmath363 count the paths from @xmath43 to @xmath94 that start with the edge @xmath60 , @xmath364 these last two lines solve to @xmath365 which we insert in the first line to obtain @xmath366 thus if we let @xmath367_{e^\alpha}^{e^\omega},\ ] ] we obtain @xmath368 and the proof is finished in the case that @xmath20 is finite , because the matrix @xmath369 is precisely that obtained from @xmath115 by substituting @xmath136 for @xmath112 . if @xmath20 has infinitely many vertices , we approximate it , thanks to lemma [ lem : limit ] , by finite graphs . denote by @xmath370 and @xmath371 the enriched path series and path series respectively in @xmath372 , and write @xmath373 to complete the proof . graphs can be studied through their _ adjacency _ and _ incidence _ matrices . we give here the relevant definitions and obtain an extension of a theorem by hyman bass @xcite on the ihara - selberg zeta function . we will use power series with coefficients in a matrix ring , and fractional expressions in matrices ; by convention , we understand ` @xmath374 ' as ` @xmath375 ' . let @xmath20 be a finite graph . the _ edge - adjacency _ and _ inversion _ matrices of @xmath20 , respectively @xmath119 and @xmath376 , are @xmath377 matrices over @xmath378 defined by @xmath379 the _ vertex - adjacency _ and _ degree _ matrices of @xmath20 , respectively @xmath115 and @xmath380 , are @xmath357 matrices over @xmath378 defined by @xmath381 a _ cycle _ is the equivalence class of a circuit under cyclic permutation of its edges . a _ proper cycle _ is a cycle all of whose representatives are proper circuits . a cycle is _ primitive _ if none of its representatives can be written as @xmath382 for some @xmath383 . the _ cyclic bump count _ @xmath384 of a circuit @xmath385 is @xmath386 where the edge @xmath387 is understood to be @xmath388 . the matrices given above are related to paths in @xmath20 as follows : consider first the matrix @xmath389 then the @xmath390 coefficient of @xmath391 is precisely @xmath392 this is clear because the series expansion of @xmath391 is the sum of sequences of @xmath393 ( contributing edges with no bump ) and @xmath394 ( contributing edges with bumps ) , with an extra factor of @xmath155 in case the path ends in @xmath395 . likewise , consider the matrix @xmath396 whose @xmath397-coefficient counts paths from vertex @xmath359 to vertex @xmath322 . we now state and prove an extension of a theorem by bass @xcite : [ thm : bass ] let @xmath398 be a set of representatives of primitive cycles in @xmath20 , and form the _ zeta function _ of @xmath20 @xmath399 ( the choice of representatives does not change the zeta function . ) then @xmath400 is a polynomial in @xmath155 and @xmath41 and can be expressed as @xmath401 where @xmath402 the special case @xmath403 of this result was stated and proved in the given sources . we will prove the general statement , using a result of shimson amitsur : let @xmath404 be square matrices of the same dimension over an arbitrary ring . let @xmath7 contain one representative up to cyclic permutation of words over the alphabet @xmath405 that are _ primitive _ , i.e. such that none of their cyclic permutations are proper powers of a word ( @xmath7 is infinite as soon as @xmath406 ) . for @xmath407 set @xmath408 . then @xmath409 considered as an equality of power series in @xmath41 over the matrix ring . the equality ( [ eq : bass1 ] ) then follows ; indeed , for all edges @xmath57 let @xmath410 be the @xmath377 matrix whose @xmath60-th row is the @xmath60-th row of @xmath411 , and whose other rows are @xmath129 . then clearly @xmath412 and , for any sequence of edges @xmath73 , @xmath413 so equality of @xmath414 and @xmath415 follows from amitsur s theorem . to prove ( [ eq : bass2 ] ) , we consider block matrices of dimension @xmath416 we write the adjacency operator and its adjoint @xmath417 so that @xmath418 whence @xmath419 let @xmath420 $ ] be the set of paths in @xmath20 from @xmath21 to @xmath138 . as the principle of inclusion - exclusion will be applied @xcite , it will be helpful to compute in @xmath421 $ ] , the @xmath378-module of functions from the set of paths to @xmath378 . we embed subsets of @xmath422 in @xmath37 by mapping a subset to its characteristic function : @xmath423 let @xmath424 be the subset of bounded non - negative elements of @xmath37 ( i.e. functions @xmath46 such that there is a constant @xmath10 with @xmath425 for all paths @xmath73 ) . if @xmath112 is a complete labelling of @xmath20 , there is an induced labelling @xmath426 given by @xmath427 note that the sum , although infinite , defines an element of @xmath106 due to the fact @xmath112 is complete . let @xmath57 and @xmath428 . squiggle along @xmath60 _ is a sequence @xmath429 . squiggle at @xmath359 _ is a squiggle along @xmath60 for some edge @xmath60 such that @xmath430 . let @xmath431 be a path of length @xmath12 in @xmath20 . a _ bump scheme _ for @xmath73 is a pair @xmath432 , with * for all @xmath433 , a finite ( possibly empty ) sequence @xmath434 of squiggles at @xmath435 ; * for all @xmath77 , a squiggle @xmath436 along @xmath437 . the _ weight _ @xmath438 of the bump scheme @xmath119 is defined as @xmath439 given a path @xmath73 and a bump scheme @xmath440 for @xmath73 , we obtain a new path @xmath441 , by setting @xmath442 where the product denotes concatenation . we now define a linear map @xmath443 $ ] by setting , for @xmath444 and @xmath445 , @xmath446 where the sum ranges over all pairs @xmath447 where @xmath448 and @xmath119 is a bump scheme for @xmath89 such that @xmath449 . note that the sum is finite because the edges of @xmath89 and of @xmath119 form subsets of those of @xmath73 . for any path @xmath73 we have @xmath450 say @xmath385 has @xmath451 bumps , at indices @xmath452 so that @xmath453 . we will show the evaluation at @xmath73 of the left - hand side of ( [ eq : lempf2 ] ) yields @xmath454 . we claim there is a bijection between the subsets @xmath455 of @xmath456 and the pairs @xmath457 where @xmath458 is a path and @xmath459 is a bump scheme for @xmath458 with @xmath460 ; and further @xmath461 . first , take a @xmath89 and a @xmath440 such that @xmath462 . the path @xmath463 is obtained by shuffling together the edges of @xmath89 and @xmath119 , and this partitions the edges of @xmath73 in two classes , namely _ ( i ) _ those coming from @xmath89 and _ ( ii ) _ those coming from @xmath464 and @xmath465 . let @xmath466 be the indices of the bumps @xmath467 in @xmath73 coming from @xmath119 , i.e. such that @xmath468 and @xmath469 belong to the class _ ( ii)_. one direction of the bijection is then @xmath470 . conversely , given a subset @xmath455 consider the set @xmath471 . split it in maximal - length runs of consecutive integers @xmath472 . for all runs @xmath473 we do the following : to @xmath474 of even cardinality we associate a squiggle @xmath475 of length @xmath476 along @xmath477 ; to @xmath478 of odd cardinality we associate a squiggle @xmath479 of length @xmath476 at @xmath480 ; then we delete in @xmath73 the edges @xmath481 . this process constructs a bump scheme @xmath440 while pruning edges of @xmath73 , giving a path @xmath465 with @xmath482 . these two constructions are inverses , proving the claimed bijection . it now follows that @xmath483 let @xmath144 $ ] be defined by @xmath484 we prove theorem [ thm : main ] by noting that @xmath485 , that @xmath486 , and that for any @xmath444 we have @xmath487 . to prove this last equality , take a path @xmath385 on vertices @xmath488 . then @xmath489 where the sum ranges over all bump schemes for @xmath73 , and @xmath490 it is clear these last two lines are equal ; for the power series expansion of the @xmath491 correspond to all the possible squiggle sequences at @xmath435 , and the power series expansion of the @xmath492 correspond to all possible squiggles along @xmath493 . we give here examples of regular graphs and when possible compute independently the series @xmath25 and @xmath26 . in some cases it will be easier to compute @xmath25 , while in others it will be simpler to compute @xmath26 first . in all cases , once one of @xmath25 and @xmath26 has been computed , the other one can be obtained from corollary [ cor : main ] . in all the examples the graphs are vertex transitive , so the choice of @xmath21 is unimportant . to simplify the computations we choose @xmath494 and the length labelling . let @xmath495 , the complete graph on @xmath496 vertices . its degree is @xmath497 . to compute @xmath25 and @xmath26 , choose three distinct vertices @xmath498 ( the choice is unimportant as @xmath499 is three - transitive ) . define growth series @xmath3 : : the growth series of circuits based at @xmath21 ; @xmath500 : : the growth series of paths @xmath73 from @xmath501 to @xmath21 with @xmath502 ; @xmath503 : : the growth series of paths @xmath73 from @xmath501 to @xmath21 with @xmath504 . then @xmath505,\\ f ' & = t\left[f '' + ( v-3+u)f'\right],\\ f '' & = t\left[1 + ( f-1)\frac{v-2+u}{v-1}\right];\end{aligned}\ ] ] indeed the first line states that a circuit at @xmath21 is either the trivial circuit at @xmath21 , or a choice of one of @xmath506 edges to another point ( call it @xmath501 ) , followed by a path from @xmath501 to @xmath21 ; this path can first go to any vertex of the @xmath507 vertices ( say ( @xmath508 ) different from @xmath21 and @xmath501 , and thus contribute @xmath509 , or go back to @xmath21 and contribute @xmath510 and a bump . the second equation says that a path from @xmath501 to @xmath21 starting by going to @xmath508 can either continue to @xmath21 , contributing @xmath509 , go to any of the @xmath511 other vertices contributing @xmath509 , or come back to @xmath501 , contributing @xmath509 and a bump . the third line says that a path from @xmath501 to @xmath21 starting by going to @xmath21 continues as a circuit at @xmath21 ; but if the circuit is non - trivial , then one out of @xmath506 times a bump will be contributed . solving the system , we obtain @xmath512 we then compute @xmath513 let @xmath514 , the cycle on @xmath291 vertices . here , as there are @xmath196 proper circuits of length @xmath12 for all @xmath12 multiples of @xmath291 , we have @xmath515 obtaining a closed form for @xmath26 is much harder . the number of circuits of length @xmath12 is @xmath516,\,i\equiv n[2 ] } \binom{n}{\frac{n+i}2},\ ] ] from which , by @xcite , it follows that @xmath517 it is not at all obvious how to simplify the above expression . a closed - form answer can be obtained from ( [ eq : c ] ) , namely @xmath518 or , expanding , @xmath519 however in general this fraction is not reduced . to obtain reduced fractions for @xmath3 ( and thus for @xmath1 ) , we have to consider separately the cases where @xmath291 is odd or even . for odd @xmath291 , letting @xmath520 , we obtain @xmath521 where @xmath522,\\ \displaystyle(-)^{\frac{m-1}2}\binom{\ell-\frac{m+1}2}{\frac{m-1}2 } & \text{if } m\equiv1[2 ] . \end{cases}\ ] ] for even @xmath291 , with @xmath523 , @xmath524 expressed as reduced fractions . the first few values of @xmath25 , where @xmath525 stands for @xmath526 , are : [ cols=">,^,>,^",options="header " , ] these rational expressions were computed and simplified using the computer algebra program maple . let @xmath20 be the @xmath0-regular tree . then @xmath527 as a tree has no proper circuit ; while direct ( i.e. , without using corollary [ cor : main ] ) computation of @xmath26 is more complicated . it was first performed by kesten @xcite ; here we will derive the extended circuit series @xmath3 and also obtain the answer using corollary [ cor : main ] . let @xmath528 be a regular tree of degree @xmath0 with a fixed root @xmath21 , and let @xmath529 be the connected component of @xmath21 in the two - tree forest obtained by deleting in @xmath528 an edge at @xmath21 . let @xmath3 and @xmath500 respectively count circuits at @xmath21 in @xmath528 and @xmath529 . for instance if @xmath530 then @xmath509 counts circuits in @xmath531 and @xmath25 counts circuits in @xmath378 . for a reason that will become clear below , we make the convention that the empty circuit is counted as ` @xmath14 ' in @xmath25 and as ` @xmath155 ' in @xmath509 . then we have @xmath532 indeed a circuit in @xmath529 is either the empty circuit ( counted as @xmath155 ) , or a sequence of circuits composed of , first , a step in any of @xmath533 directions , then a ` subcircuit ' not returning to @xmath21 , then a step back to @xmath21 , followed by a step in any of @xmath533 directions ( counting an extra factor of @xmath155 if it was the same as before ) , a subcircuit , etc . if the ` subcircuit ' is the empty circuit , it contributes a bump , hence the convention on @xmath509 . likewise , a circuit in @xmath528 is either the empty circuit ( now counted as @xmath14 ) or a sequence of circuits in subtrees each isomorphic to @xmath529 . we solve these equations to @xmath534 using ( [ eq : c ] ) and @xmath535 we would obtain @xmath536 or , after simplification , @xmath537 which could have been obtained by setting @xmath167 in @xmath538 . in particular if @xmath530 , then @xmath539 and @xmath540 note that for all @xmath0 the @xmath0-regular tree @xmath20 is the cayley graph of @xmath541 with standard generating set . if @xmath0 is even , @xmath20 is also the cayley graph of a free group of rank @xmath542 generated by a free set . we have thus computed the spectral radius of a random walk on a freely generated free group : it is , for @xmath543 or for @xmath544 , equal to @xmath545 note that for @xmath530 the series @xmath3 does have a simple expansion . by direct expansion , we obtain the number of circuits of length @xmath546 in @xmath378 , with @xmath30 local extrema , as @xmath547 , \\ 2\binom{n-1}{\frac{m}{2}}\binom{n-1}{\frac{m-2}{2 } } & \text{if } m\equiv0[2 ] . \end{cases}\ ] ] we may even look for a richer generating series than @xmath25 : let @xmath548[[t]],\ ] ] where @xmath549 denotes the graph distance . then @xmath550 and as @xmath551 is a sum of series counting paths between fixed vertices we obtain @xmath552 from @xmath553 by extending ( [ eq : b ] ) linearly : @xmath554 we could also have started by computing @xmath555 the growth series of all proper paths in @xmath528 , and using ( [ eq : c ] ) and ( [ eq : d ] ) obtain @xmath556 where @xmath557 . recall that the growth series of a graph @xmath20 at a base point @xmath21 is the power series @xmath558 where @xmath549 denotes the distance in @xmath20 . the series @xmath551 is very general in that it contains a lot of information on @xmath528 , namely * @xmath559 ; * @xmath560 is the growth series of @xmath528 ; * @xmath561 is the growth series of all paths in @xmath528 . ( note that these substitutions yield well - defined series because for any @xmath103 there are only finitely many monomials having @xmath41-degree equal to @xmath103 . ) we can also use this series @xmath551 to compute the circuit series @xmath562 of the cycle of length @xmath291 , that was found in the previous section . indeed the universal cover of a cycle is the regular tree @xmath528 of degree @xmath196 , and circuits in @xmath455 correspond bijectively to paths in @xmath528 from @xmath21 to any vertex at distance a multiple of @xmath291 . we thus have @xmath563 where the sum runs over all @xmath291th roots of unity and @xmath530 in @xmath551 . we consider next the following graphs : take a @xmath0-regular tree and fix a vertex @xmath21 . at @xmath21 , delete @xmath60 vertices and replace them by @xmath60 loops . then clearly @xmath564 as all the non - backtracking paths are constrained to the @xmath60 loops . using ( [ eq : c ] ) , we obtain after simplifications @xmath565 the radius of convergence of @xmath26 is @xmath566 ===in this subsection we outline the computations of @xmath25 and @xmath26 for more complicated graphs . they are only provided as examples and are logically independent from the remainder of the paper . the arguments will therefore be somewhat condensed . first take for @xmath20 the cayley graph of @xmath567 with generators @xmath568 , @xmath569 and @xmath570 . geometrically , @xmath20 is a doubly - infinite two - poled ladder . in subsection [ subs : trees ] we computed @xmath571 the growth of circuits restricted to one pole of the ladder . a circuit in @xmath20 is a circuit in @xmath378 , before and after each step ( @xmath572 or @xmath573 ) of which we may switch to the other pole ( with a @xmath574 ) as many times as we wish , subject to the condition that the circuit finish at the same pole as it started . this last condition is expressed by the fact that the series we obtain must have only coefficients of even degree in @xmath41 . thus , letting @xmath575 , we have @xmath576 it is then simple to obtain @xmath3 by performing the substitution ( [ eq : c ] ) . the following direct argument also gives @xmath3 : a walk on the ladder is obtained from a walk on a pole ( i.e. on @xmath378 ) by inserting before and after every step on a pole a ( possibly empty ) sequence of steps from one pole to the other . this process is expressed by performing on @xmath577 the substitution @xmath578 corresponding to replacing a step on a pole by itself , or itself followed by a step to the other pole , or itself , a step to the other pole and a step back , etc . but if the path had a bump at the place the substitution was performed , this bump would disappear when a step is added from one pole to the other . in formulas , @xmath579 finally we must add at the beginning of the path a sequence of steps from one pole to the other . therefore we obtain @xmath580 as another example , consider the group @xmath378 generated by the non - free set @xmath581 . geometrically , it can be seen as the set of points @xmath582 and @xmath583 for all @xmath584 , with edges between all points at euclidean distance @xmath196 apart ; but we will not make use of this description . the circuit series of @xmath378 with this enlarged generating set will be an algebraic function of degree @xmath280 over the rationals . define first the following series : @xmath261 : : counts the walks from @xmath129 to @xmath129 in @xmath531 ; @xmath585 : : counts the walks from @xmath129 to @xmath14 in @xmath531 ; @xmath586 : : counts the walks from @xmath14 to @xmath14 in @xmath531 . = = = = denote the generators of @xmath378 by @xmath587 , @xmath588 , @xmath589 and @xmath590 . the series then satisfy the following equations , where the generators symbol is written instead of ` @xmath41 ' to make the formulas self - explanatory : @xmath591 giving a solution @xmath46 that is algebraic of degree @xmath280 over @xmath592 . then define the following series : @xmath26 : : counts the walks from @xmath129 to @xmath129 in @xmath378 ; @xmath60 : : counts the walks from @xmath129 to @xmath14 in @xmath378 . they satisfy the equations @xmath593 giving the solution @xmath594 where @xmath595 is a root of the equation @xmath596 we perform here a computation extending the results of section [ subs : gptheory ] . the general setting , expressed in the language of group theory , is the following : let @xmath37 be a group generated by a finite set @xmath7 and let @xmath217 be any subgroup . we consider the following generating series : @xmath597 where @xmath598 is the minimal length of @xmath465 in the generators @xmath7 , and @xmath599 is the usual length of the word @xmath322 . is there some relation between these series ? in case @xmath37 is quasi - free on @xmath7 , the relation between @xmath25 and @xmath26 is given by corollary [ cor : main ] . we consider two other examples : @xmath37 quasi - free but on a set smaller than @xmath7 , and @xmath600 . let @xmath7 , @xmath339 be finite sets , and @xmath72 an involution on @xmath7 . consider the two presentations @xmath602 let @xmath217 be any subgroup , and let @xmath509 and @xmath603 be the generating series related to the first presentation . clearly @xmath604 , as both series count the same objects in @xmath37 ( regardless of @xmath37 s presentation ) ; while @xmath605 indeed any word @xmath606 in @xmath607 defining an element of @xmath38 can be uniquely decomposed as @xmath608 where @xmath609 , @xmath610 are words in @xmath339 for all @xmath103 , and @xmath611 defines an element of @xmath38 ; moreover all choices of @xmath611 defining an element of @xmath38 and words @xmath610 in @xmath339 give a distinct word @xmath322 . it then suffices to note that the generating series for any of the @xmath610 is @xmath612 . putting everything together , we obtain : let @xmath37 be as above , @xmath217 a subgroup . then @xmath613 let @xmath615 and let @xmath217 be any subgroup . we take @xmath616 . we suppose @xmath38 is torsion - free , i.e. contains no element of the form @xmath617 or @xmath618 . let @xmath20 be the schreier graph of @xmath619 relative to @xmath38 , as defined in subsection [ subs : gptheory ] ; this is a trivalent graph whose vertex set is @xmath219 . its vertices can be grouped in triples @xmath620 connected in triangles . let @xmath621 be the graph obtained from @xmath20 by identifying each triple to a vertex . explicitly , @xmath622 the involution on @xmath623 is the switch @xmath624 and the extremity functions @xmath625 are the natural projections . note that @xmath621 is a @xmath626-regular graph ( for instance , @xmath627 is connected to @xmath628 , @xmath629 and @xmath630 ) . in case @xmath210 , it is the @xmath626-regular tree . by construction we have a @xmath626-to-@xmath14 map @xmath631 . we fix an origin @xmath632 in @xmath621 , and let @xmath633 be the circuit series of @xmath634 . let @xmath635 be a triangle , @xmath636 count the circuits at a fixed vertex of @xmath635 and @xmath637 count paths between two fixed distinct vertices of @xmath635 . these series were computed in section [ subs : complete ] , with @xmath638 . circuits at @xmath21 in @xmath20 can be projected to circuits at @xmath21 in @xmath621 simply by deleting all edges of type @xmath639 and projecting the other edges through @xmath640 . conversely , circuits in @xmath621 can be lifted to @xmath20 by lifting the edges through @xmath641 , and connecting them in @xmath20 with arbitrary paths remaining inside the triples ; to lift the path @xmath385 from @xmath621 to @xmath20 , choose edges @xmath642 with @xmath643 and @xmath644 for all @xmath77 , and choose , for all @xmath433 , paths @xmath645 from @xmath646 to @xmath647 remaining inside @xmath648 , where by convention @xmath649 . then the lift corresponding to these choices is @xmath650 furthermore all circuits at @xmath21 in @xmath20 can be obtained this way . define @xmath651 as the series counting paths that start and finish at a vertex in the same triple as @xmath21 . it can be obtained using ( [ eq : shuffle ] ) by letting @xmath89 range over all paths in @xmath621 , and for each choice of @xmath89 and for each @xmath652 letting @xmath645 range over @xmath653 or @xmath654 depending on whether @xmath89 has or not a bump at @xmath103 , and letting @xmath655 and @xmath656 range over all paths inside the triple @xmath657 . in equations , this relation is expressed as @xmath658 now the series @xmath26 we wish to obtain is approximately @xmath659 : for any choice of @xmath660 there are approximately the same number of long enough paths from @xmath43 to @xmath94 . a summand of @xmath315 is the unique lifting of a summand of @xmath661 , but is twice longer in @xmath20 than in @xmath621 . two series @xmath662 , @xmath663 are _ equivalent _ , written @xmath664 , if they have the same radius of convergence @xmath89 , and there exists a constant @xmath665 such that @xmath666 then the remarks of the previous paragraph can be written as @xmath667 letting @xmath668 be the circuit series of @xmath621 , we use corollary [ cor : main ] to obtain @xmath669 ( 300,300 ) ( 0,170)(0,0)[lb ] ( 150,30)(0,0)[lb ] ( -30,-10 ) let @xmath20 be a simplicial complex such that at each vertex an edge and a ( filled - in ) triangle meet ; choose a base point @xmath21 in @xmath20 . say a circuit in the @xmath14-skeleton of @xmath20 is _ reduced _ if it contains no bump nor two successive edges in the same triangle ; thus reduced circuits are in bijection with homotopy classes of @xmath153 . let @xmath315 be the proper circuit series and @xmath1 the circuit series of @xmath20 . let @xmath670 we have proved the following theorem and corollary , similar to those in section [ subs : gptheory ] : @xmath671 . let @xmath38 be a subgroup of @xmath672 ; let @xmath251 be the spectral radius of the simple random walk on @xmath219 , and @xmath227 the `` cogrowth '' rate of @xmath219 . then provided that @xmath673 $ ] , where @xmath89 is the word growth of @xmath37 , namely @xmath674 , we have @xmath675 so @xmath676 the function @xmath236 is monotonously increasing between @xmath129 and @xmath677{2}$ ] , where it reaches its maximum . the same argument applies as that given in the proof of corollary [ cor : homo ] . we now state the same results for an arbitrary virtually free group with an appropriate generating system . let @xmath37 be a virtually free group , such that there is a split exact sequence @xmath678 where @xmath679 is a finite group and @xmath117 has a presentation @xmath680 we assume further that @xmath37 is generated by a set @xmath681 with @xmath682 in bijection through @xmath73 with @xmath683 , @xmath684 mapping through @xmath73 to @xmath685 , and @xmath686 in bijection with @xmath7 through @xmath687 . for example , consider @xmath688 . take @xmath689 and @xmath690 , take @xmath691 and @xmath692 . then the hypotheses are satisfied . with these hypotheses , the cayley graph @xmath20 of @xmath37 is a collection of complete graphs of size @xmath693 , with at each vertex @xmath694 edges leaving to other complete graphs , and such that if each of these complete graphs is shrunk to a point the resulting graph is a tree . the following theorem is then a straightforward generalization of the argument given for @xmath39 . with the notation introduced above , let @xmath38 be any subgroup of @xmath37 not intersecting @xmath695 and let @xmath315 , @xmath1 be the `` cogrowth '' series and circuit series of @xmath219 . let @xmath635 be the complete graph on @xmath693 vertices and let @xmath636 , @xmath637 count the circuits and the non - closing paths respectively in @xmath635 . define the function @xmath236 by @xmath696 then we have @xmath697 we give here a general construction combining two pointed graphs and show how to compute the generating functions for circuits in the `` product '' in terms of the generating functions for circuits in the factors . let @xmath698 and @xmath634 be two connected pointed graphs . their _ free product _ @xmath699 is the graph constructed as follows : start with copies of @xmath635 and @xmath621 identified at @xmath21 ; at each vertex @xmath359 apart from @xmath21 in @xmath635 , respectively @xmath621 , glue a copy of @xmath621 , respectively @xmath635 , by identifying @xmath359 and the @xmath21 of the copy . repeat the process , each time gluing @xmath635 s and @xmath621 s to the new vertices . if @xmath700 , @xmath701 are two groups with fixed generators whose cayley graphs are @xmath635 and @xmath621 respectively , then @xmath699 is the cayley graph of @xmath702 . we will now compute the circuit series of @xmath699 in terms of the circuit series of @xmath635 and @xmath621 . let @xmath653 , @xmath668 and @xmath703 be the generating functions counting circuits in @xmath635 , @xmath621 and @xmath704 respectively . we will use the following description : given a circuit at @xmath21 in @xmath20 , it can be decomposed as a product of circuits never passing through @xmath21 . each of these circuits , in turn , starts either in the @xmath635 or the @xmath621 copy at @xmath21 . say one starts in @xmath635 ; it can then be expressed as a circuit in @xmath635 never passing through @xmath21 , and such that at all vertices , except the first and last , a circuit starting in @xmath621 has been inserted . moreover , any choice of such circuits satisfying these conditions will give a circuit at @xmath21 in @xmath20 , and different choices will yield different circuits . let @xmath705 ( respectively @xmath706 ) be the generating function counting non - trivial circuits in @xmath635 ( respectively @xmath621 ) never passing through @xmath21 . obviously @xmath707 let @xmath708 ( respectively @xmath709 ) be the generating function counting non - trivial circuits in @xmath20 never passing through @xmath21 and starting in @xmath635 ( respectively @xmath621 ) . then @xmath710 indeed write @xmath711 . then by the description given above @xmath712 which is precisely the given formula . finally @xmath713 writing @xmath714 , and similarly for @xmath621 , we simplify these equations to @xmath715 if @xmath46 is a power series with @xmath716 and @xmath717 , let us write @xmath718 for the inverse of @xmath46 ; i.e. for the series @xmath719 with @xmath720 ( for instance , @xmath721 is equal to its inverse ; the inverse of @xmath722 is @xmath723 ) . from ( [ eq : free1 ] ) we obtain @xmath724 and @xmath725 ; so composing ( [ eq : free1 ] ) with @xmath726 we obtain the [ thm : free ] @xmath727 an equation equivalent to this one , though not trivially so , appeared in a paper by gregory quenell @xcite , and , in yet another language , in a paper by dan voiculescu ( * ? ? ? * theorem 4.5 ) . we can use ( [ eq : free ] ) to obtain by a different method the circuit series of regular trees ( see section [ subs : trees ] ) . indeed the free product of regular trees of degree @xmath0 and @xmath60 is a regular tree of degree @xmath728 . letting @xmath729 denote the circuit series of a regular tree of degree @xmath0 , we `` guess '' that @xmath730 and verify that the limit @xmath731 is indeed the circuit series of the @xmath14-regular tree . then we compute @xmath732 if we define @xmath733 by @xmath734 it satisfies @xmath735 and we have proved that our guess of @xmath729 is correct , in light of ( [ eq : free ] ) . as another application of ( [ eq : free ] ) , we will compute the circuit series @xmath1 of the cayley graph of @xmath736 with generators @xmath737 . this graph is the free product of the @xmath14-regular tree @xmath635 and of the @xmath626-cycle @xmath621 . we know from section [ subs : complete ] that @xmath738 are the circuit series of @xmath635 and @xmath621 . we then compute @xmath739 so after some lucky simplifications @xmath740 ( a closed form such as this one does not exist for @xmath741 with @xmath742 , because then the series @xmath26 is algebraic of degree greater than 2 . ) if the circuit series of @xmath635 and @xmath621 are both algebraic , then the circuit series of @xmath699 is also algebraic . sums and products of algebraic series are algebraic . if @xmath46 satisfies the algebraic relation @xmath743 , then its formal inverse satisfies the relation @xmath744 so is also algebraic . we have @xmath745 where the supremum is taken over all @xmath41 such that the series @xmath746 converges . if @xmath46 is @xmath89-recurrent , then also @xmath747 clearly @xmath748 ; if @xmath749 converges over @xmath750 then @xmath746 converges over @xmath751 where @xmath752 ; then @xmath753 . the second assertion follows because in this case @xmath754 . let the circuit series of @xmath635 , @xmath621 and @xmath704 be @xmath653 , @xmath668 and @xmath703 respectively , and suppose all three series are recurrent . then @xmath755 this follows from @xmath756 note that the corollary does not extend to non - recurrent series ; for instance , it fails if @xmath757 . indeed then @xmath758 there are two natural definitions for _ direct products _ of graphs ; they correspond to direct products of groups with generating set either the union or cartesian product of the generating sets of the factors . a general treatment of products of graphs can be found in ( * ? ? ? * and 203 ) if @xmath7 is a set , the _ stationing graph _ on @xmath7 is the graph @xmath759 with @xmath760 , where for the edges @xmath761 hold . let @xmath20 be a graph , and @xmath762 be the graph obtained by adding a loop to every vertex in @xmath20 . let @xmath719 and @xmath60 be the growth functions for circuits in @xmath20 and @xmath635 respectively . then we have @xmath763 let @xmath635 and @xmath621 be two graphs . their _ direct product _ @xmath764 is defined by @xmath765 and @xmath766 note that if the graphs @xmath635 and @xmath621 have respectively adjacency matrices @xmath767 and @xmath25 , then their product has adjacency matrix @xmath768 . in that case we have @xmath769 this is a simple application of the laplace transform , that converts an exponential generating function into an ordinary one and vice versa @xcite . indeed , if we had considered exponential generating functions , the formula would simply have been @xmath770 , as is well known @xcite . as an example , let @xmath757 , so @xmath771 then @xmath772 where @xmath665 is the complete elliptic function and @xmath25 the hypergeometric series . these functions are known to be transcendental ; thus the circuit series of @xmath773 is transcendental . this result appears in @xcite . numerical evidence suggests the growth function for @xmath774 is not even hypergeometric . let @xmath635 and @xmath621 be two graphs , and suppose that for every vertex in @xmath635 and @xmath621 there is a loop at it . then their _ direct product _ @xmath764 is defined by @xmath765 and @xmath775 note that if the graphs @xmath635 and @xmath621 have respectively adjacency matrices @xmath767 and @xmath25 , then their product has adjacency matrix @xmath776 . in that case we have , again using laplace transformations @xmath777 note that with both definitions of products it is possible that the growth function for circuits in the product be transcendental even if the growth functions for circuits in the factors are algebraic . it was mentioned in subsection [ sec : lang ] how the main result applies to languages . this a special case of a much more general problem : given a language @xmath324 and a set @xmath778 of words , define the _ desiccation _ @xmath779 of @xmath324 as the set of words in @xmath324 containing no @xmath780 as a subword . give sufficient conditions on @xmath324 and @xmath778 such that a formula exist between @xmath334 and @xmath781 . the special case we studied in this paper is that of @xmath782 and a sufficient condition is that @xmath324 be saturated . for general @xmath778 this is not always sufficient : let @xmath783 and @xmath784 be the set of words with an even number of @xmath785 s . then if @xmath786 there are @xmath787 desiccated words of length @xmath788 : @xmath789 ; and if @xmath790 there are @xmath791 desiccated words of length @xmath788 : @xmath792 . the growth series of @xmath778 and @xmath793 are the same , namely @xmath794 , but the growth series of @xmath779 and @xmath795 differ in their degree-@xmath788 coefficient . we gave in section [ sec : freeamalg ] a formula relating the circuit series of a free product to the circuit series of its factors . there is a notion of _ amalgamated product _ of graphs , that is a direct analogue of the amalgamated product of groups . what conditions on @xmath796 are sufficient so that @xmath797 where @xmath798 is an amalgamated product of @xmath635 and @xmath621 along @xmath799 ? the formula holds if @xmath799 is the trivial graph ; but it can not hold in general . if @xmath800 is the `` ladder graph '' described in section [ ex : ladder ] : the set of points @xmath801 with @xmath584 and @xmath802 , with edges connecting all pairs of vertices at euclidean distance @xmath14 , and @xmath799 is @xmath378 , embedded as a pole of the ladder , then the amalgamated product @xmath798 is isomorphic to @xmath773 . the circuit series of @xmath799 , @xmath635 and @xmath621 have been calculated explicitly and are algebraic . the circuit series of @xmath20 was shown in section [ sec : direct ] to be transcendental ; so there can exist no algebraic definition of @xmath703 in terms of @xmath803 , @xmath653 and @xmath668 . however , there exists some relations between these series , as given by ( * ? ? ? * theorem 5.5 ) . given a graph @xmath20 , one can construct a graph @xmath804 on the same vertex set , and with edge set the set of paths of length @xmath805 . is there some simple relation between the path series of @xmath20 and of @xmath804 ? this could be useful for example to obtain asymptotics about the cogrowth of a group subject to enlargement of generating set @xcite . the equation ( [ eq : free ] ) corresponds to voiculescu s @xmath184-transform @xcite . his @xmath7-transform , in terms of graphs , corresponds to @xmath699 with as edge set all sequences @xmath390 and @xmath806 , for @xmath807 and @xmath808 . is there an analogue to theorem [ thm : free ] in this context ? finally , ( [ eq : free ] ) computes the circuit series of a free product in terms of the circuit series of the factors . a more complicated formula yields the path series of a free product in terms of the path series of the factors . such considerations give another derivation of the results in section [ sec : nonfree ] . the main result of this paper was found in rome thanks to the nurturing of tullio ceccherini and his family whom i thank . many people heard or read preliminary often obscure versions and provided valuable feedback ; i am grateful to ( in order of appearance ) michel kervaire , shalom eliahou , pierre de la harpe , fabrice liardet , rostislav grigorchuk , alain valette , tienne ghys , igor lysionok , jean - paul allouche , gilles robert , and thierry vust for their help . rostislav i. grigorchuk , _ symmetric random walks on discrete groups _ , multi - component random systems ( r. l. dobrushin and ya . g. sinai , eds . ) , nauka , moscow , 1978 , english translation : adv . ( d. griffeath ed . ) vol 6 , m. dekker 1980 , 285325 , pp . 132152 . david e. mller and paul e. schupp , _ context - free languages , groups , the theory of ends , second - order logic , tiling problems , cellular automata , and vector addition systems _ , bull . amer . math . * 4 * ( 1981 ) , 331334 .	we give a simple combinatorial proof of a formula that extends a result by grigorchuk @xcite ( rediscovered by cohen @xcite ) relating cogrowth and spectral radius of random walks . our main result is an explicit equation determining the number of ` bumps ' on paths in a graph : in a @xmath0-regular ( not necessarily transitive ) non - oriented graph let the series @xmath1 count all paths between two fixed points weighted by their length @xmath2 , and @xmath3 count the same paths , weighted as @xmath4 . then one has @xmath5 we then derive the circuit series of ` free products ' and ` direct products ' of graphs . we also obtain a generalized form of the ihara - selberg zeta function @xcite
strange quarks represent the lightest quark flavor not present in the valence component of the nucleon . their study can thus provide insight into sea quark effects in the nucleon in isolated fashion . the two most fundamental properties of the nucleon are its mass and spin . the investigation presented here focuses on the extent to which those two properties are influenced by the strange quark degrees of freedom . the strange contributions to nucleon mass and spin can be characterized by the matrix elements @xmath9 \label{matel1}\ ] ] and @xmath10 respectively , where @xmath11 denotes a nucleon state with spin polarized in the @xmath12-direction . in the case of the scalar matrix element , the vacuum expectation value , i.e. , the vacuum strange scalar condensate , is subtracted ; the intention is , of course , to measure the strangeness content of the nucleon _ relative _ to the vacuum . in the case of the axial matrix element , no subtraction is necessary since the corresponding vacuum expectation value vanishes . note that @xmath1 measures specifically the contribution of strange quark spin to nucleon spin ; strange quark angular momentum constitutes a separate contribution not considered here . aside from representing a fundamental characteristic of the nucleon in its own right , the scalar strange content @xmath0 is also an important parameter in the context of dark matter searches @xcite . assuming that the coupling of dark matter to baryonic matter is mediated by the higgs field , the dark matter detection rate depends sensitively on the quark scalar matrix elements in the nucleon , cf . , e.g , the neutralino - nucleon scalar cross - section considered in @xcite . one a priori reasonable scenario is that the strange quark furnishes a particularly favorable channel @xcite , since , on the one hand , it features a much larger yukawa coupling to the higgs field than the light quarks , and , on the other hand , is not so heavy as to be only negligibly represented in the nucleon s sea quark content . as a consequence , an accurate estimate of @xmath0 is instrumental in assessing the discovery potential for dark matter candidates . the contribution of strange quark spin to nucleon spin @xmath1 is , in principle , more directly accessible to experiment than @xmath0 . @xmath1 represents the first moment of the strange quark helicity distribution @xmath13 ( including both quarks and antiquarks ) as a function of the momentum fraction @xmath14 . the helicity distribution can be determined via inclusive deep inelastic scattering and semi - inclusive deep inelastic scattering @xcite . however , its extraction in practice still has to rely to a certain extent on assumptions about the dependence of @xmath13 on @xmath14 , even in the semi - inclusive channels ( which furnish direct information on @xmath13 ) , because of the limitations in accessing small @xmath14 experimentally . complementary information about @xmath1 is obtained from the strange axial form factor of the nucleon @xmath15 , which can be extracted by combining data from parity - violating elastic electron - proton scattering and elastic neutrino - proton scattering @xcite . extrapolation to zero momentum transfer , @xmath16 , again yields an estimate of @xmath1 . depending on the specific extrapolations and/or model assumptions adopted in determining @xmath1 via the various aforementioned avenues , both significantly negative values for @xmath1 have been put forward @xcite , as well as values compatible with zero @xcite . an independent determination of @xmath1 via lattice qcd , as undertaken in the present work , thus can be useful in several ways . apart from shedding light on the fundamental question of the decomposition of nucleon spin , it can contribute constraints to phenomenological fits of polarized parton distribution functions . furthermore , it influences spin - dependent dark matter cross sections @xcite ; although more accurate determinations of the scalar matrix elements discussed further above constitute the most urgent issue in reducing hadronic uncertainties in dark matter searches , @xmath1 also plays a significant role in that context . a number of lattice qcd investigations of strange quark degrees of freedom in the nucleon have recently been undertaken@xcite , the majority of which have focused specifically on the scalar content . studies of the latter have proceeded via two avenues : on the one hand , one can directly determine the matrix element @xmath17 via the appropriate disconnected three - point function ; this methodology was adopted in @xcite and also in the present work , as described in detail further below . a study of techniques suited to improve the efficiency of this approach has been presented in @xcite . on the other hand , a somewhat less direct inference of the scalar strange quark content of the nucleon is possible via the study of the baryon spectrum , which is related via the feynman - hellmann theorem @xmath18 to the corresponding sigma terms for the baryon state @xmath19 and quark flavor @xmath20 . this avenue has been pursued in @xcite , and a related methodology , combining lattice hadron spectrum data with chiral perturbation theory , was pursued in @xcite . the characteristics of these various investigations of the scalar strange quark content of the nucleon are diverse . they include @xmath21 calculations , in which the strange quark degrees of freedom are quenched @xcite , but also @xmath22 @xcite and even @xmath23 @xcite calculations . in some cases , lattice data at only one pion mass have been obtained to date and no extrapolation to the physical point has been attempted . the most stringent results obtained at the physical point including fully dynamical strange quarks were reported in @xcite . ref . @xcite quotes @xmath24 in the @xmath22 case , and @xmath25 in the @xmath23 case ; translated to @xmath0 itself using @xmath26 in the @xmath5 scheme at a scale of @xmath6 @xcite , these correspond to @xmath27 for @xmath22 and @xmath28 for @xmath23 . on the other hand , @xcite and @xcite report significantly lower values . in @xcite , both direct three - point function calculations as well as indirect determinations via the feynman - hellmann theorem are used to arrive at a bound @xmath29 . in @xcite , a more indirect analysis using lattice hadron spectrum data and chiral perturbation theory yields @xmath30 , which translates to @xmath31 . the results obtained in the present work are of comparable accuracy , @xmath32 , cf . ( [ ftsfinal ] ) , and are more consistent with the higher values reported in @xcite . the strange axial matrix element @xmath1 , on the other hand , has also been investigated in @xcite . apart from the exploratory study @xcite , which , however , did not attempt to renormalize the results nor extrapolate them to the physical pion mass , these investigations were based on dynamical quark ensembles containing only the two light flavors in the quark sea ; the present lattice investigation , on the other hand , employs @xmath22 gauge ensembles , thus treating the strange quark degrees of freedom in dynamical fashion . the numerical results for @xmath1 obtained on this basis are renormalized and chirally extrapolated , yielding the estimate @xmath7 , cf . ( [ dsfinal ] ) , at the physical point in the @xmath5 scheme at a scale of @xmath6 . within the uncertainties , this nevertheless remains compatible with the values obtained in the aforementioned other studies , though it is about 50% larger in magnitude . this suggests that systematic adjustments to the results quoted in those works through unquenching of the strange quark degrees of freedom , renormalization , and chiral extrapolation are not severe . aside from the two quantities @xmath0 and @xmath1 considered in the present work , lattice investigations of the strange quark structure within the nucleon have also considered generalizations to non - zero momentum transfer , i.e. , form factors @xcite , including calculations of the strange electric and magnetic form factors , which are of interest in the context of corresponding experimental efforts employing parity - violating electron - proton scattering @xcite . furthermore , also the strange quark momentum fraction and strange quark angular momenta in the nucleon have been investigated @xcite . the present lattice investigation , a preliminary account of which was given in @xcite , is based on a mixed - action scheme developed and employed extensively by the lhp collaboration @xcite . the nucleon valence quarks as well as the strange quark fields appearing in the operator insertions in eqs . ( [ matel1 ] ) and ( [ matel2 ] ) are realized as domain wall fermions , propagators of which are evaluated in the background of ( hyp - smeared ) 2 + 1-flavor dynamical asqtad quark ensembles provided by the milc collaboration . though computationally expensive , domain wall fermions lead to benign renormalization properties which prove especially advantageous in the case of the @xmath0 observable ; the substantial systematic uncertainties due to strong additive quark mass renormalizations encountered in analogous calculations using wilson fermions @xcite are avoided . the computational scheme is described in detail in section [ schemesec ] . section [ resultsec ] provides the raw numerical results , the renormalization of which is discussed in section [ renormsec ] . the renormalized results are extrapolated to the physical pion mass in section [ chisec ] , and systematic uncertainties and adjustments are considered in section [ syssec ] , with conclusions presented in section [ concsec ] . the lattice calculation of the nucleon matrix elements in ( [ matel1 ] ) , ( [ matel2 ] ) proceeds in standard fashion via correlator ratios of the type @xmath33 ( \tau , t ) = \frac{\left\langle \ \left [ \gamma^{nuc}_{\alpha \beta } \ \sigma_{\vec{x } } \ n_{\beta } ( \vec{x } , t ) \bar{n}_{\alpha } ( 0,0 ) \right ] \cdot \left [ - \gamma^{obs}_{\gamma \rho } \sigma_{\vec{y } } \ s_{\rho } ( \vec{y } , \tau ) \bar{s}_{\gamma } ( \vec{y } , \tau ) \right ] \ \right\rangle } { \left\langle \ \gamma^{unpol}_{\alpha \beta } \ \sigma_{\vec{x } } \ n_{\beta } ( \vec{x } , t ) \bar{n}_{\alpha } ( 0,0 ) \ \right\rangle } \label{corrrat}\ ] ] with nucleon interpolating fields @xmath34 of the standard form ( quoting here , for definiteness in flavor structure , the proton case ) @xmath35 where @xmath36 denotes the charge conjugation matrix . the sums over spatial position @xmath37 project the nucleon states onto zero momentum , whereas the sum over spatial position @xmath38 is simply transcribed from ( [ matel1 ] ) , ( [ matel2 ] ) . since the nucleon contains no strange valence quarks , the three - point function averaged in the numerator of ( [ corrrat ] ) factorizes , as written , into the nucleon two - point function and the strange quark loop . i.e. , only the disconnected diagram , cf . [ ddiagram ] , contributes to the matrix elements under consideration . the strange quark fields have already been reordered in the numerator of ( [ corrrat ] ) such as to make the standard minus sign associated with quark loops explicit . finally , the @xmath42 matrices allow one to choose the appropriate nucleon polarization and strange quark operator insertion structures . the denominator of ( [ corrrat ] ) corresponds to the unpolarized nucleon two - point function , obtained using @xmath43 in the numerator , for the purpose of evaluating @xmath0 , unpolarized nucleon states are appropriate , corresponding to the choice @xmath44 ; furthermore , the scalar strange quark insertion is obtained by choosing @xmath45 , @xmath46 ( \tau , t ) \ - \ [ \mbox{vev } ] \right ) \ \equiv \ r\ { f_{t_s } \ } \ \stackrel{t \gg \tau \gg 0}{\longrightarrow } \ f_{t_s } \label{corrfts}\ ] ] with the vacuum expectation value @xmath47 = \langle -\sigma_{\vec{y } } \ s_{\gamma } ( \vec{y } , \tau ) \bar{s}_{\gamma } ( \vec{y } , \tau ) \rangle\ ] ] to be subtracted . on the other hand , @xmath1 is obtained by using the projector onto nucleon states polarized in the positive / negative @xmath12-direction in the numerator of ( [ corrrat ] ) , @xmath48 as well as the operator insertion structure @xmath49 . averaging over positive / negative @xmath12-direction ( with a relative minus sign ) as well as the three spatial @xmath12 to improve statistics leads one to evaluate @xmath50 ( \tau , t ) \ \equiv \ r\ { \delta s \ } \ \stackrel{t \gg \tau \gg 0}{\longrightarrow } \ \delta s \ , \label{corrds}\ ] ] where the prefactor @xmath51 compensates for the fact that the ratio ( [ corrrat ] ) is normalized using the unpolarized nucleon two - point function in the denominator , even when the numerator is restricted to a particular polarization . lastly , the prefactor @xmath52 cancels the additional factor @xmath53 which the @xmath54-matrix in the operator insertion acquires when the calculation is cast in terms of euclidean lattice correlators ; thus , the @xmath55 obtained through ( [ corrds ] ) is already wick - rotated back to minkowski space - time . note that ( [ corrds ] ) does not call for the subtraction of a vacuum expectation value , since the latter vanishes in the ensemble average . in practice , this numerical zero was nevertheless subtracted from ( [ corrds ] ) in order to reduce statistical fluctuations . the averages in the correlator ratios ( [ corrrat ] ) were carried out using the three milc @xmath56-flavor dynamical asqtad quark ensembles listed in table [ tabmilc ] . hyp - smearing was applied to the configurations . .@xmath22 , @xmath57 milc asqtad ensembles with lattice spacing @xmath58 used in the present investigation . uncertainties in the pion and nucleon masses extracted from the corresponding two - point functions are under @xmath59 . [ cols="^,^,^,^,^,^,^",options="header " , ] it should be noted that all these results were obtained at a constant lattice spacing @xmath60 . fitting parabolae to the data in table [ mrestab ] yields the following derivatives at the @xmath61-flavor symmetric point @xmath62 and at the lightest @xmath63 considered in this work , @xmath64 : @xmath65 the simplest estimate for @xmath66 from this can be obtained by noting that , at the @xmath61-flavor symmetric point , @xmath67 ( where the factor 1/2 stems from the fact that @xmath68 ) . assuming that the derivative of @xmath69 in the @xmath70-direction varies only weakly as one changes @xmath63 , one arrives at the estimate that also @xmath71 note that this is most likely an upper bound in magnitude , since one would expect the characteristics of the gauge fields entering the dirac operator to be dominated by the light quark degrees of freedom relative to the strange quark degrees of freedom as @xmath63 becomes smaller . a check on this estimate can be constructed by the following alternative chain of reasoning . first , note that the quantity @xmath72 contains two contributions , namely , one from the explicit variation of @xmath63 in the dirac operator which @xmath69 characterizes , and the other from the implicit dependence of the gauge field ensemble on @xmath63 . to estimate @xmath73 , one therefore needs to apply two correction factors to @xmath72 : a factor characterizing the proportion of the variation of @xmath69 due specifically to the implicit variation of the gauge fields , and a factor characterizing the strength of that variation with @xmath70 as opposed to @xmath63 . the order of magnitude of these correction factors can be inferred as follows . the former factor is available at the @xmath61-flavor symmetric point as the ratio @xmath74 for the purposes of the present argument , it will be assumed that this factor only varies mildly as @xmath75 is lowered to @xmath76 . on the other hand , assume also that @xmath77 i.e. , the implicit variation of @xmath78 via the dependence of the gauge fields on @xmath70 changes only mildly as a function of @xmath63 . note that , while this assumption is analogous to the one leading to ( [ simpest ] ) , it is better founded since @xmath78 itself varies less with @xmath63 than @xmath69 . again , one would expect the left - hand side to represent an upper bound for the right - hand side . noting that the left - hand side is identical to one - half the quantities in the left - hand identity in ( [ slderivs ] ) , one thus has @xmath79 where in the final step it is assumed that changes in the numerator and denominator due to changing the quark mass in the dirac operator from @xmath70 to @xmath63 approximately cancel ) and ( [ sverslvar ] ) to differ by somewhat less than a factor of two , by comparing the right - hand identity in ( [ slderivs ] ) with the right - hand identity in ( [ llderivs ] ) , corrected by the factor ( [ implicvsfull ] ) . ] . this is the desired conversion factor characterizing the strength of the implicit variation of @xmath69 with @xmath70 relative to the one with @xmath63 . correcting , as proposed above , @xmath72 by the two factors ( [ implicvsfull ] ) and ( [ sverslvar ] ) , one finally arrives at the alternative estimate @xmath80 consistent in order of magnitude with ( [ simpest ] ) , especially in view of the latter being expected to represent an overestimate . the estimate ( [ simpest ] ) for @xmath66 at the lowest light quark mass considered in the numerical calculations in this work is used in section [ scalrensec ] to constrain the influence of operator mixing effects in the renormalization of @xmath0 . 99 a. bottino , f. donato , n. fornengo and s. scopel , astropart . phys . * 18 * ( 2002 ) 205 . j. ellis , k. a. olive , y. santoso and v. c. spanos , phys . d 71 * ( 2005 ) 095007 . j. ellis , k. a. olive and c. savage , phys . rev . * d 77 * ( 2008 ) 065026 . g. bertone , ed . , _ `` particle dark matter : observations , models and searches '' _ , cambridge university press , 2010 . a. airapetian et al . ( hermes collaboration ) , phys . rev . * d 75 * ( 2007 ) 012007 . a. airapetian et al . ( hermes collaboration ) , phys . * b666 * ( 2008 ) 446 . m. alekseev et al . ( compass collaboration ) , phys . * b680 * ( 2009 ) 217 . s. pate , d. mckee and v. papavassiliou , phys . * c 78 * ( 2008 ) 015207 . m. alekseev et al . ( compass collaboration ) , phys . * b660 * ( 2008 ) 458 . g. bali , s. collins and a. schfer , comput . * 181 * ( 2010 ) 1570 . g. bali , s. collins , m. gckeler , r. horsley , y. nakamura , a. nobile , d. pleiter , p. rakow , a. schfer , g. schierholz and j. zanotti ( qcdsf collaboration ) , phys . * 108 * ( 2012 ) 222001 . g. bali , s. collins , m. gckeler , r. horsley , y. nakamura , a. nobile , d. pleiter , p. rakow , a. schfer , g. schierholz , a. sternbeck and j. zanotti ( qcdsf collaboration ) , phys . rev . * d 85 * ( 2012 ) 054502 . h. ohki , h. fukaya , s. hashimoto , t. kaneko , h. matsufuru , j. noaki , t. onogi , e. shintani and n. yamada ( jlqcd collaboration ) , phys . rev . * d 78 * ( 2008 ) 054502 . k. takeda , s. aoki , s. hashimoto , t. kaneko , j. noaki and t. onogi ( jlqcd collaboration ) , phys . * d 83 * ( 2011 ) 114506 . h. ohki , k. takeda , s. aoki , s. hashimoto , t. kaneko , h. matsufuru , j. noaki and t. onogi ( jlqcd collaboration ) , arxiv:1208.4185 . d. toussaint and w. freeman ( milc collaboration ) , phys . * 103 * ( 2009 ) 122002 . w. freeman and d. toussaint ( milc collaboration ) , arxiv:1204.3866 . t. doi , m. deka , s .- j . dong , t. draper , k .- f . liu , d. mankame , n. mathur and t. streuer , phys . d 80 * ( 2009 ) 094503 . m. gong , a. li , a. alexandru , t. draper and k .- f . liu ( @xmath81qcd collaboration ) , pos * ( lattice 2011 ) * 156 . liu , m. deka , t. doi , y. b. yang , b. chakraborty , y. chen , s. j. dong , t. draper , m. gong , h. w. lin , d. mankame , n. mathur and t. streuer ( @xmath81qcd collaboration ) , pos * ( lattice 2011 ) * 164 . s. drr , z. fodor , t. hemmert , c. hoelbling , j. frison , s. katz , s. krieg , t. kurth , l. lellouch , t. lippert , a. portelli , a. ramos , a. schfer and k. szabo , phys . rev . * d 85 * ( 2012 ) 014509 . s. dinter , v. drach , r. frezzotti , g. herdoiza , k. jansen and g. rossi ( etm collaboration ) , jhep * 1208 * ( 2012 ) 037 . r. horsley , y. nakamura , h. perlt , d. pleiter , p. rakow , g. schierholz , a. schiller , h. stben , f. winter and j. zanotti ( qcdsf and ukqcd collaborations ) , phys . * d 85 * ( 2012 ) 034506 . r. babich , r. brower , m. clark , g. fleming , j. osborn , c. rebbi and d. schaich , phys . rev . * d 85 * ( 2012 ) 054510 . r. young and a. thomas , phys . rev . * d 81 * ( 2010 ) 014503 . p. shanahan , a. thomas and r. young , arxiv:1205.5365 . c. alexandrou , k. hadjiyiannakou , g. koutsou , a. cais and a. strelchenko , comput . * 183 * ( 2012 ) 1215 . j. beringer et al . ( particle data group ) , phys . d 86 * ( 2012 ) 010001 . j. liu , r. d. mckeown and m. ramsey - musolf , phys . rev . * c 76 * ( 2007 ) 025202 . m. engelhardt , pos * ( lattice 2010 ) * 137 . p. hgler , w. schroers , j. d. bratt , r. g. edwards , m. engelhardt , g. fleming , b. musch , j. w. negele , k. orginos , a. v. pochinsky , d. renner and d. g. richards ( lhp collaboration ) , phys . rev . * d 77 * ( 2008 ) 094502 . a. walker - loud , h .- w . lin , d. g. richards , r. g. edwards , m. engelhardt , g. fleming , p. hgler , b. musch , m .- f . lin , h. b. meyer , j. w. negele , a. v. pochinsky , m. procura , s. syritsyn , c. morningstar , k. orginos , d. renner and w. schroers , phys . rev . * d 79 * ( 2009 ) 054502 . j. d. bratt , r. g. edwards , m. engelhardt , p. hgler , h .- w . lin , m .- f . lin , h. b. meyer , b. musch , j. w. negele , k. orginos , a. v. pochinsky , m. procura , d. g. richards , w. schroers and s. syritsyn ( lhp collaboration ) , phys . rev . * d 82 * ( 2010 ) 094502 . j. m. alarcn , j. martin camalich and j. a. oller , phys . rev . * d 85 * ( 2012 ) 051503 . v. furman and y. shamir , nucl . * b439 * ( 1995 ) 54 . t. blum , p. chen , n. christ , c. cristian , c. dawson , g. fleming , a. kaehler , x. liao , g. liu , c. malureanu , r. mawhinney , s. ohta , g. siegert , a. soni , c. sui , p. vranas , m. wingate , l. wu and y. zhestkov , phys . rev . * d 69 * ( 2004 ) 074502 . p. breitenlohner , d. maison and k. stelle , phys . lett . * b134 * ( 1984 ) 63 . a. skouroupathis and h. panagopoulos , phys d 79 * ( 2009 ) 094508 . c. allton , d. j. antonio , y. aoki , t. blum , p. a. boyle , n. h. christ , s. d. cohen , m. a. clark , c. dawson , m. a. donnellan , j. m. flynn , a. hart , t. izubuchi , a. jttner , c. jung , a. d. kennedy , r. d. kenway , m. li , s. li , m .- f . lin , r. d. mawhinney , c. m. maynard , s. ohta , b. j. pendleton , c. t. sachrajda , s. sasaki , e. e. scholz , a. soni , r. j. tweedie , j. wennekers , t. yamazaki and j. m. zanotti ( rbc and ukqcd collaborations ) , phys . rev . * d 78 * ( 2008 ) 114509 . m. lscher , s. sint , r. sommer and p. weisz , nucl . phys . * b478 * ( 1996 ) 365 . y. aoki , r. arthur , t. blum , p. a. boyle , d. brmmel , n. h. christ , c. dawson , j. m. flynn , t. izubuchi , x .- y . jin , c. jung , c. kelly , m. li , a. lichtl , m. lightman , m .- f . lin , r. d. mawhinney , c. m. maynard , s. ohta , b. j. pendleton , c. t. sachrajda , e. e. scholz , a. soni , j. wennekers , j. m. zanotti and r. zhou ( rbc and ukqcd collaborations ) , phys . rev . * d 83 * ( 2011 ) 074508 . w . chen and m. savage , d 66 * ( 2002 ) 074509 . t. yamazaki , y. aoki , t. blum , h .- w . lin , m .- f . lin , s. ohta , s. sasaki , r. j. tweedie and j. m. zanotti ( rbc and ukqcd collaborations ) , phys . * 100 * ( 2008 ) 171602 . j. r. green , m. engelhardt , s. krieg , j. w. negele , a. v. pochinsky and s. n. syritsyn , arxiv:1209.1687 . r. g. edwards and b. jo ( lhpc and ukqcd collaborations ) , nucl . * 140 * ( 2005 ) 832 .	contributions of strange quarks to the mass and spin of the nucleon , characterized by the observables @xmath0 and @xmath1 , respectively , are investigated within lattice qcd . the calculation employs a 2 + 1-flavor mixed - action lattice scheme , thus treating the strange quark degrees of freedom in dynamical fashion . numerical results are obtained at three pion masses , @xmath2 , @xmath3 , and @xmath4 , renormalized , and chirally extrapolated to the physical pion mass . the value extracted for @xmath1 at the physical pion mass in the @xmath5 scheme at a scale of @xmath6 is @xmath7 , whereas the strange quark contribution to the nucleon mass amounts to @xmath8 . in the employed mixed - action scheme , the nucleon valence quarks as well as the strange quarks entering the nucleon matrix elements which determine @xmath0 and @xmath1 are realized as domain wall fermions , propagators of which are evaluated in milc 2 + 1-flavor dynamical asqtad quark ensembles . the use of domain wall fermions leads to mild renormalization behavior which proves especially advantageous in the extraction of @xmath0 . pacs : 12.38.gc , 14.20.dh
the final merger of two black holes ( bhs ) having comparable masses will produce an intense burst of gravitational radiation , and is expected to be one of the strongest sources in the gravitational - wave sky . mergers of stellar black holes are key targets for ground - based detectors such as ligo , virgo , and geo600 , and knowledge of the merger waveforms is an important component of improving the detectability of such systems . the space - based lisa detector will observe mergers of massive black holes at high signal - to - noise ratios , allowing tests of general relativity in the strong - field , dynamical regime . today , numerical relativity ( nr ) studies are beginning to progress toward a full description of black - hole binary merger systems . for noneccentric inspirals , this space is spanned by seven parameters : the symmetric mass - ratio @xmath1 , and the six combined components of the black holes spin vectors . considerable study has been focused on the fiducial center point of this parameter space , the case of equal - mass nonspinning black - hole mergers . after the series of breakthroughs that ushered in an era of rapid progress in the field @xcite , several investigations assessing the accuracy of the available equal mass waveforms and applying them to data analysis were conducted @xcite . in this paper , we undertake a descriptive study of the waveforms generated in the late inspiral and merger of black - hole binaries for the subspace of nonspinning black holes , parametrized only by @xmath2 . our study is based on a series of numerical simulations , discussed in sec . [ sec : simulations ] , covering at least the last @xmath3 orbits of nonspinning black - hole binary mergers with mass ratios extending to 6:1 ( @xmath4 ) . several of the simulations presented here have already been applied in a recent paper , focusing on the development of a faithful analytic waveform model @xcite . here we provide details of these and additional simulations , together with considerable analysis , focused on providing a qualitative and quantitative picture of how the waveforms from nonspinning black - hole mergers depend on @xmath2 . nonspinning black - hole binary merger waveforms were previously examined in ref . @xcite , but our analysis is novel and complementary to that work . our descriptive presentation puts emphasis on the relationships between waveforms from the different mass - ratio cases and different harmonic modes , with references to ref . @xcite where related observations have been made . our approach to describing the inspiral - merger - ringdown transition is particularly distinct , founded in a uniform approach that describes all stages of this process in similar terms , and ultimately suggesting a complementary physical picture . black - hole - binary merger waveforms have been noted for their `` simplicity . '' for the nonspinning systems the simple physics of the coalescence is exposed by a spherical harmonic decomposition of the waveforms . in sec . [ sec : description ] we walk readers through the basic features of the radiation , characterizing amplitude and phase evolution of the multipolar components , and discussing relationships among the simulations representing different mass ratios , and among the multipolar components of each simulation . as we analyze the waveforms we develop a conceptual interpretation of the basic waveform features . in this interpretation we consider the structure of an _ implicit rotating source _ , which could have generated the measured radiation through its rotational motion . this allows a uniform interpretation that applies throughout the coalescence process : inspiral , merger and ringdown . in sec . [ sec : description2 ] , we examine the strong final burst of radiation beginning @xmath5 before the formation of a common horizon . we quantitatively describe the phasing in terms of an analytic model , based on a continuous , monotonically increasing frequency . we find , in particular , that the peak rate of change in frequency , appropriately scaled , is the same across all @xmath0 modes and mass ratios . we also identify relationships among the mode amplitudes and phases , which are connected to an approximately linear relationship between angular momentum and frequency : @xmath6 . we interpret these relationships in terms of the implicit source . finally , in sec . [ sec : neweob ] , we demonstrate the utility of what we have learned in our waveform characterization by applying some of the quantitative features we have uncovered in a new variation on the analytic waveform model in @xcite , which was based on the effective - one - body ( eob ) resummation of the post - newtonian(pn ) approximation to inspiral dynamics @xcite . in particular , we provide a distinct late - time waveform model , alternative to the common `` spectroscopic '' model @xcite based on sums of quasinormal mode overtones . we begin with some examples of gravitational strain waveforms as they might be observed by gravitational - wave instruments . in observational work , and pn analysis , it is customary to describe the radiation in terms of gravitational - wave strain , @xmath7 . in representing the strain , it is convenient to combine the two real waveform polarization components , @xmath8 and @xmath9 , into one complex strain waveform , @xmath10 we decompose the strain waveforms measured on a sphere of radius @xmath11 , into spin - weighted spherical harmonic components , @xmath12 . the details of the decomposition , and how the waveform information is extracted from the numerical simulations , are given in appendix [ appendix : radiation ] . the waveforms in this section are aligned in time and phase so that the maximum strain amplitude occurs at @xmath13 . the remaining figures of this paper will be aligned in a similar way , but with @xmath13 marking the time of peak ( 2,2 ) mode energy flux , @xmath14 ( unless stated otherwise ) . [ fig : waves ] shows waveforms from mergers of nonspinning black holes for various mass ratios , as observed at distance @xmath15 on the rotational / orbital axis of the system . the figure shows @xmath8 for each of the four mass ratios 1:1 , 2:1 , 4:1 , and 6:1 . for these observers the observed waveforms will be circularly polarized , so that @xmath16 is @xmath17 out of phase with @xmath18 . we use units in which @xmath19 and @xmath20 and express both time and spatial distances in terms of the total mass @xmath21 , where @xmath22 . more typically , the observer will not be located on the system s orbital axis . the left panel of fig . [ fig : waveform_lm ] shows @xmath8 for the 4:1 case . the strain is measured at an azimuthal angle of @xmath23 and various inclinationsthe inclination angle is defined here as the angle between the line - of - sight with respect to the detector and the orbital axis of the binary . this is the same angle referred to as `` inclination '' in the pn / nr literature , and most equations are constructed using that definition . however , the astronomical literature has often defined inclination to be the angle between the line of sight and the orbital plane of the binary , resulting in a @xmath17 inconsistency . ] . the detailed shapes of the waveforms change as the system is reoriented so that the observer moves off the system s rotational axis . for larger inclinations ( closer to being viewed edge - on ) there are notable modulations at half the base gravitational - wave frequency . the right panel of fig . [ fig : waveform_lm ] shows @xmath8 for different mass ratios oriented at an inclination of @xmath17 and an azimuthal angle of @xmath23 . for this orientation , @xmath8 constitutes the full strain waveform . for larger mass ratios , the lower frequency modulation increases . for gravitational - wave observations of sufficiently strong binary black - hole sources , the types of differences shown in fig . [ fig : waveform_lm ] could be exploited to estimate the inclination and mass - ratio of the source system . for observational purposes , the combined waveform information encoding all possible source orientations can be conveniently represented in terms of spin - weighted spherical harmonic components [ see eqs.([eqn : psi4lm_def])-([eqn : hlm_def ] ) ] , providing a neat description of the leading waveform features , the multipolar decomposition is even more valuable as a tool for exposing the hallmark simplicity of the merger radiation . the readily apparent simplicity in the waves viewed from the system s orbital axis in fig . [ fig : waves ] extends to each of the spherical harmonic components . viewed off axis , these components linearly combine to yield the more complex appearance of the waveforms in fig . [ fig : waveform_lm ] . this characterization of the gravitational radiation from a merging black - hole binary in terms of circular polarization was first recognized in the lazarus project studies @xcite . in this picture , the radiation can be represented by a slowly varying amplitude and a polarization angle ; see eq . ( [ eqn : strain_modes ] ) below . this description relies on how the radiation appears to distant observers located on the rotational axis of the system . other observers will typically see elliptically polarized waves , having a generally simple pattern that conforms to the rotational nature of the source . in the equatorial plane , the radiation reduces to the plus polarization , corresponding to the observer seeing no circulation in the source . looking along the negative rotation axis , the observer sees circular polarization with the opposite helicity . each of the spherical harmonic waveform components exhibits circular polarization with steadily varying phase and amplitude , providing a natural framework for developing a practical and intuitive understanding of binary black - hole merger radiation . our basic waveform description , in sec . [ sec : description ] , and the more detailed analysis that follows , is based on this spherical harmonic decomposition . as we describe the waveforms , we will also suggest a heuristic interpretation of what the radiation tells us about the motion and structure of the binary black - hole source . in the weak - field description of radiation from a rotating object , the multipolar waveform components of the gravitational radiation can be associated with dynamics of multipolar moments of the radiating source @xcite . it is useful , in conceptualizing the full coalescence radiation from inspiral through merger and ringdown , to think of the multipolar radiation description as providing information about the motion of a changing source object , described as a sum of several multipolar mass moments . this source object is what we will interpret as an effective rigid rotator radiation source , with a slowly changing structure . we refer to this as the _ implicit rotating source _ ( irs ) . in the process of coalescence , the source begins as a separated black - hole binary system and ends as a single distorted black hole . for nonspinning binary mergers , numerical and pn results consistently indicate that the radiation is circularly polarized , in the sense first recognized in the lazarus project studies @xcite , not only in the inspiral , but uniformly through the merger and ringdown . in our conceptual source description , this pattern of circular polarization is consistent with radiation generated by rotational motion of each source multipolar moment , where the polarization phase is tied to the instantaneous orientation of the source . similarly , we think of the amplitude of the radiation multipole as related , through some generalization of the quadrupole formula , to the magnitude of the implicit source multipole . we can write each multipolar component in a specific polar form natural for circularly polarized radiation : @xmath24 each amplitude @xmath25 is expected to be a slowly varying function of time and can be conceptually considered as a function only of the magnitude of the source multipolar moment and its rotational frequency . the additional sign for some @xmath26 cases allows a consistent interpretation for the component phases and the component amplitudes . the waveform phase is given here by @xmath27 , but in most of our analysis we refer to @xmath28 , which we call the _ rotational phase_. in our terms of our implicit source heuristic , the rotational phase for some particular @xmath29 mode can be thought of as the azimuthal orientation of the specified multipolar source component . in the inspiral , where the source can be considered as a separated binary , @xmath28 should coincide with the orbital phase , independent of @xmath29 . in the post - newtonian expansion @xcite , all the @xmath28 defined here agree to at least 2pn order , while the amplitudes @xmath25 remain real and non - negative . the equatorial - plane symmetry of these mergers ensures that @xmath30 and that @xmath31 , so that we need consider only @xmath32 modes for this analysis . the expansion ( [ eqn : strain_modes ] ) is not appropriate for the @xmath33 modes . this points to an important caveat to our implicit rotating source interpretation of the radiation , that it applies to the degree that the radiation is circularly polarized . while not strictly vanishing , the @xmath33 waveform components , and other deviations from circular polarization are generally extremely small , and largely unmeasurable at the resolutions of the simulations we study . for the most part , we will not address deviations from circular polarization , and the consequent limitations of our implicit - rotating - source interpretation in this paper , focusing for now on the dominant features of the radiation . our analysis is based on four simulations , representing mass ratios 1:1 , 2:1 , 4:1 and 6:1 . results from the 1:1 and 4:1 simulations have appeared in previous publications ( @xcite and @xcite , respectively ) . more recently , higher accuracy simulations have been presented by other groups for the 1:1 case @xcite . our older waveform is sufficiently accurate for our present purpose , to examine the general features of the nonspinning merger waveforms . our numerical simulations are carried out with the hahndol evolution code @xcite , which uses finite - differencing methods to solve a 3 + 1 formulation of einstein s equations on a cartesian grid . for initial data we solve the elliptic equation given by brandt and brgmann for conformally flat data in which the black holes are represented by punctures @xcite . this is performed numerically using the multigrid solver amrmg @xcite , which is second - order - accurate but tuned to give truncation errors typically much smaller than those produced by the evolution code . the momentum parameters are chosen according to the 2pn - accurate quasicircular approximation given by kidder @xcite , which has been found to result in low eccentricity . we evolve these data using the moving puncture method @xcite with a modified version of the baumgarte - shapiro - shibata - nakamura equations @xcite . specifically , as suggested in @xcite , we replaced the conformal factor variable @xmath34 with @xmath35 , which vanishes at the punctures . further , we added the constraint - damping terms suggested in @xcite , and the dissipation terms suggested in @xcite . for the gauge we use the specific 1+log lapse and gamma - freezing shift conditions recommended for moving punctures in @xcite . accurate simulations require adequate spatial resolution near the black holes ( length scales @xmath36 ) as well as in the wave zone where the gravitational waves are extracted ( length scales @xmath37 ) . to this end , the grid has multiple refinement levels , determined adaptively near the black holes , but fixed in regions farther away ( typically , @xmath38 ) where the waves are extracted ; all grid refinement is handled within the framework of the software package paramesh @xcite . the adaptive mesh refinement criterion near the black holes is designed to keep the scale of the square root of an invariantly defined curvature component , known as the coulomb scalar @xcite , roughly constant with respect to the grid spacing . interpolation in guard - cells between refinement regions is fifth - order - accurate , coupling with differencing stencils to yield at least fourth - order accuracy in the bulk . spatial derivatives are taken by sixth - order - accurate differencing stencils , with the exception of advection derivatives , which are handled by fifth - order - accurate mesh - adapted differencing for greater stability @xcite . time integration is performed with a fourth - order runge - kutta algorithm . the initial configurations of the simulations we analyze are given in table [ table : rundata ] . in each case , the initial separation was chosen to be large enough to result in at least five orbits . the finest resolution , @xmath39 , ranged from @xmath40 to @xmath41 , as required to adequately resolve the black hole with the smaller mass in each case . the outer boundary was typically at @xmath42 , far enough away to prevent reflections from reaching the wave - extraction region during the simulation . it is instructive to compare these values with recent predictions for the final mass and spin from analytic and numerical methods . working with a subset of the data supplied here , and also invoking the test - mass limit , @xcite suggest the one - parameter fits : @xmath43 the numerical data from table [ tab : kerrparams ] also fits well the wholly numerical formula ( 3.17a ) from @xcite , as well as the parametrized formula of @xcite . we note that when discussing the mass @xmath44 of the end - state kerr hole in secs . [ ssec : transfreq ] and later , we have consistently taken the radiation - derived value , @xmath45 , as results from @xcite indicate this has errors up to a factor of 3 less than qnm - based mass . since a far larger proportion of the initial angular momentum than mass is emitted during the inspiral and merger , there is greater uncertainty in the radiation - derived final spin than the final mass . for this reason , our preferred measure of the final dimensionless spin is the qnm - derived value @xmath46 .	we present a detailed descriptive analysis of the gravitational radiation from black - hole binary mergers of nonspinning black holes , based on numerical simulations of systems varying from equal - mass to a 6:1 mass ratio . our primary goal is to present relatively complete information about the waveforms , including all the leading multipolar components , to interested researchers . in our analysis , we pursue the simplest physical description of the dominant features in the radiation , providing an interpretation of the waveforms in terms of an _ implicit rotating source_. this interpretation applies uniformly to the full wave train , from inspiral through ringdown . we emphasize strong relationships among the @xmath0 modes that persist through the full wave train . exploring the structure of the waveforms in more detail , we conduct detailed analytic fitting of the late - time frequency evolution , identifying a key quantitative feature shared by the @xmath0 modes among all mass ratios . we identify relationships , with a simple interpretation in terms of the implicit rotating source , among the evolution of frequency and amplitude , which hold for the late - time radiation . these detailed relationships provide sufficient information about the late - time radiation to yield a predictive model for the late - time waveforms , an alternative to the common practice of modeling by a sum of quasinormal mode overtones . we demonstrate an application of this in a new effective - one - body - based analytic waveform model .
transport properties of disordered solids attracted lot of attention . the quantum percolation ( qp ) problem@xcite has been studied several decades and some questions are still open , compared to the more settled anderson model@xcite . first of all , do both models belong to the same universality class ? positive answer means that the critical exponent @xmath0 of localization length should be the same , i. e. @xmath1 in 3d and in presence of time reversal symmetry . the best data come from transfer matrix ( tm ) calculations of anderson model@xcite . another question is the validity range of single parameter scaling ( sps ) . it is possible , that sps does not work in the middle of the conduction band for 3d qp , which is the case also for 1d anderson model@xcite , but not for 3d anderson model ( at least numerically ) . contrary to the still discussed 2d qp case@xcite , all authors agree that metal - insulator transition ( mit ) is present in 3d , but they often disagree in details . the mobility edge trajectory ( mit line in energy @xmath2 vs. disorder plane ) in 3d is qualitatively clear and widely accepted @xcite , with possible jumps at special energies of incoming non - interacting transport careers ( @xmath3@xcite . we aim to comment the nature of these jumps in more detail - for site percolation , but the bond percolation behaves similarly@xcite . the common universality further requires constant @xmath0 along the mobility edge . the exponent should be the same both with respect to changing @xmath2 at fixed disorder and vice versa@xcite . we will check all these statements numerically . another widely accepted fact is that @xmath0 should be identical both for site and bond percolation . the values of the exponent @xmath0 given in literature differ heavily . some methods seem to overestimate the true value , if we accept @xmath4 as correct also for qp . renorm - group calculations gave @xmath5 @xcite or @xmath6 @xcite . thouless conductivity@xcite yielded @xmath7 on fcc lattice . on the other hand a special version of tm method with approximative reduction of matrices@xcite suggested @xmath8 and cluster expansion with pad approximation@xcite yielded @xmath9 . the last value violates even the chayes bound @xmath10 , where @xmath11 is dimension . it is known from the anderson model , that level spacing methods tend to somewhat underestimate @xmath0 compared to the tm value ; they suggest @xmath12 @xcite and @xmath13 @xcite for 3d qp . thus we can say that the latter results confirm the common universality . the same is true of calculations using retarded green s function method@xcite , with resulting @xmath14 at @xmath15 , but for small system sizes @xmath16 and rather spread data . another paper of the same group@xcite reported lower @xmath17 . the main goal of this paper is to present another , tm based confirmation of the assumption , that anderson model and qp belong to the same universality class . let us recall the tight - binding hamiltonian : @xmath18 where the first sum counts the @xmath19 lattice sites , the second sum goes over nearest neighbors and we set the hopping t@xmath20 , thus fixing the energy scale . the length scale is set by choosing the lattice constant @xmath21 . for a binary system the distribution of on - site energies @xmath22 has the form : @xmath23 we prefer to use @xmath24 , the ratio of insulating atoms b , instead of often used @xmath25 because @xmath24 is the measure of disorder , induced among perfectly conducting a atoms . the quantum percolation model can be got either by removing appropriate t@xmath26 in eq . ( [ eq:1 ] ) or by sending @xmath27 . the latter case is more simple for tm method , as we need all t@xmath20 in the transfer direction and it suits better for site percolation . thus we set the on - site energy for a - type atoms to @xmath28 and the one for the randomly positioned b - type atoms to a large number@xcite , typically @xmath29 . the prize one has to pay in numerical approach are ill - defined matrices and a non - zero , but very small tunelling conductance through @xmath30 sites with still finite barrier - though the latter disadvantage shows up only in deep insulators @xcite , behind the scope of this paper . for anderson model the @xmath31 in eq . ( [ eq:2 ] ) is continuous , e. g. box distribution from an interval of the width @xmath32 ( disorder ) : @xmath33 $ ] . it should be further mentioned , that both models are fully symmetric with respect to @xmath34 , thus we will restrict our investigation to one ( left ) half of the band : @xmath35 . the classical site percolation threshold for simple cubic lattice is @xmath36 and it is known , that the quantum critical value @xmath37 for the whole mobility edge . the quantum wave requires a better highway than the narrow path of the percolation cluster backbone . up to now the maximum calculated value@xcite was @xmath38 , but we will show that the difference from @xmath39 is smaller . our numerical calculations are based on the usual transfer matrix method , described elsewhere@xcite . the conductance @xmath40 in units @xmath41 is given by the landauer formula @xmath42 . here @xmath43 is the @xmath44 transmission matrix , specifying the transmitted ( not reflected ) part of @xmath45 planar waves , entering the sample on the left and leaving it on the right . @xmath43 is extracted from a @xmath46 transfer matrix . it was mentioned above , that we have to do with ill - defined matrices , even more than within anderson model . in transfer matrix approach we deal rather with @xmath47 matrices and the precision of their small eigenvalues gets quickly lost - but we need them accurate , as they become important after recovering the @xmath43 matrix . an efficient way how to treat this problem was described in refs . and ; we have to perform the renormalization procedure at least after each two tm multiplications , but then we are able to include larger samples up to @xmath48 . this is useful , as finite size effects turn out larger than for anderson model . the most numerically sensitive region is around the band center , i. e. for small @xmath49 , where we have to perform the procedure after each single tm multiplication@xcite . once we have the matrix @xmath50 , we can calculate also its eigenvalues @xmath51 , satisfying @xmath52 where the positive quantities @xmath53 play similar role for cubic samples , as lyapunov exponents do in quasi-1d geometry . the smallest one is labeled @xmath54 . in localized samples , the landauer conductance decays quickly with sample size , @xmath55 . the correlation length @xmath56 diverges at critical point @xmath57 as @xmath58 for @xmath59 fixed and vice versa . the sps assumption implies one combined variable from @xmath60 or @xmath61 and in the vicinity of the mit point we expect @xmath62 . if we perform taylor expansion in @xmath63 variable and add some finite - size correction , we get @xmath64 ^ 2\bigr),\ ] ] where @xmath65 is an irrelevant exponent and its term becomes negligible for large @xmath66 . analogous relations can be expected for @xmath67 , i. e. with the single variable @xmath68 and @xmath69 fixed . the conductance @xmath40 will be mostly an ensemble average @xmath70 , taken over many samples , typically 20.000 . it is known@xcite that many other quantities would fit eq . ( [ exp ] ) , e. g. the geometric mean @xmath71 , @xmath72 , percentiles of the distribution of conductances @xmath73 , @xmath74 , current moments@xcite @xmath75 , inverse participation ratios@xcite , etc@xcite . nevertheless , quantities like @xmath76 or @xmath74 are more sensitive to numerical instabilities , especially for larger @xmath77 and @xmath66 or very small @xmath49 . it was shown@xcite , that the density of states ( dos ) has sharp peaks at special energies , @xmath78 these eigenvalues belong to submatrices of the hamiltonian eq . ( [ eq:1 ] ) , created by small islands of @xmath79 sites completely surrounded by @xmath30 sites , i. e. isolated from the rest . it was found , that @xmath40 can quickly fall down at these energies and they can influence the shape of mobility edge . before going to main results , let us check our numerics in the vicinity of the special point @xmath80 . conductance @xmath40 should be @xmath81 independent for large enough @xmath81 . we can see in fig . [ fig : f1 ] , that for delocalized sample with @xmath82 , common upper line , this is true for @xmath83 even at @xmath80 . for a localized sample , all lower lines , the value depends strongly on the energy @xmath2 close to the special @xmath80 , but still a plateau develops for larger @xmath81 . exception is the special point itself , where @xmath84 , in other words @xmath85 for @xmath27 . the same @xmath86 dependence is found for a test sample with one perpendicular square barrier of @xmath30 sites completely stopping the transport through otherwise perfect cube of @xmath79 sites ; thus this dependence is an unusual form of tunelling . the picture is qualitatively the same at @xmath87 , just for much lower disorder @xmath24 . it is worth mention , that in the latter case we surely do not have to do with simple quantum tunelling through closed finite samples@xcite ( closed in the sense of classical percolation , regardless of @xmath2 ) , as we chose samples with @xmath88 , far bellow @xmath39 . alternatively formulated , the same samples become metallic for larger @xmath49 . one can also see that the geometric mean of conductance tends to zero though the arithmetic @xmath70 remains positive at special points . we chose also @xmath89 and after some effort we found a localized sample of the type as in fig . [ fig : f1 ] bottom lines , but the ratio of such specially localized samples becomes much lower than for @xmath90 . the possible deep localization at special energies was explained in several ways . in ref . it was supposed , that the clusters of insulating atoms around conducting islands could have extremely large reflectivity at these energies . another explanation@xcite argued that localization on dead ends of the spanning cluster is strong at special energies . let us add yet another possibility . the dispersion law gives @xmath91 where @xmath92 is the wavenumber in transfer direction @xmath93 and the wavenumbers in perpendicular directions @xmath94 are quantized by hardwall boundary conditions . for special values of energies the whole vector ( @xmath95 ) can point in ( body ) diagonal direction and the planar wave is then fully reflected by insulating site , siting exactly in this direction . if all planar waves are fully reflected , the sample becomes deeply localized@xcite . summarizing , with @xmath96 , we always get correct @xmath40 for delocalized samples and also for medium localization . strongly localized samples require special attention . vs. energy and ( b ) @xmath97 vs. disorder , both around the same mit point @xmath98.,title="fig : " ] ( a ) vs. energy and ( b ) @xmath97 vs. disorder , both around the same mit point @xmath98.,title="fig : " ] ( b ) now let us find one critical point in detail . for @xmath99 most samples are metallic and @xmath100 , for @xmath101 most samples are localized and @xmath102 . at mit , @xmath40 becomes @xmath66 independent . we chose the value @xmath103 , and found @xmath104 from the crossing point in fig . [ fig : f2]a ) and [ fig : f2]b ) . as @xmath69 is far enough from @xmath39 , and @xmath59 is not close to any important special value ( in the sense of relevant peaks in dos ) , we can use even the @xmath97 as order parameter . the data were fitted according to eq . ( [ exp ] ) and the resulting exponent is @xmath105 for the @xmath63 dependence and @xmath106 for the @xmath107 one , both in perfect accordance with @xmath4 . of course , taking @xmath108 instead of @xmath97 yields the same critical point and @xmath0 . having performed these calculations for almost two dozens of fixed energies , typically changing only @xmath24 , we can plot the mobility edge in fig . [ fig : f3 ] . some pairs of numerical values can be found in fig . [ fig : f6 ] . the point @xmath109 was added artificially - there are no open channels there . it was already mentioned above that the situation is mirror symmetric in @xmath110 , we do not plot the obvious interval @xmath111 . the upper border of the graph was set to the classical @xmath39 . the maximum value of our mit line is cca @xmath112 , greater than @xmath113 value reported before@xcite , but with smaller samples @xmath114 and @xmath16 , resp . - independent crossing point of arithmetic mean @xmath115 at band center.,title="fig : " ] ( a ) - independent crossing point of arithmetic mean @xmath115 at band center.,title="fig : " ] ( b ) now we will compare our results to those of ref . . they introduced the local dos for each lattice site and then performed two averages at once ( in two manners ) . the first one was ensemble average as we do , the second one was averaging with respect to all sites . the two manners were geometric and arithmetic mean . at last they calculate the ratio of geometric / arithmetic mean and plot lines , where this order parameter is constant . the line entering the point @xmath109 compares very well with our mobility edge in fig . [ fig : f3 ] , with their minimum at cca @xmath116 , i. e. our maximum at @xmath117 and a rather wide plateau around this maximum . the main difference is a discrete set of energies , where their order parameter and thus also the critical line drops to zero . but the only cases , where they could show this explicitly where integers @xmath118 . with irrational values like @xmath119 the localized case is hard to find numerically , as its region is extremely narrow in energies and also the corresponding peaks in dos are much smaller@xcite than those at three integer @xmath2 . hence we plot the detailed view of the band center in fig . [ fig : f4](a ) , where both calculations of critical @xmath69 with the use of geometric and arithmetic mean of conductance @xmath40 are presented - of course only ensemble averaged . the critical disorder @xmath69 from calculations with geometric mean of @xmath40 is systematically somewhat larger outside the band center , but only by several percent - beware that the @xmath24 axis is not drawn to zero . for very small @xmath120 the @xmath69 from geometric mean already becomes slightly smaller than that from arithmetic mean and directly at @xmath87 it drops steeply to zero . it should be stressed out , that the decay of mit line at band center down to cca @xmath121 is independent of the type of averaging , it happens within a clearly non - zero interval of energies and it should be distinguished from extremely narrow drops of mit line to zero at special energies , present only with geometric averaging . similar relatively broad interval of energies , where the mobility edge should decay , was reported@xcite around @xmath122 . with arithmetic mean and @xmath123 , we do not see it at all . we should also say that the standard mean dos , calculated up to @xmath124 , has clearly non - zero value for @xmath88 in the vicinity of the band center , apart from high central peak at @xmath87 . according to our calculations , only at cca @xmath125 the dos drops to zero for small @xmath49 and a gap opens for even larger @xmath24 . the exponent @xmath0 remains constant along the whole mobility edge line , just the error bars rise as we approach the borders of the band @xmath126 and @xmath87 . but even for @xmath127 and @xmath128 it was still @xmath129 . important exception is the point @xmath87 itself . only with geometric mean , it has @xmath130 , as suggested already in ref . . one possible explanation is that single parameter scaling ( sps ) does not work here anymore . the argument is , that geometric mean of @xmath40 differs substantially from the arithmetic one . furthermore , if we insists on sps and arithmetic @xmath70 as order parameter , standard analysis of data from fig . [ fig : f4](b ) according to eq . ( [ exp ] ) yields @xmath121 at @xmath87 and @xmath131 with @xmath63 variable . we tried also to calculate @xmath0 with @xmath132 variable , but the @xmath133 dependences changed their slope below @xmath134 and the analysis was too unstable . it is improbable , that common universality would be broken at one spectral point , the @xmath0 from fig . [ fig : f4](b ) data smaller than @xmath4 could rather mean , that sps is broken at @xmath87 . but one can also argue , that finite size effects are so strong here , that even sample sizes up to @xmath48 are insufficient . this would be exceptional , though not completely excluded . in ref . the authors had similar problems with 3d anderson model at the band edge ( close to @xmath135 ) and explained them by insufficient available sample sizes . summarizing , we always get a well defined crossing point of mobility edge , where data become @xmath66-independent as they should at criticality , even for @xmath87 , see fig . [ fig : f4](b ) . just the slopes of the crossing lines behave non - universally in band center , at least for available system sizes . let us present mean critical values of several quantities , namely @xmath136 , @xmath137 and @xmath138 . for @xmath139 we can roughly say that it is given mainly by the largest eigenvalue @xmath140 and after simple algebra @xmath141 , which is well fulfilled in fig . [ fig : f5 ] , of course only outside the band center . for small @xmath49 we get @xmath142 as a sum of several important @xmath51 . it should be emphasized , that @xmath136 remains finite , but @xmath137 and @xmath138 diverge , as we approach @xmath143 , see comments to fig . [ fig : f1 ] . the divergence is given by a minority of fully localized samples with @xmath85 for @xmath27 , which do not really influence the arithmetically averaged @xmath144 . further we present several critical conductance distributions @xmath145 and @xmath146 along the mobility edge . the distribution at @xmath87 and @xmath147 was added for comparison , too . ( a ) and @xmath148 ( b ) along the mit line.,title="fig : " ] ( a ) ( a ) and @xmath148 ( b ) along the mit line.,title="fig : " ] ( b ) we do not plot critical distributions for @xmath149 which are mutually almost overlaying , because of similar @xmath69 . further we can see typical non - analyticity@xcite at @xmath150 or @xmath151 , if @xmath73 or @xmath152 , resp . , is already decaying there . the @xmath146 at @xmath153 and those with maximal @xmath69 look similar to @xmath146 of 3d anderson model , but they are not identical for any @xmath154 - the sps works separately inside the models , not universally . this is known also for anisotropic anderson models and those in other dimension than @xmath11 . even for the bifractal@xcite with the lowest @xmath155 and @xmath156 the @xmath145 looked more gaussian than those in fig.[fig : f6](a ) on the right . this is connected with far larger @xmath157 of 3d qp compared to low - dimensional anderson model at criticality or to universal conductance fluctuations in metallic regime . let us shortly discuss the metallic regime . away from band center , say with @xmath158 , we found typical metallic behavior for medium disorder @xmath159 , i. e. @xmath73 practically gaussian , @xmath160 and var @xmath40 constant ( @xmath66 and @xmath24 independent ) , with values cca 10 percent above the theoretically predicted universal conductance fluctuations@xcite . even smaller @xmath24 brings traces of ballistic behavior , var @xmath40 becomes system - size dependent ( rising with @xmath66 ) . contrary to these standard properties , at the band center @xmath87 we did not find typical metal for any @xmath24 and available @xmath66 . we still have @xmath160 , but now @xmath161 instead of reaching the universal constant and cca one order of magnitude greater - rather like the mentioned almost - ballistic behavior ( perfect ballistics would have @xmath162 ) . also @xmath73 is somewhat deformed , not gaussian . the transition from this behavior to standard metal at larger @xmath49 is continuous , var @xmath40 decays with rising @xmath49 . if this is only finite size effect , than it is again immense at band center . we already stated recently@xcite , that 2d quantum percolation problem and 2d anderson model belong to the same universality class , meaning that @xmath163 is lower critical dimension for both models and there is no mit in time reversal symmetry case . here we add the confirmation of common universality in 3d , implying identical critical exponent @xmath0 , which we have shown within numerical accuracy along the mobility edge , with possible exception of the conduction band center @xmath87 . another new result confirms equal @xmath0 when cutting the mobility edge ( mit line ) either in disorder or in energy direction . all these properties result from single parameter scaling hypothesis and they get lost at band center , where either the sps does not work anymore , or we face immense finite - size effects , that can not be overcome with system sizes up to @xmath48 . this statement is supported also by loosing many typical properties of metallic samples at band center . concerning the mobility edge , we found the maximum critical disorder @xmath69 slightly above 0.6 , which is more than @xmath164 reported before@xcite . the shape of the mobility edge at band center depends on the choice of order parameter . the arithmetically averaged @xmath108 ( commonly preferred ) gives clearly non - zero @xmath165 at @xmath87 , whereas the geometrical mean gives @xmath130 in accordance with other numerics@xcite and theory@xcite . we present a new alternative explanation of extremely low conductance for a minority of samples , which push the geometric mean to zero . it is based on the fact , that for special energies the wavevectors of incoming planar waves point in body diagonal directions and the waves can face exactly the insulating sites , followed by full reflection . the rational special energies @xmath118 , have clearly different arithmetic and geometric means and also the resulting @xmath69 . the irrational ones , e. g. @xmath166 , seem numerically almost insensitive of the type of averaging , as the samples with nearly zero conductance become very rare . but the rather steep descent of mobility edge at band center takes place at non - zero interval of energies with sps still working for both types of averaging and it should be distinguished from the extremely narrow regions at all special energies , apparent only for geometric mean of conductances . this work was supported by grant vega nr .	we present the metal - insulator transition study of a quantum site percolation model on simple cubic lattice . transfer matrix method is used to calculate transport properties - landauer conductance - for the binary distribution of energies . we calculate the mobility edge in disorder ( ratio of insulating sites ) - energy plane in detail and we find the extremal critical disorder somewhat closer to the classical percolation threshold , than formerly reported . we calculate the critical exponent @xmath0 along the mobility edge and find it constant and equal to the one of 3d anderson model , confirming common universality class . possible exception is the center of the conduction band , where either the single parameter scaling is not valid anymore , or finite size effects are immense . one of the reasons for such statement is the difference between results from arithmetic and geometric averaging of conductance at special energies . only the geometric mean yields zero critical disorder in band center , which was theoretically predicted .
it is known observationally that about half of the observed stars are members of binary or higher order multiple systems @xcite . the fraction increases with the primary star mass and reaches up to 69% for o stars @xcite . these massive stars are supposed to end up exploding as core - collapse supernovae ( ccsne ) . it is hence natural to expect that the majority of ccsne should take place in binaries . binary interactions may be crucially important for the evolution leading to core - collapse @xcite . in particular , mass transfer will be essential to the spectral type of the supernova ( sn ) explosion . in fact , it is argued that type ib and ic sne mainly occur in interacting binaries @xcite . it has also recently been reported that the late time photometry of iptf 13bvn , a type ib sn , can not be reproduced by single star progenitors , but can naturally be explained by binary progenitor models @xcite . this may indicate that type ib sne are actually occurring in binaries although we had better wait for direct observations of the surviving companions in the future . the aftermath of the explosion could be no less important for the evolution of the binary system itself as well as of the companion star . it is well known that the binary system is disrupted if more than half of the total mass is expelled . this criterion is easily fulfilled when the primary star explodes unless it has lost most of its mass before explosion . even if the amount of mass expelled is less than half the total mass , the system may still be disrupted due to neutron star kicks . the companion star will then carry on its life as a single star . if the binary survives the sn explosion , it will have a highly eccentric orbit @xcite . some observational facts of a high mass companion actually surviving the sn explosion are known @xcite . most high mass x - ray binaries ( hmxbs ) and low mass x - ray binaries must have undergone sn at some time in their formation , which means a certain fraction of binaries need to survive from explosion . collisions of the supernova ejecta ( sne ) with the companion star may also affect the evolution of the latter if the binary separation is small enough . @xcite were the first to estimate the mass removed from the companion star by the impact . they considered momentum conservation in a simple analytical model to express the amount of mass removed with a single parameter and estimated that up to 15@xmath5 of the companion mass will be ejected . numerical computations were also carried out by @xcite , confirming the results by @xcite . their two - dimensional computational grid of @xmath6 was rather coarse by the current standard . they also assumed a planar shell as a model of sne , which may not be a good approximation for small binary separations , where the spherical geometry of sne is not negligible . @xcite followed , assuming a red giant companion of 1 @xmath7 . they treated sne as a spherical shell and used a finer mesh , enough to describe the envelope with fine zoning ( typically 107 @xmath8 65 ) . they found that almost all the envelope of the red giants were stripped off by sne . the result may be applied not only to type ib sne but also to type ia sne . the impact of sne on companion stars have been better studied for type ia sne @xcite in the single degenerate ( sd ) scenario , where a carbon - oxygen white dwarf accretes mass from its low - mass binary companion @xcite . carrying out numerical calculations with high resolutions , these authors placed strong constraints on the structure of companion stars in the sd scenario . in this paper , we perform similar simulations , but assuming more massive stars for both the primary and secondary stars , which will be appropriate for ccsne . unlike for type ia sne in the low - mass binary , there is no standard model for ccsne . in fact , the masses and structures of the primary and secondary stars as well as the binary separation are almost free parameters , since there are not many observational constraints . we pay particular attention to the mass removed from the massive companion and its dependence on separation . for the analyses of the results , we perform additional simulations of experimental nature , modifying the density and/or velocity of sne artificially . it turns out that these are indeed helpful to pin down which physical quantity is most essential in determining the amount of removed mass . this paper is structured as follows : in section 2 , we describe the models and numerical method we used . the main results are shown in section 3 , and discussions are given in section 4 . finally , we summarize our results in section 5 . in this paper we numerically investigate the impacts of the sne against the companion star in a massive close binary . we employ two - step strategy : the explosion of the primary star is computed in spherical symmetry , ignoring the existence of the secondary star , to obtain the structure of the sne ( step 1 ) ; the collision of the sne obtained in step 1 against the companion star is simulated under the assumption of axisymmetry ( step 2 ) . details of each step are given below . we use a two dimensional hydrodynamic code `` yamazakura '' for all computations performed in this paper @xcite . it is a time - explicit eulerian code based on a high resolution central scheme @xcite . spherical coordinates are employed and axisymmetry but no equatorial symmetry is assumed . in step 1 we further impose spherical symmetry , putting the origin of the coordinates at the centre of the primary star and ignoring its rotation as well as the orbital motion and the gravity of the companion star . in step 2 the existence of the primary star is ignored , with the origin of the coordinates coinciding with the centre of the companion star . the symmetry axis is directed to the primary star and rotation of the companion star as well as the orbital motion is neglected . the self - gravity of the companion star and sne in the computational domain is included by solving the poisson equation with the boundary conditions set by multipole expansion . an ideal equation of state with an adiabatic index @xmath9 is applied throughout . the masses of the primary and companion stars are set to @xmath10 and @xmath11 , respectively . note these are not the masses at birth of the binary system but the values just prior to the sn explosion of the primary star . these values are chosen so that the primary star would be massive enough to produce a sn explosion and the mass ratio should be close to unity , in which case the companion star radius @xmath12 could be largest for a given separation @xmath3 . the @xmath13 ratio is expected to be one of the most essential parameters that determine the amount of mass removed by the ejecta @xcite . in order to maximize the impact of sne on the companion star we choose the separation of binary as small as possible under the restriction that the companion should not cause roche lobe overflow ( rlof ) . assuming that the secondary star fills the roche lobe , we obtain the minimum separation @xmath14 from the eggleton relation @xcite : @xmath15 where @xmath16 is the mass ratio . the period of the keplerian motion with this separation is approximately five years , which is significantly long compared to the characteristic timescale of the collision of sne with the companion star , @xmath17 . here @xmath18 is a typical velocity of sne . it is hence justified to neglect the orbital motion in step 2 . it is one of the main objectives in this paper to find the dependence of the removed mass on the binary separation . for this purpose we have carried out several simulations with different separations @xmath19 . since we are not interested in the still unknown sn mechanism , we artificially produce a sn explosion in this study by injecting internal energy into the central region of the primary star . such a procedure is often called the `` thermal bomb '' @xcite . for the progenitor , we use a pre - sn model of heger & wellstein 1999 ( unpublished ) , which was computed according to the binary evolution models of @xcite , in which the `` standard '' wolf - rayet mass loss rate is reduced by a factor of six and mass transfer is taken into account . the mass and radius of the primary star are @xmath10 and @xmath20 , respectively , just prior to the sn explosion . it has totally lost its h layer , and is a helium star since it has undergone a significant amount of mass loss . we cover the primary star with a one - dimensional spherical grid . sixteen - hundred grid points are non - uniformly distributed in the radial direction . the inner boundary is chosen so that the central @xmath21 , corresponding to a neutron star , should be excluded from the computational region . the outer boundary is set at approximately 40 times the stellar radius . @xmath22 erg of internal energy is added to the inner 10 grids . self - gravity is included in this computation , although its effects are limited . we impose the reflective condition at the inner boundary whereas the outgoing boundary condition was used at the outer boundary . after the shock wave breaks out of the stellar surface , the sne continues to expand , converting internal energy to kinetic energy . since the hydrodynamic code can not handle vacuum , we artificially fill the volume outside the star with matter , which is so dilute that it should not affect the dynamics . when the sne reaches the outer boundary of the computational domain , it expands almost self - similarly . we record the density , pressure , velocity at an arbitrarily chosen point near the outer boundary as a function of time so that we could use them in step 2 as described shortly . as mentioned above , we place the companion star at the centre of a two - dimensional spherical mesh , whose axis is aligned with the line joining the centre of the two binary components . neglecting rotation of the companion star and the orbital motion , we assume axisymmetry in this step . the computational domain extends from the radius of @xmath23 to @xmath24 , which is roughly twice the radius of the companion star . the central portion that contains @xmath25 is cut out to avoid too severe cfl conditions . we treat it as a point gravitational source and we impose a reflective inner boundary condition . the computational region is divided into @xmath26 zones non - uniformly in the radial direction and uniformly in the @xmath27 direction . it is important to put the radial grid points appropriately in order to keep the star in hydrostatic equilibrium . we deploy at least 20 mesh points in each density scale height inside the star . on the other hand , we distribute the radial grid points uniformly outside the star , where we again put a dilute gas for numerical convenience . the locations of the grid points determined this way are shown in figure [ gridstructure ] . for the companion star we adopt the non - rotating s11.0 model from @xcite , since no information is available on the companion in heger & wellstein 1999 ( unpublished ) . this companion star model gives a mass ratio close to unity and gives the smallest value to @xmath28 among available stellar models . in fact , its radius of the companion star is @xmath29 and the @xmath13 ratio used in this study ranges from 2.7 to 6.7 . since it is a pre - sn model , i.e. , the star is near the end of its evolution and has a loosely bound envelope , so the results of our simulation possibly serve as an upper limit for the amount of mass stripped in ccsne of primary stars . it is true that the companion model is actually a single star model and not consistent with the primary star model , and may not be a realistic binary system . but massive binaries that will produce a ccsn are very much diverse and our understanding of their evolutions is not yet complete . moreover , binary - progenitor models available to us for the moment are not many , either . in this paper , we hence focus on the systematics of interactions between sn ejecta and the companion star , paying particular attention to the binary separation . our simulations are experimental in this sense . exploration of the dependences on the mass , structure and composition of the companion star for more realistic models will be left for future works . as mentioned above , the companion star is adopted from pre - sn models and its central core has already started to collapse . the computational domain excludes the central collapsing core , however . it is confirmed that the remaining envelope is almost in hydrostatic equilibrium and does not change significantly for at least 200 days after mapped onto the two - dimensional computational grid , which is shown in figure [ equil ] . sne is injected from one hemisphere of the outer boundary with its spherical geometry being taken into account properly . we estimate the density , pressure and velocity of sne from the data obtained in step 1 , assuming that the sne continues to expand self - similarly from the outer boundary of step 1 to that of step 2 . the free boundary condition is applied to the other hemisphere of the outer boundary . simulations start from the time , when the outermost part of the sne touches the surface of the companion star . to follow lagrangian motions of fluid elements in our eulerian hydrodynamics code and find which portions of the companion star are stripped by sne , we deploy tracer particles in the companion star as well as in the sne . the tracer particles are moved with the fluid velocities at their instantaneous locations , which are obtained by the hydrodynamical computations , and do not affect the dynamics of sne and stellar matter . each particle is labelled whether it originates from the star or ejecta and whether it escapes out of the computation domain before the termination of simulations or not . these information are extremely useful to understand how the mass - stripping takes place in the collision of sne and the companion star . in the following the numerical results for each step are given in turn . we artificially initiate the sn explosion in the primary star by depositing thermal energy near the inner boundary . the pressure excess then drives the shock wave outwards into the envelope . the deposited energy of @xmath30 , the canonical value for ccsne , is large enough to expel the outer envelope , since its binding energy is much smaller . the shock wave continues to propagate outward unhindered and breaks out of the stellar surface after 30 s. figure [ shock ] shows the propagation of the shock wave in the stellar envelope . it is clear that a dense shell is formed after the shock wave . it is also discernible that a reverse shock is produced when the forward shock passes through the boundary of c / o and he . it is known that the deceleration of matter by negative pressure gradients then induces a rayleigh - taylor instability in the vicinity of the co / he boundary . we ignore such multi - dimensional effects in the present study , anticipating that the dynamics of collision of the sne and the companion star , particularly the removed mass , will not be affected severely by the small - scale perturbations . after the shock break - out , the stellar envelope starts to expand into interstellar matter ( ism ) as an sne . it is initially accelerated by its own pressure , converting thermal energy to kinetic energy . when most of the thermal energy is consumed , the sne expands ballistically until it sweeps up an amount of ism comparable to the sne itself and the sedov - taylor phase commences . this is indeed observed in our simulation as shown in figure [ profile ] . in the left panels the profiles of density , pressure and velocity are shown for different times after the shock break - out whereas in the right panels the rescaled distributions are compared . self - similarity of the expansion is evident from the latter comparison . we assume in this study that this free expansion of sne continues until it reaches the companion star . the computation of step 1 is terminated when most of the sne leaves the computational domain . the temporal evolutions of density , pressure and velocity are recorded at distances @xmath31 km from the centre of the exploding star . they are shown on the left panels of figure [ explosion ] . as the dense shell of sne passes through these points , the density and pressure rise rapidly and decline gradually . the self - similarity in the expansion of sne is also reflected in these profiles as demonstrated on the right panels , where we rescale the profiles by the distance @xmath32 . we can see that the three lines almost exactly agree after this rescaling , which indicates that the sne expands in a self - similar fashion . we utilize these rescaled profiles to set the initial and boundary conditions in step 2 . we first describe the dynamics of the collisions of the sne and the companion star for the base model , in which the binary separation is set to @xmath33 . the results for other models with different binary separations will be presented in section [ secsepdep ] but qualitative features are common . the density and velocity profiles are shown at different times @xmath34 elapsed since the sn explosion of the primary star in figures [ density ] and [ pltv ] , respectively . the top left panels in these figures correspond to the initial state for step 2 . note that the lengths of vectors are scaled for visibility and amplified by a factor of @xmath35 for the right panels compared to the left panels because the velocity becomes quite small at these late stages . we can see that as the sne hits the companion star surface , a forward shock is formed and penetrates through the star . this shock wave heats the stellar matter as it propagates , and eventually reaches the other side of the star . the shock decelerates as it climbs the steep density gradient of the star toward the centre of the companion star . it then accelerates at the latter half , climbing down the density gradient from the centre to the stellar surface and depositing much smaller energies . some of the heated matter starts to expand ( panels ( d ) and ( e ) ) . a reverse shock is also formed and becomes a bow shock in the sne . it is almost stationary at the beginning as the ejecta flows past ( panels ( b ) and ( c ) ) . once the ejecta density starts to decrease , however , it begins to expand leftwards into the sne and eventually gets out of the computational domain ( panels ( d ) and ( e ) ) . after the forward shock penetrates the entire star , the whole star slowly expands almost spherically to reach a new equilibrium ( panel ( f ) ) . the sne does not strip the stellar matter directly as assumed in the previous papers @xcite . this is mainly due to the presence of the bow shock , which deflects the ejecta away . most of the removed mass is instead ablated off with the internal energy deposited by the forward shock and is also pulled off by the rarefaction wave generated by the reverse shock . these results are consistent with the previous findings for low mass companion stars in type ia sne @xcite . the shock break - out of the opposite side of the companion star also produces some mass ejections there . the above statements are more clearly vindicated in figure [ ptc ] , where the positions of tracer particles are displayed at the same instants as in figures [ density ] and [ pltv ] . in the upper half of each panel the particles originating from the star ( blue ) are distinguished from those from the sne ( red ) by colours . the lower half , on the other hand , shows whether it is `` bound '' to the companion ( light blue ) or not ( grey ) . the definition of the boundedness will be given in the next subsection in detail . we can see that as the shock propagates through the star , the heated matter becomes unbound initially . although the central region quickly becomes bound again thereafter , the outer part remains unbound and expands almost radially , an indication of thermal ablation . on the opposite side of the star , only the low - density matter near the stellar surface becomes unbound by the shock passage because the shock is weakened by that time due to the pull - back by the rarefaction wave . the central bound region then restores a spherical shape after a long time . the unbound region continues to expand , on the other hand , and eventually reaches the edge of the computational domain . since we can not determine for sure the fate of the particles that have left the computational region , we stop the simulation there . this constraint led to the termination time of 122 days for the closest separation model and that down to 105 days for the models with the farthest separations . figure [ ptc ] also points out that there is almost no ejecta contamination going on on the companion star surface . at the end of the simulation , all bound particles are of companion star origin . this indicates that we will not be able to observe imprints of the primary star on the surviving companion after sn . in previous studies on type ia sne @xcite , the stripped mass is estimated from the total energy defined locally as @xmath36 where @xmath37 is velocity , @xmath38 the specific internal energy and @xmath39 the gravitational potential . if this energy is positive , which is equivalent to having a velocity exceeding the local escape velocity , the fluid element there is judged to be unbound . we shall call this the `` escape criterion '' . the removed mass is the sum of these unbound masses . it should be noted , though , that the removed mass defined this way is time - dependent particularly at early times because the pressure inside the star is still large and interactions among fluid elements can not be neglected . it takes rather long for the removed mass calculated by this criterion to settle to the final value as shown in figure [ comparecriterion ] , in which the temporal evolution of the removed mass is shown . we find that an alternate criterion is more useful to estimate the removed mass . it is based on bernoulli s principle , which claims under certain conditions that particles along a streamline conserve the following quantity @xmath40 where @xmath41 is the pressure density and @xmath42 the mass density . strictly speaking , the principle is valid only in steady systems . regardless , we assume that a fluid element is unbound if @xmath43 is positive . note that @xmath43 should eventually approach @xmath44 when the pressure becomes negligible . the removed mass is defined as the sum of masses of unbound fluid elements determined this way . we refer to this scheme as the `` bernoulli criterion '' in the rest of this paper . in figure [ comparecriterion ] we compare the removed masses obtained by the two criteria as a function of time for the base model with @xmath33 : the escape criterion is shown as the blue line whereas the bernoulli criterion is displayed with the red line . it is evident that both lines approach the same asymptotic value , indicating that the two criteria give identical results when computed for long enough . more importantly , the red line reaches the final value earlier than the blue line , that is , the bernoulli criterion is more efficient to estimate the removed mass . this is indeed important , since the computing resource is limited and in some models we can not afford to run simulations long enough to see the settlement of @xmath44 . the shock heating of the matter facing the primary star , the main heating phase , lasts approximately 20 days . it takes about 70 days for the removed mass by the bernoulli criterion to settle to the asymptotic value , which is the time scale for the forward shock to reach the other side of the star . this is merely a few % of the orbital period and supports our neglection of the orbital motion in this study . at this time , the whole star starts expanding almost spherically and the expansion carries on for a few more months , where the effects of orbital motion and possible stellar rotation may become non - negligible . the amount of mass removed reached up to 25@xmath5 of the original companion mass in the base model . this is a significant fraction for the later evolutions of both the companion star and the binary system itself . it can be regarded as an alternate scenario of mass loss from high mass stars other than binary interaction or stellar winds . in the base model we assume the smallest possible binary separation and the removed mass will be largest . the next question is obviously how the removed mass depends on the binary separation . before proceeding to other models with different binary separations , we check here whether the spatial resolution employed in the simulation is appropriate . for this purpose we ran models with the same physical setup but with different spatial resolutions . the number of grid points was changed separately for @xmath32 and @xmath27 directions . they are summarized in table [ convtesttable ] . the dynamics of these test runs are compared with the base model in figure [ convcheck ] . the flow patterns hardly show any difference for the models with different grids in the @xmath32 direction . on the other hand , as the grid becomes coarser in the @xmath27 direction , the shock waves are slightly broadened , while complex flow patterns particularly behind the companion star are changed even qualitatively . the removed mass , which is our main concern in this paper , is hardly affected , however . its values are listed on the table for each model . note that the number of tracer particles is unchanged in these experimental computations . although there is a clear trend that the removed mass decreases as the number of grid points increases , the variation is small , @xmath45 , and does not affect our conclusion . we hence believe that the resolution employed in this study is appropriate . .spatial resolutions adopted in convergence tests [ cols="<,^,^,^",options="header " , ] the results for these tests are summarized in table [ phystests ] . it is evident from the vd - series , in which only the velocity was changed , that the removed mass increases as the velocity rises just as expected . interestingly , the comparison between the vd - series and the dd - series , in which we varied density alone , reveals that if the change in kinetic energy is the same , changing velocity has a slightly larger effect on the amount of removed mass than changing density . this is also vindicated by the ef - series , in which both density and velocity were modified but the kinetic energy was fixed . the latter result may imply that increase in the incident momentum decreases removed mass as long as the incident kinetic energy is identical . now we know how the density and/or velocity of sne affect the removed mass , our next question will be why ? we have observed that direct mass stripping was extremely inefficient in our models because the sne is directed away from the companion star once the bow shock is formed . most of the mass loss occurs instead by ablation owing to the shock heating of stellar matter . this means that the energy , not the momentum , imparted to the companion star is most important . then the shock strength will be the physical parameter that we should look into , since it determines how much the star will be heated . to see how the shock strength depends on the incident density , we consider a riemann problem as presented in figure [ riemannillu ] . this is meant to be a very crude approximation to the collision of sne against the companion star . the left side of the initial discontinuity represents the sne , taking typical values of pressure and velocity from our simulation ( @xmath46 ) . we vary @xmath47 to see how this affects the strength of the forward shock produced by collision . for the right state approximating the envelope of the companion star , we use average values of density and pressure from our companion model : @xmath48 . here cgs units are adopted . we assume that the companion star is at rest . we employ the same assumption as in @xcite to determine the maximum value of @xmath47 : the ejecta mass is distributed in a uniform shell with a thickness of @xmath49 the shock radius . this gives us the maximum density as a reference : @xmath50 where @xmath51 is the mass of sne . ( 200,60 ) ( 100,0)(0,1)60 ( 50,50)ejecta ( 125,50)star ( 10,35)@xmath47:variable ( 10,20)@xmath52erg/@xmath53 ( 10 , 5)@xmath54cm / s ( 110,35)@xmath55 ( 110,20)@xmath56erg/@xmath53 ( 110 , 5)@xmath57 cm / s the solutions of the riemann problem are given in figure [ riemannevo ] . the left panels show the pressure distributions and the right panels show the density counter part . each row hold results for different initial conditions , i.e. @xmath58 for the top panels , @xmath59 for the middle panels , and @xmath60 for the bottom panels . the different colours of lines correspond to different times . we can see jumps in density and pressure at the positions of the forward shock propagating through the companion star and sweeping up mass . there is also a reverse shock propagating away from the companion into the sne , which tries to push it back toward the star . a contact surface divides the sne from the companion and density but not pressure has a jump there . the region between the forward and reverse shocks corresponds to the matter heated by one of these shocks . figure [ riemanndep ] shows the pressure divided by the post - shock density @xmath61 , a measure of the shock strength , as a function of @xmath47 . we again find a power law , whose exponent is rather sensitive to the value of @xmath62 . in our simulations , the forward shock initially has a rather flat front perpendicular to the symmetry axis . it maintains a nearly constant speed until the reverse shock starts to expand toward the primary star . this launch of the reverse shock is accompanied by the generation of a rarefaction wave , which in turn propagates towards the forward shock . it eventually catches up and sucks the energy that would otherwise be injected to the shock . it is hence natural to expect that the total energy given to the companion is determined at this point . based on the above observation , we estimate the removed mass as follows . figure [ mheat ] displays a schematic picture that gives the idea and notations we use here . assuming that the shock front is a plane perpendicular to the symmetry axis , we approximate the shock - heated portion of the companion star by the shaded region in the picture . @xmath63 is the distance from the stellar surface to the shock front and @xmath64 is the mass contained in this shocked region . we show in figure [ slicemass ] @xmath64 as a function of @xmath63 calculated as @xmath65 it is probably by coincidence that this function can be fit by a simple power law very well , which allows us to approximate @xmath64 as @xmath66 if we assume that the shock propagates at a constant velocity , we have @xmath67 , in which @xmath34 is the time after collision . as argued earlier , the main shock heating stops at @xmath68 , when the reverse shock starts to expand outwards . we find in our models that @xmath69 roughly corresponds to the time , at which the instantaneous unbound mass reaches the maximum ( see figure [ timemub ] ) , and that this time does not differ so much among models . hence assuming that @xmath69 is constant , we obtain @xmath70 at @xmath68 . the total energy , @xmath71 , deposited by the forward shock is estimated as follows , @xmath72 where @xmath73 is the specific internal energy and is assumed to be proportional to @xmath74 ; in the second line we employed the relations @xmath70 just obtained and @xmath75 in figure [ riemanndep ] . in the third line , on the other hand , we made use of the approximation @xmath76 . this energy is redistributed in the removed mass via hydrodynamical interactions . it is hence natural to assume that the removed mass is proportional to @xmath71 . this hypothesis finally leads us to the relation @xmath77 , which is pretty close to the power law with @xmath78 ( see eq.([eq : mub_den_relation ] ) ) obtained in the simulations . in our simulations , we cut out the central core of the companion star . although this makes it difficult to accurately estimate the kick velocity imparted by the sne , it is still possible to make rough estimations on this so called `` rocket effect '' @xcite . we calculate the imbalance between the total momentum that flows in from the ejecta side and the momentum that flows out of the other side and find that it is @xmath79 for the base model with @xmath33 . this should be equal to the momentum deposited to the remnant of the companion star . the kick velocity obtained this way is @xmath80 km s@xmath81 , which is comparable to the orbital velocity @xmath82 km s@xmath81 and could destroy the binary system . note , however , this will be an upper limit , and it decreases as @xmath83 with the binary separation according to the solid angle of the companion . we have performed hydrodynamical simulations of sne in binary systems . the removed mass can reach up to approximately @xmath1 of the original mass , and decreases according to a power law of the binary separation . the power is determined by the incident density , which decides the strength of the bow shock . the effect of sne hitting the companion star can become non - negligible at close separations . such close separations can be created by common envelope evolution , or mass accretion from a tertiary star @xcite . it is also known from observation that the orbital periods of massive binaries range @xmath84 day to @xmath85 days , with a preference for short periods , meaning that binaries with close separations are popular @xcite . when such mass is removed from the companion star , it could alter the evolution from then on . it could change the type of sn if the companion is large enough to cause ccsne , or could induce kicks . simulations with more realistic binary models will be useful for future detections of surviving companions in sne such as iptf 13bvn , sn1993j etc . observations of sn1993j suggest that the companion star was a b type star @xcite , which is more compact compared with the companion model in this study . mass stripping off this compact companion star will surely be less efficient if placed at the same separation as in our simulations . this system may have been fairly close , though , since the companion must have a smaller stellar radius . since the removed mass strongly depends on incident density and/or velocity rather than the @xmath86 ratio , more mass might have been stripped off this companion by sne despite its compactness . we have not studied the further evolution of the binary as a system , in our simulations . the sudden mass loss should lead to high eccentricities , or could destroy the system . but if the system survives , the expansion of the companion may engulf the remaining neutron star , leading to common envelope evolution @xcite . this could create even closer separations , possibly forming hmxbs . after a short time , the secondary star will reach its end of evolution , and cause the second sn . if it is possible for the system to survive this explosion , it may lead to the formation of ns - ns binaries or ns - bh binaries . this study may be the first step to possibly reveal some necessary conditions in forming these compact binaries . this study was supported by the grants - in - aid for the scientific research ( a ) ( nos . 24244036 , 24740165 ) , the grants - in - aid for the scientific research on innovative areas , `` new development in astrophysics through multi messenger observations of gravitational wave sources '' ( no . 24103006 ) . 99 bersten , m.c . , benvenuto , o.g . , folatelli , g. et al . , 2014 , , in press cheng , a. 1974 , 31 , 49 de vries , n. , portegies zwart , s. , & figueira , j. , 2013 , , 438 , 1909 duchne , g. & kraus , a. 2013 , , 51 , 269d eggleton , p. , , 268 , 368 fox , o.d . , azalee bostroem , k. , van dyk , s.d . 2014 , , 790 , 17 fremling , c. , sollerman , j. , taddia , f. et al . , 2014 , , 565 , 114 fryxell , b. a. , & arnett , w. d. 1981 , , 243 , 994 geldzahler , b.j . , pauls , t. , & salter , c.j . 1980 , , 84 , 237 kasen , d. 2010 , , 708 , 1025 kurganov , a. , & tadmor , e. 2000 , jcoph , 160 , 241 liu , z - w . , kromer , m. , fink , m. et al . 2013 , , 778 , 121 liu , z - w . , pakmor , r. , seitenzahl , i.r . 2013 , , 774 , l37 livne , e. , tuchman , y. & wheeler , j.c . 1992 , , 399 , 665 marietta , e. , burrows , a. , & fryxell , b. 2000 , , 128 , 615 maund , j.r . , smartt , s.j . , kudritzki , r.p . , podsiadlowski , p. , & gilmore , g.f . 2004 , , 427 , 129 nomoto , k. 1982 , , 253 , 798 pan , k .- c . , ricker , p. , & taam , r. , 2012 , , 750 , 151p pan , k .- c . , ricker , p. , & taam , r. , 2013 , , 773 , 49p pijloo , j. t. , caputo , d. p. , & portegies zwart , s. f. 2012 , , 424 , 2914 podsiadlowski , ph . , langer , n. , poelarends , a.j.t . et al . , 2004 , , 612 , 1044 raghavan , d. , mcalister , h. a. , henry , t. j. et al . , 2010 , , 190 , 1r sana , h. , de mink , s. e. , de koter , a. et al . , 2012 , sci , 337 , 444 sawai , h. , yamada , s. , kotake , k. , & suzuki , h. 2013 , , 764 , 10 seward , f.d . , charles , p.a . , foster , d.l . 2012 , , 759 , 123 smith , n. , li , w. , filippenko , a.v . , & chornock , r. 2011 , , 412 , 1522 wellstein , s. & langer , n. , 1999 , , 350 , 148 wheeler , j. c. , lecar , m. , mckee , c. f. 1975 , , 200 , 145 woosley , s.e . , heger , a. & weaver , t.a . 2002 , , 74 , 1015 young , p. & fryer , c. , 2007 , , 664 , 1033	the majority of massive stars are formed in binary systems . it is hence reasonable to expect that most core - collapse supernovae ( ccsne ) take place in binaries and the existence of a companion star may leave some imprints in observed features . having this in mind , we have conducted two - dimensional hydrodynamical simulations of the collisions of ccsne ejecta with the companion star in an almost - equal - mass ( @xmath0 ) binary to find out possible consequences of such events . in particular we pay attention to the amount of mass removed and its dependence on the binary separation . in contrast to the previous surmise , we find that the companion mass is stripped not by momentum transfer but by shock heating . up to @xmath1 of the original mass can be removed for the closest separations and the removed mass decreases as @xmath2 with the binary separation @xmath3 . by performing some experimental computations with artificially - modified densities of incident ejecta , we show that if the velocity of ejecta is fixed , the density of incident ejecta is the single important parameter that actually determines the removed mass as @xmath4 . on the other hand , another set of simulations with modified velocities of incident ejecta demonstrate that the strength of the forward shock , which heats up the stellar material and causes the mass loss of the companion star , is actually the key parameter for the removed mass .
starburst galaxies are characterized by a nebular emission - line spectrum at optical and an absorption - line spectrum at ultraviolet wavelengths . this dichotomy arises because starbursts are powered by massive stars ( m@xmath11 10 m@xmath8 ) . these stars emit photons with energies of a few tens of ev which are absorbed and re - emitted in their stellar winds , producing ultraviolet resonance transitions , and thus , an ultraviolet absorption line spectrum . however , most of the ionizing photons emitted by the stars can travel tens of pc or more before they are absorbed by the surrounding interstellar gas . then , this ionized gas emits a nebular emission - line spectrum , mainly at optical and near - infrared wavelengths . however , around the balmer jump the spectra can show absorption features formed in the photospheres of massive stars . the radiative properties of starburst galaxies are determined by their massive stellar content . o , b and a stars can dominate the optical continuum emission of starburst galaxies . the spectra of the early - type stars are characterized by strong hydrogen balmer and neutral helium absorption lines and with only very weak metallic lines ( walborn & fitzpatrick 1990 ) . however , the detection of these stellar features at optical wavelengths in the spectra of starburst galaxies is difficult because h and hei absorption features are coincident with the nebular emission - lines that mask the absorption features . even so , the higher - order terms of the balmer series and some of the hei lines are detected in absorption in many starburst galaxies ( storchi - bergmann , kinney , & challis 1995 ; gonzlez - delgado et al . 1995 ) , in seyfert 2 galaxies with their ultraviolet and optical continuum dominated by a nuclear starburst ( shields & filippenko 1990 ; gonzlez delgado et al . 1998 ; cid fernandes , storchi - bergmann , & schmitt 1998 ) , or even in the spectra of giant hii regions ( e.g. ngc 604 , terlevich et al . these features can be seen in absorption because the strength of the balmer series in emission decreases rapidly with decreasing wavelength , whereas the equivalent width of the stellar absorption lines is constant or increases with wavelength . thus , the net effect is that h@xmath12 , h@xmath0 and h@xmath1 are mainly seen in emission , but the higher - order terms of the series are seen in absorption . very often , h@xmath12 , h@xmath0 , h@xmath1 and h@xmath2 show absorption wings superposed on the nebular emission . a similar effect occurs for the hei lines . the strongest nebular emission occurs in the hei @xmath35876 and hei @xmath36678 whereas the stellar absorption features are very weak . meanwhile , the equivalent width of the nebular emission of hei @xmath34471 or hei @xmath34026 ( @xmath6 6 , izotov , thuan , & lipovetsky 1997 ) is similar or even smaller than that of the stellar absorption lines . thus , they can be easily detected in absorption . models of the h balmer series and hei lines of a stellar population in a starburst galaxy can be used to predict the evolutionary state of a starburst , the effect of the initial mass function ( imf ) and corrections to the nebular h and hei emission - lines . accurate measurements of the nebular emission lines are needed to estimate the amount of internal interstellar reddening , star formation rate , excitation parameter and chemical abundances ( in particular in the determination of the primordial helium abundance ) in starburst galaxies . all these estimations are affected by the underlying h and hei stellar absorption . the internal interstellar reddening is derived using the balmer decrement . the emission - line ratios ( h@xmath12/h@xmath0 , h@xmath1/h@xmath0 , h@xmath2/h@xmath0 , ... ) are clearly affected by the stellar absorption . they are not affected equally because nebular emission decreases with wavelength . reddening and underlying absorption affect line ratios such as [ oii ] @xmath33727 + [ oiii ] @xmath35007/h@xmath0 and [ oiii ] @xmath35007/h@xmath0 which are used to estimate chemical abundances and the excitation parameter in starburst galaxies . the strength of the collisionally - excited emission - lines with respect to h@xmath0 can be used to estimate the evolutionary state of starburst galaxies ( garca - vargas , bressan , & daz 1995 ; garca - vargas et al . 1997 ; gonzlez delgado et al . 1999 ) ; higher line ratios are derived if h@xmath0 is not corrected for the stellar absorption . the underlying stellar absorption in the hei lines leads to an understimate of the he abundance in starburst galaxies , and potentially of the determination of the primordial helium abundance . however , high resolution spectra are needed to isolate the absorption lines from the nebular emission . on the other hand , the strength of the higher - order lines of the balmer series and some of the hei lines ( e.g. hei @xmath34026 , @xmath33819 ) , which are likely to be detected in absorption in starburst galaxies can be used to obtain information about the age and imf of their stellar population . it has been suggested that starburst galaxies have a deficit of intermediate and low - mass stars ( rieke et al . 1980 ; viallefond & thuan 1983 ; augarde & lequeux 1985 ; scalo 1990 ) . however , the evidence in favor of this top - heavy imf in starburst galaxies is ambiguous ( satyapal et al 1995,1997 ) . it should be expected that the strength of the balmer absorption lines depends on the low mass cut - off because these lines are stronger in a stars . thus , if the mass of the starburst is distributed following a top - heavy imf , we expect that the h balmer lines are weaker than for the population with a non - truncated salpeter imf . h balmer and hei lines of a stellar population can be predicted using the evolutionary synthesis technique , a powerful tool to derive the properties of a stellar population taking as a free parameter the star formation history of the starburst ( age , imf , star formation rate ) . this technique has been used previously to derive the strength of the balmer and he i absorption lines of star forming regions . daz ( 1988 ) used the synthetic profiles of h@xmath12 , h@xmath0 , h@xmath1 and h@xmath2 computed by kurucz ( 1979 ) . she followed the evolution of the stellar cluster up to 40 myr and found that the profiles are rather insentive to the age and imf . olofsson ( 1995 ) computed the equivalent widths of the hydrogen ( h@xmath12 , h@xmath0 , h@xmath1 , h@xmath2 and l@xmath12 ) and hei ( @xmath34026 , 4387 , 4471 and 4921 ) lines of a single - burst star forming regions at three different metallicities ( @xmath13)=0.001 , 0.008 and 0.020 ) , and followed the evolution up to 15 myr . he used the synthetic models of kurucz ( 1993 ) and auer & mihalas ( 1972 ) at a resolution of 5 . he found that the equivalent width of the balmer lines increases during the course of the evolution , while hei decreases monotonically . the strength of these lines are very sensitive to the imf and insensitive to the metallicity . he suggested that balmer lines are prominent even if the low mass cut - off is as high as 10 m@xmath8 . cananzi , augarde & lequeux ( 1994 ) predict the balmer lines ( h@xmath0 , h@xmath1 and h@xmath2 ) using stellar spectra of o to g type observed at a spectral resolution of 2 ( cananzi , augarde , & lequeux 1993 ) . they conclude that the equivalent widths of the lines are more sensitive to the low than to the high - mass cut - off of the imf . the results seem contradictory : it is not yet clear if the imf parameters have an effect on the strength of the h balmer and hei lines . on the other hand , none of the previous studies made predictions for the strength of the higher - order balmer series lines ( h8 to h13 ) which are least affected by the emission and the most useful to obtain information on the evolutionary state of the starburst . due to the coincidence of the nebular emission and absorption lines in the observed spectra of starbursts , models that compute the profiles of these lines at high spectral resolution are required to predict the underlying absorption through the analysis of the absorption wings . to archieve this goal , we have built a stellar library of synthetic spectra with a sampling of 0.3 , effective temperature from 50000 to 4000 k and gravity between 0.0@xmath6log [email protected] ( gonzlez delgado & leitherer 1999 , paper i ) . the profiles are computed using a set of programs developed by hubeny and collaborators ( hubeny 1988 ; hubeny & lanz , 1995a , b ; hubeny , lanz , & jeffery , 1995 ) in three stages . for t@xmath14 25000 k , the code tlusty is used to compute nlte stellar atmosphere models . these models together with kurucz ( 1993 ) lte stellar atmosphere models ( for t@xmath7 25000 k ) are used as input to synspec , the program that solves the radiative transfer equation . finally , the synthetic spectrum is obtained after performing the rotational and instrumental convolution . this library has been included in our evolutionary synthesis code ( starburst99 , leitherer et al 1999 ) to predict the strength of the balmer and he i lines for a stellar population . this paper is organized as follows : section 2 describes the models , the assumptions and computational techniques . section 3 deals with the results and discusses the effects of the imf parameters and metallicity on the strength of the lines . we compare the equivalent widths of the balmer and hei absorption lines with the values predicted for the nebular recombination emission - lines in section 4 . in section 5 , the models are compared with observations of stellar clusters of the large magellanic clouds , the super - star clusters in the starburst galaxy ngc 1569 , the nucleus of the dwarf elliptical galaxy ngc 205 and a luminous `` e+a '' galaxy . the summary and the conclusions are in section 6 . the stellar library of synthetic spectra which covers the main hydrogen balmer and neutral he lines ( paper i ) , has been implemented in our evolutionary synthesis code starburst99 ( leitherer et al . the synthesis is done in two steps . first , the code computes the population of stars in the cluster as a function of age and imf ; then the profiles of the lines are synthesized by adding the different contributions from stars . two limiting models of the star formation history are considered : an instantaneous burst and star formation proceeding continuously at a constant rate . in the instantaneous burst , all the stars are formed at the same time on the zero - age - main - sequence ( zams ) , and then evolve in the hertzsprung - russell diagram ( hrd ) until 1 gyr . these models are normalized to a total mass of 10@xmath4 m@xmath8 . the star formation rate of the continuous model is 1 m@xmath8 yr@xmath15 ; they extend to 1 gyr . the models are generated for different assumptions about the imf , which is parameterized as a power law , @xmath16(m)= c m@xmath17 ; where the constant @xmath18 is determined by the total gas mass converted into stars . the reference model is a power law with exponent @xmath12=2.35 , and with low - mass and high - mass cut - off masses of m@xmath19= 1 m@xmath8 and m@xmath20= 80 m@xmath8 . for comparison , we have computed models with @xmath12= 3.3 and @xmath12= 1.0 between @xmath21= 1 m@xmath8 and @xmath22= 80 m@xmath8 ; and a truncated salpeter ( @xmath12 = 2.35 ) imf between @xmath21= 1 m@xmath8 and @xmath22= 30 m@xmath8 , @xmath21= 5 m@xmath8 and @xmath22= 80 m@xmath8 , and @xmath21= 10 m@xmath8 and @xmath22= 80 m@xmath8 . the models are computed with the isochrone synthesis method . thus , continuous isochrones are calculated by interpolating between the tracks in the hrd on a variable mass grid . the stars evolve from the zams following the new set of evolutionary tracks of the geneva group ( schaller et al . 1992 , schaerer et al . 1993a , b , charbonnel et al . 1993 , meynet et al . 1994 ) , with standard - mass loss rate . the time resolution of the models is 0.01 myr ; however , the profile of the lines are given only at time steps during which significant changes occur . for every step , the luminosity , effective temperature , radius and surface gravity are calculated for each star . then a spectrum of our stellar library is assigned to each star . the corresponding synthetic spectrum is flux calibrated using the stellar atmosphere grid compiled by lejeune et al ( 1997 ) . this compilation which includes the 20 resolution kurucz ( 1993 ) stellar atmospheres is supplemented by schmutz , leitherer , & gruenwald ( 1992 ) models for stars with expanding envelopes . the flux of the continuum is assigned to each normalized spectrum by fitting a first order polynomial to the plane ( f@xmath23 , log @xmath3 ) , where f@xmath23 is the distribution of the continuum flux of the corresponding low resolution stellar atmosphere model . the fit is done taking points of the continuum adjacent to the absorption lines h@xmath0 , h@xmath1 , h@xmath2 and hei @xmath34471 for the spectral ranges 4820 - 4950 , 4270 - 4430 , 3990 - 4150 and 4420 - 4580 , respectively . to calibrate the stellar spectra between 3720 and 3920 , we use the result of the fit in the wavelength range of h@xmath2 ( if t@xmath14 12000 k ) or h@xmath0 ( if t@xmath7 12000 k ) extrapolated to those wavelengths . the calibration was chosen to achieve consistency between the distribution of the continuum flux of our synthesized high - resolution spectra and the spectral energy distribution which is also predicted by our code at a resolution of 20 ( see figure 1 ) . the nebular continuum has been added to the synthesized spectra . we have assumed that all the stellar far - uv photons below 912 are absorbed by gas and converted into continuous and line emission at longer wavelengths . we have computed the synthesized profiles for two metallicities , @xmath13=0.02 ( z@xmath8 ) and z=0.001 ( 1/20 z@xmath8 ) . while the input stellar evolutionary tracks are different for these two metallicities , we use the same stellar library ( generated assuming solar metallicity ) because the profiles of the balmer lines of individual stars do not change significantly with metallicity for t@xmath14 7000 k ( see paper i ) . figure 2 shows the synthetic spectra for a star cluster formed instantaneously 3 , 50 and 500 myr ago . the imf is salpeter with @xmath21= 1 m@xmath8 and @xmath22= 80 m@xmath8 ; the metallicity is solar . the general tendency is that the strength of the balmer lines increases with the evolution until 500 myr , and the hei lines until 30 - 40 myr . the equivalent widths of the balmer ( table 1 and 2 ) and he i ( table 3 and 4 ) lines are measured in the windows indicated in the figure 2 . two different measurements are made , one integrating the flux from the continuum set to unity and the other from a pseudo - continuum . this pseudo - continuum is found fitting a first - order polynomial ( except for the spectral range 3720 - 3920 , for which we use a third - order fit ) to the continuum windows defined in table 3 of paper i. the absolute values of the equivalent widths depend on the continuum level and on the width on the windows used to measure the strength of the lines . however , the relative behavior of the equivalent width with age of the star cluster is independent of the continuum and of the width of the windows . we plot the equivalent widths measured from the pseudo - continuum because these measurements are more representative of the equivalent widths we can measure in observed spectra . first , we discuss the profiles and equivalent widths of h balmer and he i lines for the reference model ( salpeter imf , m@xmath19= 1 m@xmath8 and m@xmath20= 80 m@xmath8 ) for an instantaneous burst and for continuous star formation . then , the effect of the imf parameters and finally the effect of the metallicity are discussed . the full data set is available for retrieval at http://www.iaa.es/ae/e2.html and at http://www.stsci.edu / science / starburst/. figure 3 shows the equivalent widths of the balmer lines as a function of the age for an instantaneous burst . we plotted the equivalent width measured from the pseudo - continuum and integrating the flux in the windows of 60 for h@xmath0 , h@xmath1 and h@xmath2 , and 30 for h8 , h9 and h10 , respectively . all the balmer lines have similar behavior ; the equivalent widths are in the range between 2 to 16 . during the first 4 myr , the equivalent width does not change , with h@xmath0 , h@xmath1 and h@xmath2 @xmath24 3 , which is very similar to the values predicted by olofsson ( 1995 ) , and very similar to the value estimated ( @xmath24 2 ) for the balmer absorption in giant hii regions ( shields & searle 1978 , mccall , ribsky , & shields 1985 ) . the equivalent width increases with the age of the burst until 500 myr ; then , it decreases . the maximum is at 500 myr since the turn - off stellar population in the cluster are stars of type a2v that have a life - time of several hundreds myr . these stars have t@xmath25 of 9000 k , and at this t@xmath25 the strength of the balmer lines peaks ( see e.g. figure 2 in paper i ) . the predicted value at 500 myr ( the equivalent width of h@xmath0= 11 - 12 ) is , e.g. , in agreement with the values measured in the super star clusters of ngc 1275 ( brodie et al . 1998 ) and the values predicted by population synthesis models using stellar clusters ( bica & alloin 1986a ) . for continuous star formation , the strength of the lines increases with time until 500 myr ; then , the equivalent widths are constant ( figure 4 ) . the equivalent widths range from 3 to 10 . they are very similar to the burst models at the begining of the evolution ( about 3 for continuous star formation lasting for 4 myr ) . however , the equivalent width increases less rapidly with time than in the burst model . the reason is the dilution of the balmer lines by o stars which have weaker balmer lines than a stars and which are continuously forming in the cluster . figure 5 shows the equivalent width of some of the hei lines . hei @xmath34471 and hei @xmath34026 are the strongest lines ; they have equivalent widths that range between 0.5 and 1.1 . hei @xmath34922 , hei @xmath34388 and hei @xmath33819 are about a factor two weaker than hei @xmath34471 and hei @xmath34026 . this result is opposite to that found by olofsson ( 1995 ) . however , observations of individual galactic b stars ( lennon , dufton , & fitzsimmons 1993 ) show that the lines hei @xmath34471 and hei @xmath34026 have both similar strength and are about a factor of two stronger than hei @xmath34922 and hei @xmath34388 ( these two lines have also similar strength ) . we found the same result examining the spectral atlas of walborn & fitzpatrick ( 1990 ) . the equivalent width of the hei lines is almost constant for bursts younger than 4 myr , with hei @xmath34471 and hei @xmath34026 being about 0.5 , and hei @xmath34922 , hei @xmath34388 and hei @xmath33819 about 0.3 . after 4 myr , the equivalent width increases with age ( but not monotonically ) until the burst age reaches 30 - 40 myr due to the contribution of early - type b stars that show the strongest hei absorption . this result is opposite to that found by olofsson ( 1995 ) . he followed the evolution of the starburst until 15 myr , and he found that the maximum strength of the hei lines is at 2 myr , no hei absorption is formed in his models burst older than 10 myr . lennon et al ( 1993 ) show that stars from o9 to b2 have similar strengths of the hei @xmath34471 and hei @xmath34026 lines ; these are weaker in stars hotter than o9 and cooler than b2 . by definition hei lines are not formed in the photosphere of stars cooler than b9 . these observations are in agreement with the results found in our stellar grid ( see figure 4 of paper i ) and explain why the equivalent width of these lines in the integrated light of a stellar cluster have their maximum values between 10 and 40 myr , when the cluster is dominated by early b stars . hei absorption is not formed for bursts older than 100 myr because the life - time of b9 stars is about this age . the equivalent width has some local maximum between 5 and 20 myr . this is due to the presence of post - main sequence stars with t@xmath78000 k that contribute significantly to the integrated light of the cluster at these ages ( figure 6 ) . these are the blue loops of the red super - giant in the tracks . these stars do not produce hei lines but show many metal lines close to the hei lines ( figure 7 ) . these metal lines are in the spectral windows used to measure the equivalent width . they depress the continuum and contribute to the integrated profile making it broader and asymmetric ( figure 8) . a similar effect is also seen in the h balmer lines . however , the contribution of these metallic lines to the equivalent width of the window where the balmer lines are measured is less than to the windows of the hei lines because the balmer lines are relatively stronger . for continuous star formation , the equivalent width ranges between 0.5 to 0.9 ( for hei @xmath34471 and hei @xmath34026 ) . the strength of these lines increases with time until @xmath9 100 myr ; then , the equivalent widths decrease with time ( figure 9 ) . 40 myr after the onset of the star formation , the equivalent widths of the lines ( hei @xmath34471 and hei @xmath34026 ) is 0.8 . they are weaker than in the instantaneous burst model of the same age . however , due to the presence of b stars in the cluster , the hei lines are stronger in the continuous star formation scenario than in the instantaneous burst after 100 myr . we study the influence of the slope and the low and high mass cut - off of the imf on the profile and equivalent width of the hydrogen balmer and hei lines in a single stellar population . the models computed have imf slopes @xmath12= 1.0 and @xmath12= 3.3 , low mass cut - offs @xmath21= 5 m@xmath8 and @xmath21= 10 m@xmath8 and high mass cut - offs @xmath22= 30 m@xmath8 with respect to the reference model ( @xmath12= 2.35 , @xmath21= 1 m@xmath8 and @xmath22= 80 m@xmath8 ) . figure 10 shows the change of the profiles of h@xmath2 and hei ( @xmath34026 , 4009 and 4143 ) with the slope of the imf for a 2 myr old burst . clusters formed with an imf flatter than salpeter produce hydrogen balmer lines weaker than in the reference model because more massive stars are formed in the cluster . at 2 myr , these stars dilute the strength of the balmer and hei lines since these lines are weaker in o than in b stars . however , after 4 myr the relative number of massive stars still on the main sequence with respect to b and a stars is very low . at that epoch , the profiles and equivalent widths of these lines are almost the same as those of the reference model ( figure 11 ) . hei lines are also weaker than in the reference model if the cluster is younger than 4 myr . these lines are slightly stronger if the cluster is between 20 and 100 myr old ( figure 12 ) . during the post - starburst phase , the integrated light is dominated by b stars , and for a flatter imf , the relative number of b stars with respect to a stars is slightly higher than in the reference model . this effect , however , is very small . the opposite result is found if the imf is steeper than the salpeter imf . in this case the balmer and he i lines are stronger than in the reference model if the burst is younger than 4 myr . between 20 and 100 myr , hei lines are weaker than in the salpeter model , because fewer b stars contribute to the integrated light of the cluster . however , at these ages the change of the strength of these lines ( figure 12 ) is less important than at ages younger than 4 myr , when the equivalent width can change by almost a factor of 2 ( figure 11 and 12 ) . figure 13 shows the change of h@xmath2 and hei ( @xmath34026 , 4009 and 4143 ) with @xmath21 and @xmath22 for a 2 myr old burst . the h@xmath2 and hei lines are stronger if @xmath22= 30 m@xmath8 than in the reference model ; but they are very similar to those predicted by the reference model if @xmath21= 5 m@xmath8 . this result holds as long as the burst is younger than 5 myr because the life - time of stars with mass higher than 30 m@xmath8 is less than 5 myr . thus , if the cluster is younger than 5 myr and the mass is distributed following a standard salpeter imf , the integrated light should be dominated by stars more massive than 30 m@xmath8 . these stars have balmer and hei lines that are weaker than stars less massive than 30 m@xmath8 . in this case , the balmer and hei lines are diluted with respect to the truncated ( @xmath22=30 m@xmath8 ) imf . however , after 5 myr , all the stars more massive than 30 m@xmath8 have evolved from the main sequence and the integrated light shows balmer and hei lines of the same strength . if the cluster is formed following a truncated salpeter imf with @xmath21=5 m@xmath8 , the integrated light shows the balmer and hei lines at a similar strength as in a cluster formed following the standard salpeter imf . the life - time of stars with mass of 5 m@xmath8 is about 100 myr . clusters formed without stars less massive than 5 m@xmath8 will fade after 100 myr when all the stars have evolved from the main - sequence . a stronger effect of a truncated imf is found if @xmath26 5 m@xmath8 . in these cases , balmer and hei lines are always weaker than in the reference model because fewer b and a stars are formed . the main effect of changing the parameters of the imf is observed in the first 4 myr of the evolution of the starburst . if the imf is steeper than the reference one or the upper mass cut - off is much lower than 80 m@xmath8 , fewer o stars are formed and the balmer and hei lines are stronger . if the imf is flatter than salpeter or if the low mass cut - off is higher than 5 m@xmath8 , more o stars are formed relative to b and a stars , and the lines are weaker . however , the ability of the balmer and hei lines to constrain the imf depends strongly on the resolution of the observations , and on the spatial distribution of the stellar cluster with respect to the nebular emission , because at ages younger than 5 myr starbursts show very strong hei and balmer emission - lines that fill in the absorptions . a nearby starburst , that can be spatially resolved , and where the core stellar cluster shines through a hole of nebular emission produced by the stellar winds , will show a spectrum with absorption wings around the balmer emission and with the hei lines in absorption ; ngc 604 is a good example of this ( see figure 9 in terlevich et al 1996 ) . in this cases , the hei lines that are fainter in emission ( e.g. , hei @xmath34026 , @xmath34388 and @xmath34026 ) can be dominated by the stellar absorption . thus , under these circumstances , the equivalent widths of these lines and the wing absorption profiles of the higher terms of the balmer series can be used to constrain the imf ( gonzlez delgado & prez 1999 , in preparation ) . when these circumstances do not occur , the total integrated spectrum of a starburst will be dominated by the nebular emission while it is younger than 5 myr ( see also section 4 ) , and the stellar lines will not be detected ; therefore , any information about the starburst will be obtained through the analysis of the emission - lines . in paper i we demonstrated that the strengths of the balmer and hei lines in _ individual _ stars hotter than 7000 k do not depend on metallicity . however , the strength of these lines in the _ integrated _ light of a stellar cluster depends on metallicity due to the metallicity dependent stellar population . at low metallicity , a star of a given zams mass is hotter and evolves more slowly in the hrd diagram , explaining why the balmer lines are weaker at z= 0.001 than at solar metallicity ( figure 14 , table 5 ) . at any age younger than 500 myr , main - sequence stars that dominate the luminosity of the cluster have higher effective temperature at z= 0.001 than at solar metallicity . after 500 myr , stars hotter than 9000 k are still in the main sequence at z= 0.001 metallicity . therefore , the equivalent width of the balmer lines can still increase with the age of the cluster . in contrast , at solar metallicity , main - sequence stars have effective temperature lower than 9000 k. as a result , the equivalent width decreases with age . the effect is similar for the hei lines . for ages younger than 20 myr , the lines are weaker at z= 0.001 than at solar metallicity , but between 50 - 300 myr there are still main - sequence stars with effective temperature hotter than 12000 k that contribute significantly to the formation of the hei lines . therefore , the equivalent width of the hei lines in this age range is higher at z= 0.001 than at solar metallicity ( figure 15 , table 6 ) . in this section , we compare the equivalent width of the nebular h balmer series and hei lines and the stellar absorption lines . the goal is to quantify the effect of the stellar absorption on the strength of the nebular lines and determine the evolutionary stages of the starburst during which this effect is important . this will also show which of the balmer and hei lines are more useful to age - date starbursts . we calculate the equivalent width of the h balmer and hei nebular emission - lines as the ratio of the luminosity of the emission - line to the continuum , which is estimated from the continuum adjacent to the line and includes the luminosity of the stellar and nebular continuous emission . the luminosities of the h balmer nebular lines is calculated from the relationship between the intensity of the line and the number of lyman continuum photons and the theoretical ratios ( assuming case b ; osterbrock 1989 ) h@xmath1/h@xmath0 ( 0.469 ) , h@xmath2/h@xmath0 ( 0.259 ) and h8/h@xmath0 ( 0.105 ) . the luminosity of the hei recombination lines have been calculated from the ratio hei @xmath34471/h@xmath0 and the theoretical ratio hei @xmath34026/hei @xmath34471 ( 0.474 ) , hei @xmath34922/hei @xmath34471 ( 0.274 ) and hei @xmath33819/hei @xmath34471 ( 0.264 ) . we estimate the ratio hei @xmath34471/h@xmath0 using the photoionization code cloudy ( ferland 1997 ) . the models are computed assuming that the gas is ionization bounded and is spherically distributed around the ionizing cluster with a constant density of 100 @xmath27 . the chemical composition of the gas is solar and the ionizing photon luminosity is fixed to the values predicted by the evolutionary synthesis models . we take as the radiation field the spectral energy distribution predicted by starburst99 ( leitherer et al . models are computed for an instantaneous burst with age between 0 and 20 myr and continuous star formation lasting up to 100 myr . tables 7 and 8 show the ratio hei @xmath34471/h@xmath0 and the equivalent width of the h and hei lines for burst and continuous star formation , respectively . the ratio hei @xmath34471/h@xmath0 decreases with the age of the instantaneous burst , and drops to zero for a burst older than 5 myr . these clusters do not harbor stars that produce a significant number of photons to ionize he into he@xmath28 ; therefore , the ratio should be zero . csf models lasting for more than 5 myr predict a hei @xmath34471/h@xmath0 ratio which is constant with time , because after this age the number of stars able to ionize he is constant since births and deaths are in equilibrium . the stellar absorption is more important for the higher terms of the balmer series and for older instantaneous bursts than for the csf scenario ( figure 16 ) . h@xmath0 is dominated by the stellar absorption if the burst is older than 8 myr ; however , the effect of the stellar absorption is small if the star formation proceeds continuously . thus , at 100 myr , the strength of the h@xmath0 stellar absorption is less than 20@xmath10 of the strength of the nebular emission . however , the effect of the stellar absorption is more dramatic for the higher terms of the balmer series . h@xmath2 and h8 are dominated by the stellar absorption if the burst is older than 4 and 5 myr , respectively . if the star formation proceeds continuously , after 30 myr and 100 myr , the strengths of the stellar absorption lines are equal to those of the nebular emission - lines h8 and h@xmath2 , respectively . the hei nebular emission - lines are more affected by the stellar absorption lines than the hydrogen balmer lines ( figure 17 ) . bursts older than 5 myr do not show hei in emission , and after this age they can be in absorption ( figure 17a ) . hei @xmath33819 is more affected by the stellar absorption than the other lines . the equivalent width of the emission - line equals the absorption equivalent width when the burst is only 3 myr old . as with the balmer lines , hei lines are less affected by the absorption if the star formation proceeds continuously ( figure 17b ) . hei @xmath34471 is very little affected by the absorption ; however , the equivalent width of the emission - line hei @xmath33819 equals the stellar absorption line equivalent width at 20 myr for continuous star formation , and after this time hei @xmath33819 is dominated by the stellar absorption . thus , as we anticipated , the high - order terms of the balmer series and hei @xmath33819 in absorption are good indicators of the evolutionary state of the starburst , because the nebular emission is weaker and the lines are dominated by the stellar component . in this section , we compare the profile and equivalent width of the balmer lines of our models with observations of stellar clusters in the large magellanic cloud ( lmc ) , the super - star cluster b in the starburst galaxy ngc 1569 , the nucleus of the dwarf elliptical galaxy ngc 205 , and a luminous `` e+a '' galaxy to test the applicability of our models . bica & alloin ( 1986b ) observed clusters in the lmc with ages between 10 and 500 myr . the data were grouped in four different types , called y1 , y2 , y3 and y4 , corresponding to ages of 10 myr , 25 myr , 80 myr and 200 - 500 myr , respectively . these spectra are public ( leitherer et al 1996 ) . the spectral range covers 1200 to 9800 , with a resolution in the range 7 - 17 . at optical wavelengths , the template y1 contains the cluster ngc 2004 ( age= 8 myr , [ z]=-0.25 ) ; y2 contains ngc 1847 ( age= 25 myr , [ z]=-0.4 ) , ngc 2157 ( age= 30 myr , [ z]=-0.6 ) and ngc 2214 ( age= 40 myr , [ z]=-1.2 ) ; y3 contains ngc 1866 ( age= 86 myr , [ z]=-1.2 ) ; and y4 contains ngc 1831 ( age= 300 myr , [ z]=-1.0 ) and ngc 1868 ( age= 500 myr , [ z]=-1.1 ) . figure 18 compares the equivalent width of the balmer lines of the four templates measured in the same windows as the models . the figures show the model results for a single burst assuming that the mass of the cluster is distributed with a salpeter imf between 1 and 80 m@xmath8 at z@xmath8 ( full line ) and z= 0.001 ( dashed line ) metallicity . the observations and the models are in very good agreement . however , the equivalent widths of the higher terms of the balmer series of the template y4 ( age=200 - 500 myr ) are weaker than in the models ( figure 18 c y d ) . the discrepancy could be related to the poor spectral resolution of the observations . decreased spectral resolution artificially decreases the continuum level ; thus , decreasing the equivalent width of the observed lines . this effect is more important for template y4 because the balmer lines are stronger and broader for ages between 200 and 500 myr . super - star clusters ( ssc ) are highly compact ( diameter a few pc ) , massive ( 10@xmath29 - 10@xmath4 m@xmath8 ) , young ( age younger than 1 gyr ) clusters . there are suggestions that sscs represent the present - day analogs of young globular clusters because their masses and sizes are comparable to those of milky way globular clusters ( oconnell , gallagher , & hunter 1994 ; ho & filippenko 1996 ) . they represent a basic mode of star formation in starburst galaxies ( meurer et al . 1995 ) and are building blocks of these objects . the clusters can be approximated by a coeval , single metallicity stellar population . hst images of the dwarf starburst galaxy ngc 1569 suggest that the ssc b has an age of 15 - 300 myr ( oconnell et al . 1994 ) and ground - based optical spectra suggest an age of 10 myr ( gonzlez delgado et al . 1997 ) . the latter result is based on the analysis of the optical spectral energy distribution and on the strength of the caii triplet ( see also prada et al . 1994 ) . to test our models , we have compared the synthetic profiles of the balmer lines with the observations ( see gonzlez delgado et al 1997 for details of the observations ) . the models are binned to the spectral resolution of the observations ( the dispersion is 1.4 / pix ) . the balmer lines are partially filled with nebular emission ; therefore the fitting has to be done based on the wings of the absorption features . this effect is less important for the higher terms of the balmer series because the nebular emission drops quickly with decreasing wavelength . figure 19 plots the observed lines and the synthetic models for a burst of 10 and 50 myr at z@xmath8/5 metallicity ( assuming salpeter imf , @xmath21=1 m@xmath8 and @xmath22=80 m@xmath8 ) . the profiles indicate that the balmer lines are more compatible with a 10 myr old burst than with one that is 50 myr old . ages older than 10 myr produce profiles which are wider than the observed ones . this comparison confirms the result of gonzlez delgado et al . ( 1997 ) that the age of the ssc b is about 10 myr and shows that this technique can discriminate well between a young and an intermediate - age population . we have observed the nucleus of the dwarf elliptical galaxy ngc 205 to provide a spectral template of an intermediate age population in order to analyze the stellar population of a sample of seyfert 2 nuclei ( gonzlez delgado et al 1998 ) . the equivalent widths of the balmer lines measured in our spectrum ( ew(h@xmath0)=7.7 , ew(h@xmath1)=7.4 , ew(h@xmath2)=8.8 , ew(h@xmath30)=7.4 ) are weaker than the values predicted by a population in the range 100 - 500 myr ( see figure 14 ) . bica , alloin , & schmidt ( 1990 ) have undertaken the population synthesis of the integrated optical light of the nucleus of ngc 205 . they conclude that the dominant population is in the range 100 - 500 myr ; however , this population only produces @xmath9 50@xmath10 of the optical light . the remaining 50@xmath10 is due to young ( @xmath9 10 myr ) , intermediate ( 1 - 5 gyr ) and old ( @xmath11 10 gyr ) components that contribute with 10@xmath10 , 20@xmath10 and 20@xmath10 of the optical light , respectively . these populations have equivalent widths of the balmer lines that are weaker than the intermediate population of 100 - 500 myr ( at these ages , the balmer lines have their maximum strength ) . thus , the equivalent width of the balmer lines is diluted by the contribution of the other components with respect to a single population of several hundred myrs . we have combined our models contributing to the total light in the fraction derived by bica et al ( 1990 ) and we found that the combined synthesis profile fits very well the observations ( figure 20 ) . oegerle , hill , & hoessel ( 1991 ) reported the serendipitous discovery of a relatively low redshift galaxy in the post - starburst phase that is one of the best examples of the so - called `` e+a '' galaxies ( dressler & gunn 1983 ) . the spectra of these galaxies show very strong balmer absorption lines produced by a large number of a stars . these stars result from a burst of star formation that probably occurred @xmath241 gyr ago , and are mixed with the underlying old stellar population typical of an elliptical galaxy . the equivalent widths of the balmer lines measured in the spectrum of the `` e+a '' galaxy discovered by oegerle et al . ( 1991 ) are ew(h@xmath0)=9.5 , ew(h@xmath1)= 10.6 , ew(h@xmath2)= 8.7 , and ew(h8)= 8.6 . these values are plotted as a horizontal dotted lines in figure 18 . with the exception of h@xmath2 , the strength of the balmer lines indicates that the galaxy is dominated by a population of a stars . the equivalent widths of the lines are well fitted by a burst that occurred in the galaxy 10@xmath31 to 10@xmath5 year ago . profile shapes in spectra of higher resolution could discriminate between the 10@xmath31 and 10@xmath5 year age . we have computed evolutionary synthesis models that predict high - resolution spectra of a stellar popultation in the wavelength range : 3720 - 3920 , 3990 - 4150 , 4300 - 4400 , 4420 - 4580 and 4820 - 4950 . these models predict the absorption line profiles of the hydrogen balmer series ( h@xmath0 , h@xmath1 , h@xmath2 , h9 , h10 , h11 , h12 and h13 ) and the neutral helium lines ( hei @xmath34922 , hei @xmath34388 , hei @xmath34144 , hei @xmath34121 , hei @xmath34026 , hei @xmath34009 and hei @xmath33819 ) as a function of the age , imf parameters and metallicity in two different star formation scenarios : an instantaneous burst and star formation proceeding continuously at a constant rate . models are generated using the isochrone synthesis method assuming that stars evolve from the zams following the set of evolutionary tracks of the geneva group . at each age , the spectrum is computed assigning the corresponding high - resolution spectrum from our stellar library to each star . the library was computed assuming that the atmosphere of the stars hotter than 25000 k is in nlte and cooler stars in lte and spans a range of t@xmath25 between 50000 to 4000 k and gravity 0.0 @xmath6 log g @xmath6 5.0 ( see paper i for details ) . the computed profiles of the balmer and hei lines are useful for comparison with spectra of star - forming regions . high spectral resolution profiles of these lines ( in the models and observations ) are required to constrain the evolutionary state of a very young stellar population through the analysis of the balmer absorption wings under the nebular emission - lines . evolutionary synthesis profiles of the higher terms of the balmer series are very useful , as anticipated , to age - date the starburst because the nebular balmer decrement decreases very rapidly with decreasing wavelength whereas the equivalent widths of the stellar balmer absorption lines are almost constant with wavelength . a 4 myr old burst shows a h8 stellar line of equal strength to the nebular emission - line ; in contrast , at this age h@xmath0 is dominated by nebular emission . hei @xmath33819 is also very useful to age - date starbursts because the nebular emission is weaker than in the other hei lines and is dominated by the stellar absorption if the burst is older than 3 myr . the profiles of the absorption features have been computed for the post - starburst phase up to 1 gyr . these profiles can be used to date super star clusters and nuclei of galaxies dominated by an intermediate age population ( younger than 1 gyr ) . the strength of the balmer absorption lines is sensitive to evolutionary effects . their equivalent widths increase with age until about a few hundred myr when the stellar population is dominated by early - a type stars ; they range between 2 and 16 if the cluster is formed at solar metallicity . during the first 4 myr , the balmer lines are not sensitive to the age because the equivalent widths of the lines are almost constant , h@xmath0 , h@xmath1 and h@xmath2 @xmath9 3 , and h9 and h10 @xmath9 2 , respectively . the balmer lines are always weaker than 6 if the starburst is in the nebular phase ( younger than 10 myr old ) . balmer lines are also sensitive to the change of the imf parameters during the first 4 myr of the evolution of the starburst . stronger lines are formed in the cluster if the mass of the cluster is distributed following a truncated salpeter imf with @xmath22=30 m@xmath8 or if the imf is steeper than salpeter . however , the strength of the lines is less sensitive to the low mass cut - off , contrary to previous suggestions . balmer lines are prominent ( @xmath9 2 ) even when the low mass cut - off is as high as 10 m@xmath8 . weaker lines are formed if the slope of the imf is flatter than salpeter because more massive stars are formed , which have weaker balmer lines than early - type b stars . however , this effect is only important in the first few myr of the evolution of the starburst when massive stars are still on the main sequence . hei lines are also sensitive to the evolution of the starburst . they increase ( but not monotonically ) with age until about 30 - 50 myr ; after this age , they decrease . no hei lines are formed after 100 myr because stars hotter than b8 have evolved from the main sequence , and stars cooler than b8 ( @xmath9 12000 k ) do not show hei lines . the equivalent widths of hei @xmath34471 and hei @xmath34026 , which are the strongest hei lines , range between 0.5 and 1.1 . during the first 4 myr of the evolution of the starburst , the strength of these lines is constant ( @xmath9 0.5 for hei @xmath34471 and hei @xmath34026 and 0.3 for hei @xmath34929 , hei @xmath34388 and hei @xmath33819 ) . we estimate that the correction to the hei @xmath34471 nebular emission - lines due to the underlying absorption is between 5 and 25@xmath10 if the nebular emission has an equivalent width between 10 and 2.5 ( a young burst ) . the hei lines are sensitive to the imf parameters in the same way as the balmer absorption lines . for continuous star formation , the strength of the balmer lines increases with time until 500 myr . the equivalent widths of the hei lines increase with time until 100 myr , then they turn over . the equivalent width of the stellar absorption lines ranges between 3 and 10 , and 0.5 and 0.9 for the balmer and hei lines , respectively . the higher - order balmer absorption lines are more useful to date - age starbursts than h@xmath0 because the latter is mainly dominated by the nebular emission ; in contrast , h8 is dominated by the stellar absorption if continuous star formation lasts for more than 30 myr . hei @xmath33819 is also very useful to date - age starbursts because the line is dominated by the stellar absorption if continuous star formation lasts for more than 20 myr . the advantages of using balmer and hei lines in absorption to date starbursts with respect to using nebular emission - lines are twofold : a ) the age can be constrained in a much wider range , including the nebular and the post - nebular phases ; and b ) the strengths of the absorption lines are not affected by extinction or by the leaking of ionizing photons . the models have been tested by comparing the strength of the balmer lines with those of the stellar clusters in the lmc , ssc b in ngc 1569 , the nucleus of ngc 205 and a luminous `` e+a '' galaxy . very good agreement is found . the full set of models is available at our website http://www.iaa.es/ae/e2.html and at http://www.stsci.edu/science/starburst/ , or on request from the authors at [email protected] we thank ivan hubeny for his help in the initial phase of this project and for making his code available to the community , enrique prez , angeles daz and grazyna stasiska for their helpful suggestions and comments , and bill oegerle for sending us the spectrum of the `` e+a '' galaxy . rgd thanks the stsci for financial support during the preparation of part of this paper . this work was supported the spanish dgicyt grant pb93 - 0139 , and by the nasa ltsa grant nagw-3138 . auer , l. h. , & mihalas , d. 1972 , apjs , 205 , 24 augarde , r. , & lequeux , j. 1985 , a&a , 147 , 273 1986a , a&a , 162 , 21 bica , e. , & alloin , d. 1986b , a&as , 66 , 171 bica , e. , alloin , d. , & schmidt , a. 1990 , a&a , 228 , 23 brodie , j.p . , schroder , l.l . , huchra , j. p. , phillips , a.c . , kissler - patig , m. , & forbes , d. 1998 , aj , 116 , 691 cananzi , k. , augarde , r. , & lequeux , j. 1993 , a&as , 101 , 599 - . 1994 , spac . , 66 , 75 charbonnel , c. , meynet , g. , maeder , a. , schaller , g. , & schaerer , d. 1993 , a&as , 101 , 415 . cid fernandes , r. , storchi - bergmann , t. , & schmitt , h. 1998 , mnras , 297 , 579 daz , a. i. 1988 , mnras , 231 , 57 dressler , a. , & gunn , j.e . 1983 , 270 , 7 ferland , g.j . 1997 , hazy , a brief introduction to cloudy , university of kentucky , department of physiscs and astronomy internal report garca - vargas , m.l . , gonzlez - delgado , r.m . , prez , e. , alloin , d. , daz , a.i . , & terlevich , e. 1997 , apj , 478 , 112 garca - vargas , m.l . , bressan , a. , & daz , a. 1995 , a&as , 112 , 35 gonzlez delgado , r.m . , garca - vargas , m.l . , goldader , j. , leitherer , c. , & pasquali , a. 1999 , apj , 513 , 707 gonzlez delgado , r.m . , heckman , t. , leitherer , c. , meurer , g. , krolik , j , wilson , a.s . , kinney , a.l . , & koratkar , a. 1998 , apj , 505 , 174 gonzlez delgado , r.m . , leitherer , c. , heckman , t. , & cervio , m. 1997 , apj , 483 , 705 gonzlez - delgado , r.m . , prez , e. , daz , a.i . , garca - vargas , m.l . , terlevich , e. , & vlchez , j.m . 1995 , apj , 439 , 604 ho , l. c. , & filippenko , a. v. 1996 , apj , 466 , l83 hubeny , i. 1988 , compt . comm . , 52 , 103 hubeny , i. , & lanz , t. 1995a , apj , 439 , 875 - . 1995b , tlusty -a user s guide hubeny , i. , lanz , t. , & jeffery , c.s . 1995 , synspec -a user s guide kurucz , r.l . 1979 , apjs , 40,1 . 1993 , cd - rom 13 , atlas9 stellar atmosphere programs and 2 km / s grid ( cambridge : smithsonian astrophys . izotov , y. i. , thuan , t. x. , & lipovetsky , v. a. 1997 , apjs , 108 , 1 leitherer , c. , schaerer , d. , goldader , j. d. , gonzlez delgado , r.m . , robert , c. , foo kune , d. , de mello , d. f. , devost , d. , heckman , t. m. 1999 , apjs , in press leitherer , c. , et al . 1996 , pasp , 108 , 1996 lejeune , t. , buser , r. , & cuisinier , f. 1997 , a&as , 125 , 229 lennon , d. j. , dufton , p.l . , & fitzsimmons , a. 1993 , a&as , 97 , 559 mccall , m. l. , rybski , p. m. , & shields , g.a . 1985 , apjs , 57 , 1 meurer , g. , heckman , t. , leitherer , c. , kinney , a. , robert , c. , & garnett , d. 1995 , aj , 110 , 2665 meynet , g. , maeder , a. , schaller , g. , schaerer , d. , & charbonnel , c. 1994 , a&as , 103 , 97 oconnell , r. , gallagher , j. , & hunter , d. 1994 , apj , 433 , 65 oegerle , w. r. , hill , j. m. , & hoessel , j. g. , 1991 , apj , 381 , l9 olofsson , k. 1995 , a&as , 111 , 57 osterbrock , d.e . , 1989 , _ astrophysics of gaseous nebulae and active galactic nuclei _ , university science press prada , f. , greve , a. , & mckeith , c. 1994 , a&a , 288 , 396 rieke , g. , lebofsky , m. , thompson , r. , low , f. , & tokunaga , a. 1980 , apj , 238 , 24 satyapal , s. , et al 1995 , apj , 448 , 611 satyapal , s. , et al 1997 , apj , 483 , 148 scalo , j.m . 1990 , in _ windows on galaxies _ , ed . a. renzini , g. fabbiano , & j.s . gallagher ( dordrecht : kluwer ) 125 schaerer , d. , charbonnel , c. , meynet , g. , maeder , a. , & schaller , g. 1993a , a&as , 102 , 339 schaerer , d. , meynet , g. , maeder , a. , & schaller , g. 1993b , a&as , 98 , 523 schaller , g. , schaerer , d. , meynet , g. , & maeder , a. 1992 , a&as , 96 , 269 schmutz , w. , leitherer , c. , & gruenwald , r. 1992 , pasp , 104 , 1164 shields , g. a. , & searle , l. 1978 , apj , 222 , 821 shields , j.c . , & filippenko , a.v . 1990 , aj , 100 , 1034 storchi - bergmann , t. , kinney , a. , & challis , p. 1995 , apj , 98 , 103 terlevich , e. , daz , a.i . , terlevich , r. , gonzlez delgado , r.m . , prez , e. , & garca - vargas , m.l . 1996 , mnras , 279 , 1219 viallefond , f. , & thuan , t.x . 1983 , apj , 269 , 444 walborn , n. r. , & fitzpatrick , e. l. 1990 , pasp , 102 , 379 0 & 2.7 & 2.8 & 3.2 & 2.5 & 2.0 & 1.7 1 & 2.6 & 2.8 & 3.1 & 2.4 & 2.0 & 1.7 2 & 2.6 & 2.7 & 3.1 & 2.3 & 2.0 & 1.7 3 & 2.7 & 2.7 & 3.2 & 2.4 & 2.1 & 1.9 4 & 3.4 & 3.2 & 4.0 & 3.1 & 2.8 & 2.5 5 & 4.9 & 4.2 & 5.4 & 4.3 & 4.6 & 3.0 6 & 4.6 & 3.8 & 5.0 & 3.8 & 3.7 & 3.0 7 & 5.8 & 3.8 & 5.9 & 4.3 & 4.1 & 3.3 8 & 4.0 & 4.3 & 5.7 & 4.6 & 4.3 & 3.6 9 & 4.5 & 4.4 & 6.1 & 4.7 & 4.4 & 3.7 10 & 4.5 & 4.2 & 6.0 & 4.6 & 4.2 & 3.7 11 & 4.5 & 4.1 & 6.1 & 4.7 & 4.2 & 3.8 12 & 4.6 & 4.5 & 6.2 & 4.8 & 4.2 & 3.8 13 & 5.5 & 4.8 & 6.7 & 5.1 & 4.6 & 4.0 14 & 5.8 & 5.6 & 7.2 & 5.6 & 5.7 & 4.2 15 & 6.8 & 6.8 & 8.1 & 6.7 & 6.5 & 4.9 16 & 5.7 & 5.7 & 6.7 & 5.5 & 5.1 & 4.1 17 & 5.6 & 5.7 & 6.6 & 5.3 & 4.9 & 4.1 18 & 5.7 & 5.7 & 6.7 & 5.4 & 4.9 & 4.1 19 & 5.9 & 6.0 & 7.0 & 5.6 & 5.2 & 4.3 20 & 6.1 & 6.1 & 7.1 & 5.8 & 5.3 & 4.4 25 & 6.7 & 6.7 & 7.8 & 6.4 & 6.0 & 4.8 30 & 7.2 & 7.4 & 8.7 & 6.9 & 6.4 & 5.2 35 & 7.5 & 7.8 & 9.1 & 7.2 & 6.6 & 5.4 40 & 7.5 & 7.9 & 9.2 & 7.2 & 6.7 & 5.3 45 & 7.7 & 8 & 9.4 & 7.3 & 6.8 & 5.4 50 & 7.6 & 7.9 & 9.3 & 7.1 & 6.7 & 5.3 60 & 7.6 & 7.9 & 9.4 & 7.1 & 6.6 & 5.3 70 & 7.9 & 8.2 & 9.7 & 7.4 & 6.9 & 5.6 80 & 8.1 & 8.4 & 10 & 7.6 & 7 & 5.6 90 & 8.5 & 8.8 & 10.4 & 7.9 & 7.2 & 5.8 100 & 8.7 & 9.0 & 10.6 & 8.1 & 7.4 & 5.9 200 & 10.5 & 10.9 & 13.0 & & 9.1 & 6.9 300 & 11.5 & 12.4 & 14.8 & 11.1 & 10.1 & 7.5 400 & 12.3 & 13.0 & 15.8 & 11.7 & 10.5 & 7.5 500 & 11.6 & 13.0 & 16.2 & 11.5 & 10.6 & 7.3 600 & 11.1 & 12.4 & 15.8 & 11.2 & 10.4 & 7.1 700 & 10.4 & 11.1 & 14.5 & 10.5 & 10.2 & 6.8 800 & 9.7 & 10.2 & 14.5 & 10.0 & 10.0 & 6.6 900 & 8.8 & 8.8 & 12.1 & 9.3 & 9.8 & 6.2 1000 & 8.3 & 8.1 & 11.6 & 8.8 & 9.7 & 6.1 0 & 2.6 & 2.8 & 3.2 & 2.5 & 2.0 & 1.7 1 & 2.6 & 2.8 & 3.1 & 2.4 & 2.0 & 1.7 2 & 2.6 & 2.8 & 3.1 & 2.4 & 2.0 & 1.8 3 & 2.6 & 2.7 & 3.1 & 2.4 & 2.0 & 1.8 4 & 2.7 & 2.9 & 3.3 & 2.5 & 2.2 & 1.9 5 & 3.2 & 3.0 & 3.7 & 2.9 & 2.6 & 2.2 6 & 3.4 & 3.1 & 3.8 & 3.0 & 2.8 & 2.2 7 & 3.6 & 3.3 & 4.1 & 3.2 & 3.0 & 2.4 8 & 3.8 & 3.3 & 4.2 & 3.3 & 3.1 & 2.4 9 & 3.8 & 3.4 & 4.3 & 3.4 & 3.2 & 2.5 10 & 3.8 & 3.5 & 4.4 & 3.4 & 3.2 & 2.5 15 & 4.1 & 3.7 & 4.8 & 3.7 & 3.5 & 2.8 20 & 4.3 & 4.1 & 5.1 & 4.0 & 3.7 & 3.0 30 & 4.6 & 4.4 & 5.6 & 4.3 & 4.1 & 3.3 40 & 4.8 & 4.7 & 6.0 & 4.7 & 4.3 & 3.6 50 & 5.0 & 5.1 & 6.3 & 4.9 & 4.5 & 3.8 60 & 5.1 & 5.3 & 6.5 & 5.0 & 4.7 & 3.9 70 & 5.1 & 5.5 & 6.6 & 5.2 & 4.8 & 4.0 80 & 5.2 & 5.7 & 6.7 & 5.2 & 4.9 & 4.1 90 & 5.4 & 5.9 & 6.9 & 5.3 & 4.9 & 4.1 100 & 5.6 & 6.1 & 7.1 & 5.4 & 5.0 & 4.2 200 & 6.5 & 6.6 & 7.9 & 6.3 & 5.7 & 4.7 300 & 7.1 & 7.2 & 8.6 & 6.6 & 6.0 & 4.8 400 & 7.5 & 7.6 & 9.2 & 7.0 & 6.4 & 5.1500 & 7.8 & 8.0 & 9.6 & 7.2 & 6.6 & 5.2600 & 8.0 & 8.2 & 10.0 & 7.4 & 6.8 & 5.3700 & 8.2 & 8.3 & 10.2 & 7.5 & 6.9 & 5.3800 & 8.2 & 8.4 & 10.3 & 7.6 & 7.0 & 5.4900 & 8.2 & 8.5 & 10.3 & 7.6 & 7.0 & 5.4 1000 & 8.2 & 8.4 & 10.3 & 7.6 & 7.1 & 5.3 0 & 0.25 & 0.47 & 0.19 & 0.52 & 0.29 1 & 0.24 & 0.47 & 0.18 & 0.52 & 0.29 2 & 0.28 & 0.51 & 0.19 & 0.55 & 0.32 3 & 0.40 & 0.60 & 0.25 & 0.60 & 0.40 4 & 0.43 & 0.61 & 0.32 & 0.64 & 0.41 5 & 0.96 & 0.91 & 0.73 & 0.94 & 0.64 6 & 0.90 & 0.87 & 0.64 & 0.81 & 0.59 7 & & 1.2 & 0.96 & 1.0 & 0.77 8 & 0.52 & 0.80 & 0.47 & 0.87 & 0.65 9 & 0.61 & 0.89 & 0.50 & 0.94 & 0.71 10 & 0.60 & 0.94 & 0.49 & 0.97 & 0.75 11 & & 0.94 & 0.43 & 1.0 & 0.73 12 & 0.49 & 0.94 & 0.40 & 1.0 & 0.74 13 & & 1.2 & 0.70 & 1.1 & 0.85 14 & 0.92 & 1.1 & 0.81 & 1.2 & 0.92 15 & 0.67 & 0.90 & 0.69 & 1.0 & 0.64 16 & 0.49 & 0.76 & 0.52 & 0.82 & 0.53 17 & 0.49 & 0.80 & 0.49 & 0.89 & 0.59 18 & 0.49 & 0.81 & 0.50 & 0.90 & 0.59 19 & 0.49 & 0.82 & 0.52 & 0.91 & 0.59 20 & 0.51 & 0.83 & 0.54 & 0.92 & 0.62 25 & 0.58 & 0.90 & 0.60 & 1.0 & 0.66 30 & 0.58 & 0.93 & 0.62 & 1.1 & 0.70 35 & 0.61 & 0.94 & 0.64 & 1.1 & 0.70 40 & 0.63 & 0.94 & 0.65 & 1.1 & 0.69 45 & 0.63 & 0.92 & 0.65 & 1.1 & 0.69 50 & 0.61 & 0.91 & 0.63 & 1.1 & 0.65 60 & & 0.86 & 0.62 & 1.0 & 0.60 70 & 0.56 & 0.80 & 0.60 & 0.90 & 0.61 80 & & 0.76 & 0.58 & 0.86 & 0.56 90 & & 0.69 & 0.56 & 0.78 & 0.46 100 & & 0.68 & 0.56 & 0.74 & 0.44 0 & 0.25 & 0.47 & 0.21 & 0.50 & 0.29 1.0 & 0.25 & 0.46 & 0.21 & 0.50 & 0.29 2.0 & 0.25 & 0.48 & 0.22 & 0.51 & 0.30 3.0 & 0.29 & 0.52 & 0.24 & 0.53 & 0.32 4.0 & 0.34 & 0.55 & 0.25 & 0.57 & 0.35 5.0 & 0.44 & 0.59 & 0.32 & 0.61 & 0.38 6.0 & 0.54 & 0.64 & 0.37 & 0.64 & 0.42 7.0 & 0.60 & 0.68 & 0.43 & 0.68 & 0.45 8.0 & 0.64 & 0.70 & 0.44 & 0.70 & 0.45 9.0 & 0.64 & 0.71 & 0.44 & 0.70 & 0.47 10 & 0.64 & 0.73 & 0.45 & 0.72 & 0.48 15 & 0.65 & 0.76 & 0.47 & 0.76 & 0.54 20 & 0.63 & 0.77 & 0.48 & 0.78 & 0.53 30 & 0.63 & 0.79 & 0.52 & 0.80 & 0.56 40 & 0.62 & 0.82 & 0.54 & 0.81 & 0.58 50 & 0.62 & 0.85 & 0.58 & 0.83 & 0.59 60 & 0.62 & 0.90 & 0.59 & 0.83 & 0.59 70 & 0.62 & 0.90 & 0.59 & 0.84 & 0.58 80 & 0.62 & 0.91 & 0.59 & 0.82 & 0.59 90 & 0.61 & 0.90 & 0.60 & 0.82 & 0.57 100 & 0.62 & 0.89 & 0.61 & 0.83 & 0.57 200 & 0.62 & 0.75 & 0.57 & 0.82 & 0.53 300 & 0.61 & 0.71 & 0.56 & 0.77 & 0.53 400 & 0.60 & 0.69 & 0.54 & 0.71 & 0.47 500 & 0.60 & 0.67 & 0.55 & 0.71 & 0.49 600 & 0.61 & 0.67 & 0.56 & 0.74 & 0.48 700 & 0.62 & 0.67 & 0.59 & 0.72 & 0.50 800 & 0.63 & 0.70 & 0.60 & 0.72 & 0.49 900 & 0.63 & 0.70 & 0.60 & 0.72 & 0.50 1000&0.63 & 0.70 & 0.60 & 0.72 & 0.50 0 & 2.2 & 2.5 & 2.8 & 2.2 & 1.8 & 1.4 1 & 2.2 & 2.4 & 2.8 & 2.1 & 1.7 & 1.4 2 & 2.1 & 2.3 & 2.5 & 2.0 & 1.6 & 1.35 3 & 2.2 & 2.3 & 2.5 & 2.0 & 1.6 & 1.37 4 & 4.2 & 3.8 & 4.3 & 3.9 & 3.7 & 3.0 5 & 3.5 & 3.4 & 3.9 & 3.1 & 2.9 & 2.5 6 & 3.3 & 3.3 & 3.9 & 3.0 & 2.7 & 2.4 7 & 3.4 & 3.3 & 4.2 & 3.1 & 2.7 & 2.5 8 & 3.6 & 3.5 & 4.4 & 3.3 & 2.8 & 2.7 9 & 3.7 & 3.7 & 4.5 & 3.4 & 2.9 & 2.7 10 & 3.8 & 3.6 & 4.6 & 3.4 & 3.0 & 2.8 15 & 4.6 & 4.5 & 5.5 & 4.2 & 3.8 & 3.4 20 & 5.2 & 4.8 & 6.2 & 4.8 & 4.3 & 3.7 25 & 5.6 & 5.1 & 6.5 & 5.2 & 5.1 & 4.0 30 & 5.8 & 5.6 & 6.7 & 5.3 & 4.8 & 4.1 40 & 6.4 & 6.2 & 7.3 & 5.9 & 5.4 & 4.4 50 & 6.9 & 6.8 & 8 & 6.3 & 5.8 & 4.8 60 & 7.3 & 7.3 & 8.4 & 6.7 & 6.3 & 5.0 70 & 7.5 & 7.5 & 8.7 & 7.0 & 6.5 & 5.2 80 & 7.8 & 7.9 & 9.2 & 7.4 & 6.9 & 5.5 90 & 8.0 & 8.1 & 9.5 & 7.6 & 7.1 & 5.7 100 & 8.3 & 8.5 & 9.9 & 7.8 & 7.3 & 5.8 200 & 9.5 & 10.0 & 11.8 & 9.1 & 8.5 & 6.6 300 & 9.6 & 10.0 & 12.3 & 9.2 & 8.7 & 6.5 400 & 9.7 & 10.2 & 12.6 & 9.3 & 9.0 & 6.4 500 & 9.7 & 10.4 & 12.9 & 9.6 & 9.2 & 6.7 600 & 9.9 & 10.5 & 13.3 & 9.7 & 9.5 & 6.8 700 & 10.2 & 11.0 & 13.7 & 10.1 & 9.8 & 7.0 800 & 10.4 & 11.3 & 14.2 & 10.2 & 10.0 & 7.1 900 & 10.6 & 11.5 & 14.5 & 10.5 & 10.3 & 7.4 1000 & 10.8 & 11.8 & 15.1 & 10.8 & 10.5 & 7.3 0 & 0.43 & 0.18 & 0.47 & 0.25 1 & 0.42 & 0.17 & 0.46 & 0.25 2 & 0.38 & 0.16 & 0.43 & 0.22 3 & 0.43 & 0.17 & 0.44 & 0.24 4 & 0.36 & 0.35 & 0.38 & 0.19 5 & 0.60 & 0.36 & 0.64 & 0.40 6 . & 0.68 & 0.38 & 0.70 & 0.40 7 & 0.79 & 0.41 & 0.74 & 0.49 8 & 0.81 & 0.43 & 0.80 & 0.54 9 & 0.82 & 0.45 & 0.84 & 0.55 15 & 0.83 & 0.50 & 0.88 & 0.57 20 & 0.88 & 0.54 & 1.0 & 0.62 25 & 0.93 & 0.57 & 1.0 & 0.63 30 & 1.0 & 0.62 & 1.0 & 0.68 40 & 1.0 & 0.68 & 1.0 & 0.70 50 & 0.97 & 0.70 & 1.0 & 0.67 60 & 0.93 & 0.67 & 1.0 & 0.64 70 & 0.91 & 0.69 & 1.0 & 0.62 80 & 0.89 & 0.70 & 0.95 & 0.61 90 & 0.86 & 0.70 & 0.93 & 0.57 100 & 0.85 & 0.67 & 0.90 & 0.57 200 & 0.75 & 0.70 & 0.85 & 0.48 300 & 0.72 & 0.75 & 0.80 & 0.38 400 & 0.65 & 0.72 & 0.74 & 0.39 500 & 0.60 & 0.70 & 0.63 & 0.37 600 & 0.61 & 0.69 & 0.64 & 0.45 700 & 0.59 & 0.73 & 0.60 & 0.47 800 & 0.57 & 0.70 & 0.60 & 0.47 900 & 0.59 & 0.73 & 0.58 & 0.51 1000 & 0.59 & 0.81 & 0.57 & 0.52 0 & 52.7 & 40.4 & 0.048 & 400 & 134 & 64 & 22 & 5.3 & 14.7 & 5.3 & 2.6 1 & 52.7 & 40.4 & 0.046 & 367 & 124 & 58 & 20 & 4.8 & 13.1 & 4.6 & 2.3 2 & 52.7 & 40.4 & 0.037 & 308 & 102 & 48 & 16 & 3.2 & 8.8 & 3.1 & 1.5 3 & 52.5 & 40.2 & 0.028 & 120 & 39 & 18 & 6 & 0.9 & 2.5 & 0.9 & 0.4 4 & 52.2 & 39.9 & 0.046 & 68 & 23 & 11 & 3.6 & 0.9 & 2.4 & 0.8 & 0.4 5 & 51.9 & 39.6 & 0.050 & 40 & 16 & 8 & 2.8 & 0.5 & 1.7 & 0.7 & 0.35 6 & 51.4 & 39.1 & 0.002 & 13 & 5 & 2.5 & 0.9 & 0.006 & 0.02 & 0.007 & 0.0047 & 51.0 & 38.7 & 0 & 8 & 3.7 & 1.8 & 0.7 & 0 & 0 & 0 & 0 8 & 50.7 & 38.4 & 0 & 5 & 2.0 & 1.0 & 0.4 & 0 & 0 & 0 & 0 9 & 50.5 & 38.1 & 0 & 3.6 & 1.3 & 0.6 & 0.2 & 0 & 0 & 0 & 0 10 & 50.2 & 37.9 & 0 & 2.3 & 0.9 & 0.4 & 0.15 & 0 & 0 & 0 & 0 15 & 49.4 & 37.1 & 0 & 0.3 & 0.1 & 0.06 & 0.02 & 0 & 0 & 0 & 0 20 & 48.9 & 36.6 & 0 & 0.1 & 0.05 & 0.02 & 0.008 & 0 & 0 & 0 & 0 1 & 52.7 & 40.4 & 0.047 & 379 & 127 & 60 & 21 & 4.9 & 14 & 4.9 & 2.4 2 & 53.0 & 40.7 & 0.046 & 356 & 120 & 56 & 19 & 4.5 & 13 & 4.5 & 2.2 3 & 53.2 & 40.8 & 0.044 & 295 & 98 & 46 & 16 & 3.5 & 9.9 & 3.5 & 1.7 4 & 53.2 & 40.9 & 0.045 & 227 & 75 & 35 & 12 & 2.8 & 7.8 & 2.7 & 1.3 5 & 53.3 & 40.9 & 0.045 & 182 & 61 & 29 & 10 & 2.2 & 6.3 & 2.3 & 1.1 6 & 53.3 & 41.0 & 0.045 & 155 & 54 & 26 & 9 & 1.9 & 5.5 & 2.0 & 1.00 7 & 53.3 & 41.0 & 0.045 & 135 & 48 & 23 & 8 & 1.7 & 4.9 & 1.8 & 0.90 8 & 53.3 & 41.0 & 0.045 & 125 & 45 & 21 & 7.5 & 1.5 & 4.6 & 1.7 & 0.85 9 & 53.3 & 41.0 & 0.045 & 119 & 43 & 20 & 7.1 & 1.5 & 4.4 & 1.6 & 0.80 10 & 53.3 & 41.0 & 0.045 & 114 & 41 & 20 & 6.9 & 1.4 & 4.2 & 1.5 & 0.78 15 & 53.3 & 41.0 & 0.045 & 96 & 35 & 17 & 5.9 & 1.2 & 3.6 & 1.3 & 0.66 20 & 53.3 & 41.0 & 0.045 & 83 & 30 & 14 & 5.0 & 1.0 & 3.1 & 1.1 & 0.57 30 & 53.3 & 41.0 & 0.045 & 70 & 25 & 12 & 4.2 & 0.86 & 2.6 & 0.94 & 0.48 40 & 53.3 & 41.0 & 0.045 & 62 & 23 & 11 & 3.8 & 0.77 & 2.3 & 0.84 & 0.43 50 & 53.3 & 41.0 & 0.045 & 57 & 21 & 10 & 3.5 & 0.71 & 2.1 & 0.77 & 0.40 60 & 53.3 & 41.0 & 0.045 & 54 & 19 & 9 & 3.3 & 0.66 & 2.0 & 0.73 & 0.38 70 & 53.3 & 41.0 & 0.045 & 51 & 18 & 8.7 & 3.2 & 0.63 & 1.9 & 0.69 & 0.36 80 & 53.3 & 41.0 & 0.045 & 49 & 18 & 8.4 & 3.0 & 0.60 & 1.8 & 0.66 & 0.35 90 & 53.3 & 41.0 & 0.045 & 47 & 17 & 8.0 & 2.9 & 0.58 & 1.7 & 0.63 & 0.33 100 & 53.3 & 41.0 & 0.045 & 45 & 16 & 7.8 & 2.8 & 0.56 & 1.6 & 0.61 & 0.32	we present evolutionary stellar population synthesis models to predict the spectrum of a single - metallicity stellar population , with a spectral sampling of 0.3 in five spectral regions between 3700 and 5000 . the models , which are optimized for galaxies with active star formation , synthesize the profiles of the hydrogen balmer series ( h@xmath0 , h@xmath1 , h@xmath2 , h8 , h9 , h10 , h11 , h12 and h13 ) and the neutral helium absorption lines ( hei @xmath34922 , hei @xmath34471 , hei @xmath34388 , hei @xmath34144 , hei @xmath34121 , hei @xmath34026 , hei @xmath34009 and hei @xmath33819 ) for a burst with an age ranging from 10@xmath4 to 10@xmath5 yr , and different assumptions about the stellar initial mass function ( imf ) . continuous star formation models lasting for 1 gyr are also presented . the input stellar library includes nlte absorption profiles for stars hotter than 25000 k and lte profiles for lower temperatures . the temperature and gravity coverage is 4000 k @xmath6t@xmath750000 k and 0.0@xmath6log [email protected] , respectively . the metallicity is solar . it is found that the balmer and hei line profiles are sensitive to the age , except during the first 4 myr of the evolution , when the equivalent widths of these lines are constant . at these early stages of the evolution , the profiles of the lines are also sensitive to the imf . however , strong h balmer and hei lines are predicted even when the low mass cut - off of the imf is as high as 10 m@xmath8 . the equivalent widths of the balmer lines range from 2 to 16 and the hei lines from 0.2 to 1.2 . during the nebular phase ( cluster younger than about 10 myr ) , h@xmath0 ranges from 2 to 5 and hei @xmath34471 between 0.5 and 1.2 . the strength of the lines is maximum when the cluster is a few hundred ( for the balmer lines ) and a few ten ( for the hei lines ) myr old . in the continuous star formation scenario , the strength of the balmer and hei lines increases monotonically with time until 500 myr and 100 myr , respectively . however , the lines are weaker than in the burst models due to the dilution of the balmer and hei lines by the contribution from very massive stars . the high spectral resolution of the profiles is useful to reproduce the absorption wings observed in regions of recent star formation and to estimate the effect of the underlying absorption on the nebular emission - lines . the strength of the nebular emission balmer and hei lines compared with the stellar absorption components indicates that h@xmath2 and the higher - order terms of the balmer series and hei are dominated by the stellar absorption component if an instantaneous burst is older than @xmath9 5 myr . some of the hei lines ( e.g. hei @xmath33819 , hei @xmath34922 ) are more favorable than others ( e.g. hei @xmath34471 ) for the detection of stellar features in the presence of nebular emission . we estimate that the correction to the hei @xmath34471 nebular emission line due to the stellar absorption is between 5 and 25@xmath10 , if the nebular emission has equivalent width between 10 and 2.5 ( corresponding to a burst age between 1 - 3 myr ) . the models can be used to date starburst and post - starburst galaxies until 1 gyr . they have been tested on data for clusters in the lmc , the super - star cluster b in the starburst galaxy ngc 1569 , the nucleus of the dwarf elliptical ngc 205 and a luminous `` e+a '' galaxy . the full data set is available for retrieval at http://www.iaa.es/ae/e2.html and at http://www.stsci.edu/science/starburst/ , or on request from the authors at [email protected]
ultra - luminous x - ray sources ( ulxs ) are very bright extragalactic x - ray point sources , with observed fluxes which would correspond to luminosities greater than @xmath0 ergs per second if they were radiating isotropically . since this is above the eddington limit for normal stellar - mass compact objects , it has been widely thought that they are associated with intermediate - mass black holes ( imbhs ) @xcite , although non - spherical accretion and beamed emission could give rise to inferred luminosities significantly above the eddington limit ( see , for example * ? ? ? * ) , allowing also for lower masses . @xcite have reported nustar observations of ulx m82 x-2 ( also known as nustar j095551 + 6940.8 ) which reveal periodic changes in the hard x - ray luminosity of this source , indicative of a rotating magnetized neutron star being involved rather than a black hole . the measured peak flux ( in the @xmath1 band ) would correspond to @xmath2 if the radiation were isotropic , and is challenging to explain with a neutron star . the period ( taken to be the neutron - star spin period ) was found to be @xmath3 , with a 2.53-day sinusoidal modulation , interpreted as being an orbital period @xmath4 corresponding to motion around an unseen companion which would be the mass donor in the accreting system . the time derivative of the spin period @xmath5 was also measured . values for this coming from different individual observations show considerable variations but a relevant underlying spin - up tendency was found , with @xmath6 . the mass donor is indicated as having a mass larger than @xmath7 , so that the system should be categorized as a high mass x - ray binary ( hmxb ) . taking canonical neutron star parameters as a rough guide ( @xmath8 and @xmath9 ) , the luminosity relation @xmath10 , gives the mass accretion rate corresponding to @xmath11 as being @xmath12 . there are three main mechanisms by which the mass transfer might occur : ( i ) via a spherical wind ( as for o - type hmxbs ) , ( ii ) via a disc - shaped wind ( as for be - type hmxbs ) , or ( iii ) by roche lobe overflow ( rlof ) . because of the large inferred @xmath13 , the third option was suggested as the mechanism by @xcite and subsequent studies @xcite . here , we investigate each of these scenarios in turn to see which may be appropriate for ulx m82 x-2 . in section 2 , we discuss the strength required for the neutron - star magnetic field , and show that it needs to be moderately strong but not at a magnetar level . in section 3 , we discuss the applicability of scenarios ( i)-(iii ) , finding that ( i ) is excluded but that ( ii ) and ( iii ) could be viable possibilities . in section 4 , we discuss the role of the propeller effect and transient behaviour , and section 5 contains conclusions . in the standard picture for hmxbs , the system has to be fairly young because the companion donor star is massive enough to have only a rather short main - sequence life - time . matter coming from the donor star falls towards its neutron star companion , becomes included in a keplerian accretion disc , and eventually becomes entrained by the neutron star s magnetic field , creating hot x - ray emitting accretion columns above the magnetic poles ( cf . sufficiently young neutron stars typically have magnetic - field strengths above @xmath14 ( see , for example , the data in the atnf pulsar catalogue , * ? ? ? * ) , with a tail of the distribution extending beyond @xmath15 and eventually joining with the magnetar regime at @xmath16 . this can be relevant for explaining how this source can be so luminous , because if the magnetic field is stronger than the quantum limit , @xmath17 , the scattering cross - section would be suppressed , reducing the opacity of matter in the accretion columns above the magnetic poles and allowing higher luminosities . with this in mind , @xcite suggested that this source might contain a magnetar . the computational results of @xcite and the evidence of propeller effect from @xcite support this idea . in other works , however , @xcite , @xcite and @xcite have explored different scenarios with standard pulsar fields @xmath18 , while other authors have advocated weaker fields , @xmath19 @xcite . as the present work was being completed , we have seen a new paper by @xcite , advocating a model with strong beaming and a magnetic field of @xmath20 . the work presented here represents a line of study parallel to theirs . we focus here on a scenario with a field at the top end of the range for standard pulsars . in the rest of this section , we apply some simple assumptions for testing the relevance of a solution of this type . we take the full entrainment of the accreting matter by the magnetic field to occur close to the magnetic radius , @xmath21 , where the magnetic pressure balances the ram pressure of the infalling matter . using the condition of mass continuity , we then obtain the following expression for the magnetic radius : @xmath22 here @xmath23 is the field strength at the surface of the neutron star , and @xmath24 is the ratio of the accretion velocity to the free - fall velocity @xcite . at @xmath25 , the accreting matter is taken to come into corotation with the neutron star , with the corotation speed being @xmath26 . since the system is probably close to spin equilibrium @xcite , it is reasonable to take @xmath27 as being approximately equal to the keplerian velocity at @xmath25 . we can then estimate the appropriate value of the field strength in order to be consistent with the observed spin period for ulx m82 x-2 . this gives @xmath28 interestingly , this value is just above the critical quantum limit @xmath29 at which electron scattering is suppressed , as mentioned earlier : @xmath30 @xcite . on the other hand , we can also consider the spin - up rate , focusing on the measured underlying tendency @xmath31 @xcite , rather on the variations seen in the individual measurements , as mentioned earlier . we use the classical ghosh & lamb model @xcite , in order to identify appropriate parameter values ; the spin - up rate is then given by eq . ( 15 ) in their paper . assuming that the moment of inertia of the neutron star has the canonical value , @xmath32 , and using the spin - up rate given by @xcite , we find that the magnetic field strength corresponding to @xmath33 would be @xmath34 , which is very similar to the value obtained above , in eq . ( [ eq : bns ] ) , from considerations of the spin period . based on these two independent estimates , we conclude that having a field of @xmath35 can be consistent with the interpretation that m82 x-2 could be an example ( albeit a rather extreme one ) of previously - known classes of hmxbs ( see , for example , @xcite ) . a field of this strength can be just large enough to permit the quantum suppression of electron scattering as invoked by @xcite . in this section , we examine the different possibilities mentioned in the introduction for the accretion mode of ulx m82 x-2 . the two mechanisms operating for standard hmxbs are : ( i ) accretion from a spherical wind emitted by the donor star , and ( ii ) accretion from a disc - shaped wind , thought to come from a `` decretion disc '' around the donor star ( in a decretion disc , angular momentum is continually added to the inner edge of the disc from the rapidly rotating central star , and the matter drifts outwards rather than inwards ) . type ( i ) occurs for ob - type hmxbs , while type ( ii ) fuels be - type hmxbs . additionally , although there is no observational evidence for this in standard hmxb studies , it is possible that roche lobe overflow ( rlof ) could be relevant here ; we are referring to this as type ( iii ) . @xcite and @xcite both suggested rlof as a likely accretion mode for m82 x-2 , but did not discuss it in detail . in a recent paper , however , @xcite presented an argument suggesting that a certain proportion of ulxs should indeed be accreting neutron stars with rlof mass - transfer . here , we will consider each of the mechanisms in turn to assess which of them may be appropriate for our case . in order to give a luminosity as high as that observed , the accretion rate needs to be at least as high as @xmath36 , as shown earlier . first , we consider whether a spherical stellar wind coming from the massive donor could successfully fuel ulx m82 x-2 ( type ( i ) accretion ) . in simple stellar wind models , the wind velocity is frequently described by an expression of the form @xmath37 @xcite , where @xmath38 is the terminal velocity of the wind , @xmath39 is the stellar radius , @xmath40 is the distance from the centre of the donor star , and @xmath41 is a parameter whose value is close to unity ( we will take @xmath42 exactly here ) . the accretion rate from a spherical wind is usually described by the hoyle - lyttleton formula @xcite : @xmath43 where @xmath44 is the accretion radius defined by @xmath45 and @xmath46 is the velocity of the neutron star relative to the accreting matter : @xmath47 , where @xmath48 is the orbital velocity of the neutron star . the wind density can be written as @xmath49 where @xmath50 is the rate of mass loss from the donor star . if one specifies a particular value for @xmath50 , the accretion rate @xmath13 can be calculated from eq . ( [ eq : dotm ] ) , and then the corresponding x - ray luminosity of the neutron star can be obtained from @xmath51 . @xcite have given analytic formulae from which @xmath50 can be calculated for high - mass stars in terms of their mass @xmath52 , luminosity @xmath53 , effective temperature @xmath54 and the ratio @xmath55 , where @xmath38 is the wind s terminal velocity and @xmath56 is the escape velocity from the stellar surface . a change in behaviour occurs when @xmath54 passes a value @xmath57 and separate formulae are given for temperatures above and below that . here we use the mass loss rates given by equations ( 24 ) and ( 25 ) in their paper . the ratio @xmath55 is known to be @xmath58 for @xmath59 and @xmath60 for @xmath61 @xcite while values for @xmath53 and @xmath54 can be obtained as functions of @xmath52 using a stellar evolution code ( we used the analytic fitting formulae from the paper by @xcite ) . putting this data into eqs . ( [ eq : dotm ] ) ( [ eq : rhow ] ) , we could then calculate predicted values of the x - ray luminosity for a sequence of masses of the donor star , and our results are shown in fig . [ fig : wind ] . the two curves in fig . [ fig : wind ] correspond to different evolutionary phases ; the zero - age main sequence phase ( solid curve ) and the terminal main sequence phase ( dashed curve ) . for stars more massive than 30 @xmath62 , the stellar radius becomes larger than that of the binary orbit , @xmath63 ^{1/3 } , \label{eq : a}\ ] ] before it reaches the terminal main sequence ( here @xmath64 is the mass of the donor star ) . it is clear , from this figure , that even if we consider the most efficient conditions , and an optimal viewing angle , an x - ray luminosity of ulx level ( @xmath65 ) can not be achieved . hence , we conclude that ulx m87 x-2 can not be fed by a spherical wind because the density of the wind would not be high enough . the corbet diagram ( where @xmath4 is plotted against @xmath66 ) is frequently used in studies of hmxbs , and it can be useful to consider it also in the present context . ob - type and be - type hmxbs show clearly different distributions in these diagrams , with the majority of the be - type ones being close to the diagonal from lower - left to upper - right ( they are the points marked with a @xmath67 in fig . [ fig : be ] ) . @xcite obtained an expression for the sequence followed by the be - type hmxbs by assuming the hoyle - lyttleton accretion rate given by eq . ( [ eq : dotm ] ) . the decretion disc wind parameters @xmath68 and @xmath69 could be taken from any appropriate wind model ; @xcite adopted a simple disc - shaped model : @xmath70 where @xmath71 and @xmath72 are the density and velocity at the stellar radius @xmath73 , and @xmath74 is a constant which needs to be fixed . in order to explain the positions of known standard be - hmxbs in the corbet diagram , @xcite used @xmath75 , and we will also use this value in the following this choice is roughly consistent with recent numerical computations of be - disc winds @xcite . with this wind model and any specified value of the magnetic field strength , one can obtain a corresponding relationship between @xmath66 and @xmath4 by substituting equation ( [ eq : dotm ] ) into equation ( [ eq : bns ] ) ( @xmath4 is related to @xmath76 via equation ( [ eq : a ] ) ) . this gives @xmath77 where @xmath78 in fig . [ fig : be ] we show @xmath4-@xmath66 curves , obtained from eq . ( [ eq : pp ] ) , for three values of both @xmath79 and @xmath71 . the values used for @xmath72 and the radius of the neutron star are the same as those used by @xcite : @xmath80 and @xmath81 . in the two frames , we show the results for two different donor masses : @xmath82 and @xmath83 , with the corresponding radii being those for the terminal main sequence phase on the stellar evolution tracks of @xcite : @xmath84 for the @xmath83 donor and @xmath85 for the @xmath82 donor . it can be seen that the results in the two frames are very similar . in each of them , ulx m82 x-2 ( marked with the circles ) takes a rather extreme position in comparison with the be - hmxbs but , for an appropriate set of parameters , it does fit with the be - hmxb sequences . as shown in the figure , assuming a magnetic field strength of @xmath86 , the model given by eq . ( [ eq : pp ] ) with the density @xmath87 does fit the position of ulx m82 x-2 very well in both cases . of course , the position of ulx m82 x-2 on the figure could also be explained by models with lower @xmath79 and extremely low @xmath71 , or higher @xmath79 and extremely high @xmath71 . from the limited available polarization data , @xcite suggested that the @xmath71 for discs around be - stars is between @xmath88 and @xmath89 , based on the nine systems which they studied . @xcite tested values of @xmath71 between @xmath90 and @xmath91 for fitting the ir radiation from be discs , and suggested that the best fit value is @xmath92 . in view of these results , our best fit density ( @xmath93 ) seems slightly high but still a reasonable value . additionally , if we take this value for @xmath71 , eq . ( [ eq : dotm ] ) gives @xmath94 for the mass accretion rate which is again quite reasonable for ulx m82 x-2 . we note that a field as low as @xmath20 would be difficult to accommodate within this picture . with this large mass loss rate , however , the donor be - star can not maintain a high density disc for very long and so , after the bright ( high accretion rate ) phase , it would run out of decretion disc matter . hence , in this scenario , the accreting system would inevitably become a transient . since , according to the archival survey , ulx m82 x-2 does show transient tendencies , this prediction seems to be consistent with the observations @xcite . we note , however , that transient behaviour could also be understood in the context of propeller switching @xcite , as we will discuss later . plotted against the spin period of the neutron star @xmath66 . the upper and lower panels show results for donor masses of @xmath82 and @xmath83 , respectively . the @xmath4-@xmath66 relation given by equation ( [ eq : pp ] ) is shown by the continuous curves with the three sets of curves in each panel showing results for different values of @xmath79 ( measured in gauss ) , and the different curves of each set being for different disc densities ( in @xmath95 ) . the stellar radii are set to the terminal main sequence values obtained from the stellar evolution tracks of @xcite . the positions on the diagram of known standard be - type hmxbs are shown with crosses , while that for ulx m82 x-2 is shown with a circle . the data for the be - hmxbs is taken from @xcite and @xcite . , title="fig:",width=302 ] plotted against the spin period of the neutron star @xmath66 . the upper and lower panels show results for donor masses of @xmath82 and @xmath83 , respectively . the @xmath4-@xmath66 relation given by equation ( [ eq : pp ] ) is shown by the continuous curves with the three sets of curves in each panel showing results for different values of @xmath79 ( measured in gauss ) , and the different curves of each set being for different disc densities ( in @xmath95 ) . the stellar radii are set to the terminal main sequence values obtained from the stellar evolution tracks of @xcite . the positions on the diagram of known standard be - type hmxbs are shown with crosses , while that for ulx m82 x-2 is shown with a circle . the data for the be - hmxbs is taken from @xcite and @xcite . , title="fig:",width=302 ] @xcite suggested that the accretion mode for ulx m82 x-2 may involve roche lobe overflow ( rlof ) and we need to consider this scenario as well . an approximate value for the roche radius @xmath96 is given by @xmath97 @xcite , where @xmath98 is the mass ratio @xmath99 ( with @xmath64 being the mass of the donor star ) and @xmath76 is an orbital separation given by eq . ( [ eq : a ] ) . for having rlof accretion , the donor radius needs to be larger than the roche radius but smaller than the orbital radius . [ fig : radii ] shows evolutionary tracks for the radii of stars with selected masses ( again using data from @xcite ) with the dotted lines linking the points where each track crosses its corresponding roche radius ( lower curve ) and orbital radius ( upper curve ) . the shaded region in fig . [ fig : rl ] shows where rlof can occur and is plotted as a function of the mass of the donor star . the upper bound of this shaded region corresponds to the orbital radius , and the lower bound shows the roche radius . stellar radii at several evolutionary stages ( zero - age main sequence , the end of the main sequence and the beginning of the giant branch ) are also shown . if @xmath64 is larger than @xmath100 , the donor star already fills its roche lobe during the main sequence stage , leading to the type of mass transfer known as case a @xcite . this occurs on the thermal timescale @xmath101 for a massive star , this timescale is shorter than @xmath102 and the mass transfer rate reaches @xmath103 , which is sufficient for feeding ulx m87 x-2 . however , for the most massive stars ( @xmath104 ) , the main - sequence stellar radius would be larger than the orbital radius as well , so that a common envelope would form around the binary in a short time scale . in that case , the system would be luminous at infra - red frequencies rather than in x - rays @xcite ; the donor must therefore be less massive than this in order to give the right sort of rlof accretion . furthermore , the mass of the donor ought to be restricted by the fact that the companion is a neutron star : the progenitor of the neutron star would have been the primary when this binary system was born and , according to @xcite , it should have had an original mass of less than @xmath105 in order to give rise to a neutron star rather than a black hole . the present primary should be less massive than the original one ( even if its mass may have increased slightly during the first mass transfer stage ) , and hence the donor mass is constrained to be less than roughly @xmath106 . another issue which needs to be taken into account is that if the mass ratio @xmath107 is above a certain limit ( @xmath108 for a red giant , and @xmath109 for a high - mass main sequence star ) , the binary would be subject to the darwin instability , meaning that it could not sustain a circular orbit as required in order for rlof to be a viable candidate mechanism here @xcite . for @xmath64 between @xmath110 , the star will be inside its roche lobe during its main sequence lifetime but may overflow it during its expansion at the end of the main sequence ( hertzsprung gap phase ) , giving an early case b type of mass transfer . this also proceeds on the thermal timescale , as given above , and could in principle be sufficient to feed ulx m87 x-2 . however , in general the timescale on which the star crosses the hertzsprung gap is quite short . for instance , an @xmath111 star evolves from the end of the main sequence to the giant branch in 0.35 myr @xcite and would have a radius between @xmath96 and @xmath76 for only some fraction of that ( determining which would require a detailed stellar evolution calculation ) . the relevant time is then very short compared with the overall life - time of the star ( @xmath112 for @xmath113 ) . however , the duration of the hertzsprung gap phase depends sensitively on the mass of the donor star and for the minimum donor mass envisaged , @xmath7 @xcite , it becomes @xmath114 . while it would still require a lucky chance to see any ulx fed by this case b type of rlof accretion , it would become less unlikely with the least massive possible donor stars . in connection with this : according to the population synthesis simulation of @xcite , the number of ulxs containing a neutron star should peak for systems having donor stars in a small mass range around @xmath115 , and so it is possible that the short ulx lifetime of these systems could be compensated by there being so many of them , and that they might actually predominate in the observations . to summarise : accretion via rlof might possibly be the mechanism for ulx m82 x-2 if the donor is an evolved star of @xmath116 passing through the hertzsprung gap . an rlof scenario with a main sequence donor of @xmath117 could also be a possibility although , for the higher masses , such a system would be prone to the orbit becoming eccentric due to the darwin instability . an rlof scenario with the donor being evolved up to the giant phase is ruled out completely . in the previous sections , we have discussed similarities between ulx m82 x-2 and known standard hmxbs , in terms of the strength of the neutron star s magnetic field and the accretion mode . in section 2 , we concluded that a fairly strong magnetic field , of around @xmath118 is favoured for this object . while our considerations have been quite simple ones , this conclusion is consistent with more detailed analyses @xcite . in section 3 , we considered the possible feeding mechanisms , concluding that accretion modes similar to those of standard be - hmxbs are possible here , and that some mechanisms involving roche lobe overflow would also be possible . in the present section , we discuss two further related issues . the 1.37 second spin period of the neutron star in ulx m82 x-2 is shorter than those of most hmxbs ( which are typically in the range of @xmath119 ) . in view of this and the rather high magnetic field which we are inferring , one needs to consider whether it would be able to continue accreting matter from its disc , and continue to spin up , or whether it would enter the propeller regime where the disc matter is instead forced away @xcite . the critical period for onset of the propeller regime is given by @xmath120 where @xmath121 is the magnetic moment measured in units of @xmath122 , as before , and @xmath123 is the x - ray luminosity in units of @xmath124 @xcite . [ fig : prop ] shows the spin period - luminosity plane , with the straight lines marking the relation given by eq . ( [ eq : ps ] ) for various values of @xmath79 . also shown are the locations on the plot of known hmxbs ( marked with crosses ) and of ulx m82 x-2 ( marked with the circle ) . for the latter , we are taking the values @xmath125 and @xmath126 . below the line for the relevant value of @xmath79 , the propeller regime is operative ; above it , accretion can proceed . according to this , ulx m82 x-2 would avoid the propeller regime if it has @xmath127 and so field strengths between the quantum limit of @xmath128 and the maximum of @xmath129 could give accretion and spin - up , but magnetar - strength fields of above @xmath130 could not do so . the allowed range for @xmath79 to exceed the quantum critical limit but avoid the propeller effect , is small but non - vanishing . ( dashed ) , @xmath128 ( thick solid ) and @xmath129 ( dotted ) are shown . to avoid the propeller effect operating , the location of ulx m82 x-2 needs to be above the line corresponding the surface magnetic field strength of its neutron star , i.e. @xmath79 needs to be smaller than @xmath129 for the parameter values given in the text , which is nevertheless significantly greater than the quantum limit @xmath131 . , width=302 ] it has been reported that archival data for m82 x-1/x-2 from xmm newton shows only tiny fluctuations at a level below 2.2 % @xcite . this could be understood in two different ways . firstly , the pulsations seen by @xcite could be a transient behaviour so that while xmm newton did not see the pulsations during the time - span of its observations , nustar did see them . the second possibility is that the pulsations might be observable only in the high energy range above 10 kev , so that nustar could detect them while xmm newton could not . @xcite , noting the variations seen by @xcite in different measurements of @xmath5 , have suggested that the system may be marginal for exhibiting the propeller effect , switching it on and off , and that the strong x - ray emissions and pulsations may only be observed when the propeller mechanism is not operating whereas during the propeller phase , only radiation from the hot disc is seen . in their latest and more extended analysis @xcite , they conclude that the magnetically threaded disc model with @xmath132 could explain the observed properties of this object very well . @xcite also argue that the observed bimodal distribution of x - ray luminosity can be interpreted as being due to the propeller effect switching on and off . in their picture , even when the system is in the propeller regime , a small fraction of matter still leaks through the magnetic field lines and continues to give rise to lower luminosity emission . another point is that hmxbs can often show different pulsation behaviour in different energy ranges , with the higher energy x - rays being emitted as a _ pencil beam _ , while lower energy ones are emitted as a _ fan beam _ @xcite . in this case , the emission directions would be different for the different energy ranges . in fact , more than 20% of hmxbs show different light - curve shapes in high and low energy bands @xcite . at present , the discussion on the transient behavior of m82 x-2 can not be concluded at all . more data is needed for deciding between these different possibilities . also one needs to consider the effect of irradiation of the accreting matter by the strong x - ray emission coming from the neutron star . the direct effect of the radiation pressure coming from this is , of course , to act as another obstacle for the accretion . in order to achieve a high accretion rate in luminous x - ray systems , some means is required for avoiding this obstacle , such as anisotropy of the x - ray radiation ( see , for example , * ? ? ? * ; * ? ? ? * ) . in the wind accretion case , however , there are effects of the irradiation which act in the opposite direction , enhancing the accretion rate onto the neutron star . in detached wind - fed binaries , intra - binary matter can not avoid some photo dissociation under these circumstances , and this reduces the efficiency of the acceleration mechanism driving the line - accelerated stellar wind away from the donor star @xcite . from equation ( [ eq : racc ] ) one can see that the resulting slower wind velocity would cause the accretion radius to be larger , and lead to _ increasing _ the accretion rate onto the neutron star , as given by equation ( [ eq : dotm ] ) . however , the wind velocity is limited by @xmath133 , where @xmath56 is the escape velocity at the donor surface , and so it can not be slow enough to significantly affect our discussion here . another potential positive effect of x - ray irradiation is that it could in principle cause the roche lobe to become swamped with heated outer - envelope matter which could then overflow the roche lobe even if the stellar radius was smaller than roche radius @xcite . this would require very strong irradiation though ; it could only work in tight systems with orbital periods of a few hours , and would not play any significant role for m82 x-2 . in this paper , we have discussed the magnetic field strength and accretion mode for the recently identified neutron star in the source ulx m82 x-2 . we have considered the conditions required for producing the observed values of the key parameters : @xmath134 , @xmath135 , @xmath136 and @xmath137 , and have argued that a consistent explanation can be given involving a moderately strong magnetic field at around the quantum limit @xmath138 . having a field strength above the quantum limit is favourable for explaining the high observed luminosity , because of the reduced opacity in this case . however , we note that there have been a number of other suggestions regarding the magnetic field strength @xcite ; further observations and discussions are required for resolving the issue . we then went on to examine whether the standard accretion modes for hmxbs can be appropriate for this object . we concluded that spherical wind accretion , which drives ob - type hmxbs , can not be the mechanism here but that an extension of the standard mechanism for be - type hmxbs can provide a natural explanation . we have shown that if the neutron star has a moderately strong magnetic field ( @xmath139 ) and there is wind accretion from a reasonably high - density ( @xmath140 ) decretion disc around the be - companion , then all of the main properties of ulx m82 x-2 , including its position on the corbet diagram , would be consistent with the relations followed by standard be - hmxbs . roche lobe overflow accretion is also a possibility . if the donor star has sufficiently high mass , it should still be on the main sequence and case a rlof accretion could then feed the ulx ; if it has lower mass , it would need to be undergoing case b rlof while passing through the hertzsprung gap during its post - main - sequence expansion . we gratefully acknowledge helpful discussions with odele straub and hitoshi yamaoka during the course of this work .	periodic pulsations have been found in emission from the ultra - luminous x - ray source ( ulx ) m82 x-2 , strongly suggesting that the emitter is a rotating neutron star rather than a black hole . however , the radiation mechanisms and accretion mode involved have not yet been clearly established . in this paper , we examine the applicability to this object of standard accretion modes for high mass x - ray binaries ( hmxbs ) . we find that spherical wind accretion , which drives ob - type hmxbs , can not apply here but that there is a natural explanation in terms of an extension of the picture for standard be - type hmxbs . we show that a neutron star with a moderately strong magnetic field , accreting from a disc - shaped wind emitted by a be - companion , could be compatible with the observed relation between spin and orbital period . a roche lobe overflow picture is also possible under certain conditions . accretion , accretion disk stars : neutron x - rays : binaries x - rays : individuals : m82 x-2
the fluorescence iron k@xmath2 line revealed by a 4.2-day asca observation in the seyfert 1 galaxy mcg 6 - 30 - 15 has distinct features which are unique characteristics of the emission line from a relativistic disk ( tanaka et al . the line is peaked at around 6.4 kev , with a broad red wing extending down to 4 kev . the skewness of the profile towards red and the sharp blue edge near the rest frame energy are a consequence of the combination of relativistic doppler effect and strong gravitational redshift , and are unique to a low - inclination relativistic accretion disk ; other mechanisms of the line formation are less successful to produce these features ( fabian et al . however , the alternative models such as thermal comptonization ( misra and kembhavi 1998 ) can not be ruled out completely . both the disk line fit ( tanaka et al . 1995 ) and the frequency extrema analysis ( bromley , miller & pariev 1998 ) suggested that the inner edge of line formation region is within a few schwarzschild radii , providing ever strongest evidence for a supper - massive black hole in this object . similar broad line profiles with lower statistics have also been seen in other agns ( mushotzky et al . 1995 , tanaka et al . 1995 , yaqoob et al . 1995 , nandra et al . 1997 , turner et al . 1998 and references therein ) . the line profile is strongly dependent on the inclination of the relativistic accretion disk ( fabian et al . 1989 ) , thus providing a method to estimate the inclination of inner accretion disk from line profile fitting . the inclinations from this method for a sample of seyfert 1 galaxies are strongly constrained to be nearly face - on ( nandra et al . 1997 ) , consistent with the expectation of the unification model of the two types of seyfert galaxies ( see antonucci 1993 for a review ) . however , there are complications to this simple picture . turner et al . ( 1998 ) found that type 2 seyfert galaxies also possess similar iron k@xmath2 line profiles , indicative of face - on accretion disk as well . this might be due to the strong contribution of a narrow component in these objects ( weaver & reynolds 1998 ) or due to complex absorption ( wang et al . moreover , the rapid variability of the line equivalent width ( ew ) and profile observed in mcg 6 - 30 - 15 ( iwasawa et al . 1996 ) and ngc 4051 ( wang et al . 1999b ) can not be readily explained by any current simple accretion disk line model . ngc 4151 is a bright nearby ( z = 0.0033 ) seyfert 1.5 galaxy . the edge - on orientation of its nucleus is strongly supported by the biconical geometry of the [ o iii ] @xmath55007 region ( evans et al . 1993 , pedlar et al . 1993 ) and the best estimated inclination is 65@xmath3 . yaqoob et al . ( 1995 ) presented the first measurement of its broad fe k line profile from the asca observations performed in 1993 may , november and december . the apparent line profile is asymmetric , consisting of a peak at @xmath06.3 6.4 kev and a broad red wing extending to @xmath04 5 kev . when a disk line model is fitted , the inclination angle of the disk ( @xmath6 = 0@xmath7 degrees ) is strongly constrained to be face - on , in contrary to the edge - on geometry of this source . this problem can be eased , but not solved , by assuming an additional narrow - line contribution , presumably from the torus , to the fe k line ( @xmath6 = 25@xmath8 degrees , see yaqoob & weaver 1997 ) . a much longer asca observation ( 3 days ) of ngc 4151 was carried out in 1995 , which provides data with better statistics ( leightly et al . 1997 ) . in this letter , we report measurement of the complex broad iron k line profile and its implication . ngc4151 was observed by asca from 1995 may 10 to 12 with the solid - state imaging spectrometer ( sis ) in 1ccd mode and the gas imaging spectrometer ( gis ) in ph mode . the gis data are contaminated by a nearby bl lac object due to their worse spatial resolution . in this letter , we will concentrate on the sis data , also because they offer much better energy resolution ( inoue 1993 ) , which is crucial to the fe k line profile analysis . the data were reduced with the asca standard software xselect and cleaned using the following screening criteria : satellite not passing through the south atlantic anomaly ( saa ) , geomagnetic cutoff rigidity greater than 6 gevc@xmath9 , and minimum elevation angle above earth s limb of 10@xmath3 and 20@xmath3 for nighttime and daytime observations , respectively . source counts were extracted from a circular area of radius 3.4 for the sis0 and sis1 , and the background counts were estimated from the blank sky data . spectra extracted from sis0 and sis1 were combined together and grouped to have @xmath1020 counts in each bin to allow the use of @xmath11 statistics . spectral analysis was carried out using xspec . the asca observation lasted for about three days , and the sis0 detector received a net exposure time of 93,000s and an average count rate of 1.49 cts / s in the 0.4 10.0 kev band , while these two parameters for the well - known 4.2-day asca observation on mcg 6 - 30 - 15 in 1994 july are 150,000s and 1.82 cts / s , respectively . although shorter exposure time and lower count rate may probably lead to worse statistics , we noticed that , because of its much harder spectrum ( due to strong absorption ) , the actual total counts ( 32,000 ) for ngc 4151 in the 5.0 7.0 kev band are almost three times of those ( 11,000 ) for mcg 6 - 30 - 15 in the same energy band . thus , the ngc 4151 fe k spectrum used in this paper has better statistics than the average mcg 6 - 30 - 15 spectrum and a best ever quality fe k line profile . following weaver et al . ( 1994 ) , we fit the underlying continuum in the 1.0 4.0 and 8.0 10.0 kev bands ( to exclude the possible broad iron line region ) with a model which consists of a dual absorbed power law with some fraction ( a best - fit value for this letter is @xmath05% ) of the direct continuum scattered into our line of sight and absorbed only by the galactic column of 2 @xmath12 10@xmath13 @xmath14 . we do not include a compton reflection component in the spectral fits because a clear reflection component has never been detected ( maisack & yaqoob 1991 ; yaqoob et al . 1993 ) . zdziarski et al . ( 1996 ) claimed that they detected a reflection component . however , considering the poor statistical significance of their detection ( they used only a delta - chi square of 2.7 for the their errors yet the model was complicated ) , the complex intrinsic continuum which was unknown , and the cross - calibration of two different satellites ( ginga and gro / osse ) which is not perfect , we think that their result is questionable . our model can describe the data reasonably well ( @xmath11 = 364 for 325 degrees of freedom ) , and the best - fit model parameters converge to @xmath15 = 1.30@xmath16 , n@xmath17(low ) @xmath18 3.2 @xmath12 10@xmath19 @xmath14 ( covering @xmath0 40% of the source ) , and n@xmath17(high ) = @xmath0 10.9 @xmath12 10@xmath19 @xmath14 ( covering @xmath0 60% of the source ) . the index is slightly flatter than that seen previously for ngc 4151 as derived from data weighted toward higher energies ( yaqoob et al . 1993 ) . we have also tried a model consisting of a power - law absorbed by ionized material ( zdziarski et al . 1995 ) plus a fraction of scattered underlying continuum , but failed to get a satisfactory fit ( @xmath15 = 1.00 , @xmath11 = 393 for 326 degrees of freedom ) . the profile of the iron k line is shown in figure 1a , which is similar to , and much better defined than , the one presented by yaqoob et al . the line shows a strong narrow peak around 6.4 kev and a huge red wing , containing a second slightly weak peak at around 5.2 kev , extending to @xmath0 4.5 kev . in this respect , the line profile of ngc 4151 is remarkably similar to that seen in mcg 6 - 30 - 15 ( see figure 1b for comparison ) . we also find a weak blue wing extending to about 8 kev , which is just visible in figure 2c of yaqoob et al . ( 1995 ) , is now clearly seen , thanks to the greatly improved statistics . the brightness of this source rules out the possibility that the blue wing is caused by improper background subtraction ( only 1@xmath20 of the count rate at 7 . kev is due to the background ) . first , we fit the line with a disk - line model ( fabian et al . 1989 ) plus a narrow core at 6.4 kev which is presumably from the torus and also statistically required ( @xmath21 = -46 ) . though expected in theory , the fluorescence fe k@xmath1 line is always ignored when fitted to data with much lower quality . considering the high quality of the data in this letter , we also include the corresponding fe k@xmath1 components ( disk - line plus narrow core ) in all of our models by amounts of 11.3% of fe k@xmath2 ( george & fabian 1991 ) . in fact , fe k@xmath1 components are also statistically required here ( @xmath21 = - 36 ) . the outer radius ( @xmath22 ) of the disk is fixed at 1000 @xmath23 ( @xmath23 = gm / c@xmath24 ) to minimize the numbers of free parameters because the fits are not sensitive to the value for the outer radius . the best fit value is also consistent with 1000 @xmath23 when it is allowed to vary . considering the strong peak at 6.4 kev , we fixed the disk - line energy at 6.4 kev in the source rest frame . results of this fit are given in table 1 ( model a ) . the inclination of the disk is 27@xmath3 , similar to the value obtained by yaqoob & weaver ( 1997 ) . however , the blue - wing of the line beyond the 6.8 kev has not been well fitted ( see figure 2a ) . in the accretion disk model context , the high energy blue wing can be produced by an accretion disk with high inclination angle . the blue wing , as well as the edge - on geometry of ngc 4151 inferred from optical and uv observations ( evans et al . 1993 ) , motivated us to add an extra component of a disk line with large inclination to model a ( model b , fe k@xmath1 components also included ) . we assume the inner and outer radii of the disk for these two disk - line components are the same , while the emissivity index q ( f@xmath25 ) is allowed to vary independently considering the fact the line photons from different parts of the disk may have different scattering fraction . the results are also given in table 1 . the model can fit the observed line profile fairly well ( see figure 2b ) . the improvement of model b to model a is significant ( @xmath21 = -19 ) . as expected , the second disk component requires a high inclination of @xmath26 degrees , while the first component requires a face - on disk ( @xmath27 degrees ) . alternatively , the high energy blue wing between 6.8 and 8.0 kev may also be produced by the ni k@xmath2 line emission at 7.48 kev . in order to reduce the excesses in the residuals between 6.8 and 8.0 kev in figure 2a , corresponding ni k@xmath2 components by amounts of 12@xmath28% of fe k@xmath2 have to be added to model a ( see figure 2c , @xmath21 = -13 ) . zdziarski et al . ( 1996 ) argued that the asca data of ngc 4151 obtained in may 1993 can be modeled by complex absorption and a narrow fe k line only , with no broad line required , contrary to the results of yaqoob et al . ( 1995 ) and this letter . however , the asca data used by zdziarski et al have much lower signal - to - noise . through fitting the unprecedented high quality spectra of ngc 4151 obtained by asca in 1995 , we find a prominent broad fe k line profile in ngc 4151 ( see figure 1 ) , which is remarkably similar to that of mcg 6 - 30 - 15 . a model consisting a dual absorbed power law plus a scattered component and a narrow fe k line at 6.4 kev ( @xmath29 = 0.11@xmath30 ) fits the data ( 1.0 10.0 kev ) poorly ( @xmath11 = 706 for 598 degrees of freedom ) and results broad systematic positive residuals around 6.0 kev ( see figure 2d ) . an extra broad line is statistically required ( @xmath21 = -78 ) . the high signal - to - noise profile of ngc 4151 obtained in this letter shows a strong narrow peak around 6.4 kev and a huge red wing , containing a second slightly weak peak at around 5.2 kev , extending to @xmath0 4.5 kev . in this respect , the line profile of ngc 4151 is remarkably similar to that seen in mcg 6 - 30 - 15 . we also find a weak blue wing extending to about 8.0 kev , which is just visible in figure 2c of yaqoob et al . ( 1995 ) , is now clearly seen ( figure 1a ) , thanks to the greatly improved statistics . there may also be such weak blue wing in the fe k profile of mcg 6 - 30 - 15 ( see figure 1b ) . when fitted by the disk - line model plus a narrow core , a face - on disk is also required ( @xmath6 = 27@xmath3 ) , contrary to the edge - on geometry inferred from optical and uv observations ( evans et al . 1993 , pedlar et al . we consider that the compton mirror of poutanen et al . ( 1996 ) , who proposed that the central source of ngc 4151 is completely hidden from our line of sight and observed x - ray radiation is produced by scattering in the higher , cooler parts of the accretion disk corona , or in a wind , could solve the problem well . we also want to point out that scattering by an `` ionizing cone '' would also explain the similarity of fe k line profiles seen in seyfert 2 galaxies and in seyfert 1 galaxies ( turner et al . 1998 ) , indicating reprocessing by face - on disks and contradicting the expectation of the unification scheme . the variability properties of the fe k line are important to test such scattering hypothesis . besides the broad profile of the fe k line , the blue wing extending to about 8.0 kev ( see figure 1a ) , which could not be well fitted by a simple disk line plus a narrow core even after the corresponding fe k@xmath1 components included , should also be paid much attention to . one possible explanation is that we observe not only a broad fe k line in ngc 4151 , but also a corresponding ni k@xmath2 line by an amount of 12@xmath28% of fe k@xmath2 . this may be the first detection of ni k@xmath2 fluorescence line emission in the x - ray spectrum of agns . however , we notice that the energy of ni k@xmath2 line is just above that of the fe k absorption edge . the line flux , which is sensitive to the ni abundance and x - ray spectral slope , can be substantially reduced due to photo - electron absorption by fe atoms . so , the contribution of ni k@xmath2 emission line might be much weaker than expected . an alternative explanation to the weak blue wing is an extra fe k disk line with high inclination angle ( it is also possible that both the ni k@xmath2 components and the extra disk line are important in reality ) . poutanen et al . ( 1996 ) assumed that the central source is completely hidden from our line of sight by the thick part of the accretion disk wind or by the broad emission - line clouds . however , it is plausible to suppose that the thick material covers the x - ray only partially , and part of the central continuum is seen directly . the fast variability and the complex absorption of the x - ray continuum seem to support this idea . the results of two - disk component model fit ( model b ) is fully consistent with this picture . poutanen et al . ( 1996 ) assumed that the scattering region is situated along the axis of the accretion disk in the form of a cone . the half opening angle of the `` ionized cone '' determined from the optical and uv observation is about 35 degree ( see fig . 10 of evans et al . 1993 ) , the angle - averaged line profile in the scattered light should resemble the one from a face - on disk , which corresponds to the low inclination line component in our two disk - line fit . the direct component should show the same inclination as the system , which is 65@xmath3 inferred from optical and uv observations . we got an inclination of 58@xmath31 degrees for this line component , which is in good agreement with interpreting it as a direct component . comparable ew of the two line components suggests that the flux of the x - ray scattered should be commensurate roughly with the directly observed flux . the little lower parameter @xmath32 of the direct disk - line indicates that more direct fe k photons can be observed from large disk radii . this work is supported by chinese national natural science foundation , pandeng project and foundation of ministry of education . we thank the referee for many critical comments , especially on the possility of ni k@xmath2 contribution , which significantly improve the presentation of this paper . + + @xmath33errors are given at 90% level . the energy of narrow core is fixed at 6.4 kev in the galaxy rest frame and the line width fixed at zero . model a : a single cold disk line ( 6.4 kev in the galaxy rest frame ) plus a narrow core ; model b : two cold disk line with different inclinations plus a narrow core . fe k@xmath1 components by amounts of 11.3% of fe k@xmath2 are included in both models .	we present an analysis of the fe k line profile of ngc 4151 by using long asca observation data obtained in may 1995 . the unprecedented good data quality , which is much better in the energy band around 6.4 kev than that of the famous 4.2-day asca observation of mcg 6 - 30 - 15 in july 1994 , offers a unique opportunity to study the details of fe k line profile . apart from those characteristics already noticed in earlier asca observations on this object ( yaqoob et al . 1995 ) : a broad and skewed profile , with a strong peak at about 6.4 kev and a large red wing extending to @xmath04 5 kev , which is remarkably similar to that of mcg 6 - 30 - 15 , we also find a weak blue wing extending to about 8 kev , thanks to the good quality of the data . when fitted by a relativistic accretion disk line plus a narrow core at 6.4 kev , the data constrain the accretion disk to be nearly face - on , contrary to the edge - on geometry inferred from optical and uv observations . however , the extended blue wing can not be well fitted even after we include corresponding fe k@xmath1 components . ni k@xmath2 line emission by an amount of 12% of fe k@xmath2 is statistically required . an alternative explanation is a model consisting of a narrow core and two disk lines with inclinations of 58@xmath3 and 0@xmath3 , respectively . we suppose that the component with inclination of 58@xmath3 was observed directly , consistent with its edge - on geometry , and the component with inclination of 0@xmath4 was scattered into our line of sight by a compton mirror , which might be the cool accretion disk corona proposed by poutanen et al . ( 1996 ) . 1.cm 1.cm
computational electromagnetism is concerned with the numerical study of maxwell s equations . the yee scheme is known as finite difference time domain and is one of the most successful numerical methods , particularly in the area of microwave problems@xcite . it preserves important structural features of maxwell s equations@xcite . bossavit et al present the yee - like scheme and extend yee scheme to unstructured grids . this scheme combines the best attributes of the finite element method ( unstructured grids ) and yee scheme ( preserving geometric structure)@xcite . stern et al @xcite generalize the yee scheme to unstructured grids not just in space , but in 4-dimensional spacetime by discrete exterior calculus(dec)@xcite . this relaxes the need to take uniform time steps . in this paper , we generalize the yee scheme to the discrete space manifold and the time . the spacetime manifold used here is split as a product of @xmath0d time and @xmath1d or @xmath2d space manifold . the space manifold can be approximated by triangular and tetrahedrons depending on dimension , and the time by segments . so the spacetime manifold is approximated by prism lattice , on which the discrete lorentz metric can be defined . * with the technique of discrete exterior calculus , the @xmath3 value discrete connection , curvature and bianchi identity are defined on prim lattice . with discrete variation of an inner product of discrete @xmath4forms and their dual forms , the discrete source equation and continuity equation are derived . * those equations compose the discrete maxwell s equations in vacuum case , which just need the local information of triangulated manifold such as length and area . the discrete maxwell s equations here can be re - grouped into two sets of explicit iterative schemes for te and tm waves , respectively . those schemes can directly use acute triangular , rectangular , regular polygon and their combination , which has been implemented on c++ plateform to simulate the electromagnetic waves propagation and interference on manifold . maxwell s equations can be simply expressed once the language of exterior differential forms is used . the electric and magnetic fields are jointly described by a curvature @xmath5form @xmath6 in a 4-d spacetime manifold . the maxwell s equations reduce to the bianchi identity and the source equation @xmath7 where @xmath8 denotes the exterior differential operator , @xmath9 denotes the hodge star operator , and 1-form @xmath10 is called the electric current form satisfying the continuity equation @xmath11 as the exterior derivative is defined on any manifold , the differential form version of the bianchi identity makes sense for any 3d or 4d spacetime manifold , whereas the source equation is defined if the manifold is oriented and has a lorentz metric . now , we introduce the discrete counterpart of those differential geometric objects to derive the numerical computational schemes for maxwell s equations . the spacetime manifold used here is split as a product of @xmath0d time and @xmath1d or @xmath2d space manifold . the @xmath1d or @xmath2d space manifold can be approximated by triangular or tetrahedrons , and the time by segments . the length of edge and area of triangular and volume of tetrahedrons gives the discrete riemann metric on space grids . the metric on time grid is the minus of length square . the spacetime manifold is approximated by prism lattice , on which the discrete lorentz metric can be defined as the product of discrete metric on space and time . a discrete differential @xmath12-form , @xmath13 , is the evaluation of the differential @xmath12-form on all @xmath12-simplices . dual forms , i.e. , forms that we evaluate on the dual cell . suppose each simplex contains its circumcenter . the circumcentric dual cell @xmath14 of simplex @xmath15 is @xmath16 where @xmath17 is all the simplices which contains @xmath15 , ... , @xmath18 , and @xmath19 is the circumcenter of @xmath17 . the two operators in eqs.(1 ) can be discretized as follows : * discrete exterior differential operator @xmath8 , this operator is the transpose of the incidence matrix of @xmath12-cells on @xmath20-cells . * discrete hodge star @xmath9 , the operator scales the cells by the volumes of the corresponding dual and primal cells . discrete connection @xmath4form or gauge field @xmath21 assigns to each element in the set of edges @xmath22 an element of the gauge group @xmath3 : @xmath23 discrete curvature @xmath5form is the discrete exterior derivative of the discrete connection @xmath4form @xmath24 the value of @xmath6 on each element in the set of triangular @xmath25 is the coefficient of holonomy group of this face . the @xmath5form @xmath6 automatically satisfies the discrete bianchi identity @xmath26 note that since the gauge group @xmath3 used here is abelian , we need not pick a starting vertex for the loop . we may traverse the edges in any order , so long as taking orientations into account . for source case , we need discrete current @xmath4form @xmath10 . let @xmath27 and the lagrangian functional be @xmath28 where @xmath29 supposing that there is a variation of @xmath30 , vanishing on the boundary , we have @xmath31 the hamilton s principle of stationary action states that this variation must equal zero for any such vary of @xmath30 , implying the euler - lagrange equations @xmath32 which is the discrete source equation @xmath33 where @xmath34 . since @xmath35 , the discrete continuity equation can express as : @xmath36 the equations of discrete bianchi identity ( 2 ) , source equation ( 3 ) , and continuity equation ( 4 ) are called discrete maxwell s equations . discrete gauge transformations are maps @xmath37 for any @xmath38form or scalar function @xmath39 on vertex . since the discrete exterior derivative maps @xmath40 the discrete maxwell s equations(2 - 4 ) are invariant under discrete gauge transformations . since the discrete continuity equation ( 4 ) ensures @xmath41 so we have @xmath42 that is to say the lagrangian function is also invariant under discrete gauge transformations . the discrete current @xmath4form , discrete curvature @xmath5form and its dual can be written as @xmath43 where @xmath44 and @xmath45 denote the coordinate of the time , @xmath46 ( electric field ) is discrete @xmath4form on space , @xmath47 ( magnetic field ) is discrete @xmath5form on space , @xmath48 ( magnetizing field ) is the dual of @xmath49 on space , @xmath50(electric displacement field ) is the dual of @xmath22 on space , @xmath51 ( charge density ) is the discrete @xmath4form on time , @xmath52 ( electric current density ) is the discrete @xmath4form on space . the discrete maxwell s equations can be rewritten as @xmath53 where @xmath54 , @xmath55 are the restriction of @xmath8 , @xmath56 on space , and @xmath57 if allowing for the possibility of magnetic charges and current discrete @xmath58form @xmath59 the symmetric scheme can be written as @xmath60 where * @xmath61(magnetic charges ) is discrete 3-form on space , * @xmath62 ( current ) is discrete @xmath5form on space . the compact form of eqs.(6 ) can be written as @xmath63 with discrete continuity equations or integrability conditions @xmath64 [ th : fundmat ] if the initial condition satisfies the first and third equations in eqs.(6 ) , the solution of the second and fourth equations in eqs.(6 ) automatically satisfy eqs.(6 ) . proof . because the dimension of spacetime is @xmath65 , therefore @xmath66 and the continuity equations can be reduced to @xmath67 so we have @xmath68 @xmath69 now we show the scheme ( 5 ) on the product of 2d discrete space manifold and time . the second and fourth equations in eqs.(5 ) based on fig.1 are @xmath70 where @xmath71 denotes the measure of forms and dual . the summation on the right is orient , that is to say , inverse the orientation of @xmath72 , then multiply @xmath73 with @xmath74 . eqs.(7 ) can be implemented on 2d discrete manifold directly(see eq.(7 ) on rectangular gird is just the yee scheme . @xmath75{qumiandec/2weifangcheng.eps } \end{center } \begin{center } { figure 1 : edge and face with direction } \end{center } \end{minipage}\ ] ] in the absence of magnetic or dielectric materials , the relations are simple : @xmath76 where @xmath77 and @xmath78 are two universal constants , called the permittivity of free space and permeability of free space , respectively . with relations ( 8) , eqs.(7 ) can be rewritten into an explicit iterative scheme for te wave . @xmath79 the symmetric te wave scheme induced from eqs.(6 ) can be written as follows . @xmath80 if writing @xmath81 where @xmath82 is the discrete @xmath4form on space , @xmath83 is the discrete @xmath5form on space , @xmath84 is the dual of @xmath85 on space , @xmath86 is the dual of @xmath87 on space , @xmath88 is the discrete @xmath58form on space , @xmath89 is the discrete @xmath5form on space , @xmath90 is the discrete @xmath4form on time , @xmath91 is the discrete @xmath4form on space , the discrete maxwell s equations can be rewritten as @xmath92 [ th : fundmat ] if the initial condition satisfies the first and third equations in eqs.(11 ) , the solution of the second and fourth equations in eqs.(11 ) automatically satisfy eqs.(11 ) . proof.because the dimension of spacetime is @xmath65 or @xmath93 , therefore @xmath94 and the continuity equations can be reduced to@xmath95 so we have @xmath96 @xmath69 now we show the scheme ( 11 ) on the product of 2d discrete space manifold and time . the second and fourth equations in eqs.(11 ) based on fig.1 are @xmath97 with relations ( 8) , eqs.(12 ) can be rewritten into an explicit iterative scheme for tm wave . @xmath98 for real world materials , the constitutive relations are not simple proportionalities , except approximately . the relations can usually still be written : @xmath99 but @xmath100 and @xmath101 are not , in general , simple constants , but rather functions . with ohm s law @xmath102where @xmath103 is the electrical conductivity and @xmath104 is magnetic conductivity . the dec schemes can be written as @xmath105 @xmath106 the courant - friedrichs - lewy condition is a necessary condition for convergence while solving certain partial differential equations numerically . now , we find this condition for scheme ( 10 ) . condition for scheme ( 13 ) can be induced in the same way . first , we decompose dec algorithm into temporal and spacial eigenvalue problems . the temporal eigenvalue problem : @xmath107 it can approximated by difference equation @xmath108 supposing @xmath109 and @xmath110 and substituting those into eq.(14 ) , we obtain @xmath111 @xmath112 therefore @xmath113 this is the stabile condition for the temporal eigenvalue problem . the spacial eigenvalue problem : @xmath114 it can be approximated by difference equation ( 15 ) based on fig.2 . @xmath115 @xmath116{qumiandec / wending.eps } \end{center } \begin{center } { figure 2 : face and dual face } \end{center } \end{minipage}\ ] ] let @xmath117 and substitute into eq.(15 ) to obtain @xmath118 @xmath119 so we have @xmath120 in order to keep the stability of scheme ( 13 ) , we need @xmath121 or @xmath122 by the definition of truncation error , the exact solution of maxwell s equations satisfy the same relation as dec scheme except for an additional term @xmath123 . this expresses the consistency , and so convergence for dec scheme by lax equivalence theorem ( consistency @xmath124 stability @xmath125 convergence ) . the derivative of maxwell s equations is approximated by first order difference in schemes ( 10 ) and ( 13 ) . equivalently , @xmath87 and @xmath22 are approximated by linear interpolation functions . consulting the definition about accuracy of finite volume method , we can also say that schemes ( 10 ) and ( 13 ) have first order spacial and temporal accuracy , and have second order spacial and temporal accuracy on rectangular grid with equivalent space and time steps . the dec algorithm of maxwell s equations was implemented in c++ platform . the fig.3 shows the flowchart of dec schemes for maxwell s equations . @xmath126{qumiandec / liuchengtu.eps } \end{center } \begin{center } { figure 3 } \end{center } \end{minipage}\ ] ] in the common practice , not every simulation step needs to be visualized , especially when the time step size is too small . fig.4 exhibit gaussian pluses waveforms simulated by dec . @xmath127{qumiandec / tuzi1.eps } \includegraphics[scale=0.335]{qumiandec / tuzi2.eps } \includegraphics[scale=0.335]{qumiandec / tuzi3.eps } \end{center } \begin{center } { figure 4 : simulation of gaussian pluse on stanford bunny by dec } \end{center } \end{minipage}\ ] ] fig.5 exhibits two sources gaussian pluses interference simulated by dec . our algorithm can simulate more complex situation on surface and 3d - space manifold . @xmath127{qumiandec / gans1.eps } \includegraphics[scale=0.335]{qumiandec / gans2.eps } \includegraphics[scale=0.335]{qumiandec / gans3.eps } \end{center } \begin{center } { figure 5 : simulation of gaussian pluse interference on sphere by dec } \end{center } \end{minipage}\ ] ] a. nicolet , f. zolla , y.o . agha and s. guenneau , geometrical transformations and equivalent materials in computational electromagnetism , international journal for computation and mathematics in electrical and electronic engineering compel , vol . 27 , 806 - 819 , ( 2008 ) .	in this paper , the method of discrete exterior calculus for numerically solving maxwell s equations in space manifold and the time is discussed , which is a kind of lattice gauge theory . the analysis of its stable condition and error is also accomplished . this algorithm has been implemented on c++ plateform for simulating te / m waves in vacuum . 0.2 cm * keywords : * discrete exterior calculus , discrete variation , maxwell s equations , lattice gauge theory . 0.2 cm * pacs(2010 ) : * 41.20.jb , 02.30.jr , 02.40.sf , 02.60.cb .
planetary nebulae ( pne ) are a key pathway in the evolution of low to intermediate mass stars , and their central stars are the immediate precursors of white dwarfs . studies of pn central stars ( herein cspne ) are motivated by : the desire to understand the origin of the rich variety of pn morphologies ; to establish the mass - loss process via fast winds driven by radiation pressure by spectral lines ; and to secure fundamental stellar parameters that can test post - agb stellar evolution models . time - series spectroscopy is an important diagnostic tool in developing our understanding of cspne . recently , far - uv and uv datasets have revealed signatures of large - scale wind structures and evidence for modulated temporal behaviour that may provide a handle on the central star rotation rates ( e.g. prinja et al . 2012a , 2012b ) . similarities between the wind properties of h - rich cspne and those of massive population i ob stars ( which also have line - driven winds ) suggest that instabilities in variable fast winds may result in shock heated gas which emits x - rays in the central cavities of pne ( e.g. guerrero 2006 ; kastner et al . 2012 ) . in the optical waveband , time - series data are requisite for establishing systematic radial velocity shifts in cspne absorption lines . de marco et al . ( 2004 ) have for example conducted a radial velocity survey of 11 cspne to provide constraints on the binary properties of the parent agb population and thus the extent to which binarity may play a causal role in shaping non - spherical nebulae . in this paper we present time - series optical spectra of the central star of ngc 2392 ( eskimo nebula ) . our study is motivated by several interesting characteristics , discrepancies and scenarios for this pn : ( i ) the central star of ngc 2392 exhibits high he , n and low c , o abundances suggesting that the photosphere has been processed ( mndez et al . a possible scenario is that the abnormal central star abundances are due to a common - envelope evolutionary phase thus implying a close binary companion ; ( ii ) danehkar et al . ( 2011 ) employ photoionization models of high excitation pn emission lines to argue that ngc 2392 has a hot white dwarf ( @xmath0 1 m@xmath2 ) companion ; ( iii ) detailed kinematic modelling of the ( eskimo ) nebula by garcia - diaz et al . ( 2012 ) supports a near - pole orientation , complex nebula morphology with multiple kinematic components , and an evolution path that may invoke a common - envelope binary ; ( iv ) the extended and point x - ray emission from ngc 2392 ( e.g. kastner et al . 2012 ) is not entirely consistent with the predicted thermal energy converted from the kinetic energy of the fast wind . additional coronal energy from a binary companion may explain the observed high x - ray temperatures . despite all the implications of the above studies , there is no definitive evidence so far of a binary nucleus in ngc 2392 , and the time - variable and geometric characteristics of its fast wind are not established . in this study we present the analysis of high - resolution optical time - series datasets secured over two epochs in 2006 and 2010 using the 3.6 m eso and canada - france - hawaii ( cfht ) observatories . our goal is to investigate for the first time @xmath0 hourly changes in the fast wind of ngc 2392 and fluctuations close to the surface of the central star . we characterise here evidence for evolving structure in the outflow and indications of radial velocity changes in deep - seated absorption lines . a log of the time - series spectra of the central star of ngc 2392 is given in table 1 . thirteen spectra were secured over 3 consecutive nights in 2010 march using the espadons echelle spectrograph ( manset & donati , 2003 ; donati et al . 1997 ) on the 3.6 m cfht at mauna kea , hawaii . the continuum signal - to - noise ratio ( s / n ) of an individual spectrum is @xmath0 100 for 30 min exposures , with a spectral resolution , r , @xmath0 68000 . the data were reduced using the standard cfht pipeline upena . the cfht data are complemented in this study by 18 spectra obtained at the eso la silla 3.6 m telescope using the harps echelle spectrograph ( mayor et al . , 2003 ) . the eso observations were carried out by us during 2006 march over 3 consecutive nights ( table 1 ) . typical individual harps spectra have s / n @xmath0 40 ( for 30 min exposures ) and r @xmath0 110,000 . the eso automatic online pipeline was used for homogeneous reduction . lll observatory & mjd range ( days ) & no . of spectra + eso harps & 53818.002 @xmath1 53818.109 & 6 + & 53818.995 @xmath1 53819.101 & 6 + & 53820.002 @xmath1 53820.108 & 6 + + cfht espadons & 55256.358 @xmath1 55256.447 & 5 + & 55257.271 @xmath1 55257.337 & 4 + & 55258.270 @xmath1 55258.336 & 4 + all the spectra were normalised by fitting a low - order polynomial through continuum windows . the line profiles discussed here have been corrected for a radial velocity of 82 km s@xmath3 , measured in weak absorption lines . the fundamental central star parameters adopted in this study are listed in table 2 . .ngc 2392 adopted central star parameters . [ cols="<,<,<",options="header " , ] there are no archival fuv or uv _ time - series _ datasets available for ngc 2392 that permit an investigation of systematic variability on time - scales comparable to the wind flow time ( i.e. @xmath0 hours ) . patriarchi & perinotto ( 1995 ) show sparse @xmath4 spectra separated by @xmath0 7 years which suggest some fluctuations in the civ and nv lines , and guerrero & de marco ( 2013 ) comment on the two _ fuse _ spectra available for the central star . the optical time - series data of the nuclei of ngc 2392 that we present here provide access to the fast wind via excited transitions arising from hei and heii . ( note that the h@xmath5 and h@xmath6 lines are heavily affected by nebular emission . ) the he lines are primarily the result of recombination from a higher ionization stage followed by a radiative de - excitation . since recombination is sensitive to the square of wind density ( @xmath7 ) , the optical lines form in the densest region of the wind ( in contrast to the uv lines ) . furthermore the hei @xmath85876 ( 2@xmath9p@xmath10 @xmath1 3@xmath9d ) transition is a potential exception if large populations are acquired in its lower level such that it effectively becomes the ground level and consequently behaves as a scattering or resonance line . we find that the stellar lines in the optical spectra of ngc 2392 are undoubtedly variable within @xmath0 hours , with peak - to - peak amplitude changes of @xmath0 5% in hei @xmath85876 and heii @xmath84686 . the corresponding equivalent width and standard deviations for the combined eso and cfht datasets are @xmath0 0.32 s.d . 0.12 and @xmath0 @xmath13.8 s.d . 0.6 . the line profile variability is quantified in fig . 1 where we employ the temporal variance spectrum ( tvs ) method ( e.g. fullerton et al . 1996 ) to estimate the statistical significance of the changes , having accounted for differences in the data quality at spectrum and pixel levels . the variability level is higher in heii @xmath84686 than hei @xmath85876 or hei @xmath84472 ; note that numerous other stellar lines are also weakly variable in the night - to - night data . significant variability extends to @xmath0 @xmath1180 km s@xmath3 ( i.e. @xmath0 0.45 @xmath11 ) in the eso ( 2006 ) and cfht ( 2010 ) spectra of heii @xmath84686 , but the redward extend of these changes is lower in 2010 compared to 2006 . for both epochs the hei @xmath85876 weak p cygni profile varies between @xmath0 @xmath1150 km s@xmath3 to 100 km s@xmath3 . an important feature of the tvs in fig . 1 is that it is _ double - peaked _ in the heii and hei lines for the eso and cfht datasets . typically the peaks in the tvs are separated by @xmath0 100 to 120 km s@xmath3 . the amplitude of the tvs peaks are not systematically stronger or weaker in either the blue or red sides . for line profile variability dominated by clumps we may expect small - scale structure to be distributed evenly over the whole inner wind thus presenting a broadly symmetrical single - peak tvs . one possibility for the double - peak tvs is that it results from radial velocity shifts perhaps due to a binary nature in ngc 2392 . we return to the issue of radial velocity changes in sect . the systematics of the heii @xmath84686 and hei @xmath85876 profile changes are examined further in figs . 2 and 3 where the data are displayed in two - dimensional velocity@xmath1time dynamic spectra ( images ) . the intensity levels of individual spectra are determined as _ differences _ with respect to the average spectrum for each observing night , and represented by a colour scale from minimum ( black ) to maximum ( white ) intensity cut level . the features seen in figs . 2 and 3 therefore represent pseudo - absorption and pseudo - emission features relative to the mean . the fast wind in ngc 2392 is undoubtedly variable on very short time - scales and can switch between an overall pseudo - emission to pseudo - absorption pattern in the dynamic spectra in less than 30 minutes ( i.e. time lengths comparable to the characteristic wind flow time , @xmath0 @xmath12 ) . the evidence from figs . 2 and 3 is that episodes of strong blue and red changes in heii @xmath84686 are mimicked in hei @xmath85876 at the same velocities . for example enhanced redward emission in heii @xmath84686 is accompanied by an increase in the ( weak ) p cygni emission component in hei @xmath85876 . a substantial low velocity blueward emission increase in heii @xmath84686 is ` matched ' by an increase in the p cygni absorption strength in hei @xmath85876 . the overall impression from our optical time - series is that the fast wind in ngc 2392 is stochastically variable over @xmath0 hours as opposed to revealing coherent modulated behaviour . we do not for example see signs of blueward and/or redward migrating features during each night , nor is there evidence for much slower evolving ( large - scale ) structures in the wind that persist over @xmath0 3 nights . 5411 , hei @xmath85876 and heii @xmath84686 line profiles taken @xmath0 1 day apart . the left- and right - hand panels show examples from the eso and cfht data , respectively . ] the overall line profile morphology can change in complex manners from night - to - night . we show in fig . 4 corresponding pairs of hei @xmath85876 , heii @xmath85411 and heii @xmath84686 spectra . each pair is separated by @xmath0 1 day . the left - hand panel in fig . 4 shows a case where as the total hei @xmath85876 absorption increases across all velocities , in the corresponding heii @xmath84686 pair the overall emission decreases . for this same pair , the changes in heii @xmath85411 are however clearly _ asymmetric _ toward redward velocities . in a further twist , the right - hand panels in fig . 4 show a case where the hei @xmath85876 line profile transforms from an absorption profile to a clear p cygni profile with redward emission . in this case the corresponding heii @xmath84686 pair exhibit almost no change in total equivalent width but the peak emission shifts blueward by @xmath0 30 km s@xmath3 . but , once more , the heii @xmath85411 profile changes are almost entirely at redward velocities . to gain further insights into the implications of these overall profile changes , we examined model predictions using the unified non - lte , line - blanketed model atmosphere code cmfgen ( e.g. hillier & miller 1998 ) . briefly , cmfgen solves the non - lte radiation transfer problem assuming a chemically homogeneous , spherically symmetric , steady - state outflow . each model is defined by the stellar radius , the luminosity , the mass - loss rate , the wind terminal velocity ( @xmath11 ) , the stellar mass and by the abundances of the species included in the calculations . the code does not solve for the hydrodynamical structure , hence the velocity field has to be defined using the output of a plane - parallel model ( tlusty in this case ; see hubeny & lanz , 1995 ) to define the pseudo - static photosphere , connected just below the sonic point to a beta - type velocity law to describe the wind regime . we calculated a substantial grid of cmfgen model atmosphere spectra and explored sequences of different mass - loss and clumping . the mass - loss rate has been incremented between @xmath0 7 @xmath13 10@xmath14 m@xmath2 yr@xmath3 to @xmath0 1.6 @xmath13 10@xmath15 m@xmath2 yr@xmath3 . we also adjusted the clumping factor , @xmath16 , which measures the over - density inside the clumps with respect to the average wind density . finally , the velocity at which optically thin clumping starts ( @xmath17 ) near the stellar surface was adjusted between 10 km s@xmath3 to 40 km s@xmath3 , using an exponential law defined by the parameters @xmath16 and @xmath17 . in each sequence all but one of the above parameters are kept fixed . the model sequences are shown in fig . 5 for heii @xmath84686 , heii @xmath85411 and hei @xmath85876 . comparisons between this grid with variable mass - loss rates and wind clumping sequences lead us to conclude that the absorption versus emission strength changes seen in the cfht and eso data in the left - hand panels in fig . 4 may qualitatively be explained by ( global ) mass - loss differences , but not the asymmetric behaviour evident in heii @xmath85411 . furthermore the relatively large emission line shift seen in the right - hand panels in fig . 4 _ can not _ be explained by line - synthesis model predictions with a spherically homogeneous wind . ordinarily the search for radial velocity shifts in symmetric photospheric absorption lines would be a relatively straightforward exercise in hot stars . trends in velocity changes may then betray for example the causal role of stellar binarity or multiplicity . the situation in ngc 2392 is more complex however for two key reasons ; ( i ) we have demonstrated in sect . 3 that the nucleus of ngc 2392 drives a highly variable fast wind , and the origin of these changes are deep - seated and close to the stellar surface . this means that a large selection of the he i , he ii and metal absorption lines in the optical spectra may be disturbed by the effects of a variable wind ; ( ii ) in their analysis of the 3-d and kinematic structure of the nebula , garcia - diaz et al . ( 2012 ) present evidence that the inner nebular shell has an almost pole - on orientation , such that the inclination angle with respect to the line - of - sight is only @xmath18 . assuming that the central star has the same orientation , the geometry is obviously not favourable for the detection of binary induced radial velocity motion of the nucleus . taking up this challenge nevertheless , we have examined weak , relatively symmetrical stellar absorption lines in ngc 2392 for evidence of radial velocity shifts . the cmfgen line - synthesis models ( sect . 3 ) suggest that most of the optical lines in ngc 2392 are potentially affected by mass - loss and wind clumping changing . our grid of models points to the niv @xmath86380.8 absorption line as a rare example of a primarily photospheric absorption line , suitable for a study of subtle radial velocity fluctuations . inspection of the time - series data suggests that the niv line _ does _ in fact shift in central velocity by @xmath0 10 to 15 km s@xmath3 in the eso and cfht datasets , over hourly timescales . pairs of niv profiles separated by @xmath0 2 hours are shown in fig . 6 where the central minimum and blue and red wings are shifted by @xmath0 15 km s@xmath3 . a particularly useful consistency check is provided by h@xmath19 , since the caiih @xmath83968.5 interstellar / circumsystem lines provide an excellent nearby fiducial for the accuracy of the wavelength scale close to h@xmath19 . we note in fig . 6 that the velocity behaviour seen in niv is systematically mimicked in h@xmath19 , thus firming confidence in the notion that there is a genuine radial velocity shift in ngc 2392 . the niv line profiles were fitted with gaussian model profiles with least squares to estimate the central absorption velocities . for a typical internal fitting error of @xmath20 1 km s@xmath3 , the mean and median absolute deviation ( mad ) of the eso and cfht data are 6.4 ; mad @xmath0 3.7 km s@xmath3 and 3.9 ; mad @xmath0 3.4 km s@xmath3 . the time - series data are unfortunately not intensive enough over a sufficiently extensive time - scale to allow for a confident search for periodic behaviour in the recorded radial velocities . with this caveat , fourier power spectra were calculated independently for the cfht and harps measurements and the results are shown in fig . 7 ( upper panel ) . there are undoubtedly spurious features and aliases throughout the power spectrum , but it is interesting to note that in both datasets the maximum power occurs at very similar frequencies i.e. @xmath0 8.1 d@xmath3 . 6380.8 line profiles secured @xmath0 2 hours apart in cfht and eso spectra . simultaneous velocity changes are also evident h@xmath19 , where the sharp interstellar feature of caiih @xmath83968.5 is an excellent wavelength scale fiducial . ] for a slightly more sophisticated treatment we also employed the clean algorithm ( roberts et al . 1987 ) which is more suited to the temporal analysis of unequally spaced finite data samples . using a gain of 0.5 and 100 iterations , the spectral window function was iteratively subtracted in the fourier domain from the raw power spectrum shown in the upper panel in fig . the resultant ` cleaned ' power spectrum is shown in the bottom panel in fig . 7 , where the strongest peak frequency is at @xmath0 8.13 d@xmath3 , _ for both datasets_. we estimate an uncertainty of at least @xmath0 10% in this frequency based on the half - width at half - maximum of the main peak in the window function . these results provide tentative evidence for a period of @xmath0 0.123-day in the niv radial velocity changes , with a semi - amplitude of @xmath0 10 km s@xmath3 . figure 8 shows the niv central velocities phased on 0.123-day . ( in each case phase 0 is arbitrarily set to the timing of the first observation . ) the results in fig . 8 suggest some coherency on the 0.123-day period _ independently _ for the eso ( 2006 ) and cfht ( 2010 ) datasets , though there is relatively larger scatter ( in the very small velocity displacements ) at some phases . ultimately a more intensive , very high - resolution time - series dataset of the central star is needed to confirm the significance of this signal . 6380.8 stellar radial velocity changes in ngc 2392 . the cfht ( solid black ) and eso ( dotted red ) power spectra are shown in the upper panels , and corresponding cleaned versions are shown in the lower panel . in _ both _ datasets a peak frequency of @xmath0 8.1 d@xmath3 is apparent . ( power is in arbitrary units . ) ] 6380.8 absorption line phased on the @xmath0 0.123-d period . the eso harps and cfht espadons data are separated by @xmath0 4 years . pairs of representative fitting error bars are shown in each panel . ] we have provided evidence and demonstrated that in two independently secured optical spectroscopic time - series separated by @xmath0 3 years , both datasets reveal for the central star of ngc 2392 ; ( i ) stochastic variations in the fast wind - formed recombination lines on timescales down to @xmath0 30 min . , ( ii ) changes in the overall morphology of the hei and heii line profiles that can not be accounted for by 1-d line - synthesis predictions for a spherically homogeneous wind , ( iii ) radial velocity shifts of semi - amplitude @xmath0 10 km s@xmath3 in niv @xmath86380.8 ( and tentatively in h@xmath19 ) that show maximum fourier power spectra signal at @xmath0 0.123-d in eso ( 2006 ) and cfht ( 2010 ) data . our detection of radial velocity motion of the central star in ngc 2392 is obviously tentative . cross - correlation with the majority of stellar absorption lines in the optical range is not fruitful since most of features are ` contaminated ' by the imprints of stellar wind variability which we have established here . in advance of more definitive optical spectroscopy and photometry being secured , we can only speculate on the potential binary components in ngc 2392 : for an assumed circular orbit and line - of - sight inclination @xmath0 @xmath21 ( e.g. garcia - diaz et al . 2012 ) ; semi - amplitude of 10 km s@xmath3 and period = 0.123-d ( figs . 8 and 9 ) ; assumed ( primary ) central star mass = 0.6 m@xmath2 , the implied dynamical mass of the secondary in circular keplerian orbit is @xmath0 0.1 m@xmath2 . such a low - mass companion would for example correspond to a late m - dwarf of @xmath22 @xmath0 2000 - 3000k . the morphologies of the fuv and uv lines are complex in ngc 2392 and provide additional signatures for an asymmetric geometry . the cmfgen line - synthesis models ( sect . 3 ) do not provide consistent matches to doppler widths , and absorption and emission strengths across all the uv ion stages observed in the fast wind . a selection of uv wind line profiles in ngc 2392 is presented in fig . 9 , ranging from ovi @xmath81031.9 and svi @xmath8944.5 , to pv @xmath81118.0 and nv @xmath81238.8 , and civ @xmath81548.2 and mgii @xmath82795.5 . ( the data have been retrieved from the _ fuse _ and _ iue _ archives . ) there are some key points to note in fig 9 : ( i ) all the wind lines get weaker with increasing outflow velocity . it may be that the wind plasma is shifting to a very high ionization state ( beyond ovi ) as it travels to larger radii . alternatively , the line shapes in fig . 9 could be an indication that the fast wind of ngc 2392 is moving out of the line - of - sight , somewhat as may be expected for a polar , high - latitude wind in an asymmetric geometry ; ( ii ) the low excitation mgii line is very narrow ( @xmath0 100 km s@xmath3 ; as is siiii @xmath81206.5 ) and consistent with a low - velocity equatorial wind ( see e.g. bjorkman et al . 1994 ; massa 1995 ) ; ( iii ) the presence of pv most likely indicates that the wind is optically very thick , since phosphorous has a low cosmic abundance and this line would otherwise not be so clearly detected . however civ and nv are weak at intermediate velocities ( @xmath0 200 to 300 km s@xmath3 ) . an optically very thick wind that causes weak absorption can arise in a scenario where the wind is not covering the entire stellar disk , as may be expected from a polar wind . we conclude that uv lines provide evidence for an asymmetric , two - component outflow in ngc 2392 , where high - speed high - ionization gas forms preferentially in the polar region . slower , low ionization material is then confined primarily to a cooler equatorial component of the outflow . 1031.9 , svi @xmath8944.5 , pv @xmath81118.0 , nv @xmath81238.8 , civ @xmath81548.2 , mgii @xmath82795.5 . based on observations obtained at the canada - france - hawaii telescope ( cfht ) which is operated by the national research council of canada , the institut national des sciences de lunivers of the centre national de la recherche scientique of france , and the university of hawaii , and on observations collected at the european southern observatory , la silla ( programme i d eso 076.d-0207(a ) . we thank derck massa for discussions about the fast wind of ngc 2392 . we acknowledge the helpful comments of the referee . bjorkman , j , e . , ignance , r. , tripp , t.m . , cassinelli , j.p . 1994 , apj , 435 , 416 danehkar , a. , frew , d.j . , parker , q.a . , de marco , o. 2011 , in interacting binaries to exoplanets : essential modeling tools , iau symp . 282 , eds . m. t. richards and i. hubeny , p. 470 donati , j .- f . , semel , m. , carter , b.d . , rees , d.e . , collier cameron , a. 1997 , mnras , 291 , 658 fullerton , a.w . , gies , d.r . , bolton , c.t 1996 , apjs , 103 , 475 garcia - diaz , m.t . , lpez , j.a . , steffen , w. , richer , m.g . 2012 , apj , 761 , 172 guerrero , m.a . , de marco , o. 2013 , a&a , 553 , 126 guerrero , m.a . 2006 , in planetary nebulae in our galaxy and beyond , proceedings of the international astronomical union , symposium 234 . barlow and r.h . mndez , cambridge university press , 2006 . , p.153 herald , j.e . , bianchi , l. 2011 , mnras , 417 , 2440 hillier , d.j . , miller , d. , 1998 , apj , 496 , 407 hubeny , i. , lanz , t. 1995 , apj , 439 , 875 kaschinski , c.b . , pauldrach , a.w . a. , hoffmann , t.l . , 2012 , a&a , 542 , 45 kastner , j.h . , montez , r. , balick , b. 2012 , aj , 144 , 58 manset , n. , donati , j .- f . , 2003 , in polarimetry in astronomy , proceedings of the spie . s. fineschi , 4843 , p. 425 massa , d.l . 1995 , apj , 438 , 376 mayor , m. et al . 2003 , the messenger 114 , 20 mndez , r.h . 1991 , in evolution of stars : the photospheric abundance connection , iau symp . 145 , eds . g. michaud and a.v . tutukov , kluwer academic publishers , dordrecht , p.375 mndez , r.h . , urbaneja , m.a . , kudritzki , r .- p . , prinja , r.k . 2012 , in planetary nebulae : an eye to the future , iau symp . 283 , eds . a. manchado , l. stanghellini and d. schoenberner , cambridge university press , p. 436 de marco , o. , bond , h.e . , harmer , d. , fleming , a.j . 2004 , apj , 602 , 93 patriarchi , p. , perinotto , m. 1995 , a&s , 110 , 353 prinja , r.k . , massa , d.l . , urbaneja , m.a . , kudritzki , r .- 2012a , mnras , 422 , 3142 prinja , r.k . , massa , d.l . , cantiello , m. 2012b , apj , 759 , l28 roberts , d.h . , lehr , j. , dreher , j.w . 1987 , aj , 93 , 968	we report on high - resolution optical time - series spectroscopy of the central star of the ` eskimo ' planetary nebula ngc 2392 . datasets were secured with the eso 2.3 m in 2006 march and cfht 3.6 m in 2010 march to diagnose the fast wind and photospheric properties of the central star . the hei and heii recombination lines reveal evidence for clumping and temporal structures in the fast wind that are erratically variable on timescales down to @xmath0 30 min . ( i.e. comparable to the characteristic wind flow time ) . we highlight changes in the overall morphology of the wind lines that can not plausibly be explained by line - synthesis model predictions with a spherically homogeneous wind . additionally we present evidence that the uv line profile morphologies support the notion of a high - speed , high - ionization polar wind in ngc 2392 . analyses of deep - seated , near - photospheric absorption lines reveals evidence for low - amplitude radial velocity shifts . fourier analysis points tentatively to a @xmath0 0.12-d modulation in the radial velocities , independently evident in the eso and cfht data . we conclude that the overall spectroscopic properties support the notion of a ( high inclination ) binary nucleus in ngc 2392 and an asymmetric fast wind . [ firstpage ] stars : mass - loss @xmath1 stars : evolution @xmath1 stars : individual : ngc 2392 @xmath1 optical : stars
i thank the cern and desy theory groups for their hospitality during this work , and w. buchmller and m. lscher for fruitful discussions . this work was supported in part by the united states department of energy under contract no . de fg02 90er40560 . # 1#2#3am . j. phys . * # 1 * ( # 3 ) # 2 # 1#2#3acta phys . austriaca suppl . * # 1 * ( # 3 ) # 2 # 1#2#3ann . ( n.y . ) * # 1 * ( # 3 ) # 2 # 1#2#3acta phys . polonica * # 1 * ( # 3 ) # 2 # 1#2#3ann . nucl . part . sci . * # 1 * ( # 3 ) # 2 # 1#2#3commun . # 1 * ( # 3 ) # 2 # 1#2#3comments on nucl . part . # 1 * ( # 3 ) # 2 93_lepton and photon interactions : xvi international symposium , ithaca , ny august 1993 _ , aip conference proceedings no . 302 , ed . by p. drell and d. rubin ( aip , new york , 1994 ) 89_cp violation , _ edited by c. jarlskog ( world scientific , singapore , 1989 ) 94dpf 94 meeting , albuquerque , nm , aug . 26 , 1994 # 1#2#3europhys . # 1 * ( # 3 ) # 2 79_proceedings of the 1979 international symposium on lepton and photon interactions at high energies , _ fermilab , august 23 - 29 , 1979 , ed . by t. b. w. kirk and h. d. i. abarbanel ( fermi national accelerator laboratory , batavia , il , 1979 87_proceeding of the 1987 international symposium on lepton and photon interactions at high energies , _ hamburg , 1987 , ed . by w. bartel and r. rckl ( nucl . b , proc . suppl . 3 ) ( north - holland , amsterdam , 1988 ) # 1#2#3 * # 1 * ( # 3 ) # 2 72_proceedings of the xvi international conference on high energy physics _ , chicago and batavia , illinois , sept . 613 , 1972 , edited by j. d. jackson , a. roberts , and r. donaldson ( fermilab , batavia , il , 1972 ) # 1#2#3int . j. mod . phys . a * # 1 * ( # 3 ) # 2 # 1#2#3j . # 1 * ( # 3 ) # 2 # 1#2#3j . g * # 1 * ( # 3 ) # 2 87_selected topics in electroweak interactions _ ( proceedings of the second lake louise institute on new frontiers in particle physics , 1521 february , 1987 ) , edited by j. m. cameron ( world scientific , singapore , 1987 ) 85_proceedings of the international symposium on lepton and photon interactions at high energy , _ kyoto , aug . 19 - 24 , 1985 , edited by m. konuma and k. takahashi ( kyoto univ . , kyoto , 1985 ) # 1#2#3mod . phys . lett . a * # 1 * ( # 3 ) # 2 # 1#2#3nuovo cim . * # 1 * ( # 3 ) # 2 # 1#2#3nucl . # 1 * ( # 3 ) # 2 # 1#2#3#4pisma zh . eksp . fiz . * # 1 * ( # 3 ) # 2 [ jetp lett . * # 1 * ( # 3 ) # 4 ] # 1#2#3phys . * # 1 * ( # 3 ) # 2 # 1#2#3phys . b * # 1 * ( # 3 ) # 2 # 1#2#3phys . # 1 * ( # 3 ) # 2 # 1#2#3phys . rev . a * # 1 * ( # 3 ) # 2 # 1#2#3phys . d * # 1 * ( # 3 ) # 2 # 1#2#3phys . lett . * # 1 * ( # 3 ) # 2 # 1#2#3phys . rep . * # 1 * ( # 3 ) # 2 # 1#2#3prog . # 1 * ( # 3 ) # 2 # 1#2#3rev . # 1 * ( # 3 ) # 2 # 1 rp # 1 9025th international conference on high energy physics , singapore , aug . 2 - 8 , 1990 87_proceedings of the salt lake city meeting _ ( division of particles and fields , american physical society , salt lake city , utah , 1987 ) , ed . by c. detar and j. s. ball ( world scientific , singapore , 1987 ) 89_proceedings of the xivth international symposium on lepton and photon interactions , _ stanford , california , 1989 , edited by m. riordan ( world scientific , singapore , 1990 ) 82_proceedings of the 1982 dpf summer study on elementary particle physics and future facilities _ , snowmass , colorado , edited by r. donaldson , r. gustafson , and f. paige ( world scientific , singapore , 1982 ) 90_research directions for the decade _ ( proceedings of the 1990 summer study on high energy physics , june 25 july 13 , snowmass , colorado ) , edited by e. l. berger ( world scientific , singapore , 1992 ) 90_testing the standard model _ ( proceedings of the 1990 theoretical advanced study institute in elementary particle physics , boulder , colorado , 327 june , 1990 ) , edited by m. cveti and p. langacker ( world scientific , singapore , 1991 ) # 1#2#3#4yad . # 1 * ( # 3 ) # 2 [ sov . j. nucl . # 1 * ( # 3 ) # 4 ] # 1#2#3#4#5#6zh . # 1 * ( # 3 ) # 2 [ sov . - jetp * # 4 * ( # 6 ) # 5 ] # 1#2#3zeit . c * # 1 * ( # 3 ) # 2	when a heavy quark and the corresponding antiquark are separated by more than @xmath0 fm , it becomes energetically favorable for a light quark - antiquark pair to be produced , leading to fragmentation into a pair of flavored mesons . the relation of this critical quark separation to other dimensional constants of the strong interactions ( such as the pion decay constant , the qcd scale , and the light - quark constituent mass ) is discussed . 5.8 in -0.8 in -0.2 in * critical spacing for heavy quarkonium dissociation * + + + + + _ university of chicago , chicago , il 60637 usa _ + the dimensional nature of quantum chromodynamics is manifested in a number of different ways , all of which are plausibly equivalent to one another . for example : ( 1 ) the scale of low - energy pion - pion interactions is set by the ratios @xmath1 , where @xmath2 are pion 4-momenta and @xmath3 mev is the pion decay constant . ( 2 ) a qcd scale @xmath4 mev whose specific value depends on renormalization scheme is necessary in order to define the strong - coupling constant @xmath5 at a suitable momentum scale . ( 3 ) the masses of @xmath6 and @xmath7 quarks as manifested in hadrons can be regarded as `` constituent - quark '' values @xmath8 of order 300 400 mev . the scale of @xmath8 , while not as precisely defined as @xmath9 or @xmath10 , must be related to these two quantities if , as expected , the limit of zero bare mass for @xmath6 and @xmath7 quarks makes sense . in the present note we add a quantity to this list . there appears to be a critical separation between a heavy quark and a heavy antiquark for which it is energetically favorable to produce a pair of flavored mesons [ as , for example , in @xmath11 . this feature was noted in early studies of the @xmath12 system @xcite . however , the explicit magnitude of the interquark separation leading to heavy quarkonium dissociation was not presented . we find that it is of order @xmath0 fm . this number is of current interest for several reasons . \(1 ) lattice gauge theories are approaching a stage where the dimensional quantities just mentioned ( as well as others ) can be related to one another . the study of light - quark production between color centers represented by a heavy quark and antiquark is becoming feasible @xcite as one learns to cope with the `` unquenched '' version of qcd in which light quark - antiquark pairs are properly treated . \(2 ) the quarkonium systems to which the critical dissociation distance applies include @xmath13 and @xmath14 systems , for which there remain prospects for discovering a few more p - wave and d - wave states , and @xmath15 states , for which there are extensive theoretical studies of the spectroscopy @xcite and a hint of the ground state @xcite . the top quark is too heavy to have a quarkonium spectroscopy ( since its decay width will be greater than the expected level spacing ) , but the critical - separation parameter should still apply to @xmath16 pairs or their products @xmath17 , @xmath18 , or @xmath19 @xcite . \(3 ) one might expect the critical separation of a pair of static color centers for dissociation into a pair of flavored mesons to be a fundamental parameter in theories @xcite linking heavy - quark physics and chiral perturbation theory . we assume that a flavor - independent potential @xmath20 describes all bound states @xmath21 of a heavy quark @xmath22 and its antiquark as well as flavored mesons @xmath23 and @xmath24 involving the heavy quark @xmath22 and a light quark @xmath25 . in the notation of ref . @xcite , we define @xmath26 . this quantity should tend to a finite limit @xmath27 as @xmath28 . the quantity @xmath29 is the same as the parameter @xmath30 of heavy quark effective theory @xcite . we seek the value of the critical threshold @xmath21 separation @xmath31 for which @xmath32 . a trivial modification permits the description of bound states such as @xmath15 involving unlike - mass quarks . the value of @xmath33 will be universal to the extent that ( a ) @xmath27 is well - defined ( as expected in heavy quark effective theory ) , and ( b ) a flavor - independent potential @xmath20 actually provides a good description of the @xmath21 interaction near flavor threshold . the parameters @xmath29 and the light - quark constituent mass @xmath8 are closely related ; neither parameter can be chosen arbitrarily in a theory with given @xmath9 or @xmath10 . the assumption of a flavor - independent potential has been reasonably well borne out @xcite by comparison of @xmath13 and @xmath14 systems , and will be tested further in studies of @xmath15 states @xcite . we calculate the critical separation using a simple phenomenological potential @xcite . we then compare it with values obtained in a more model - independent manner , and discuss its relation to other dimensional constants in qcd . a satisfactory interpolation between charmonium ( @xmath13 ) and upsilon ( @xmath34 ) states is provided by a potential of the form @xmath35 , with @xmath36 gev and @xmath37 depending upon the specific choice of @xmath38 and @xmath39 quark masses @xcite . the schrdinger equation for the reduced radial wave function @xmath40 of s - wave bound states is [ eqn : schr ] - + c ( ) u = e u , where @xmath41 is the reduced mass : @xmath42 for a @xmath21 bound state , and @xmath43 . in terms of dimensionless variables @xmath44 and @xmath45 , ( [ eqn : schr ] ) becomes - + u = [ + ( r_0 ) ] u . the lowest eigenvalue of this equation is @xmath46 , based on a numerical solution @xcite . consequently , one may eliminate the parameter @xmath37 in favor of the ground state mass , the parameter @xmath47 , and the reduced mass . recalling the definition of the radius @xmath31 and setting @xmath48 , we finally have /c + 1.0443 = ( ) . we begin by neglecting spin - dependent effects in both @xmath21 and @xmath49 systems , and reduced - mass effects in the @xmath49 system . this approximation is most reliable for the @xmath12 levels . consequently , taking @xmath50 gev , @xmath51 = 9.460 $ ] gev , @xmath52 gev , and a range of @xmath53 between 4.5 and 5 gev , we find the values of @xmath31 shown as the dashed line in fig . 1 . to account for the hyperfine splittings in the @xmath21 and @xmath49 systems and the reduced - mass effect in the @xmath49 system , we assume that the @xmath54 level is 10 mev above the spin - averaged @xmath55 mass @xcite . furthermore , we apply the corrections @xcite ( m_q)_hfs = ( m_b^ * - m_b ) = 69 mev ( ) and ( m_q)_red . mass = - c c m_q ( - ) to estimate - ( m_b ) 10 mev + 69 mev - 46 mev = 33 mev , where the first term is our estimate of the @xmath56 hyperfine term , the second is the contribution of the hyperfine term in @xmath57 , and the third is the reduced - mass effect , estimated for the logarithmic potential . one is thus assuming such a potential to hold not only for @xmath21 but also for @xmath58 states . we have used a light - quark constituent mass @xcite @xmath59 mev . with the above corrections , we now estimate the range of @xmath31 shown as the solid line in fig . 1 . the corrected values of @xmath31 range between about 7.6 and 7.2 gev@xmath60 ( 1.5 1.42 fm ) for @xmath61 gev . for comparison , the dot - dashed line in fig . 1 depicts the values of @xmath31 obtained from an inverse - scattering construction of the interquark potential @xcite using the @xmath12 levels . these values have not been corrected for hyperfine or reduced - mass effects , so they should be compared with the uncorrected values obtained above . a power - law potential @xcite fitting charmonium and @xmath12 spectra with @xmath62 gev gives @xmath63 gev@xmath60 ( uncorrected ) and 7.6 gev@xmath60 ( corrected ) . the agreement of the various estimates is fairly good , indicating that model - dependent effects are unlikely to affect the estimate significantly . as shown in inverse - scattering @xcite and explicit potential @xcite calculations , the shape of any smooth potential which reproduces a given set of energy levels is fairly well specified for the range of energies corresponding to the known levels . once a potential is required to reproduce the @xmath64 levels and their leptonic widths , that potential s shape is specified between roughly 0.1 fm = 0.5 gev@xmath60 and the interquark separation corresponding to @xmath65 - 2 m_b$ ] , which is just above flavor threshold . lattice gauge theories are able to estimate the distance at which a string of chromoelectric flux breaks by the deviation from a linear behavior of the field energy as a function of interquark separation . it is not clear that such a distance corresponds to flavor threshold , since discrete @xmath14 resonances ( and indeed , series of light - quark resonances ) persist well above flavor threshold . nonetheless , our estimate of @xmath31 is in accord with an upper bound of @xmath66 fm @xcite ( quoted in ref . @xcite as 1.7 fm on the basis of a calculation in ref . @xcite ) for the breaking of a qcd string obtained using a quenched approximation in lattice gauge theory . in a quenched lattice of size ( 1.5 fm)@xmath67 , no breaking of a qcd string has been observed ; the linear behavior persists out to the maximum accessible interquark separation . it would be interesting to see whether at slightly larger distances a linear behavior could actually _ coexist _ with production of light quark - antiquark pairs . how might the critical spacing parameter @xmath31 be related to other dimensionful quantities in qcd ? it is clearly related to the constituent - quark mass since a certain amount of chromoelectric energy is required to create the @xmath68 pair . in turn , the constituent - quark mass scale is set @xcite by the need to agree with such quantities as @xmath69 and @xmath70 , whose values are related to the qcd scale @xmath10 . it has been argued @xcite ( cf . , however , refs . @xcite ) that the mass scale of resonances like the @xmath71 meson can be regarded as a number of order @xmath72 in a chirally symmetric theory involving massless pions . in any event a direct relation between @xmath9 and @xmath73 in a theory of massless pions is highly likely . further dimensionful quantities in the strong interactions which might bear a relation to those mentioned include the universal string tension describing the long - distance interquark interaction @xmath74 , @xmath75 gev@xmath76 , and the universal slope @xmath77 of regge trajectories for light - quark systems , @xmath78 gev@xmath79 @xcite . the production of light quark - antiquark pairs in a linear potential has been considered some time ago @xcite . to conclude , we have argued that once a heavy color triplet and antitriplet become separated by more than 1.4 1.5 fm , the chromoelectric flux lines joining them contain sufficient energy to produce a light quark - antiquark pair , leading to the decay of the heavy quarkonium system into a pair of flavored mesons . it would be interesting to see whether current lattice gauge and chiral theories of non - perturbative qcd could relate this quantity to others which set the qcd scale .
the distribution of mass in galaxy clusters , groups and massive galaxies provides a powerful tool for cosmological studies . explicit predictions from our current @xmath0cdm cosmological paradigm for the number , size and radial mass distribution of dark matter halos can now be tested against high - quality constraints from studies employing lensing , sunyaev - zeldovich , stellar dynamics and , in particular , x - rays @xcite . the relative distribution of dark and baryonic mass , coupled with the thermodynamic state of the hot intracluster medium , similarly provides a unique insight into the uncertain baryonic physics of galaxy formation , such as the role of feedback in shaping the nascent structures , and the complex interplay between adiabatic contraction and dynamical friction @xcite . spherical , hydrostatic x - ray techniques are an appealing method for measuring such mass distributions due to their computational simplicity , given the isotropy of the gas pressure tensor , and the small biases introduced by the spherical approximation ( * ? ? ? * and references therein ) , particularly if the spherically averaged mass profile is close to a singular isothermal sphere @xcite . while the hot gas permeating the potential well is not expected to be exactly hydrostatic , theoretical arguments and observational constraints suggest only modest ( 30% ) biases on the inferred gravitating mass distribution , provided care is taken to study systems with relaxed x - ray morphologies @xcite . with the current generation of x - ray observatories , x - ray methods are especially appealing as they can provide mass measurements over @xmath13 orders of magnitude in virial mass , or more , and the radial mass distribution inferred can span a similarly large dynamical range in radius @xcite . for spherically distributed hot gas in hydrostatic equilibrium , the radial mass profile can be uniquely inferred provided the gas density and temperature profiles are known @xcite . prior to the launch of and , temperature profiles were typically sparsely sampled , at best . in such circumstances , isothermality is a convenient approximation , since the gas temperature does not vary dramatically with radius . this then implies a one - to - one relation between the gravitational potential and the density profile , and hence the surface brightness distribution , provided the abundance profile is known ( or , more usually , assumed to be flat ) . for a @xcite gravitational potential , this leads to the ubiquitous `` isothermal @xmath2-model '' @xcite . the simple analytical form of the @xmath2-model has guaranteed its longevity as a convenient ad hoc fitting function even though the underlying assumptions of the model are no longer believed to hold strictly @xcite . with the advent of and , spatially resolved spectroscopy has largely superseded wide - band surface brightness photometry as a means for measuring the mass ( although see * ? ? ? * ) , at least for high signal - to - noise ( s / n ) data @xcite . a range of techniques have evolved for transforming the spectra into mass constraints ( * ? ? ? * for a review ) , most of which first entail fitting a single - phase plasma model to spectra from different regions of sky in order to obtain binned temperature ( and , possibly , density ) profiles . this process often introduces correlations between the binned temperature or density points , especially if deprojection techniques are employed or if coarser binning is used for the temperature or abundance than the density . care should be taken to account for these , for example by using the full covariance matrix to compute @xmath3 when model - fitting downstream , rather than the common practice of just using the leading diagonal @xcite . even for gas that is strictly single phase in any infinitesimal volume , temperature or abundance variations over the spectral extraction aperture violate the single phase approximation in that bin and can lead to biases in the inferred temperature , abundance or density profiles @xcite . attempts to mitigate these issues have been made by modifying the spectral fitting procedure . for example , @xcite and @xcite fitted stacks of coarsely - binned , narrow - band images ( `` data cubes '' ) by adopting parametrized models for the temperature , abundance , and either the gas density or gravitating mass profiles . ( in the latter case , the gas density profile was then derived under the hydrostatic approximation . ) given the physical state of the gas as a function of position predicted by this model , spectra were generated in a series of shells that were , in turn , projected along the line of sight and fitted directly to the data cube . this circumvents the intermediate step of measuring the binned temperature profile . similar approaches , albeit emphasizing the simultaneous fitting of full - resolution spectra obtained from concentric annuli , were advocated by @xcite and @xcite . in objects with lower surface brightness it is often impossible to obtain sufficient photons to enable high - quality spectral analysis in as many bins as required . in these cases , it is common practice to measure coarse , global quantities such as the emission - weighted luminosity ( determined , for example , from a @xmath2-model fit ) , temperature or @xmath4 ( the product of temperature and gas mass : * ? ? ? * ) , and apply scaling relations to transform these into mass estimates @xcite . the calibration of these scaling relations is generally empirical , and , to be reliable , requires high resolution spectroscopy of objects similar to those under scrutiny . any given object can not , in practice , be guaranteed to obey these relations , and assuming this behaviour can , therefore , restrict discovery space . as a compromise between these two extremes ( global scaling relations and spatially resolved spectroscopy ) , we propose a powerful alternative method , _ narrow band photometry _ , the simultaneous fitting of radial x - ray surface brightness profiles in multiple pulse height ( pha ) bands , as an appropriate choice for modest - s / n data . this involves first adopting models for the three - dimensional gas temperature , density and abundance profiles , which are then used to derive the predicted surface brightness distribution in each band . while this has philosophical similarities to the data cube fitting of @xcite , in this paper , we demonstrate that only a limited number of energy bands are necessary to constrain the mass of the system . by emphasizing radial surface brightness photometry , the treatment of the background is also simplified . if the source and background models are constrained together , spectral based methods generally require the arbitrary parametrization of the background spectrum , as well as usually employing assumptions over its spatial variation @xcite . in our approach , we demonstrate that minimal assumptions need to be made about the spectral shape of the background . an essential ingredient of our approach is a physical model for the three dimensional distribution of the gas density and temperature . in our recent work , we have demonstrated the utility of an entropy - based model , in which the gas density and temperature profiles are inferred self - consistently in the hydrostatic approximation from a model for the ( non - gas ) gravitating mass and the gas entropy @xcite . part of the advantage of this parametrization is that it allows convective stability ( a necessary condition for hydrostatic equilibrium ) to be rigorously enforced , which imposes additional important constraints on the mass distribution @xcite . similar , entropy - based approaches have been independently explored in the context of sunyaev - zeldovich studies @xcite , and in a more non - parametric manner by @xcite . we have found that a powerlaw model ( with arbitrary , multiple breaks ) and a central plateau is sufficient to fit typical entropy profiles measured from high - quality , spatially resolved spectroscopy @xcite . when using such a model , a major source of uncertainty is the behaviour of the entropy at the largest physical scales , where the s / n of the data is lowest , which can translate into errors on the recovered global properties , such as the gas fraction @xcite . in addition to the usual powerlaw parametrization , in this paper , we therefore also explore a class of model that mitigates this uncertainty by exploiting , in a somewhat revised form , the recently discovered correlation between the gas fraction and entropy profile , which spans a wide range of halo mass @xcite . we discuss the practical implementation of narrow band photometry in x - ray analysis in [ sect_narrow_band ] . in [ sect_entropy_model ] we illustrate how this technique can be used in conjunction with the entropy - based , hydrostatic gas model to constrain the gravitating mass over a range of mass scales by employing simulations tailored to match two well - studied objects from the literature . finally , in [ sect_entropy_scaling ] , we show how additional constraints on the entropy profile lead to a new class of model that is optimal for fitting low s / n data , and reach our conclusions in [ sect_discussion ] . in this section , we discuss the practical implementation of narrow band photometry as a tool for x - ray astronomy . let us define the x - ray surface brightness ( in @xmath5 ) measured for a given object in a particular pha ( energy ) channel band ( @xmath6@xmath7 ) over some region of the detector @xmath8 @xmath9 where @xmath10 is the area of the region measured in square arc minutes . @xmath11 is the photon count rate detected in pha channel @xmath12 and region @xmath8 . assuming the source emission comes from an optically thin plasma , we can write ( see appendix [ sect_sb ] ) : @xmath13 \label{eqn_ci_definition}\end{aligned}\ ] ] where @xmath14 and @xmath15 are detector coordinates ( in arc minutes ) , @xmath16 if @xmath17 is within region @xmath8 or is 0 otherwise , @xmath18 is the solid angle , @xmath19 is the line - of - sight coordinate to the object , @xmath20 is the photon frequency in the source rest frame , @xmath21 is the ( rest frame ) count rate of photons emitted by the source per unit volume , per unit frequency range , @xmath22 is the source redshift , and @xmath23 is the likelihood that a photon of energy @xmath24 incident at angle @xmath18 will be detected in pha channel @xmath12 at detector coordinate @xmath17 . @xmath25 contains information about the telescope and detector sensitivity , point spread function and spectral redistribution function . our aim is to invert eqn [ eqn_ci_definition ] to infer @xmath21 as a function of position in the object , and thus the physical state of the emitting medium . forward fitting is a powerful way to achieve this . this entails adopting a reasonable parametrization for @xmath21 as a function of ( @xmath26 ) , and evaluating eqn [ eqn_ci_definition ] numerically . the adopted parameters are adjusted until good agreement is found with the measured @xmath27 . as we discuss below ( [ sect_hardness ] ) , it is generally insufficient to use a single photometric band @xmath6@xmath7 , and so multiple surface brightness profiles ( @xmath28 ) will need to be fitted simultaneously . to simplify the problem , spherical symmetry is generally assumed . in appendix [ sect_sb ] we outline a practical scheme for evaluating the integral efficiently in that case . directly fitting a coronal plasma model to the full resolution spectrum of the hot gas provides the most reliable means for measuring the temperature and metallicity . provided the abundance ratios of key species with respect to fe are known ( for example , if they match the solar abundance pattern ) , however , hardness ratios ( the ratio of the detected count - rates in two energy - bands ) can also be used to extract temperature and abundance information . for representative acis - s3 observations , we show in fig [ fig_hardness ] the loci in hardness - hardness space of apec thermal plasma models having different temperatures and abundances . we have carefully chosen the energy - bands under scrutiny to maximize the separation of the models , but it is clear that two hardness ratios ( three energy bands ) are sufficient to constrain the gas temperature and abundance ( and hence density ) in these observations . still , this approach requires careful calibration of the hardness ratios on an observation by observation basis , taking into account the spatial variation of the instrumental response , the galactic absorbing column and the redshift of the source . for reference , we also show the corresponding hardness ratios in a series of radial bins for two simulated datasets ( discussed in [ sect_simple_entropy_performance ] ) , illustrating that , for real data , this provides a feasible means to measure the gas properties in multiple annuli . for an optically thin coronal plasma , we expect @xmath21 to depend only on the photon frequency ( @xmath20 ) , the gas density ( @xmath29 ) , temperature ( expressed in energy units as kt ) , and abundances ( @xmath30 ) of key species ( denoted here as @xmath31 ) . as a simple illustration of the method , we therefore consider the case of such a plasma . assuming spherical symmetry and isothermal gas with a @xmath2-model density profile , @xmath32 where @xmath33 and @xmath34 are the central electron and hydrogen number densities , respectively , @xmath35 relates the gas emissivity at frequency @xmath20 to the temperature for the species @xmath31 , which has abundance @xmath30 . here @xmath36 and @xmath2 are the standard parameters of the @xmath2-model ( the `` core radius '' and @xmath2 parameter ) . in appendix [ sect_sb ] , we derive a generalized expression for the surface brightness in terms of @xmath21 . substituting eqn [ eqn_beta ] into eqn [ eqn_n_nu_j1 ] and integrating , we obtain : @xmath37 where @xmath38 \end{aligned}\ ] ] if @xmath39 . here , @xmath40 is the gamma function . if @xmath41 , then @xmath42 the integral in eqn [ eqn_beta_model ] does not depend on @xmath2 or @xmath36 , and instead only depends on the ( known ) pha bands used for surface photometry , the ( known ) annuli definitions and the ( unknown ) gas temperature . by fitting surface brightness profiles in multiple bands simultaneously , therefore , it is possible to constrain the temperature and abundance , along with @xmath2 and @xmath36 . for speed it is often convenient to adopt a linear approximation for @xmath35 ; : @xmath43 where @xmath44 is chosen such that @xmath45 . ( we note that this same approximation is used within the x - ray spectral fitting package for the computation of the apec plasma emissivity . ) this approximation means that kt can be taken outside of the integral in eqn [ eqn_beta_model ] , allowing the integrals over @xmath20 to be pre - computed for speed . to illustrate the utility of this approach , we fitted the model to realistic , simulated surface brightness profiles . we assumed the hot gas was distributed as a @xmath2-model with @xmath46 , @xmath47 kpc ( which corresponds to 3.4 , adopting a distance of 25.7 mpc ) and a central gas density of @xmath48 . these parameters were chosen to be similar to the properties of the isolated , milky way - mass elliptical galaxy ngc720 @xcite . adopting 1 kev for the gas temperature and abundances which are 0.5 times solar @xcite , we simulated an artificial events file , using the algorithm outlined in @xcite . briefly , we computed the gas emissivity in a series of ( r , z ) bins in cylindrical polar coordinates , assuming an apec thermal plasma model . folding in the effective area and spectral response function of , we then simulated a corresponding spectrum from each region for a 100 ks exposure time , and the resulting photons were projected onto the sky to build up the event list . in @xcite , we demonstrated the reliability of this method for producing artificial data . the resulting events file contains @xmath120000 photons from the hot gas within a 4 region . we added a background component with a constant surface brightness profile and a plausible spectrum ( two soft apec thermal plasma components , and a powerlaw , normalized to match the sky background in the vicinity of ngc720 ) . for simplicity , we ignored the spatial dependence of the effective area and spectral response , and ignored the spatial point spread of the telescope . we extracted surface brightness profiles in three energy bands , 0.50.9 kev , 0.92.0 kev and 2.07.0 kev , corresponding to those shown in fig [ fig_hardness ] . we fitted a model comprising the isothermal @xmath2-model described above ( eqn [ eqn_beta_model ] ) for the hot gas and a constant background component in each band . we allowed @xmath2 , @xmath36 , the central gas density ( @xmath29 ) , kt and the gas abundance to be fitted freely . fits were performed by minimizing the c - statistic , which is appropriate given the poisson distributed nature of the data @xcite using dedicated software based around the minuit software library . by visual inspection , the overall fit was good and each parameter was recovered accurately , specifically within 2-@xmath49 of the true value ( kt=@xmath50 kev , = @xmath51 , @xmath52 , @xmath53 kpc , and @xmath54 ) . the ability to constrain both gas temperature and abundance by using these three energy bands is unsurprising given the temperature of the gas ( fig [ fig_hardness ] ) . [ fig_n720_sb ] we next consider a more physically motivated model for the surface brightness profile , based on a self - consistent solution to the equation of hydrostatic equilibrium for the density and temperature profiles . if the gas is spherically distributed and in an equilibrium state ( although not necessarily static ) , we can write @xcite : @xmath55 where @xmath56 is the universal gravitational constant , @xmath57 is the radius within which the mean mass density of the system is 500 times the critical density of the universe , @xmath58 is the mass within @xmath57 , @xmath59 is the mean molecular weight ( @xmath60 for a fully ionized plasma ) , @xmath61 is the mass of hydrogen , @xmath62 is an arbitrary reference energy ( typically 1 kev ) , @xmath63 is the radius from the centre of the object divided by @xmath57 , @xmath64 is the gas mass ( divided by @xmath58 ) enclosed within radius @xmath63 , @xmath65 is the enclosed non - gas mass , @xmath66 is the `` effective mass '' codifying the bulk gas motions , including rotation ( in the notation of @xcite , @xmath67 . for pure rotation , @xmath68 is the spherically averaged rotation velocity at radius r ) . we define @xmath69 , corresponding to the usual astronomers entropy proxy , where @xmath70 is the nonthermal pressure fraction ( including the effects of magnetic pressure , cosmic ray pressure and turbulence ) , and @xmath71 where @xmath72 is the cosmological baryon fraction ( 0.17 ; @xcite ) , @xmath73 is the hubble constant in units of @xmath74 , and @xmath75 is the cosmological matter density parameter . we also define @xmath76 , which is proportional to the thermal gas pressure to the power of @xmath77 . the true values of @xmath57 and @xmath58 ( which include the gas mass ) are , technically , not known prior to solving the equations for @xmath78 . however , since the scaling factors are arbitrary , we instead adopted the values of @xmath57 and @xmath58 appropriate for the dark matter halo alone , which can be computed analytically for a ( * ? ? ? * hereafter nfw ) profile . by adopting parametrized models for @xmath79 , @xmath80 , @xmath70 , and @xmath81 , we can solve eqns [ eqn_zeta][eqn_mg ] numerically ( with a fourth order runge - kutta method ) for any given set of input parameters , and hence determine the radial density and temperature profiles . a reasonable model for @xmath65 comprises a stellar mass component ( assuming mass follows light , this can be derived by deprojecting the stellar light ; @xcite ) with adjustable m / l ratio , a central supermassive black hole , and a dark matter halo ( nfw ) . if the gas is close to hydrostatic , @xmath82 and @xmath83 , leaving it necessary only to adopt a parametrized form for @xmath79 , and provide two boundary conditions ( on @xmath64 and @xmath84 ) . we begin the integration from a small radius , @xmath85 , which is sufficiently small that we can impose the boundary condition @xmath86 , we assume there is no gas within @xmath85 . the boundary condition on @xmath84 is an adjustable parameter , which significantly affects the shape of the resulting gas temperature and density distribution . in particular , we found that the temperature profile shape can change from radially falling to radially rising ( a `` cool core '' profile ) by increasing @xmath87 . that changing the pressure boundary condition can produce such diverse behaviour is unsurprising @xcite . the outer radius of integration is arbitrary , but it should be sufficiently large that it accounts for almost all of the emission that will be projected into the line - of - sight ( see [ sect_sb ] ) . we advocate some fraction ( or multiple ) of the virial radius as the optimal choice here , although it is important to understand that this requires that the gas remains approximately spherically distributed and hydrostatic ( or , at least , accurately described by the adopted forms for @xmath70 and @xmath66 ) out to this scale . in our recent work , we have assumed that the entropy depends on radius as a ( multiply broken ) powerlaw , plus a constant ( see @xcite ) , @xmath88 and so on , adding in as many breaks as required . the parameters @xmath89 , @xmath90 , @xmath91 , @xmath92 and @xmath93 are adjustable . empirically , observed entropy profiles are often well parametrized by similar models @xcite . theoretically , powerlaw - like entropy profiles are expected from adiabatic structure formation models @xcite , which may be modified by feedback , resulting in profiles that resembles ( multiply ) broken powerlaw distributions , with a constant offset @xcite . in order to convert the measured gas density and temperature into a surface brightness profile , it is necessary to know the three dimensional radial dependence of the abundance pattern . in practice , one can adopt a simple parametrization , the free parameters of which can be fitted along with the parameters controlling the mass model and thermodynamic properties of the gas . unfortunately there is no generally accepted functional form for such a distribution , and so we recommend experimentation with different metallicity profiles as a means of exploring the sensitivity of the results to the details of this assumption . for the purpose of validating the surface brightness model as an adequate tool for fitting , it is convenient to adopt flat ( constant ) metallicity profiles in this paper . we allowed only the total ( fe ) abundance to be fitted , while fixing the abundance ratios of other metals with respect to fe to be at their solar values @xcite . we return to the problem of a radially varying abundance profile in [ sect_discuss_abundance ] . [ fig_kt_rho_n720 ] to illustrate the performance of the model , we fitted simulated data for two objects , spanning almost two orders of magnitude in mass . these objects , henceforth termed objects a and b , are tailored to resemble , respectively , the isolated , milky way - mass elliptical galaxy and the fossil group/ cluster rxj1159 + 5531 @xcite . these systems were chosen as we have previously studied them over a wide range of physical scales using the more traditional approach of spatially resolved spectroscopy . starting with the published models for each real system , and assuming = 0.5 times solar , we simulated artificial events files corresponding to a 100 ks exposure with acis - s3 . these files were generated as described in [ sect_fit_beta ] ( see also * ? ? ? * ) . to represent better real data , we included an additional component to represent the unresolved x - ray binary contribution to each system . specifically , we assumed the combined light from the lmxbs was distributed as de vaucouleurs profiles , with effective radii 25.2 and 6.4 , respectively , and a 7.3 kev bremsstrahlung spectrum . we added a background component with a constant surface brightness profile and a plausible spectrum ( two soft apec thermal plasma components , and a powerlaw , normalized to match the sky background in the vicinity of ) . this is somewhat idealized as it omits the non x - ray background that is important for data at high energies . for simplicity , we also ignored the spatial dependence of the effective area and spectral response , and ignored the spatial point spread of the telescope . we extracted surface brightness profiles in three energy bands , 0.50.9 kev , 0.92.0 kev and 2.07.0 kev , corresponding to those shown in fig [ fig_hardness ] , which were fitted simultaneously . the model comprised hot gas , a background component and a devaucouleurs component to model the unresolved point sources . the hot gas model was derived by numerically solving eqns [ eqn_zeta][eqn_mg ] , assuming @xmath94 and @xmath95 ( no nonthermal pressure ) , and modelling @xmath80 as the sum of a stellar light model ( as discussed in * ? ? ? * ; * ? ? ? * ) , an nfw dark matter halo and a central black hole . the stellar m / l ratio and the mass and concentration of the nfw model were allowed to vary in the fit , and the black hole mass was fixed at @xmath96 for and @xmath97 for rxj1159 + 5531 . the model assumed a flat abundance profile , but we allowed the overall abundance to fit freely , fixing the abundance ratios with respect to fe to their solar values . to compute the surface brightness , we evaluated the gas emissivity at a series of logarithmically spaced radial points ( from @xmath85 to 12 times the virial radius ) , and projected it using a cubic spline approximation and eqn [ eqn_projection ] . to enable direct comparison with our previous work , parameter space exploration was carried out with a bayesian monte carlo code ( version 2.7 of the multinest code : * ? ? ? we adopted the appropriate poisson likelihood function for a set of independent data points , and used uniform priors for all variable parameters , similar to our default analysis in @xcite . the model gave a good fit to both datasets ( see fig [ fig_rxj1159_sb ] ) , and the abundance was measured reasonably well for each system ( = @xmath98 , @xmath99 ) in fig [ fig_compare_derived ] , we compare various derived physical parameters obtained from our fits to the known , true value , demonstrating good overall agreement , and indicating that the hydrostatic , narrow band photometric model provides a viable means for constraining the global properties of elliptical galaxies , groups and clusters . the constraints are comparable to those obtained from our fits to the full spatially resolved spectroscopic data of and rxj1159 + 5531 @xcite ; for example we measured @xmath100 and @xmath101 in rxj1159 + 5531 to 0.05 dex and 6% , respectively , with both the surface brightness fitting approach and the full spectral analysis , if restricted only to the data @xcite . in fig [ fig_kt_rho_rxj ] , we show the derived temperature and density profiles for each object , which agree well with the true profile . for object a , the temperature at large scales is slightly over - estimated , albeit with large uncertainty . at these scales , the background is comparable to , or exceeds , the hot gas flux , at least at energies 1 kev , which limits the ability to constrain the temperature precisely from the hardness ratios . nevertheless , the global properties of the system are very well recovered . this illustrates the potential of narrow band photometry for mass measurements over a range of mass scales . [ sect_scam ] while eqn [ eqn_powerlaw ] is appealing in its simplicity and has been used successfully to model the x - ray emission from galaxies and groups @xcite , it has significant drawbacks . in particular , extrapolating the model to large radii involves ad hoc assumptions in a region where the data are often sparse , at best , or missing altogether . furthermore , as the entropy profile shape becomes more and more complex , additional adjustable parameters are required , resulting in a large number of potentially degenerate parameters . these problems are amplified if the s / n of the data is low . instead , we propose here to employ an alternative parametrization for the entropy distribution . outside the very central regions , gravity - only models of structure formation predict a `` baseline '' distribution , @xmath102 @xcite , which is distorted by non - gravitational heating , especially in the inner parts of low - mass systems @xcite . nevertheless , the principal effect of the entropy injection is to lower the gas density , rather than raise the temperature @xcite , which is supported by the universality of cluster and group temperature profiles @xcite . thus , to zeroth order , @xmath103 , where @xmath104 is the gas fraction . this scaling has been found to give a much better match to the observed entropy profiles than the baseline model when the enclosed gas fraction , @xmath105 is employed @xcite . as we show below , this model fails to match the data successfully in lower mass objects , and in the central parts of all systems ( roughly , within the cooling radius ) . we therefore begin by refining the parametrization of this scaling relation . [ cols= " < , < , < , < , < , > , > , > " , ] having eliminated @xmath79 by application of eqn [ eqn_scaled1][eqn_scaled3 ] , we solved eqn [ eqn_zeta][eqn_mg ] and computed surface brightness profiles , which we fitted to the data for and rxj1159 + 5531 , exactly as described in [ sect_simple_entropy_performance ] . in table [ table_parameters ] we summarize the fitted and fixed parameters . in fig [ fig_compare_derived ] we compare the derived properties for the simulated objects a and b with the true values , finding overall excellent agreement . the gas abundance was measured accurately ( = @xmath98 and @xmath106 , respectively ) in fig [ fig_kt_rho_rxj ] , we show the derived temperature and density profiles for both systems . overall , the fits are reasonable , but in the inner regions , we see statistically significant ( albeit not dramatic ) deviations of the best fitting model from the true profiles . nevertheless , at large scales the scam model does a good job of capturing the shapes of the profiles , explaining why the global parameters are so well recovered . these results give us confidence in the utility of the scaled adiabatic model ( scam ) to derive global physical parameters from narrow - band photometry . since the scam model imposes restrictions on the entropy profile distribution at large scales , it is ideally suited for fitting low - to - modest s / n data . to demonstrate its usefulness , we simulated data corresponding to a 5 ks snapshot of object a , corresponding to only 5% of the exposure used in our previous discussion . fitting the surface brightness profiles , we found that the global parameters were accurately recovered . while the precision of the constraints was poorer than when fitting the same model to the 100 ks exposure , the global parameters were , nevertheless , recovered to within @xmath10.1 dex or better . we have demonstrated that narrow band photometry provides an effective alternative to spatially resolved spectroscopy for determining detailed mass profiles of x - ray luminous galaxies , groups and clusters . since surface photometry is , in general , less demanding of data than spectroscopy , the narrow band photometric method presented here is ideally suited to systems with modest s / n , for which spatially resolved spectroscopy is not realistic . without the intermediate step of measuring temperature and density profiles ( from single phase plasma model fits ) , biases due to the projection of different temperature gas components along the line of sight are effectively minimized , and the difficulty of accounting for covariance between the temperature and density data points downstream is circumvented . in our fully bayesian forward - fitting analysis , a model for the surface brightness profile in each band is constructed based on a physical model for the mass distribution and the entropy profile , assuming hydrostatic equilibrium , a single phase gas and spherical symmetry , and is directly fitted to the data . although real galaxy clusters are not exactly spherical , on average the spherical approximation does not introduce significant bias into their inferred global properties , such as the virial mass , halo concentration , and gas fraction @xcite . similarly , the hydrostatic approximation , provided the system is sufficiently relaxed in its x - ray morphology , is likely to be accurate at least to the @xmath125% level ( * ? ? ? * and references therein ) . in regions where the gas is morphologically relaxed , spatially resolved spectroscopy is generally consistent with a single phase plasma , provided the temperature gradient is not strong across the aperture @xcite . similarly , the range of temperatures in the cores of clusters measured within the rgs aperture @xcite is broadly consistent with the observed central temperature gradients . limited multiphase gas has , nevertheless , been inferred in regions of strong interaction between the hot gas and ejecta from a central agn @xcite . recent work has also invoked multiphase gas as a possible explanation of the unexpected flattening of cluster entropy profiles outside @xmath1@xmath107 @xcite , although observations of relaxed clusters suggest such behaviour ( and , hence , the postulated clumping ) is not ubiquitous @xcite . while it is relatively straightforward to identify and exclude data from the disturbed cores of clusters ( where the hydrostatic and single phase approximations may break down ) , the effects of possible clumping at large scales are more troubling , due to projection effects . still , if one focuses on regions well within projected @xmath107 , the projected emission from outside @xmath107 is unlikely to contribute significantly along the line of sight , in which case the results should not depend sensitively on the behaviour of the entropy in the outskirts of the cluster . this can be tested explicitly by varying the maximum radius used during the projection calculation . application of the photometry method to the outskirts of clusters can , of course , also be used to test the need for clumping or deviations from hydrostatic equilibrium in those regions . if necessary , given a physical model for the clumping , the expression for @xmath21 can be modified to compensate and allow a useful measurement of the gravitating mass even in those regions . while the number of energy bands used is entirely arbitrary , we found that three standard bands ( 0.50.9 kev , 0.92.0 kev and 2.07.0 kev ) were sufficient to constrain the gas temperature and total abundance from data over a fairly wide range of parameter space . limiting the number of photometric bands is desirable , since it minimizes the number of adjustable parameters needed to account for the background . nevertheless , if the data permit , the number of energy bands can be expanded to focus on regions of particular interest that may help constrain the relative abundance of other species , such as o , ne , mg and si . we present two classes of hydrostatic model that enable the gravitating mass to be inferred . both involve a physical model for the ( non - gas ) gravitating mass ( comprising dark matter halo , stars and central black hole ) , and solving the hydrostatic equations ( eqn [ eqn_zeta][eqn_mg ] ) by imposing a constraint on the entropy distribution . the performance of the former approach , which entails adopting an arbitrary parametrization for the entropy profile , has been demonstrated in our previous , spatially resolved spectrocopic fits to x - ray data @xcite , and , independently , in fits to sunyaev - zeldovich data @xcite . combined with narrow - band photometry , we have demonstrated that precise , accurate constraints can be obtained on the global properties of objects over a wide mass range . the latter entropy parametrization is more suitable for lower s / n data , since it involves exploiting an empirical scaling relation between the gas fraction and entropy profile outside the cooling core , effectively eliminating the arbitrariness in the extrapolation of the entropy profile to large scales . this scaling relation is unlikely to be exact in any given system , which introduces systematic errors into the recovered temperature and density profiles , especially in the system cores . however , with the extra constraints at large scale , we found that the reduction in the error - space did not give rise to appreciable biases in the derived global system properties . the utility of this method is illustrated by the fact that , even for a shallow , snapshot observation of a realistic low - mass object , we were still able to recover the global parameters to within @xmath10.1 dex . for simplicity , our simulations were for systems with a constant metallicity profile . in reality morphologically relaxed systems tend to have centrally - peaked metallicity distributions @xcite and so , if the properties in the centre of the system are important it is necessary to modify the model to incorporate such a gradient . as there is no _ de facto _ standard model for such purposes , we recommend a simple ansatz , such as : @xmath108 where @xmath109 and @xmath110 , @xmath111 and @xmath112 are adjustable fit parameters , and we initially assume that the ratio of the abundance of the other metals to fe does not vary with radius . to verify that adopting such a parametrization does not appreciably degrade the constraints on the global system parameters , we re - simulated object b , assuming eqn [ eqn_abund_gradient ] . we set @xmath113 , @xmath114 , @xmath115 [ kpc ] and @xmath116 . we fitted the surface brightness profiles , using the scam model ( this time incorporating eqn [ eqn_abund_gradient ] ) and allowing the parameters @xmath109 , @xmath111 and @xmath112 to be fitted . we summarize the constraints on the global parameters in fig [ fig_compare_derived ] , demonstrating that incorporating an abundance gradient into our fits does not significantly degrade the level of constraints we can obtain . furthermore , with this approach , we were also able to constrain the abundance profile reliably ( we measured @xmath117 , @xmath118 [ kpc ] and @xmath119 ) . while we have focused on the use of the scam model in the context of narrow - band x - ray photometry , it actually has broader implications . in particular , it is a potentially powerful tool for constraining the global properties of galaxy clusters from the sunyaev - zeldovich ( s - z ) effect ( * ? ? ? * for a review ) . the key observable is the compton y - parameter , @xmath120 where @xmath121 is the thompson scattering cross - section , @xmath122 is the mass of the electron , @xmath123 is the speed of light , @xmath124 is the electron number density , and the integral is along the line - of - sight ( los ) . in comparison to x - ray emission , for which the emissivity depends on the density _ squared _ , the y - parameter is more sensitive to the properties of the gas at large radii . in order to derive the physical properties of a system from the inferred y - parameter , it is necessary to construct a model for the spatial variation in both density and temperature . models employing _ ad hoc _ parametrization for the thermodynamical state of the gas , for example the isothermal @xmath2-model , or even the simple entropy - based model discussed in [ sect_entropy_model ] ( for the application of this kind of model to s - z data , see * ? ? ? * ) , can not be guaranteed to capture the true behaviour of the gas when extrapolated to large scales . on the other hand , the thermodynamic behaviour of the gas at large radii predicted by the scam model is determined entirely by the underlying scaling relation , and it is clear from fig [ fig_kt_rho_rxj ] that it is sufficiently accurate to model morphologically relaxed objects with very different mass distributions , density and temperature profiles . while it remains to be seen whether there exist clusters for which the empirical scam scaling does not ( at least approximately ) hold , the well - determined asymptotic behaviour of the model is very desirable in s - z studies , and so we strongly advocate its use . we are grateful to christina topchyan for helping extensively test the models . we thank wenhao liu for providing his temperature and density profiles . pjh and dab acknowledge partial support from nasa under grant nnx10ad07 g , issued through the office of space science astrophysics data program . pjh also acknowledges support through a gary mccue fellowship , offered through uc irvine . the distance along the line of sight to the parcel is denoted as @xmath19 , and the inner and outer radii of an arbitrary shell of gas ( shown in red ) correspond to incidence angles @xmath125 and @xmath126 , respectively . [ fig_projection],width=336 ] consider a small parcel of gas of volume dv , which is emitting a total ( rest frame ) photon flux ( photons s@xmath127 ) of @xmath128 at frequency @xmath20 . let the system be spherically symmetric ( see fig [ fig_projection ] ) , so that @xmath21 is a function of distance , r , from the centre of the system . the total photon flux incident on a telescope aperture of area a placed at angular size distance @xmath129 from the gas parcel will be : @xmath130 where @xmath22 is the observed redshift of the gas parcel , @xmath131 is the _ observed _ photon frequency , @xmath132 is the angular size subtended by the gas parcel at the telescope , and @xmath19 is the line of sight coordinate in the rest frame of the gas parcel , so that @xmath133 . the arriving photons pass through the telescope optics and are ( possibly ) detected in a focal plane detector . the detected photon flux at position ( x , y ) in `` pulse height channel '' ( energy bin ) @xmath12 is given by : @xmath134 where @xmath135 codifies the telescope spatial point spread function at frequency @xmath136 for photons arriving with incident angle @xmath137 and azimuthal angle @xmath138 ( it represents the likelihood that such a photon would be detected at position ( x , y ) ) , @xmath139 is the combined detection efficiency of the optics and detector , and @xmath140 codifies the spectral redistribution function , indicating the likelihood that a photon with frequency @xmath136 that is detected at position ( x , y ) will be assigned to pulse height channel i. to make further progress , we divide the sky into a series of concentric annuli , the centres of which coincide with the projected centre of the object . annulus @xmath141 has inner and outer radii @xmath142 and @xmath143 , respectively . now , let us compute the total counts @xmath144 detected in pulse height channel @xmath12 within some ( arbitrary ) region of the detector , defined by the function @xmath145 , which is 1 in the region , or 0 elsewhere . it follows that : @xmath146\end{aligned}\ ] ] where x and y are coordinates on the detector . now , let annulus @xmath141 be sufficiently narrow that @xmath147 and @xmath148 , so that we can rewrite @xmath149 \nonumber \\ \times \bigg [ \int_{\theta_j}^{\theta_{j+1}}\sin\theta d\theta \int_{-\infty}^{\infty } \frac{\epsilon_{\nu^\prime}(r)}{4\pi(1+z)^3}dz \bigg]\end{aligned}\ ] ] it is convenient to rewrite this as : @xmath150 \times \bigg [ n_\nu^j \bigg]\end{aligned}\ ] ] where the quantity @xmath151 is the spectral redistribution function , @xmath152 is the spatial `` point response function '' linking sky annulus @xmath141 to detector region @xmath8 , @xmath153 is the `` effective area '' linking annulus @xmath141 to detector region @xmath8 , and @xmath154 changing the order of integration , this becomes : @xmath155 where @xmath156 is introduced due to the change of the order of integration and the requirement that @xmath157 , and @xmath158 , and @xmath159 , applying the small angle approximation @xmath160 and integrating . now , let us define the x - ray surface brightness in a particular pha channel band ( @xmath6@xmath7 ) , @xmath161 where @xmath162 is the area of detector region @xmath8 . provided the pha channel range @xmath6@xmath7 is sufficiently narrow , and the redistribution function @xmath163 is not too broad , it is often sufficient to approximate @xmath164 , where @xmath165 is the mean frequency of photons that would be detected in pha bins @xmath6@xmath7 . substituting in for @xmath166 from eqn [ eqn_n_nu_j ] , we get : @xmath167 to evaluate eqn [ eqn_sigma ] efficiently , it is convenient to approximate the @xmath136 integral as a piecewise polynomial function of r ( for example , a cubic spline function ) , @xmath168 where @xmath169 if @xmath170 , and @xmath171 otherwise , @xmath172 are a series of arbitrarily spaced reference points where the integral is explicitly evaluated , and @xmath173 are a set of coefficients that are chosen to ensure that the approximation is exact at each reference point . in general , @xmath174 will be a function of the gas density and temperature at radius @xmath172 , which can be computed from eqn [ eqn_zeta][eqn_mg ] , as well as the gas abundance , for which one can adopt an arbitrary parametrization . in practice , the integral is generally computed approximately at each radius @xmath172 ( for example , within a spectral fitting package such as ): @xmath175 where @xmath176 , @xmath177 are a set of reference frequencies and @xmath178 is an appropriate averaged frequency between @xmath179 and @xmath180 . the integrals can be evaluated analytically . if we define @xmath184 we note that @xmath185 to compute the remaining terms , we integrate eqn [ eqn_jn ] by parts and rearrange it to give the recurrence relation @xmath186 thus , the computation of @xmath27 reduces to the sum of a finite series .	we explore the utility of narrow band x - ray surface photometry as a tool for making fully bayesian , hydrostatic mass measurements of clusters of galaxies , groups and early - type galaxies . we demonstrate that it is sufficient to measure the surface photometry with the x - ray observatory in only three ( rest frame ) bands ( 0.50.9 kev , 0.92.0 kev and 2.07.0 kev ) in order to constrain the temperature , density and abundance of the hot interstellar medium ( ism ) . adopting parametrized models for the mass distribution and radial entropy profile and assuming spherical symmetry , we show that the constraints on the mass and thermodynamic properties of the ism that are obtained by fitting data from all three bands simultaneously are comparable to those obtained by fitting similar models to the temperature and density profiles derived from spatially resolved spectroscopy , as is typically done . we demonstrate that the constraints can be significantly tightened when exploiting a recently derived , empirical relationship between the gas fraction and the entropy profile at large scales , eliminating arbitrary extrapolations at large radii . this `` scaled adiabatic model '' ( scam ) is well suited to modest signal - to - noise data , and we show that accurate , precise measurements of the global system properties are inferred when employing it to fit data from even very shallow , snapshot x - ray observations . the well - defined asymptotic behaviour of the model also makes it ideally suited for use in sunyaev - zeldovich studies of galaxy clusters . xrays : galaxies galaxies : elliptical and lenticular , cd galaxies : ism dark matter methods : data analysis
it is well know that the ` hot big bang ' model @xcite , despite its numerous successes , is plagued by a number of ` initial conditions ' problems , of which the horizon , flatness and unwanted relic ones are the best known . the standard way to solve them is to invoke an epoch of cosmological inflation @xcite , a relatively brief period of exponential ( or quasi - exponential ) cosmological expansion . the way inflation solves these problems is , loosely speaking , by erasing all traces of earlier epochs and re - setting the universe to a rather simple state . indeed , inflation is so efficient in this task that a number of people have wondered if one can ever hope to probe the physics of a pre - inflationary epoch . there are , however , a small number of possible pre - inflationary relics . for example , the recent work by turok and collaborators @xcite shows that curvature can , in some sense , survive inflation . another class of inflationary survivors are topological defects @xcite , formed at phase transitions either before or during inflation @xcite ) . it is known ( see , e.g. @xcite ) that one needs about 20 e - foldings of inflation to solve the monopole problem . one can reverse the argument and say that monopoles can survive about 20 e - foldings of inflation . the inflationary epoch itself will obviously push the monopoles outside the horizon , but the subsequent evolution of the universe tends to make them come back inside , so if the inflationary epoch is not too long they can still have important cosmological consequences . cosmic strings are even more successful , being able to survive about 50 e - foldings . the reason for this difference is that their non - trivial dynamics @xcite makes them come back inside the horizon faster than one might naively have expected . the above two numbers are typical , but there are specific models where defects can survive even longer . one example is that of open inflation scenarios @xcite . in this case the universe undergoes two different inflationary epochs roughly speaking , a period of ` old inflation ' followed by one of ` new inflation ' . as pointed out by vilenkin @xcite , one can expect that defects will form between the two inflationary epochs . in this case , a collaboration @xcite including the present authors has recently shown that not only will cosmic strings survive the entire second inflationary epoch , regardless of how long it lasts , but they will in fact be back inside the horizon by the time of equal matter and radiation densities . in such models , monopoles can survive up to about 30 e - foldings . now , given that defects seem to be so successful surviving inflation , and that one expects them to be frozen out while they are outside the horizon , one can think of a further interesting possibility . for the best - studied case of cosmic strings , it is well known that the scaling properties of the network depend on the background cosmology @xcite . moreover , in some cases ( typically when their evolution is friction - dominated ) they can retain a ` memory ' of the initial conditions , or the general properties of the cosmology in which they find themselves at early times , for quite a large number of orders of magnitude in time @xcite . it is therefore conceivable that if such an imprint of an early cosmological epoch is retained by a defect network which manages to survive inflation , we might still be able to observe it today . we believe that this is a general feature of defect models , and a number of non - trivial pieces of information about the very early universe can probably be preserved in this way . in the present paper we will restrict ourselves to a simple example . we will discuss the possibility of a domain wall network retaining information about an early anisotropic phase of the universe . there are very strong constraints @xcite on the mass of domain walls formed after inflation , due to the fact that their density decays more slowly than the radiation and matter densities . however , these can be evaded by walls forming before or during inflation . in a subsequent paper , we will discuss the more interesting , but also more complicated , case of cosmic strings . the plan of the paper is as follows . in section [ evequations ] we briefly describe our background ( bianchi i ) cosmology and the basic evolutionary properties of the domain walls . in particular , we focus on the approach to isotropy during inflation , which is discussed through both analytic arguments and numerical simulations . we emphasise that these simulations _ do not _ include the defects . however , they serve an important purpose , as they are used in the subsequent discussion to show that the timescale needed for isotropization is compatible with the ` survival ' on anisotropic defect networks . we provide a description of our numerical simulations of domain wall evolution in section [ nums ] . these are analogous to those of press , ryden & spergel @xcite , and the interested reader is referred to this paper for a more detailed discussion of some relevant numerical issues . here defect networks are evolved in an _ isotropic _ , matter - dominated ( ie , post - inflationary universe ) , and their main purpose is to show that isotropic and anisotropic networks will evolve in different ways , so two such networks can in principle be observationally distinguished as they re - enter the horizon . our main results are presented and discussed in section [ redi ] , and finally we present our conclusions and discuss future work in section [ conc ] . throughout this paper we will use fundamental units in which @xmath0 . we consider the evolution of a network of domain walls in a k=0 anisotropic universe of bianchi type i with line element @xcite : @xmath1 where @xmath2,@xmath3 and @xmath4 are the cosmological expansion factors in the @xmath5 , @xmath6 and @xmath7 directions respectively , and @xmath8 is the physical time . the dynamics of a scalar filed @xmath9 is determined by the lagrangian density , @xmath10 where we will take @xmath11 to be the generic @xmath12 potential with two degenerate minima given by @xmath13 this obviously admits domain wall solutions @xcite . by varying the action @xmath14 with respect to @xmath9 we obtain the field equation of motion : @xmath15 where @xmath16 with @xmath17 and @xmath18 . the dynamics of the universe is described by the einstein field equations . here we shall seek perfect fluid solutions . the time component of the einstein equation then becomes @xmath19 while the spatial components give @xmath20 with @xmath21 , @xmath22 and @xmath23 , @xmath24 and @xmath25 and @xmath26 . it is straightforward to combine equations ( [ einstein0],[einsteini ] ) to obtain : @xmath27 in the following discussion we will make the simplification that @xmath28 ( and therefore @xmath29 ) and consider the dynamics of the universe during an inflationary phase with @xmath30 . in this case @xmath31 and the einstein field equations ( [ einstein0],[einsteini],[einconst ] ) imply : @xmath32 while @xmath33 can be found from the suggestive relation @xmath34 equation ( [ dyna ] ) has two solutions , depending on the initial conditions . if @xmath35 , then @xmath36 is the smaller of the two dimensions and the shape of spatial hyper - surfaces is similar to that of a rugby ball . then the solution is @xmath37 , \label{soldynasmall}\ ] ] with @xmath38 . on the other hand , if @xmath39 , then @xmath36 is the larger of the dimensions and the shape of spatial hyper - surfaces is similar to that of a pumpkin . then the solution is @xmath40\ , . \label{soldynalarge}\ ] ] note that in both cases the ratio @xmath41 tends to unity exponentially fast , and hence the same happens with the ratio @xmath42 . in other words , inflation tends to make the universe more isotropic , as expected . an easy way to see this is to consider the ratio of the two different dimensions , @xmath43 , and to study its evolution equation . one easily finds @xmath44 which has an obvious attractor at @xmath45 . note that even though we have so far assumed ( for simplicity ) that @xmath46 , the same analysis can be carried out for an inflating universe with @xmath47 with @xmath48 by numerically solving the conservation equation @xmath49 together with equations ( [ einsteini ] ) and ( [ einconst ] ) . indeed , the more general case will be relevant for what follows . in figure [ figure1 ] we plot the evolution of the asymmetry parameter @xmath50 , according to eqns . ( [ einstein0 ] ) and ( [ einsteini ] ) , for several values of @xmath51 assuming @xmath52 and @xmath53 ( note that @xmath54 is the maximum value of @xmath55 which violates the strong energy condition ) . note that we do not include the defect network in the simulation . ( we assume that the network at the initial time @xmath56 is statistically isotropic . ) we take @xmath57 . we can see that depending on the initial degree of anisotropy , specified by @xmath58 , the value of @xmath59 can grow to be very large , especially if @xmath58 is large . moreover , although for @xmath52 , the value of @xmath59 becomes approximately constant in one hubble time that does not happen so rapidly for inflating universes with larger @xmath55 . this removes the necessity of producing the domain walls right at the onset of the inflationary era . + -0.5 in 0.5 in [ figure1 ] what about the evolution of the domain walls ? based on rather general grounds , we expect it to have a number of similarities with the much better studied case of cosmic strings @xcite . in particular , one can define a ` characteristic length - scale ' , which we shall denote by @xmath60 , that can be roughly interpreted as a typical curvature radius or a correlation length of the wall network . it is also a length - scale that measures the total energy of the domain wall network per unit volume , since we can define @xmath61 where @xmath62 is the domain wall energy per unit area . note that in a more rigorous treatment that allowed for the expected build - up of small - scale ` wiggles ' on the walls ( in analogy with what happens for the case of cosmic strings @xcite ) each of these three length scales would be different . however , for our present purposes it is adequate to suppose that they are all similar . then we can expect to find two different evolution regimes . while the network is non - relativistic , we expect it to be conformally stretched by the cosmological expansion , and hence @xmath63 in this case there is essentially no dynamics . an extreme example of this regime happens during inflation we can see from eqn . ( [ dynamics ] ) that due to the very rapid expansion which occurs in the inflationary regime the time derivatives of the field @xmath9 rapidly approach zero so that the network of domain walls will simply be frozen in comoving coordinates . on the other hand , once the network becomes relativistic , one expects it to evolve in a linear scaling regime where @xmath64 this is the case of ` maximal ' dynamics , in the sense that the network is evolving ( in particular , losing energy by wall collisions and re - connections ) as fast as allowed by causality . we note that previous work of press , ryder and spergel @xcite suggests that there may be logarithmic corrections to this linear regime . at late times ( after the inflationary epoch ) the universe is homogeneous and isotropic with @xmath65 with the average dynamics of the universe being specified by the evolution of the scale - factor @xmath66 . we now consider the evolution of isotropic and anisotropic defect networks in this background . in particular , we are interested in determining how the networks evolve as they re - enter the horizon , since if one finds differences in the dynamics of the two cases then this should translate into observational tests that will allow us to discriminate between then and hence probe pre - inflationary physics . it is useful for numerical purposes to re - write equation ( [ dynamics ] ) as a function of the conformal time @xmath67 defined by @xmath68 . in this case equation ( [ dynamics ] ) becomes @xmath69 with @xmath70 when making numerical simulations of the evolution of domain wall networks ( or indeed other defects ) it is also often convenient to modify the equation of motion for the scalar field @xmath9 in such a way that the comoving thickness of the walls is fixed in comoving coordinates . this is known as the prs algorithm @xcite , and it is generally believed not to significantly affect the large - scale dynamics of domain walls . we note , however , that recent high - resolution simulations @xcite have revealed that the accuracy of this algorithm is not as good as has been claimed . this effect is expected to increase with increasing dynamic range . in particular , the prs algorithm artificially prevents the build - up of small - scale features on the domain walls ( or , for that matter , any other defect ) . this turns out to be crucial for a quantitatively accurate description of their evolution , and hence for a reliable analysis of their observational consequences . for our purposes in the present work , however , the prs algorithm is enough as an approximation to the true wall dynamics . in a subsequent , more detailed publication we shall compare results obtained using this algorithm with those from the true wall dynamics . having clarified this point , we will modify the evolution equation for the scalar field @xmath9 in the isotropic phase according to the prs prescription : @xmath71 where @xmath72 and @xmath73 are constants . we choose @xmath74 in order for the walls to have constant comoving thickness and @xmath75 by requiring that the momentum conservation law for how a wall slows down in an expanding universe is maintained @xcite . we perform two - dimensional simulations of domain wall evolution for which @xmath76 . these have the advantage of allowing a larger dynamic range and better resolution than tree - dimensional simulations . we solve equation ( [ dynamics2 ] ) numerically assuming a matter - dominated einstein - de - sitter cosmology with @xmath77 . we used a standard difference scheme second - order accurate in space and time and periodic boundary conditions ( see @xcite for a more detailed description of the algorithm and other related numerical issues ) . the initial properties of the network of domain walls depend strongly on the details of the phase transition which originated them . it is conceivable that the initial network is already formed asymmetric with the walls being elongated along preferred directions . however , this is beyond the scope of the present paper . for our present purposes , we can ignore this possibility and assume that the initial domain wall network is statistically isotropic . this assumption will not modify the conclusions of the paper if anything , any _ ab initio _ anisotropies would only enhance the effects we are describing . hence , we assume the initial value of @xmath9 to be a random variable between @xmath78 and @xmath79 and the initial value of @xmath80 to be equal to zero everywhere . we normalise the numerical simulations so that @xmath81 . we set the conformal time at the start of the simulation and the comoving spacing between the mesh points to be respectively @xmath82 and @xmath83 . the wall thickness , defined by @xmath84 is set to be equal to @xmath85 . the kinetic energy of the field @xmath9 is calculated by @xmath86 on the other hand , the rest energy of the walls is calculated by multiplying the comoving area of the walls , a , by the energy density per comoving area , which can be written as @xmath87 with @xmath88 , and @xmath89 is the value of the physical velocity of the domain walls . finally the total area of the walls is defined as the area of the surfaces on which @xmath90 and is computed using the method described in ref . as pointed out above , it will be of fundamental importance to study the dynamics of the wall network at late times , as the hubble length becomes larger than the typical size of the major axis of a domain wall . a crucial issue will be the timescale required for the wall network to switch from the non - relativistic regime to the relativistic one . for our present purposes , the main difference between these two regimes is that a friction - dominated network can remain anisotropic if it froze out that way , whereas a relativistic network will rapidly become isotropic and erase any imprints from the earlier anisotropic phase . in our simulations we ignore the possibility that the network can be friction dominated due to particle scattering @xcite when the domain walls come back inside the horizon again , this would only enhance the effects we are describing . we consider three simulations with different initial conditions . in the first one ( case i ) we evolve the initial network generated in the manner specified in the previous section from the conformal time @xmath82 . in the second one ( case ii ) the initial conditions at the time @xmath82 were specified by the network configuration of the previous simulation at the conformal time @xmath91 , with the velocities reset to zero . physically , this corresponds to starting with the network outside the horizon . finally , the case iii is similar to the second one but with the initial network of case ii stretched in the @xmath6 direction by a factor of @xmath92 ( see fig . [ fig_boxes ] ) , and corresponds to the anisotropic case . we have performed @xmath93 simulations for each of the three cases , plus an additional @xmath94 run of case i , in order to test for possible box effects . + 0.5 in [ figure3 ] for each run we plot ( see fig . [ figure3 ] ) the ratios @xmath95 and @xmath96 ( note that @xmath36 and @xmath97 are the _ comoving _ area and volume , respectively ) , as well as the ratio of the kinetic and rest energies , as in press , ryden and spergel @xcite . these are plotted from the beginning of the simulation until the time when the horizon becomes one half ( for the @xmath93 runs ) or one quarter ( for the @xmath94 run ) of the box size . in addition to these ( which we plot mainly for the purposes of comparison with previous work @xcite ) we plot a ` scaling coefficient ' which will be our main analysis tool . we will define it by analogy with the cosmic string case @xcite , as follows . assume that @xmath98 then what we plot is the ` instantaneous ' or ` effective ' value of @xmath99 as a function of conformal time . for a given @xmath99 , the _ physical _ network correlation @xmath60 length will be evolving as @xmath100 note that @xmath99 can , in general , be a time - dependent quantity . however , for the two scaling regimes discussed above , we expect it to be a constant , namely @xmath101 in the non - relativistic limit where the network is being conformally stretched , and @xmath102 in the linear scaling regime . from these it is trivial to deduce the behaviour of @xmath95 and @xmath96 in both scaling regime . one expects @xmath103 in the non - relativistc regime and @xmath104 in the linear scaling regime . similarly , the ratio @xmath105 should be a constant in the linear scaling regime ( with its numerical value providing a measure of the characteristic network scaling speed ) , and it should approach zero in the non - relativistic limit . firstly , we note that the two case i runs produce very similar results : significant differences can only be seen at late times . this is an indication that the resolution we are using is adequate for our present purposes . as expected , the network in case i becomes relativistic very quickly , while those of cases ii and iii start in the extreme non - relativistic regime and only evolve away from it fairly slowly , after they re - enter the horizon . more importantly , there are two non - trivial observations to be made . firstly , we confirm that there is a correction to the linear scaling regime . we find @xmath106 which corresponds to to evolve in a linear scaling regime where @xmath107 in agreement with the previous result by press , ryden and spergel @xcite . this means that the network is not straightening out as fast as allowed by causality . secondly , the rates at which the networks in cases ii and iii approach the relativistic regime are different . one might expect this on physical grounds : if the network is stretched in one direction , then there are in fact different ` network correlation lengths ' for each direction , and interactions between the domain walls will tend to occur faster along the directions with smaller correlation lengths , and more slowly in the others . another way of saying this is that the network will only start evolving towards the relativistic regime when its larger axis has re - entered the horizon . note that this mechanism also tends to make the domain wall network more isotropic . so one can naively say that the approach to the linear scaling regime takes longer in an anisotropic universe because the dynamics of the walls must accomplish two tasks ( make the wall network relativistic and isotropic ) rather than just one . in this paper we have discussed a simple example of what we believe to be a rather generic feature of topological defect models , namely that they can easily retain information about the properties of the very early universe . this information is encoded in the scaling ( ie , ` macroscopic ' ) and statistical ( ie , ` microscopic ' ) properties of the defect networks . this is even more relevant given the fact that defects can survive significant amounts of inflation . hence , they can provide a unique probe of the pre - inflationary universe . the two crucial scales in the problem are the defect mass scale and the epoch when the defects come back inside the horizon . specifically , we have discussed the role of domain walls . we have highlighted the existence of two scaling regimes for the domain wall network , in agreement with previous work @xcite . furthermore , we have shown that an anisotropic network re - entering the horizon will take longer to approach scaling than an isotropic one . hence , if the very early universe had an anisotropic phase which was erased by an inflationary epoch , and if domain walls are present , then the walls can retain an imprint of the earlier phase , and this can have important observational consequences , eg for structure formation scenarios . as is well known , there are quite strong constraints @xcite on the mass of domain walls formed after inflation . these are basically due to the fact that their density will decay more slowly than the radiation and matter densities . however , essentially all of these can be evaded ( or at least significantly relaxed ) by walls forming before or during inflation ( and also by walls evolving in a friction - dominated regime ) . having said this , how could these anisotropies be detected ? the most naive answer would be through their imprint on cmb , but this is only true if their energy density is not too low , and such models are constrained in a variety of other ways ( not only from the cosmology side , but also from the high - energy physics side ) . the case of ` light ' walls is therefore more interesting : note that just like in the case of ` light strings ' @xcite , these are expected to be friction - dominated throughout most of the cosmic history . here the observational detection of the effects we have described becomes somewhat non - trivial . the best way of doing it should be through observations of numbers of objects as a function of redshift in different directions ( assuming that one has a reliable understanding of other possible evolutionary effects ) . two specific examples would be large - scale velocity flows @xcite and gravitational lensing statistics of extragalactic surveys @xcite . finally , there is also an important implication of our work if at least one of the minima of the scalar field potential has a non - zero energy density , which is an anisotropic non - zero vacuum density . in a subsequent , more detailed publication , we shall discuss this scenario in more detail , as well as the analogous one for cosmic strings . to conclude , we have shown that the importance of topological defects as a probe of cosmological physics goes well beyond structure formation . even if defects turn out to be unimportant for structure formation they can still ( if detected ) provide us with extremely valuable information about the physical conditions of the very early universe . kolb & m.s . turner , _ the early universe _ , ( addison - wesley , 1994 ) . guth , _ phys . rev . _ * d23 * , 347 ( 1981 ) . linde , _ particle physics and inflationary cosmology _ ( harwood , chur , switzerland , 1990 ) . a.r . liddle & d.h . lyth , _ phys . rep . _ * 231 * , 1 ( 1993 ) ; d.h . lyth & a. riotto _ phys . rep . _ * 314 * , 1 ( 1999 ) . s. gratton , t. hertog & n.g . turok , _ astro - ph/9907212 _ ( 1999 ) . a. vilenkin & e. p. s. shellard , _ cosmic strings and other topological defects _ , ( cambridge university press : cambridge , 1994 ) . q. shafi & a. vilenkin , _ phys . rev . lett . _ * 52 * , 691 ( 1984 ) ; j. yokoyama , _ phys . _ * 63 * , 712 ( 1989 ) . hodges & j.r . primack , _ phys . _ * d43 * , 3155 ( 1991 ) . avelino , r.r . caldwell & c.j.a.p . martins , _ phys . rev . _ * d59 * , 123509 ( 1999 ) . martins & e.p.s . shellard , phys . rev . * d53 * , 575 ( 1996 ) ; c.j.a.p . martins & e.p.s . shellard , phys . rev . * d54 * , 2535 ( 1996 ) . avelino , r.r . caldwell & c.j.a.p . martins , phys . rev . * d56 * , 4568 ( 1997 ) . martins , _ quantitative string evolution _ , ph.d . thesis , university of cambridge ( 1997 ) . m. bucher , a.s . goldhaber & n.g . turok , _ phys . rev . _ * d52 * , 3314 ( 1995 ) . a. vilenkin , _ phys . rev . _ * d56 * , 3258 ( 1997 ) . zeldovich , i. kobzarev & l.b . soviet phys . jetp _ , * 40 * , 1 ( 1975 ) . w.h . press , ryden & d.n . spergel , _ ap . j. _ * 347 * , 590 ( 1994 ) . m. carmeli , ch . charach & s. malin , _ phys . rep . _ * 76 * , 79 ( 1981 ) ; j. wainwright & g.f.r . ellis , _ dynamical systems in cosmology _ ( cambridge university press , 1997 ) . shellard , _ private communication _ ( 1999 ) . i. zehavi & a. dekel , astro - ph/9904221 ( 1999 ) . r. quast & p. helbig , _ astron . astroph . _ * 344 * , 721 ( 1999 ) .	we consider the evolution of domain walls produced during an anisotropic phase in the very early universe , showing that the resulting network can be very anisotropic . if the domain walls are produced during an inflationary era , the network will soon freeze out in comoving coordinates retaining the imprints of the anisotropic regime , even though inflation makes the universe isotropic . only at late times , when the typical size of the major axis of the domain walls becomes smaller than the hubble radius , does the network evolve rapidly towards isotropy . hence , we may hope to see imprints of the anisotropic era if by today the typical size of the major axis of the domain walls is of the order of the hubble radius , or if the walls re - entered it only very recently . depending on the mass scale of the domain walls , there is also the possibility that they re - enter at earlier times , but their evolution remains friction - dominated until recently , in which case the signatures of the anisotropic era will be much better preserved . these effects are expected to occur in generic topological defect models . # 1#2 epsf
traffic sign recognition system play a important role in autonomous driving environment and for advanced driver assistance systems . for driving safely and avoiding accident in daily life s traffic signs are very important . a smart system that can help human driver by automatic recognition and classification and giving warning will ease driver task . with the advancement of modern vehicles design and safety this have got a considerable attention . for a human driver because of fatigue , divergence of attention , occlusion of sign due to road obstruction and natural scenes , related to different problems may lead to miss some important traffic sign , which may result in severe accident . also automatic recognition system will help in proper navigation and to follow traffic rules . there are several challenges involved in developing a complete traffic sign recognition system , because of occlusion of signs due to different background , specifically trees , building etc . old damaged signs , weather condition , viewpoint variation , also illuminant changes , in day and night etc . a complete traffic sign system consist of detection and classification a wide variety of signs . classification of different class signs having same shape is always a difficult part , since there are very small difference in between lots of traffic signs . in this work we will present a novel classification techniques based on probabilistic latent semantic analysis . also we have built a shape classification system based of pyramid of hog features and template matching . for feature representation of final plsa system well known sift is used . previously different method which have got state of art result on gtsrb and gtsdb are very complex to train and need supervision . method related to convolutional neural network , neural network , support vector machine are computationally complex and need effort to implement in system like fpga device or other low computationally powerful devices . in our method related to plsa is computationally flexible than previous method , and it is an unsupervised method and use knn classifier at final step . we will show our result on publicly available german traffic sign ( gtsrb ) database . we have an accuracy of around 96.86 - 100% over sub category . rest of the paper is organised as follows , in section 2 we have reviewed existence literature in this area , section 3 describes our main algorithm and section 4 gives detail about dataset and experimental result along with comparison with other existing method . considerable research work exist on detection and classification traffic signs for addressing the challenged involved in real life problems . even though it is not possible go through all this research works in this paper we will give brief overview of some relevant works . most of the work used computer vision and machine learning based techniques by using data from several camera sensors mounted on car roof at different angles . in some of the work researchers explores detection based on colour features , such as converting the colour space from rgb to hsv etc . and then using colour thresholding method for detection and classification by well - known support vector machines . in colour thresholding approach morphological operation like connected component analysis was done for accurate location . colour , shape , motion information and haar wavelet based features was used in this work [ 12 ] . by using svm based colour classification on a block of pixels le et all [ 13 ] addressed the problems of weather variation . features like , sift , hog and haar wavelet etc . was used by some of this work . in german traffic sign recognition benchmark ( gtsrb ) competition , top performing algorithm exceeds best human classification accuracy . by using committee of neural networks [ 3 ] achieved highest ever performance of 99.46% , where best human performance was 98.84% . multiscale convolutional network [ 15 ] has achieved 98.31% accuracy in this dataset . also other algorithm based on k - d trees and random forest [ 16 ] and lda on hog1 [ 2 ] have got very high accuracy . for recognition purpose group sparse coding was used by [ 14 ] for rich feature learning of traffic signs recognition . the aim of the works is to develop a topic based classification frameworks for fast and reliable traffic sign categorization . discovering hidden topics in images and correlating them with similar topics images form a successful classification algorithm . since each of traffic sign category can be assumed very well as combination of one or more topics , so we have choose this method over other learning based method . for classification of the images we have used two step method , in first processing we will classify the shape of the traffic signs and after that we will classify its actual class . image may have undergone through different type of rotational effect due to viewpoint variation , alignment . in fig . [ 1 ] main idea concept of our method is depicted . as a prepossessing task we will use affine invariant transform on the images for getting rid of rotational effect . in this step traffic sign we will be divided into six class , specifically tringle , square , circle , single - circle , rectangle and hexagon as shown in fig . [ 2 ] . dividing the images in terms of shape help in classification since traffic sign topic most of the time depends on their shape . different shape class images associated with different levels of dangers in road . getting correct shape of the images is a boost for further classification . for shape representation hog [ 7 ] is very successful . hog accurately capture structures of an image , in low resolution of an image it capture high level features leaving fine grained pixels . this will help us representing the images as different skull shape and getting rid of its inside content . we have formed first five hog template for reference of five shapes . for classification a pyramid of hog feature at different resolution of the image is developed . then a template matching based method is used for classification of its shape to a specific shape from the templates . this is based on normalization correlation value . @xmath0\epsilon r}(t(i , j)-im(i , j))^{2}\ ] ] in this step we have image of known shape , for classifying their actual sign on the basis of hidden topic , a probabilistic latent semantic analysis based classification techniques will be used . since image is a very high dimensional data , we will prepossess it to reduce its dimensionality by using visual codebook formation method . each image will be represented as a bag of visual words . each visual words will be relevant information stored in those images . for extracting meaningful information from images we will use sift descriptor . for object recognition sift is a very popular local descriptor having dimension of 128 . this descriptor is scale invariant and suitable for object recognition and key - point matching purpose . sift computation based on construction of scale space . it uses difference of gaussian filter for locating potential corner points . actual implementation used 16x16 image patch , and further divided into 4x4 descriptors , resulting a 4x4x16 = 128 dimensional vector . for the formation of visual words locality constrained linear coding [ 18 ] based method is used . this algorithm generate similar codes for similar descriptors by sharing bases . locality also leads to sparsity . idea based locality importance more than sparsity is used and given by below optimization problem . here x is a d dimensional local descriptors , @xmath1 \in \re^{dxn}$ ] and @xmath2 \in \re^{dxm}$ ] is m dimensional codebook . @xmath3 @xmath4 @xmath5 where @xmath6 denotes the element - wise multiplication , and @xmath7 is the locality adaptor that gives different freedom for each basis vector proportional to its similarity to the input descriptor@xmath8 . also @xmath9 is the euclidean distance between @xmath10 and b. shift invariance nature is confirmed by the constraint equation ( 3 ) . in our case we will have 128 dimension sift descriptors @xmath10 for each image . and we will form a codebook b of size 300 words . each image will be expressed as a combination of these words . each image will be represented as histogram of visual words . in fig . [ 4 ] codebook formation idea is depicted . probabilistic latent semantic analysis [ 7 ] is a topic discovery model , its concept based on latent variable analysis . an image also can be considered as a collection of topics . every image can be considered as a text document with words from a specific vocabulary . the vocabulary will be formed by using k - means algorithm on sift features from training image . suppose we have a collection of n images(document ) @xmath11 , and corresponding vocabulary with size n1 is @xmath12 . and let there be n2 topic @xmath13 model parameter are computed using expectation maximization method . in fig . [ 3 ] and fig . [ 5 ] plsa algorithm idea and typical scenario is explained . @xmath14 @xmath15 @xmath16 @xmath17 @xmath18 by using histogram of words concept , each image will be converted to a document with previously designed vocabulary . since our framework based on topic modelling and discovery , fuzzy knn classification [ 19 ] techniques is used for getting appropriate topic af a test image corresponding to training images . fuzzy knn perform better than traditional knn algorithm , because it depends on weight of neighbours . in the training stage we will calculate @xmath19 , which will be used as input for testing algorithm to compute @xmath20 . after that a k- nearest neighbour algorithm is used to classify these image by using probability distribution @xmath21 . @xmath22 @xmath23 is value for membership strength to be computed , and @xmath24 is previously labelled value of i th class for j th vector . final class label of the query point x is computed as follows : @xmath25 * 20c algorithm & speed limits & prohibitions & derestrictions & mandatory & danger & unique + committee of cnns[3 ] & 99.47 & 99.93 & 99.72 & 99.89 & 99.07 & 99.22 + human & 98.32 & 99.87 & 98.89 & 100.00 & 99.21 & 100.00 + multi - scale cnn[15 ] & 98.61 & 99.87 & 94.44 & 97.18 & 98.3 & 98.63 + * plsa * & 98.82 & 98.27 & 97.93 & 96.86 & 96.95 & 100.00 + random forest[16 ] & 95.95 & 99.13 & 87.50 & 99.27 & 92.08 & 98.73 + lda on hog2 [ 2 ] & 95.37 & 96.80 & 85.83 & 97.18 & 93.73 & 98.63 + for testing accuracy of our algorithm we have used well known , german traffic sign recognition benchmark ( gtsrb ) [ 17 ] , having 12000 training image of different class , it has 43 different class of signs . this dataset contain wide variety of image with the consideration of different challenge may involve in recognition . we have also used belgium traffic sign recognition dataset [ 5 ] and used extra 19 classes of sign which were not included in gtsrb . this classes mainly related to different warning . different deformation due to viewpoint variation , occlusion due to different obstacles like trees , building etc . , natural degrading , different weather condition are considered in these dataset . also for training the system we have used different version of same images , specifically rotation of a image approximately @xmath26 in left and right direction also translated version of original image . in addition to that images are prepossed using contrast adjustment method , both the original and contrast adjusted images are used for training . total number of images per category is shown in table . [ 2 ] . . [ cols="^,^,^",options="header " , ] before running main algorithm on these image we will resize our image to @xmath27 . for testing purpose we will use affine transformation of same image to produce different version of it for getting rid of viewpoint variation . plsa is an iterative algorithm , for convergence of parameters we have used 182 iterations . in fig . [ 7 ] we have shown log - likelihood variation with respect to number of iteration . first these image are classified on the basis of their shape by using hog features . once we have the prior information about the shape , we have applied our classification algorithm which is based on sift features . all testing image will be pre - processed and converted to combination of visual words using llc codebook formation method as described in earlier section for training images . our method final result is very promising . a comparison with different state of art methods is shown in table.[3 ] . also we will compare accuracy of some sub classes , like speed limits , other prohibitions , derestrictions , danger , mandatory and priority . we have almost very good classification result across sub classes . in table.[1 ] a comparison of results obtained by different methods is explored . the algorithm is implemented in a matlab platform also with some c++ rappers . all the testing was carried out in a quad core intel i7 linux machine . performance of our algorithm vary with numbers of topic used in classification . in fig . [ 6 ] we have shown a comparison of accuracy over topics for speed limits subclass . similar accuracy variation over topics is obtained for other subcategory . we have maximum accuracy over 29 - 44 number of topics . we have chosen 35 number of topics as optimal for classification of our system . in this work we have developed a novel and fast approach for traffic sign recognition problems . because of little variability between different same shaped sign it is very difficult problems in classification . we exploit the concept of representation of images as collection of words and topic , by using the visual codebook formation concepts . for computation of probability of topics in images we have used probabilistic latent semantic analysis . our main contribution is using novel plsa model for getting near state of art performance in classification problems and developing a shape based classifier . this system is independent of colour . hardware implementation of this approach is much easier than other svm , neural network and deep learning based approach . markus mathias , radu timofte , rodrigo benenson and luc van gool , `` traffic sign recognition - how far are we from the solution ? '' proceedings of ieee international joint conference on neural networks ( ijcnn 2013 ) dallas , usa , august 2013 csurka , g. , bray , c. , dance , c. , fan , l. : visual categorization with bags of key- points . slcv workshop , eccv , ( 2004 ) 1 - 22 mikolajczyk , k. , schmid , c. : scale and afine invariant interest point detectors . ijcv 60 ( 2004 ) 63 - 86 liansheng zhuang , lanbo she , yuning jiang , ketan tang , nenghai yu,``image classification via semi - supervised plsa '' , fifth international conference on image and graphics,2009 anna bosch , andrew zisserman , and xavier munoz,``scene classification using a hybrid generative / discriminative approach'',ieee trans . 712727 , 2008 . c. bahlmann , zhu , v. ramesh , m . pellkofer , t. koehler `` a system for traffic sign detection , tracking , and recognition using color , shape , and motion information'',intelligent vehicles symposium , 2005 j. stallkamp , m. schlipsing , j. salmen , and c. igel . `` the german traffic sign recognition benchmark : a multi - class classification competition '' , in proceedings of the ieee international joint conference on neural networks , pages 14531460 .	* in this work we developed a novel and fast traffic sign recognition system , a very important part for advanced driver assistance system and for autonomous driving . traffic signs play a very vital role in safe driving and avoiding accident . we have used image processing and topic discovery model plsa to tackle this challenging multiclass classification problem . our algorithm is consist of two parts , shape classification and sign classification for improved accuracy . for processing and representation of image we have used bag of features model with sift local descriptor . where a visual vocabulary of size 300 words are formed using k - means codebook formation algorithm . we exploited the concept that every image is a collection of visual topics and images having same topics will belong to same category . our algorithm is tested on german traffic sign recognition benchmark ( gtsrb ) and gives very promising result near to existing state of the art techniques . * + * keywords * : adas , traffic signs classification , plsa .
although legged robots have already been studied in robotics @xcite , only a very few papers consider the motion planning problem amidst obstacles @xcite . in @xcite some heuristic approaches are described while , in @xcite efficient and provably correct geometric algorithms are described for a restricted type of legged robots , the so - called spider robots to be defined precisely below , and for finite sets of point footholds . a _ legged robot _ consists of a body with legs . each leg has one end attached to the body and the other end ( called the foot ) that can lie on the ground ( or move in space between two positions on the ground ) . compared to the classic piano movers problem , legged robots introduce new types of constraints . we assume that the environment consists of regions in the plane , called _ foothold regions _ , where the robot can safely put its feet . a _ foothold _ is a point in a foothold region . the legged robot must satisfy two different constraints : the accessibility and the stability constraints . a foothold is said to be _ accessible _ from a _ placement _ ( position of the body of the robot ) if it can be reached by a leg of the robot . a placement is called _ stable _ if there exist accessible footholds and if the center of mass of the robot lies above the convex hull of these accessible footholds . the set of stable placements is clearly relevant for planning the motion of a legged robot : we call this set _ the free space _ of the legged robot . note that a legged robot has at least four legs , three legs ensure the stability of a placement and a fourth leg permits the motion of the robot . a first simple instance of a legged robot is the _ spider robot _ ( see figure [ spider - plan ] ) . the spider robot was inspired by ambler , developed at carnegie mellon university @xcite . the body of the spider robot is a single point in the euclidean plane and all its legs are attached to the body . the legs are retractable and their lengths may vary between @xmath13 and a constant @xmath1 . we also assume that the center of mass of the robot is its body . it follows that a placement is stable if the body of the robot lies above the convex hull of the accessible footholds . the constraint that the body of the spider robot lies in the plane ( instead of in 3d ) is not really restrictive . indeed , consider a legged robot for which that constraint is relaxed . then , if a placement @xmath14 of such a legged robot is stable then , any placement @xmath15 , @xmath16 is also stable . reciprocally , it can be shown that if @xmath17 is in the interior of the free space of the spider robot , then there exists @xmath18 such that @xmath14 is a stable placement of the corresponding legged robot . the problem of planning the motion of a spider robot has already been studied by boissonnat et al . however , their method assumes that the set of footholds is a finite set of points and can not be generalized to more complex environments . this paper proposes a new method for computing the free space of a spider robot in the presence of polygonal foothold regions . this method is based on a transformation between this problem and the problem of moving a half - disk amidst obstacles . our method requires the computation of some parts of the free space of the half - disk . these computations are rather technical and complicated . consequently , for the sake of clarity , we first present our algorithm for the simple case of discrete footholds , then we show how it can be generalized to the case of polygonal foothold regions . once the free space of the spider robot has been computed , it can be used to find trajectories and sequences of legs assignments allowing the robot to move from one point to another . indeed , once the free space is known , a trajectory of the body can be found in the free space . then , a sequence of legs assignments can be computed as follows ( see @xcite for details ) . given an initial legs assignment , the body of the robot moves along its trajectory until it crosses the convex hull of its ( three ) feet that are on the ground or one leg reaches its maximal extension . then , a suitable foothold is found for the fourth leg and one leg leaves its foothold . the paper is organized as follows : some notations and results of @xcite are recalled in the next section . section [ fsrthdr ] shows the transformation between the spider robot problem and the half - disk problem . we present in section [ computation_of_f ] our algorithm for computing the free space of a spider robot for a discrete set of footholds . section [ polygonal - foothold - regions ] shows how to extend the algorithm to polygonal foothold regions . in sections 2 , 3 and 4 , denotes a discrete set of distinct footholds @xmath19 in the euclidean plane ( will denote in section 5 a set of disjoint polygonal regions ) . point @xmath20 denotes the body of the robot ( in the same plane ) and @xmath21 $ ] is the length range of each leg . the free space @xmath0 is the set of all stable placements of @xmath20 . a placement is said to be at the _ limit of stability _ if it lies on the boundary of the convex hull of its accessible footholds . notice that @xmath0 is a closed set and contains the placements at the limit of stability . let @xmath22 denote the circle of radius @xmath1 centered at @xmath23 . @xmath24 is the arrangement of the circles @xmath22 for @xmath25 , i.e. , the subdivision of the plane induced by the circles . this arrangement plays an important role in our problem and we will express the complexity results in term of @xmath26 , the size of @xmath27 . in the worst - case , @xmath28 but if @xmath29 denotes the maximum number of disks that can cover a point of the plane , among the disks of radius @xmath1 centered at the @xmath30 , it can be shown that @xmath31 @xcite . clearly @xmath29 is not larger than @xmath5 and in case of sparse footholds , @xmath26 may be linearly related to the number of footholds . for any set @xmath32 , let @xmath33 denote its boundary , @xmath34 its convex hull , @xmath35 its relative interior in a space @xmath36 is the interior of @xmath32 in the space @xmath32 for the topology induced by @xmath36 . for example , the relative interior of a closed line segment in @xmath37 is the line segment without its endpoints , though its interior in @xmath37 is empty . ] , @xmath38 its closure , and @xmath39 its complementary set . let @xmath40 denote the set of angles @xmath41 . we denote by @xmath42 $ ] the equality of @xmath43 and @xmath44 modulo @xmath45 . we say in the sequel that two objects _ properly intersect _ if and only if their relative interiors intersect . the algorithm described in @xcite is based on the following observation : for @xmath20 in a cell @xmath46 of @xmath27 , the set of footholds that can be reached by the robot is fixed ; the portion of @xmath46 that belongs to @xmath0 is exactly the intersection of @xmath46 with the convex hull of the footholds that can be reached from @xmath47 . therefore , the edges of @xmath48 are either circular arcs belonging to @xmath24 or portions of line segments joining two footholds . moreover , a vertex of @xmath48 incident to two straight edges is a foothold ( see figure [ exemple - f-2 ] ) . the complexity of @xmath0 has been proved to be @xmath49 @xcite . the algorithm presented in @xcite computes the free space @xmath0 in @xmath50 time . it uses sophisticated data structures allowing the off - line maintenance of convex hulls . the algorithm described in this paper has the same time complexity , uses simple data structures and can be extended to the case where the set of footholds is a set of polygonal regions and not simply a set of points . for simplicity , we consider first the case of point footholds and postpone the discussion on polygonal foothold regions to section [ polygonal - foothold - regions ] . to simplify the presentation of this paper , we make the following general position assumptions . all these hypotheses can be removed by a careful analysis . recall that we consider here that the set of footholds is discrete . no two footholds lie at distance exactly @xmath1 or @xmath51 . among the circles @xmath52 and the line segments joining two footholds , the intersection between three circles or , two circles and a line segment or , one circle and two line segments , is empty . in this section , we establish the connection between the free space of the spider robot and the free space of a half - disk robot moving by translation and rotation amidst @xmath5 point obstacles . [ thfl1 ] the spider robot does not admit a stable placement at point @xmath53 if and only if there exists a half - disk ( of radius @xmath1 ) centered at @xmath53 that does not contain any foothold of ( see figure [ spider - to - hd ] ) . let @xmath54 be the set of all the footholds that are reachable from placement @xmath53 . by definition , @xmath53 is not stable if and only if the convex hull of @xmath54 does not contain @xmath53 ( see figure [ spider - to - hd ] ) . that is equivalent to say that there exists an open half - plane through @xmath53 containing @xmath54 , or that there exists a closed half - disk of radius @xmath1 centered at @xmath53 which does not contain any foothold . [ def - hd ] let @xmath55 be the half - disk of radius @xmath1 centered at @xmath53 ( see figure [ def - dd ] ) defined by : @xmath56 [ def - hi ] @xmath57 let : @xmath58 @xmath59 @xmath60 @xmath61 will be called the helicoidal volume centered at @xmath30 ( see figure [ hi ] ) . notice the typographical distinction between the circle @xmath22 defined in @xmath62 and the torus @xmath63 defined in @xmath64 . for convenience , we will often identify @xmath40 and the interval @xmath65 $ ] of @xmath66 . this allows us to draw objects of @xmath64 in @xmath37 and to speak of the @xmath67-axis . @xmath68 denotes the `` plane '' @xmath69 . [ def - space - l ] the free space of a half - disk robot moving by translation and rotation amidst the set of obstacles is the set of @xmath70 such that the half - disk @xmath71 does not intersect . [ propl ] @xmath72 . @xmath73 , the set @xmath74 is the free space of the half - disk @xmath71 moving by translation only , amidst the obstacle @xmath75 . since the set of points @xmath53 such that @xmath71 contains a @xmath30 is @xmath76 , @xmath74 is the complementary set of the union of the @xmath76 . thus , @xmath77 is the complementary set of the union of the @xmath78 , that is @xmath79 . let @xmath80 denote the mapping ( called `` orthogonal projection '' ) : @xmath81 . [ thfhi ] @xmath82 by definition of @xmath77 , @xmath83 is the set of points @xmath84 such that there exists an angle @xmath85 such that the half - disk @xmath55 does not intersect @xmath86 . by theorem [ thfl1 ] , it is equivalent to say that there exists @xmath85 such that @xmath55 does not intersect @xmath86 , or that @xmath53 is not a stable placement of the spider robot . thus , @xmath83 is the set of points @xmath53 where the robot does not admit a stable placement , i.e. , @xmath87 . the result then follows from proposition [ propl ] . [ remfhi ] @xmath88 is the largest `` cylinder '' included in @xmath79 , whose axis is parallel to the @xmath67-axis ( in grey in figure [ cpc ] ) . the basis of this cylinder is @xmath12 . .8 cm [ remgeneralize ] the results of this section do not depend on the fact that the footholds are discrete points . for more general foothold regions , we simply need to replace the helicoidal volumes by their analog . this will be done in section [ polygonal - foothold - regions ] . in this section , we propose an algorithm for computing @xmath0 based on theorem [ thfhi ] . a first attempt to use theorem [ thfhi ] may consist in computing @xmath72 and projecting it onto the horizontal plane . the motion planning of a convex polygonal robot in a polygonal environment has been extensively studied ( see for example @xcite ) . such algorithms can be generalized to plan the motion of a half - disk . it should lead to an algorithm of complexity @xmath89 , where @xmath90 is an almost linear function of @xmath5 . the projection can be done using classical techniques , such as projecting all the faces of and computing their union . since the complexity of the 3d object is not directly related to the complexity of its projection , this approach does not provide a combinatorial bound on @xmath12 . however , assuming @xmath91 ( which will be proved in this paper ) the time complexity of the algorithm of kedem et al . is @xmath92 . in this paper , we present a direct computation of @xmath12 . this approach provides an upper bound on the size of @xmath12 , namely @xmath91 . it also provides an algorithm for computing @xmath0 in @xmath93 time . as in @xcite and contrary to @xcite , the algorithm proposed here is sensitive to @xmath26 which is usually less than quadratic . another advantage of our direct computation is to avoid the explicit construction of the 3d object which is useless for our application . our algorithm manipulates only two - dimensional arrangements or lower envelopes and we provide a detailed description of the curves involved in the construction . let us now detail the computation of @xmath0 in the case of point footholds . we know that each arc of the boundary @xmath48of @xmath0 is either a straight line segment belonging to a line joining two footholds or an arc of a circle @xmath22 ( see section [ notations_and_previous_results ] ) . the circular arcs @xmath94 are computed first ( sections [ computation_of_df_cap_a ] , [ properties_of_zi ] and [ construction_of_df_cap_cio ] ) and linked together with the line segments in a second step ( sections [ cotaodfifaf ] and [ construction_of_f ] ) . in the sequel , the _ contribution _ of an object @xmath96 to another object @xmath97 is @xmath98 . we compute the contribution of each circle @xmath99 , @xmath100 , to @xmath48 in turn . recall that @xmath101 denote the torus @xmath102 . the contribution of each circle @xmath99 to @xmath48 will be obtained by computing the intersection of all the @xmath78 , @xmath103 , with the torus @xmath104 . let @xmath105 , @xmath103 , denote these intersections : + @xmath106 . we first show how to compute the contribution of @xmath99 to @xmath48 in term of the @xmath105 , and leave the studies of the shape and properties of @xmath105 to section [ properties_of_zi ] . figures [ example - zi - and - corresponding - hd ] and [ construction - df - cap - co-1-v2 ] show some ( hatched ) @xmath107 ( @xmath108 ) where @xmath101 is parameterized by @xmath109 ( @xmath110 and @xmath67 parameterize @xmath99 and @xmath40 respectively ) ; the dark grey region shows @xmath111 . [ contribution ] the contribution of @xmath99 to @xmath48 is : + @xmath112 . since @xmath0 is a closed set , @xmath113 \setminus [ c_{i_0 } \cap int({\ensuremath{{\cal f}}})]$ ] . according to theorem [ thfhi ] , @xmath114 . one can easily prove that for any set @xmath115 , @xmath116 , @xmath117 , and @xmath118 . it then follows from the expression of @xmath12 that @xmath119 . recall that for any sets @xmath120 , @xmath121 , @xmath122 , and @xmath123 . that implies + @xmath124 . we now consider that equation with @xmath96 equal to @xmath79 or @xmath125 , and @xmath97 equal to the torus @xmath104 . since @xmath126 is the circle @xmath99 we get : + @xmath127 and @xmath128 . since @xmath129 and @xmath106 by definition , @xmath130 and @xmath131 . by the general position assumption , no two footholds lie at distance @xmath51 , thus for @xmath132 , @xmath133 denotes the relative interior , thus @xmath134 is the interior of @xmath61 in @xmath135 but @xmath136 denotes the interior of @xmath137 in @xmath104 . ] . as @xmath138 , we get @xmath139 . the study of the shape of @xmath137 will yield ( see lemma [ inter - int - zi ] ) that @xmath140 . therefore , @xmath141 and @xmath142 . using @xmath143\setminus [ c_{i_0 } \cap int({\ensuremath{{\cal f}}})]$ ] , we get the result . thus , the contribution of @xmath99 to @xmath48 comes from the computation of @xmath144 and @xmath145 . geometrically , @xmath146 is the vertical projection ( along the @xmath67-axis ) of the largest vertical strip @xmath147 included in @xmath144 ( see figure [ construction - df - cap - co-1-v2 ] ) . similarly , @xmath148 is the projection of the largest vertical strip @xmath149 included in @xmath150 . thus , @xmath151 is the vertical projection onto @xmath99 of the vertical strip @xmath152 . in order to compute @xmath0 efficiently , we need to compute the union of the regions @xmath105 efficiently . more precisely , we will show that the union of the regions @xmath105 can be computed in @xmath153 time where @xmath154 is the number of helicoidal volumes @xmath61intersecting @xmath104 . this is possible because the @xmath137 have special shapes that allow us to reduce the computation of their union to the computation of a small number of lower envelopes of curves drawn on @xmath104 , with the property that two of them intersect at most once . the geometric properties of the @xmath137 are discussed in section [ properties_of_zi ] and , in section [ construction_of_df_cap_cio ] , we present and analyze the algorithm for constructing @xmath151 . we study here the regions @xmath155 . recall that we parameterize @xmath156 by @xmath109 where @xmath110 and @xmath67 parameterize @xmath99 and @xmath40 respectively ( @xmath157 corresponds to the point of @xmath99 with maximum @xmath43-coordinate ) . figures [ example - zi - and - corresponding - hd ] and [ construction - df - cap - co-1-v2 ] show examples of such regions @xmath105 . for convenience , we will use the vocabulary of the plane when describing objects on the torus @xmath104 . for instance , the curve drawn on the torus @xmath101 with equation @xmath158 will be called a line . the line @xmath159 will be called vertical and oriented according to increasing @xmath67 . lower and upper will refer to this orientation . the discussion below considers only non empty regions @xmath105 ( such that @xmath160 ) . we introduce first some notations . let @xmath161 be the half - circle of the boundary of @xmath76 , i.e. , @xmath162 . let @xmath163 be the spoke of @xmath22 that makes an angle @xmath67 with the @xmath43-axis , i.e. , @xmath164\}$ ] where @xmath165 is the unit vector whose polar angle is @xmath67 . the boundary of @xmath61 is composed of the three following patches : @xmath166 let @xmath167 and @xmath168 denote the curves @xmath169 and @xmath170 , respectively . since @xmath171 and @xmath172 are translated copies of one another , i.e. , @xmath173 , we have : [ translated - copies ] @xmath167 and @xmath168 are translated copies of one another , i.e. , + @xmath174 . [ u - monotone ] the curves @xmath175 are monotone in @xmath110 . assume for a contradiction that a curve @xmath175 is not monotone in @xmath110 . then , there exists @xmath110 and @xmath176 in @xmath40 such that @xmath109 and @xmath177 parameterize points of @xmath175 . by the definition of @xmath178 , it then follows that the point @xmath179 parameterized by @xmath110 belongs to the two spokes @xmath163 ( or @xmath180 ) and @xmath181 ( or @xmath182 ) . the intersection between any two of these spokes is exactly @xmath23 . thus , @xmath183 , which contradicts ( since @xmath179 ) the general position assumption saying that the distance between @xmath30 and @xmath184 is not @xmath1 . [ zi0 ] the region @xmath185 is the subset of @xmath101 parameterized by @xmath186 ( shown in grey in figures [ example - zi - and - corresponding - hd ] and [ construction - df - cap - co-1-v2 ] ) . for any @xmath85 , the intersection between @xmath187 and the `` horizontal plane '' @xmath188 is the half - disk @xmath189 . similarly , the intersection between @xmath101 and that plane is @xmath99 . thus , the intersection between @xmath185 and @xmath188 is @xmath190 , which is parameterized on @xmath99 by @xmath191 . that intersection is actually on the plane @xmath188 and is therefore parameterized on @xmath101 by @xmath186 . [ zi ] @xmath137 is a connected region bounded from below by @xmath167 and from above by @xmath168 , i.e. , @xmath192 , ( u,\theta - x)\in \rho_i^- , ( u,\theta - x+\pi)\in \rho_i^+\}$ ] ( see figures [ example - zi - and - corresponding - hd ] , [ construction - df - cap - co-1-v2 ] ) . by cutting @xmath101 and @xmath78 by the `` horizontal plane '' @xmath188 , we get that a point parameterized by @xmath109 on @xmath101 belongs to @xmath78 if and only if the point @xmath193 parameterized by @xmath110 on @xmath99 belongs to @xmath194 . since @xmath194 can be seen as the union of the spokes @xmath195\}$ ] , @xmath196 if and only if there exists @xmath197 $ ] such that @xmath198 , or equivalently , @xmath199 with @xmath200 $ ] . since @xmath201 , it follows from @xmath199 that the point of @xmath101 parameterized by @xmath202 belongs to @xmath171 and thus to @xmath203 . from lemma [ translated - copies ] , we get that the point parameterized by @xmath204 belongs to @xmath205 . therefore , @xmath137 is a connected region bounded from below by @xmath167 and from above by @xmath168 . we want to compute the union of the @xmath137 by computing the `` lower envelope '' are not actually defined . ] of the lower edges @xmath206 , and the `` upper envelope '' of the upper edges @xmath207 . it is unfortunately impossible to do so because some upper edges @xmath205 may possibly be `` below '' or intersect some lower edges @xmath208 . however , we can subdivide the regions @xmath137 into blocks @xmath209 , @xmath210 , and separate these blocks into two sets @xmath211 and @xmath212 such that the union of the @xmath209 in @xmath211 ( resp . @xmath212 ) is the region bounded from above by the upper envelope of the upper edges of the @xmath213 and bounded from below by the lower envelope of the lower edges of the @xmath213 ( resp . @xmath212 ) . such property can be realized by showing that all the upper edges of the @xmath213 belong to the strip @xmath214\}$ ] and all the lower edges of the @xmath213 belong to the strip @xmath215\}$ ] ( a similar property is shown for @xmath212 ) . note that the upper and lower envelopes are then defined since they are considered in @xmath216 . we subdivide @xmath137 into blocks @xmath209 when @xmath217 . that subdivision is performed such that the upper and lower edges of the @xmath209 are @xmath67-monotone . recall that the upper edge @xmath205 of @xmath105 is the intersection of @xmath218 and @xmath104 . the spoke @xmath163 intersects @xmath99 twice ( for some @xmath67 ) when @xmath219 , which implies that @xmath205 is not @xmath67-monotone . we cut the spoke @xmath163 into two pieces such that each piece intersects @xmath99 at most once . let @xmath220 be the intersection point between @xmath99 and on one of the two lines passing through @xmath30 and tangent to @xmath99 ( see figure [ def - t - v2 ] ) . let @xmath221 be the point on @xmath163 at distance @xmath222 from @xmath23 . cutting @xmath163 at @xmath223 defines two sub - spokes @xmath224 and @xmath225 that intersect @xmath99 in at most one point each ; without loss of generality , let @xmath224 denote the sub - spoke joining @xmath23 to @xmath223 . the set of @xmath85 for which @xmath224 intersects @xmath99 is clearly connected but the set of @xmath85 for which @xmath225 intersects @xmath99 consists of two connected components . we denote by @xmath226 the intersection @xmath227 and by @xmath228 and @xmath229 the two connected components of the intersection @xmath230 ( see figure [ construction - df - cap - co-1-v2 ] ) . since @xmath224 and @xmath225 intersect @xmath99 at most once for any @xmath85 , the curves @xmath228 , @xmath226 and @xmath231 are @xmath67-monotone . the lower edges @xmath232 , @xmath233 can be defined similarly or in a simpler way as the translated copies of @xmath234 , @xmath233 , i.e. , @xmath235 . we denote by @xmath209 , @xmath233 , the subset of @xmath137 bounded from above by @xmath234 and from below by @xmath232 . we can now prove the following proposition that will allow us to compute the union of the @xmath137 by computing the upper and lower envelopes of their upper and lower edges . [ droitecoupe ] if @xmath236 , the line @xmath237 properly intersects @xmath105 , and the lines @xmath238 properly intersect neither @xmath239 nor @xmath240 . if @xmath219 , the line @xmath241 properly intersects @xmath242 , and the line @xmath237 properly intersects @xmath243 and @xmath244 . furthermore , the lines @xmath245 properly intersect none of the edges @xmath228 , @xmath246 , @xmath226 , @xmath247 , @xmath231 and @xmath248 . if @xmath249 , the line @xmath241 properly intersects @xmath105 , and the lines @xmath238 properly intersect neither @xmath205 nor @xmath250 . let @xmath251 parameterize a point of a curve @xmath252 . let @xmath53 denote the point of @xmath99 with parameter @xmath253 and @xmath254 $ ] ( see figure [ gamma - v2 ] ) . one can easily show that @xmath255 $ ] . we prove that @xmath256 $ ] , except possibly when @xmath251 is an endpoint of @xmath252 ( or @xmath257 when @xmath219 ) , which implies , since @xmath258 $ ] , that the lines @xmath238 intersect neither @xmath239 nor @xmath259 ( resp . @xmath260 nor @xmath261 ) , except possibly at their endpoints . * case 1 * : @xmath236 . since @xmath30 belongs to the disk of radius @xmath1 centered at @xmath262 , @xmath263 for any @xmath264 ( see figure [ gamma - v2 ] ) . thus , the lines @xmath245 properly intersect neither @xmath205 nor @xmath206 . finally , the point of @xmath101 @xmath265 , where @xmath266 $ ] , belongs to the line @xmath237 and also to the relative interior of @xmath137 since it belongs to the interior of @xmath61 ( see figure [ points - dans - zi]a ) . therefore , the line @xmath267 properly intersects @xmath105 . * case 2 * : @xmath219 . let @xmath268 parameterize the point connecting @xmath269 and @xmath226 , and @xmath270 parameterize the point connecting @xmath226 and @xmath229 . let @xmath271 and @xmath272 denote the points of @xmath99 parameterized by @xmath273 and @xmath274 respectively . according to the construction of @xmath269 , @xmath226 and @xmath229 , the tangent lines to @xmath99 at @xmath271 and @xmath272 pass through @xmath30 . at most two tangent lines to @xmath99 pass through @xmath30 , thus @xmath271 and @xmath272 are the only points of @xmath99 where @xmath275 $ ] . since @xmath205 is @xmath110-monotone by lemma [ u - monotone ] , @xmath268 and @xmath270 are the only points of @xmath205 where @xmath275 $ ] . therefore , the lines @xmath276 do not properly intersect @xmath234 , @xmath233 . similarly , the lines @xmath277 do not properly intersect @xmath232 , @xmath233 . let @xmath278 and @xmath279 be the parameters on @xmath99 of the intersection points between @xmath99 and @xmath22 ( see figures [ points - dans - zi]c and d ) ; to differentiate @xmath278 from @xmath279 , assume without loss of generality that , for any @xmath280 small enough , the points of @xmath99 parameterized by @xmath281 and @xmath282 are in the disk of radius @xmath1 centered at @xmath23 . then , the points @xmath283 and @xmath284 of @xmath104 belong to @xmath243 and @xmath244 ( or to @xmath244 and @xmath243 ) respectively ( see figures [ points - dans - zi]c and d ) . however , these points do not belong to the relative interior of @xmath243 and @xmath244 ( because they lie on the border of @xmath285 and @xmath286 ) . nevertheless , there clearly exists @xmath280 small enough such that the point parameterized by @xmath281 ( resp . @xmath282 ) on @xmath99 belongs to the interior of the half - disk @xmath287 ( resp . @xmath288 ) . thus , the points @xmath289 and @xmath290 of @xmath104 belong to the relative interior of @xmath243 and @xmath244 respectively . therefore , the line @xmath237 properly intersects @xmath243 and @xmath244 . on the other hand , @xmath291 ( where @xmath292 $ ] ) belongs to relative interior of @xmath242 because the point of @xmath99 parameterized by @xmath293 belongs to the relative interior of the sub - spoke @xmath294 ( see figure [ points - dans - zi]b ) which belongs to interior of @xmath295 . therefore , the line @xmath296 properly intersects @xmath242 . * case 3 * : @xmath249 . since @xmath163 intersects @xmath99 at most once , @xmath297 $ ] ( see figure [ gamma - v2 ] ) . moreover , @xmath298 @xmath299 $ ] only when @xmath300 , but then , @xmath53 is at distance @xmath1 from @xmath30 which implies that @xmath251 is an endpoint of @xmath252 . thus , the lines @xmath301 intersect neither @xmath205 nor @xmath250 , except possibly at their endpoints . finally , the point @xmath291 of @xmath101 ( where @xmath302 $ ] ) belongs to the line @xmath241 and also to the relative interior of @xmath137(see figures [ points - dans - zi]b and [ example - zi - and - corresponding - hd ] ) . therefore , the line @xmath296 properly intersects @xmath105 . by proposition [ droitecoupe ] , we can compute the union @xmath303 by separating the @xmath105 , @xmath209 into two sets @xmath211 and @xmath212 ( where @xmath105 , @xmath209 belongs to @xmath211 if and only if @xmath205 , @xmath234 belongs to the strip @xmath304\}$ ] and @xmath206 , @xmath232 belongs to the strip @xmath215\}$ ] ) and computing the union of the @xmath105 , @xmath209 in @xmath211 ( resp . @xmath212 ) by computing the upper envelope of their upper edges and the lower envelope of their lower edges . in order to compute efficiently these upper and lower envelopes , we show that the curves @xmath205 , @xmath206 , @xmath234 and @xmath232 intersect each other at most once . however , we need for that purpose to split the regions @xmath137 when @xmath305 into two blocks @xmath243 and @xmath242 separated by the vertical line @xmath306 ; it also remains to split the @xmath67-interval ( or the @xmath110-interval ) over which @xmath252 is defined into two intervals of equal length over which @xmath307 and @xmath308 are defined ( see figure [ construction - df - cap - co-1-v2 ] ) . note that proposition [ droitecoupe ] still holds if we replace ( when @xmath309 ) @xmath137 by @xmath209 and @xmath310 by @xmath311 , @xmath312 . for consistency , we split @xmath185 into two blocks @xmath313 and @xmath314 separated by a vertical line ( chosen arbitrarily , say @xmath315 ) . also for consistency , the curves @xmath316 when @xmath249 are occasionally denoted in the sequel @xmath307 . [ unefois ] let @xmath317 and @xmath318 be some connected portions of @xmath319 and @xmath320 respectively ( @xmath321 ) . if @xmath317 or @xmath318 is monotone in @xmath67 and defined over a @xmath67-interval smaller than @xmath322 , then @xmath317 and @xmath318 intersect at most once . let @xmath323 be a point of intersection between @xmath317 and @xmath318 and @xmath324 be the point of the circle @xmath99 with parameter @xmath325 . since @xmath317 is a portion of the intersection between @xmath101and @xmath178 , @xmath326 is a point of intersection between @xmath99 and the diameter of @xmath327 . therefore , the line passing through @xmath30 and @xmath324 has slope @xmath328 . by applying the same argument to @xmath318 , we obtain that @xmath30 and @xmath329 belong to the same straight line of slope @xmath328 . therefore , if @xmath317 and @xmath318 intersect twice , at @xmath323 and @xmath330 , then @xmath331 $ ] . it follows , if @xmath317 or @xmath318 is defined over a @xmath67-interval smaller than @xmath322 , that @xmath332 $ ] . furthermore , if @xmath317 or @xmath318 is monotone in @xmath67 , then @xmath323 and @xmath333 are equal . [ inter - int - zi ] @xmath334 , @xmath335 . we assume that @xmath321 because otherwise the result is trivial . one can easily show that @xmath336 only if the boundaries of @xmath137 and @xmath337 partially coincide , i.e. , the dimension of @xmath338 is 1 . by proposition [ zi ] , @xmath339 consists of the edges @xmath207 and @xmath206 and of two vertical line segments joining the endpoints of @xmath205 and @xmath206 when these endpoints exist ( which is the case when @xmath132 ) . moreover , these vertical line segments are clearly supported by the vertical lines @xmath340 and @xmath341 where @xmath278 and @xmath279 parameterize on @xmath99 the points of intersection between @xmath99 and @xmath22 ( see figure [ example - zi - and - corresponding - hd ] ) . by lemma [ unefois ] , the edges @xmath316 and @xmath320 do not partially coincide . by the general position assumption , no three distinct circles @xmath99 , @xmath22 and @xmath342 have a common intersection point . thus , for any @xmath321 , @xmath343 and @xmath344 are disjoint . therefore , the vertical lines @xmath345 and @xmath346 do not partially coincide . finally , since @xmath316 is nowhere partially supported by a vertical line by lemma [ u - monotone ] , @xmath316 and the vertical lines @xmath346 do not partially coincide . [ une - intersection ] any two curves among the curves @xmath347 intersect at most once ( where @xmath348 if @xmath236 , @xmath349 if @xmath217 , and @xmath350 if @xmath249 ) . by lemma [ unefois ] , it is sufficient to prove that all the curves @xmath347 , @xmath132 , are monotone in @xmath67 and defined over @xmath67-intervals smaller than @xmath322 . indeed , the curves @xmath351 , @xmath352 , @xmath353 and @xmath354 clearly do not pairwise intersect more than once , by lemma [ zi0 ] . if @xmath355 , any spoke of @xmath22 intersects @xmath99 at most once . hence , @xmath316 is monotone in @xmath67 . @xmath316 is defined over a @xmath67-interval greater than @xmath322 but smaller than @xmath356 . since we have split that interval in two equal parts , @xmath307 and @xmath308 are defined over a @xmath67-interval smaller than @xmath322 ( see @xmath357 in figure [ construction - df - cap - co-1-v2 ] ) . if @xmath219 , the @xmath67-interval where @xmath163 ( or @xmath358 ) intersects @xmath99 is smaller than @xmath322 , which implies that @xmath252 is defined over a @xmath67-interval smaller than @xmath322 . the curves @xmath234 , @xmath233 , are defined as the connected components of @xmath227 and @xmath230 . since the sub - spokes @xmath224 and @xmath225 intersect @xmath99 at most once for any @xmath85 , the curves @xmath234 , @xmath233 , are @xmath67-monotone . if @xmath249 , @xmath163 ( and also @xmath359 ) intersects @xmath99 in at most one point , which proves that @xmath252 is monotone in @xmath67 . furthermore , the @xmath67-interval where @xmath252 is defined is smaller than @xmath322 because the @xmath67-interval where @xmath163 ( or @xmath359 ) intersects @xmath99 is smaller than @xmath322 . we first show how to compute @xmath360 . let @xmath211 and @xmath212 be the following sets of @xmath209 : + @xmath361 , + @xmath362 . by proposition [ droitecoupe ] , the line @xmath296 properly intersects all the @xmath213 but the lines @xmath301 properly intersect none of their upper and lower edges @xmath234 and @xmath232 . thus , the regions @xmath213 can be seen as regions of @xmath363\}$ ] such that all their upper edges @xmath234 lie in @xmath214\}$ ] and all their lower edges @xmath232 lie in @xmath215\}$ ] . therefore , the union of the @xmath364 is the region of @xmath365\}$ ] bounded from above by the upper envelope of their @xmath234 and bounded from below by the lower envelope of their @xmath232 . similarly , the union of the @xmath366 is the region of @xmath367\}$ ] bounded from above by the upper envelope of the @xmath234 and bounded from below by the lower envelope of the @xmath232 . the union of @xmath211 and @xmath212 , which is @xmath368 , can be achieved by computing , on one hand , the intersection between the upper edge chain of @xmath369 with the lower edge chain of @xmath370 ( which both belong to @xmath371\}$ ] ) , and on the other hand , the intersection between the upper edge chain of @xmath370 with the lower edge chain of @xmath369 ( which both belong to @xmath372\}$ ] ) . these intersections can simply be performed by following the two edge chains for @xmath110 from @xmath13 to @xmath356 , since they are monotone in @xmath110 by lemma [ u - monotone ] . let us analyze the complexity of the above construction . the @xmath154helicoidal volumes @xmath61 that intersect @xmath101 can be found in @xmath373 amortized time once the delaunay triangulation of the footholds has been computed , which can be done in @xmath374 time @xcite . by proposition [ une - intersection ] , the upper and lower envelopes can be computed in @xmath153 time using @xmath375 space where @xmath376 is the pseudo inverse of the ackerman s function @xcite . also by proposition [ une - intersection ] , the union of @xmath377 and @xmath378 can be done in linear time in the size of the edge chains , that is @xmath375 time . thus , we can compute @xmath368 in @xmath153 time using @xmath375 space after @xmath374 preprocessing time . we can compute @xmath379 similarly by removing @xmath313 and @xmath314 from @xmath378 . the contribution of @xmath99 to @xmath48 is , according to proposition [ contribution ] , @xmath380 . by remark [ remfhi ] , + @xmath146 and @xmath148 are the projections onto @xmath99 of the largest vertical strips @xmath381 and @xmath382 included in @xmath383 and @xmath384 , respectively ( see figure [ construction - df - cap - co-1-v2 ] ) . these projections are easily computed because the edges of @xmath383 and @xmath384 are monotone with respect to @xmath110 ( lemma [ u - monotone ] ) . these projections , and therefore the computation of @xmath385 , can thus be done in linear time and space in the size of @xmath383 and @xmath386 , that is @xmath375 . moreover , we label an arc of @xmath48 either by @xmath387 if the arc belongs to the circle @xmath22 or by @xmath388 if the arc belongs to the straight line segment @xmath389 $ ] . the labels of the edges of @xmath48 incident to @xmath99 can be found as follows , without increasing the complexity . an arc of @xmath151 corresponds to a vertical strip @xmath390 . an endpoint @xmath53 of such an arc is the projection of a vertical edge , or the projection of a point of intersection between two curved edges . in the first case , @xmath53 is the intersection of @xmath99 with some @xmath22 and in the second case , @xmath53 is the intersection of @xmath99 with some line segment @xmath391 $ ] . by the general position assumption , among the circles @xmath52 and the line segments joining two footholds , the intersection between three circles or , two circles and a line segment or , one circle and two line segments , is empty . thus , @xmath53 is the intersection between @xmath99 and either a unique @xmath22 or a unique line segment @xmath391 $ ] . therefore , the edge of @xmath48 incident to @xmath99 at @xmath53 is either a circular arc supported by @xmath22 or a line segment supported by @xmath391 $ ] . hence , the labels of the edges of @xmath48 incident to @xmath99 can be found at no extra - cost during the construction . since @xmath24 is the arrangement of the circles of radius @xmath1 centered at the footholds , @xmath392 . the above considerations yield the following theorem : [ r12 ] we can compute @xmath393 and the labels of the edges of @xmath48 incident to the arcs of @xmath393 in @xmath50 time using @xmath394 space . the previous section has shown how to compute all the vertices of @xmath0that are incident to at least one circular arc . it remains to find the vertices of @xmath0 incident to two straight edges . as we have seen in section [ notations_and_previous_results ] , a vertex of @xmath12 incident to two straight edges of @xmath395 is a foothold . furthermore , considering a foothold @xmath184 in a cell @xmath46 of @xmath27 , @xmath184 is a vertex of @xmath12 incident to two straight edges of @xmath395 if and only if @xmath184 is a vertex of the convex hull of the footholds reachable from @xmath184 . the @xmath396 footholds contained in the disk of radius @xmath1 centered at @xmath262 can be found in @xmath397 amortized time because we have already computed the delaunay triangulation of the footholds @xcite . thus , we can decide if @xmath184is a vertex of the convex hull of these @xmath396 footholds in @xmath397 time and space . when @xmath184 is a vertex of the convex hull , we can also find the two edges of the convex hull adjacent to @xmath184 in @xmath397 time and space . as the sum of the @xmath398 for @xmath399 is bounded by the size of @xmath27 , we obtain the following theorem : [ r34 ] the footholds belonging to @xmath48 and the labels of the arcs of @xmath48issued from these footholds can be found in @xmath400 time and space . [ r ] the free space of the spider robot can be computed in @xmath50 time using @xmath394 space . by theorem [ r12 ] , we have computed all the circular arcs of @xmath48and the labels of the edges of @xmath48 incident to them . by theorem [ r34 ] , we have computed all the vertices of @xmath48 that are incident to two straight edges of @xmath48 and the label of these two edges . it remains to sort the vertices of @xmath48 that appear on the line segments @xmath391 $ ] . we only consider the line segments @xmath391 $ ] such that the corresponding label @xmath388 appears during previous computations . then , we sort the vertices of @xmath48 that belong to each such relevant line . since @xmath401 @xcite , sorting all these vertices can be done in @xmath50 time . a complete description of @xmath395 then follows easily . we consider now the case where the set of footholds is no longer a set of points but a set of pairwise disjoint polygonal regions bounded by @xmath5 line segments @xmath402 . clearly , is a subset of the free space @xmath0 of the spider robot . let @xmath403 denote the free space of the spider robot using as foothold regions only the edges @xmath402 . suppose that the spider robot admits a stable placement outside with its feet inside some polygonal footholds ; then the placement remains stable if it retracts its legs on the boundary of these polygonal regions . hence , @xmath404 . we show how to compute @xmath405 . as observed in remark [ remgeneralize ] , the results of section [ fsrthdr ] remains true if the foothold regions are line segments provided that @xmath61 is replaced by @xmath406 the generalized helicoidal volume defined by ( see figure [ ehi ] ) : @xmath407 the helicoidal volume associated to a point site @xmath30 will be , henceforth , denoted by @xmath408 . .7 cm similarly , we define the generalized circle @xmath409 as the set of points at distance @xmath1 from @xmath410 . let @xmath411 denote the arrangement of the @xmath5 generalized circles @xmath412 . notice that @xmath413 . each arc of the boundary @xmath414 of @xmath403 is either an arc of @xmath409 corresponding to a maximal extension of one leg , or an arc corresponding to placements at the limit of stability of the spider robot . similarly to what we did in section [ computation_of_f ] , we compute first the contribution of each @xmath409 to @xmath414(sections [ computation_of_dfp_cap_ea ] ) . thereafter , we compute the arcs of @xmath414 that correspond to placements where the spider robot is at the limit of stability ( section [ unstable - equilibrium ] ) . finally , we show how to construct @xmath403 ( and @xmath2 ) in section [ construction_of_f_2 ] . figure [ f - poly ] shows an example of free space @xmath403 for polygonal foothold regions . we compute the contribution to @xmath414 of each generalized circle @xmath409 in turn . we consider the contribution of @xmath416 to @xmath414 for some @xmath417 . @xmath416 is composed of two half - circles and two straight line segments . in order to compute the contribution of @xmath416 to @xmath418 , we evaluate first the contribution of the half - circles and then the contribution of the straight line segments . for convenience , we will not compute the contribution of the half - circles to @xmath414 but the contribution of the whole circles . similarly , we will compute the contribution of the whole straight lines supporting the line segments of @xmath419 . let @xmath184 and @xmath420 denote the two endpoints of the line segment @xmath421 , and let @xmath422 and @xmath423 denote the unit circles centered at @xmath184 and @xmath420 respectively . let @xmath424 and @xmath425 denote the two straight line segments of @xmath416 , and @xmath426 and @xmath427 their supporting lines . we show how to compute the contributions of @xmath422 and @xmath426 to @xmath418 ; the contributions of @xmath423 and @xmath427 can be computed likewise . let @xmath428 and @xmath429 . basically , we compute @xmath430 and @xmath431 , as explained in section [ computation_of_df_cap_a ] , by computing @xmath432 , @xmath433 , @xmath434 and @xmath435 . the properties of the new regions @xmath436 and @xmath437 are different though similar to the properties of @xmath438 described in section [ properties_of_zi ] . the analysis of @xmath439 and @xmath440 are subdivided into two parts : first , we consider the line @xmath441 supporting @xmath442 and we examine the regions @xmath443 and @xmath444 where @xmath445 is the generalized helicoidal volume induced by @xmath441 : @xmath446 then we deduce @xmath439 ( resp . @xmath440 ) from @xmath447 , @xmath448 and @xmath449 ( resp . @xmath450 , @xmath451 and @xmath452 ) where @xmath30 and @xmath453 are the two endpoints of @xmath442 . thereafter , we compute the contribution of @xmath416 to @xmath414 in a way similar to what we did in section [ construction_of_df_cap_cio ] . the following theorem sums up these results : [ thae ] we can compute @xmath454 and the labels of the edges of @xmath414incident to the arcs of @xmath454 in @xmath455 time using @xmath456 space . the proof of this theorem , omitted here , is a direct generalization of the proof of theorem [ r12 ] . details are given in @xcite or @xcite . we now have to compute the edges of @xmath403 that do not belong to @xmath457 . the arcs of @xmath454 correspond to placements at the limit of accessibility of the spider robot , and vice versa . thus , other edges of @xmath403 correspond to placements at the limit of stability of the spider robot . we denote by @xmath458 the set of those edges . a placement @xmath53 of the spider robot is at the limit of stability if and only if there exists a closed half - disk of radius @xmath1 centered at @xmath53 that does not contain any foothold except at least two footholds located on the diameter of the half - disk such that @xmath53 is between these footholds ( see figure [ limit - stab ] ) . therefore , the edges of @xmath458 are portions of the curves drawn by the midpoint of a ladder of length @xmath51 moving by translation and rotation such that the ladder touches the boundary of the foothold regions in two points but does not intersect the interior of the foothold regions . hence , the edges of @xmath458 are supported by the projection ( onto @xmath459 ) of the edges of the boundary of the free space of the ladder moving by translation and rotation amidst the foothold regions considered as obstacles , i.e. , the set of @xmath460 such that the ladder of length @xmath51 that has its midpoint at @xmath53 and makes an angle @xmath67 with the @xmath43-axis does not properly intersect the interior of the foothold regions . according to @xcite , the edges of the boundary of the free space of the ladder can be computed in @xmath461 time using @xmath462 space . the projection ( onto @xmath459 ) of each edge can easily be computed in constant time . thus , we can compute , in @xmath461 time and @xmath462 space ( using @xcite ) , a set of curves in @xmath459 that contains the arcs of @xmath414 that correspond to placements at the limit of stability of the spider robot . however , it remains to compute the portions of these curves that belong to @xmath418 . the relative interior of an @xmath442 is called a _ wall_. an endpoint of an @xmath442 is called a _ corner _ ( when several walls share an endpoint , we define only one corner at that point ) . the _ ladder _ is a line segment of length @xmath51 . a _ placement _ of the ladder is a pair @xmath460 where @xmath53 is the location of the midpoint of the ladder and @xmath67 is the angle between the @xmath43-axis and the ladder . a _ free placement _ of the ladder is a placement where the ladder does not properly intersect the walls or partially lies on some walls and does not properly intersect the others ( if none of the polygonal regions of @xmath463 are degenerated into line segments or points , then a free placement of the ladder is a placement where the ladder does not intersect the interior of the polygonal regions of @xmath463 ) . a _ placement of type corner - ladder _ is a placement of the ladder such that the relative interior of the ladder touches a _ corner_. a _ placement of type wall - endpoint _ is a placement of the ladder such that an endpoint of the ladder touches a _ wall_. a _ placement of type corner - endpoint _ is a placement of the ladder such that an endpoint of the ladder touches a corner . we now define _ @xmath29-contact placements _ of the ladder . 1-contact placement _ is a free placement of type corner - ladder or wall - endpoint . 2-contact placement _ is either the combination of two 1-contact placements or a free placement of type corner - endpoint . a 2-contact placement is said to be of type ( corner - ladder)@xmath464 , ( corner - ladder , wall - endpoint ) , ( wall - endpoint)@xmath464 , or ( corner - endpoint ) , in accordance to the types of placements involved in the 2-contact placement ( see figure [ examples-2-contact - placements ] ) . given two walls ( resp . a wall and a corner , two corners , one corner ) , the set of 2-contact placements induced by these two walls ( resp . the wall and the corner , the two corners , the single corner ) is called a _ 2-contact curve_. the type of a 2-contact curve is the type of the 2-contact placement defining the curve . note that the 2-contact curves are defined in @xmath465 . 3-contact placement _ is a combination of a 1-contact placement and a 2-contact placement . the types of 3-contact placements are naturally given by ( corner - ladder)@xmath466 , ( corner - endpoint , wall - endpoint)@xmath467 with this definition , we unfortunately can not guarantee that all the 2-contact curves end at 3-contact placements . indeed , a 2-contact curve defined by the ladder sliding along a wall ( see figure [ examples-2-contact - placements]b ) ends on one side ( if no other wall blocks the sliding motion ) at a 2-contact placement of type ( corner - endpoint ) , where the ladder is collinear with the wall , without properly intersecting it . in order to ensure that all the 2-contact curves end at 3-contact placements , we consider these 2-contact placements as 3-contact placements , and denote their type by ( corner - endpoint , @xmath468 ) . a _ k - contact placement _ , @xmath469 , is the combination of @xmath45 1-contact placements , @xmath470 2-contact placements and @xmath471 3-contact placements such that @xmath472 . now , we define a _ 2-contact tracing _ as the projection onto @xmath459 of a 2-contact curve . similarly as above , we define the types of the 2-contact tracings . notice that , to any point @xmath53 on a given 2-contact tracing @xmath473 , corresponds a unique placement @xmath474 on the 2-contact curve that projects onto @xmath473 . it follows that , to any point @xmath53 on a 2-contact tracing @xmath473 , corresponds a unique pair @xmath475 of points of contact between the ladder at @xmath474 and the walls ( @xmath476 and @xmath477 are equal when @xmath473 is a 2-contact tracing of type ( corner - endpoint ) ) ; when @xmath53 is an endpoint of @xmath473 , a 3-contact placement corresponds to @xmath53 , however , @xmath475 is uniquely defined by continuity . the points @xmath476 and @xmath477 are called the _ contact points corresponding to @xmath478 . _ we also define the three contact points corresponding to a 3-contact placement . a 2-contact tracing is either a straight line segment , an arc of ellipse , an arc of conchoid or a circular arc . indeed ( see figures [ ce ] , [ we - we ] , [ cl - we ] and [ cl - cl ] ) , a 2-contact tracing of type ( corner - endpoint ) is a circular arc ; a 2-contact tracing of type ( wall - endpoint)@xmath464 is an arc of ellipse ; a 2-contact tracing of type ( corner - ladder , wall - endpoint ) is an arc of conchoid ( see @xcite ) ; a 2-contact tracing of type ( corner - ladder)@xmath464 is a straight line segment . as we said before , we can compute all these 2-contact tracings in @xmath461 time using @xmath462 space @xcite , and it remains to compute the portions of these curves that belong to @xmath418 . we first show that only some portions of the 2-contact tracings correspond to positions at the limit of stability of the spider robot ( section [ subsubsection - relevant-2-contact - tracings ] ) . these portions are called the relevant 2-contact tracings . then , we prove that we do not have to take into consideration the intersections between the relative interior of relevant 2-contact tracings ( proposition [ a - end - point - of - proper-2-contact - curves ] ) . we also show that , if a point @xmath479 is an endpoint of several relevant 2-contact tracings , only two of them can support edges of @xmath458 in the neighborhood of @xmath479 ( propositions [ at - most - two - edges ] ) . finally ( section [ construction - of - delta ] ) , we compute a graph whose edges are relevant 2-contact tracings and where the degree of each node is at most two . this graph induces a set @xmath480 of curves supporting @xmath458 ( theorem [ thc ] ) that will allow us to compute @xmath458 in section [ construction_of_f_2 ] . as mentioned above , a placement @xmath53 of the spider robot is at the limit of stability if and only if there exists a closed half - disk of radius @xmath1 centered at @xmath53 that does not contain any foothold except at least two footholds located on the diameter of the half - disk , one on each side of @xmath53 . thus , a point @xmath53 of a 2-contact tracing @xmath473 belongs to an arc of @xmath458 only if @xmath53 lies between the two contact points corresponding to @xmath481 . the portions of the 2-contact tracings for which that property holds are called the _ relevant 2-contact tracings_. the other portions are called the _ irrelevant 2-contact tracings_. we now show how to compute the relevant 2-contact tracings for each type of contact . let @xmath473 denote a 2-contact tracing , let @xmath478 and let @xmath476 and @xmath477 be the two contact points corresponding to @xmath478 . in figures [ ce ] , [ we - we ] , [ cl - we ] and [ cl - cl ] , the walls and the relevant 2-contact tracings are thick , the irrelevant 2-contact tracings are dashed thick , and the ladder is thin . we now show that the intersections between the relative interiors of the relevant 2-contact tracings are not interesting for the spider robot motion problem . we recall that , if a vertex @xmath479 of @xmath414 belongs to @xmath457 , then we know by theorem [ thae ] the labels of the edges of @xmath414 incident to @xmath479 . otherwise , if @xmath482 , then the two edges of @xmath414 that end at @xmath479 correspond to placements at the limit of stability of the spider robot . [ a - end - point - of - proper-2-contact - curves ] any vertex @xmath479 of @xmath418 , such that @xmath482 , is an endpoint of the two relevant 2-contact tracings supporting the edges of @xmath414 ending at @xmath479 . since the two edges of @xmath414 that end at @xmath479 correspond to placements at the limit of stability of the spider robot , they are both supported by some relevant 2-contact tracings . thus , we only have to prove that @xmath479 is an endpoint of these two relevant 2-contact tracings . let @xmath483 and @xmath484 be these two relevant 2-contact tracings and assume for a contradiction that @xmath479 is not an endpoint of @xmath483 ( nothing is assumed for @xmath479 with respect to @xmath484 ) . let @xmath485 ( resp . @xmath486 ) be the placement of the ladder that correspond to @xmath487 ( resp . @xmath488 ) and let @xmath489 and @xmath490 ( resp . @xmath491 and @xmath492 ) be the corresponding contact points ( see figure [ mmmmeeee ] ) . first , notice that @xmath493 . indeed , otherwise , @xmath494 is at least a 3-contact placement and then , @xmath479 must be an endpoint of @xmath483 , which contradicts our assumption . by the definition of the relevant 2-contact tracings , @xmath479 is between @xmath489 and @xmath490 . moreover , @xmath479 can not be equal to @xmath489 or @xmath490 since @xmath479 is not an endpoint of @xmath483 . it follows that neither @xmath491 nor @xmath492 is equal to @xmath479 , because otherwise @xmath494 would be a 3-contact placement . therefore , @xmath479 is strictly between @xmath489 and @xmath490 , and strictly between @xmath491 and @xmath492 . thus , @xmath479 is strictly inside the polygon @xmath495 . on the other hand , since @xmath482 , @xmath479 does not belong to any @xmath496 , and therefore , the walls supporting @xmath489 , @xmath490 , @xmath491 and @xmath492 intersect the open disk @xmath497 of radius @xmath1 centered at @xmath479 . thus , there exists four points @xmath498 , @xmath499 , @xmath500 and @xmath501 on these walls and in @xmath497 , that are close enough to @xmath489 , @xmath490 , @xmath491 and @xmath492 respectively to ensure that @xmath479 belongs to the interior of the polygon @xmath502 . since the distances from @xmath479 to @xmath498 , @xmath499 , @xmath498 and @xmath501 , are strictly smaller than @xmath1 , @xmath479 belongs to the interior of @xmath405 . this contradicts our assumption that @xmath479 is a vertex of @xmath414 and yields the result . consider now the adjacency graph @xmath503 of the relevant 2-contact tracings such that two relevant 2-contact tracings are connected in @xmath503 if and only if they have a common endpoint ( the intersections between the relative interiors of the relevant 2-contact tracings are not considered ) . notice that , given the set of relevant 2-contact tracings , @xmath503 can be easily computed in @xmath461 time . now , given two vertices of @xmath415 that are connected along @xmath414 by arcs of @xmath458 , we want to compute these arcs . for computing these arcs , we can not simply use the graph @xmath503 because the degree of some nodes of @xmath503 may be arbitrarily large ( see figure [ nodes - of - degree - k_simplifie - v2 ] ) . we show in the next proposition that we can deduce from @xmath503 a graph @xmath504 such that the degree of each node of @xmath504 is at most two and that @xmath504 supports any portion of @xmath414 which is the concatenation of arcs of @xmath458 . we consider four hypotheses ( h1,@xmath467,h4 ) that obviate the need to consider degenerate cases . they are not essential but substantially simplify the proof of the following proposition . the first three hypotheses are made to ensure that the degree of each vertex of the free space of the ladder is three . h1 : : the line segments @xmath402 compose the boundary of a set of non degenerated polygons ( i.e. , no polygon is reduced to a line segment or to a point ) . h2 : : the ladder does not admit any 4-contact placement . h3 : : the arc ( of conchoid ) drawn by an endpoint of the ladder when its other endpoint moves along a wall while the ladder remains in contact with a corner , is not tangent to any other wall . h4 : : the ladder does not admit any 3-contact placement when its midpoint is located at a corner . [ at - most - two - edges ] for any node @xmath479 of @xmath503 of degree @xmath29 such that @xmath482 , at most two relevant 2-contact tracings can support @xmath414 in a sufficiently small neighborhood of @xmath479 . moreover , we can determine these at most two curves in @xmath505 time using @xmath506 space . let @xmath482 be a node of @xmath503 of degree @xmath29 . we assume that @xmath507 , otherwise proposition [ at - most - two - edges ] is trivial . let @xmath508 be the relevant 2-contact tracings that end at @xmath479 , and let @xmath509 be the placement of the ladder that corresponds to @xmath510 . @xmath497 is the open disk of radius @xmath1 centered at @xmath479 . we distinguish two cases whether @xmath479 is a corner or not . * case 1 : @xmath479 is a corner . * ( see figure [ nodes - of - degree - k_simplifie - v2 ] . ) the 2-contact tracing @xmath511 involves at least another contact than the corner - ladder contact at @xmath479 . this contact can not be of type corner - endpoint by hypothesis h4 . if the contact is of type wall - endpoint , we define @xmath512 as the contact point between this wall and the ladder at placement @xmath513 ( see figure [ wedge ] ) . since @xmath482 , the wall must intersect @xmath497 and we define @xmath514 as a point close to @xmath512 in that intersection . if the contact is of type corner - ladder , we define @xmath515 as the corner ( distinct from @xmath479 ) involved in this contact ( notice that @xmath516 by hypothesis h4 ) . * fact : * @xmath517 . + otherwise , @xmath518 is a 3-contact placement contradicting hypothesis h4 . * fact : * _ @xmath519 is a non - flat vertex of @xmath520 or belongs to the interior of @xmath521_. + assume that @xmath522 . then , @xmath479 lies on the boundary of @xmath523 because , otherwise , the @xmath514 provide footholds such that the spider robot can move in a neighborhood of @xmath479 . furthermore , @xmath479 must be a non - flat vertex of @xmath523 , by hypothesis h4 . assume now that @xmath522 , and let @xmath524 and @xmath525 be the two vertices of @xmath526 @xmath527 such that @xmath524 , @xmath479 and @xmath525 are consecutive along the boundary of @xmath526 @xmath527 ( see figure [ wedge ] ) . we will exhibit a stable placement for the spider robot at any position @xmath53 inside the intersection of the wedge @xmath528 and a neighborhood of @xmath479 . let @xmath529 and @xmath530 be two points in the wedge @xmath528 such that the wedges @xmath531 and @xmath532 are right ( see figure [ wedge ] ) . + if @xmath53 is in the wedge @xmath533 , and is close enough to @xmath479 , the footholds @xmath479 , @xmath524 and @xmath534 yield a stable placement for the spider robot . + if @xmath53 is in the wedge @xmath535 , and is close enough to @xmath479 , footholds @xmath479 , @xmath536 and @xmath534 yield a stable placement for the spider robot . + if @xmath53 is in the wedge @xmath537 , and is close enough to @xmath479 , footholds @xmath479 , @xmath536 and @xmath525 yields a stable placement for the spider robot . * fact : * _ @xmath538 , @xmath539 , can not support an edge of @xmath414 incident to @xmath479 . _ + we assume that @xmath522 because , otherwise , the claim is obvious . it follows that @xmath479 is a non - flat vertex of @xmath523 . a 2-contact tracing @xmath538 , @xmath539 , can not be an arc of ellipse because , otherwise , @xmath513 is a 3-contact placement ( because @xmath479 is here a corner ) contradicting hypothesis h4 . then , @xmath538 can be either the segment @xmath540 or an arc of conchoid . if @xmath538 is an arc of conchoid , then , by the general properties of conchoids ( see @xcite ) , @xmath538 is tangent to the segment @xmath540 at @xmath479 . thus , @xmath538 is always tangent to the segment @xmath540 at @xmath479 . the point @xmath512 strictly belongs to the wedge @xmath528 , because we have shown that @xmath541 . thus , in a neighborhood of @xmath479 , @xmath538 is strictly inside the wedge @xmath528 and thus strictly inside @xmath405 . therefore , @xmath538 can not support @xmath418 , in a neighborhood of @xmath479 . hence , by sorting the @xmath512 by their polar angles around @xmath479 , we can determine , in @xmath505 time , if @xmath479 is a non - flat vertex of @xmath523 , and if so , determine @xmath542 and @xmath543 . if @xmath479 is a non - flat vertex of @xmath523 , then , only @xmath544 and @xmath545 can support an edge of @xmath414 incident to @xmath479 . otherwise , @xmath479 belongs to the interior of @xmath403 and none of the 2-contact tracings @xmath546 can support an edge of @xmath414 incident to @xmath479 . * case 2 : @xmath479 is not a corner . * * fact : * _ if there exists @xmath321 such that @xmath547 , then @xmath479 belongs to the interior of @xmath405 . _ + for each relevant 2-contact placement @xmath548 , there exists two contact points @xmath549 and @xmath550 on each side of @xmath479 at distance less or equal to @xmath1 . since @xmath479 is not a corner , neither @xmath549 nor @xmath550 is equal to @xmath479 , thus @xmath479 belongs to the relative interior of the segment @xmath551 . it follows , when @xmath547 , that @xmath479 belongs to the interior of the polygon @xmath552 ( see figure [ mmmmeeee ] ) . similarly as in the proof of proposition [ a - end - point - of - proper-2-contact - curves ] , since @xmath482 , there exists four footholds @xmath553 in @xmath497 and in some neighborhoods of @xmath554 , respectively , such that @xmath479 belongs to the interior of the polygon @xmath555 . thus , @xmath479 belongs to the interior of @xmath405 . hence , if there exists @xmath321 such that @xmath547 , none of the 2-contact tracings @xmath546 can support an edge of @xmath414 incident to @xmath479 . we now assume that @xmath556 , @xmath334 . * fact : * _ there are at most six 2-contact tracings incident to @xmath479_. + the general position hypothesis h2 forbid @xmath29-contacts for @xmath469 , thus @xmath479 corresponds to a 3-contact placement . the three possible choices of two contacts among three , give three 2-contact tracing intersecting in @xmath479 and thus , six arcs incident to @xmath479 . * fact : * _ there are three 2-contact tracings incident to @xmath479_. + if the 3-contact placement @xmath557 is of type ( corner - endpoint , @xmath468 ) , then there are only three 2-contact tracings incident to @xmath479 , that are two circular arcs and one line segment . otherwise , it comes from the general position hypotheses h1 , h2 and h3 ( designed to ensure that property ) that a 2-contact tracing can not be valid on both side of the 3-contact , i.e. , on one side of the 3-contact placement , the placements are not free . the proof that the hypotheses ensured that fact is detailed in @xcite . * fact : * _ there are two relevant 2-contact tracings incident to @xmath479_. + since @xmath479 is not a corner , at the 3-contact placement @xmath557 , two contact points are on the same side of @xmath479 . thus , only two of the three 2-contact tracings incident to @xmath479 are relevant . now , consider the graph @xmath503 and each node @xmath479 in turn . if @xmath558 , we disconnect all the edges of @xmath503 that end at @xmath479 . notice that for each such node @xmath479 , we know , by theorem [ thae ] , whether @xmath522 and , in such a case , the labels of the edges of @xmath414 incident to @xmath479 . if @xmath482 , we disconnect the edges ending at @xmath479 except those ( at most two ) that may support @xmath414 in a neighborhood of @xmath479 ( see proposition [ at - most - two - edges ] ) . in this way , we obtain a graph @xmath504 such that the degree of each node is one or two . we consider each connected component of this new graph as a curve . let @xmath480 be this set of curves . these curves are represented in @xmath504 as chains ( open or closed ) . it follows that , even if a curve is not simple , there exists a natural order along the curve . then , according to propositions [ a - end - point - of - proper-2-contact - curves ] and [ at - most - two - edges ] , we get the following theorem : [ thc ] we can compute , in @xmath461 time using @xmath462 space , a set @xmath480 of curves that support the edges of @xmath414corresponding to placements at the limit of stability of the spider robot . moreover , any portion @xmath559 of @xmath414 either intersects @xmath411or belongs to a unique curve of @xmath480 . we can now construct @xmath403 and @xmath12 . let @xmath560 denote the maximum length of the davenport - schinzel sequence of order @xmath29 on @xmath5 symbols and @xmath561 . note that @xmath562 . [ thm - tri ] given , as foothold regions , a set of @xmath5 non intersecting straight line segments that satisfies hypotheses h1 , h2 , h3 and h4 , we can compute the free space @xmath403 of the spider robot in @xmath563 time using @xmath456 space . by theorem [ thae ] , we can compute the contribution of @xmath457 to @xmath414 and the label of the edges of @xmath414 incident to them in @xmath455 time using @xmath456 space . by theorem [ thc ] , we can compute , in @xmath461 time using @xmath462 space , a set @xmath480 of curves that support the edges of @xmath414 that do not belong to @xmath457 . moreover , any portion @xmath559 of @xmath414 such that @xmath564 belongs to a unique curve of @xmath480 . thus , by sorting all the vertices of @xmath565 on the relevant curves of @xmath480 , we obtain all the edges of @xmath414 that belong to a connected component of @xmath414 intersecting @xmath457 . indeed , for each vertex @xmath566 , we know , in a neighborhood of @xmath479 , the portion of the curve of @xmath480 that belongs to @xmath414 because we can simply determine , for each edge , a side of the edge that belongs to @xmath405 ( the contact points corresponding to the edges determine a side that necessarily belongs to @xmath405 ) , its two sides belong to @xmath405 . ] . then , it is an easy task to deduce all the connected components of @xmath414 that intersect @xmath457 . it remains to compute the connected components of @xmath414 that do not intersect @xmath457 . each of these components must be a closed curve of @xmath480 . moreover , all the curves of @xmath480 belong to @xmath405 . thus , according to theorem [ thc ] , any closed curve @xmath473 of @xmath480 that does not intersect @xmath411 is either a connected component of @xmath414 or is strictly included in @xmath405 . therefore , by considering , in addition , all the closed curves of @xmath480 that do not intersect @xmath457 , we finally obtain a set @xmath567 of closed curves that contains @xmath414 and such that any curve of @xmath567 is either a connected component of @xmath414 or is strictly included in @xmath405 . at last , as we can simply determine , for each curve of @xmath567 , a side of the edge that belongs to @xmath405 , we can easily deduce from @xmath567 the free space @xmath405 . that concludes the proof since all these computations can be done in @xmath563 time using @xmath456 space . as we said at the beginning of section [ polygonal - foothold - regions ] , the free space of the spider robot using as foothold regions a set of polygonal regions is obtained by adding these polygonal regions to @xmath405 . this does not increase the geometric complexity of the free space nor the complexity of the computation . thus , we get the following theorem : [ er ] given a set of pairwise disjoint polygonal foothold regions with @xmath5 edges in total that satisfies hypotheses h1 , h2 , h3 and h4 , we can compute the free space @xmath0 of the spider robot in @xmath563 time using @xmath456 space . the function @xmath10 is extremely slowly growing and can be considered as a small constant in practical situations . this result is almost optimal since , as shown in @xcite , @xmath568 is a lower bound for the size of @xmath12 . we have seen in theorem [ r ] that , when the foothold regions are @xmath5 points in the plane , the free space of the spider robot can be computed in @xmath50 time using @xmath394 space where @xmath6 is the pseudo inverse of the ackerman s function and @xmath24the arrangement of the @xmath5 circles of radius @xmath1 centered at the footholds . by @xcite the size of @xmath0 is known to be @xmath569 . the size of @xmath24 is @xmath570 but it has been shown in @xcite that @xmath571 , where @xmath29 denotes the maximum number of disks of radius @xmath1 centered at the footholds that can cover a point of the plane . thus , in case of sparse footholds , the sizes of @xmath24 and @xmath0 are linearly related to the number of footholds . when the foothold regions are polygons with @xmath5 edges in total , the free space of the spider robot can be computed in @xmath563 time using @xmath456 space , where @xmath572 is the maximum length of a davenport - schinzel sequence of order @xmath29 on @xmath5 symbols , and @xmath411 is the arrangement of the @xmath5 curves consisting of the points lying at distance @xmath1 from the straight line edges . note that the size of @xmath411 is @xmath570 . it should be observed that , in the case of point footholds , our algorithm implies that @xmath394 is an upper bound for @xmath573 . however , this bound is not tight since @xmath49 @xcite . in the case of polygonal footholds , our analysis implies that @xmath456 is an upper bound for @xmath573 . we leave as an open problem to close the ( small ) gap between this upper bound and the @xmath568 lower bound . once the free space @xmath0 is known , several questions can be answered . in particular , given two points in the same connected component of @xmath12 , the algorithm in @xcite computes a motion of the spider robot , i.e. , a motion of the body and a corresponding sequence of leg assignments that allows the robot to move from one point to the other . the motion planning problem for other types of legged robots remains to be studied . the case where all the legs are not attached at the same point on a polygonal / polyhedral body is particularly relevant . a spider robot for which all the legs are not of the same length is also an interesting model . we would like to thank joseph orourke for helpful comments .	we study the problem of computing the free space @xmath0 of a simple legged robot called the spider robot . the body of this robot is a single point and the legs are attached to the body . the robot is subject to two constraints : each leg has a maximal extension @xmath1 ( accessibility constraint ) and the body of the robot must lie above the convex hull of its feet ( stability constraint ) . moreover , the robot can only put its feet on some regions , called the foothold regions . the free space @xmath2 is the set of positions of the body of the robot such that there exists a set of accessible footholds for which the robot is stable . we present an efficient algorithm that computes @xmath0 in @xmath3 time using @xmath4 space for @xmath5 discrete point footholds where @xmath6 is an extremely slowly growing function ( @xmath7 for any practical value of @xmath5 ) . we also present an algorithm for computing @xmath0 when the foothold regions are pairwise disjoint polygons with @xmath5 edges in total . this algorithm computes @xmath0 in @xmath8 time using @xmath9 space ( @xmath10 is also an extremely slowly growing function ) . these results are close to optimal since @xmath11 is a lower bound for the size of @xmath12 . legged robots , computational geometry , motion planning 68u05
since the observation of the current cosmic expansion speed up through the light - curve of type ia supernovae @xcite , which was confirmed independently by cosmic microwave background ( cmb ) @xcite and large scale structures observations @xcite , it became clear that a pure friedmann - lemaitre - robertson - walker ( flrw ) cosmology with matter and radiation could not explain all the large scale properties of our universe ( for reviews see @xcite ) . the missing piece was called dark energy and , surprisingly , a single constant , the once despised cosmological constant @xmath0 , was capable of describing its behavior with great success ; being now part of the standard @xmath0cdm model @xcite . however , the curious smallness of its energy density ( @xmath1 ) and its lack of correspondence with fundamental physics has raised many questions , and many alternative models have appeared@xcite . while the cosmological constant always leads to an equation of state ( eos ) parameter for the vacuum with constant value @xmath2 ( @xmath3 ) , in alternative models its value can vary with time and can be strictly @xmath4 ( e.g. , quintessence @xcite ) , strictly @xmath5 ( phantom energy @xcite ) or cross from one region to another depending on time ( e.g. , quintom @xcite ) . moreover , the possibility of nonlinear eos s is also under current research @xcite . although they are quite different , all these approaches assume that dark energy is isotropic . the possibility of anisotropic dark energy has just started to be evaluated @xcite . in this work , motivated by the anomalies found in the cmb anisotropies @xcite , which appear to indicate violation of statistical isotropy @xcite , and on the increasing interest on bianchi cosmologies ( for a recent review see @xcite ) , we analyze a nondynamical anisotropic form of dark energy which is generated by the usual cosmological constant @xmath0 in combination with a certain parameter @xmath6 . to this end we employ the bianchi type i line element as the space - time background . a null @xmath6 corresponds to the usual cosmological constant and hence leads to the usual isotropic cosmology due to the cosmic no - hair theorem @xcite . being @xmath6 non - null , the angular dependence of the hubble parameter does not fade away as the universe expands ; opening the possibility of generating an eccentricity in the last scattering surface @xcite . this is achieved without the need of others structures and evades problems with isotropization during inflation or during the current epoch dominated by dark energy . the price " of this interesting behavior is violation of the null energy condition ( nec ) ; in particular , this type of dark energy crosses the phantom divide depending on the direction . some different approaches to cosmology with different types of noncommutativity have appeared @xcite , these are commonly motivated by general quantum - gravity expectations and on the possibility of finding traces of it in the cmb anisotropies @xcite , which are infrared ( ir ) consequences of the original ultraviolet ( uv ) modification @xcite . on the other hand , except for a small fraction of a second , all the universe history is classical and the formation of the last scattering surface occurred when the universe had @xmath7 of its current age ; thus the possibility and consequences of an infrared noncommutativity " @xcite , i.e. , a direct modification of the classical space - time canonical structure , should also be verified . here we show that , in the context of a bianchi i cosmology , the most simple poisson structure deformation that does not violate rescaling of the scale factors evolves in time and admits a natural interpretation as describing a new form of dark energy , precisely the one generated by the anisotropic cosmological constant @xmath8 , where @xmath6 regulates the deformation magnitude . in the following section we introduce the poisson structure deformation in the bianchi i cosmology , unveiling some of its general features . in sec . iii an axial symmetry is introduced and exact solutions are presented . from the latter , considering the vacuum eos experimental value and limits on the cmb temperature anisotropies , upper bounds for @xmath9 are found . consistently with these bounds , the @xmath6 value is estimated in order to generate an appropriate eccentricity in the last scattering surface capable of solving the quadrupole anomaly @xcite . in the last section we present our conclusions and perspectives . the general relativity action in 4d space - time is given by @xmath10 \sqrt{- g } ~ d^4 x,\ ] ] where @xmath11 , @xmath12 is the ricci scalar , @xmath0 is the cosmological constant , @xmath13 is the metric determinant and @xmath14 the lagrangian for the matter part . the flrw cosmology is an application of general relativity which starts from the very beginning with the assumption of spacial homogeneity and isotropy . in a flat universe this leads to the well known robertson - walker metric @xmath15 . bianchi classified a broader class of cosmologies which are homogeneous but not necessarily isotropic @xcite . in the presence of a positive cosmological constant all the nine bianchi cosmologies ( or eight of the nine depending on curvature conditions on the bianchi ix ) isotropize themselves and become de sitter spaces asymptotically @xcite . to provide the framework for analyzing cosmological deviations from isotropy , we will use the bianchi i line element , which reads @xmath16 we modify the cosmological classical evolution , avoiding late - time isotropization , by deforming the poisson structure between the scale factors conjugate momenta . this procedure can be pictured as the ir analogue of the noncommutativity between the scale factors , the latter was explored in @xcite . the deformation of the poisson structure to achieve an ir noncommutativity " has already been explored in other contexts @xcite . we consider the following relation @xmath17 where @xmath6 is a real constant , @xmath18 , @xmath19 , @xmath20 , @xmath21 , @xmath22 and @xmath23 is the @xmath24 conjugate canonical momentum : @xmath25 , @xmath26 , @xmath27 , where @xmath28 . considering as usual @xmath29 = m^{-1}$ ] and @xmath30 = 1 $ ] , then @xmath31 = 1 $ ] and @xmath32 = 1 $ ] . on the other hand , for dimensional analyzes it is more useful to adopt the convention @xmath33 = m^{-1}$ ] , @xmath34 = 1 $ ] and @xmath30 = m^{-1}$ ] , which implies @xmath31 = m$ ] and @xmath32 = m^3 $ ] ; showing that @xmath6 should indeed be regarded as an ir deformation . independently on the convention on dimensions one selects ( but preserving @xmath35 = m^{-2}$ ] ) , the only possible way to express @xmath6 in terms of a dimensionless constant @xmath36 , without inserting new dimensional parameters , is @xmath37 . in the equations of motion and derivative relations , @xmath6 only occurs in the form @xmath38 . considering that @xmath30 = m^{-1}$ ] , @xmath39 and @xmath40 have the same dimension . hereafter the tilde will not be used . ( [ nc ] ) should not be interpreted as a poisson structure deformation induced by @xmath41 = i b^{ij}$ ] , since the latter would insert an improper quantum behavior into classical observables . thus , by implementing ( [ nc ] ) into the cosmological dynamics , we are not ignoring higher powers of @xmath6 , that is the full deformation . one important property of ( [ nc ] ) is its invariance by the rescaling of the scale factors [ i.e. , @xmath42 when @xmath43 ( no sum , the @xmath44 s are non - null constants ) ] . a constant @xmath45 , for instance , violates this symmetry . this violation would turn the scale factors into physical observables , in particular it would not be possible to select coordinates such that @xmath46 at a given time . the equations of motion of a bianchi i cosmology with matter , cosmological constant and the poisson deformation @xmath47 read @xmath48 \frac { \ddot a } a + \frac { \ddot c}c + h_a \ ; h_c & = & \lambda + b_0 ~ \sqrt \lambda ~ \left ( h_a - h_c \right ) - \frac 1 k t_2 ^ 2 , \\[.1 in ] \frac { \ddot a } a + \frac { \ddot b } b + h_a \ ; h_b & = & \lambda + b_0 ~ \sqrt \lambda \left ( h_b - h_a \right ) - \frac 1 k t_3 ^ 3 , \\[.1 in ] h_a \ ; h_b + h_a \ ; h_c & + & h_b \ ; h_c = \lambda - \frac 1 k t_0 ^ 0,\end{aligned}\ ] ] where @xmath49 is the matter energy - momentum tensor and @xmath50 is the hubble parameter in the @xmath51 direction . the vacuum energy momentum - tensor @xmath52 now depends on @xmath0 and @xmath36 . the vacuum energy is given by the former , while the anisotropic vacuum pressure depends on both , @xmath36 parametrize the pressure anisotropies . the eos parameter is anisotropic , but its mean value is exactly @xmath2 . + for the bianchi i line element , the zero - component of any diagonal energy - momentum tensor divergence is @xmath53 this expression is trivially null if @xmath54 and @xmath55 , but it also admits an anisotropic pressure solution , @xmath56 where @xmath57 is a constant . setting @xmath58 , the above pressures are exactly the vacuum pressures that come from the bianchi i cosmology with the poisson deformation . this solution also satisfies @xmath59 . regarding lorentz invariance violation , we consider that the relation ( [ nc ] ) holds in a certain orientation at the cosmic preferential referential ( the one which the cmb looks maximal isotropic ) and that observer lorentz transformations are not violated @xcite ; so the anisotropic observables transform normally under rotations . in particular , the pressure in an arbitrary direction , given by the angles @xmath60 and @xmath61 ( usual definitions ) , is found from a rotation of the vacuum energy - momentum tensor and reads @xmath62 so the mean equation of state parameter is @xmath63 in the following section the symmetries of the above model will be augmented by adding an axial symmetry , thus proceeding we find exact solutions and estimate the value of @xmath6 . recently the bianchi i cosmology with axial symmetry received much attention , particularly due to evidences that it provides a reasonable framework to solve cmb large - angle anomalies @xcite . to implement this symmetry , we use @xmath64 , @xmath65 and @xmath66 , therefore @xmath67 \label{eqma2 } \frac { \ddot a } a + \frac { \ddot b } b + h_a ~ h_b & = & \lambda + b_0 ~ \sqrt \lambda ~ h_a - \frac 1k t_2 ^ 2 , \\[.1 in ] \label{eqma3 } 2 ~ h_a ~ h_b + h_b^2 & = & \lambda - \frac 1k t_0 ^ 0.\end{aligned}\ ] ] for the vacuum ( with @xmath68 ) , the above equations admit a simple exact solution which reads @xmath69 \label{bvac } b & = & b_0 ~ e^ { \sqrt { \frac { \tilde \lambda } 3 } ~ \ ( 1 ~ - ~ \tilde b_0 \ ) ~ t } , \end{aligned}\ ] ] where @xmath70 , @xmath71 and @xmath72 and @xmath73 are constants with the suitable dimension . in the limit @xmath74 , we recover the usual de sitter space solution ( @xmath75 ) . the solutions ( [ avac ] , [ bvac ] ) are expected to be good approximations to the universe dynamics only for @xmath76 . we now consider the presence of barionic and cold dark matter with matter density @xmath77 , where @xmath78 is a constant . the @xmath79 dependence on the scale factors comes from @xmath80 . the exact solutions read @xmath81 \label{bmat } b(t ) & = & a_0 e^ { - \frac 13 b_0 ~ \sqrt \lambda ~ ( t - t _ * ) } ~ \sinh^{\frac 23 } \ ( \frac 32 ~\sqrt{\frac { \tilde \lambda } 3 } ~ t \).\end{aligned}\ ] ] @xmath82 is an arbitrary constant . the corresponding flrw solution is recovered in the limit @xmath74 [ @xmath83@xcite ] . a natural choice is to set @xmath84 , the present time , so @xmath85 . + the scale factor solution for the flrw cosmology , in the presence of a cosmological constant @xmath86 , describes an average behavior of the proposed anisotropic universe in the following sense @xmath87 hence @xmath88 thus we set the @xmath86 value to the experimental cosmological constant value ( @xmath89 @xcite ) . wmap experimental results assert that @xmath90 @xcite . this imposes an upper bound for @xmath91 in the axial case . from the eqs . ( [ ptp ] [ eqma1 ] [ eqma2 ] [ eqma3 ] [ avac ] [ bvac ] ) , @xmath92 therefore we find the modest upper bound @xmath93 . the universe spacial anisotropy can be quantified by the shear anisotropic parameter @xcite , @xmath94 @xmath95 , contrary to the usual approach to anisotropic cosmologies @xcite , increases with time ( avoiding isotropization ) and it has a maximum given by @xmath96 . cosmologies with a similar behavior were recently studied in ref . @xcite . the cmb temperature anisotropies put strong bounds on anisotropic cosmologies since , at most , @xmath97 . @xmath98 is the mean measured cmb temperature , which , in the flrw cosmology , is related with the mean original temperature @xmath99 by @xmath100 ( for @xmath101 ) . in the model here proposed the space - time background is a source of temperature anisotropies , so @xmath102 , where @xmath103 is the variation solely due to the background . for the preferential ( @xmath104-axis ) direction , @xmath105 , where @xmath106 is the original temperature of the cmb photons in that direction . we consider that the anisotropic effects were not significant for the universe dynamics by the decoupling time ( in accordance with @xmath107 ) , so the lss original temperature is almost the same in all directions . from eq . ( [ amed ] ) , the relation @xmath100 also holds in the anisotropic case as a mean value , thus we set @xmath108 and find @xmath109 since the above expression should be of the same order or lesser than @xmath110 , we find @xmath111 . the x - axis is the direction with greater redshift variation from the mean one for @xmath112 ; for negative @xmath36 this upper bound analysis is done in a perpendicular direction and leads to the above bound . refer to the following scale factors : @xmath113 ( dot - dashed ) , @xmath114 ( dashed ) and @xmath115 ( solid ) . the other solid line is the eccentricity ( @xmath116 for @xmath117 and @xmath118 for @xmath119 ) . the other dashed line is @xmath120 . plotted with @xmath121 , @xmath122 , @xmath123 , @xmath124 . the horizontal axis is the time in @xmath125 unities . ] recently it was shown that a non - null eccentricity solves the quadrupole power anomaly of the cmb spectrum if the eccentricity of the last scattering surface ( lss ) is @xmath126 @xcite ( considering a system of coordinates with null eccentricity at present time ) . the present model describes a universe with eccentricity given by @xmath127 \sqrt{1 - \ ( \frac b a \)^2 } = \sqrt{1 - e^{2 b_0 \sqrt \lambda ~ ( t _ * - t ) } } , & \mbox{\small for $ b_0 ( t - t _ * ) \ge 0 $ } \end{array } \right . \nonumber\end{aligned}\ ] ] within the approach of @xcite we set @xmath84 , where @xmath125 is the age of the universe , and find @xmath128 . this value is in agreement with the previous upper bounds values . in accordance with ref . @xcite , @xmath129 leads to the upper bound on the eccentricity , @xmath130 . using this eccentricity value , the same bound @xmath111 is found . to second order on the eccentricity , the background alone induces the following contribution that appears here eliminates the monopole contribution and fixes in what directions @xmath131 is null . this term is deduced in the appendix and is in conformity with the previous @xmath131 analysis , @xmath132 . ] @xcite , @xmath133 where @xmath134 and @xmath135 is the projection of @xmath136 into the axis of symmetry . in the simplest case in which the axis of symmetry coincides with the @xmath104-axis , we write @xmath137 , so @xmath138 depends only on @xmath139 . it is easy to see that @xmath140 , implying that this elliptical background , to first order on @xmath141 , modifies the quadrupole but no other multipole moment . for a coordinate system not aligned with the axis of symmetry , @xmath142 is written as a linear combination of @xmath143 , thus the previous conclusion remains valid . in order to evaluate the influence of the elliptical background to the total temperature fluctuations @xmath144 , following @xcite , we employ@xcite @xmath145 where @xmath146 refers to the usual statistically isotropic contribution ( i.e. , it satisfies @xmath147 ) . consequently , @xmath148 and the total multipole power @xmath149 differs from the statistically isotropic one only for @xmath150 . this exclusive quadrupole modification , as implied by ( [ gtr ] , [ ans_sta ] ) , guarantees no changes to the power of others multipoles . this is welcome since others large - angle multipole powers found from the @xmath0cdm model have good agreement with experiment . on the other hand , it does not help on the understanding of another cmb quadrupole anomaly , namely its alignment with the octopole@xcite . @xcite introduces a vector @xmath151 for each multipole as the direction @xmath152 in which @xmath153 is maximum refers to the observed @xmath154 multipole moment computed with the @xmath155 axis in the @xmath152 direction . ] . it was shown that @xmath156 ( see also @xcite ) . the above presented results for an elliptical universe induced by an anisotropic cosmological constant rely on the first order on @xmath141 approximation , so _ a priori _ there is hope of solving the alignment if higher powers on the eccentricity are considered . the problem with this approach is that @xmath157 and the alignment is evident from the first digits of @xmath158 and @xmath159 , so a new very small contribution is not helpful . in the appendix we comment more on the extension of the previous analyses to higher eccentricity powers . currently it is unclear whether the cmb large - angle anomalies originate from some unknown systematic error ( present in both the cobe and wmap data ) or if they have a physical origin . the latter points toward new physics and , in particular , to small deviations of the @xmath0cdm model such that statistical isotropy is violated @xcite . in this work we proposed a @xmath0cdm model extension whose dark energy component preserves its nondynamical character @xmath160 but wield a vacuum anisotropic pressure . we presented exact solutions for the cosmological scale factors ( [ avac ] , [ bvac ] , [ amat ] , [ bmat ] ) ; upper bounds on the anisotropy magnitude @xmath36 , considering the experimental vacuum eos parameter and the smallness of the cmb temperature anisotropies ( @xmath161 ) ; and lastly the @xmath36 value was estimated to be @xmath128 , consistently with the previous bounds and in accordance with the quadrupole anomaly solution in an elliptical universe @xcite . the axial form of the anisotropic cosmological constant here proposed , to second order on the eccentricity , only modifies the quadrupole coefficients and is capable of improving the experimental concordance for the quadrupole power . for higher orders on the eccentricity , as commented in the appendix , the influence to others multipoles is model dependent , but such extension can not induce the observed quadrupole - octopole alignment . in particular , for the presented model , the dipole and the octopole deviations from the @xmath0cdm predictions are exactly null . it is difficult to conceive a single physical mechanism capable of solving all the cmb anomalies , imprints of physics beyond the @xmath0cdm model probably are related with a fraction of them . the presented cosmological model can be achieved by two equivalent ways from a bianchi i cosmology with cosmological constant @xmath0 : @xmath162 ) by considering the most straightforward implementation of anisotropic vacuum pressure consistent with energy - momentum tensor conservation ( [ div0 ] , [ amomenta ] ) ; @xmath163 ) by implementing a poisson structure deformation between canonical momenta such that rescaling of the scale factors is not violated ( [ nc ] , [ nc2 ] ) . the latter procedure can be pictured as a kind of ir analogue of the scale factor noncommutativity @xcite , it has already been explored in other contexts with constant deformation parameter @xcite . bianchi cosmologies in the presence of a positive cosmological constant usually isotropize themselves with the universe expansion . this property was originally welcomed and motivated an interesting discussion on the naturalness of the universe initial conditions @xcite , in particular it opened the possibility that the universe has started under more general conditions than those given by the flrw cosmology and latter evolved to it . however , as above commented , imprints of some anisotropic ( bianchi ) cosmology were found in the cmb spectrum ; thus , _ a priori _ , being these of true cosmological nature , inflation and/or dark energy need to be fine tuned to do not wipe out the anisotropic imprints . this issue emerge for instance when considering primordial magnetic fields as a source of current cosmological anisotropies @xcite . here we employed an anisotropic model outside the flrw equivalent class " ( i.e. , not in conformity with the cosmic no - hair theorem conditions @xcite ) and it was shown that such model is viable and can also solve the quadrupole power anomaly . while this paper was being prepared , some similar ideas appeared in ref . @xcite . this recently explored approach in which the universe anisotropy after inflation increases with time ( e.g. , @xcite ) is not only consistent with observations , it also improves the concordance with experiment and suggest other interpretations for dark energy . new phenomenological anisotropic alternatives for the dark energy modeling might , we hope , inspire and lead to more fundamental models . * acknowledgments * the author would like to thank jorge gamboa for suggesting the application of an ir noncommutativity to cosmology , for useful discussions and for the kind reception in chile . the author also thanks fernando mndez , clvis wotzasek and marcelo s. guimares for useful discussions . this work was supported by fondecyt - chile grant n. 3070008 . here we derive eq.([gtr ] ) and comment on the background influence to others multipoles if higher eccentricity powers are considered . the axisymmetric bianchi i line element can be written in terms of an anisotropic perturbation @xmath164 and a conformal scale factor @xmath165 as @xmath166.\ ] ] considering ( [ amed ] ) , we set @xmath167 , thus @xmath168 & & h_{22 } = h_{33 } = \ ( \frac b a \)^{\frac 23 } -1 \\ & & ~~~~~~~ = \frac 13 \mathsf{e}^2 + \frac 29 \mathsf{e}^4 + o(\mathsf{e}^6 ) , \nonumber\end{aligned}\ ] ] with @xmath169 . let @xmath170 , with @xmath171 . @xmath172 is traceless ( and therefore in accordance with the standard tensorial perturbation theory @xcite ) . some redefinitions are needed in order to eliminate the traceful part of , they read : @xmath173 , @xmath174 , @xmath175 . the @xmath176 redefinition is necessary to preserve @xmath177 . @xmath178 indeed works as a good mean scale factor , but , as implied by eq . ( [ amed ] ) , it is not an exact relation , a small time ( or @xmath0 ) redefinition is also necessary ] @xmath179 . in the following we analyze the background influence to the dipole and octopole if higher eccentricity powers are considered . in fact , the background alone contribution to these multipoles is exactly null . to any order on the eccentricity , consider a system of coordinates in which the @xmath104-axis coincides with the preferential direction , hence @xmath182 has even parity symmetry in the interval @xmath183 $ ] ( induced by the exact elliptical symmetry ) and one concludes that @xmath184 rotations do not change this conclusion . any odd @xmath185 modification in the multipole moments induced by the background comes from the eq . ( [ ans_sta ] ) extension to higher eccentricity powers , which depends on how the gaussian fluctuations interact with the anisotropic perturbation . in the presented approach , since the source of anisotropies is non - dynamical , the linearity of eq . ( [ ans_sta ] ) is preserved and , in particular , there is no change to the dipole or the octopole . one may extend the presented approach in order to include the possibility of dark energy perturbations . in that case , since both @xmath186 and @xmath182 are roughly of the same order , the interaction should be properly found by an evaluation of a second order sachs - wolfe effect that considers both the anisotropic and the isotropic perturbations together . the second order sachs - wolfe effect of the isotropic " part alone is under current research , in particular it induces non - gaussianity ( e.g. , @xcite ) . the details of this issue are outside this paper purposes , we only add that , up to @xmath187 ( second order ) , the eq . ( [ gtr ] ) should have the _ aspect _ @xmath188 , leading to @xmath189 . hence , background contributions to the dipole and octopole are expected if the mentioned interaction is added becomes null for @xmath190 , generating no contribution to the cmb microphysics , as expected . ] . this tiny influence is not related with the quadrupole - octopole alignment , nonetheless future high precision cmb observations should be sensible to such influence . 99 a. g. riess _ et al . _ [ supernova search team collaboration ] , `` observational evidence from supernovae for an accelerating universe and a cosmological constant , '' astron . j. * 116 * , 1009 ( 1998 ) [ arxiv : astro - ph/9805201 ] . c. b. netterfield _ et al . _ [ boomerang collaboration ] , `` a measurement by boomerang of multiple peaks in the angular power spectrum of the cosmic microwave background , '' astrophys . j. * 571 * , 604 ( 2002 ) [ arxiv : astro - ph/0104460 ] . d. n. spergel _ et al . _ [ wmap collaboration],``wilkinson microwave anisotropy probe ( wmap ) three year results : implications for cosmology , '' astrophys . j. suppl . * 170 * , 377 ( 2007 ) [ arxiv : astro - ph/0603449 ] . v. sahni and a. a. starobinsky , `` the case for a positive cosmological lambda - term , '' int . j. mod . d * 9 * , 373 ( 2000 ) [ arxiv : astro - ph/9904398 ] ; v. sahni,``dark matter and dark energy , '' lect . notes phys . * 653 * , 141 ( 2004 ) [ arxiv : astro - ph/0403324 ] ; s. m. carroll,``the cosmological constant , '' living rev . rel . * 4 * , 1 ( 2001 ) [ arxiv : astro - ph/0004075 ] ; l. perivolaropoulos , `` accelerating universe : observational status and theoretical implications , '' aip conf . proc . * 848 * , 698 ( 2006 ) [ arxiv : astro - ph/0601014 ] ; e. j. copeland , m. sami and s. tsujikawa,``dynamics of dark energy , '' int . j. mod . d * 15 * , 1753 ( 2006 ) [ arxiv : hep - th/0603057 ] ; t. padmanabhan , `` dark energy and gravity , '' arxiv:0705.2533 [ gr - qc ] ; m. s. turner and d. huterer,``cosmic acceleration , dark energy and fundamental physics , '' arxiv:0706.2186 [ astro - ph ] . m. zaldarriaga , `` introduction to cmb anisotropies , '' www.cfa.harvard.edu/@xmath191mzaldarr/astro202b/lectures.pdf ; w. hu and s. dodelson,``cosmic microwave background anisotropies , '' ann . rev . astrophys . * 40 * , 171 ( 2002 ) [ arxiv : astro - ph/0110414 ] . b. feng , x. l. wang and x. m. zhang , `` dark energy constraints from the cosmic age and supernova , '' phys . b * 607 * , 35 ( 2005 ) [ arxiv : astro - ph/0404224 ] ; z. k. guo , y. s. piao , x. m. zhang and y. z. zhang , `` cosmological evolution of a quintom model of dark energy , '' phys . b * 608 * , 177 ( 2005 ) [ arxiv : astro - ph/0410654 ] . s. nojiri and s. d. odintsov , `` inhomogeneous equation of state of the universe : phantom era , future singularity and crossing the phantom barrier , '' phys . d * 72 * , 023003 ( 2005 ) [ arxiv : hep - th/0505215 ] ; h. stefancic , `` expansion around the vacuum equation of state : sudden future singularities and asymptotic behavior , '' phys . rev . d * 71 * , 084024 ( 2005 ) [ arxiv : astro - ph/0411630 ] ; d. f. mota and c. van de bruck , `` on the spherical collapse model in dark energy cosmologies , '' astron . astrophys . * 421 * , 71 ( 2004 ) [ arxiv : astro - ph/0401504 ] . t. koivisto and d. f. mota , `` dark energy anisotropic stress and large scale structure formation , '' phys . rev . d * 73 * , 083502 ( 2006 ) [ arxiv : astro - ph/0512135 ] ; r. a. battye and a. moss , `` anisotropic perturbations due to dark energy , '' phys . d * 74 * , 041301 ( 2006 ) [ arxiv : astro - ph/0602377 ] ; d. f. mota , j. r. kristiansen , t. koivisto and n. e. groeneboom,``constraining dark energy anisotropic stress , '' arxiv:0708.0830 [ astro - ph ] . ralston and p. jain,``the virgo alignment puzzle in propagation of radiation on cosmological scales , '' int . j. mod . d * 13 * , 1857 ( 2004 ) [ arxiv : astro - ph/0311430 ] ; d. j. schwarz , g. d. starkman , d. huterer and c. j. copi , `` is the low - l microwave background cosmic ? , '' phys . lett . * 93 * , 221301 ( 2004 ) [ arxiv : astro - ph/0403353 ] . c. j. copi , d. huterer and g. d. starkman , `` multipole vectors a new representation of the cmb sky and evidence for statistical anisotropy or non - gaussianity at @xmath192 , '' phys . rev . d * 70 * , 043515 ( 2004 ) [ arxiv : astro - ph/0310511 ] ; c. j. copi , d. huterer , d. j. schwarz and g. d. starkman , `` on the large - angle anomalies of the microwave sky , '' mon . not . * 367 * , 79 ( 2006 ) [ arxiv : astro - ph/0508047 ] ; c. copi , d. huterer , d. schwarz and g. starkman,``the uncorrelated universe : statistical anisotropy and the vanishing angular correlation function in wmap years 1 - 3 , '' phys . rev . d * 75 * , 023507 ( 2007 ) [ arxiv : astro - ph/0605135 ] . h. k. eriksen , f. k. hansen , a. j. banday , k. m. gorski and p. b. lilje , `` asymmetries in the cmb anisotropy field , '' astrophys . j. * 605 * , 14 ( 2004 ) [ erratum - ibid . * 609 * , 1198 ( 2004 ) ] [ arxiv : astro - ph/0307507 ] ; h. k. eriksen , a. j. banday , k. m. gorski , f. k. hansen and p. b. lilje , `` hemispherical power asymmetry in the three - year wilkinson microwave anisotropy probe sky maps , '' astrophys . j. * 660 * , l81 ( 2007 ) [ arxiv : astro - ph/0701089 ] . k. land and j. magueijo , `` the axis of evil , '' phys . * 95 * , 071301 ( 2005 ) [ arxiv : astro - ph/0502237 ] ; `` is the universe odd ? , '' phys . d * 72 * , 101302 ( 2005 ) [ arxiv : astro - ph/0507289 ] ; `` the axis of evil revisited , '' mon . not . soc . * 378 * , 153 ( 2007 ) [ arxiv : astro - ph/0611518 ] . l. campanelli , p. cea and l. tedesco , `` ellipsoidal universe can solve the cmb quadrupole problem , '' phys . rev . * 97 * , 131302 ( 2006 ) [ erratum - ibid . * 97 * , 209903 ( 2006 ) ] [ arxiv : astro - ph/0606266 ] ; `` cosmic microwave background quadrupole and ellipsoidal universe , '' arxiv:0706.3802 [ astro - ph ] ; p. cea,``ellipsoidal universe induces large scale cmb polarization , '' arxiv : astro - ph/0702293 . m. visser,``scale anomalies imply violation of the averaged null energy condition , '' phys . b * 349 * , 443 ( 1995 ) [ arxiv : gr - qc/9409043 ] ; s. nojiri and s. d. odintsov , `` effective equation of state and energy conditions in phantom / tachyon inflationary cosmology perturbed by quantum effects , '' phys . b * 571 * , 1 ( 2003 ) [ arxiv : hep - th/0306212 ] ; i. y. arefeva and i. v. volovich , `` on the null energy condition and cosmology , '' arxiv : hep - th/0612098 . r. brandenberger and p. m. ho , `` noncommutative spacetime , stringy spacetime uncertainty principle , and density fluctuations , '' phys . rev . d * 66 * ( 2002 ) 023517 [ aapps bull . * 12n1 * ( 2002 ) 10 ] [ arxiv : hep - th/0203119 ] ; f. lizzi , g. mangano , g. miele and m. peloso,``cosmological perturbations and short distance physics from noncommutative geometry , '' jhep * 0206 * , 049 ( 2002 ) [ arxiv : hep - th/0203099 ] ; s. tsujikawa , r. maartens and r. brandenberger,``non - commutative inflation and the cmb , '' phys . lett . b * 574 * , 141 ( 2003 ) [ arxiv : astro - ph/0308169 ] ; g. calcagni and s. tsujikawa,``observational constraints on patch inflation in noncommutative spacetime , '' phys . d * 70 * , 103514 ( 2004 ) [ arxiv : astro - ph/0407543 ] . h. garcia - compean , o. obregon and c. ramirez,``noncommutative quantum cosmology , '' phys . lett . * 88 * , 161301 ( 2002 ) [ arxiv : hep - th/0107250 ] ; b. vakili , n. khosravi and h. r. sepangi,``bianchi spacetimes in noncommutative phase - space , '' class . * 24 * , 931 ( 2007 ) [ arxiv : gr - qc/0701075 ] ; n. khosravi , s. jalalzadeh and h. r. sepangi,``non - commutative multi - dimensional cosmology , '' jhep * 0601 * , 134 ( 2006 ) [ arxiv : hep - th/0601116 ] ; `` quantum noncommutative multidimensional cosmology , '' gen . grav . * 39 * , 899 ( 2007 ) [ arxiv : gr - qc/0702067 ] . j. gamboa and j. lopez - sarrion,``u(1 ) noncommutative gauge fields and magnetogenesis , '' phys . d * 71 * , 067702 ( 2005 ) [ arxiv : hep - th/0501034 ] ; a. das , j. gamboa , j. lopez - sarrion and f. a. schaposnik , `` gauge field theory in the infrared regime , '' phys . d * 72 * , 107702 ( 2005 ) [ arxiv : hep - th/0510002 ] ; h. falomir , j. gamboa , j. lopez - sarrion , f. mendez and a. j. da silva,``vortices , infrared effects and lorentz invariance violation , '' phys . b * 632 * , 740 ( 2006 ) [ arxiv : hep - th/0504032 ] ; j. m. carmona , j. l. cortes , a. das , j. gamboa and f. mendez , `` matter - antimatter asymmetry without departure from thermal equilibrium , '' mod . phys . lett . a * 21 * , 883 ( 2006 ) [ arxiv : hep - th/0410143 ] .	there are evidences that the cosmic microwave background ( cmb ) large - angle anomalies imply a departure from statistical isotropy and hence from the standard cosmological model . we propose a @xmath0cdm model extension whose dark energy component preserves its nondynamical character but wield anisotropic vacuum pressure . exact solutions for the cosmological scale factors are presented , upper bounds for the deformation parameter are evaluated and its value is estimated considering the elliptical universe proposal to solve the quadrupole anomaly . this model can be constructed from a bianchi i cosmology with cosmological constant from two different ways : i ) a straightforward anisotropic modification of the vacuum pressure consistently with energy - momentum conservation ; ii ) a poisson structure deformation between canonical momenta such that the dynamics remain invariant under scale factors rescalings . + + pacs numbers : 95.36.+x , 98.70.vc .
data with a matrix - valued response for each experimental unit are commonly encountered in contemporary statistical applications . for example , a longitudinal multivariate response can be treated integrally as a matrix - valued variable by designating rows and columns to be time and variates . temporal and spatial data , multivariate growth curve data , image data and data from cross - over designs also generate matrix - valued responses . for example , in a twin cross - over bioassay of insulin by the rabbit blood sugar method ( v@xmath3lund , 1980 ) , each rabbit received two different treatments on two days . blood sugar was measured at hourly intervals for six hours each day . in this case , the response for each rabbit is a @xmath4 matrix , with rows and columns indicating treatments and time respectively . the eeg data ( li et al . 2010 ) is another example that contains temporal and spatial matrix - formed variates from 77 alcoholic subjects and 45 non - alcoholic subjects . the electrical records of each subject form a matrix of dimensions @xmath5 that can be treated as a matrix - valued response variable when we investigate the association between the brain signals and alcoholism . in these examples , the components of the matrix - variates are dependent among rows and columns . this dependence distinguishes longitudinal data in a matrix - valued response from traditional longitudinal modeling of vector - valued responses in which independent units are each measured over time . vectorizing a matrix - valued response , or modeling the row or column vectors separately , typically loses dependency information and fails to capture the data structure . tackling matrix - variates directly can circumvent this issue . research into this topic has gained considerable interest in recent years . ( 2010 ) proposed a class of sufficient dimension reduction ( sdr ) methods , called dimension folding sdr , for data with matrix - valued predictors . pfeiffer et al . ( 2012 ) extended sliced inverse regression ( sir ) to longitudinal predictors . ding and cook ( 2014 ) developed model - based dimension folding methods for matrix - valued predictors . xue and yin ( 2014 ) introduced dimension folding sdr for conditional mean functions . on another track , hung and wang ( 2013 ) , zhou et al . ( 2013 ) , and zhou and li ( 2014 ) extended generalized linear models ( glm ) to matrix- and tensor - valued predictors for analyzing image data . all these methods , however , address data with matrix or tensor - valued predictors . methods for dealing directly with matrix - valued responses are relatively limited in the literature . viroli ( 2012 ) proposed special matrix - variate regressions assuming independent rows for error terms or covariates . yet in many applications such assumptions are violated as dependency can exist among both rows and columns of a matrix - valued response . we recently became aware that li and zhang ( 2015 ) studied tensor response regression where a linear relationship between a tensor - valued response and a predictor vector is considered . their method is more suited for tensor responses and vector - valued predictors . however , our motivation , rationale , model formulation , and estimation algorithms are different from those of li and zhang ( 2015 ) . firstly , our work was motivated by contemporary matrix - variate data and was geared to develop demanding yet limited matrix - variate analysis tools for such data . thus , it is tailored for matrix - valued responses and can handle matrix - valued predictors simultaneously . secondly , our method can fully utilize intrinsic matrix - variate data structure and allow a matrix linear model to reduce number of parameters in matrix - variate regressions . moreover , our method of estimation is different than that used by li and zhang ( 2015 ) . as discussed in section [ sec:7.2 ] this can have important consequences in applications . in this article , we propose matrix - variate regressions in a general framework , where the response @xmath0 on each experimental unit is a random matrix and the predictor @xmath1 can be either a scalar , a vector , or a matrix , treated as non - stochastic in terms of the conditional distribution @xmath2 . neither the rows nor the columns of the matrix - valued variables are required to be independent . thus , the proposed methods can capture intrinsic dependent data structures . moreover , they can reduce the number of parameters and improve efficiency in estimation compared to conventional methods . as matrix - variate responses may have sizable dimensions , extraneous error variation may often occur beyond that anticipated by the model . to allow for such situations , we further propose envelope methods for efficient estimation in matrix - variate regressions . by applying the idea of enveloping , one can achieve dimension reduction in the analysis , by extracting only the useful information and eliminating redundant information from estimation . this can lead to substantial efficiency gains in estimation . the remainder of this article is organized as follows . in section [ sec:2 ] we propose a new class of matrix - variate regressions and connect them with conventional regression models . section [ sec:3.1 ] reviews the idea of enveloping and illustrates it with a real example . section [ sec:3 ] is devoted to the development of envelope methods for matrix - variate regressions . section [ sec:5 ] studies theoretical properties of matrix regression models and envelopes . sections [ sec:7 ] and [ sec:8 ] are devoted to illustrations with simulations and real data analyses . technical details and proofs are included in a supplement . generally , a two - way measurement layout can be treated integrally as a matrix - valued variable , denoted as @xmath6 . in application , modeling the relationship between @xmath0 and certain covariates is often of interest , where the covariates can be matrix - valued @xmath7 , vector - valued @xmath8 , or univariate , depending on the specific data . in the insulin assay data , for example , the covariates are formed as a @xmath9 matrix with elements indicating different treatment and dose levels . the goal is to investigate how the treatment and dose levels influence blood sugar concentration , a matrix - valued response . as an obvious first step , we might use the @xmath10 operator that stacks the columns of a matrix to transform the problem into a standard vector - variate regression : @xmath11 where @xmath12 , @xmath13 , and the error vector @xmath14 has mean 0 and covariance matrix @xmath15 . in view of the potential for many parameters , this model may be hard to estimate well and difficult to interpret , unless perhaps some type of sparse structure is imposed . for problems in which a sparsity assumption is untenable , it seems to be a rather blunt instrument that neglects the matrix structure and may be grossly over - parameterized . we propose instead a relatively parsimoniously parameterized model that reflects the underlying structure of the matrix - variate response and predictor . in preparation , the following operators for matrix - valued variables are predefined . for @xmath6 , the expectation of @xmath0 is @xmath16 , the covariance matrix of @xmath0 is assumed to be of the form @xmath17 = { \mathrm{cov}}_r({\textbf{y}})\otimes { \mathrm{cov}}_c({\textbf{y}})$ ] , where ` @xmath18 ' stands for the kronecker product , and @xmath19/\mathrm{tr}[{\mathrm{cov}}_c({\textbf{y}})]$ ] and @xmath20/\mathrm{tr}[{\mathrm{cov}}_r({\textbf{y}})]$ ] are defined as the row and column covariance matrices of @xmath0 ( de waal 1985 ) . here ` tr ' means the trace operation . the two covariance matrices @xmath21 and @xmath22 are uniquely defined up to a proportionality constant . for @xmath23 , we correspondingly define the covariances between @xmath24 and @xmath25 in terms of the row covariance and column covariance , designated @xmath26 $ ] and @xmath27 $ ] . given a matrix - variate predictor @xmath28 , we define the matrix regression of @xmath6 on @xmath1 as the bilinear model @xmath29 where @xmath30 is the overall mean , and @xmath31 and @xmath32 are the row and column coefficient matrices . here @xmath33 and @xmath34 are uniquely defined only up to a proportionality constant . for identifiability , the column coefficient matrix @xmath34 is defined to have frobenius norm 1 and positive element in its first row and column . the distribution of the matrix - valued random error @xmath35 is assumed be independent of @xmath1 , and have zero mean and covariance matrix @xmath36 = { \mathrm{\boldsymbol\sigma}}_2\otimes { \mathrm{\boldsymbol\sigma}}_{1}$ ] , where @xmath37 and @xmath38 are the column and row covariance matrices of @xmath35 . similarly , we require @xmath39 to have unit frobenius norm and positive first diagonal element in order to uniquely identify the two covariance matrices @xmath40 and @xmath39 . if no constraints are required on @xmath41 and @xmath39 , the kronecker products @xmath42 and @xmath43 are still identifiable but the individual row and column parameter matrices are not . the kronecker covariance structure supposes a relational characteristic of the matrix - variate @xmath0 , as the covariances of the column vectors of @xmath35 are all proportional to @xmath40 and the covariances of the row vectors of @xmath35 are all proportional to @xmath44 . such relationships are usually desirable for matrix - valued variables , especially for multivariate repeated measures and multivariate longitudinal data because of the intrinsic relationships among elements . for instance , the eeg data contains measurements of each subject from different time ( row ) and different scalp locations ( column ) . it seems reasonable to model the data with similar variations among measurements over rows and measurements over columns . another advantage of formulating the kronecker covariance structure for such data is that the number of parameters in @xmath36 $ ] can be dramatically reduced when the matrix dimension is high . when the kronecker structure does not hold , a general covariance matrix @xmath36={\mathrm{\boldsymbol\sigma}}$ ] can be applied . hypothesis tests for the kronecker covariance structure can be found in shitan and brockwell ( 1995 ) , lu and zimmerman ( 2005 ) , and roy and khattree ( 2005 ) . let @xmath45 denote the @xmath46 indicator vector with a @xmath47 in the @xmath48-th element and zeros elsewhere , and imagine for the moment that @xmath34 is known . then the multivariate regression implied by ( [ eq : reg1 ] ) for the @xmath49-th column of @xmath0 , @xmath50 , can be seen as standard multivariate regression with response vector @xmath51 , coefficient matrix @xmath33 and predictor vector @xmath52 . the matrix model ( [ eq : reg1 ] ) can be interpreted in the same way , except the predictors for each column of @xmath0 are estimated as linear combinations of the columns of @xmath1 determined by the rows of @xmath34 . the matrix regression ( [ eq : reg1 ] ) is a new model formulation that as far as we know has not been discussed in the literature . it incorporates simple regression , multiple regression and multivariate multiple regression as special cases under different settings of @xmath1 and @xmath0 . for instance , when the response is univariate and the predictor is vector - valued , ( [ eq : reg1 ] ) reduces to a multiple regression . when both the response and the predictor are vector - valued , ( [ eq : reg1 ] ) coincides with the usual multivariate regression model . by capturing the row and column relationships , model ( [ eq : reg1 ] ) reduces the number of parameters by @xmath53 in comparison to ( [ eq : reg0 ] ) , where @xmath54 is the parameter reduction from the coefficients and the rest is the parameter reduction from the covariance matrices . the total number of the reduced parameters could be very large when the matrix dimensions of @xmath1 and @xmath0 are relatively high . model ( [ eq : reg1 ] ) can be seen as a new regression tool for matrix - variate analyses in chemometrics . a variant on partial least square ( pls ) regression is the primary method used to study matrix - variate problems in chemometrics ( smilde et al . , 2004 ) , which regresses both response and predictor on a common latent term with a bilinear coefficient structure . the relationship between the matrix - variate regression ( [ eq : reg1 ] ) and matrix pls is similar to the relationship between regression and pls in the conventional univariate and multivariate settings . * concomitant columns. when the column dimension of @xmath1 is the same as the corresponding dimension of @xmath0 , and in particular when the columns of @xmath1 and the columns of @xmath0 represent repeated measures or similar characteristics , it might be appropriate to build ( [ eq : reg1 ] ) with @xmath55 , an @xmath56 identity matrix , since each column of @xmath0 is usually associated with the corresponding column of @xmath1 . in this case , the matrix regression model can be simplified to : @xmath57 so the regression of each column of @xmath0 on the corresponding column of @xmath1 has the same coefficient matrix @xmath33 . in application , one can apply a likelihood ratio test or other model selection methods to select between ( [ eq : reg1 ] ) and ( [ eq : reg2 ] ) . similar comments apply to the row dimensions of @xmath1 and @xmath0 . univariate predictor . * when the predictor @xmath58 is univariate , a matrix regression might allow a different linear regression for each element of @xmath0 on @xmath59 : @xmath60 where @xmath61 . in this case , ( [ eq : reg1 ] ) still has a multiplicative coefficient structure but now the column dimensions of @xmath62 and @xmath41 need not be restricted to match the dimension of @xmath59 : @xmath63 where @xmath64 and @xmath65 , for some @xmath66 . it is easy to see that ( [ eq : reg3 ] ) is a reduced rank ( izenman , 1975 ; reinsel and velu , 1998 ) form of ( [ eq : reg4 ] ) . again , model selection tools can be applied to select the proper model between ( [ eq : reg4 ] ) and ( [ eq : reg3 ] ) . * multivariate pod model . * the matrix - variate regression ( [ eq : reg1 ] ) can be seen as an extension of the partial one - dimensional regression model ( pod ) studied by cook and weisberg ( 2004 ) . the pod was proposed to study data with one response variable , one grouping variable or factor , and several covariates . a typical pod model represents the response @xmath67 at factor level @xmath68 as @xmath69 , where @xmath70 and @xmath71 represent the @xmath68-th level factor effect , and @xmath72 represents the one - dimensional covariate effect ( @xmath73 ) . if each subject contains measurements over all @xmath74 categories , then a multivariate response @xmath75 will be observed for each subject . in this case , the pod model can be extended as @xmath76 where @xmath77 and @xmath78 are @xmath74 dimensional vectors . the mean function of this model has the exactly the same coefficient structure as shown in ( [ eq : reg1 ] ) except the response and the predictor are vector - valued . when @xmath79 is matrix - valued , for example , @xmath80 response variables measured at each factor level , ( [ eq : pod1 ] ) can be extended to the exact matrix - variate regression setting . in this section , we focus on the estimation of ( [ eq : reg1 ] ) . the estimation of the special formulations can be similarly derived . without a specific parametric distribution on the random error , one can estimate the coefficient parameters in ( [ eq : reg1 ] ) by using a loss function , like a squared loss function , and estimate the covariance matrices using moments . however , for statistical inference , as in conventional linear regressions , we assume that the random error @xmath35 follows a matrix normal distribution @xmath81 . background on the matrix normal distribution can be found in dawid ( 1981 ) and de waal ( 1985 ) . we next describe maximum likelihood estimation of the parameters in ( [ eq : reg1 ] ) assuming the matrix normal . assume that @xmath82 , @xmath83 , are sampled from the conditional distribution @xmath84 , and that they are independent . without loss of generality , suppose @xmath85 . then the mle of @xmath86 is @xmath87 . in matrix - variate analysis , explicit mles of @xmath62 , @xmath41 , @xmath40 and @xmath39 are generally unobtainable . we propose a two - step iterative algorithm to construct the estimators . we first hold @xmath41 and @xmath39 fixed and estimate the remaining parameters . we then use an iterative algorithm to obtain the full set of the mles . let @xmath88 and @xmath89 . according to the log - likelihood function ( in the supplement ) , given @xmath41 and @xmath39 fixed , the estimators @xmath62 and @xmath40 are @xmath90 similarly , let @xmath91 and @xmath92 . given @xmath62 and @xmath40 fixed , @xmath41 and @xmath39 can be estimated by @xmath93 let @xmath94 , @xmath95 , @xmath96 and @xmath97 be the mles of @xmath62 , @xmath41 , @xmath40 and @xmath39 respectively . therefore , by initializing @xmath98 with the required frobenius normalization and initializing @xmath99 , one can obtain @xmath94 , @xmath95 , @xmath96 and @xmath97 by iterating between ( [ eq : mrmle1 ] ) and ( [ eq : mrmle2 ] ) with updated values until the log - likelihood function of ( [ eq : reg1 ] ) meets a convergence criterion . then normalize @xmath95 as @xmath100 and rescale @xmath94 as @xmath101 , where @xmath102 with @xmath103 being the first element of @xmath95 and ` @xmath104 ' representing the frobenius norm ; and normalize @xmath97 as @xmath105 and rescale @xmath96 as @xmath106 , where @xmath107 and @xmath108 is the first diagonal element of @xmath97 . if unique identification is not required for the individual parameter matrices @xmath109s and @xmath110s , the last step of normalization and rescaling in the algorithm is not necessary . in this case , we obtain mles for @xmath42 and @xmath111 but not the individual @xmath109s and @xmath110s . in addition , the aforementioned estimation procedure provides the same results as estimating the likelihood directly under the normalization constraints on @xmath41 and @xmath39 , because the corresponding mles with and without constraints are proportional . the justification is similar to the matrix normal estimation shown in dutilleul ( 1999 ) and srivastava et al . ( 2008 ) . the existence of the estimators @xmath94 , @xmath95 , @xmath96 and @xmath97 depends on the existence of the inverse matrices @xmath112 and @xmath113 , which requires at least @xmath114 ( dutilleul , 1999 ) , and on the existence of @xmath115 and @xmath116 , which requires @xmath117 and @xmath118 respectively , when @xmath94 and @xmath95 are full rank matrices . in general , these conditions are fairly mild and require only the row dimensions of the response and the predictor less than their column dimensions multiplied by sample size and vice versa . in section [ sec:5 ] , we further investigate the asymptotic properties of the mles in ( [ eq : reg1 ] ) and show their asymptotic efficiency relative to estimators from conventional vector regression . to guage how well the matrix regression ( [ eq : reg1 ] ) fits the observed data , one can test the goodness of fit of the model compared to an alternative . for instance , to compare ( [ eq : reg1 ] ) with the vector - variate regression of @xmath119 on @xmath120 given in ( [ eq : reg0 ] ) assuming that @xmath121 , one can test the hypothesis : @xmath122 versus @xmath123 is not true . since ( [ eq : reg1 ] ) is nested within ( [ eq : reg0 ] ) , the likelihood ratio test can be applied . let @xmath124 and @xmath125 be the maximized likelihoods ( [ eq : reg1 ] ) and ( [ eq : reg0 ] ) . then the test statistic @xmath126 follows a chi - squared distribution with degree of freedom @xmath127 , which is equal to the number of reduced parameters from ( [ eq : reg0 ] ) to ( [ eq : reg1 ] ) . goodness of fit can also be evaluated by other model selection methods , such as aic , bic or cross validation . the goodness of fit of ( [ eq : reg1 ] ) compared to other models , such as to the special formulations in section [ sec:2.2 ] , can be similarly derived . since the envelope methodology introduced by cook et al . ( 2010 ) is still relatively new , we provide a brief review in this section , before turning to envelopes for matrix - variate regressions in section [ sec:3 ] . to facilitate the discussion , we introduce the following notations that will be used in the rest of the article . for an @xmath128 matrix @xmath129 , @xmath130 is the subspace of @xmath131 spanned by the columns of @xmath129 , @xmath132 is the projection onto @xmath130 , and @xmath133 is the orthogonal projection , where @xmath134 is the moore - penrose inverse . let @xmath135 . the projections @xmath136 and @xmath137 can be equivalently denoted as @xmath138 and @xmath139 . a subspace @xmath140 is said to be a reducing subspace of @xmath141 if @xmath142 decomposes @xmath143 as @xmath144 . if @xmath142 is a reducing subspace of @xmath143 , we say that @xmath142 reduces @xmath143 . envelope methodology was proposed originally to improve efficiency in the vector - variate linear model @xmath145 where the response vector @xmath146 , the predictor vector @xmath147 , @xmath148 , @xmath149 , and the random error @xmath150 is independent of @xmath1 . the motivation for enveloping in this context arose from asking if there are linear combinations of @xmath0 whose distribution is invariant to changes in the non - stochastic predictor @xmath1 . we refer to such linear combinations as @xmath1-invariants . to gain an operational version of this notion , suppose that the subspace @xmath151 satisfies the two conditions ( i ) the marginal distribution of @xmath152 does not depend on @xmath1 , and ( ii ) given the predictor @xmath1 , @xmath153 and @xmath152 are independent . then a change in @xmath1 can affect the distribution of @xmath0 only via @xmath153 ( cook et al . informally , we think of @xmath153 as the part of @xmath0 that is material to the regression , while @xmath152 is @xmath1-invariant and thus immaterial . let @xmath154 . cook et al . ( 2010 ) showed that the statistical conditions ( i ) and ( ii ) are equivalent to the algebraic conditions : ( a ) @xmath155 , and ( b ) @xmath156 . therefore , @xmath142 is a reducing subspace of @xmath157 that contains @xmath158 . the intersection of all reducing subspaces of @xmath157 that contain @xmath158 is called the @xmath157-envelope of @xmath158 , denoted as @xmath159 , or @xmath160 when used as a subscript . the envelope @xmath159 serves to distinguish @xmath161 and the maximal @xmath1-invariant @xmath162 in the estimation of @xmath163 and can result in substantial increases in efficiency , sometimes equivalent to taking thousands of additional observations . reparameterizing ( [ eq : m1 ] ) in terms of a semi - ortogonal basis matrix @xmath164 for @xmath159 , we have its envelope version @xmath165 with @xmath166 . the standard errors of elements of the maximum likelihood estimator of @xmath167 are often substantially smaller than the corresponding standard errors based on ( [ eq : m1 ] ) . the overarching premise of the envelope @xmath159 that there may be linear combinations of @xmath0 whose distribution is invariant to changes in @xmath1 can be seen as inducing a type of generalized sparsity in both @xmath163 and @xmath157 . since @xmath168 , @xmath169 and thus linear combinations of @xmath163 are 0 rather than individual elements . envelopes go a step further by simultaneously requiring a generalized form of sparsity in @xmath166 . we employ data on concho water snakes ( johnson and wichern , 2007 ) to illustrate the idea of enveloping intuitively . the data contain measurements on the gender , age , tail length ( mm ) , and snout to vent length ( mm ) for 37 female concho water snakes and 29 male concho water snakes . we want to investigate the gender effect on the tail length and snout to vent length for young snakes ( age @xmath170 years ) . the data can be modeled using ( [ eq : m1 ] ) , where @xmath171 is a bivariate response vector with @xmath172 and @xmath173 representing the tail length and snout to vent length , respectively , and @xmath58 is an indicator variable with zero indicating a male snake and with one indicating a female snake . thus , the intercept @xmath174 is the mean of the female population , and the coefficient vector @xmath175 is mean difference between the female population and the male population , which implies the gender effect . figure 1(a ) shows the data , where the circles represent the male snakes and the cross points represent the female snakes . under model ( [ eq : m1 ] ) , inference on the gender difference in tail lengths proceeds by first projecting the data onto the horizontal axis of figure 1(a ) and then using univariate inference methods with the projected data . the two dashed curves represent the marginal distributions of the male tail length and the female tail length . these curves overlap substantially , so it may take a large sample size to detect the gender effect . the estimated envelope and its orthogonal complement are represented by the solid and the dashed lines in figure 1(a ) . we can see that the two samples differ notably in the direction of the envelope , while differing little if at all in the direction of its orthogonal complement . accordingly , envelope estimation proceeds by first projecting the data onto the envelope to remove @xmath162 , the @xmath59-invariant part of @xmath0 , and then projecting onto the horizontal axis to focus on the gender difference in tail length . this process is depicted in figure 1(b ) , which shows the approximated marginal distributions of the male s and the female s tail lengths after enveloping , represented by the solid curves . the variances of two marginal distributions after enveloping are considerably reduced compared to the original ones ( the dashed curves ) , which reflects the efficiency gain . the efficiency gains can be massive when the immaterial variation @xmath176 is substantially larger than the material variation @xmath177 . although the matrix - variate regression ( [ eq : reg1 ] ) is parsimoniously parameterized relative to the naive model ( [ eq : reg0 ] ) , there still may be linear combinations of the rows or columns of @xmath178 whose distribution is invariant to changes in the predictors . to allow for the possibility of @xmath1-invariant linear combinations , we extend the vector - variate model described in section [ sec:3.1 ] to matrix - variate regression . the rationale for doing so is generally the same as that described for vector - variate regressions : we hope to achieve efficiency in estimation and prediction better than that for model ( [ eq : reg1 ] ) . we first establish the envelope structure for the matrix - variate regression model ( [ eq : reg1 ] ) . envelope models for the special cases are given in the supplement . to allow for the possibility that there are @xmath1-invariants in both the rows and columns of @xmath0 , we suppose that there exist subspaces @xmath179 and @xmath180 so that @xmath181 where @xmath182 and @xmath183 were defined in section [ sec:2.1 ] . these conditions are similar to conditions ( i ) and ( ii ) introduced in section [ sec:3.1 ] for defining @xmath1-invariants of @xmath0 in vector - variate regressions . let @xmath184 and @xmath185 . conditions ( [ eq : mel]a ) and ( [ eq : mer]a ) mean that the marginal distributions of @xmath186 and @xmath187 do not depend on @xmath1 , which is equivalent to requiring that @xmath188 and @xmath189 . conditions ( [ eq : mel]b ) and ( [ eq : mer]b ) indicate that @xmath186 does not respond to changes in @xmath1 through a linear association with @xmath190 and that @xmath191 does not respond to changes in @xmath1 through a linear association with @xmath192 . further , condition ( [ eq : mel]b ) holds if and only if @xmath193 , which is equivalent to requiring that @xmath194 reduce @xmath195 , so @xmath196 . similarly , @xmath197 must reduce @xmath198 , so @xmath199 . the intersection of all reducing subspaces @xmath194 of @xmath40 that contain @xmath200 is the @xmath40-envelope of @xmath200 , denoted as @xmath201 or @xmath202 when used as a subscript . similarly , the intersection of all reducing subspaces @xmath197 of @xmath39 that contain @xmath203 is the @xmath39-envelope of @xmath203 , denoted as @xmath204 or @xmath205 when used as a subscript . these subspaces @xmath201 and @xmath204 always exist , are uniquely defined and serve as the fundamental constructs that allow row and column reduction of @xmath0 . in concert , they imply that @xmath0 responds to changes in @xmath1 only via @xmath206 and they hold the promise of much better estimation of @xmath62 and @xmath41 . a theoretical justification regarding the variance reduction of enveloping is given by proposition [ prop4 ] in section [ sec:5 ] . because the column spaces @xmath207 and @xmath208 , and the orthogonal decompositions on @xmath40 and @xmath39 , are all invariant under a multiplicative constant , the column and row envelopes @xmath209 and @xmath210 are always uniquely defined , whether normalization is required or not . we next use @xmath201 and @xmath204 to reparameterize ( [ eq : reg1 ] ) and to establish the envelope model . let @xmath211 and @xmath212 be semi - orthogonal bases of @xmath201 and @xmath204 , respectively , where @xmath213 and @xmath214 are the dimensions of the corresponding row and column envelopes and are temporarily assumed to be known . the determination of @xmath213 and @xmath214 will be discussed in section [ sec:6 ] . by definition , we know that @xmath215 and @xmath216 , so there exist two coordinate matrices @xmath217 and @xmath218 such that @xmath219 and @xmath220 . let @xmath221 and @xmath222 be orthogonal matrices . then the matrix regression model ( [ eq : reg1 ] ) can be re - parameterized as the following envelope model : @xmath223 where @xmath224 , @xmath225 , and @xmath226 and @xmath227 are unknown , @xmath228 . as @xmath229 and @xmath230 are overparameterized , the matrices @xmath231 and @xmath232 themselves are not identifiable but their column spaces @xmath233 and @xmath234 are identifiable . the two parameter spaces are grassmannians of dimension @xmath213 and @xmath214 in @xmath131 and @xmath235 with the numbers of unknown real parameters @xmath236 and @xmath237 respectively . therefore , the total number of real parameters in ( [ eq : envreg1 ] ) is @xmath238 , which is equal to the sum of the numbers of parameters @xmath239 in @xmath86 , @xmath240 in @xmath241 , @xmath242 in @xmath243 , @xmath236 in @xmath231 , @xmath237 in @xmath232 , @xmath244 in @xmath245 , @xmath246 in @xmath247 , and @xmath248 in @xmath249 and @xmath250 , whereas ( [ eq : reg1 ] ) has @xmath251 parameters . here the normalization constraints on @xmath41 and @xmath39 eliminate two free parameters in each model . we see that the envelope model further reduce @xmath252 parameters from ( [ eq : reg1 ] ) . efficiency gain of envelope model ( [ eq : envreg1 ] ) over matrix model ( [ eq : reg1 ] ) can arise in two distinct ways . the first is through parameter reduction , particularly when the number of real parameters in ( [ eq : reg1 ] ) is large relative to that in ( [ eq : envreg1 ] ) . but the largest efficiency gains are often realized when the variation in the @xmath1-invariant part of @xmath0 is large relative to the material variation @xmath0 . letting @xmath253 denote the spectral norm of a matrix , this happens when @xmath254 or @xmath255 . for vector - variate regression , this is reflected in figure [ ch3_fig1 ] , since the variation along @xmath256 is notably larger than that along @xmath159 . in this section , we describe the mles for the unknown parameters in ( [ eq : envreg1 ] ) . as @xmath85 , the mle of @xmath86 is still @xmath87 but the remaining parameter estimates can not be expressed in closed form . consequently , we propose a two - step iteration algorithm to obtain the mles . let @xmath257 , @xmath258 , @xmath259 , @xmath260 , @xmath261 and @xmath262 be the envelope mles of the corresponding parameters . since closed - form expressions for these estimators are unavailable , we first hold @xmath34 and @xmath198 fixed and give the estimators of the remaining parameters . we later describe an iterative algorithm for computing the full set of mles . assuming then that @xmath34 and @xmath198 are given , let @xmath263 , and let @xmath264 and @xmath265 be defined in ( [ eq : mrmle1 ] ) . then @xmath231 can be estimated as @xmath266 where the objective function @xmath267 and @xmath268 is taken over all semi - orthogonal matrices @xmath269 . the estimators of @xmath33 and @xmath195 are then @xmath270 and @xmath271 , where @xmath272 is orthogognal . similarly , given @xmath62 and @xmath40 , let @xmath273 , and let @xmath274 and @xmath275 be defined in ( [ eq : mrmle2 ] ) . then @xmath276 where @xmath277 and @xmath278 is taken over all semi - orthogonal matrices @xmath279 . then the intermediate estimators of @xmath34 and @xmath198 are @xmath280 and @xmath281 , where @xmath282 is orthogognal . the full mles can now be obtained by alternating between @xmath283 and @xmath284 , and then rescaling them : 1 . initialize @xmath285 and @xmath286 as the mles from ( [ eq : reg1 ] ) . let @xmath287 and let @xmath288 . 2 . given @xmath260 and @xmath262 , estimate @xmath231 by ( [ eq : envlme1 ] ) , @xmath289 and @xmath290 . 3 . given @xmath259 and @xmath261 , estimate @xmath232 by ( [ eq : envlme2 ] ) , @xmath291 and @xmath292 . 4 . iterate 2 - 3 until the log - likelihood function of ( [ eq : envreg1 ] ) converges . then normalize @xmath260 as @xmath293 and rescale @xmath259 as @xmath294 , where @xmath295 and @xmath296 is the first element of @xmath260 , and normalize @xmath262 as @xmath297 and rescale @xmath261 as @xmath298 , where @xmath299 and @xmath300 is the first diagonal element of @xmath262 . similar to the discussion in section [ sec:2.3 ] , the proposed estimation procedure provides the same results as estimating the parameters directly under constrained @xmath41 and @xmath39 . if one is interested only in estimating the entire kronecker products @xmath42 and @xmath111 , but not individual @xmath109s and @xmath110s , the last rescaling step in the algorithm is unnecessary . moreover , the optimization of @xmath301 and @xmath302 are attained by a non - grassman algorithm ( cook et al . 2015 ) that is more efficient and faster than existing grassmannian optimization approaches . in particular , cook et al . ( 2015 ) demonstrated its superiority over the 1d algorithm used by li and zhang ( 2015 ) . the resulting envelope estimators @xmath303 and @xmath304 in steps 2 and 3 can be viewed as the projections of their corresponding estimators , @xmath264 and @xmath305 , from ( [ eq : reg1 ] ) onto the estimated row and column envelopes @xmath306 and @xmath307 . in addition , the envelope mles @xmath261 and @xmath262 are both partitioned into the estimated @xmath1-variant and @xmath1 invariant parts of @xmath0 . similar to the intuitive arguments in figure [ ch3_fig1 ] , with the envelope projections , the formulations here can lead to more efficient estimation of the parameters . so far , we have assumed that the dimensions @xmath213 and @xmath214 of the row and column envelopes are known . extending the methods described by cook et al . ( 2010 ) , these dimensions , which are essentially model selection parameters , can be determined by using an information criterion , say aic or bic , or using other model selection methods such as cross validation and likelihood ratio tests . for the information criteria , the envelope dimensions @xmath213 and @xmath214 can be selected by minimizing the objective function @xmath308 , where @xmath309 is the maximized log - likelihood function of an envelope matrix regression model , @xmath310 is equal to log@xmath311 for bic and is equal to 2 for aic , and @xmath312 is the total number of parameters in the target model . when the matrix regression model ( [ eq : reg1 ] ) holds , these procedures for selecting @xmath313 and @xmath314 are in effect methods for determining the appropriateness of a proper envelope model . if it is inferred that @xmath315 and @xmath316 , then the data do not support a proper envelope model . on the other hand , the inference that @xmath315 and @xmath316 may itself be useful depending on the application , because it indicates the absence of @xmath1-invariant linear combinations of the row and columns of @xmath0 . in this section , we investigate the asymptotic properties of the mles from ( [ eq : reg1 ] ) and ( [ eq : envreg1 ] ) . we show that the mles from ( [ eq : reg1 ] ) are asymptotically more efficient than the mles obtained from ( [ eq : reg0 ] ) and that the envelope estimators from ( [ eq : envreg1 ] ) can gain efficiency over the mles from ( [ eq : reg1 ] ) . we neglect @xmath86 in the theoretical development , since its mle is @xmath87 in all the three models , which is asymptotically independent of the mles of all other parameters in the models . to show the relative efficiency of ( [ eq : reg1 ] ) to ( [ eq : reg0 ] ) , we need to consider only the joint asymptotic distribution of the mles of @xmath317 and @xmath318 , whereas the individual parameter matrices @xmath109s and @xmath110s are not of interest . in this case , the matrix regression ( [ eq : reg1 ] ) can be considered as an over - parameterized structural model of ( [ eq : reg0 ] ) . hence we can apply proposition 4.1 in shapiro ( 1986 ) to derive the target property . for notation convenience , let @xmath319 and @xmath320 , where ` vech ' denotes the half - vectorization operator . then under ( [ eq : reg1 ] ) , we can rewrite @xmath164 as @xmath321 , where @xmath322 and @xmath323 . define @xmath324 to be the standard mle of @xmath164 from ( [ eq : reg0 ] ) and define @xmath325 to be the mle of @xmath164 from the matrix regression ( [ eq : reg1 ] ) . for a general statistic @xmath326 , denote the asymptotic variance of @xmath327 as @xmath328 . let @xmath329 be the fisher information function of @xmath164 and let @xmath330 be the gradient matrix . proposition [ prop1 ] establishes the asymptotic distribution of @xmath325 and its asymptotic efficiency relative to @xmath324 . under ( [ eq : reg1 ] ) with a normal error , @xmath331 converges in distribution to a normal random vector with mean zero and covariance matrix @xmath332 . moreover , @xmath333 . [ prop1 ] consequently , when the matrix regression structure holds , omitting row and column information in the coefficient and covariance formulation can result in loss of efficiency in parameter estimation . we next establish asymptotic properties for the mle from ( [ eq : envreg1 ] ) and show that enveloping can further improve efficiency in estimating ( [ eq : reg1 ] ) . let @xmath334 be the collection of the envelope parameters . under ( [ eq : envreg1 ] ) , @xmath335 is reformulated as @xmath336 , where @xmath337 , @xmath338 , @xmath339 , and @xmath340 . thus , @xmath341 . correspondingly , the envelope mle of @xmath164 is @xmath342 . let @xmath343 be a gradient matrix . under ( [ eq : envreg1 ] ) with a normal error , @xmath344 converges in distribution to a normal random vector with mean zero and covariance matrix @xmath345 . moreover , @xmath346 . [ prop2 ] so far we have derived the asymptotic distributions for the estimators of @xmath164 under different model settings and compared the relative efficiency among different estimators . we next investigate the asymptotic properties of the individual row and column parameter estimators of @xmath335 in ( [ eq : reg1 ] ) and ( [ eq : envreg1 ] ) , given that these parameters are uniquely defined by normalization as discussed in the model settings . let @xmath347 denote the mle of @xmath335 from ( [ eq : reg1 ] ) , and let @xmath348 denote the envelope mle of @xmath335 from ( [ eq : envreg1 ] ) . under ( [ eq : reg1 ] ) with a normal error and uniquely defined row and column parameter matrices , @xmath349 converges in distribution to a normal random vector with mean zero and covariance matrix @xmath350 , where @xmath351 and @xmath352 are matrix - valued functions of @xmath335 given in the supplement . [ prop3 ] under ( [ eq : envreg1 ] ) with a normal error and uniquely defined row and column parameter matrices , @xmath353 converges in distribution to a normal random vector with mean zero and covariance matrix @xmath354 . moreover , @xmath355 . [ prop4 ] propositions [ prop3 ] and [ prop4 ] demonstrate the asymptotic normality for both enveloped and non - enveloped estimators of @xmath335 . proposition [ prop4 ] further indicates that as long as there exists @xmath1-invariants in ( [ eq : reg1 ] ) , one can gain potential efficiency in estimation by enveloping the matrix regression model . since the row and column coefficient matrices , @xmath356 and @xmath357 , are often of interest in regression analysis , we further derived asymptotic variances for @xmath358 and @xmath359 , which are given in the supplement . when the random error in ( [ eq : reg1 ] ) is not normal , the estimator of @xmath335 under a type of generalized least square estimation ( as shown in the supplement ) is the same as the mle of ( [ eq : reg1 ] ) under normality . we thus use the same notation @xmath347 . under mild conditions , the estimator @xmath347 is still asymptotically normal , except having a more complex asymptotic covariance matrix . the terms @xmath360 and @xmath361 described in proposition [ prop3 ] will be used in the following propositions . under ( [ eq : reg1 ] ) with uniquely defined row and column parameter matrices , suppose that the random error @xmath35 follows a general matrix - valued distribution with zero mean and column and row covariance matrices @xmath40 and @xmath39 . if @xmath352 is invertible at the true value @xmath335 and @xmath362 is @xmath363 at some @xmath364 between the estimator @xmath347 and the true @xmath335 , where @xmath365/\partial { \boldsymbol\theta}_1^t\\|_{{\mathrm{f}}}$ ] , then @xmath366 converges in distribution to @xmath367 , where @xmath368 is a matrix of second - order limiting estimating functions given in the supplement . [ prop5 ] for the envelope model ( [ eq : envreg1 ] ) with a non - normal error @xmath35 , a good way to achieve parameter estimation is to use the same objective functions ( [ eq : envlme1 ] ) and ( [ eq : envlme2 ] ) to estimate the row and column envelopes , and then estimate the remaining parameters as described in section [ sec:3.3 ] . the next proposition shows that without normality on the error , the envelope estimators @xmath348 obtained from this procedure still retain desired asymptotic properties . under ( [ eq : envreg1 ] ) and the conditions in proposition [ prop5 ] , @xmath353 converges in distribution to @xmath369 , where @xmath370 moreover , the envelope estimators are asymptotically more efficient than the estimators from ( [ eq : reg1 ] ) if @xmath371 is a reducing subspace of @xmath372 . in this section , we evaluate the performance of matrix - variate regression ( [ eq : reg1 ] ) and envelope matrix - variate regression ( [ eq : envreg1 ] ) numerically , and compare the two models along with the vector regression model ( [ eq : reg0 ] ) . we first generated data based on ( [ eq : envreg1 ] ) with @xmath373 , @xmath374 , @xmath375 , @xmath376 , @xmath377 , and @xmath378 . here @xmath379 and @xmath380 . the semi - orthogonal matrices @xmath231 and @xmath232 were generated by orthogonalizing matrices of independent uniform ( 0,1 ) random variables . the elements of @xmath86 , @xmath241 , @xmath243 and the predictors were selected as independent standard random normal variables . we then fitted three models : ( [ eq : reg0 ] ) , ( [ eq : reg1 ] ) , and ( [ eq : envreg1 ] ) to the data and evaluated their estimation accuracy of the coefficient parameters , according to the criteria : @xmath381 for the first two matrix regression models , and @xmath382 for the vector regression ( [ eq : reg0 ] ) . a corresponding evaluation of the covariance estimators showed results similar to the coefficient and is thus omitted . we used six different sample sizes , @xmath383 . within each sample size , 200 replicates were simulated . the average estimation errors were computed over the 200 random samples for each sample size , under each model . the envelope dimensions were set to the true dimensions @xmath374 . with this setup models ( [ eq : reg0 ] ) , ( [ eq : reg1 ] ) and ( [ eq : envreg1 ] ) have 975 , 105 and 75 real parameters to be estimated . figure [ fig:2 ] shows the comparison results from the three models . in particular , the left panel of figure [ fig:2 ] shows a comparison of the three models , and the right panel highlights the improvement of enveloping . over all selected sample sizes , the two matrix regression models provided much smaller estimation errors than did the vector regression . in addition , by effectively removing @xmath1-invariants from estimation , the envelope model further improved estimation accuracy in comparison to the matrix regression model without enveloping . figure [ fig:3 ] shows the efficiency gains by assessing the asymptotic , actual , and bootstrap standard errors of the first elements in @xmath384 ( or @xmath385 ) from the three models . similar results were obtained for other elements . the asymptotic standard errors were estimated according to the results presented in propositions [ prop1 ] and [ prop2 ] . the actual standard errors were computed using the sample standard errors of @xmath384 ( or @xmath385 ) over 200 samples for each selected sample size . the bootstrap standard errors were obtained by bootstrapping one sample for each sample size . again , the left panel in figure [ fig:3 ] provides a comparison among the three models and the right panel highlights the comparison between the two matrix regression models with and without enveloping . clearly , the asymptotic , actual and bootstrap standard errors are very close , indicating that the asymptotic covariance matrices proposed in section [ sec:5 ] are accurate . in addition , over all selected sample sizes , the matrix regression without enveloping was asymptotically more efficient in parameter estimation than the vector regression , and imposing additional envelope structure further improved the efficiency gains . the importance of enveloping can be seen clearly in the right panel of figure [ fig:3 ] . indeed , the ratios of the asymptotic standard errors between the estimators from ( [ eq : reg0 ] ) and ( [ eq : reg1 ] ) range from 23.7 to 54.9 , while the ratios of the asymptotic standard errors between the estimators from ( [ eq : reg0 ] ) and ( [ eq : envreg1 ] ) range from 309.0 and 523.9 . therefore , effectively utilizing matrix - variate information in model fitting , and at the same time , removing @xmath1-invariants from estimation , one can gain substantial efficiency in parameter estimation . we next assess the performance of the two matrix regression models ( [ eq : reg1 ] ) and ( [ eq : envreg1 ] ) in terms of the individual row and column parameter estimations , particularly the estimation of @xmath62 and @xmath41 . the data were generated as in the previous example but to make the row and column parameter matrices identifiable , we normalized @xmath41 and @xmath39 such that each of them has frobenius norm 1 and positive element in its first row and column . we then evaluated the estimation accuracy of @xmath62 and @xmath41 according to the criterion @xmath386 , @xmath387 for both models . the results are shown in figure [ fig:4 ] wherein the envelope model notably outperformed the matrix regression ( [ eq : reg1 ] ) for both @xmath62 and @xmath41 at all sample sizes . figure [ fig:5 ] demonstrates the substantial efficiency gains of the envelope model by comparing the standard errors of the first elements in @xmath259 and @xmath260 obtained from ( [ eq : envreg1 ] ) and ( [ eq : reg1 ] ) . the comparison results similarly hold for other elements . again , asymptotic , actual and bootstrap standard errors are reported for each model . the asymptotic standard errors were estimated based on the results from propositions [ prop3 ] and [ prop4 ] . the actual and bootstrap standard errors were computed as in the previous example . from figure [ fig:5 ] , we see that the asymptotic standard errors of the estimators of @xmath62 and @xmath41 are very close to their corresponding actual and bootstrap standard errors for both models . this again supports the accuracy of the asymptotic covariance matrices given in section [ sec:5 ] . in addition , the envelope model shows asymptotic efficiency in comparison to ( [ eq : reg1 ] ) over all selected sample sizes . the ratios of the standard errors between the estimators from ( [ eq : reg1 ] ) and ( [ eq : envreg1 ] ) over all elements in @xmath62 range from 2.7 to 11.8 . the ratios of the standard errors between the corresponding estimators of @xmath41 range from 2.5 to 27.2 . enveloping improved both accuracy and efficiency in parameter estimation . we applied the proposed matrix - variate regression models to a cross - over assay of insulin based on rabbit blood sugar concentration ( v@xmath3lund , 1980 ) . the design partitioned the animals into four groups of nine rabbits each , with a different treatment assigned to each group . hence @xmath388 . let @xmath389 and @xmath390 denote the low and high dose levels , 0.75 units and 1.5 units of the standard treatment , and let @xmath391 and @xmath392 denote the same two dose levels of the test treatment . the treatment assignment is shown in table [ ch4_tb1 ] . after injection of the insulin dose each day , the blood sugar concentration of each rabbit was measured at 0 , 1 , 2 , 3 , 4 , and 5 hours . we consider the percentage decreases of the blood sugar concentrations at 1 , 2 , 3 , 4 , and 5 hours relative to the initial concentration at 0 hours . the measurements on each rabbit then form a matrix @xmath395 , where the two rows represent the percentage decreases under two different treatments assigned on day 1 and day 2 , and the columns indicate the hourly percentage decreases each day . we modeled the insulin assay data with treatment effects and dose levels as covariates to explore how these factors affect the percentage decreases of the blood sugar concentrations . we defined the covariates of each rabbit as a matrix @xmath396 , where the two rows of @xmath1 represent standard and the test treatments , and the two columns of @xmath1 represent day 1 and day 2 . for example , for the rabbits in group 1 , the covariate matrix @xmath397 , a diagonal matrix with elements 0.75 and 1.5 . that is because the rabbits in group 1 received the standard treatment at the low dose level 0.75 in day 1 , and received the test treatment at the high dose level 1.5 in day 2 . similarly , for the rabbits in groups 2 , 3 , and 4 , the covariate matrices can be formed as @xmath398 , @xmath399 and @xmath400 . we then fitted the relationship between the response @xmath401 and the predictor @xmath396 based on ( [ eq : reg1 ] ) , ( [ eq : reg2 ] ) and ( [ eq : reg0 ] ) as : @xmath402 @xmath403 @xmath404 where @xmath405 is the intercept matrix , @xmath406 , @xmath407 , @xmath408 , and @xmath409 are the corresponding coefficient matrices in the three models , and @xmath35 is the random error that is assumed to be matrix normal in the matrix - variate regression models ( [ eq : rabbitreg1 ] ) and ( [ eq : rabbitreg3 ] ) , and @xmath410 is assumed to be multivariate normal under the multivariate regression ( [ eq : rabbitreg2 ] ) . since the blood sugar concentration levels observed in the first day can not be affected by the treatment received in the second day , we set the elements in the upper - right corner of @xmath41 in ( [ eq : rabbitreg1 ] ) and of @xmath411 in ( [ eq : rabbitreg2 ] ) to be zero , that is , @xmath412 where @xmath413 are scalars , and @xmath414 . in this setting , the percentage decreases of the blood sugar concentrations in day 2 can depend on the treatments received in both day 1 and day 2 , whereas the percentage decreases in day 1 can only depend on the treatment received in day 1 . since ( [ eq : rabbitreg2 ] ) , ( [ eq : rabbitreg1 ] ) and ( [ eq : rabbitreg3 ] ) are nested models , we applied likelihood ratio tests to select the best model among the three and concluded that ( [ eq : rabbitreg3 ] ) was sufficient . we then performed envelope fitting for ( [ eq : rabbitreg3 ] ) to achieve efficiency gains for the parameter estimation . using bic , the envelope dimension was selected to be @xmath415 . the estimates ( est ) and the asymptotic standard errors ( se ) for each element in @xmath416 are summarized in table [ tab3 ] , with subscripts ` e ' and ` s ' indicating the results from the envelope fit and the fit from ( [ eq : rabbitreg3 ] ) without enveloping . c\|ccc\|ccc\|c + & & & ratio + & @xmath417 & @xmath418 & @xmath419 & @xmath420 & @xmath421 & @xmath422 & @xmath423 + & 7.74&1.78&4.35&6.34&2.21&2.87&1.24 + & 13.07&2.48&5.28&14.42&2.76&5.22&1.12 + & 18.98&3.22&5.9&21.12&3.36&6.28&1.04 + & 17.41&3.11&5.6&16.67&3.35&4.98&1.08 + & 9.63&2.17&4.43&6.03&2.65&2.28&1.22 + & 7.33&1.73&4.23&6.33&2.21&2.86&1.28 + & 12.37&2.44&5.06&12.45&2.76&4.51&1.13 + & 17.96&3.21&5.6&20.1&3.36&5.98&1.05 + & 16.47&3.09&5.34&16.42&3.35&4.9&1.09 + & 9.11&2.12&4.3&5.68&2.65&2.15&1.25 + in comparison to the fit from ( [ eq : rabbitreg3 ] ) , the envelope model shows smaller standard errors for all the elements in @xmath416 . all the parameters in @xmath424 are significant based on the two models . yet the magnitudes of the ratios @xmath425 from the envelope fitting are mostly higher than those from the fitting without enveloping . we further evaluated the sum of squared prediction error ( sspe ) by partition the data into training sets of seven subjects within each group and testing sets of two subjects within each group . the results based on the prediction error @xmath426 are summarized in figure [ ch4_fig7 ] , which shows that the envelope regression dominates for all choices of the envelope dimension @xmath213 . the smallest prediction error occurs at the optimal dimension @xmath415 . when @xmath427 , the envelope model coincides with the non - enveloped model and so the prediction errors of the two models are identical . in addition , we also modeled the data with group effects and compared estimates and standard errors with and without enveloping . the model and the results are given in the supplement . in this case , enveloping provides substantial efficient gains in estimation and can successfully identify the effects of different treatments , while the model without enveloping fails to detect such effects . the eeg data was briefly introduced in the introduction section . it contains two groups of subjects : 77 subjects from the alcoholic group and 45 subjects from the control group . each subject has measurements of electrical scalp activity , which form a @xmath5 matrix . to explore the influence of the alcoholism on the brain activity , let @xmath428 if a subject is alcoholic and @xmath429 if a subject is nonalcoholic . we applied ( [ eq : reg4 ] ) to model the data as @xmath430 where @xmath431 is the matrix - valued brain measurements , @xmath432 is the mean brain activity of the nonalcoholic population ( control group ) , @xmath433 is the difference between the mean brain activity of the alcoholic subjects , denoted as @xmath434 , and the mean brain activity of the nonalcoholic subjects , @xmath435 . therefore , the coefficient matrix , @xmath436 , compares the alcoholic subjects and the nonalcoholic subjects based on their brain signals . the random error @xmath35 is assumed to follow @xmath437 . a saturated model such as ( [ eq : reg0 ] ) is not estimable for the eeg data as the dimension of @xmath119 is much higher than the sample size , and the the covariance matrix of @xmath119 given @xmath59 is not likely to be sparse . matrix - variate regression provides one way to handle such data . in ( [ eq : eeg ] ) , it is easy to see that the mle of @xmath163 without enveloping is @xmath438 , the difference between the sample means of alcoholic brain activities and nonalcoholic brain activities . to achieve estimation efficiency , we considered fitting an envelope model on ( [ eq : eeg ] ) as @xmath439 where @xmath231 and @xmath232 are the bases of the @xmath40-envelope of @xmath440 and the @xmath39-envelope of @xmath441 respectively , @xmath442 is the the coordinate matrix , and the rest of the parameters are similarly defined as in ( [ eq : envreg1 ] ) . the bic chose the envelope dimensions to be @xmath443 and @xmath444 . similar to the results observed in the bioassay example , the envelope model can be more effective at identifying the differences between the alcoholic and nonalcoholic brain activities because it shows smaller standard errors for the estimators of @xmath163 . with @xmath443 and @xmath444 , we calculated the relative ratios of the standard errors of the estimators from ( [ eq : eeg ] ) to those from ( [ eq : eegenv ] ) , and 90% of them fall in the range of 1.2 and 5.5 . moreover , since brain regions associated with alcoholism could of interest to researchers , we further investigated the effect of alcoholism on different brain locations ( channels ) by averaging the two estimators of @xmath163 over time ( 256 rows ) . we then obtained location - based estimators from ( [ eq : eeg ] ) and ( [ eq : eegenv ] ) , each of dimension 64 . bootstrap standard errors were computed for each element of the estimators and corresponding t - tests were applied . because this is a multiple hypothesis testing problem , we employed the approach proposed by benjamini and yekutieli ( 2001 ) to control the false discovery rate ( fdr ) . the fdr adjusted p - values ( in @xmath445 scale ) for the estimators with and without enveloping are reported in the left and right panels of figure [ fig:7 ] . the dashed lines in both panels represent the significance levels @xmath446 and @xmath447 . some plot characteristics are due to the adjustment process . for instance , about 25 p - values were adjusted to 1 , which corresponds to @xmath448 in figure [ fig:7 ] . at the level @xmath446 , the envelope model suggests that 35 out of 64 brain regions are associated with alcoholism , while the model without enveloping was unable to identify relevant regions with only a few borderline significance . li and zhang ( 2015 ) also analyzed these data , but computationally their method of estimation had to first reduce the number of time points from 256 to 64 by averaging adjacent groups of four . no such reduction was necessary with our estimation method . this is an important distinction . by averaging adjacent groups of four time points , a tuning parameter was essentially introduced into their estimation method . different result would surely be obtained by varying the averaging group size . for instance , using bic their method estimated envelope dimensions of @xmath449 and @xmath450 , while using the full data we estimated @xmath443 and @xmath444 . by averaging adjacent groups , they could lose dimensions . zhao , q. , caiafa , c. f. , mandic , d. p. , chao , z. c. , nagasaka , y. , fujii , n. , zhang , l. , and cichocki , a. ( 2013 ) . higher order partial least squares ( hopls ) : a generalized multilinear regression method . _ ieee trans . pattern anal . machine intell . _ * 35 * , 16601673 .	modern technology often generates data with complex data structures in which both response and explanatory variables are matrix - valued . existing methods in the literature are able to tackle matrix - valued predictors but are rather limited for matrix - valued responses . in this article , we study matrix - variate regressions for such data , where the response @xmath0 on each experimental unit is a random matrix and the predictor @xmath1 can be either a scalar , a vector , or a matrix , treated as non - stochastic in terms of the conditional distribution @xmath2 . we propose models for matrix - variate regressions and then develop envelope extensions of these models . under the envelope framework , redundant variation can be eliminated in estimation and the number of parameters can be notably reduced when the matrix - variate dimension is large , possibly resulting in significant gains in efficiency . # 1 * key words : * matrix - valued response ; matrix - variate regression ; reducing subspace ; sufficient dimension reduction .
we know of @xmath220 x - ray binaries with dynamically confirmed black hole ( bh ) accretors ; these include 15 low mass x - ray binaries ( lmxbs ) and 3 high mass x - ray binaries ( hmxbs ) in the milky way and magellanic clouds ( see e.g. * ? ? ? * and references within ) , as well as bh + wolf - rayet binaries in ic10 @xcite and ngc300 @xcite . these systems were identified using x - ray and optical observations ; the mass function is calculated from periodic radial velocity shifts in emission lines from the optical counterpart . all of the bh lmxbs identified this way are necessarily transient , because the optical spectra of bright lmxbs are dominated by the accretion disc ( see e.g. * ? ? ? we have established a method for identifying bhcs from their x - ray properties alone . this makes use of the `` low / hard '' emission state seen in bh and neutron star ( ns ) xbs @xcite , that is only seen at 0.011000 kev luminosities @xmath010% eddington in ns xbs @xcite ; @xcite recently found that the low / hard state is limited to luminosities @xmath010% eddington in bh xbs also . bh xbs may exhibit low / hard emission states at considerably higher luminosities than ns xbs , due to the higher accretor mass . however , it is necessary to differentiate between our bhcs and distant active galactic nuclei ( agn ) , since agn and xb emission spectra are often similar . we have identified 10 bhcs from their high luminosity low states to date . of these , 9 are associated with m31 gcs @xcite , and are therefore likely lmxbs . only one of these is transient ; however , persistently bright gc bh xbs are consistent with tidal capture of a main sequence donor ( * ? ? ? * although the donor may be disrupted in the process ) , or with an ultra - compact system with a degenerate donor @xcite . we identified our first bhc outside of a gc from its x - ray spectra , long term ( @xmath212 year ) behaviour , and a serendipitous hst observation @xcite ; the faint optical counterpart ( @xmath5 @xmath6 @xmath70.4 ) suggests a low mass donor for this system also . we cover 8 of the 9 gc bhcs in our survey : bhcs 1 , 2 , 20 , 25 , 28 , 31 , 32 , and 34 . the field bhc described in @xcite is bhc3 . we obtained accurate positions for our x - ray sources by registering 27 x - ray sources associated with m31 gcs to the m31 field 5 b band image provided by @xcite . the r.m.s . offsets were 0.11@xmath8 in r.a . and 0.09@xmath8 in dec @xcite ; this would be extremely unlikely unless the 27 x - ray sources used for calibration were indeed associated with the gcs . the x - ray spectra of bh binaries are usually described with two components : a thermal component ( often modelled as a multi - temperature disk blackbody ) , and a power law component to represent unsaturated , inverse - compton scattering of cool photons on hot electrons @xcite . the hard state is classified by a power law component with photon index ( @xmath9 ) = 1.42.1 , and a thermal component that contributes @xmath10 of the 220 kev flux @xcite . our very first gc bhc ( xb045 , xbo 45 in * ? ? ? * ) was associated with the m31 gc b045 , named following the revised bologna catalogue v.3.4 ( rbc , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? its @xmath217,000 count xmm - newton / pn spectrum was well described by an absorbed power law with line - of - sight absorption ( @xmath11 ) = [email protected]@xmath13 atom @xmath14 , and @xmath9 = [email protected] ; @xmath15/dof = 517/487 ( good fit probability , gfp , = 0.17 ) . adding a blackbody component improved the fit somewhat , but the thermal contribution to the flux was too small to be constrained ; we therefore considered xbo 45 to be in its hard state . since its unabsorbed 0.310 kev luminosity was [email protected]@xmath16 erg s@xmath17 , 140% eddington for a 1.4 @xmath18 ns , we considered it a bhc . the 8 remaining gcs bhcs are included in this survey , unlike xbo 45 , so we do nt describe each spectrum in detail here . however , we note that fitting a two component model to xb144 ( xbo 144 in * ? ? ? * ) yielded k@xmath19 = [email protected] kev , indicating a complete lack of thermal component in the spectrum . we also note that even though xb082 was best described by @xmath9 = [email protected] , the @xmath15/dof for that fit was @xmath20.9 , and we were able to obtain fits where @xmath9=1.4 and @xmath15/dof @xmath201 @xcite . the unabsorbed 0.310 kev luminosities of our gc bhcs exhibiting hard state spectra range over @xmath2545@xmath21 erg s@xmath17 @xcite . comparison with the 0.510 kev agn flux distribution obtained by @xcite yields a 2.4@xmath22 probability that our gc bhcs are coincident agn . the probability that our brightest gc bhc , xb135 , is a coincident agn is 1.2@xmath23 . we will present evidence in a separate paper that xb135 may contain the most massive stellar mass bh known to date ; it may have been formed by direct collapse of a high mass , metal poor star ( r. barnard et al . , 2013 , in prep ) . we found that the gcs hosting these very bright x - ray sources were significantly more massive and/or metal rich than the other gcs in m31 , agreeing with previous work . however , two gcs had particularly low metalicities ; one of these is b135 , consistent with the direct collapse formation scenario for xb135 ( see * ? ? ? * and references within ) . @xcite have shown that while bh masses are limited to @xmath215 @xmath18 for solar metalicties , they could theoretically reach @xmath280 @xmath18 for metalicities @xmath20.01 solar . high accretion rate ns xbs exhibit multi - component emission that may appear to be hard state spectra in extragalactic x - ray sources , where the spectra have relatively few photons . to compare the emission spectra of our bhcs with ns xrb , we use the double thermal ( disk blackbody + blackbody ) model used by lin et al . ( 2007 , 2009 , 2012 ) to describe hundreds of rxte observations of ns xbs in all their varied emission states . we note that the physical interpretation of their model is contradicted for persistent xbs by a substantial body of work ; however , lin et al . ( 2007 , 2009 , 2012 ) have sampled spectra from the full range of ns lmxb emission states in a consistent manner , allowing us to compare our bhcs with ns xbs in a single parameter space . for a long time , ns xbs were divided into two types , z - sources ( high lumonisity ) and atoll sources ( low luminosity ) , based on their luminosities and color - color diagrams ( cds ) ; the cds of z - sources had three branches ( horizontal , normal , and flaring ) , and evolved along these branches without ever jumping from one branch to the other ; the cds of the atoll sources were more fragmented @xcite . furthermore , the z - sources were split into those like cygnus x-2 , and those like scorpius x-1 @xcite . it was believed that the differences were due to more than just the accretion rate , because the z - sources varied by a factor of a few when tracing their z - shaped cds , while atoll source intensities varied by 12 orders of magnitude @xcite . however , we now know of two transient ns systems that exhibited both types of z - source behavior before evolving to atoll source behavior during decay @xcite . @xcite examined the spectral evolution of two galactic x - ray transients , aql x-1 and 4u1608@xmath752 , over many rxte observations covering @xmath620 outbursts . they devised a new double - thermal model ( disk blackbody + blackbody ) to describe a ns transient soft state that is analagous to the bh soft state described by @xcite . they have since applied their model to rxte observations of xtej1701@xmath7462 , one of the transients that exhibits cyg - like and sco - like z - source behavior as well as atoll behavior @xcite , and also to the sco - like z - source gc 17 + 2 @xcite . they have applied their model to hundreds of rxte spectra from galactic ns binaries including the full gamut of ns spectral behavior . they found that their disk blackbody + blackbody was unsuccessful in two situations . firstly , they found the hard state spectra to be power law dominated , as expected @xcite . we therefore expect our bhcs to inhabit a separate parameter space to the ns xbs because fitting the double thermal model to hard state spectra will yield unphysical results . secondly , they found that z - sources required a three - component spectrum ( disk blackbody + blackbody + power law ) on the horizontal branch @xcite . @xcite also performed broadband analysis of suzaku and bepposax observations of the persistently bright galactic ns xb 4u1705@xmath744 , with energy ranges 0.1600 kev and 0.1300 kev respectively . when using a disk blackbody + blackbody model to the soft state spectra , they found temperatures that were similar to those they observed in the rxte observations @xcite . fitting an additional comptonization component , via either a power law or simpl convolution model ( following * ? ? ? * ) resulted only in small changes in the thermal components ; this is because the comptonized component only contributed @xmath010% of the total flux in the soft state ( see fig . 6 in * ? ? ? hence the parametric differences between our bhcs and the ns xbs should not be due to differences in energy bands used in the observations , or the lack of a third ( power law ) component . evidence against the double thermal model indicates an extended corona that contributes a substantial portion of the x - ray flux . this evidence includes the ingress times of periodic absorption dips in the x - ray lightcurves of the high inclination xbs ( the dipping sources , see e.g. * ? ? ? * ; * ? ? ? * and references within ) , and also broadened emission lines in a chandra grating spectroscopy of cyg x-2 @xcite . furthermore , compact corona models where the inner disk temperature is tied to the seed photon energy for comptonization are rejected for ulxs in ngc253 , and the confirmed bh+wolf - rayet binary ic10 x-1 @xcite , as well as bhc3 in the steep power law state @xcite . in @xcite , we applied structure function analysis to xbs for the first time , following @xcite , who created an ensemble structure function for agn . they used a structure function ( sf ) to estimate the mean intensity deviation for data separated by time @xmath24 : @xmath25 where @xmath26 is the photon noise and @xmath27 is the x - ray flux . they grouped the sf into logarithmic bins with width 0.5 ; each bin in the range log(@xmath24 ) = 0.03.0 contained more than 100 measurements . our sample consisted of 37 x - ray sources associated with objects in the rbc ; these were classified by @xcite as 30 confirmed gcs , 4 gc candidates , 1 star , and 2 agn . the sfs of gc xbs with 0.310 kev luminosities @xmath2250@xmath28 erg s@xmath17 tended to show significantly more variability than agn over a wide range of time - scales . the sfs of brighter xbs generally showed comparable or less variability than agn , despite their high signal to noise ; however , their high fluxes made them unlikely agn . sfs provide an effective mechanism for distinguishing between xbs and agn @xcite . in this work , we combine these techniques to search for bhcs in the whole region covered by 152 chandra observations from our monitoring programme ; the roll angle is unrestricted , resulting an approximately circular field of view with radius @xmath29 . we surveyed 530 x - ray sources in this region . we also obtain spectra from archival xmm - newton observations , to strengthen the cases for several bhcs . the 100@xmath3 region surrounding m31 * is particularly interesting , because @xcite found an excess of x - ray binaries over the radial distribution expected from k band light ( tracing stellar mass ) ; this excess had the distribution expected of dynamically formed xbs @xcite . dynamical xb formation requires stellar densities rarely seen outside gcs , and @xcite proposed that the m31 bulge is sufficiently large and dense to form a significant number of xbs dynamically . however , since the stellar velocities in the m31 bulge are @xmath2510 times higher than in gcs , @xcite expect only short period binaries to survive , with most of these containing bh accretors . evidence for bhcs in this region would therefore provide strong support for their theory . in the next section , we discuss the observations and data analysis ; this is followed by our results , then by our discussion . the central region of m31 has been observed with chandra on a @xmath2 monthly basis for the last @xmath213 years in order to monitor transients ; we exclude periods when m31 can not be observed due to satellite constraints ( @xmath2 march april ) . we have analyzed 98 acis observations and 54 hrc observations . we determined the position of each source from a merged 0.37.0 kev acis image , using the iraf tool imcentroid . the x - ray positions were registered to the field 5 b band image of m31 from the local group galaxy survey ( lgs , see * ? ? ? * ) , using 27 gc x - ray sources and following the procedure described in @xcite . the r.m.s uncertainties in registration were 0.11@xmath8 in r.a . and 0.09@xmath8 in dec . we obtained 0.37.0 kev lightcurves and spectra from circular source and background regions for each source . the background region was the same size as the source region , and at a similar off - axis angle . the extraction radius varied between sources , because larger off - axis angles resulted in larger point spread functions . we used the ciao v4.5 software suite , with corresponding caldb to reduce the data , and xspec v12.7 to analyze the spectra . for acis observations , we obtained corresponding response matrices and ancillary response files . we initially estimated the conversion from flux to luminosity by assuming a power law emission spectrum with photon index 1.7 , with @xmath11 = 7@xmath30 atom cm@xmath17 , then determining the unabsorbed 0.310 kev luminosity equivalent to 1 count s@xmath17 at the location of the source . after correcting for the exposure , vignetting and background , multiplying the source intensity by this conversion factor gave the source luminosity . source spectra with @xmath6200 net counts were freely fitted . when good spectral fits were found for a particular observation of a source , the parameters of these fits replaced the default parameters when estimating the source luminosity for observations of the source with @xmath20200 photons . cccccccccccc bhc & position & @xmath32 & @xmath33 & @xmath34 & @xmath15/dof & @xmath35 + 1 & 00:42:15.786 + 41:01:14.24 & 18 & 48 & [email protected] & 191/65 & 2145 + 2 & 00:42:18.648 + 41:14:01.85 & 83 & 54 & [email protected] & 1479/136 & 9316 + 3 & 00:42:22.954 + 41:15:35.23 & 91 & 54 & [email protected] & 23850/144 & 10440 + 4 & 00:42:26.070 + 41:19:15.03 & 93 & 54 & [email protected] & 3576/146 & 10731 + 5 & 00:42:28.193 + 41:10:00.30 & 85 & 54 & [email protected] & 486/138 & 9591 + 6 & 00:42:28.285 + 41:12:22.95 & 89 & 54 & [email protected] & 6113/142 & 10153 + 7 & 00:42:31.147 + 41:16:21.67 & 93 & 54 & [email protected] & 1123/146 & 10585 + 8 & 00:42:34.669 + 40:57:14.20 & 13 & 12 & [email protected] & 696/24 & 300 + 9 & 00:42:39.585 + 41:16:14.30 & 70 & 21 & [email protected] & 3480/90 & 4095 + 10 & 00:42:40.654 + 41:13:27.32 & 93 & 53 & [email protected] & 4724/145 & 10585 + 11 & 00:42:42.177 + 41:16:08.23 & 74 & 34 & [email protected] & 2331/107 & 5778 + 12 & 00:42:44.831 + 41:11:37.89 & 92 & 54 & [email protected] & 3220/145 & 10585 + 13 & 00:42:45.122 + 41:16:21.68 & 96 & 54 & [email protected] & 3291/149 & 11175 + 14 & 00:42:45.946 + 41:10:36.53 & 54 & 27 & [email protected] & 1759/80 & 3240 + 15 & 00:42:46.969 + 41:16:15.58 & 95 & 54 & [email protected] & 3202/148 & 11026 + 16 & 00:42:47.176 + 41:16:28.41 & 92 & 54 & [email protected] & 40403/145 & 10585 + 17 & 00:42:47.893 + 41:15:32.87 & 95 & 54 & [email protected] & 3278/148 & 11026 + 18 & 00:42:48.529 + 41:15:21.12 & 95 & 54 & [email protected] & 2369/148 & 11026 + 19 & 00:42:48.546 + 41:25:22.12 & 60 & 51 & [email protected] & 8701/110 & 6105 + 20 & 00:42:52.030 + 41:31:07.87 & 11 & 54 & [email protected] & 343/64 & 2080 + 21 & 00:42:52.534 + 41:18:54.46 & 95 & 54 & [email protected] & 1271/148 & 11026 + 22 & 00:42:54.935 + 41:16:03.19 & 95 & 54 & [email protected] & 18769/148 & 11026 + 23 & 00:42:57.900 + 41:11:04.65 & 93 & 54 & [email protected] & 826/146 & 10731 + 24 & 00:42:59.675 + 41:19:19.35 & 93 & 54 & [email protected] & 887/146 & 10731 + 25 & 00:42:59.872 + 41:16:05.72 & 95 & 54 & [email protected] & 1074/148 & 11026 + 26 & 00:43:02.937 + 41:15:22.53 & 95 & 54 & [email protected] & 4838/148 & 11026 + 27 & 00:43:03.220 + 41:15:27.69 & 95 & 54 & [email protected] & 11960/148 & 11026 + 28 & 00:43:03.876 + 41:18:04.91 & 94 & 54 & [email protected] & 2920/147 & 10878 + 29 & 00:43:05.667 + 41:17:02.43 & 77 & 29 & [email protected] & 10972/105 & 4851 + 30 & 00:43:09.866 + 41:19:00.76 & 85 & 48 & [email protected] & 2675/132 & 8646 + 31 & 00:43:10.622 + 41:14:51.30 & 90 & 54 & [email protected] & 915/143 & 10296 + 32 & 00:43:17.593 + 41:27:44.87 & 17 & 0 & [email protected] & 409/16 & 136 + 33 & 00:43:34.334 + 41:13:23.08 & 45 & 54 & [email protected] & 1380/98 & 4851 + 34 & 00:43:37.320 + 41:14:43.09 & 34 & 54 & [email protected] & 209/87 & 3828 + 35 & 00:43:44.593 + 41:24:10.25 & 32 & 39 & [email protected] & 2296/71 & 2346 + for hrc observations , we we included only pi channels 48293 , thereby reducing the instrumental background . we used the webpimms tool to find the unabsorbed luminosity equivalent to 1 count s@xmath17 , assuming the same emission model as for the acis observations with @xmath20200 photons . we created a 1 kev exposure map for each observation , and compared the exposure within the source region with that of an identical on - axis region , in order to estimate the estimate the necessary exposure correction . we multiplied the background subtracted source intensity by the correction factor to get the 0.310 kev luminosity . we created long term 0.310 kev lightcurves for each source , using the luminosities obtained from each observation as described above ; all luminosities assume a distance of 780 kpc @xcite . we only included observations with net source counts @xmath36 0 after background subtraction . we fitted each long term lightcurve with a line of constant intensity , in order to ascertain the source variability . we note that the ratio of hrc to acis luminosity depends strongly on the spectral model , and should be 1.0 if the model is correct . differences between actual and assumed emission spectra during hrc and faint acis observations may lead to systematic offsets between luminosities from adjacent hrc and acis observations . we derived sfs from the 0.54.5 kev fluxes of each observation of every target by assuming a power law spectrum with the same photon index as for the hrc and faint acis observations ; an m31 x - ray source with a 0.310 kev unabsorbed luminosity of 1.0@xmath21 erg s@xmath17 , @xmath11 = 7@xmath30 atom @xmath14 and @xmath9 = 1.7 has a 0.54.5 kev flux of 0.80@xmath37 erg @xmath14 s@xmath17 . @xcite calculated the noise component from @xmath38 assuming that @xmath39 = @xmath40 , and @xmath41 is the number of photons . our lightcurves are background subtracted , and arf - corrected ; furthermore , uncertainties in the luminosities of bright sources include uncertainties in the spectral parameters . as a result , our uncertainties are not simply due to photon counting noise . hence in our case , @xmath42 ^ 2 + \left [ \frac{\sigma f_{\rm x}\left(t\right)}{f_{\rm x}\left(t\right ) } \right]^2\right>,\ ] ] where @xmath43 is the uncertainty in x - ray flux and often significantly larger than @xmath44 . in the simplest case where only 1 pair has a particular separation , then @xmath45,\ ] ] where @xmath46 and @xmath47 are the fluxes of the two observations . in addition to our chandra observations , we also obtained spectra for some of our bhcs from archival xmm - newton observations . their higher sensitivity and longer exposure times often yielded superior spectra for bright sources . we used only the pn instrument , because the mos detectors are more prone to pile up . we used sas v10 for the data reduction . we filtered out background flares by creating a lightcurve with the selection `` ( pattern==0)&&(flag==0)&&(pi in [ 10000:12000 ] ) '' and 100 s binning , then excluding intervals with @xmath60.4 count s@xmath17 . source and background spectra were obtained from circular regions with 2040@xmath8 radius , along with corresponding response files ; the filter `` ( pattern@xmath20=4)&&(flag==0 ) '' was applied along with the good time filter and extraction region . our bhc identification process involved three steps . first , we had to establish that the x - ray source was an xb rather than an agn . then we checked the acis observations of each xb , to see if it exhibited a high luminosity hard state ( @xmath9 @xmath48 2.1 at a 0.310 kev unabsorbed luminosity @xmath63@xmath21 erg s@xmath17 , i.e. 10% eddington for a 2 @xmath49 ns ) . finally we found the highest quality observation that exhibited a high luminosity hard state , fitted it with the double - thermal model , and compared the results with the known range of parameters exhibited by the ns systems studied by lin et al . ( 2007 , 2009 , 2012 ) . likely xbs were identified from their associations with gcs , variable sfs or relatively high fluxes . there are @xmath2170 gcs within our field @xcite ; using the 0.510 kev agn flux distribution @xcite , we expect to find 0.0005 agn with flux @xmath61.4@xmath37 erg @xmath14 s@xmath17 ( 10@xmath50 erg s@xmath17 ) within 1@xmath8 of any gc . hence any bright x - ray source associated with a gc is a likely xb . we also identify xbs using sfs that are strikingly more variable than the ensemble agn sf . furthermore , only @xmath21 agn per square degree is expected with a flux equivalent to @xmath51 erg s@xmath17 ; hence brighter x - ray sources are likely xbs too . lin et al . ( 2007 , 2009 ) found that they were unable to successfully fit hard state spectra with their double thermal model . this is because the disk blackbody component must account for the low - energy flux , resulting in an unphysically low temperature . to demonstrate this , we created a fake hard state spectrum using the xspec command fakeit ; we used a power law with @xmath9 = 1.7 , with absorption equivalent to 7@xmath30 atom @xmath14 . we present the best fit double thermal model for this spectrum in fig . [ example ] , which consists of a 0.67 kev disk blackbody and a 1.76 kev blackbody ; the blackbody component contributes @xmath280% of the 210 kev flux . by contrast , the soft states of ns xbs studied by lin et al . ( 2007 , 2009 , 2012 ) yielded disk blackbody temperatures @xmath521 kev , with the blackbody component contributing less than 50% of the 210 kev flux . since lin et al . ( 2007 , 2009 , 2012 ) failed to fit hard state spectra with the double thermal model , we expect disk blackbody + blackbody fits to our bhcs to inhabit a different parameter space from the ns binaries if they really are in the hard state . if a bhc exhibits a blackbody temperature @xmath201.5 kev , a disk blackbody temperature @xmath201.0 kev ( @xmath201.2 kev for disk blackbody 210 kev luminosity @xmath62@xmath21 erg s@xmath17 ) , and/or a blackbody contribution to the total 210 kev flux @xmath645% at a 3@xmath53 level , then we consider it a strong bhc ; otherwise , we label it a plausible bhc . these criteria are drawn from the hundreds of rxte spectra of aql x-1 , 4u1608@xmath752 , xtej1701@xmath7462 , and gx 17 + 2 analyzed by @xcite . table [ props1 ] gives the location of each of our bhcs , followed by the number of observations with acis and hrc . next we provide the best fit line of constant 0.310 kev luminosity over @xmath213 years , with the corresponding @xmath15 and number of degrees of freedom ( dof ) . finally , we show the number of observation pairs used to create the structure function . usually this is @xmath54 , where @xmath55 is the total number of observations ; however , bhcs 7 , 29 , 30 , and 35 included observations with zero luminosity that could not be included in the sf . all uncertainties in this work are quoted at the 1@xmath53 level unless specified otherwise . for nearly all bhcs , closely - spaced acis and hrc observations give consistent luminosities , although some small systematic offsets occur due to differences between assumed and actual spectra when fitting is not possible . however , bhc32 appears to have substantial signal in the hrc observations even when this transient source is in quiescence ; the hrc is considerably more sensitive than acis to low energy photons , hence the hrc obsevations of bhc32 must be contaminated by soft x - rays . bhc32 is located near 2 variable stars ( 16@xmath8 and 17@xmath8 distant ) ; since the extraction radius for this souce is 15@xmath8 due to the high off - axis angle , our hrc observations are likely contaminated by one or both foreground stars . as a result , we excluded all hrc observations of bhc32 from our analysis . . figure [ lgsbh ] shows a detail of the b band image of m31 field 5 from the lgs ( massey et al . , 2006 ) , superposed with the positions of our bhcs ; north is up , east is left . the circle encloses the region within 100@xmath3 ( @xmath2380 pc ) of m31 * ; this region is enlarged in the top right portion of the figure . bhcs 1 and 8 lie below the southern edge of field 5 . we find 7 bhcs within 100@xmath3 of m31@xmath56 , 20% of our total sample : bhcs 9 , 11 , 13 , 15 , 16 , 17 , and 18 . this result appears to be consistent with the predictions of @xcite . bhc9 is a recurrent transient with a peak 0.310 kev luminosity of @xmath28@xmath21 erg s@xmath17 . bhc11 was bright for @xmath212 years , then turned off . bhc18 appears to be persistently bright ( @xmath57 erg s@xmath17 ) , and may vary rapidly if the hrc luminosities are correct ; however , the emission from bhc18 is variable ( @xmath9 @xmath21.42.4 for bright acis observations ) , meaning that the hrc luminosities may be systematically offset due to differences between assumed and true spectra . bhcs 13 , 15 , 16 , and 17 are extremely variable , with bhcs 13 , 15 , and 17 luminosities ranging over @xmath216@xmath21 erg s@xmath17 , and bhc16 exhibiting luminosities from @xmath25@xmath28 to @xmath22@xmath58 erg s@xmath17 . we observed similar behavior in the x - ray source associated with b158 @xcite , a high inclination lmxb with a @xmath210000 s period @xcite , and an asymmetric , precessing disc @xcite . such behavior is seen in low mass ratio ( short period ) systems where the outer disc reaches the 3:1 resonance with the donor star , causing additional tidal torques that lead to elongation and precession of the disc ( see e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? b158 exhibited rapid variation over 440@xmath21 erg s@xmath17 , which we speculated might be due to varying accretion rates over the disc precession cycle ; we suggested that other xbs exhibiting such behaviour may also be in precessing ( i.e. short period ) systems @xcite . therefore we propose that bhcs 13 , 15 , 16 , and 17 are short period bh xbs , as predicted by @xcite . two of our new bhcs are associated with confirmed old gcs ; bhc4 is associated with b096 ( following the rbc naming convention ) , and bhc24 with b143 . @xcite rank b096 as the 36th most massive and 34th most metal rich out of 379 gcs ( [ fe / h ] = @xmath70.28 ) ; b143 is ranked 52nd most massive and 17th most metal rich , with [ fe / h ] = @xmath70.07 . we have found that the gcs hosting our previous bhcs were particularly massive , or metal rich , or both ( barnard et al . , 2011a , 2012 ) ; hence it is encouraging that the new gc bhcs are associated with gcs that are amongst the most massive and most metal rich of the m31 gcs studied by caldwell et al . ( 2009 , 2011 ) . in @xcite we classified the semi - regular transient associated with hii rich gc candidate sk059a as a possible hmxb associated with the hii , because the intervals between its outbursts were consistent with @xmath59 cycles of 120 days ; hence the accretor could make an eccentric , @xmath2120 day orbit around the high mass donor , with mass transfer restricted to times near periastron . this system is bhc30 . cccccccccccc bhc & obs & cnts & @xmath11/10@xmath60 @xmath14 & @xmath9 & @xmath15/dof & @xmath61 /10@xmath50 + 1 & 0109270101 & 3229 & [email protected] & [email protected] & 133/155 [ 0.89 ] & [email protected] + 2 & 303 & 704 & [email protected] & [email protected] & 33/29 [ 0.29 ] & [email protected] + 3 & 13827 & 5597 & 0.07 f & [email protected] & 161/182 [ 0.87 ] & [email protected] + 4 & 0690600401 & 2370 & [email protected] & [email protected] & 153/146 [ 0.32 ] & [email protected] + 5 & 13825 & 1425 & [email protected] & [email protected] & 8/11 [ 0.72 ] & [email protected] + 6 & 0112570101 & 10384 & [email protected] & [email protected] & 428/411 [ 0.26 ] & [email protected] + 7 & 0112570101 & 14523 & [email protected] & [email protected] & 519/490 [ 0.18 ] & [email protected] + 8 & 9524 & 352 & 0.07 f & [email protected] & 16/15 [ 0.37 ] & 17@xmath123 + 9 & 1575 & 3097 & [email protected] & [email protected] & 106/109 [ 0.55 ] & [email protected] + 10 & 13827 & 1679 & [email protected] & [email protected] & 72/67 [ 0.32 ] & [email protected] + 11 & 303 & 704 & [email protected] & [email protected] & 58/45 [ 0.10 ] & [email protected] + 12 & 13825 & 2620 & [email protected] & [email protected] & 90/98 [ 0.71 ] & [email protected] + 13 & 14197 & 1593 & [email protected] & [email protected] & 72/64 [ 0.24 ] & [email protected] + 14 & 9521 & 183 & 0.07 f & [email protected] & 6/7 [ 0.56 ] & 10@xmath123 + 15 & 14198 & 1860 & [email protected] & [email protected] & 75/74 [ 0.47 ] & [email protected] + 16 & 13825 & 4003 & [email protected] & [email protected] & 161/165 & [email protected] + 17 & 14197 & 1954 & [email protected] & [email protected] & 94/76 & [email protected] + 18 & 13825 & 4207 & [email protected] & [email protected] & 140/154 [ 0.78 ] & [email protected] + 19 & 0112570101 & 5032 & [email protected] & [email protected] & 210/226 [ 0.77 ] & [email protected] + 20 & 0402560901 & 20025 & [email protected] & [email protected] & 771/769 [ 0.44 ] & [email protected] + 21 & 13827 & 3847 & [email protected] & [email protected] & 127/142 [ 0.82 ] & [email protected] + 22 & 0112570101 & 10145 & [email protected] & [email protected] & 490/407 [ 3 e-3 ] & [email protected] + 23 & 14198 & 2508 & [email protected] & [email protected] & 111/95 [ 0.12 ] & [email protected] + 24 & 0112570101 & 6804 & [email protected] & [email protected] & 265/294 [ 0.89 ] & [email protected] + 25 & 0112570101 & 5116 & [email protected] & [email protected] & 254/236 [ 0.20 ] & [email protected] + 26 & 13825 & 1425 & [email protected] & [email protected] & 54/57 [ 0.60 ] & [email protected] + 27 & 14197 & 2099 & [email protected] & [email protected] & 69/75 [ 0.67 ] & [email protected] + 28 & 13826 & 635 & 0.07 f & [email protected] & 23/27 [ 0.69 ] & [email protected] + 29 & 1577 & 1375 & 0.07 f & [email protected] & 79/58 [ 0.04 ] & 31@xmath122 + 30 & 13299 & 274 & 0.07 f & [email protected] & 10/10 [ 0.42 ] & [email protected] + 31 & 0112570101 & 14338 & [email protected] & [email protected] & 450/488 [ 0.89 ] & [email protected] + 32 & 8184 & 521 & 0.07 f & [email protected] & 27/23 [ 0.25 ] & [email protected] + 33 & 13825 & 1411 & [email protected] & [email protected] & 57/60 [ 0.57 ] & [email protected] + 34 & 0112570101 & 5181 & [email protected] & [email protected] & 246/230 [ 0.22 ] & [email protected] + 35 & 14198 & 1646 & [email protected] & [email protected] & 85/69 [ 0.10 ] & [email protected] + we initially identified most of our bhcs from their long term ( @xmath213 year ) light curves and spectral histories . they exhibited spectra consistent with the hard state ( @xmath9 @xmath202.1 ) at 0.310 kev luminosities @xmath63@xmath21 erg s@xmath17 in at least one observation . for such spectra , we expect the 0.011000 kev luminosities to be @xmath623@xmath21 erg s@xmath17 , our limit for neutron star hard states following @xcite . we present the @xmath213 year , corrected , 0.310 kev luminosity lightcurve for each bhc in fig . [ lcs ] , along with its spectral history . the top panel in each case shows the lightcurve , with acis and hrc observations represented by circles and crosses respectively . the bottom panel shows @xmath9 for each acis observation ; sufficiently bright observations have freely fitted @xmath9 , while faint observations have @xmath9 fixed to the mean value ; this approach should be valid because we expect the faint observations to be in the hard state also , with similar a emission spectrum . figure [ hist ] shows histograms of the number of observation pairs with separation @xmath24 for the most frequently observed bhc , bhc13 . the histogram with filled circles includes acis and hrc observations , while the open circle histogram contains only acis observations . the observation pairs were logarithmically binned by @xmath24 , with a bin width of 0.2 dex . we present the 0.54.5 kev structure function for each of our bhcs in fig . [ sfs ] , with same scheme as for fig . the ensemble sf for typical agn obtained by vagnetti et al . ( 2011 ) is represented by a dashed line in each panel : sf(@xmath24 ) = 0.11@xmath63 . we included an acis only sf for each source in addition to an acis + hrc sf because faint hrc observations add extremely large noise components for some bhcs , suppressing the variation ; bhcs 8 and 35 only show significant excess variability over typical agn in their acis only sfs . we find that 28 out of 35 bhcs exhibit sfs with significantly more variation than the vagnetti et al . ( 2011 ) ensemble agn sf ( @xmath64@xmath53 excess in at least one @xmath24 channel , @xmath63@xmath53 excess in at least 2 channels , or @xmath62@xmath53 excess in at least 3 channels over the 3@xmath53 limit for the agn sf , assumed to be 0.11@xmath64 ) . this distinction is important , because agn and xbs often have very similar spectra . the x - ray sources with sfs exhibiting similar or less variation than the ensemble agn sf ( bhcs 2 , 5 , 7 , 21 , 23 , 24 , 25 , and 34 ) had mean 0.310 kev fluxes @xmath525.5@xmath37 erg cm @xmath65 s@xmath17 . we have found 24 x - ray sources in our observations that exhibited mean 0.310 kev fluxes @xmath525.5@xmath6610@xmath67 erg @xmath14 s@xmath17 over the @xmath213 years of chandra monitoring ; this includes 8 gc xbs and 1 possible supernova remnant . the 0.510 kev flux distribution of agn found by @xcite predicts 0.6 agn at this flux level within our observed region ( approximated by a circle with radius @xmath220@xmath1 or @xmath24 kpc ) . since the observed number of unidentified x - ray sources at these flux levels is larger than the predicted agn number by a factor @xmath225 , we conclude that they are likely xbs . after the initial identification of candidates , we sought the observations that best represented the bhc case for each object ; these were the highest quality spectra that appeared to exhibit high luminosity hard states . table [ bestspec ] provides the details of these observations . for each bhc we give the observation number , and net sources counts . we then give the absorption ( @xmath11 ) , photon index ( @xmath9 ) , @xmath15/dof and 0.310 kev luminosity for the best fit absorbed power law model . observations 30313827 are chandra acis observations , while 0109270101 , 0112570101 , and 0402560901 are xmm - newton observations . we note that the spectrum of bhc 22 rejected simple absorbed power law models . it is well described by an absorbed disk blackbody + power law model with @xmath68 = 1.97@xmath69 kev , and @xmath9 = 2.16@xmath70 , with the power law contributing 61@xmath1212% of the [email protected] @xmath21 erg s@xmath17 0.310 kev luminosity ; it is consistent with either of two canonical black hole states : the hard state , or a steep power law state similar to grs 1915 + 105 ( mcclintock & remillard , 2006 ) . we applied a disk blackbody + blackbody emission model to our bhc spectra . the disk blackbody components were generally better constrained than the blackbody components , particularly for spectra with relatively few net source counts ; this is simply because there are fewer high energy photons than low energy photons and the disk blackbody is cooler than the blackbody . our best fits yielded blackbody contributions of 6599% to the 210 kev flux , in stark contrast to the neutron star systems where the disk blackbody dominates . the blackbody component dominates the 210 kev flux most when the disk blackbody temperature is lowest . four of our bhcs only exhibited apparent high luminosity hard states during 5 ks acis observations , and we were unable to constrain double thermal models for the resulting spectra . however , future deep observations may strengthen their bhc cases . the 31 bhcs that we did fit with the double thermal model yielded k@xmath68 0.628@xmath53 outside the ns parameter space ; the combined @xmath15 was 212 , for 30 dof ( probability @xmath23@xmath71 ) . clearly our bhcs are inconsistent with being soft ns xbs . ccccccccccccc bhc & @xmath11 / 10@xmath60 @xmath14 & k@xmath72 / kev & @xmath73/10@xmath50 & k@xmath74 / kev & @xmath75/10@xmath50 & @xmath76 & bh + 1 & [email protected] & 2.00@xmath77 & 21@xmath122 & 0.88@xmath78 & [email protected] & [email protected] & p + 2 & 0.07f & 1.12@xmath79 & [email protected] & 0.50@xmath80 & [email protected] & [email protected] & p + 3 & 0.07f & 1.18@xmath81 & [email protected] & 0.49@xmath82 & [email protected] & [email protected] & s + 4 & [email protected] & 1.50@xmath83 & [email protected] & 0.62@xmath84 & [email protected] & [email protected] & p + 5 & 0.07f & 1.3@xmath85 & [email protected] & 0.7@xmath86 & [email protected] & [email protected] & p + 6 & 0.07f & 1.34@xmath87 & [email protected] & 0.510@xmath88 & [email protected] & [email protected] & s + 7 & 0.07f & 1.47@xmath89 & [email protected] & 0.51@xmath90 & [email protected] & [email protected] & s + 8 & & & & & & & p + 9 & 0.07f & 0.93@xmath82 & [email protected] & 0.38@xmath91 & [email protected] & [email protected] & s + 10 & 0.07f & 2.4@xmath92 & [email protected] & 0.86@xmath93 & [email protected] & [email protected] & p + 11 & 0.07f & 1.11@xmath79 & [email protected] & 0.50@xmath80 & [email protected] & [email protected] & p + 12 & 0.07f & 1.24@xmath94 & [email protected] & 0.59@xmath95 & [email protected] & [email protected] & s + 13 & 0.07f & 1.5@xmath96 & [email protected] & 0.67@xmath97 & [email protected] & [email protected] & p + 14 & & & & & & & p + 15 & 0.07f & 1.12@xmath79 & [email protected] & 0.56@xmath98 & [email protected] & [email protected] & p + 16 & 0.07f & 2.2@xmath99 & [email protected] & 1.03@xmath93 & [email protected] & [email protected] & p + 17 & 0.07f & 0.96@xmath100 & [email protected] & 0.43@xmath87 & [email protected] & [email protected] & s + 18 & 0.07f & 1.05@xmath101 & [email protected] & 0.41@xmath102 & [email protected] & [email protected] & s + 19 & [email protected] & 1.7@xmath103 & [email protected] & 0.83@xmath93 & [email protected] & [email protected] & p + 20 & [email protected] & 1.61@xmath104 & [email protected] & 0.83@xmath105 & [email protected] & [email protected] & s + 21 & [email protected] & 1.7@xmath106 & [email protected] & 0.77@xmath107 & [email protected] & [email protected] & p + 22 & 0.07f & 1.31@xmath87 & [email protected] & 0.55@xmath108 & [email protected] & [email protected] & s + 23 & 0.07f & 0.99@xmath109 & [email protected] & 0.45@xmath110 & [email protected] & [email protected] & s + 24 & 0.07f & 1.12@xmath87 & [email protected] & 0.43@xmath90 & [email protected] & [email protected] & s + 25 & 0.07f & 1.56@xmath95 & [email protected] & 0.61@xmath111 & [email protected] & [email protected] & s + 26 & 0.07f & 2.3@xmath112 & [email protected] & 0.78@xmath113 & [email protected] & [email protected] & p + 27 & 0.07f & 0.94@xmath114 & [email protected] & 0.46@xmath81 & [email protected] & [email protected] & s + 28 & 0.07f & 1.4@xmath115 & [email protected] & 0.63@xmath97 & [email protected] & [email protected] & p + 29 & 0.07f & 1.00@xmath116 & [email protected] & 0.29@xmath102 & [email protected] & [email protected] & s + 30 & & & & & & & p + 31 & 0.07f & 1.36@xmath117 & [email protected] & 0.45@xmath118 & [email protected] & [email protected] & s + 32 & & & & & & & p + 33 & 0.07f & 1.15@xmath119 & [email protected] & 0.52@xmath120 & [email protected] & [email protected] & s + 34 & 0.07f & 1.56@xmath121 & [email protected] & 0.63@xmath111 & [email protected] & [email protected] & s + 35 & 0.07f & 1.44@xmath122 & [email protected] & 0.68@xmath123 & [email protected] & [email protected] & p + in figs . [ dbbvl ] and [ bbvl ] we plot the 2.010 kev luminosity vs. temperature for the disk blackbody and blackbody components of the best fit double thermal models for each of our bhcs . we overlay the parameter space inhabited by the neutron star systems observed by lin et al . ( 2007 , 2009 , 2012 ) for spectral states where comptonization is not required in their models , i.e. the atoll soft states and z - source normal and flaring branches . we note that lin et al . quote luminosities from broader energy ranges ; however , our disk blackbody components make a negligible contribution above 10 kev , while the luminosities of our blackbody components are typically @xmath210% higher in the 2.020 kev band than the 2.010 kev band ; hence our observed 2.010 kev luminosity for our bhcs provide a fair comparison with the neutron star systems observed by lin et al . ( 2007 , 2009 , 2012 ) . as predicted , our bhcs inhabit a separate region of the disk blackbody parameter space to the ns systems studied by lin et al . ( 2007 , 2009 , 2012 ) , although some sources are consistent with ns values for @xmath68 within 3@xmath53 . we also see a correlation between the 210 kev luminosity and temperature of the disk blackbody component ; this is due to lower disk temperatures resulting in lesser contributions to the 210 kev flux . the bhcs are scattered more widely in the blackbody space , but are also systematically offset from the ns systems . remembering that lin et al . ( 2009 , 2012 ) failed to fit the horizontal branches of xte j1701@xmath7462 and gx 17 + 2 with their double thermal model , we also modeled our brightest bhcs ( @xmath124 @xmath6 10@xmath125 erg s@xmath17 ) with the disk blacbkody + blackbody + comptonization emission model that they employed for the horizontal branch spectra . @xcite used a cut - off power law to model the comptonization in gx 17 + 2 over the 2.960 kev band ; they required simultaneous fitting of several spectra to constrain the parameters of the cut - off power law , which yielded @xmath9 = [email protected] , and a cut - off energy of [email protected] kev . we therefore modeled the 0.310 kev xmm - netwon pn spectra , or 0.37.0 kev chandra acis spectra of our brightest bhcs with a blackbody , disk blackbody and a power law with @xmath9 fixed to 1.4 . @xcite found that the compotonized component contributed @xmath21050% of the total flux on the horizontal branch of gx 17 + 2 ; we fixed the flux of the power law component in our fits to @xmath230% of the total 210 kev flux . we fitted the spectra of bhcs 1 , 3 , 16 , 20 , 21 , and 31 in this way ; the results are summarized in table [ bdp ] . we find that this three component model results in lower temperatures for both thermal components than the double thermal model fits to the same spectra . i.e. , the addition of the power law component drives our bhcs away from the ns parameter space . increasing the power law contribution to 50% of the flux decreased the temperatures further . since none of the spectra from our brightest bhcs have three component fits that are consistent with the fits that successfully described the horizontal branch ( lin et al . , 2009 , 2012 ) , we conclude that none of our bhcs are z - sources . we have identified 26 new black hole candidates in the central region of m31 , using their structure functions or luminosities to identify them as x - ray binaries , and their high luminosity hard state spectra to classify them as bhcs . of these , 12 are strong candidates , and 14 are plausible candidates that may benefit from further observations . we were previously limited to identifying bhcs in globular clusters , due to the similarities between xb and agn spectra . this brings the total number of bhcs within 20@xmath1 of m31@xmath56 identified by their high luminosity hard states to 35 . ccccc bhc & @xmath11 / 10@xmath60 @xmath14 & k@xmath72 / kev & k@xmath74 / kev + 1 & 0.35 f & 1.99@xmath126 & 1@xmath103 + 3 & 0.07 f & 1.07@xmath127 & 0.42@xmath128 + 16 & 0.07 f & 1.44@xmath129 & 0.8@xmath106 + 20 & [email protected] & 1.29@xmath107 & 0.73@xmath130 + 21 & [email protected] & 1.6@xmath131 & 0.7@xmath132 + 31 & 0.07 f & 1.27@xmath133 & 0.43@xmath108 + the structure functions of most of our bhcs reveal them to be substantially more variable than typical agn ( as measured by vagnetti et al . , 2011 ) over a wide range of time scales . those bhcs with comparable or less variability than the agn have 0.510 kev luminosities matched by @xmath20.6 agn within the observed region , according to @xcite . it is therefore unlikely that our new bhcs are agn . we have found that our bhc spectra exist in a separate parameter space to galactic neutron star systems when we compare our best fits for absorbed blackbody + disk blackbody emission models with the systems studied by lin et al . ( 2007 , 2009 , 2012 ) . this is expected because this double thermal model fails to fit hard state spectra ; the disk blackbody component is forced low temperatures in order to provide the low energy flux . indeed , the probability that our bhcs all lie in the ns parameter space when fitted with the double - thermal model is just @xmath23@xmath71 . seven of our bhcs ( 20% of our total sample ) were found within 100@xmath3 of the m31 nucleus , lending support to the hypothesis that the m31 bulge is sufficiently dense to form a significant number of xbs dynamically , as seen in globular clusters ; since the stellar velocities in the m31 bulge are considerably higher than in gcs , surviving xbs are expected to have short periods and are likely to contain black hole accretors @xcite . we thank the referee for their thoughtful comments . we thank z. li for merging the chandra data . this research has made use of data obtained from the chandra data archive , and software provided by the chandra x - ray center ( cxc ) . we also include analysis of data from xmm - newton , an esa science mission with instruments and and contributions directly funded by esa member states and the us ( nasa ) . is funded by chandra grants go0 - 11106x and go1 - 12109x , along with hst grants go-11833 and go-12014 . m.r.g . and s.s.m are partially supported by nasa grant nas-03060 .	we have previously identified 10 m31 black hole candidates ( bhcs ) in m31 , from their x - ray properties alone . they exhibit `` hard state '' emission spectra that are seen at luminosities @xmath010% eddington in x - ray binaries ( xbs ) containing a neutron star ( ns ) or black hole ( bh ) , at luminosities that significantly exceed the ns threshold . nine of these are associated with globular clusters ( gcs ) ; hence , these are most likely low mass x - ray binaries ( lmxbs ) ; eight are included in this survey . we have recently discovered that analysis of the long term 0.54.5 kev variability of xbs via structure functions allows us to separate xbs from agn , even though the emission spectra are often similar ; this has enabled us to search for bhcs outside of gcs . we have identified 26 new bhcs ( 12 strong , 14 plausible ) within 20@xmath1 of the m31 nucleus ( m31 * ) , using 152 chandra observations spaced over @xmath213 years ; some of our classifications were enhanced with xmm - newton observations . of these , 7 appear within 100@xmath3 of m31 * ; this supports the theory suggesting that this region experiences enhanced xb production via dynamical processes similar to those seen in gcs . we have found a parameter space where our black hole candidates are separated from galactic neutron star binaries : we show that modelling a simulated hard state spectrum with a disk blackbody + blackbody model yields parameters that lie outside the space occupied by neutron star binaries that are modeled this way . the probability that our bhcs all lie within the ns parameter space is @xmath4 .
un rsultat classique d brouwer nonce que tout homomorphisme du plan @xmath3 qui prserve lorientation et possde une orbite priodique , possde galement un point fixe . dans le mme ordre dides , on peut montrer @xcite quun tel homomorphisme @xmath0 possde un point fixe li cette orbite priodique , en ce sens quil nexiste pas de courbe de jordan @xmath4 , bordant un disque @xmath5 contenant lorbite priodique mais ne contenant pas le point fixe , et telle que @xmath6 soit homotope @xmath4 dans le complmentaire de lorbite priodique et du point fixe . une question pose par john franks dans @xcite demeure toujours sans rponse : tant donn un homomorphisme @xmath7 prservant lorientation , existe - t - il pour toute orbite priodique de @xmath0 un point fixe ayant un nombre denlacement non nul avec cette orbite priodique ? on sait que la rponse cette question est positive pour les orbites de priode @xmath8 ( voir @xcite ) ou de priode @xmath9 @xcite . dautre part , dans une prpublication rcente , franks @xcite utilise la rponse affirmative cette question comme tant un thorme de handel ( sans rfrence ) : on peut donc supposer que cette question est , soit rsolue , soit en passe de ltre . nous montrons ici quun raisonnement trs simple et trs rapide permet de rpondre par laffirmative la question de franks pour les orbites de toutes les priodes des homomorphismes @xmath0 de @xmath3 pour lesquels @xmath1 vrifie un condition de lipschitz . plus prcisment : soit @xmath7 un homomorphisme qui prserve lorientation et tel que @xmath1 soit lipschitzienne de rapport @xmath10 $ ] . alors , pour toute orbite priodique @xmath11 de @xmath0 , il existe un point fixe de @xmath0 , @xmath12 , ayant un nombre denlacement non nul avec @xmath2 . outre lintrt du rsultat , la simplicit de la preuve met en valeur lavantage quil y a tester sur cette classe ( pas trop petite ) dhomomorphismes , les conjectures concernant les homomorphismes des surfaces . soit @xmath7 un homomorphisme . on note @xmath13 lensemble de ses points fixes . un point @xmath14 est priodique de priode @xmath15 si @xmath16 mais @xmath17 pour @xmath18 . on note @xmath19 lorbite de @xmath20 sous @xmath0 . soit @xmath21 , @xmath20 un point priodique de priode @xmath15 et @xmath22 un arc joignant @xmath20 et @xmath23 dans @xmath24 , on note @xmath25 la courbe ferme obtenue en joignant bout bout les arcs , @xmath26 . on note @xmath27 le nombre denroulement de @xmath25 autour de @xmath12 cest dire le nombre dintersection algbrique dune demi - droite gnrique issue de @xmath12 avec @xmath25 ( ce nombre est souvent appel indice de @xmath12 par rapport @xmath25 ) . [ lem1 ] soient @xmath22 et @xmath28 deux arcs quelconques joignant @xmath20 et @xmath23 dans @xmath29 . si @xmath0 prserve lorientation , alors @xmath30 . on a @xmath31 or si @xmath0 prserve lorientation @xmath32 do @xmath33 . on peut montrer galement que la valeur de @xmath34 ( mod @xmath15 ) ne dpend pas du choix du point @xmath20 de @xmath2 choisi pour le construire . avec les notations ci - dessus , on note @xmath35 lunique entier @xmath36 tel que @xmath37 pour un choix quelconque de @xmath22 et on lappelle le _ nombre denlacement _ ( ou _ linking number _ ) du point fixe @xmath12 avec lorbite priodique @xmath2 . lappellation nombre denlacement est justifie par la remarque suivante : @xmath38 est aussi le nombre denlacement des deux orbites fermes @xmath39 et @xmath40 du champ de vecteurs canonique induit dans la suspension @xmath41 de @xmath0 . le point fixe de la rotation dangle @xmath42 a pour nombre denlacement @xmath43 avec lune quelconque de ses orbites priodiques . soit @xmath44 une application continue . si lensemble @xmath45 est non vide , on note @xmath46 sa borne infrieure . sinon , on pose @xmath47 . soient @xmath20 et @xmath48 deux points de @xmath3 , on note @xmath49 $ ] le segment de droite joignant @xmath20 et @xmath48 . [ lem2 ] soit @xmath7 un homomorphisme tel que @xmath50 et @xmath51 . alors @xmath52=\emptyset$ ] . par labsurde , supposons quil existe @xmath53 . on a alors : @xmath54 dautre part @xmath55 car @xmath0 est injective , ce qui conclut . [ lem3 ] soit @xmath56 un homomorphisme tel que @xmath57 et @xmath51 . alors , @xmath58)$ ] et @xmath59 $ ] sont homotopes relativement @xmath23 , @xmath60 dans @xmath61 . soit @xmath62^{2}\rightarrow { \mathbb r}^{2}$ ] dfinie par : @xmath63 limage de @xmath64 est la runion des segments de droite @xmath65(y\in[x , f(x)])$ ] . daprs le lemme [ lem2 ] @xmath66 $ ] nest pas un point fixe de @xmath0 , donc @xmath65\subset { \mathbb r}^{2}\setminus fix(f)$ ] . do @xmath67 . dautre part , @xmath68 . on obtient donc une application du disque : @xmath69^{2}/(s , 0)\sim(0 , s)-{\mathbb r}^{2}\setminus fix(f)\ ] ] telle que @xmath70\cup f([x , f(x)])$ ] . on a ainsi ralis lhomotopie . soient @xmath71 , @xmath72 deux vecteurs de @xmath3 . on note @xmath73 $ ] langle non orient des vecteurs @xmath71 et @xmath72 . [ lem4 ] soit @xmath7 un homomorphisme tel que @xmath50 et @xmath51 . alors pour tout @xmath66 $ ] : @xmath74 on a : @xmath75 et par suite : @xmath76 . si langle est @xmath77 , cela implique @xmath78 ce qui est impossible en vertu du lemme [ lem2 ] . soit @xmath79 lorbite dun point priodique de priode @xmath15 dun homomorphisme de @xmath3 qui prserve lorientation et tel que @xmath50 . soit @xmath80 la courbe polygonale obtenue en joignant bout bout les segments @xmath81 $ ] , @xmath59 $ ] , @xmath82 , @xmath83 $ ] ( voir figure [ fig1 ] ) et soit @xmath84 $ ] . lassertion ( 1 ) rsulte du lemme [ lem2 ] . en raisonnant par rcurrence et en utilisant le lemme [ lem3 ] , on tablit que @xmath80 est homotope @xmath25 dans @xmath24 , do lgalit @xmath87 . par ailleurs , le nombre denroulement de @xmath80 par rapport @xmath12 est aussi le nombre algbrique de croisements dune demi - droite issue de @xmath12 avec @xmath80 . ce nombre est donc ncessairement infrieur @xmath88 en valeur absolue . soient @xmath89 les composantes connexes bornes de @xmath90 ( remarquer quil en existe au moins une , sinon @xmath80 serait rduit un segment de droite ce qui est exclu en vertu du lemme [ lem4 ] ) et @xmath91 la composante connexe non borne . si @xmath92 on dfinit lindice de @xmath93 en posant : @xmath94 o @xmath95 dsigne une dtermination continue de langle du vecteur @xmath96 avec une direction fixe et o @xmath97 est le bord orient de @xmath93 . dans chaque composante dindice non nul il existe au moins un point fixe de @xmath0 . soient @xmath98 les sommets de @xmath97 et @xmath99 ses arrtes munies de lorientation induite par celle de @xmath80 . en un sommet @xmath100 , il y a deux configurations possibles quant lorientation des artes adjacentes @xmath100 : ou bien ces deux arrtes ont des orientations compatibles , ou bien il y a un changement dorientation ( voir figure [ fig2 ] ) . on a : @xmath103 o @xmath104 et @xmath105 sont les valeurs respectives en @xmath100 et @xmath106 ( @xmath107 ) dune dtermination continue @xmath108 de langle @xmath109 ( @xmath110 ) avec la tangente oriente @xmath111 . il rsulte alors du lemme [ lem4 ] que @xmath112 $ ] ( autrement dit le vecteur @xmath109 ne dcrit pas de tour complet lorsque @xmath48 parcourt @xmath111 ) . en dsignant alors par @xmath113 $ ] langle intrieur @xmath93 en @xmath100 , on a ( voir figure [ fig3 ] ) : @xmath114 par suite : @xmath115\\ & = \frac{1}{2\pi}\left[\sum_{i=0}^{m-1}(\varphi_{i-1}^{1}-\varphi_{i}^{0})\right]\\ & = \frac{1}{2\pi}\left[\sum_{i=0}^{m-1}(\pi-\beta_{i})-2p\pi\right]\\ & = 1-p.\end{aligned}\ ] ] nous pouvons maintenant tablir le thorme . le nombre denroulement @xmath116 dun point @xmath117 ne dpendant que de la composante @xmath93 laquelle il appartient , on notera @xmath118 cette valeur commune . il nous reste donc tablir lexistence dune composante @xmath119 de @xmath90 telle que : @xmath120 remarquons dabord que @xmath124 et que @xmath125 pour toute composante @xmath119 adjacente @xmath91 . il existe donc bien @xmath126 tel que : @xmath127 par labsurde , supposons que @xmath128 . daprs le lemme [ lem6 ] , il existe alors au moins un changement dorientation en un des sommets @xmath129 de @xmath130 . alors lune des composantes @xmath131 adjacente @xmath132 en @xmath129 vrifie @xmath133 et lautre @xmath134 vrifie @xmath135 ( voir figure [ fig4 ] ) . en effet , les valeurs de @xmath136 et @xmath137 ne dpendent que de lorientation avec laquelle on franchit @xmath80 pour passer de @xmath132 @xmath131 et @xmath134 . il existe donc @xmath138 ( @xmath139 ) tel que : @xmath140 ce qui contredit lhypothse faite sur @xmath141 .	soit @xmath0 un homomorphisme du plan qui prserve lorientation et tel que @xmath1 soit une contraction . sous ces hypothses , on tablit lexistence , pour toute orbite priodique @xmath2 , dun point fixe ayant un nombre denlacement non nul avec @xmath2 .
it is important to have good leading order ( or leading - log order ) estimates of hadron - hadron collider observables . of course , precise comparisons with data require beyond leading order calculations . if the leading order estimates are within a factor of two of the data , higher order calculations might be expected to improve the agreement . on the other hand , if the leading order estimates are off by more than about a factor of two of the data , one can not expect higher order calculations to improve the situation . in this case , even if the higher order corrections were large enough to bring agreement , one could not trust a perturbative series in which the second term is greater than the first . if a leading order estimate is off by more than a factor of two , it usually means that one has overlooked some important physics . for this reason good leading - log order estimates are important . -to-@xmath3 reaction results in the creation of a @xmath2 pair via the subprocess @xmath4 or @xmath5 . the event also contains possible initial and final - state radiation and particles that result from the break - up of the initial proton and antiproton ( _ i.e. _ , beam - beam remnants " ) . ] -quark total cross section ( @xmath6 , @xmath7 ) for proton - antiproton collisions at @xmath8 compared with the qcd monte - carlo model predictions of herwig 5.9 , pythia 6.115 , and isajet 7.32 for the flavor creation " subprocesses illustrated in fig . [ prd_fig1 ] . all three monte - carlo models were generated using the parton distribution functions cteq3l and @xmath9 . ] in this analysis the leading - log order qcd hard scattering monte - carlo models of herwig @xcite , isajet @xcite , and pythia @xcite are used to study the sources of @xmath0-quarks at the tevatron . the reactions responsible of producing @xmath0-quarks are separated into three categories ; flavor creation , flavor excitation , and shower / fragmentation . flavor creation corresponds to the production of a @xmath2 pair by gluon fusion or by the annihilation of light quarks via the two @xmath3-to-@xmath3 parton subprocesses , @xmath4 , and @xmath5 , and is illustrated in fig . [ prd_fig1 ] . the data from cdf and d0 @xcite for the integrated inclusive @xmath0-quark cross section for @xmath7 at @xmath8 are compared with the qcd monte - carlo model predictions for flavor creation in fig . [ prd_fig2 ] , where @xmath10 is the rapidity of the @xmath0-quark . here the parton distribution functions cteq3l have been used for all three monte - carlo models and , as is well know , the leading order predictions are roughly a factor of four below the data . the leading order estimates of the flavor creation contribution to @xmath0-quark production at the tevatron are so far below the data that higher order corrections ( _ even though they may be important _ ) can not be the whole story . -to-@xmath3 reaction corresponds to the scattering of a @xmath0-quark out of the initial - state into the final - state by a gluon or a light quark or light antiquark via the subprocess @xmath11 , @xmath12 , or @xmath13 . these subprocesses together with the corresponding @xmath1 terms are referred to as flavor excitation " . the event also contains possible initial and final - state radiation and particles that result from the break - up of the initial proton and antiproton ( _ i.e. _ , beam - beam remnants " ) . ] an additional source of @xmath0-quarks at the tevatron comes from the scattering of a @xmath0 or @xmath1 quark out of the initial - state into the final - state by a gluon or by a light quark or antiquark via the subprocesses ; @xmath11 , @xmath14 , @xmath12 , @xmath15 , @xmath13 , and @xmath16 . this is referred to as flavor excitation " and is illustrated in fig . [ prd_fig3 ] . flavor excitation is , of course , very sensitive to the number of @xmath0 and @xmath1 quarks within the proton ( _ i.e. _ , the structure functions ) . the @xmath0 and @xmath1 quarks are generated through the @xmath17 evolution of the structure functions . even with no intrinsic " @xmath2 pairs within the proton , at high @xmath17 @xmath2 pairs are produced by gluons and populate the proton sea " . the number of @xmath2 pairs within the proton is related , through the @xmath17 evolution , to the gluon distribution within the proton . none of the structure functions considered in this analysis include intrinsic " @xmath2 pairs within the proton . the @xmath2 pair content within the proton is generated entirely through the @xmath17 evolution of the structure functions . pair is created within a parton - shower or during the fragmentation process of a gluon or a light quark or antiquark . here the qcd hard @xmath3-to-@xmath3 subprocess involves only gluons and light quarks and antiquarks ( no heavy quarks in the @xmath3-to-@xmath3 hard scattering subprocess ) . the event also contains possible initial and final - state radiation and particles that result from the break - up of the initial proton and antiproton ( _ i.e. _ , beam - beam remnants " ) . ] another source of @xmath0-quarks at the tevatron comes from reactions which have a @xmath2 in the final - state but with only gluons and light quarks and light antiquarks participating in the @xmath3-to-@xmath3 hard parton scattering subprocess ( _ i.e. _ , no heavy quarks in the @xmath3-to-@xmath3 hard scattering subprocess ) . this is referred to as shower / fragmentation " and is illustrated in fig . [ prd_fig4 ] . here the subprocesses are all qcd @xmath3-to-@xmath3 gluon , light quark , and light antiquark subprocesses . the shower / fragmentation " contribution comes from @xmath2 pairs produced within parton - showers or during the fragmentation process . this category includes the gluon splitting " subprocess , @xmath18 , as modeled by the qcd leading - log monte - carlo models . section ii discusses methods of generating the three sources of @xmath0-quarks ; flavor creation , flavor excitation , and shower / fragmentation . in section iii the leading - log qcd monte - carlo model predictions are compared with data on the @xmath0-quark inclusive cross section at the tevatron . correlations between the @xmath0 and @xmath1 quark are very different for the three sources and in section iv we study ways to isolate the individual contributions using @xmath2 correlations . section v is reserved for summary and conclusions . -quark total cross section ( @xmath6 , @xmath7 ) for proton - antiproton collisions at @xmath8 compared with the qcd monte - carlo model predictions of pythia 6.115 ( cteq3l , @xmath9 ) . the four curves correspond to the contribution from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] -quark total cross section ( @xmath6 , @xmath7 ) for proton - antiproton collisions at @xmath8 compared with the qcd monte - carlo model predictions of isajet 7.32 ( cteq3l , @xmath9 ) . the four curves correspond to the contribution from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] -quark total cross section ( @xmath6 , @xmath7 ) for proton - antiproton collisions at @xmath8 compared with the qcd monte - carlo model predictions of herwig 5.9 ( cteq3l , @xmath9 ) . the four curves correspond to the contribution from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] it is not easy to arrive at the qcd monte - carlo model predictions for @xmath0-quark production at the tevatron . some of the plots presented here contain @xmath19 million generated events . in addition , one must handle each of the monte - carlo models differently in order to get all three contributions ; flavor creation , flavor excitation , and shower / fragmentation . the flavor creation contribution to @xmath0-quark production at the tevatron is the easiest to generate . isajet , herwig , and pythia all allow the user to select and generate separately the flavor creation contribution . for isajet if one runs all qcd and then selects the flavor creation terms by monitoring the @xmath3-to-@xmath3 subprocess , one gets the same answer as running the flavor creation contribution separately . however , this is not true for herwig and pythia . these two monte - carlo models only include the @xmath0-quark mass when one runs the flavor creation terms separately ( _ i.e. _ , heavy quark production ) . for herwig and pythia , when one runs all qcd the @xmath0-quark mass is set to zero . isajet always includes the @xmath0-quark mass in the flavor creation subprocesses . setting the @xmath0-quark mass equal to zero increases the @xmath0-quark inclusive flavor creation cross section for @xmath20@xmath21 and @xmath7 by roughly @xmath22 at @xmath8 . in summary , one can produce the flavor creation contribution in two ways with isajet , either by running them separately or by running all qcd . for herwig and pythia one must generate the flavor creation terms separately by running the heavy quark option ( iproc = 1705 or msel = 5 ) . for pythia this produces only the flavor creation contribution . however , for herwig the heavy quark option produces both the flavor creation terms and the flavor excitation terms and one must monitor the @xmath3-to-@xmath3 subprocesses to separate them . for isajet one can run the flavor excitation contribution separately or one can run all qcd and then select out the flavor excitation contribution ( one gets the same answer either way ) . for herwig running the heavy quark option gets both the flavor creation and the flavor excitation contributions ( with the @xmath0-quark mass included ) . for herwig , if one runs all qcd and then selects out the flavor excitation contribution one gets a different answer since here the @xmath0 mass is set to zero . for pythia the only way to get the flavor excitation contribution is to run all qcd and then select out the flavor excitation contribution , but here one only gets the massless approximation . actually , in leading order it is not clear how to handle the @xmath0-quark mass in the flavor excitation subprocesses since one is going from an off - shell @xmath0 ( or @xmath1 ) quark in the initial state to an on - shell massive @xmath0 ( or @xmath1 ) quark in the final state . this limits the accuracy of the leading order estimate of this contribution . for all three monte - carlo models the only way to get the shower / fragmentation contribution to @xmath0-quark production is to run all qcd @xmath3-to-@xmath3 subprocesses and then select out the shower / fragmentation contribution by monitoring the subprocesses , which requires producing large numbers of events . the shower / fragmentation contribution differs greatly among the monte - carlo models for two reasons . the first is due to different fragmentation schemes . isajet uses independent fragmentation , while herwig and pythia do not . the second difference arises from the way the qcd monte - carlo produce parton showers . isajet uses a leading - log picture in which the partons within the shower are ordered according to their invariant mass . kinematics requires that the invariant mass of daughter partons are less than the invariant mass of the parent . herwig and pythia modify the leading - log picture to include color coherence effects " which leads to angle ordering " within the parton shower . the flavor excitation and shower / fragmentation @xmath3-to-@xmath3 subprocesses diverge as the parton transverse momentum , @xmath23 , goes to zero . to get a finite result one must regulate the divergences or specify a minimum value for @xmath23 . for herwig and isajet we take @xmath9 . pythia regulates the perturbative @xmath3-to-@xmath3 parton scattering differential cross sections , which allow one to take @xmath23@xmath24 . for pythia @xmath0-quark predictions with @xmath20(@xmath0-quark ) @xmath21 are independent of the whether one takes @xmath9 or @xmath23@xmath24 . -quark differential cross section , @xmath25 , for @xmath7 for proton - antiproton collisions at @xmath8 resulting from pythia 6.158 ( cteq4l , @xmath23@xmath24 ) . the four curves correspond to the contribution from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] -quark total cross section ( @xmath6 , @xmath7 ) for proton - antiproton collisions at @xmath8 compared with the qcd monte - carlo model predictions of pythia 6.158 ( cteq4l , @xmath23@xmath24 ) . the four curves correspond to the contribution from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] the data from cdf and d0 @xcite on the integrated inclusive @xmath0-quark total cross - section are compared with the qcd monte - carlo model predictions in figs . [ prd_fig5]- [ prd_fig7 ] , with the cteq3l structure functions and @xmath9 . the four curves in each of the plots correspond to the contribution to @xmath0-quark production from flavor creation , flavor excitation , shower / fragmentation , and the resulting overall total . after adding the contributions from all three sources phythia ( cteq3l ) agrees fairly well with the data . isajet ( cteq3l ) is a bit below the data because of a smaller shower / fragmentation contribution . herwig ( cteq3l ) is also slightly below the data due to a smaller flavor excitation component . one does not expect precise agreement from these leading - log estimates . however , the qualitative agreement indicates that probably nothing unusual is happening in @xmath0-quark production at the tevatron . [ prd_fig8 ] and fig . [ prd_fig9 ] show the differential inclusive @xmath0-quark @xmath20 cross section and the integrated total cross section , respectively , from pythia ( cteq4l ) , with @xmath23@xmath24 . pythia regulates the divergences in the perturbative @xmath3-to-@xmath3 parton scattering differential cross sections and these plots show a smooth behavior down to @xmath20(@xmath0-quark ) equal to zero . -quark total cross section ( @xmath26 , @xmath7 ) for proton - antiproton collisions at @xmath8 . the contributions from flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) are shown together with the resulting sum ( overall height of box ) . ] [ prd_fig10 ] shows the predictions of herwig ( do1.1 ) , herwig ( cteq3l ) , isajet ( cteq3l ) , pythia ( cteq3l ) , pythia ( cteq4l ) , and pythia ( grv94l ) for the integrated @xmath0-quark total cross section ( @xmath20@xmath27 , @xmath7 ) at @xmath8 . the contributions from flavor creation , flavor excitation , and shower / fragmentation are shown together with the overall resulting sum ( overall height of box ) . the do1.1 structure functions include only four flavors and the probability of finding a @xmath0 ( or @xmath1 ) quark within the proton is set identically equal to zero . thus , running with the do1.1 structure functions results in no flavor excitation . the differences among the monte - carlo models for flavor excitation are due to the different ways the models handle the @xmath0-quark mass in this subprocess . the monte - carlo model predictions for the shower / fragmentation contribution differ considerably . this is not surprising since isajet uses independent fragmentation , while herwig and pythia do not ; and herwig and pythia modify the leading - log picture of parton showers to include color coherence effects " , while isajet does not . the flavor creation contribution to inclusive @xmath0-quark production ( @xmath20@xmath27 , @xmath7 ) is predicted by herwig ( cteq3l ) and pythia ( cteq3l ) to be about @xmath28 of the overall @xmath0-quark production rate from all three sources at @xmath8 , while for isajet ( cteq3l ) it is about @xmath29 of the total . ) for the probability of finding a @xmath1-quark with @xmath30 and @xmath7 in an event with a @xmath0-quark with @xmath30 and @xmath7 for proton - antiproton collisions at @xmath8 . the contribution from the toward " ( @xmath31@xmath32 ) and the away " ( @xmath31@xmath33 ) region of the @xmath0-quark are shown for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , and shower / fragmentation ( fig . [ prd_fig4 ] ) , where @xmath31 is the azimuthal angle between the @xmath0 and @xmath1-quark . ] ) for the probability of finding a @xmath1-quark with @xmath30 and @xmath7 in an event with a @xmath0-quark with @xmath30 and @xmath7 for proton - antiproton collisions at @xmath8 . the contribution from the toward " ( @xmath31@xmath32 ) and the away " ( @xmath31@xmath33 ) region of the @xmath0-quark are shown for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , and shower / fragmentation ( fig . [ prd_fig4 ] ) , where @xmath31 is the azimuthal angle between the @xmath0 and @xmath1-quark . ] ) for the probability of finding a @xmath1-quark with @xmath30 and @xmath7 in an event with a @xmath0-quark with @xmath30 and @xmath7 for proton - antiproton collisions at @xmath8 . the contribution from the toward " ( @xmath31@xmath32 ) and the away " ( @xmath31@xmath33 ) region of the @xmath0-quark are shown for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , and shower / fragmentation ( fig . [ prd_fig4 ] ) , where @xmath31 is the azimuthal angle between the @xmath0 and @xmath1-quark . ] clearly the three sources of @xmath0-quarks ; flavor creation , flavor excitation , and shower / fragmentation have quite different topological structures and the correlations between the @xmath0 and @xmath1 quarks are quite different . figs . [ prd_fig11 ] - [ prd_fig13 ] show the qcd monte - carlo model predictions for some simple correlations . here one requires a @xmath0-quark to be in the region @xmath20@xmath21 and @xmath7 and then asks for the probability of finding a @xmath1-quark in the same region , @xmath20@xmath21 and @xmath7 . furthermore , one breaks this probability into two terms , toward " and away " . the toward " region corresponds to @xmath31@xmath34 and the away " region has @xmath31@xmath35 , where @xmath36 is the azimuthal angle between the @xmath0 and @xmath1 quark . the models predict that for flavor creation the probability that both the @xmath0 and @xmath1 quark lie in the region @xmath20@xmath21 and @xmath7 is about @xmath22 , with the @xmath1-quark almost always on the away - side " of the @xmath0-quark . the qcd monte - carlo models differ considerably on the correlations for the shower / fragmentation contribution . for this contribution all three models predict that both the @xmath0 and @xmath1 quark lie in the region @xmath20@xmath21 and @xmath7 around @xmath37-@xmath22 of the time , which is comparable to the flavor creation contribution . however , isajet predicts that for the shower / fragmentation contribution that the @xmath1-quark is almost always on the toward - side " of the @xmath0-quark , while herwig and pythia predict about equal amounts of toward " and away " for this contribution . as one would expect , all the models predict that it is not very likely to find both the @xmath0 and the @xmath1-quark in the region @xmath20@xmath21 and @xmath7 for the flavor excitation contribution . for flavor excitation either the @xmath0-quark or the @xmath1-quark is part of the underlying " event ( see fig . [ prd_fig3 ] ) . the models differ greatly on the correlations for the flavor excitation contribution . for this contribution pythia predicts that both the @xmath0 and @xmath1-quark lie in the region @xmath20@xmath21 and @xmath7 around @xmath38 of the time , while isajet predicts @xmath39 and herwig gives @xmath40 for this probability . all three models predict that for flavor excitation when both the @xmath0 and @xmath1-quark are found in the region @xmath20@xmath21 and @xmath7 that the @xmath1-quark is about equally likely to be on toward " or away " side of the @xmath0-quark . @xmath41 ) for the azimuthal angle , @xmath36 , between a @xmath0-quark with @xmath42 and @xmath43 and a @xmath1-quark with @xmath44 and @xmath45 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath46 ( @xmath47b/@xmath48 ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] , and the resulting total . the toward " ( @xmath31@xmath32 ) and the away " ( @xmath31@xmath33 ) region of the @xmath0-quark are labeled . ( _ note the logarithmetic scale . _ ) ] @xmath41 ) for the azimuthal angle , @xmath36 , between a @xmath0-quark with @xmath42 and @xmath43 and a @xmath1-quark with @xmath49 and @xmath45 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath46 ( @xmath47b/@xmath48 ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . the toward " ( @xmath31@xmath32 ) and the away " ( @xmath31@xmath33 ) region of the @xmath0-quark are labeled . ( _ note the logarithmetic scale . _ ) ] fig . [ prd_fig14 ] shows the predictions of pythia for the azimuthal angle , @xmath36 , between a @xmath0-quark with @xmath50 and @xmath43 and @xmath1 quark with @xmath51 and @xmath45 . [ prd_fig15 ] is the same as fig . [ prd_fig14 ] except both the @xmath0 and @xmath1 quark are required to have @xmath30 . notice that the relative amounts of flavor excitation and shower / fragmentation decrease compared to flavor creation when one demands that both the @xmath0 and @xmath1 quarks to have @xmath30 . integrating the curves in fig . [ prd_fig15 ] over the toward " and away " regions correspond to the simple probabilities discussed in section iv(a ) . clearly , the toward " region is very sensitive to the presence of the flavor excitation and shower / fragmentation terms . @xmath41 ) for the distance , @xmath52 , in @xmath53-@xmath54 space between a @xmath0-quark with @xmath42 and @xmath43 and a @xmath1-quark with @xmath49 proton - antiproton collisions at @xmath8 . the curves correspond to @xmath55 ( @xmath47b ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ( _ note the logarithmetic scale . _ ) ] fig . [ prd_fig16 ] shows the predictions of pythia for the distance , @xmath52 , in @xmath53-@xmath54 space between a @xmath0-quark with @xmath50 and @xmath43 and @xmath1-quark with @xmath56 , where @xmath57 and @xmath53 is the pseudo - rapidity . one can see that the region @xmath58 isolates the shower / fragmentation contribution . [ prd_fig17 ] shows the predictions of pythia for the rapidity , @xmath59 of a @xmath0-quark with @xmath50 for events with a @xmath1-quark with @xmath56 and @xmath60 . here the flavor excitation contribution behaves much differently than the other two terms . however , the resulting sum of the three contributions looks similar to the flavor creation term . @xmath41 ) for the rapidity , @xmath59 , of a @xmath0-quark with @xmath42 for events with a @xmath1-quark with @xmath49 and @xmath60 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath61 ( @xmath47b ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] @xmath41 ) for the @xmath20 asymmetry , @xmath62 , for events with a @xmath0-quark with @xmath63 and @xmath64 and a @xmath1-quark with @xmath49 and @xmath65 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath66 ( @xmath47b ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ] fig . [ prd_fig18 ] shows the predictions of pythia for the @xmath20 asymmetry , @xmath62 , for events with a @xmath0-quark with @xmath67 and @xmath43 and a @xmath1-quark with @xmath56 and @xmath45 . if there were no initial - state radiation ( see fig . [ prd_fig1 ] ) and no intrinsic " transverse momentum , the flavor creation terms would produce zero asymmetry . also , as one would expect ( fig . [ prd_fig3 ] ) , the flavor excitation terms have , on the average , a large transverse momentum asymmetry . requiring @xmath20 asymmetries with a magnitude greater than around @xmath68 isolates the flavor excitation contribution . @xmath41 ) for the transverse momentum , @xmath69 , of a @xmath1-quark with @xmath65 for events with a @xmath0-quark with @xmath42 and @xmath70 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath71 ( @xmath47b / gev / c ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ( _ note the logarithmetic scale . _ ) ] @xmath41 ) for the integrated pair cross section for a @xmath1-quark with @xmath72 and @xmath65 and a @xmath0-quark with @xmath42 and @xmath70 in proton - antiproton collisions at @xmath8 . the curves correspond to @xmath73 ( @xmath47b ) for flavor creation ( fig . [ prd_fig1 ] ) , flavor excitation ( fig . [ prd_fig3 ] ) , shower / fragmentation ( fig . [ prd_fig4 ] ) , and the resulting total . ( _ note the logarithmetic scale . _ ) ] fig . [ prd_fig19 ] shows the predictions of pythia for the transverse momentum , @xmath69 , of a @xmath1-quark with @xmath45 for events with a b - quark with @xmath50 and @xmath43 . as the transverse momentum asymmetry in fig . [ prd_fig18 ] indicated , the flavor creation contribution peaks at @xmath74 ( _ i.e. _ , small @xmath20 asymmetry ) , while the flavor excitation and shower / fragmentation contributions peak at low @xmath20 ( _ i.e. _ , larger @xmath20 asymmetry ) . [ prd_fig20 ] shows the predictions of pythia for the integrated @xmath2 pair cross section for a @xmath1-quark with @xmath75 and @xmath45 and b - quark with @xmath50 and @xmath43 . notice how the three terms contribute in different proportions to the @xmath2 pair cross section in fig . [ prd_fig20 ] and the inclusive single @xmath0-quark cross section in fig . [ prd_fig9 ] . flavor excitation is a major contributor to the single inclusive cross section , while flavor creation dominates the pair cross section . the leading - log order qcd hard scattering monte - carlo models of herwig , isajet , and pythia have been used to study the sources of @xmath0-quarks at the tevatron . the reactions responsible of producing @xmath0-quarks are separated into flavor creation , flavor excitation , and shower / fragmentation . the production of a @xmath2 pair via gluon fusion or annihilation of light quarks is easy to generate and all three qcd monte - carlo models ( _ with the same structure functions _ ) predict roughly the same cross section and similar @xmath2 correlations from these terms . at the tevatron all three monte - carlo models predict that the flavor creation contribution to @xmath0-quark production is less than @xmath29 of the overall inclusive b - quark production rate ( @xmath20@xmath27 , @xmath7 ) from all three sources . the source of @xmath0-quarks resulting from the scattering of a @xmath0-quark or @xmath1-quark out of the initial - state into the final - state by a gluon or by a light quark or antiquark is difficult to generate and depends sensitively on the parton distribution functions . the qcd monte - carlo model predictions differ somewhat , however , it seems likely that at the tevatron the flavor excitation contribution to the inclusive @xmath0-quark cross section is comparable to or greater than the contribution from flavor creation . the @xmath2 correlations resulting from the flavor excitation term are much different than those arising from flavor creation . in particular , the distribution in the azimuthal angle , @xmath36 , between the @xmath0 and the @xmath1-quark peaks sharply in the backward " ( @xmath31@xmath35 ) region for flavor creation , while flavor excitation produces an approximately flat @xmath31 distribution . also , the flavor excitation terms have , on the average , a large transverse momentum asymmetry between the @xmath0 and @xmath1-quark , while the flavor creation terms produce a @xmath20 asymmetry near zero . the production of @xmath2 pairs within a parton - shower or during the fragmentation process of gluons and light quarks and antiquarks is an important source at the tevatron . the qcd monte - carlo models predictions differ considerably for this contribution . however , at the tevatron the shower / fragmentation contribution to the inclusive b - quark cross section might be comparable to the contribution from flavor creation . this source can be isolated by looking for @xmath2 pairs with @xmath58 , where @xmath52 is the distance in @xmath53-@xmath54 space between the @xmath0 and @xmath1-quark . 0.2 in the qcd monte - carlo model leading - log estimates of the flavor excitation and the shower / fragmentation contributions to @xmath0-quark production are uncertain and should not be taken too seriously . however , it seems likely that at the tevatron all three sources of @xmath0-quarks , flavor creation , flavor excitation , and shower / fragmentation are important @xcite . one does not expect precise agreement from leading - log estimates . on the other hand , the qualitative agreement shown indicates that probably nothing unusual is happening in @xmath0-quark production at the tevatron . furthermore , in run ii at the tevatron we should be able to isolate experimentally the individual contributions to @xmath0-quark production by studying @xmath2 correlations in detail .	the leading - log order qcd hard scattering monte - carlo models of herwig , isajet , and pythia are used to study the sources of @xmath0-quarks at the tevatron . the reactions responsible for producing @xmath0 and @xmath1 quarks are separated into three categories ; flavor creation , flavor excitation , and parton - shower / fragmentation . flavor creation corresponds to the production of a @xmath2 pair by gluon fusion or by annihilation of light quarks , while flavor excitation corresponds to a @xmath0 or @xmath1 quark being knocked out of the initial - state by a gluon or a light quark or antiquark . the third source occurs when a @xmath2 pair is produced within a parton - shower or during the fragmentation process of a gluon or a light quark or antiquark ( includes gluon splitting ) . the qcd monte - carlo models indicate that all three sources of @xmath0-quarks are important at the tevatron and when combined they qualitatively describe the inclusive cross - section data . correlations between the @xmath0 and @xmath1 quark are very different for the three sources and can be used to isolate the individual contributions .
in 1979 siemens and rasmussen formulated a model describing the hadron production in ne + na f reactions at the beam energy of 800 mev per nucleon @xcite . the physical picture behind the model was that the fast hydrodynamic expansion of the produced hadronic matter leads to a sudden decoupling of hadrons and freezing of their momentum distributions , which retain their thermal character ( although modified by the collective expansion effects ) until the observation point . in their own words , siemens and rasmussen described the collision process as follows : `` central collisions of heavy nuclei at kinetic energies of a few hundred mev per nucleon produce fireballs of hot , dense nuclear matter ; such fireballs explode , producing blast waves of nucleons and pions '' . in this way , with ref . @xcite , the concept of the blast waves of hadrons and the blast - wave model itself entered the field of relativistic heavy - ion collisions . although the model of siemens and rasmussen was motivated by an earlier hydrodynamic calculation by bondorf , garpman , and zimanyi @xcite , the results presented in ref . @xcite were not obtained by solving the hydrodynamic equations but followed from the specific assumptions on the freeze - out conditions . the most important ingredient of the model was the spherically symmetric expansion of the shells of matter with constant radial velocity . with an additional assumption about the times when such shells disintegrate into freely streaming hadrons ( this point will be discussed in a greater detail in sect . [ sect : rad ] ) siemens and rasmussen obtained the formula for the momentum distribution of the emitted hadrons @xcite @xmath0 . \label{sr1}\ ] ] in eq . ( [ sr1 ] ) @xmath1 is a normalization factor , @xmath2 denotes the hadron energy , @xmath3 is the temperature of the fireball ( the same for all fluid shells ) , and @xmath4 is the lorentz gamma factor with @xmath5 denoting the radial collective velocity ( radial flow ) . a dimensionless parameter @xmath6 is defined by the equation @xmath7 small values of @xmath5 ( and @xmath6 ) correspond to small expansion rate and , as expected , a simple boltzmann factor is obtained from eq . ( [ sr1 ] ) in the limit @xmath8 , @xmath9 the fits to the data based on the formula ( [ sr1 ] ) gave @xmath3 = 44 mev and @xmath5 = 0.373 . interestingly , the value of the radial flow @xmath5 turned out to be quite large suggesting the strong collective behavior . this was an unexpected feature summarized by the authors with the statement : `` monte carlo studies suggest that ne + na f system is too small for multiple collisions to be very important , thus , this evidence for a blast feature may be an indication that pion exchange is enhanced , and the effective nucleon mean free path shortened in dense nuclear matter '' . below we shall analyze the formal steps leading to eq . ( [ sr1 ] ) . our starting point is the expression defining the momentum distribution of particles as the integral of the phase - space distribution function @xmath10 over the freeze - out hypersurface @xmath11 , i.e. , the renowned cooper - frye formula @xcite , @xmath12 the three - dimensional element of the freeze - out hypersurface in eq . ( [ cf1 ] ) may be obtained from the formula @xmath13 where @xmath14 is the levi - civita tensor and @xmath15 are the three independent coordinates introduced to parameterize the hypersurface . we note that for systems in local thermodynamic equilibrium we have @xmath16 where the function @xmath17 is the equilibrium distribution function @xmath18^{-1}. \label{eq}\ ] ] here the case @xmath19 corresponds to the fermi - dirac ( bose - einstein ) statistics , and the limit @xmath20 yields the classical ( boltzmann ) statistics . for a static fireball one finds @xmath21 and eq . ( [ cf2 ] ) is reduced to the formula @xmath22 where @xmath23 is the volume of the system . ( [ cf3 ] ) agrees with eq . ( [ sr2 ] ) in the classical limit if the normalization constant @xmath1 is taken as @xmath24 + for spherically symmetric freeze - outs it is convenient to introduce the following parameterization of the space - time points on the freeze - out hypersurface @xcite @xmath25 the freeze - out hypersurface is completely defined if a curve , i.e. , the mapping @xmath26 in the @xmath27 space is given . this curve defines the ( freeze - out ) times when the hadrons in the shells of radius @xmath28 stop to interact , see fig . [ fig : tr ] . the range of @xmath29 may be always restricted to the interval : @xmath30 . the three coordinates : @xmath31 , \theta \in [ 0,\pi]$ ] , and @xmath32 $ ] play the role of the variables @xmath15 appearing in eq . ( [ d3sigma ] ) . hence , the element of the spherically symmetric hypersurface has the form @xmath33 where the prime denotes the derivatives taken with respect to @xmath29 . besides the spherically symmetric hypersurface we introduce the spherically symmetric ( hydrodynamic ) flow @xmath34 where @xmath35 is the lorentz factor , @xmath36 . in a similar way the four - momentum of a hadron is parameterized as @xmath37 , \label{pmurad}\ ] ] and we find the two useful expressions : @xmath38 @xmath39 we note that the spherical symmetry allows us to restrict our considerations to the special case @xmath40 . in the case of the boltzmann statistics , with the help of eqs . ( [ cf1 ] ) , ( [ purad ] ) and ( [ sigmaprad ] ) , we obtain the following form of the momentum distribution @xmath41 r^2(\zeta ) d\zeta . \label{dnd3prad1}\ ] ] here @xmath42 and @xmath43 are functions of @xmath29 , and the parameter @xmath6 is defined by eq . ( [ a ] ) . the thermodynamic parameters @xmath3 and @xmath44 may also depend on @xmath29 . to proceed further we need to make certain assumptions about the @xmath29-dependence of these quantities . in particular , to obtain the model of siemens and rasmussen we assume that the thermodynamic parameters as well as the transverse flow velocity are constant @xmath45 moreover , we should assume that the freeze - out curve in the @xmath46 space satisfies the condition @xmath47 in this case we obtain the formula @xmath48 \int\limits_0 ^ 1 r^2(\zeta ) { dr \over d\zeta } d\zeta . \label{dnd3prad2}\ ] ] equation ( [ dnd3prad2 ] ) coincides with eq . ( [ sr1 ] ) if we use eq . ( [ z ] ) and make the following identification @xmath49 note that the quantity @xmath50 does not necessarily denote the maximum value of the radius of the system , see the dotted line on the left - hand - side of fig . [ fig : tr ] . an interesting and perhaps unexpected feature of the model proposed by siemens and rasmussen is the relation between the times and positions of the freeze - out points , see eq . ( [ dtvdr ] ) illustrated on the right - hand - side part of fig . [ fig : tr ] . ( [ dtvdr ] ) indicates that the fluid elements which are further away from the center freeze - out later . moreover , taking into account eq . ( [ dtvdr ] ) in the formula ( [ d3sigmarad ] ) we find that the four - vector describing the hypersurface is parallel to the four - vector describing the flow , compare eqs . ( [ d3sigmarad ] ) and ( [ umurad1 ] ) giving @xmath51 in this case . as we shall see the same features are assumed in the single - freeze - out model @xcite . it is worth to emphasize that in the hydrodynamic approach the @xmath27 freeze - out curves contain the space - like and time - like parts . the treatment of the space - like parts leads to conceptual problems since particles emitted from such regions of the hypersurface enter again the system and the hydrodynamic description of such regions ( combined with the use of the cooper - frye formula ) is inadeqate . recently much work has been done to develop a consistent description of the freeze - out process from the space - like parts @xcite . however , very often only a quantitative argument is presented @xcite that the contributions from the space - like parts are small and may be neglected compared to the contributions from the time - like regions . the choice of siemens and rasmussen seems to have anticipated such arguments . the model presented above is appropriate for the low - energy scattering processes where the two nuclei completely merge at the initial stage of the collision and further expansion of the system is , to large extent , isotropic . at higher energies such a picture is not valid anymore and , following the famous paper by bjorken @xcite , the boost - invariant and cylindrically symmetric models have been introduced to describe the collisons . moreover , the assumption about the cylindrical symmetry is valid only for the most central data . ] . the boost - invariance ( symmetry with respect to the lorentz transformations ) may be incorporated in the hydrodynamic equations , kinetic equations , and also in the modeling of the freeze - out process . in the latter case , the appropriate formalism was developed by schnedermann , sollfrank , and heinz @xcite . the ansatz for the boost - invariant , cylindrically symmetric freeze - out hypersurface has the form @xmath52 here , the parameter @xmath53 is the space - time rapidity . at @xmath54 the longitudinal coordinate @xmath55 is also zero and the variable @xmath56 coincides with the time coordinate @xmath43 . similarly to the spherical expansion discussed in sect . [ sect : rad ] , the boost - invariant freeze - out hypersurface is completely defined if the freeze - out curve @xmath57 is given . this curve defines the freeze - out times of the cylindrical shells with the radius @xmath58 . because of the boost - invariance it is enough to define this curve at @xmath59 , since for finite values of @xmath55 the freeze - out points may be obtained by the lorentz transformation . the volume element of the freeze - out hypersurface is obtained from eq . ( [ d3sigma ] ) , @xmath60 similarly to eq . ( [ xmubinv ] ) the boost - invariant four - velocity field has the structure @xmath61 we note that the longitudinal flow is simply @xmath62 ( as in the one - dimensional bjorken model ) , whereas the transverse flow is @xmath63 . with the standard parameterization of the particle four - momentum in terms of rapidity @xmath64 and the transverse mass @xmath65 , @xmath66 we find @xmath67 and @xmath68 \rho(\zeta){\tilde \tau}(\zeta ) d\zeta d\alpha_\parallel d\phi . \label{sigmapbinv}\ ] ] for the boltzmann statistics , with @xmath69 , the cooper - frye formalism gives the following momentum distribution @xmath70 \nonumber \\ & & \times \exp\left[-\beta m_\perp \hbox{cosh}(\alpha_\perp ) \hbox{cosh}(\alpha_\parallel - y)+ \beta p_\perp \hbox{sinh}(\alpha_\perp ) \cos(\phi-\varphi ) \right ] . \label{bmdn1}\end{aligned}\ ] ] the form of eq . ( [ bmdn1 ] ) shows explicitly that the distribution @xmath71 is independent of @xmath64 and @xmath72 , in accordance with our assumptions of the boost - invariance and cylindrical symmetry . the integrals over @xmath53 and @xmath73 in eq . ( [ bmdn1 ] ) are analytic and lead to the bessel functions @xmath74 and @xmath75 , @xmath76 i_0 \left[\beta p_\perp \hbox{sinh}(\alpha_\perp)\right ] \int\limits_0 ^ 1 d\zeta \,\ , \rho(\zeta ) { \tilde \tau}(\zeta ) { d\rho \over d\zeta } \nonumber \\ & & - { e^{\beta \mu } \over 2 \pi^2 } p_\perp k_0\left[\beta m_\perp \hbox{cosh}(\alpha_\perp ) \right ] i_1 \left[\beta p_\perp \hbox{sinh}(\alpha_\perp)\right ] \int\limits_0 ^ 1 d\zeta \,\ , \rho(\zeta ) { \tilde \tau}(\zeta ) { d{\tilde \tau } \over d\zeta}. \nonumber \\ \label{kandi}\end{aligned}\ ] ] in the spirit of the blast - wave model of siemens and rasmussen we have assumed here that the radial velocity is constant , @xmath77 , otherwise the bessel functions should be kept under the integral over @xmath29 . in order to achieve the simplest possible form of the model , the common practice is to neglect the second line of eq . ( [ kandi ] ) . this procedure means that one assumes implicitly the freeze - out condition @xmath78 . in this case the boost - invariant blast - wave model is reduced to the formula @xmath79 i_0 \left[\beta p_\perp \hbox{sinh}(\alpha_\perp)\right ] . \label{k1i0a}\ ] ] where the constant has absorbed the factor @xmath80 . ( [ k1i0a ] ) forms the basis of numerous phenomenological analyses of the transverse - momentum spectra measured at the sps and rhic @xcite energies . the main drawback of the formalism outlined above is that it neglects the effect of the decays of hadronic resonances . such an approach may be justified at lower energies but should be improved at the relativistic energies where most of the light particles are produced in the decays of heavier resonance states . the expressions giving the rapidity and transverse momentum spectra of particles originating from two- and three - body decays of the resonances with a specified momentum distribution were worked out by sollfrank , koch , and heinz @xcite . their formulae may also be used to account for the feeding from the resonances in the blast - wave model , as already proposed in ref . @xcite . in other words , we wish to stress that the choice of the freeze - out hypersurface and of the flow profile are elements _ completely independent _ of the treatment of the resonances . both are important with the latter being the basic ingredient in the calculation of particle abundances and the key to success of thermal models . at the sps and rhic energies it is important to include not only the decays of the most common resonances such as @xmath81 or @xmath82 , but also of much heavier states . although their contributions are supressed by the boltzmann factor , their number increases strongly with the mass @xcite , hence their role can be easily underestimated . the effects of sequential decays of heavy resonances were first realized in statistical analyses of the ratios of hadronic abundances / multiplicities ( for recent results see @xcite ) which showed that the statistical models give a very good description of the data , provided most of the hadrons appearing in the particle data tables are included in the calculations . in order to discuss the role of the sequential decays of the resonances it is convenient to start with a general formalism giving the lorentz - invariant phase - space density of the measured particles @xcite @xmath83 . \nonumber \\ \label{ornik1}\end{aligned}\ ] ] here the indices @xmath84 label hadrons in one chain of the sequential decays . the first resonance is produced on the freeze - out hypersurface and has the label @xmath85 . the final hadron has the label @xmath86 , for more details see @xcite . the function @xmath87 is the probability distribution for a resonance with momentum @xmath88 to produce a particle with momentum @xmath89 in a two - body decay @xmath90 the function @xmath87 satisfies the normalization condition @xmath91 where @xmath92 is the branching ratio for a given decay channel and @xmath93 is the momentum ( energy ) of the emitted particle in the resonance s rest frame ( a generalization to three - body decays is straightforward and explained in refs . @xcite ) . integration of eq . ( [ ornik1 ] ) over all space - time positions gives the formula for the momentum distribution @xmath94 . \nonumber \\ \label{ornik2}\end{aligned}\ ] ] equation ( [ ornik2 ] ) serves as the starting point to prove that for constant values of the thermodynamic parameters on the freeze - out hypersurface the ratios in the full phase - space ( @xmath95 ) are the same as in the local fluid elements . in this way , a connection between the measured ratios and the local thermodynamic parameters is obtained @xcite . one may also check that for the boost - invariant systems it is enough to consider the ratios at any value of the rapidity to infer the values of the thermodynamic parameters @xcite . the experimental rhic data show , however , that the rapidity distributions are of gaussian shape and the thermodynamic parameters vary with rapidity ( the measured @xmath96 ratio depends on @xmath64 ) , hence , the system created at rhic is , strictly speaking , not boost - invariant . in this situation the relation between the measured ratios and thermodynamic parameters is not obvious . fortunately , the rhic data show also a rather flat rapidity distribution and constant ratios in the rapidity range @xmath97 @xcite . in this region ( the central part of the broad gaussian ) the system to a good approximation may be treated as boost - invariant and the standard analysis of the ratios may be performed to obtain the thermodynamic parameters at @xmath98 . the analysis of the ratios of hadron multiplicities measured at rhic gives a typical temperature of 170 mev . on the other hand , the analysis of the spectra based on eq . ( [ k1i0a ] ) gives a lower temperature of about 100 - 140 mev . such a situation was observed already at the sps energies , which motivated the introduction of the concept of two different freeze - outs . certainly , if the spectra contain important contributions from high lying states , the value of @xmath3 obtained from the blast - wave formula fitted to the spectra can not be interpreted as the temperature of the system in the precise thermodynamic sense . first , the contributions from the resonances ( feeding mostly the low - momentum region ) should be subtracted from the spectra of light hadrons , giving the insight to the properties of the primordial particles . using other words , we may argue that the calculation of the ratios should include the same number of the resonances as the corresponding calculation of the spectra . an example of such a calculation is the single - freeze - out model formulated in refs . @xcite . in this model the decays of the resonances as well as the transverse flow change the spectra of the primordial particles in such a way that it is possible to describe well the spectra and the ratios with a single value of the temperature . the basic effect here is that the hadronic decays lead to effective cooling of the spectra . similarly to the original blast - wave models discussed above , the single freeze - out model assumes a certain form of the freeze - out hypersurface in the minkowski space . in this case it is defined by the constant value of the proper time @xmath99 the transverse size of the system is defined by the parameter @xmath100 , @xmath101 and the velocity field at freeze - out is taken in the hubble - like form ) - ( [ hubflow ] ) with hydrodynamic calculation see ref . ] @xmath102 the natural parameterization of the freeze - out hypersurface has the form @xmath103 which may be considered as the special case of the formula ( [ xmubinv ] ) . ( [ d3sigma ] ) leads to the following expression defining the volume element @xmath104 a very important feature of the choice ( [ tau ] ) - ( [ hubflow ] ) is that the volume element is proportional to the four - velocity field . this feature holds also in the model of siemens and rasmussen . in this case the treatment of the resonance is very much facilitated . in particular , eq . ( [ ornik2 ] ) may be rewriten in the form @xmath105 \nonumber \\ & = & \int d\sigma \left ( x_{n}\right ) p_{1}\,\cdot u\left ( x_{n}\right ) \,f_{1}\left [ p_{1}\cdot u\left ( x_{n}\right ) \right ] , \label{ornik3}\end{aligned}\ ] ] where we have introduced the notation @xmath106 \label{trsim } \nonumber \\ & & = \int \frac{d^{3}p_{i}}{e_{p_{i}}}b\left ( p_{i},p_{i-1}\right ) p_{i}\,\cdot u\left ( x_{n}\right ) \,f_{i}\left [ p_{i}\cdot u\left ( x_{n}\right ) \right ] . \label{acta1}\end{aligned}\ ] ] in the local rest frame , the iterative procedure defined by eq . ( [ acta1 ] ) becomes a simple one - dimensional integral transform @xmath107 where @xmath108 . ( [ acta1 ] ) and ( [ acta2 ] ) allow us to deal with a very large number of decays in the very efficient way , very similar to that used in the calculation of the hadron abundances . the model described above may be generalized to the non boost - invariant version in the minimal way by the modification of the system boundaries . introducing a dependence of the transverse size on the longitudinal coordinate @xmath55 ( or @xmath53 ) , we break explicitly the assumption of the boost - invariance . at the same time , however , the local properties of the hypersurface and flow remain unchanged allowing us to treat the resonances in the same simple way as described in the previous section . since the measured rapidity distributions are approximately gaussian , it is natural to start with the gaussian ansatz for the dependence of the transverse size on the parameter @xmath53 and restrict the region of the integration over @xmath109 to the interval @xmath110.\ ] ] the original boost - invariant version is recovered in the limit @xmath111 . using the values of the thermodynamic parameters obtained from the boost - invariant version of the model applied to the hadronic ratios measured at midrapidity ( @xmath3 = 165.6 mev and @xmath112 mev ) , we are left with three extra parameters ( @xmath113 , @xmath114 , and @xmath82 ) , which should be fitted to the @xmath115-spectra collected at different values of the rapidity . the result of such a fit to the available brahms data on @xmath116 , @xmath117 , @xmath118 , and @xmath119 production are shown in fig . [ fig : brpt ] . the optimal values of the parameters found in the fit are : @xmath120 fm , @xmath121 , and @xmath122 . one can see that the model reproduces the data very well in a wide range of the transverse - momentum and rapidity . in fig . [ fig : bry ] we show the model rapidity distributions compared to the data . small discrepancies ( of about 10% ) between the model and the data may be seen for the pions at @xmath98 . note that the comparison of the rapidity distributions in fig . [ fig : bry ] is done with a linear scale ; definitely , small discrepancies may be expected for a such simplified description of the freeze - out . it should be emphasized that the non - boost invariant version of the model presented above is not capable of describing correctly the @xmath96 ratio . in the present framework this requires an introduction of the rapidity dependence of the baryon chemical potential . we have discussed several parameterizations of the freeze - out conditions in the relativistic heavy - ion collisions . we have argued that the single freeze - out model used to describe the rhic data is a natural development of the blast - wave models worked out , among others , by siemens , rasmussen , heinz , schnedermann , and sollfrank . the main advantage of the single - freeze - out model is that it includes all well established resonance decays , allowing us to treat the chemical and thermal freeze - out as essentially one phenomenon . in this respect , the single - freeze - out model is very similar to the original blast - wave model . further similarities concern the shape of the freeze - out hypersurface ( only the time - like parts are considered ) and the strict use of the cooper - frye formula . due to the limited space , we have not discussed here the variety of models where , instead of the cooper - fry formula , the so - called emission functions are introduced and modeled . an example of such an a approach is the buda - lund model @xcite .	popular parameterizations of the freeze - out conditions in relativistic heavy - ion collisions are discussed . similarities and differences between the blast - wave model and the single - freeze - out model , both used recently to interpret the rhic data , are outlined . a non - boost - invariant extension of the single - freeze - out model is proposed and applied to describe the recent brahms data .
anderson localization is the absence of diffusive wave transport in highly disordered scattering media @xcite . it was first introduced in a pioneering theoretical study in 1958 by philip warren anderson @xcite , who investigated the behavior of spin diffusion and electronic conduction in random lattices . it took more than ten years for the scientific community to recognize the importance of anderson s work . however , it has remained at the forefront of physics research since 1968 . there are still many uncertainties and unanswered questions in the linear and nonlinear behavior of disordered systems in various dimensions . the model that anderson studied involved an electron on a potential lattice with a random spread in the energies of the sites caused by a source of disorder . the electron was allowed to hop between sites via nearest neighbor potential coupling terms . anderson showed that the wavefunction of electron localizes to only few sites at all times , provided that the amount of randomness is sufficiently large . it did not take long for anderson and others to realize that the novel localization phenomenon was due to the wave nature of the quantum mechanical electrons scattering in a disordered potential , and that similar behavior should also be observed in other coherent wave systems , including classical ones @xcite . the disorder - induced localization in electronic systems was shown to be inhibited by thermal fluctuations and nonlinear effects ; therefore , it was encouraging to find other avenues in which the disorder - induced anderson localization could be observed . subsequently , localization was studied in various classical wave systems including acoustics , elastics , electromagnetics , optics @xcite , and various quantum optical systems , such as atomic lattices @xcite and propagating photons @xcite . optical systems have played a unique role in the fundamental understanding and experimental observation of anderson localization . optical phenomena are easy to `` visualize , '' and there are many advanced tools and techniques in optics that can be used to study the physics of localization . optical studies of anderson localization can often be done with tools that are widely accessible and can be performed in a single laboratory . in addition , anderson localization already has device - level applications in the optics @xcite , and optics can `` illuminate '' the path to localization - based devices in other disordered classical and quantum wave systems . it has been shown that coherent waves in one - dimensional ( 1d ) and and two - dimensional ( 2d ) unbounded disordered systems are always localized @xcite . for bounded 1d and 2d systems , if the sample size is considerably larger than the localization radius , the boundary effects are minimal and can often be ignored @xcite . however , in three - dimensional ( 3d ) coherent wave systems , the scattering strength needs to be larger than a threshold value for the localization to happen @xcite . the scattering strength is characterized by the wave scattering transport length @xmath0 ( shorter @xmath0 means stronger scattering ) , and the ioffe - regel condition @xcite states that in order to observe anderson localization , the disorder must be strong enough that the wave scattering transport length becomes on the order of the wavelength . the ioffe - regel condition is often cast in the form of @xmath1 , where @xmath2 is the effective wavevector in the medium . it is notoriously difficult to satisfy in 3d disordered - media . for example , for the optical field to localize in 3d , very large refractive index contrasts are required that are not generally available in low - loss optical materials @xcite . _ the fact that anderson localization is hard to achieve in 3d optical systems may be a blessing in disguise ; otherwise , no sunlight would reach the earth on highly cloudy days _ @xcite . in order to observe anderson localization of light , strongly scattering materials at optical and near infrared frequencies such as tio@xmath3 , gaas , gap , si , and ge nanoparticles can be used . careful measurements are required because bulk absorption can easily lead to experimental signatures similar to anderson localization @xcite . unlike 3d lightwave systems , in which observation of localization is prohibitively difficult , observation of anderson localization in quasi-2d and -1d optical systems ( transverse anderson localization ) is readily possible , as was first shown by abdullaev _ et al_. @xcite and de raedt _ et al_. @xcite . there have since been many reports on the observation of transverse anderson localization of light in 1d and 2d , which is the main focus of this tutorial review , as well . transverse anderson localization is attractive because of its relative simplicity , ease of experimentation , and the rewarding physical insights it brings about on many fronts . moreover , the longitudinal coordinate along the direction of propagation plays the role of time in a 2d disordered system ; therefore , controlled temporal variations can also be studied in these systems . there are many excellent reviews that cover various aspects of wave propagation in disordered systems as well as anderson localization @xcite . this tutorial review is neither intended to be comprehensive , nor is it intended to explore each area it covers in great depth . rather , the intent is to provide a pedestrian and intuitive approach to anderson localization , mainly focused on the transverse localization of light . the coverage of topics is inevitably slanted toward those of particular interest to the author . the pedagogical approach is intended to benefit both newcomers to this rewarding research field , as well as outsiders who are interested to learn about anderson localization . the author seeks forgiveness from those whose work is not mentioned here , as well as for any technical errors or omissions . a good way to build an intuition about the relationship between randomness and localization is to use random matrices @xcite . the following example shows that the extended eigenvectors of an ordered matrix become very localized when some randomness is added to the elements of a matrix . consider a symmetric tridiagonal @xmath4 matrix @xmath5 defined as @xmath6 for all possible values of @xmath7 . for definiteness in this numerical example , we consider @xmath8 . matrix @xmath5 has @xmath9 real eigenvectors , and each eigenvector is an @xmath9-element vector . we identify the @xmath7th eigenvector as @xmath10 , and @xmath11 represents its @xmath12th element . defined in eq . [ eq : defmorederd ] are extended over the entire element - position domain . eigenvectors @xmath13 , and @xmath14 are plotted in ( a ) ; and @xmath15 is plotted in ( b ) as a function of the element position @xmath12 . , width=480 ] in figure [ fig : orderedmatrix1 ] , a few eigenvectors @xmath11 are plotted as a function of their element number @xmath12 . in figure [ fig : orderedmatrix1](a ) , @xmath13 , and @xmath14 are plotted and are all oscillatory functions of @xmath12 , where the @xmath7th eigenvector is identified by @xmath16 oscillation nodes . @xmath15 is also plotted in figure [ fig : orderedmatrix1](b ) . the relevant observation here is that all of these eigenvectors are extended over the entire element - position domain . we now would like to show that randomness can localize the eigenvectors of @xmath5 over the element - position domain . for this part , lets preserve the tridiagonal character of @xmath5 as presented in eq . [ eq : defmorederd ] , but add a small random number to each off - diagonal element . the new , slightly randomized @xmath5 is formally defined as @xmath17 , \label{eq : defmrandom1}\end{aligned}\ ] ] where @xmath18 is a number randomly selected from a real uniform distribution in the range @xmath19 $ ] . similar to the case of the ordered @xmath5 in figure [ fig : orderedmatrix1 ] , @xmath13 , and @xmath14 are plotted in figure [ fig : randommatrix1](a ) and @xmath15 is plotted in figure [ fig : randommatrix1](b ) . this time , all of these eigenvectors appear to be localized over the element - position domain . , except the matrix @xmath5 is * slightly * randomized according to eq . [ eq : defmrandom1 ] and the eigenvectors are localized.,width=480 ] not only does the randomness result in localized eigenvectors , but also the relative strength of randomness compared with the average off - diagonal values determines the size of the localization ( localization length ) . this can be verified by increasing the range of the random numbers @xmath18 to @xmath20 $ ] . the new , strongly randomized @xmath5 is formally defined as @xmath21 . \label{eq : defmrandom2}\end{aligned}\ ] ] again , the same eigenvectors are plotted in figures [ fig : randommatrix2](a ) and [ fig : randommatrix2](b ) . the eigenvectors are strongly localized over the element - position domain because of the strong randomness in the off - diagonal elements of @xmath5 . and figure [ fig : randommatrix1 ] , except the matrix @xmath5 is * strongly * randomized according to eq . [ eq : defmrandom2 ] and the eigenvectors are * strongly * localized.,width=480 ] the impact of disorder on the distribution of the width of the eigenvectors on the element - position domain can be easily visualized in a histogram . in figure [ fig : histogramrandommatrix ] , the distribution of the width of the eigenvectors is plotted for the weak disorder of eq . [ eq : defmrandom1 ] and the strong disorder of eq . [ eq : defmrandom2 ] . the width is calculated using the second moment method , which is the standard deviation around the mean position calculated using the absolute - value - squared of the eigenvectors . the width @xmath22 for the eigenvector @xmath10 is given by @xmath23 each probability distribution is presented in a histogram and is the result of averaging over 100 independent random simulations . from the distributions in figure [ fig : histogramrandommatrix ] , it is clear that stronger disorder results in stronger localization of the eigenvectors ; moreover , it is clear that such statements can only be made in a statistical sense . therefore , in the case of strong disorder , although the majority of the eigenvectors are more localized , a minority of the eigenvectors may actually be less localized than those of the weak disorder . , and the strong disorder related to eq . [ eq : defmrandom2 ] . ] the exercise presented here shows how off - diagonal disorder results in localization . similar localization behavior can be observed for diagonal disorder , where only the diagonal elements of @xmath5 are randomized , and also for mixed diagonal and off - diagonal disorder . * highlights : * * the eigenvectors of random matrices can be localized in the element position space . * some modes are very narrow and some are wide , and localization is only meaningful in a statistical sense . * by calculating the width of the eigenvectors of a large ensemble of random matrices , it is possible to calculate the probability distribution for the width of the eigenvectors . * a stronger level of randomness shifts the eigenvector - width probability distribution to smaller width values , hence a stronger localization . another interesting example that links randomness to localization is the problem of light transmission through a random stack of dielectrics . the normalized transmission through a stack of dielectrics is shown in figure [ fig:1d - stack ] . the light , which is incident from the left , is partially reflected from the stack , while the rest is transmitted through the stack . the dielectrics are assumed to be lossless . and @xmath24 is shown , ( b ) is similar , except the thickness of the layers is chosen randomly.,width=432 ] figure [ fig:1d - stack](a ) shows a periodic array of two different dielectric materials identified with refractive indexes @xmath25 and @xmath24 , more commonly referred to as a bragg grating . figure [ fig:1d - stack](b ) is similar , except the thickness of the layers is chosen randomly . in figure [ fig:1d - pc](a ) , the relative optical power transmission is plotted as a function of the normalized frequency . the thickness of each layer is @xmath26 ( identical for all layers ) , and @xmath27 is the wavevector , where @xmath28 is the optical wavelength in vacuum . the stack is made of 200 air - glass layers ( 400 layers total ) , where @xmath29 and @xmath30 is assumed . the usual bandgaps and bandpasses are observed in the transmission plot of this periodic bragg grating . the setup for figure [ fig:1d - pc](b ) is identical to that of figure [ fig:1d - pc](a ) , except the thickness of each layer is chosen from a uniform random distribution in the range @xmath31 $ ] . unlike the periodic bragg grating , transmission through the random stack is nearly zero except for very small values of the normalized frequency @xmath32 . therefore , the random stack behaves like a nearly perfect mirror beyond a certain frequency . for ( a ) a periodic layer thickness and ( b ) random layer thickness . @xmath26 is the periodicity in case ( a ) , while the thickness of each layer in case ( b ) is chosen from a uniform random distribution in the range @xmath31$].,width=460 ] the calculation of the optical transmission in figure [ fig:1d - pc ] is carried out using the transmission matrix @xmath33 defined in eq . [ eq : mmatrix ] related to figure [ fig : m - matrix ] : @xmath34 ( n_2-n_1)e^{-i\varphi } & ( n_2+n_1)e^{-i\varphi } \end{bmatrix } , \quad \varphi = n_1k_0d . \label{eq : mmatrix}\ ] ] and eq . [ eq : umu].,height=134 ] in figure [ fig:1d - pc ] , the @xmath33 matrix can be used to relate the right- and left - moving components of the optical field in dielectric @xmath24 to those in dielectric @xmath25 according to @xmath35 u^-_2 \end{bmatrix } = m. \begin{bmatrix } u^+_1 \\[0.3em ] u^-_1 \end{bmatrix}. \label{eq : umu}\ ] ] the total transmission and reflection can be calculated by cascading all the @xmath33 matrices for different random layers of the dielectric stack , as discussed in detail in ref . @xcite . , except with 10,000 layers.,width=460 ] it is instructive to repeat the exercise of figure [ fig:1d - pc ] with a larger number of layers . in figure [ fig:1d - pc-10000 ] , the number of layers is increased to 10,000 in each case , and the relative optical power transmission is plotted in db units . the logarithmic scaling on the vertical axis shows that the bandgap frequency ranges in the ordered stack of figure [ fig:1d - pc-10000](a ) , although not perfect due to the finite number of layers , attenuate the transmitted light by more than 20,000 db , making the stack a nearly perfect mirror over this frequency range . the random stack in figure [ fig:1d - pc-10000](b ) , except for low frequency values , is also a very good reflector . although the attenuation in transmission at 3,000 db or more is not as good as the bandgap region in the periodic case , it is very broadband and is not limited to bandgap range . the observations in figure [ fig:1d - pc-10000 ] bring about a somewhat philosophical issue . in practice , any stack of optical dielectrics has some inevitable randomness ; therefore , given a sufficiently large number of layers , it acts as a perfect mirror beyond a certain frequency . of course , at very low frequencies , the wavelength is so large that it hops over the stack and results in large transmission . therefore , in practice , any stack of optical dielectrics with a sufficiently large number of layers is a practically a broadband bandgap structure . the periodic structure analyzed in figure [ fig:1d - pc-10000](a ) can be viewed as the limiting case of the random structure , when the randomness goes to zero . when this limit is taken , the bandpasses emerge out of the broadband bandgap of the more general disordered stack . therefore , the philosophical point of view is that the magic of a periodic structure is not in its bandgap , because the bandgap comes naturally in any practical 1d stack ; rather , it is the emergence of the bandpasses that makes periodic structure so special . while this may seem like an inconsequential philosophical issue , it is necessary to alert students who learn about coherent waves in periodic systems in optics or condensed matter physics to ensure that they develop proper early intuition about the true implications of periodicity . _ the bandpasses are at least as glorious as the bandgaps_. the low - frequency behavior , as well as the periodic `` very weak '' resurrection of transmission in figure [ fig:1d - pc-10000](b ) , can be intuitively understood based on the work of berry _ et al_. @xcite . they have shown that the optical transmission through a random stack of @xmath9 bi - layered transparent plates , each contributing a random phase @xmath36 belonging to a uniform distribution @xmath37 $ ] , is given by @xmath38 using @xmath39 , @xmath29 , and @xmath30 , one obtains @xmath40 , which is in agreement with the numerical simulation in figure [ fig:1d - pc-10000](b ) . in the simulations presented in figures [ fig:1d - pc ] and [ fig:1d - pc-10000 ] , it was assumed that @xmath41 $ ] ; therefore , @xmath42 in order to obtain the near uniform distribution of @xmath37 $ ] , the upper bound in eq . [ eq : berry2 ] can be set to @xmath43 , where one obtains @xmath44 as the condition for the disorder - induced localization , exerting its full power and agreeing with the assumptions of ref . it should be noted that this value is very close to the value of @xmath32 in figure [ fig:1d - pc-10000](b ) , beyond which the attenuation is strong . therefore , the main reason that the attenuation is not strong at low @xmath32 is that the random phase @xmath36 does not cover the entire range of @xmath45 $ ] . for @xmath46 , although the range of the phase @xmath36 covers the entire @xmath45 $ ] , it partially folds over and makes the distribution of @xmath36 non - uniform over @xmath45 $ ] . for example , consider the case of @xmath47 $ ] . @xmath36 is a phase variable for which @xmath48\equiv [ 0,\pi]$ ] ; therefore , @xmath36 covers the entire @xmath45 $ ] range but with twice the probability in the @xmath49 $ ] range compared with @xmath50 $ ] range and is no longer uniformly distributed over @xmath45 $ ] . this is the main reason behind the partial resurrections of transmission observed in figure [ fig:1d - pc-10000](b ) for @xmath46 . a final and important point is that the exponential decay of the optical amplitude in the disordered stack is a manifestation of the `` coherent '' superposition of partial waves that reflect and transmit at each boundary . the fact that the naive ray theory , which is based on incoherent transmissions and reflections , fails to predict the anderson localization behavior observed in these examples= defies the intuition at first glance . the interested reader can consult ref . @xcite for a more detailed discussion on these points . a simple and elegant experiment using a laser pointer and a stack of overhead transparencies is also presented in ref . @xcite , which can be excellent demonstration in an undergraduate laboratory in optics . * highlights : * * optical transmission through a stack of dielectric layers with differing refractive indexes and random thicknesses drops , on the average , exponentially with the number of layers . * this exponential drop is observed for all wavelengths that are comparable or shorter than the mean thickness of the dielectric layers . the random stack behaves like an ultra - broadband mirror . * the optimum localization is obtained when the cumulative optical phase in each layer is from a uniformly distributed random number in the range @xmath45 $ ] . if the domain is smaller , or even larger such as @xmath51 $ ] , the localization is weaker . the example presented in section [ sec : normalstack ] can be readily generalized to the case of light incident at an angle on the dielectric stack , as illustrated in figure [ fig:1d - stack - angle ] . the calculation can be carried out using a simple generalization of the transmission matrix @xmath33 defined in eq . [ eq : mmatrix ] to the case of incidence at an angle ( see for example ref . @xcite ) . angle for the te polarization for ( a ) a periodic dielectric stack and ( b ) a random dielectric stack.,width=460 ] in figure [ fig:1d - pc-10000-te-45degrees ] , the relative optical power transmission is plotted as a function of the normalized frequency for oblique incidence at a 45@xmath52 angle ( as measured in the air layer ) for the te ( transverse electric ) polarization of light . figure [ fig:1d - pc-10000-te-45degrees ] should be compared with figure [ fig:1d - pc-10000 ] , which was for the case of normal incidence . in the case of the periodic dielectric stack in figure [ fig:1d - pc-10000-te-45degrees](a ) , the bandgaps are shifted compared with normal incidence in figure [ fig:1d - pc-10000](a ) and are wider and deeper . for the random dielectric stack in figure [ fig:1d - pc-10000-te-45degrees](b ) , the transmission is lower over the entire frequency band compared with figure [ fig:1d - pc-10000](b ) ; however , besides the lower value and frequency shifts of the features in transmission , no substantial qualitative difference exists between oblique incidence at a 45@xmath52 angle and normal incidence . angle for the tm polarization for ( a ) a periodic dielectric stack and ( b ) a random dielectric stack.,width=460 ] the case of a tm ( transverse magnetic ) polarization is explored in figure [ fig:1d - pc-10000-tm-45degrees ] , where the periodic - stack bandgaps in figure [ fig:1d - pc-10000-tm-45degrees](a ) are substantially narrower and more shallow compared with the case of a te polarization at 45@xmath52 angle in figure [ fig:1d - pc-10000-te-45degrees](a ) and normal incidence in figure [ fig:1d - pc-10000](a ) . similarly , the transmission of the tm polarization at 45@xmath52 angle in figure [ fig:1d - pc-10000-tm-45degrees](b ) through a random stack is stronger in the case of a tm polarization compared with the te polarization in figure [ fig:1d - pc-10000-te-45degrees](b ) and the case of normal incidence in figure [ fig:1d - pc-10000](b ) . angle for the te polarization for a ( a ) periodic dielectric stack and a ( b ) random dielectric stack.,width=460 ] for the larger incidence angle of 85@xmath52 , the difference between transmission through a periodic dielectric stack and a random dielectric stack is more pronounced ; and so is the difference between te and tm poalizations . the bandgaps are wider and deeper ( note the @xmath53 db label in the vertical scale ) for the te polarization incident on a periodic dielectric stack in figure [ fig:1d - pc-10000-te-85degrees](a ) compared with the case of 45@xmath52 and the case of normal incidence . the gaps are also shifted in normalized frequency and are more widely separated . similar behavior is observed for the case of the te polarization incident on a random dielectric stack in figure [ fig:1d - pc-10000-te-85degrees](b ) , where the full strength of the localization begins at a higher normalized frequency than what is shown in figure [ fig:1d - pc-10000-te-85degrees](b ) ; this behavior agrees with the general shift of the spectral features to the higher normalized frequency observed in the case of a periodic dielectric stack in figure [ fig:1d - pc-10000-te-85degrees](a ) . angle for the tm polarization for ( a ) a periodic dielectric stack and ( b ) a random dielectric stack.,width=460 ] for the tm polarization incident on a periodic dielectric stack in figure [ fig:1d - pc-10000-tm-85degrees](a ) , many narrow and closely separated gaps appear in the studied normalized frequency range . the gaps are deeper compared with the case of 45@xmath52 and the case of normal incidence . similarly , the spectral features shift to lower values in the case of the tm polarization incident on a random dielectric stack in figure [ fig:1d - pc-10000-tm-85degrees](b ) ; therefore , the full strength of the localization begins at a lower normalized frequency compared with te polarization and smaller incidence angles ( for both polarizations ) . in the above examples , it is observed that the tm polarization behaves differently from the te polarization . the root of this distinction is in the different boundary conditions for the electric field in the te and tm polarizations at the interface of each layer , and the difference is most apparent at brewster s angle , while the tm polarization shows perfect transmission . in figure [ fig:1d - pc-10000-tm - vs - angles ] , the relative optical power transmission is plotted as a function of the incidence angle of the tm polarized light on a random dielectric stack for ( a ) @xmath54 , and ( b ) @xmath55 . perfect transmission is observed at brewster s angle @xmath56 through 10,000 glass - air layers of random thickness . of course , like other simulations so far , intrinsic attenuation of glass is neglected in these simulations . the transmission is lower ( localization is stronger ) for @xmath55 in figure [ fig:1d - pc-10000-tm - vs - angles](b ) compared with the case of @xmath54 in figure [ fig:1d - pc-10000-tm - vs - angles](a ) , as expected from previous simulations and related arguments . , and ( b ) @xmath55 . perfect transmission is observed at brewster s angle @xmath56.,width=460 ] , and ( b ) @xmath55.,width=460 ] for comparison , in figure [ fig:1d - pc-10000-te - vs - angles ] , the relative optical power transmission is plotted as a function of the incidence angle of te polarized light on a random dielectric stack for ( a ) @xmath54 , and ( b ) @xmath55 . compared with figure [ fig:1d - pc-10000-tm - vs - angles ] , the absence of perfect transmission at brewster s angle is notable . it can also be concluded that below brewster s angle , transmission is weaker in the tm polarization compared with the te . this is expected , because at normal incidence te and tm polarizations have the same transmission ; increasing the angle results in a monotonic increase in transmission for the tm polarization all the way up to 100% at brewster s angle , while the transmission for the te polarization monotonically decreases . beyond brewster s angle , the te polarization continues its slow decline ; however , the tm polarization goes through a steep decline . depending on the value of @xmath32 , at large angles , the tm polarization will have a lower transmission compared with te ( compare figure [ fig:1d - pc-10000-tm - vs - angles](a ) with figure [ fig:1d - pc-10000-te - vs - angles](a ) ) , or vice versa ( compare figure [ fig:1d - pc-10000-tm - vs - angles](b ) with figure [ fig:1d - pc-10000-te - vs - angles](b ) ) . the interested reader is encouraged to consult a detailed account of the polarization dependence of an obliquely incident light on layered media in ref . @xcite . based on the detailed discussions of the oblique incidence e.g. , the results in figure [ fig:1d - pc-10000-tm - vs - angles ] and figure [ fig:1d - pc-10000-te - vs - angles ] , it is conceivable to make a 1d waveguide in the form shown in figure [ fig:1d - waveguide - angle ] . for a periodic dielectric stack , a waveguide can efficiently operate over a narrow range of wavelengths in the bandgaps corresponding to the specific incident wavelength @xcite . the operating bandwidth can be increased by considering a chirped grating as shown in ref . alternatively , a design based on a random dielectric stack is very broadband ; the trade off is that the leakage is slightly higher than that of a bandgap design for a periodic dielectric stack . therefore , if ultra - broadband operation is desired , a random stack is likely the solution . * highlights : * * considerable differences are observed between the transmission of the te and tm polarizations , when a random dielectric stack is illuminated at an angle . this hints at the possibility that the vectorial nature of the electromagnetic field may play an important role in setting the localization behavior in certain situations . * the strongest difference between the te and tm polarizations can be observed for oblique incidence at brewster s angle , at which the tm polarization is entirely transmitted through the random dielectric stack . * below brewster s angle , the te polarization usually has a lower transmission ; however , above brewster s angle , the situation depends on the ratio of the size of the optical wavelength to the average lattice size . * the concept of a broadband waveguide is introduced , where the random dielectric stacks are used as broadband reflecting walls for the waveguide . consider the one dimensional array of @xmath9 identical single - mode optical fibers sketched in figure [ fig:1d - array - fiber ] . the propagation constant of each fiber is @xmath57 and the direction of the propagation of the optical wave is assumed to be into the page . each fiber is weakly coupled to its nearest neighbor , and the strength of the coupling is determined by the separation between the fibers . the coupled mode equations for the propagation of the optical field through this optical fiber array can be expressed as @xcite @xmath58 @xmath59 is the amplitude of the optical field in the @xmath12th fiber , @xmath60 ( @xmath61 ) is the coupling strength of the @xmath12th fiber to its right neighbor @xmath62 ( left neighbor @xmath63 ) , and we assume symmetric coupling , so @xmath64 . of course , we need to assume @xmath65 and @xmath66 , because the 1st ( @xmath9th ) fiber does not have a neighbor to its left ( right ) . identical single - mode optical fibers , referred to as a coupled waveguide array . ] here , we consider the simple case , where the input light is only coupled to the middle fiber at @xmath67 . as the optical field propagates through the middle fiber , it couples to its neighboring fibers ; those will couple to their neighbors as well , and this cascading event results in a discrete diffraction pattern . this scenario is shown in figure [ fig : randomcoupledwaveguides1d](a ) , where the intensity of the propagating light is plotted as a function of the propagation distance and the waveguide number . the parameters used in this simulation are : @xmath68 , @xmath69 , @xmath70 , and @xmath71 . the input boundary condition is set to @xmath72 , as mentioned above . the situation is quite different if the coupling between the waveguides is randomized . for example , consider perturbing the array slightly , so that the average coupling remains @xmath73 , but individual couplings vary with a uniform distribution according to @xmath74 , where @xmath75 $ ] . in figure [ fig : randomcoupledwaveguides1d](b ) , we observe that the light in the middle waveguide can not spread as efficiently to its neighbors and the diffraction slows down as the beam propagates farther in @xmath76 . this behavior is certainly induced by the disorder and randomness introduced in the cross coupling of fibers . therefore , it is reasonable to expect that more randomness should result in slower diffraction . this is verified in figure [ fig : randomcoupledwaveguides1d](c ) by repeating the experiment for @xmath77 $ ] . let s take a closer look at the optical field pattern in figure [ fig : randomcoupledwaveguides1d](c ) . the propagating light initially follows a diffracting pattern , but after a certain propagation distance the diffraction halts and the total width of the beam remains more or less the same for the rest of the propagation . it appears that the fibers near the edges of the array always remain dark . this behavior is referred to as transverse anderson localization . in fact , similar behavior can be observed in figure [ fig : randomcoupledwaveguides1d](b ) after a sufficiently long propagation distance ; however , the transition to a stable width occurs at a longer propagation distance and the beam width is larger because of the smaller amount of disorder . the light propagation in a one - dimensional disordered coupled fiber array is transversely localized for any amount of disorder ; however , if the disorder is too small , the stable localized beam width may be larger than the transverse size of the structure and the localization effect can not be observed in practice . the coupled fiber array of figure [ fig:1d - array - fiber ] described by eq . [ eq : coupledmode1 ] can be analyzed alternatively in the language of normal modes . the advantage of the normal mode description is that the equations describing light propagation in a disordered coupled fiber array can be directly mapped to a random matrix discussed in section [ sec : randommatrix ] : the transverse anderson localization is nothing but the localization of the eigenvectors in the element - position domain index @xmath12 of @xmath59 . in order to see this , we define the vector @xmath78 as @xmath79 and rewrite eq . [ eq : coupledmode1 ] @xmath80 where all diagonal elements of the symmetric tridiagonal matrix @xmath81 are equal to @xmath57 and the off - diagonal elements are @xmath74 . the boundary condition is set to @xmath82 , because the input light is only coupled to the middle fiber . the real tridiagonal matrix @xmath81 is symmetric ; therefore , all of its eigenvectors are real and mutually orthogonal and its eigenvalues are real . we identify the @xmath7th eigenvalue as @xmath83 and the corresponding eigenvector as @xmath10 , and @xmath11 represents the @xmath12th element of the @xmath10 eigenvector . furthermore , without loss of generality , we assume that the eigenvectors are properly scaled such that they are orthonormal , i.e. , @xmath84 . using the eigenvectors , we can construct a real orthogonal matrix @xmath85 such that @xmath86 , where @xmath87 , and carry out an orthogonal transformation to rewrite eq . [ eq : coupledmode2 ] as @xmath88 the orthogonal transformation of the matrix @xmath81 results in a diagonal matrix whose elements are the eigenvalues of @xmath81 , i.e. , @xmath89 . using the above information , we can rewrite eq . [ eq : coupledmode3 ] as @xmath90 where we have defined @xmath91 the solutions to the @xmath9 independent first - order differential equations [ eq : coupledmode4 ] for the unidirectional propagation of light can be simply written as @xmath92,\quad j=1,\cdots , n . \label{eq : coupledsol1}\end{aligned}\ ] ] using the orthogonal transformation and a few lines of simple algebra , eq . [ eq : coupledsol1 ] can be expressed as @xmath93,\quad j=1,\cdots , n.\\ & = \sum^n_{k=1}\ q_{k , j}q_{k,101}\exp[-i{\bar \beta}_k z],\\ & = \sum^n_{k=1}\ { \mathbb v}^{(k)}_j{\mathbb v}^{(k)}_{101}\exp[-i{\bar \beta}_k z ] . \label{eq : coupledsol2}\end{aligned}\ ] ] equation [ eq : coupledsol2 ] brings the problem to its final form , using which we can now discuss the transverse anderson localization in terms of the random matrix @xmath81 in the language of section [ sec : randommatrix ] . recall that we showed in figure [ fig : randommatrix1 ] and the corresponding discussions that the eigenvectors of a symmetric tridiagonal matrix with random off - diagonal elements are localized over the element - position domain . therefore , only a few elements of each @xmath94 in eq . [ eq : coupledsol2 ] are non - zero . we also observed in figure [ fig : randommatrix1 ] that the localization of the eigenvectors occurs at different locations in the element - position domain ; therefore , only very few eigenvectors @xmath94 have non - zero elements in the @xmath95 position . as a result , the sum in eq . [ eq : coupledsol2 ] is practically limited to only a few eigenvectors ( modes ) @xmath94 that have non - zero @xmath95 elements and the other non - zero elements of these few eigenvectors are all concentrated around @xmath95 due to localization . therefore , @xmath96 will be equal to zero for all @xmath76 if @xmath12 is very different from @xmath97 . from the above discussions we can conclude that for a stronger disorder , the localization becomes stronger as seen in figure [ fig : randommatrix1 ] , and @xmath96 will remain zero unless @xmath12 is even closer to @xmath97 . the picture portrayed here is consistent with figure [ fig : randomcoupledwaveguides1d ] , especially in describing the difference between figure [ fig : randomcoupledwaveguides1d](b ) and figure [ fig : randomcoupledwaveguides1d](c ) . lets recap these observations in the more common modal language . the initial beam only couples to those guided modes that are localized in the vicinity of the center fiber . near the entrance , the optical excitation closely resembles the in - coupling beam , as a result of a coherent sum of the excited modes . each excited mode propagates with a different phase velocity determined by the propagation constant @xmath98 ; therefore the detailed balance between the excitation amplitudes of the guided modes that is responsible for the narrow excitation at the entrance @xmath72 is broken as the relative phases between the modes change when the beam propagates along the fiber array . as the beam propagates and the detailed balance is further broken , the beam expands ; however , it can never expand beyond the size of its constituent modes , which are only a few , each being very localized . therefore , the expansion is eventually halted , as observed in figure [ fig : randomcoupledwaveguides1d](b ) and figure [ fig : randomcoupledwaveguides1d](c ) . using the modal language we can explain a few other interesting phenomena . one question that is often asked is what happens if the initial excitation is wider than the typical width of an individual mode ? the answer is that a wider beam excites a few more modes ; therefore , the propagating light eventually localizes to a slightly wider beam . another interesting fact is that because in eq . [ eq : coupledsol2 ] only a few terms in the sum play a role , after a sufficiently long propagation distance @xmath99 , the beam refocuses back to its initial intensity distribution ( self - imaging ) @xcite . this revival distance can be calculated by noting that at @xmath99 , @xmath100 must be satisfied for all the relevant terms in the sum , where each @xmath101 is an integer . the intensity pattern at @xmath99 will be identical to that at @xmath67 . in practice , if the number of excited modes is sufficiently large , the revival distance can be much longer than the waveguide and the refocusing is not observed . it must be noted that a similar localization behavior is obtained if instead of randomizing the coupling coefficients between the fibers in the array @xcite , the propagation constants of individual fibers are slightly randomized @xcite . this was briefly discussed in section [ sec : randommatrix ] as diagonal localization . in figure [ fig:1d - waveguide - prop - const ] we show the distribution of the propagation constants of the eigenmodes of the coupled waveguide array of figure [ fig:1d - array - fiber ] ; i.e. , @xmath102 in eq . [ eq : coupledsol1 ] , for @xmath103 . the parameters used for the disorder relate to the case discussed in figure [ fig : randomcoupledwaveguides1d](b ) , where we used @xmath68 , @xmath69 , @xmath70 , and @xmath74 , with @xmath75 $ ] . the calculated modes are numbered from @xmath104 to @xmath105 , sorted in a descending order of the value of the propagation constant . as expected , the calculated propagation constants are in the vicinity of @xmath68 . the range of variation of @xmath102 is determined by the range of @xmath106 ( @xmath107 $ ] ) , while the exact values vary for each different random waveguide . to @xmath70 , sorted in a descending order of the value of the propagation constant.,width=240 ] in figure [ fig:1d - mode - profiles ] we plot the shape of a few modes of the waveguide . of course , the coupled waveguide array supports a total of @xmath70 modes . in figure [ fig:1d - mode - profiles ] each mode is normalized to unity for easier comparison . the horizontal axis is the waveguide number in figure [ fig:1d - array - fiber ] . figures [ fig:1d - mode - profiles](a ) , ( b ) , ( c ) , ( d ) , ( e ) , ( f ) correspond to mode numbers 1 , 25 , 99 , 101 , 170 , 201 , respectively . recall that the mode numbers here correspond to the mode numbers used in figure [ fig:1d - waveguide - prop - const ] . therefore , figure [ fig:1d - mode - profiles](a ) is the mode with the largest propagation constant in figure [ fig:1d - waveguide - prop - const ] , which is the top edge of the propagation constant band . are plotted . subfigures ( a ) , ( b ) , ( c ) , ( d ) , ( e ) , ( f ) , correspond to mode numbers 1 , 25 , 99 , 101 , 170 , 201 , respectively . the modes whose propagation constants belong to the region near the top edge of the band in figure [ fig:1d - waveguide - prop - const ] are highly localized with no or few oscillations as in ( a ) and ( b ) . the modes with propagation constants near the middle of the band spread over many waveguides and oscillate as in ( c ) and ( d ) . the modes with propagation constants near the bottom of the band are highly localized and oscillate so rapidly that the sign of the mode profile flips between adjacent waveguides as in ( c ) and ( d ) . ] the modes that correspond to the region in the vicinity of the top edge of the band are highly localized with very few oscillations . as the propagation constant is increased e.g. , for mode number 25 the mode profile oscillates a few times as in figure [ fig:1d - mode - profiles](b ) . the modes near the middle of the band in the vicinity of @xmath108 generally oscillate rapidly and spread over many waveguides , as can be seen for modes number 99 and 101 in figures [ fig:1d - mode - profiles](c ) and ( d ) , respectively . these modes become more localized if the disorder is increased . once the mode number gets closer to the bottom edge of the band , the oscillations become so rapid that the sign of the mode profile flips between adjacent waveguides , and the modes localize again . this behavior can be clearly seen for modes number 170 and 201 in figures [ fig:1d - mode - profiles](e ) and ( f ) , respectively . the mode width is plotted against the propagation constant of each mode in figure [ fig:1d - waveguide - prop - const - mode - widths ] , averaged over 1000 simulations for ( a ) @xmath75 $ ] and ( b ) @xmath77 $ ] . it is clear that the band edges result in the narrowest modes , while the modes near the middle of the band are more spread out . also , the stronger disorder in figure [ fig:1d - waveguide - prop - const - mode - widths](b ) results in smaller mode widths compared with the weaker disorder case in figure [ fig:1d - waveguide - prop - const - mode - widths](a ) . $ ] and ( b ) @xmath77$].,width=384 ] the disorder in the model studied here is off - diagonal . although the effects of diagonal and off - diagonal disorder are generally quite similar , there exist subtle differences between the two types of disorder . for example , it has been argued that while all modes are exponentially localized for diagonal disorder , this may not be strictly true for off - diagonal disorder @xcite . specifically , the modes at the center of the band in a 1d off - diagonally disordered lattice fall off asymptotically as @xmath109 instead of the simple exponential form @xmath110 , where @xmath111 represents the distance from the center of the mode at @xmath112 . the @xmath109 form is also responsible for the divergence of the mode width in the middle of the band in both subfigures of fig . [ fig:1d - waveguide - prop - const - mode - widths ] . we emphasize that even in the middle of band , the mode is localized , although the weaker localization form results in a divergent mode width . there are also differences between diagonal and off - diagonal disorder in 2d that will be pointed out later in section [ sec : aamechanism ] . , except the _ k - space _ beam width is averaged over 100 independent statistical realizations of the disordered waveguides . the error bars signify the one standard deviation for the beam width over the 100 samples . ] , except the _ k - space _ beam width is averaged over 100 independent statistical realizations of the disordered waveguides . the error bars signify the one standard deviation for the beam width over the 100 samples . ] , except the _ k - space _ beam width is averaged over 100 independent statistical realizations of the disordered waveguides . the error bars signify the one standard deviation for the beam width over the 100 samples . ] , except the _ k - space _ beam width is averaged over 100 independent statistical realizations of the disordered waveguides . the error bars signify the one standard deviation for the beam width over the 100 samples . ] in figures [ fig : beamexpansion - samplex - waveguides1d ] , [ fig : beamexpansion - samplef - waveguides1d ] , [ fig : beamexpansion - xspace - waveguides1d ] , [ fig : beamexpansion - fspace - waveguides1d ] , we explore in some detail the expansion and localization of the optical beam that is coupled to the waveguide array of figure [ fig:1d - array - fiber ] , as a function of the longitudinal coordinate . the vertical axis is the beam width , which is determined by the second moment method of eq . [ eq : xi ] . the parameters used for the disorder relate to the cases already discussed in figure [ fig : randomcoupledwaveguides1d ] . for these figures we have used @xmath68 , @xmath69 , and @xmath70 . any figure labeled with ( a ) corresponds to the disorder - free periodic array ; labeled with ( b ) corresponds to the weakly disordered case of @xmath74 , where @xmath75 $ ] ; and labeled with ( c ) corresponds to the strongly disordered case of @xmath77 $ ] . the only minor differences are that @xmath113 , and the boundary condition is set to @xmath114,\quad j=1,\cdots , n , \quad { \cal w}_0=\sqrt{2}. \label{eq : exponentialinput}\ ] ] we note that the beam width formula of eq . [ eq : xi ] ( presented later in this article ) also gives the value of @xmath115 for the beam width at @xmath67 . figure [ fig : beamexpansion - samplex - waveguides1d](a ) shows the expansion of the beam width in the disorder - free lattice . the expansion is ballistic , i.e. , the width grows linearly with propagation distance @xmath76 ( at large @xmath76 ) . in figure [ fig : beamexpansion - samplex - waveguides1d](b ) we plot the expansion of the beam width for three sample realizations of the random waveguide for the weakly disordered case , as explained above . the tendency to localize can be seen in each of these samples , but one can immediately see the random nature of the process , where different random realizations of the coupled waveguide system result in different rates of initial expansion and final beam localization width . lastly , the expansion of the beam width for three sample realizations of the random waveguide for the strongly disordered case is shown in figure [ fig : beamexpansion - samplex - waveguides1d](c ) . note the different vertical scale in these three subfigures . it is clear that despite the random variation the disorder slows and eventually halts the expansion , and a stronger disorder results in a smaller eventual localization width _ on average_. it is interesting to study the beam expansion in the fourier @xmath2_-space _ ( momentum space ) as well . a well - known characteristic of the ballistic expansion of the beam is that the beam width in the @xmath2_-space _ remains invariant under propagation , as shown in figure [ fig : beamexpansion - samplef - waveguides1d](a ) . the @xmath2_-space _ expansion of the cases plotted in figures [ fig : beamexpansion - samplex - waveguides1d](b ) , ( c ) are shown in figures [ fig : beamexpansion - samplef - waveguides1d](b ) , ( c ) . interestingly , the beam widths in the @xmath2_-space _ conform well to our intuition that narrower beams in @xmath116_-space _ are wider in @xmath2_-space _ ; of course , this simplistic intuition is not always true , especially in the presence of spatial chirp . as was noted above , the statistical nature of anderson localization means that the beam expansion rate and the eventual localization width varies depending on the particular `` random '' realization of the disordered waveguide . therefore , no two propagations will be identical . however , some general conclusions can still be drawn , in a statistical sense , about the average beam width , as well as the variation around the average . in figure [ fig : beamexpansion - xspace - waveguides1d ] we show the average beam width over 100 independent statistical realizations of the disordered waveguide , plotted as a function of the longitudinal propagation coordinate @xmath76 . figure [ fig : beamexpansion - xspace - waveguides1d](a ) is the disorder - free ballistic propagation and is plotted again for comparison with figures [ fig : beamexpansion - xspace - waveguides1d](b ) and ( c ) , which correspond to the weakly and strongly disordered cases , respectively . the vertical scales in these subfigures are all chosen identically , for easier comparison . it is clear that , on average , the localized beam width is narrower for the case of a stronger disorder , as expected . the error bars in figures [ fig : beamexpansion - xspace - waveguides1d](b ) and ( c ) signify one standard deviation for the beam width around the mean over the 100 independent random samples . strong disorder results , not only in a smaller beam width , but also in a smaller variation around the average beam width . therefore , a stronger disorder is equivalent to a better predictability : when the disorder is strong , the beam width is smaller and is almost the same in all random realizations of the disordered waveguide . finally , in figure [ fig : beamexpansion - fspace - waveguides1d ] we plot the corresponding average beam width in @xmath2_-space _ , where the error bars are now defined as the one standard deviations of the beam width in @xmath2_-space_. the interested reader is urged to consult refs . @xcite for a more detailed account of the beam expansion in 1d and 2d disordered lattices e.g. , it is argued that in a 2d disordered waveguide the expansion starts as ballistic then goes through a diffusive phase , and eventually the localization takes over . however , in 1d the diffusive transport regime is absent and the expansion turns from ballistic directly to localized . * highlights : * * the light propagating through a coupled array of identical optical waveguides remains confined to only a few waveguides ( transversely localized ) , if the waveguide - to - waveguide couplings are random . * similar localization behavior can be observed if the individual waveguides are randomized , regardless of whether the waveguide - to - waveguide couplings are random or not . * the stronger the disorder , the more localized the beam is . transverse anderson localization in two transverse dimensions can be explored in a similar fashion to 1d transverse localization discussed in the previous section . consider the 2d array of optical fibers placed in a hexagonal lattice in figure [ fig:2d - array - fiber ] . the hexagonal lattice is special because each fiber has six nearest neighbors separated by an equal distance @xmath26 , and the next - to - nearest separations are @xmath117 . in contrast , a fiber in a square lattice has four nearest neighbors and the next - to - nearest separations are @xmath118 . therefore , the next - to - nearest couplings are weaker in a hexagonal lattice and the simplifying assumption of the nearest neighbor coupling is more reliable . fibers . ] similar to eq . [ eq : coupledmode1 ] , the coupled mode equations for the propagation of the optical field through this optical fiber lattice can be expressed as @xmath119 where the sum is only on the nearest neighbors , and the coupling matrix is symmetric . as we discussed in the previous section , the transverse anderson localization can be explored by studying the distributions of the elements of eigenvectors of the @xmath81 matrix over the element position space . here , @xmath81 is the effective propagation constant matrix of the coupled fiber system and is defined as @xmath120 , noting that @xmath121 are non - zero only when @xmath12 and @xmath2 are the nearest neighbor fibers . the fiber coupling constants are assumed to be random in the general form of @xmath122 , where @xmath73 is a fixed value of @xmath123 and belongs to a random distribution . here , we take @xmath68 and @xmath124 on a hexagonal lattice of @xmath125 fibers . in figure [ fig : hex - array - uniform ] we consider the non - random deterministic case with @xmath126 . the magnitude of the four eigenvectors of the @xmath81 matrix with the largest eigenvalues are shown as density plots on the element - position domain of the hexagonal lattice . for the reader familiar with coupled mode theory , the calculated eigenvectors are the supermodes of the entire waveguide structure @xcite . the left most figure shows an azimuthally symmetric distribution of non - zero elements that decrease monotonically from the center of the lattice toward the boundary . other eigenvectors have different distributions and symmetry properties , but they all nearly fill the entire element - position domain , i.e. , the non - zero elements of the eigenvectors spread over nearly the entire lattice . , except the propagation constants of the individual fibers are randomized ( diagonal disorder ) . note that each eigenvector is localized in a certain region on the lattice similar to figure [ fig : hex - array - random - offdiagonal ] , as expected from the transverse anderson localization . ] , except the propagation constants of the individual fibers are randomized ( diagonal disorder ) . note that each eigenvector is localized in a certain region on the lattice similar to figure [ fig : hex - array - random - offdiagonal ] , as expected from the transverse anderson localization . ] , except the propagation constants of the individual fibers are randomized ( diagonal disorder ) . note that each eigenvector is localized in a certain region on the lattice similar to figure [ fig : hex - array - random - offdiagonal ] , as expected from the transverse anderson localization . ] let s consider what happens when the coupling constants are randomized according to @xmath127 $ ] in figure [ fig : hex - array - random - offdiagonal ] . similar to the above , the magnitude of four eigenvectors of the @xmath81 matrix are shown as density plots on the element - position domain of the hexagonal lattice . all figures show that the non - zero elements of the eigenvectors are localized in certain regions on the lattice , as expected from the transverse anderson localization . of course , the localization region for each eigenvector is different . as mentioned before , random variations of the diagonal elements of the @xmath81 matrix can also induce localization of the eigenvectors in the element - position domain . this is shown in figure [ fig : hex - array - random - diagonal ] , where the off - diagonal elements are assumed to be constant values of @xmath124 and the diagonal elements ( propagation constants of the individual optical fibers ) are randomized according to @xmath128 , where @xmath129 $ ] . therefore , the effect of the diagonal disorder is similar to that of the off - diagonal disorders , at least qualitatively . of course , transverse anderson localization is similarly observed if both diagonal and off - diagonal disorder are implemented simultaneously . transverse anderson localization in a 2d lattice of optical fibers with random variations of the parameters was first proposed by s. s. abdullaev and f. kh . abdullaev of the heat physics department of uzbekistan academy of sciences in 1980 @xcite . earlier in this tutorial review , we briefly highlighted the subtle differences between diagonal and off - diagonal disorders in a 1d disordered lattice . the weaker localization in the middle of the band for off - diagonally disordered lattices persists in 2d as well . for example , it has been argued in ref . @xcite that in the middle of the band for a 2d disordered square lattice , the wave amplitude falls off with distance as @xmath130 where @xmath131 is the distance from the center of the mode and @xmath28 is an exponent that depends on the amount of disorder . the power - law localization is also attributed to the geometry of the square lattice ( and the logarithmic singularity of the density of states in the middle of the band ) . therefore , for the triangular lattice that has been studied in this section ( which does not have a logarithmic singularity of the density of states in the middle of the band ) , all modes are exponentially localized . * highlights : * * transverse anderson localization is observed for a 2d disordered coupled waveguide array , just as in 1d . the 2d transverse anderson localization of light that was proposed by abdullaev and abdullaev and discussed in the previous section strongly resembles the original proposal by p. w. anderson . we will refer to this as the aa mechanism for the rest of this article , not to be confused with the aubry - andre model @xcite . in either case , the starting point is a lattice - periodic potential where each site on the lattice is characterized by a `` bound - state '' energy ( @xmath57 for fibers ) and off - diagonal couplings ( @xmath121 for fibers ) . in the absence of randomness , the solutions to the wave equations are bloch - periodic solutions that extend over the entire lattice . if sufficient randomness is introduced in the energy and/or couplings , the solutions of the wave equation no longer extend over the entire lattice ; rather , they localize to certain regions on the lattice . an alternative method to obtain the 2d transverse anderson localization of light was independently proposed by de raedt , lagendijk , and de vries in 1989 @xcite . recall that for the aa mechanism , the disorder was introduced on top of an existing ordered lattice . however , in the suggested method by de raedt , lagendijk , and de vries ( rlv mechanism ) , the randomness is not superimposed on top of an existing ordered lattice ; rather , the underlying potential is completely random @xcite . de raedt _ et al_. @xcite proposed that this scheme can be realized in a quasi-2d optical system in a dielectric with a transversely random and longitudinally uniform refractive index profile . this is sketched in figure [ fig : beam - prop - sketch]a , where a long dielectric `` waveguide '' has a random refractive index profile that is invariant along the fiber . de raedt _ et al_. suggested a binary random system , where each pixel is randomly chosen to have a refractive index of @xmath25 or @xmath24 with equal probabilities , where @xmath25 and @xmath24 are marked by red and blue color , respectively in figure [ fig : beam - prop - sketch]a . using extensive numerical simulations , de raedt _ et al_. showed that for a properly designed random fiber e.g. with @xmath132 , @xmath133 ; and @xmath134 , where @xmath135 is the width of each pixel and @xmath28 is the wavelength of light , they obtain quasi-2d localization in the transverse plane of the waveguide . an optical field that is launched in the longitudinal direction initially expands until it reaches a terminal localization radius , after which the freely propagating beam fluctuates around a relatively stable radius . in figure [ fig : beam - prop - sketch]b , a slice of the intensity distribution in the @xmath116-@xmath76 plane is shown for a representative localized propagation through a random binary waveguide , where @xmath116 is a transverse coordinate and @xmath76 is the longitudinal coordinate . * highlights : * * one way to observe transverse anderson localization is to use a 2d disordered coupled waveguide array . we refer to this as the aa mechanism because it was originally proposed by abdullaev and abdullaev . * alternatively , transverse anderson localization can be observed in a optical fiber - like medium with a fully random refractive index profile . we refer to this as the rlv mechanism because it was originally proposed by de raedt , lagendijk , and de vries . the rlv mechanism is distinct from the aa mechanism by the absence the underlying lattice . the first experimental observation of the transverse anderson localization of light was carried out in segev s group in 2007 @xcite . the experiment was more in line with the aa mechanism discussed earlier in section [ sec : aamechanism ] , where the disorder was superimposed on top of an existing triangular lattice of waveguides . the ordered lattice and the superimposed disorder were formed , by means of the optical induction technique @xcite , in a photo - refractive crystal ( sbn:60 ) using a laser beam at 514 nm wavelength . in order to write the underlying ordered lattice in the photo - refractive crystal , schwartz _ et al_. @xcite used the interference pattern generated by three symmetrically arranged lasers and obtained a fully periodic hexagonal interference pattern with periodicity of 11.2 m inside the 10 mm - long crystal . the disorder is generated from a speckled beam that is invariant in the longitudinal direction , as required for the observation of transverse anderson localization . the speckled beam was formed from a bessel beam , created by passing a gaussian laser beam through an axicon , and then passed through a 4-f imaging system , where a diffuser was placed at the joint fourier plane between the two lenses . the 4-f imaging system with the diffuser transformed the bessel beam into a broad and speckled beam that was coherently combined with the original lattice - forming beam to create the desired refractive index fluctuations on top of the ordered lattice in the photo - refractive crystal . in order to investigate the localization , schwartz _ et al_. used another probe beam at 514 nm wavelength . the beam width was 10.5 m full - width at half - maximum ( fwhm ) and was always launched into the crystal at the same location . in the absence of disorder , they observed diffraction patterns after 10 mm of propagation . they observed clear transverse localization of the probe beam when the disorder level was increased to more than 30% . in order to obtain an appropriate ensemble to calculate the mean localization radius ( localization length ) , the diffuser was rotated by a step - motor , such that the laser beam passed through a different location on the diffuser in each step , where a new intensity measurement was taken . for the large disorder level of 45% , et al_. fitted the properly averaged intensity distribution of the localized beam , over 100 independent experiment , to an exponential of the form @xmath136 , and obtained @xmath137=64 m for the localization length of the anderson localized beam . the experiment carried out by schwartz _ et al_. was quite interesting , as they were able to vary the disorder level by controlling the intensity level of the disorder - inducing lasers , and also obtained an ensemble of independent measurements for the statistical analysis of the localization phenomenon by rotating the diffuser . however , the variations of the refractive index of random sites in ref . @xcite were on the order of @xmath138 . as we will discuss later in greater detail , the radius of the localized beam ( localization length ) depends on the magnitude of the refractive index fluctuations . in order to obtain a more localized beam for device applications e.g. , comparable to the beam radius in a conventional optical fiber , it is necessary to increase the magnitude of the index fluctuations @xcite . moreover , when the magnitude of the index fluctuation is larger , the sample - to - sample variation in the beam radius becomes smaller ; therefore , each element of the ensemble closely resembles the average . this self - averaging behavior alleviates the need for averaging over a large ensemble , and for a sufficiently large magnitude of the index fluctuation the average is almost identical to each element . we will discuss a recent realization of the transverse anderson localization in the presence of a large index contrast in the next section . * highlights : * * the first observation of transverse anderson localization was reported by schwartz _ et al _ in an aa - mechanism setup . * the underlying lattice and the overlaying disorder were optically induced in a photo - refractive crystal . a separate beam was used to probe the localization behavior . * the setup allowed them to vary the amount of disorder on demand . in section [ sec : rlv ] we discussed the rlv mechanism for the transverse anderson localization of light . the quasi-2d system proposed by de raedt _ et al_. @xcite strongly resembles an optical fiber . in 2012 , karbasi _ et al_. @xcite designed and fabricated an optical fiber that functioned based on the rlv mechanism , i.e. , transverse anderson localized in the presence of an entirely random refractive index profile . in order to obtain the random and pixelated refractive index profile of the rlv mechanism , karbasi _ et al_. @xcite used 40,000 pieces of a low index polymer fiber polymethyl methacrylate ( pmma ) with refractive index of 1.49 and 40,000 pieces of a high index polymer fiber polystyrene ( ps ) with refractive index of 1.59 . each fiber was 8 inches long with an approximate diameter of 200 m. the fibers were randomly mixed , assembled to a square preform as shown in figure [ fig : fiber - strands ] , fused together , and redrawn to a fiber with a nearly square profile and approximate side width of 250 m @xcite . we recall that one of the requirements for transverse anderson localization is the longitudinal invariance . some of the randomly mixed optical fibers may have crossed over each other during the assembly and redraw process ; however , the large draw ratio of the fiber guarantees that the refractive index profile remains relatively unchanged along the fiber . moreover , small perturbations are likely not going to disturb the transverse localization noticeably . figure [ fig : palof - profile - localization ] shows the scanning electron microscope ( sem ) image of a polished polymer anderson localized optical fiber ( palof ) . figure [ fig : palof - profile - localization]a is the image of the cross section of the optical fiber with an approximate side width of 250 m , where the high and low index regions are not distinguishable in this figure . figure [ fig : palof - profile - localization]b is a zoomed - in sem image of a 24 m wide region on the tip of palof exposed to 70% ethanol solvent to to dissolve the pmma , so that the pmma region can be differentiated by the darker color . the sizes of the random features ( pixels ) in figure [ fig : palof - profile - localization]b are around 0.9 m. in order to investigate the guidance and localization properties of the palof , light from a single mode optical fiber was directly launched ( nearly butt - coupled ) into the palof . the output near - field image was collected by a 40x objective and was projected onto a ccd camera . figure [ fig : palof - profile - localization]c shows an experimental measurement of the near - field intensity profile of the localized beam after 60 cm of propagation through palof , where the wavelength of light is 633 nm . the localization was observed to be strong : when the input beam was scanned across the input facet , the output beam clearly followed the transverse position of the incoming beam @xcite . we encourage the interested reader to consult ref . @xcite for another method to obtain transverse anderson localization based on the rlv mechanism . the optically induced randomized potential is created by a computer controlled spatial light modulator ( slm ) in a photorefractive crystal . * highlights : * * experimental observation of transverse anderson localization for the rlv mechanism was reported by karbasi _ et al _ in a disordered optical fiber medium . * the disordered fiber was drawn from a random preform . the preform consisted of 80,000 strands of optical fibers with different refractive indexes that were randomly mixed and fused together . * when the input beam was scanned across the input facet , the output beam clearly followed the transverse position of the incoming beam an attractive feature of the palof is the large index contrast of 0.1 that is helpful in reducing the localization radius ( localization length ) of the beam . the measured value for the localization radius was reported as @xmath13931 at 633 nm wavelength . the measured radius was only slightly larger than the calculated value ; however , the measured standard deviation of @xmath14014 m was considerably larger than the variations calculated numerically and reported in figure 3 of ref . the measured standard deviation for the beam radius was taken from 100 separate measurements of the beam profile radius with 20 different fiber samples and 5 different locations across the fiber . this discrepancy between the theoretical and experimental variation of the beam radius was mainly attributed to the quality of the polishing of the palof tip as the surface quality of a polished polymer optical fiber tip is generally lower than what is routinely achievable for glass optical fibers . the simulation of light propagation in a disordered fiber is carried out by numerically solving the wave propagation equation eq . [ eq : bpm ] using the finite difference beam propagation method ( fd - bpm ) @xcite . @xmath141=0.\ ] ] eq . [ eq : bpm ] is the paraxial approximation to the helmholtz equation , where @xmath142 is the slowly - varying envelope of the primarily transverse electric field @xmath143 $ ] centered around frequency @xmath144 and @xmath27 . @xmath145 is the ( random ) refractive index of the optical fiber , which is a function of the transverse coordinates , and @xmath146 is average refractive index of the fiber . the forward propagation scheme is implemented using the fourth order runge - kutta method @xcite . the stability condition for the fourth order runge - kutta method limits the size of the steps in the longitudinal direction as @xmath147 , where @xmath148 is assumed to be the size of the transverse grid in the finite - difference numerical scheme , and @xmath149 in a uniform medium . in the simulations in ref . @xcite , @xmath150 is chosen in ref . @xcite to ensure stability and no power dissipation for reliable long distance propagation . while transparent boundary condition @xcite is implemented , the size of the simulation domain is taken to be large enough to ensure that the total power in the simulation region remains unchanged along the fiber for anderson localized beams . a typical reliable simulation of transverse anderson localization e.g. for palof , requires a transverse area in the range of @xmath151 . the random refractive index pixel size is on the order of a wavelength , and each pixel must be resolved by roughly 10 points in each direction in the finite difference scheme ; therefore , @xmath152 points are required in the transverse domain with @xmath153 . this results in the fourth order runge - kutta stability criterion of the form @xmath154 , which is @xmath155 for @xmath150 . for @xmath156500 nm , @xmath1571 nm , and for a typical propagation distance of 1 cm required for reliable localization , @xmath152 steps in the longitudinal direction are required , which is computationally intensive . the effective beam radius ( localization length ) is calculated by the variance method @xcite as @xmath158 where the angle brackets denote integration over transverse @xmath159 coordinates . @xmath160 is the transverse position vector and @xmath161 is the vector pointing to the center of the beam , defined as the mean intensity position by @xmath162 . the optical field is assumed to be normalized according to @xmath163 . we note that an estimate of the localization length is sometimes given using a quantity called the inverse participation ratio ( ipr ) @xcite . ipr is defined as @xmath164 where @xmath165 is the optical intensity defined in the @xmath159 transverse plane , at the longitudinal coordinate @xmath76 . ipr has units of inverse area and the average effective width is defined as @xmath166 , where @xmath167 represents the statistical averaging over the ensemble . we note that ipr as defined by eq . [ eq : ipr ] is originally rooted in the definition of the mode effective area in nonlinear optics ( see e.g. @xcite ) . equation [ eq : xi ] used by de raedt _ et al_. @xcite is likely a better representation of the localization phenomena and more true to the random - walk nature of the scattered wave . therefore , we prefer to use the second moment method of eq . [ eq : xi ] , rather than the inverse participation ratio to calculate the localization length . the experimental realization of the rlv scheme by karbasi _ et al_. @xcite showed that the transversely disordered optical fibers can be embraced as a completely new class of optical fibers that guide light , not in a conventional core - cladding setting , but by means of the anderson localization , where any location across the transverse profile of the fiber can be used to guide light . as we will see later in image transport applications of this disordered optical fiber , a small beam localization radius is likely its main desired attribute . a reduced variance is also desired , because as a device it is not helpful to have variations in the beam radius that depend on the transverse location of the beam in a random and unpredictable way . in other words , although the underlying waveguiding mechanism is based on disorder and randomness , the main measurable device attribute , the localized beam radius , should be predictable . these considerations lead one to ask what can be done , in the design of these disordered fiber in order to minimize the localization radius and also reduce the sample - to - sample variations in the localization radius . the localized beam radius can depend on : the input characteristics of the optical beam such as the wavelength and the in - coupling beam diameter ; the dimensionful fiber characteristics such as the transverse dimensions of the fiber , the width of each pixel ; and the dimensionless fiber characteristics such as the refractive index of the components , especially the index difference , and the geometrical ratio of the components used in the fiber characterized by the fill - fraction . given the intuition we established based on the modal picture , we learned that if the initial excitation width is smaller than the radius of a typical mode of the disordered waveguide ( localization radius or localization length ) , the propagating light eventually localizes to a radius that is comparable to the localization length ; and if the initial excitation is wider than the typical width of an individual mode , the propagating light eventually localizes to a slightly wider beam than the original size . therefore , the design program to optimize the disordered waveguide for best localization must be independent of the input beam profile and must focus on the transverse dimensions of the localized beams @xcite . another important property of the disordered fiber is the transverse size of the fiber . if the fiber cross section is not large enough , the boundaries of the fiber strongly affect the localization radius of the modes . the impact of the boundary on the localization radius of the modes that reside near the boundary is inevitable , but in an ideal case the fiber cross section must be large enough so that the modes in the interior region are shielded from the boundary to prevent excessive scattering loss ; therefore , the physical characteristics of the majority or nearly all of the interior modes are solely determined by the transverse anderson localization mechanism . in the ideal case , when the disordered fiber cross section is much larger than the light wavelength , feature size ( width of each index pixel ) , and the localization radius of the typical modes , the total transverse size of the fiber becomes irrelevant . in that case , the only two dimensionful fiber design parameters are the light wavelength and the width of each pixel . considering the fact that maxwell s equations are scale invariant , if the wavelength and feature size ( fluctuating index pixel size ) are scaled by some factor , the localized beam radius of each mode is scaled by the same factor . the feature size of the palof studied in refs . @xcite are approximately 0.9 m. experimental measurements and numerical simulations have confirmed that the localization radius is much smaller at 405 nm wavelength than at 633 nm wavelength , as shown in figure [ fig : w633nmvs405 ] . it is speculated that the optimum feature size for strong localization using the rlv scheme is around @xmath168 . this statement is merely an educated guess and the optimum value of the feature size relative to the wavelength has yet to be found . intuitively , if the feature size is too small compared with the wavelength , the optical field would hop over the random sites and merely average the refractive index , with very weak scattering . on the other hand , if the feature size is too large compared with the wavelength , the mean free path for wave scattering will be large , resulting in a large localization radius . therefore , there must be an optimum value of the feature size for a given wavelength . finding the optimum value is computationally very challenging . another design parameter that affects the localized beam radius in the binary disordered rlv scheme is the portion of each random refractive index component , characterized by the fill - fraction @xmath169 . in ref . @xcite , @xmath169 is defined as the portion of the low - index material in the higher index host medium . using an intuitively plausible argument that the maximum transverse scattering is obtained when there is an equal amount of low - index and high - index material in the disordered matrix , and also a few instances of simulations relevant to the experimental regime of interest , karbasi _ et al_. @xcite have argued that @xmath170 should be regarded as the ideal design target . a sample simulation is shown in figure [ fig:40p50p ] , where the evolution of the effective beam radius versus the propagation distance is shown for different values of the fill - fraction of @xmath171 and @xmath172 ; and the latter provides a lower effective beam radius and localization length . , and @xmath172 , where the latter provides a lower effective beam radius and localization length . adapted with permission , copyright 2012 , optical society of america @xcite.,width=240 ] the last , but definitely not the least important design parameter to reduce the localization radius is the index difference between the random components of the disordered fiber . it is generally believed that a stronger localization ( smaller average mode radius ) is obtained if the refractive index difference between the random constituents is increased . this is confirmed in the 1d model of the transverse anderson localization of ref . however , it is also shown that the dependence of the average mode radius on the index difference asymptotically saturates beyond a certain value of the index difference ; therefore , the payback in the reduction of the localization length may be quite small beyond a certain threshold index difference . while these issues are well studied in the 1d model of ref . @xcite , the 2d analysis remains to be done . for the 2d geometry , karbasi _ et al_. @xcite compared the localization radius in a glass - air disordered fiber with palof . the refractive index of sites in the glass - air disordered fiber are randomly picked as @xmath173 and @xmath174 for the glass host and random air - hole sites , respectively . the index difference of 0.5 in the glass - air disordered fiber compared with 0.1 in palof results in a considerable reduction in the localization radius . in figure [ fig : fillfractionglass ] , the calculated beam radius versus propagation distance for different values of fill - fraction are plotted , and thelocalized beam radius is clearly smaller than that provided by the index difference of 0.1 for palof . , in glass disordered optical fibers with random air holes . the index difference of 0.5 between the random sites results in a very small localization radius . adapted with permission , copyright 2012 , optical society of america @xcite.,width=240 ] finally , we have repeatedly observed that whenever the scattering strength is increased to reduce the localization length ( localized beam radius ) , whether by selecting the right wavelength , or by bringing the fill - factor close to 50% , or by increasing the index difference , the statistical variation of the beam radius decreases . therefore , we have the luxury of solving two problems with one solution ( a more animal friendly expression than _ killing two birds with one stone ! _ ) . the possibility of spatial beam multiplexing in palof was studied in ref . @xcite in 2013 . it was shown , both numerically and experimentally , that a palof can be used to simultaneously transport multiple beams @xcite . in figure [ fig : beam - multiplex](a ) beam multiplexing in palof is studied numerically at 405 nm wavelength , where the intensity profile of a multiplexed beam is plotted after 5 cm of propagation along the fiber ; the four exterior beams are launched at a distance of 70 m from the central beam . the output beams do not show any appreciable drift and remain in the same transverse location across the fiber . experimental verification of beam multiplexing are shown in figure [ fig : beam - multiplex](b ) , where both beams are 405 nm wavelength , and in figure [ fig : beam - multiplex](c ) , where the upper beam is at 405 nm wavelength and the lower beam is at 633 nm wavelength . the localization is clearly stronger at 405 nm due to the choice of the design parameters of palof , as discussed before . in section [ sec : bm ] the possibility of beam multiplexing in an disordered fiber was discussed . given the unconventional nature of the beam confinement in a disordered optical fiber , one must worry about the possibility that the spatially multiplexed beams drift across the fiber when the fiber is subjected to substantial macro - bending . this issue was studied in detail in ref . using numerical simulations , it was shown that transverse anderson localization is very robust and can withstand , at least in theory , a degree of macro - bending that is beyond what is acceptable for conventional fibers . to explore the macro - bending experimentally , a 10 cm section of a 15 cm - long palof was wrapped 16 times around a mandrel with an approximate radius of 1 mm , where no appreciable loss of walk - off effect was observed in the localized beam . motivated by the successful demonstration of beam multiplexing , karbasi _ et al_. used palof for endoscopic fiber - optic imaging . to their pleasant surprise , the image transport quality was comparable to or better than some of the best commercially available multicore imaging fibers , with less pixelation and higher contrast @xcite . figure [ fig : experimental405 ] shows some of the transported images in the form of numbers from a section of the 1951 u.s . air force resolution test chart through palof . the test - target , in the form of a stencil in which numbers and lines were carved , was butt - coupled to the hand - polished input facet of palof and was illuminated by white light . the near - field output was projected onto a ccd camera with a 40@xmath175 microscope objective . the minimum resolution of the images is determined by the width of the point spread function of the disordered optical fiber imaging , which was calculated to be smaller than 10 m at 405 nm wavelength @xcite . in practice , the imaging resolution in palof is limited by the quality of the cleave and polishing of the fiber . the fiber surface quality is partially responsible for the distortions in the transported images in figure [ fig : experimental405 ] . the high quality image transport in the proof - of - concept experiment has been achieved without any optimization in the design . highly multicore optical fibers , similar to those introduced in section [ sec : aamechanism ] , have been used for direct transportation of images in various configurations @xcite . the highly multicore fiber , often referred to as the `` coherent fiber bundle , '' is commonly used in medical and industrial endoscopy @xcite . for an anderson localized fiber , a higher amount of disorder and a larger level of fluctuation in the refractive index provides stronger beam localization , resulting in an improved image resolution . a similar statement can be said for multicore imaging fiber as well : the coherent core - to - core coupling is detrimental and blurs the image ; therefore , core - to - core coupling must be suppressed by varying the size of the cores so that neighboring cores can not couple resonantly . also , in a multicore imaging fiber , the cores must be as close as possible to create a less pixelated image , while being very different in size so that they can not couple efficiently even when they are close . one can view the disordered anderson localized fiber as taking the two limits of the high - packing of the cells and the large variation of the sizes to an extreme , so that the individuality of each core is completely lost : _ all neighboring sites are strongly coupled , but the extreme randomness prevents light leakage and blurring . _ m - long and the same scale bar can be used for ( a ) and ( c ) . each fiber is approximately 5 cm long . adapted with permission , copyright 2014 , nature communications @xcite.,width=321 ] the imaging performance of the `` unoptimized '' palof compares with some of the best commercially available multicore imaging optical fibers , as is further confirmed in figure [ fig : fig-5 ] . the transported images over 5 cm of the number `` 6 '' from group 5 of the 1951-aftt test chart are compared between palof in figure [ fig : fig-5](a ) , fujikura figh-10 - 350s in figure [ fig : fig-5](b ) , and fujikura figh-10 - 500n in figure [ fig : fig-5](c ) . the image quality of the transported image through the palof is clearly better than figh-10 - 350s and is comparable with figh-10 - 500n . the feature sizes in figure [ fig : fig-5 ] are on the order of 10 - 20 m. the rayleigh range for this level of resolution is approximately 0.8 - 3 mm , which is substantially shorter than the typical propagation length in these imaging fibers . therefore , the imaging results are non - trivial and can not be obtained using bulk propagation or conventional multimode fibers . * highlights : * * in an anderson localized disordered optical fiber , a smaller localized beam radius is obtained via increasing the differences between the refractive indexes of the random dielectric constituents . * the feature size should be @xmath176 , and a fill - fraction of 50% is preferred . * a stronger localization is generally accompanied by a smaller variation in the localized beam radius , resulting in a more uniform and predictable beam radius . * beam multiplexing and high - quality image transport are featured as device - level applications of the transverse anderson localization in a disordered optical fiber . the first observation of anderson localization in a silica fiber was reported in ref . the main motivation for using a glass - air structure has been the larger index contrast that results in a smaller beam diameter ( better image transport resolution ) , as well as a lower sample - to - sample variation in the value of the beam diameter ( better image uniformity ) @xcite . the reported glass - air disordered fiber was drawn at clemson university . the preform was made from `` satin quartz '' ( heraeus quartz ) , which is a porous artisan glass . by drawing the preform , the airholes ( bubbles ) in the glass are stretched to form the hollow air - rods required for transverse anderson localization . the large draw ratio sufficiently preserves the longitudinal invariance , without significant disturbance over typical lengths used in the experiments . the cross - sectional sem image of the disordered glass - air optical fiber is shown in figure [ fig : glass - profile]a , and a zoomed - in sem image is shown in figure [ fig : glass - profile]b . the sem images provide a good estimate of the refractive index profile of the fiber ; the light gray background matrix is glass and the black dots represent the random airholes . the diameter of the disordered glass fiber is measured to be 250 m. the diameters of the airholes vary between 0.2 m and 5.5 m. unfortunately , the airhole fill - fraction was shown to be as low as 2% in the central regions , which was far below the ideal value of 50% , so strong localization was not observed near the center . however , anderson localization was observed near the boundary , where the airhole fill - fraction is 8% . more recently , transverse anderson localization has been reported by scientists from corning incorporated in random airline fibers ( ralf ) @xcite . the preform for ralf is fabricated using the outside vapor deposition ( ovd ) process . the consolidation process is done in the presence of 100% nitrogen gass , during which nitrogen is trapped in the blank to form glass with randomly distributed air bubbles . when the preform is drawn , the random air bubbles stretch to form random airlines . ralfs with 150 , 250 and 350 m diameters were fabricated , where the averaged air line diameters were 177 , 247 and 387 nm in these fibers , respectively . the maximum airhole fill - fractions measured in ralfs is reported to be 1.473% , which is considerably lower than that of ref . @xcite . ref . @xcite reports the observation of transverse anderson localization of light , despite the low airhole fill - fraction , and attributes its observation to the substantially smaller diameter of the airholes compared with that reported in ref . @xcite . we also encourage the interested reader to consult a pioneering work by pertsch _ @xcite on light propagation in a disordered 2d array of mutually coupled optical fibers . they observed both localized and delocalized modes and also analyzed the impact of nonlinearity on these modes . * highlights : * * glass - air random optical fibers are highly desired because the large refractive index difference between glass and air can result in narrowly localized beams , and also small variations around the mean beam radius . * all attempts at glass - air random fibers have so far resulted in undesirably small air fill - fractions . as mentioned in section [ sec : numerics ] , the paraxial propagation of an optical beam in a longitudinally invariant medium is described by the paraxial approximation to the helmholtz equation , eq . [ eq : bpm ] . this equation is formally equivalent to the schrdinger equation in quantum mechanics , where the time variable @xmath177 is replaced with the longitudinal coordinate @xmath76 and the potential is replaced by the term proportional to the refractive index profile @xmath178 . the longitudinal invariance of the refractive index profile is taken into account by explicitly showing that @xmath179 is only a function of the transverse coordinates @xmath116 and @xmath180 . the requirement for the longitudinal invariance of the refractive index profile is equivalent to the potential being constant in time in the schrdinger equation . presence of some form of time dependence in the random potential can _ potentially _ change the localization picture . this issue has been studied extensively over the years e.g. , in refs . @xcite for temporally random or correlated potentials . an interesting finding has been the possibility of hyper - transport in certain temporally varying random potentials , where the wavefunction expands even more rapidly than the ballistic expansion observed in free space . we recall our earlier discussions on ballistic , diffusive , and localized propagation of light in disordered optical waveguides . for the propagation of the optical beam , the beam width generally grows with the propagation distance according to @xmath181 , where @xmath182 for ballistic , @xmath183 for diffusive , and @xmath184 for localized propagation . an example of the ballistic expansion is the familiar formula for the width of a gaussian beam as a function of the propagation distance @xcite @xmath185 where @xmath186 for @xmath187 . for a temporally random potential , ref . @xcite showed analytically that the width of the wavefunction , as measured by the second - moment method , expands with time according to @xmath188 . hyper - expansion was later shown for correlated temporally fluctuating potentials as well in refs . @xcite . hyper - transport for an optical beam was studied experimentally and theoretically in ref . @xcite and ref . @xcite , respectively , where the temporal variations were replaced by rapid longitudinal fluctuations in the refractive index profile of the waveguide . the experiment reported in ref . @xcite was performed in a similar setting to that of the first observation of the aa mechanism earlier reported in section [ sec : segevaa ] and ref . the disordered lattice was formed by means of the optical induction technique @xcite in a photo - refractive crystal . however , unlike ref . @xcite , where special care was taken to ensure the longitudinal invariance of the interference speckles and the resulting transversely random index profile , the refractive index profile created in the nonlinear crystal in ref . @xcite was made to vary with @xmath76 . the experiment proved that in the presence of a sufficiently large longitudinal variation of the index fluctuation , not only did the transverse localization cease to exist , but also the beam expanded at a rate faster than ballistic as it propagated through the random - index waveguide . * highlights : * * in the presence of rapid longitudinal fluctuations in the refractive index profile of a disordered waveguide , it is possible for the optical beam to expand even more rapidly than the ballistic expansion observed in free space . the interplay between nonlinearity and disorder is of great interest to the studies of transverse anderson localization . the main question that is often asked is whether the presence of nonlinearity perseveres , enhances , undermines , or destroys anderson localization . this issue has been explored over the years , where some of the studies have benefited from the existing literature on systems with similar dynamical equations , such as the bose - einstein condensate in the presence of disorder ( see e.g. ref . @xcite and the references therein ) . here , we briefly highlight some of the results that are more relevant to the concept of transverse anderson localization . for more details , we refer the interested reader to an excellent review on this subject by fishman , krivolapov , and soffer , in ref . @xcite . earlier , in section [ sec : segevaa ] , we reviewed the numerical and experimental work of schwartz _ et al_. @xcite that resulted in the observation of transverse anderson localization of light for the aa mechanism . the authors also investigated the transverse anderson localization of light in the presence of kerr nonlinearity , both numerically and experimentally . the defining equation for the nonlinear propagation of light is the nonlinear schrdinger equation ( nlse ) and is identical to eq . [ eq : bpmnl ] , with the addition of a kerr nonlineaity term @xcite : @xmath189+k_0 n_2 \|a\|^2 a=0,\ ] ] where @xmath24 is the nonlinear index , which is positive for self - focusing and negative for self - defocusing nonlinearity . as a case study , the authors considered a disordered lattice where the maximum contribution of the nonlinear term to the index change `` @xmath190 '' was assumed to be a maximum of 15% of the index contrast of the underlying periodic waveguide . they also varied the disorder level from 0% to 30% , where the disorder level was defined as the magnitude of random index fluctuations relative to the index contrast of the underlying periodic waveguide . they observed that over this range , the self - defocusing nonlinearity ( @xmath24 ) results in a moderate ( nearly negligible ) widening of the average beam profile . however , the self - focusing nonlinearity ( @xmath191 ) resulted in a substantial reduction of the average localized beam diameter . the enhancement of localization due to the self - focusing nonlinearity was particularly noticeable when the disorder level was less than 15% . the experiments were carried out at 15% disorder level . however , the maximum nonlinear contribution was taken to be equal or higher ( up to a factor of 3 ) , compared with the index contrast of the underlying periodic waveguide . this was achieved by making the probe intensity equal to or higher than the interference maxima of the lattice - writing beams ( for a perfect lattice ) . the statistical analysis of the localized beam radius clearly confirmed the expected reduction in the average beam radius due to the self - focusing nonlinearity . similar results were reported by lahini _ et al_. @xcite using disordered one - dimensional waveguide lattices . their experiment consisted of a one - dimensional lattice of coupled optical waveguides patterned on an algaas substrate . light was injected into one or a few waveguides at the input , and light intensity distribution was measured at the output . we recall our earlier discussions in section [ sec : modeshapes ] : we identified the highly localized eigenmodes near the top edge of the propagation constant band . the amplitude of these flat - phased modes , as they are referred to in ref . @xcite , are in - phase at all sites ; perhaps only with a few flips as seen in figures [ fig:1d - mode - profiles](a ) and ( b ) . another set of highly localized modes were shown to exist near the top edge of the propagation constant band . the amplitude of these staggered modes , as they are referred to in ref . @xcite , have phase flips between adjacent sites , as shown in figures [ fig:1d - mode - profiles](e ) and ( f ) . in the weak nonlinear regime , lahini _ et al _ observed that nonlinearity enhances localization in flat - phased modes and induces delocalization in the staggered modes . this behavior is explained as follows : the presence of the weak nonlinearity perturbatively shifts ( increases ) the value of the propagation constant of each localized mode . for the flat - phased modes , the nonlinearity shifts the modes outside the original linear spectrum . however , for the staggered , which belong to the bottom edge of the propagation constant band , a perturbative increase in the value of the propagation constant shifts it further inside the original linear spectrum . therefore , the propagation constant of a staggered mode can cross and resonantly couple with other modes of the lattice , resulting in delocalization @xcite . in 2008 , pikovsky and shepelyansky presented a somewhat different account of the interaction between disorder and nonlinearity @xcite . here we rephrase their main findings in a language more consistent with our notation so far . in a disordered coupled waveguide lattice , they demonstrated that above a certain critical strength of nonlinearity the anderson localization is destroyed and turns into a subdiffusive spreading . they focused on the discrete anderson nonlinear schrdinger equation , which is essentially the same as eq . [ eq : coupledmode1 ] with the addition of a third - order diagonal kerr nonlinear term , expressed as @xmath192 + \gamma \|a_j(z)\|^2 a_j(z)= 0,\quad j=1,\cdots , n . \label{eq : coupledmodenl}\end{aligned}\ ] ] they also assumed that the waveguide coupling coefficients @xmath73 is deterministic and is identical for all waveguides . the disorder is introduced through the diagonal propagation constant terms , where they are assumed to be randomly distributed according to @xmath193.\end{aligned}\ ] ] for the linear case of @xmath194 , the modes are exponentially localized due to the disorder . in the nonlinear case where @xmath195 , pikovsky and shepelyansky demonstrated that above a certain critical strength of nonlinearity , the anderson localization is destroyed and the field spreads in a subdiffusive form indefinitely across the optical lattice . they also showed that the mode width calculated using the second moment method ( see eq . [ eq : xi ] ) grows with the propagation distance as @xmath196 , where @xmath197 . pikovsky and shepelyansky presented a theoretical argument in support of the subdiffusive spreading of the beam with @xmath198 @xcite ; however , their work is mainly based on numerical integration of eq . [ eq : coupledmodenl ] and monitoring the results , up to @xmath199 . for the numerical simulation they used the boundary condition @xmath200 , where @xmath201 represents the middle waveguide , and the integration is performed by the operator splitting method . they also assumed that @xmath202 , without any loss of generality . in a sample set of simulations they chose the nonlinearity strength to be @xmath203 for two cases : case 1 with @xmath204 ; and case 2 with @xmath205 . the second moment was evaluated according to @xmath206 as expected , the initial expansion was ballistic for either case , but after some distance @xmath207 , the expansion became subdiffusive . they fit the subdiffusive expansion to @xmath208 over the range @xmath209 . in case 1 , for different instances of randomness , they obtained @xmath210 ; and for case 2 they reported @xmath211 . upon averaging over 8 independent realizations , they reported a fit of the form @xmath212 for case 1 and @xmath213 for case 2 over the subdiffusive range . they also reported a critical value of nonlinearity @xmath214 above which this subdiffusive behavior is observed . intuitively speaking , one may think that in a nonlinear disordered coupled waveguide system , the dynamics of the beam is initially influenced by nonlinearity ; and as the beam spreads , the effect of nonlinearity becomes weaker and the disorder dynamics takes over . therefore , one should always expect anderson localization after sufficiently long propagation . this is clearly in contrast with the findings of pikovsky and shepelyansky reported above . et al_. @xcite present a thorough survey of the many subtleties involved rearding the interaction of nonlinearity and disorder . the conclusion is that the situation can best be described as inconclusive at this point . for example , when using the numerical simulations , they caution that eq . [ eq : coupledmodenl ] is chaotic with an exponential sensitivity to numerical errors . for long - distance propagation , it is not clear that reducing the @xmath76 step size can control the cumulative numerical error , given that the limit of zero @xmath76 step may be singular . details are beyond our intended scope and can best be found in ref . @xcite . other forms on nonlinearity besides kerr can also interact with the disorder - induced localization . for example , it was recently shown that a beam of light propagating in a palof ( introduced in section [ sec : mafirlv ] ) exhibits self - focusing properties due to a thermal nonlinearity @xcite . the larger light absorption strength in pmma than ps results in a inhomogeneous temperature distribution . the higher temperature in pmma translates into a decrease of its refractive index . the result is an increased refractive index mismatch and stronger localization . the results are quite counter - intuitive , because the polymer materials used in the experiment have defocusing intrinsic nonlinear coefficients ( @xmath215 ) . in ref , @xcite , leonetti _ et al_. demonstrated that transversally localized modes shrink when the pump intensity is increased despite the fact that @xmath215 for the polymers . in a subsequent publication @xcite , the authors provided further evidence of this behavior by analyzing the direct relation between the optical intensity and the localization length , and also demonstrated the disorder - induced focusing by a monochromatic continuous wave ( cw ) laser . the interested reader is also urged to study other aspects of the interaction between nonlinearity and disorder not covered in this tutorial review e.g. , on soliton propagation in random media @xcite . * highlights : * * the main question of interest is whether the presence of nonlinearity perseveres , enhances , undermines , or destroys anderson localization . * the answer to the above question can best be described as inconclusive at this point . in the previous sections anderson localization was explored for a temporally and spatially coherent light , coupled to a disordered optical lattice . in this section some of the main issues related to the propagation of partially incoherent light are explored . a discussion on the propagation of single photons and correlated photons is also presented . anderson localization of waves with imperfect coherence was reported in 2011 by apeta , _ et al _ the main issue that was answered in their paper was the extent to which transverse anderson localization is affected by the partial coherence of the in - coupling beam in a disordered linear lattice . they observed that if all the eigenmodes of the disordered waveguide are exponentially localized , any partially incoherent beam exhibits localization with exponentially decaying tails , after sufficiently long propagation distances . the reported observation conforms with intuition , because an incoherent wave can be thought of as a superposition of coherent modes with stochastically varying coefficients . because each coherent mode is expected to undergo localization , the entire beam should localize as well . however , localization is delayed by incoherence compared with the case of a coherent in - coupling beam : the more incoherent the wave is , the longer it diffusively spreads while propagating in the medium . for an optical beam propagating in a 1d disordered optical lattice described by the propagation eqs . [ eq : coupledmode1 ] or [ eq : bpm ] , the state of the coherence of the beam is determined by the mutual coherence function @xmath216 @xcite , where @xmath217 is the ensemble average , and @xmath218 is the stochastic field . for the disordered waveguide array of figure [ fig:1d - array - fiber ] , @xmath219 and @xmath220 are the waveguide array indexes . for example , the mutual coherence function of the partially coherent extension of the input gaussian beam of eq . [ eq : exponentialinput ] can be written as @xmath221 \exp[-\dfrac{(j - k)^2}{{\cal s}^2_0}],\quad j , k=1,\cdots , n , \label{eq : incoherentinput}\ ] ] where @xmath112 is the index of the waveguide at the center of the input gaussian beam ; and @xmath222 and @xmath223 are the spatial and the coherence widths of the beam , respectively . we note that the optical intensity over the coupled waveguide array is given by the diagonal element of the mutual coherence function @xmath224 ( @xmath12 is the waveguide index ) @xcite . using eq . [ eq : coupledmode1 ] and following a similar procedure that resulted in eq . [ eq : coupledsol2 ] , we conclude that the propagation of the mutual coherence function is described by @xmath225,\end{aligned}\ ] ] where the propagation coefficients are determined at @xmath67 according to @xmath226 let s consider the disordered coupled waveguide array in figure [ fig : randomcoupledwaveguides1d ] , for @xmath68 , @xmath69 , @xmath70 , @xmath227 $ ] , and @xmath228 , according to the notation used in the discussion following eq . [ eq : coupledmode1 ] . the input light is assumed to be a partially coherent beam defined by eq . [ eq : incoherentinput ] , with @xmath229 and @xmath230 . the average beam width as a function of the propagation distance is plotted in figure [ fig : beam - width - coherence ] for the case of highly coherent @xmath231 in red , semi - coherent @xmath232 in blue , and near - incoherent @xmath233 in cyan . the beam widths are averaged over 100 independent simulations . the results agree with the observations of apeta , _ et al _ @xcite , where localization happens for all three cases , but is strongest for the more coherent input beam . in red , semi - coherent @xmath232 in blue , and near - incoherent @xmath233 in cyan . @xmath229 and @xmath230 have been used and the beam widths are averaged over 100 independent simulations . the disordered coupled waveguide array is defined by @xmath68 , @xmath69 , @xmath70 , @xmath227$],,width=240 ] the propagation of non - classical light in disordered waveguides has also been a subject of interest over the past few years . for example , the propagation of a fock number state , coherent input state , as well as a squeezed state in a disordered waveguide array was studied by thompson _ et al _ @xcite . the quantum mechanical analogue of the disordered tight - binding model used for the propagation of light in a disordered waveguide array in eq . [ eq : coupledmode1 ] can be heuristically constructed by simply elevating the amplitudes of the optical field @xmath234 to photon annihilation operators @xmath235 in waveguide @xmath12 , subject to the following commutation relations @xmath236=0,\quad \left[{\hat a}_{j},{\hat a}^\dagger_{k}\right]=\delta_{kj},\quad j , k=1,\cdots , n . \label{eq : commutation}\end{aligned}\ ] ] the quantum mechanical analogue of eq . [ eq : coupledmode1 ] is the `` linear '' heisenberg equation for the evolution of @xmath237 and @xmath238 and the creation and annihilation operators at the output ports can be obtained from those at the input ports using the green s function of the disordered medium as @xmath239 the green s function , in the language of the disordered couple waveguide array and eq . [ eq : coupledsol2 ] , can be expressed as @xmath240.\end{aligned}\ ] ] using eq . [ eq : greens1 ] , the input first - order quantum coherence function can be mapped to that at the output according to @xmath241 where @xmath242 . the averaging on the green s functions are carried over multiple realizations of _ disordered waveguides _ hence the subscript `` dw '' in @xmath243 . equation [ eq : quantumcoherence1 ] is exactly what one would write for the propagation of the first - order classical coherence function as well . the quantum effects enter through the first order coherence function , which is defined by the density matrix @xmath244 as @xmath245 for example , if the input state to each waveguide is a glauber coherent state , and the quantum state of the input light is defined by @xmath246 , we will have @xmath247 alternatively , if the input state consists of fock number states defined by @xmath248 , we will have @xmath249 the average output intensity in port @xmath12 , @xmath250 is the diagonal element of the output first - order coherence function . for the case of the coherent state above , we will have @xmath251 this formula is identical to the intensity we would expect to obtain at port @xmath12 when classical optical field of amplitude @xmath252 are coupled at the input port @xmath2 . for the fock number states , the intensity is given by @xmath253 the special case where the photons are coupled into a single input port , say middle port @xmath33 , results from eqs . [ eq : intalpha1 ] and [ eq : intfock1 ] can be simplified as @xmath254 equation [ eq : intalphafock2 ] is notable it shows that regardless of the statistics of the input photons , if it is injected to a single port we will observe the same localization behavior . in fact , localization is solely dictated by the coherent averaging of the green s function of the disordered waveguide @xmath255 . this is not the case when light is coupled into multiple ports , as can be observed in the difference between eq . [ eq : intalpha1 ] and eq . [ eq : intfock1 ] : the coherent interference of the input coherent states in eq . [ eq : intalpha1 ] certainly affects the localization , while eq . [ eq : intfock1 ] merely consists of multiple copies of eq . [ eq : intalphafock2 ] assembled incoherently together because of the total ambiguity in the phase information in the fock number states . the bottom line is that the localization is dictated by both the input photon statistics as well as the disordered waveguides , and their interaction can be complicated or simple depending on the situation . in figure [ fig : g1-coherence ] we plot the @xmath256 at @xmath257 for the disordered coupled waveguide array in figure [ fig : randomcoupledwaveguides1d ] , for @xmath68 , @xmath69 , @xmath70 , @xmath227 $ ] . we also assume that @xmath258 ; therefore , the input light is assumed to be coupled only to the middle waveguide as discussed in eq . [ eq : intalphafock2 ] . @xmath256 in figure [ fig : g1-coherence ] is plotted as a function of @xmath12 , which labels the output waveguide number . the dashed blue line represents a single simulation , while the solid red line shows the result of averaging @xmath255 over 1000 simulations . the exponential localization is clearly observed in figure [ fig : g1-coherence ] . for @xmath258 is plotted as a function of the output waveguide number for a disordered coupled waveguide array . the dashed blue line represents a single simulations , while the solid red line shows the result of averaging @xmath255 over 1000 simulations.,width=240 ] higher order quantum coherence functions can also be studied in disordered waveguides . in particular , the second - order quantum coherence function gives us information about two photon correlations and hanbury brown twiss ( hbt ) effect @xcite . the second - order quantum coherence function at the output of the disordered waveguide can be written as @xmath259 we are interested in two particular scenarios . the first scenario is when the two photons are coupled to the same waveguide , say middle port @xmath33 , and the input state is given by @xmath260 . in this case , the second - order quantum coherence function is given by @xmath261 the second scenario is when the two photons are coupled to two different waveguides , say ports @xmath33 and @xmath262 , and the input state is given by @xmath263 . in this case , the second - order quantum coherence function is given by @xmath264 in either scenario , we have indistinguishable photons that co - propagate in the lattice , and are subject to the `` averaged '' correlations of the propagators of the disordered lattice , resulting in rather exotic quantum statistics behavior , as shown for example in ref . @xcite . in the following , we will adopt the same disordered waveguide array as that of figure [ fig : g1-coherence ] and illustrate the behavior of the second - order quantum coherence function . of eq . [ fig : g2same ] is plotted for @xmath265 ( @xmath266 waveguides in the lattice ) as a function of the output waveguide numbers @xmath12 and @xmath267 for a disorder - free coupled waveguide array for ( a ) @xmath268 , and ( b ) @xmath257.,width=432 ] in figure [ fig : g2-same - waveguide - no - disorder ] we plot @xmath269 of eq . [ fig : g2same ] for the case when the two photons are coupled to the middle waveguide @xmath265 ( we use @xmath266 waveguides for easier simulation ) , in the absence of any disorder ( @xmath270 ) , where @xmath268 and @xmath257 is assumed in figures [ fig : g2-same - waveguide - no - disorder](a ) and ( b ) , respectively . the plots indicate the ballistic expansion of the photons , with larger probability to find the photons in the edge waveguides ( figure [ fig : g2-same - waveguide - no - disorder](a ) ) , and also reflections from the boundary for longer propagation ( figures [ fig : g2-same - waveguide - no - disorder](b ) ) , resulting in strong interference effects . in figure [ fig : g2-same - waveguide - with - disorder ] we present plots of the same scenario of figure [ fig : g2-same - waveguide - with - disorder](a ) , except in the presence of disorder , averaged over 1000 independent simulations . we have used @xmath271 $ ] for figure [ fig : g2-same - waveguide - with - disorder](a ) and @xmath272 $ ] for figure [ fig : g2-same - waveguide - with - disorder](b ) . we have also presented both density and 3d plots in each case for easier comparison . note that we have taken @xmath268 in order to prevent reflections from the outer waveguides ; such reflections complicate the analysis and are not observed as long as there is a sufficiently large number of coupled waveguides in the array for a given propagation distance . of eq . [ fig : g2same ] is plotted for @xmath265 ( @xmath266 waveguides in the lattice ) as a function of the output waveguide numbers @xmath12 and @xmath267 for a disordered coupled waveguide array for ( a ) @xmath271 $ ] and ( b ) @xmath272 $ ] . both density and 3d plots are presented in each case for easier comparison.,title="fig:",width=432 ] of eq . [ fig : g2same ] is plotted for @xmath265 ( @xmath266 waveguides in the lattice ) as a function of the output waveguide numbers @xmath12 and @xmath267 for a disordered coupled waveguide array for ( a ) @xmath271 $ ] and ( b ) @xmath272 $ ] . both density and 3d plots are presented in each case for easier comparison.,title="fig:",width=432 ] in the presence of low disorder as in figure [ fig : g2-same - waveguide - with - disorder](a ) , it is possible for both photons to remain localized ( center of the plot ) , one remains localized and one propagates freely ( edge - centers of the plot ) , or both freely propagate ( corners of the plot ) , in decreasing order of probability . once the disorder is increased as in figure [ fig : g2-same - waveguide - with - disorder](b ) , it is almost only possible to have localized photons with small non - zero chance of having one localized and one freely propagating photon . once the disorder is further increased e.g. , @xmath273 $ ] ( not plotted here ) , the only reasonable non - zero probability is to have both photons highly localized . we recall that the disorder - free case in figure [ fig : g2-same - waveguide - no - disorder](a ) favors _ both freely propagating _ to _ one localized and one freely propagating _ to _ both localized _ , respectively , and the presence of disorder reverses this behavior . of eq . [ fig : g2different ] is plotted for ( a ) @xmath274 , @xmath275 ( @xmath266 waveguides in the lattice ) , and ( b ) @xmath274 , @xmath276 , as a function of the output waveguide numbers @xmath12 and @xmath267 for a disorder - free @xmath266 coupled waveguide array and @xmath268.,width=432 ] in figure [ fig : g2-more - separation - no - disorder ] we plot @xmath269 of eq . [ fig : g2different ] for the case where the photons are coupled to different waveguides : @xmath274 and @xmath275 in figure [ fig : g2-more - separation - no - disorder](a ) and @xmath274 and @xmath276 in figure [ fig : g2-more - separation - no - disorder](b ) . either case is in the absence of any disorder ( @xmath270 ) for @xmath268 . the plots indicate the ballistic expansion of the photons , with larger probability to find the photons in the edge waveguides ( figure [ fig : g2-same - waveguide - no - disorder](a ) ) , and also reflections from the boundary for longer propagation ( figures [ fig : g2-same - waveguide - no - disorder](b ) ) , resulting in strong interference effects . when the two photons are coupled to neighboring waveguides as in figure [ fig : g2-more - separation - no - disorder](a ) , the most likely scenario is that the two photons ballistically propagate to the same edge of the lattice , and it is impossible to find the two photons in the opposite edges of the lattice ; however , there is a small yet non - vanishing probability to find one photon remaining in the center waveguide and the other in the edge lattice . when the two in - coupled photons are separated by one waveguide as in figure [ fig : g2-more - separation - no - disorder](b ) , the most likely scenario is that the two photons ballistically propagate to either edge of the lattice , and it is impossible to find one photon remaining in the center waveguide and the other in the edge . of eq . [ fig : g2different ] is plotted for @xmath265 and @xmath276 ( @xmath266 waveguides in the lattice ) as a function of the output waveguide numbers @xmath12 and @xmath267 for a disordered coupled waveguide array for ( a ) @xmath271 $ ] and ( b ) @xmath272 $ ] . both density and 3d plots are presented in each case for easier comparison.,title="fig:",width=432 ] of eq . [ fig : g2different ] is plotted for @xmath265 and @xmath276 ( @xmath266 waveguides in the lattice ) as a function of the output waveguide numbers @xmath12 and @xmath267 for a disordered coupled waveguide array for ( a ) @xmath271 $ ] and ( b ) @xmath272 $ ] . both density and 3d plots are presented in each case for easier comparison.,title="fig:",width=432 ] in figure [ fig : g2-neighbor - waveguide - with - disorder ] we present plots of the same scenario of figure [ fig : g2-more - separation - no - disorder](a ) , except in the presence of disorder , averaged over 1000 independent simulations . we have used @xmath271 $ ] for figure [ fig : g2-neighbor - waveguide - with - disorder](a ) and @xmath272 $ ] for figure [ fig : g2-neighbor - waveguide - with - disorder](b ) . we have also presented both density and 3d plots in each case for easier comparison . we note that increasing the level of disorder compared with the disorder - free case in figure [ fig : g2-more - separation - no - disorder](a ) decreases the probability of ballistic co - propagation of photons to the same edge , while increasing the probability of the localization of both photons in the center waveguide . therefore , the disorder - free case of figure [ fig : g2-more - separation - no - disorder](a ) favors _ both freely propagating to the same edge _ to _ one localized and one freely propagating _ to _ both localized _ , respectively , and the presence of disorder reverses this behavior . we encourage the interested reader to consult ref . @xcite for a more detailed account of the behavior of @xmath269 , upon which much of the preceding discussion is based . the above discussion only covers the case of off - diagonal disorder in a disordered coupled array waveguide . the comparison with the case of the diagonal disorder is performed in ref . it is shown that the evolution and localization of the photon density ( optical intensity ) is similar in the two cases of diagonal and off - diagonal disorder , as expected . however , the intensity correlation ( photon density - density correlation or @xmath269 ) carries a distinct signature of the type of disorder . the propagation of an entangled - photon pairs in a disordered waveguide array has also been explored in ref . it has been shown that while neither photon is localized , the two - photon separation in coincidence space is : this behavior is called anderson colocalization . the increase in entanglement is accompanied by a gradual evolution from anderson - localization to anderson colocalization . * highlights : * * if all the eigenmodes of a disordered waveguide are exponentially localized , any partially incoherent beam exhibits localization with exponentially decaying tails , after a sufficiently long propagation distances . * the more incoherent the in - coupling beam is , the longer it diffusively spreads while propagating in the medium . therefore , localization is delayed by incoherence compared with the case of a coherent in - coupling beam . * the observed localization behavior in a disordered waveguide is affected by the quantum statistics of the in - coupling photons . the effect can be mainly observed in the second - order quantum coherence function , which gives us information about two photon correlations . a tutorial review of the transverse anderson localization of light in disordered waveguides is presented . in addition to the intriguing fundamental science behind anderson localization , there are potential applications for light interaction with disordered dielectric systems . examples such as spatially - multiplexed beam delivery and high - quality image transport were explored . there are still plenty of open questions in both linear and nonlinear dynamics of disordered systems that have yet to be answered . there has been a recent surge of interest among the optics community in anderson localization , and it would have been impossible to cover them all even briefly in this survey . this brief tutorial review is intended to demystify anderson localization for the broader audience and provide sufficient background to newcomers in this field to follow up with studying other more in - depth reviews and research articles . the author acknowledges support by grant number 1029547 from the national science foundation . the author is also grateful for the opportunity to collaborate on disordered optical fibers with salman karbasi , ryan frazier , craig mirr , karl koch , john ballato , thomas hawkins , claudio conti , and marco leonetti , over the past four years . j. billy , v. josse , z. zuo , a. bernard , b. hambrecht , p. lugan , d. clment , l. sanchez - palencia , p. bouyer , and a. aspect , `` direct observation of anderson localization of matter waves in a controlled disorder , '' nature * 453 * , 891894 ( 2008 ) . s. karbasi , r. j. frazier , k. w. koch , t. hawkins , j. ballato , and a. mafi . `` image transport through a disordered optical fibre mediated by transverse anderson localization , '' nature communications * 5 * , 3362 ( 2014 ) . a. szameit , y. v. kartashov , p. zeil , f. dreisow , m. heinrich , r. keil , s. nolte , a. tunnermann , v. a. vysloukh , and l. torner , `` wave localization at the boundary of disordered photonic lattices , '' , 11721174 ( 2010 ) . s. ghosh , r. k. varshney , b. p. pal , and g. monnom , `` a bragg - like chirped clad all - solid microstructured optical fiber with ultra - wide bandwidth for short pulse delivery and pulse reshaping , '' optical and quantum electronics , * 42 * , 114 ( 2010 ) . l. martin , g. di giuseppe , a. perez - leija , r. keil , f. dreisow , m. heinrich , s. nolte , a. szameit , a. f. abouraddy , d. n. christodoulides , and b. e. a. saleh , `` anderson localization in optical waveguide arrays with off - diagonal coupling disorder , '' , 1363613646 ( 2011 ) . s. ghosh , n. d. psaila , r. r. thomson , b. p. pal , r. k. varshney , and a. k. kar , `` ultrafast laser inscribed waveguide lattice in glass for direct observation of transverse localization of light , '' , 101102 ( 2012 ) . y. lahini , a. avidan , f. pozzi , m. sorel , r. morandotti , d. n. christodoulides , and y. silberberg , `` anderson localization and nonlinearity in one - dimensional disordered photonic lattices , '' , 013906 ( 2008 ) . s. karbasi , r. j. frazier , c. r. mirr , k. w. koch , and a. mafi , `` fabrication and characterization of disordered polymer optical fibers for transverse anderson localization of light , '' j. vis . exp . * 77 * ( 2013 ) . s. karbasi , c. r. mirr , ry . j. frazier , p. g. yarandi , k. w. koch , and a. mafi , `` detailed investigation of the impact of the fiber design parameters on the transverse anderson localization of light in disordered optical fibers , '' * 20 * , 1869218706 ( 2012 ) . m. boguslawski , s. brake , j. armijo , f. diebel , p. rose , and c. denz , `` analysis of transverse anderson localization in refractive index structures with customized random potential , '' , 3171331724 ( 2013 ) . s. ghosh , g. p. agrawal , b. p. pal , and r. k. varshney , `` localization of light in evanescently coupled disordered waveguide lattices : dependence on the input beam profile , '' opt . commun . * 284 * , 201206 ( 2011 ) . s. ghosh , b. p. pal , r. k. varshney , and g. p. agrawal , `` transverse localization of light and its dependence on the phase - front curvature of the input beam in a disordered optical waveguide lattice , '' j. opt . ( iop ) * 14 * , 075701:15 ( 2012 ) . arash mafi is an associate professor of physics and astronomy and a member of the center for high technology materials ( chtm ) at the university of new mexico . he received his undergraduate degree in physics from sharif university of technology in 1995 , and his ph.d . degrees in physics from the ohio state university in 2001 . following his postdoctoral appointments at the university of arizona in physics and the optical sciences center , he joined corning inc . in 2005 as a senior research scientist working on optical fibers and liquid crystal displays . he moved to the university of wisconsin - milwaukee in 2008 , where he was an associate professor of electrical engineering and computer science , before joining the university of new mexico in 2014 . he was a recipient of the early career development ( career ) award from the national science foundation in 2013 . his research interests include quantum and nonlinear behavior of optical waveguides , and light propagation in disordered media with applications in optical image transport .	this tutorial review gives an overview of the transverse anderson localization of light in one and two transverse dimensions . a pedagogical approach is followed throughout the presentation , where many aspects of localization are illustrated by means of a few simple models . the tutorial starts with some basic aspects of random matrix theory , and light propagation through and reflection from a random stack of dielectric slabs . transverse anderson localization of light in one- and two - dimensional coupled waveguide arrays is subsequently established and discussed . recent experimental observations of localization and image transport in disordered optical fibers are discussed . more advanced topics , such as hyper - transport in longitudinally varying disordered waveguides , the impact of nonlinearity , and propagation of partially coherent and quantum light , are also examined .
this article is based on a covariant causal set approach to discrete quantum gravity @xcite . for background and more details , we refer the reader to @xcite . we call a finite partially ordered set a _ causet _ and interpret the order @xmath4 in a causet @xmath1 to mean that @xmath5 is in the causal future of @xmath6 . we denote the cardinality of a causet @xmath1 by @xmath7 . if @xmath1 and @xmath8 are causets with @xmath9 then @xmath1 _ produces _ @xmath8 ( written @xmath10 ) if @xmath8 is obtained from @xmath1 by adjoining a single maximal element @xmath6 to @xmath1 . if @xmath10 we call @xmath8 an _ offspring _ of @xmath1 . a _ labeling _ for a causet @xmath1 is a bijection @xmath11 such that @xmath12 with @xmath4 implies that @xmath13 . if @xmath14 is labeling for @xmath1 , we call @xmath15 an @xmath14-_causet_. two @xmath14-causets @xmath1 and @xmath8 are _ isomorphic _ if there exists a bijection @xmath16 such that @xmath4 in @xmath1 if and only if @xmath17 in @xmath8 and @xmath18}=\ell ( a)$ ] for every @xmath19 . isomorphic @xmath14-causets are considered identical as @xmath14-causets . we say that a causet is _ covariant _ if it has a unique labeling ( up to @xmath14-causet isomorphism ) and call a covariant causet a @xmath0-_causet_. we denote the set of a @xmath0-causets with cardinality @xmath20 by @xmath21 and the set of all @xmath0-causets by @xmath22 . it is easy to show that any @xmath23 with @xmath24 has a unique producer and that any @xmath23 has precisely two offspring @xcite . it follows that @xmath25 , @xmath26 . two elements @xmath12 are _ comparable _ if @xmath4 or @xmath27 . we say that @xmath6 is a _ parent _ of @xmath5 and @xmath5 is a _ child _ of @xmath6 if @xmath4 and there is no @xmath28 with @xmath29 . path from _ @xmath6 _ to _ @xmath5 in @xmath1 is a sequence @xmath30 , @xmath31 , @xmath32 where @xmath33 is a parent of @xmath34 , @xmath35 . the _ height _ @xmath36 of @xmath19 is the cardinality minus one of the longest path in @xmath1 that ends with @xmath6 . if there are no such paths , then @xmath37 by convention . it is shown in @xcite that a causet @xmath1 is covariant if and only if @xmath12 are comparable whenever @xmath6 and @xmath5 have different heights . if @xmath23 we call the sets @xmath38 _ shells _ and the sequence of integers @xmath39 , @xmath3 , is the _ shell sequence_. a @xmath0-causet is uniquely determined by its shell sequence and we think of @xmath2 as describing the `` shape '' or geometry of @xmath1 @xcite . the tree @xmath40 can be thought of as a growth model and an @xmath41 is a possible universe among many universes at step ( time ) @xmath20 . an instantaneous universe @xmath41 grows one element at a time in one of two ways . to be specific , if @xmath41 has shell sequence @xmath42 , then @xmath1 will grow to one of its two offspring @xmath43 , @xmath44 , where @xmath45 and @xmath46 have shell sequences @xmath47 respectively . in the microscopic picture , we view a @xmath0-causet @xmath1 as a framework or scaffolding for a possible universe . the vertices of @xmath1 represent small cells that can be empty or occupied by a particle . the shell sequence that determines @xmath1 gives the geometry of the framework . a labeling of @xmath1 gives a `` birth order '' for the elements of @xmath1 and since @xmath1 is covariant , its structure is independent of birth order . in the previous section , @xmath0-causets were employed to describe possible universes within a multiverse . in the growth model @xmath48 a universe grew one element at each time step which created a history for a possible evolved universe . each history began with one vertex and its early stages contained only a few vertices . it is reasonable to assume that those early stages were inhabited by various elementary particles.we therefore propose that an elementary particle is described by a small @xmath0-causet . in this section , instead of thinking of a causet as a possible instantaneous universe , we will think of it as an elementary particle . a _ link _ in a @xmath0-causet @xmath1 is a set @xmath49 where @xmath6 is a parent of @xmath5 ( order is immaterial ) . a @xmath0-causet @xmath1 together with its links can be thought of as a graph . a _ graph path _ from vertex @xmath0 to vertex @xmath50 in @xmath1 is a sequence of distinct incident links @xmath51 . a @xmath0-causet @xmath1 is 2-_connected _ if for any two comparable vertices @xmath52 of @xmath1 there exist at least two disjoint graph paths from @xmath6 to @xmath5 . if @xmath19 with @xmath1 a @xmath0-causet , then the pairs @xmath53 are called _ preons_. we think of a preon as an indivisible particle with electric charge @xmath54 , @xmath55 or @xmath56 . a @xmath0-causet is _ occupied _ if each of its vertices is replaced by a preon . we will classify certain occupied @xmath0-causets as elementary particles . for such @xmath0-causets the vertices will represent preons and the edges ( links ) will represent forces between pairs of preons . a photon is represented by an occupied @xmath0-causet having an even number of vertices with shell sequence @xmath57 and the preons are alternatively charged @xmath58 . the reason a photon has this form is that a photon should move along a geodesic in its containing universe and we have seen in a previous study that geodesics have this structure @xcite . we consider a photon to be an elementary particle . the other elementary particles will be represented by certain occupied 2-connected @xmath0-causets . these elementary particles and their combinations will correspond to matter - energy . occupied @xmath0-causets that are not 2-connected ( 2-disconnected ) correspond to dark matter - energy . the following table lists the number of @xmath0-causets with @xmath20 vertices that are 2-connected and 2-disconnected , @xmath59 . this indicates that dark matter - energy dominates over matter - energy . in this article we only consider matter - energy ; specifically , elementary particles and their combinations . .25 in [ cols="^,^,^ " , ] we now classify the elementary particles according to their representative occupied @xmath0-causets . this is mainly guesswork but we present some evidence for this scheme . except for photons , we postulate that because elementary particles are forms of matter - energy , they are described by 2-connected @xmath0-causets . the reason for this is that 2-connected @xmath0-causets have multiple graph paths so there are strong connections between their preons . this enables them to stay together and be more stable . photons are the exception but their preons are charged so they have electromagnetic attraction while the preons of dark matter - energy presumably are electrically neutral . we see from table 1 that there are no 2-connected @xmath0-causets with less than four vertices . the only 2-connected @xmath0-causets with four vertices are the @xmath0-causets with shell sequences @xmath60 and @xmath61 . we postulate that the first corresponds to the electron @xmath62 and the second to the up quark @xmath63 . these are illustrated in figure 1 . in figure 1 , the vertices are preons with the labeled charges . of course , the positron @xmath64 and the up anti - quark @xmath65 are like @xmath62 and @xmath63 except the preon charges are reversed . also , the electron neutrino @xmath66 is like the electron except the preon charges are zero . we propose that in the next generations , the muon and tauon leptons have the shell sequences @xmath67 , @xmath68 . the preons in @xmath69 and @xmath70 have charges that give the correct total charge of @xmath71 . as with @xmath62 and @xmath63 , these charges are not uniquely distributed . as before , we define the antiparticles @xmath72 , @xmath73 to be the same as @xmath69 , @xmath74 with the preon charges reversed and the neutrinos @xmath75 , @xmath76 to be the same as @xmath69 , @xmath70 with preon charges 0 . finally , we define the quarks @xmath77 as having the shell sequences , @xmath78 , @xmath79 , @xmath80 , @xmath81 , @xmath82 . the preon charges are distributed to give the usual total charges @xmath56 , @xmath56 , @xmath83 , @xmath56 , @xmath83 , respectively . the antiquarks @xmath84 , @xmath85 , @xmath86 , @xmath87 , @xmath88 again have reverse preon charges . this completes our list of elementary particles . we shall not consider the higg s boson and the weak and strong force mediating bosons here . the mesons are formed as usual from quark - antiquark pairs and the baryons are formed from quark triplets . these will be considered in the next section . a vertex of a @xmath0-causet is _ lonesome _ if it is the only vertex in its shell in @xmath1 . for example , @xmath62 has shell sequence @xmath60 so @xmath62 has two lonesome vertices . in general , @xmath1 contains lonesome vertices if and only if @xmath1 has at least one 1 in its shell sequence . as another example , the up quark @xmath63 with shell sequence @xmath61 has no lonesome vertices . we say that a link is _ weak _ if at least one of its two vertices is lonesome . otherwise , a link is _ strong_. we associate a weak link with the weak force and a strong link with the strong force . notice that a photon has only weak links . since photons mediate the electromagnetic force , we see that this force is described by a sequence of an even number of weak links . in a sense , this gives a unification of these three forces of nature . finally , the links are equivalent to the causal order which determines the geometry of @xmath0-causets . as we have described in @xcite , this geometry is responsible for the force of gravity . continuing this analysis , we see that all the leptons @xmath62 , @xmath69 , @xmath70 , their antiparticles and associated neutrinos all contain weak links . except for @xmath62 , @xmath64 , @xmath66 they also contain strong links . moreover , all the quarks have only strong links . we now present a mass formula for the elementary particles and their compositions . recalling that the @xmath20th bohr orbit has energy proportional to @xmath89 , we postulate that the part of the mass of a preon in the @xmath20th shell due to the weak force is also proportional to @xmath89 . we next postulate that the part of the mass of a preon in the @xmath20th shell due to the strong force is proportional to @xmath90 . moreover , we assume that even if a particle has only strong links there are weak forces involved . to be precise , we make the following postulate where masses are in mev . * mass formula postulate * : let @xmath91 be a lepton , quark or hadron with shell sequence @xmath92 . if @xmath91 has only weak links the _ weak - strong mass _ of @xmath91 is @xmath93 and if @xmath91 has strong links the _ weak - strong mass _ of @xmath91 is @xmath94 of course , we do not expect @xmath95 to give the exact experimental mass because the electromagnetic force is not included . also , this formula is unit dependent and we might get a little better agreement with experiment if we change the @xmath96 factor slightly and multiply the first summation term in by a factor close to one . we now apply the mass formula to the leptons and quarks . since the electron has shell sequence @xmath60 we have @xmath97 where we have placed the experimental value in parenthesis . since the muon has shell sequence @xmath98 we have @xmath99 the tauon has shell sequence @xmath100 which gives @xmath101 for quarks we have @xmath102 we do not compare these to experimental values because isolated quarks have not been found so such values are not clear .. we now consider some mesons . until a more complete theory is established we are only guided by their quark constituency and their masses . the @xmath103-meson is composed of a @xmath104 pair so we assume it has the nine vertices of @xmath63 and @xmath84 distributed in some order . we propose that @xmath103 has the shell sequence @xmath105 . this gives @xmath106 we propose that the @xmath107-meson has the shell sequence @xmath108 which gives @xmath109 we assume that the @xmath110-meson has shell sequence @xmath111 which gives @xmath112 finally , the @xmath113-meson is composed of an @xmath114 pair so it has 12 vertices and we assume it has shell sequence @xmath115 which gives @xmath116 we next consider some baryons . the neutron @xmath117 is a @xmath118 quark triple so we assume that @xmath117 has the 14 vertices of @xmath119 distributed in some order . we propose that @xmath117 has the shell sequence @xmath120 . this gives @xmath121 the proton @xmath122 is a @xmath123 quark triple which gives 13 vertices and we assume its shell sequence is @xmath124 . this gives @xmath125 the @xmath126-baryon is a @xmath127 quark triple with 16 vertices and we assume its shell sequence is @xmath128 . this gives @xmath129 the @xmath130-baryon is a @xmath131 quark triple with 17 vertices and we assume its shell sequence is @xmath132 . this gives @xmath133 the @xmath134-baryon is a @xmath135 quark triple with 12 vertices and we assume its shell sequence is @xmath136 . this gives @xmath137 the @xmath138-baryon is a @xmath139 quark triple with 15 vertices and we assume its shell sequence is @xmath140 . this gives @xmath141 finally , @xmath142 is an @xmath143 quark triple with 18 vertices and we assume its shell sequence is @xmath144 . this gives @xmath145 the author admits that this is fairly arbitrary and that one can get practically any mass he / she wants by manipulating the shell sequence . nevertheless , it is somewhat striking that the above examples give masses so close to the experimental values . moreover , there is a hint of a pattern in these baryon shell sequences . notice that if there is no @xmath146 quark , then the last two shell numbers add to 6 , if there is one @xmath146 quark , they add to 7 or 8 , if there are two @xmath146 quarks , they add to 9 and if there are 3 @xmath146 quarks , they add to 11 . it is hoped that a strengthening of this theory will point directly to the shell sequence of various particles . we can also continue this investigation by studying atoms . for example , a hydrogen atom should be represented by a proton with the photon - electron `` kite '' of figure 2 attached to one of its highest shell preons . 99 s. gudder , a covariant causal set approach to discrete quantum gravity , arxiv : gr - qc 1311.3912 ( 2013 ) . s. gudder , a unified approach to discrete quantum gravity , arxiv : gr - qc 1403.5338 ( 2014 ) . s. gudder , the universe as a quantum computer , in preparation . j. henson , quantum histories and quantum gravity , arxiv : gr - qc 0901.4009 ( 2009 ) . r. sorkin , causal sets : discrete gravity , arxiv : gr - qc 0309009 ( 2003 ) . s. surya , directions in causal set quantum gravity , arxiv : gr - qc 1103.6272 ( 2011 ) .	in a previous paper , the author introduced a covariant causet ( @xmath0-causet ) approach to discrete quantum gravity . a @xmath0-causet is a finite partially ordered set that is invariant under labeling . the invariant labeling of a @xmath0-causet @xmath1 enables us to uniquely specify @xmath1 by a sequence @xmath2 , @xmath3 , of positive integers called a shell sequence of @xmath1 . a @xmath0-causet @xmath1 describes the microscopic structure of a possible universe at a particular time step . in general , @xmath1 represents one of many universes in a multiverse and @xmath1 grows by a single element at each time step . since early stages of a universe were probably composed of elementary particles , we propose that elementary particles can be described by simple @xmath0-causets . although we do not have a rigorous theory for such a description , we present our guess as to how it might appear . the shell sequence can be applied to find theoretical masses of particles and these seem to approximately agree with known masses . we point out that the causal order provides a unification of the strong and weak forces for elementary particles and also determines the geometry of @xmath0-causets which describes gravity . moreover , it appears that some particles correspond to dark matter - energy .
the outer edge of saturn s b - ring is confined by the strongest resonance in saturn s ring system : the 2:1 inner lindblad resonance ( ilr ) with mimas @xcite . that resonance is expected to impose a variation in the radial position of the edge with azimuthal wavenumber @xmath2 ( i.e. , a saturn - centered ellipse ) , with one minimum locked to mimas @xcite . voyager observations were consistent with those predictions @xcite , implying a radial amplitude of about 75 km , though relatively large residuals indicated additional unresolved structure . a preliminary analysis of images taken early in the cassini mission confirmed the existence of the saturn - centered ellipse , but found mimas lagging the radial minimum by about 28@xmath3 , suggesting either a static lag between the ring response and the direction to the satellite , or a libration of the periapse direction relative to the satellite , or both . we also reported higher - wavenumber radial oscillations and additional time variations that were difficult to characterize , though an @xmath4 pattern did seem to explain part of that variation @xcite . here , we use cassini imaging data sets spanning about 4 years to analyze the b - ring outer edge in greater detail than previously attainable , with the objective of understanding its complicated shape . for the kinematical modeling in this paper , we examined 18 data sets spanning the interval from day 2005 - 174 to day 2009 - 207 , consisting of a total of nearly 2000 cassini narrow - angle camera ( nac ) images ( see table [ tbl : data ] ) , with pixel scales about 10 km or better ( fig . [ fig : maps ] ) , employing the techniques and types of image sequences ( azimuthal scans and ansa - staring movies ) described in our earlier work @xcite . we also examined several smaller data sets taken near 2009 saturn equinox , when the sun elevation was low enough for shadows cast by out - of - plane material to be detectable . [ fig : maps ] shows map - projected mosaics in radius vs. longitude for each data set used in the kinematical modeling . an additional data set , set 64 , is shown as well because it spanned a significant range of co - rotating longitudes , though it was excluded from the kinematical modeling because most of the images did not contain the fiducial feature ( feature number 13 from @xcite ) used for all of the other data sets to determine radial positions . the expected 2-lobed shape is apparent in many of the mosaics in fig . [ fig : maps ] , with additional higher wavenumber variations superimposed . however , the amplitude of the @xmath2 pattern varies , and the pattern is absent , or barely discernible in some data sets ( e.g. , sets 6 , 8 , 9 , and 23 ) . as in almost all previous analyses of planetary rings , including our recent study of the outer edge of the a - ring @xcite , we modeled the ring edge using a sinusoidal model of a streamline to describe each mode whose presence we investigated . after some experimentation with linear combinations of various low - wavenumber modes , we found the best - fitting model to be one comprising 4 normal modes : an @xmath5 moving at a rate appropriate for free precession near the outer b - ring edge , an @xmath2 moving with a pattern speed @xmath6 expected for forcing by the mimas 2:1 ilr ( i.e. , mimas average mean motion during the period of the investigation , see sec . [ sec : m=2 ] ) , a second @xmath2 moving slightly faster than mimas , and an @xmath4 mode . parameters for the adopted solution are given in table [ tbl : elem_m=1224 ] , and fig . [ fig : plots ] plots each component . the large @xmath7/dof value indicates that there is significant additional structure to be characterized that we were unable to model with simple normal modes . however , note that the largest deviations from the four - component model ( colored red and blue in the figure ) tend to comprise radial excursions of narrow ( @xmath030@xmath3 ) azimuthal extent , which are reminiscent of the `` spikes '' pointed out in our recent study of the a - ring outer edge @xcite . for reasons discussed below , those spikes were considered to be independent of the edge modes , and were thus excluded from the kinematical fits . therefore , the quoted statistics do not include those large excursions . we interpret the presence of an @xmath5 normal mode at the outer b - ring edge as an unforced self - excited unstable normal mode or standing density wave . dense rings , like the outer b ring , behave like non - newtonian fluids and the close packing of the particles can lead to collective behaviors similar to granular flow . under these circumstances density waves are expected to become viscously overstable ; in the presence of feedback , as occurs upon double reflection within a resonant cavity , they become non - linear @xcite . very short wavelength structures of order @xmath0 150 to 400 m have been observed in cassini imaging @xcite and occultation data @xcite within the high optical depth ( @xmath9 ) b - ring , and have been interpreted as spontaneously formed , axisymmetric , non - linear waves or `` viscous overstabilities '' in both cases . these axisymmetric modes are just one form of overstability ; non - axisymmetric modes are possible as well and are believed to be responsible for the multi - mode shapes of the narrow rings of both saturn and uranus @xcite . the density waves with the longest wavelengths are those with @xmath5 . however , without feedback and significant amplification , unforced density wave trains , especially of such long wavelength , will damp . the existence of an @xmath5 disturbance at the outer b - ring edge whose radial amplitude ( @xmath10 km ) is comparable to that of the @xmath2 resonant perturbation ( @xmath11 km ; see below ) argues that sufficient feedback and hence amplification must exist for this mode to be observable . amplification requires a resonant cavity within which the wave can reflect and grow . the outer b - ring edge , at @xmath12 km , provides a sharp outer boundary for a resonant cavity ; the inner boundary is formed by the mode s ilr whose location is determined by solving @xmath13 for @xmath14 , where @xmath15 and @xmath16 are the mean motion and apsidal precession rate at semimajor axis @xmath14 respectively . for the @xmath5 speed given in table [ tbl : elem_m=1224 ] , the ilr location is @xmath17 = 117314.92(8 ) km . the resonant cavity is consequently 253(4 ) km wide . if the wavelength of an @xmath5 disturbance propagating away from a lindblad resonance is known , we can estimate the surface mass density required to keep the wave from shearing due to differential precession . in the general formulation for the behavior of linear density waves driven at a resonance in a ring of surface mass density @xmath18 , the first wavelength @xmath19 is given by @xcite : @xmath20 where @xmath21 , the fractional distance from an ilr location @xmath17 is equal to @xmath22 and @xcite : @xmath23 ( m - 1 ) % + \frac{21}{2 } j_2 \left(\frac{r_p}{a_l}\right)^2\right\}. % \label{eq : cp_x}\ ] ] also , @xmath24 is the universal constant of gravitation , @xmath25 is saturn s second zonal gravity harmonic and @xmath26 is saturn s reference radius . if the ring edge at @xmath27 provides a fixed outer boundary to the cavity , imposing a node at that radial location then , for the case in which no radial nodes exist within the cavity , 1/2 of the first wavelength of the density wave must fit exactly within the cavity . consequently , @xmath28 must equal the cavity width @xmath29 , where @xmath30 = 1/2 . however , the ring edge can not be fixed because it is not circular , its shape being perturbed by these non - axisymmetric density waves , including the one created by the mimas 2:1 resonance . therefore the condition for a node at the ring edge does not exist and the cavity may actually support a longer wave , in which case @xmath30 would be less than 1/2 . we solve for the wavelength at the location where @xmath31 , taking @xmath32 , to obtain : @xmath33 ( m - 1 ) % + \frac{21}{2 } j_2 \left(\frac{r_p}{a_l}\right)^2\right\ } } , % \label{eq : cp_b}\ ] ] where @xmath34 is saturn s mass . setting @xmath28 = 253(4 ) km yields a lower bound on the average surface mass density across this region of @xmath18 = 43.5(7 ) g @xmath1 , for @xmath30 = 1/2 . the cassini radio science optical depth for this putative resonant cavity is @xmath35 1.2 for the outermost @xmath36 km , and @xmath35 0.6 interior to that ( fig . 13.15 in @xcite ) . regions with @xmath35 0.8 in the a - ring have been found to have @xmath37 40 g @xmath1 @xcite , comparable to our lower bound for the b ring . if @xmath30 is equal to , say , 1/4 because the edge is not fixed , eq . [ eq : cp_b ] yields a surface density of 87.0(6 ) g @xmath1 . as support for the presence of an @xmath5 standing wave in the outer few hundred km of the b - ring , we note that this is one of the few places in saturn s rings where multiple occultation profiles diverge significantly from one another ( see fig . 11 of @xcite ) , as would be expected if a non - axisymmetric structure like an @xmath5 standing wave were present . in a previous analysis of the inner cassini division @xcite , we noted that a wavenumber-3 pattern moving at @xmath0 505.5 @xmath3/day provided a good description for the residuals in the data sets that we have labeled 3 and 5 in this work . here , we find that such a mode , moving at @xmath0 507.7@xmath3/day , is consistent with the entire set of observations , as shown in table [ tbl : elem_m=1224 ] and fig . [ fig : plots ] . again , we interpret this pattern as an unstable free oscillation , whose best - fit pattern speed implies an ilr at 117544.2(4 ) km . the resonant cavity within which the @xmath4 mode resides is therefore @xmath39 km wide . with @xmath30 = 1/2 , we calculate from eq . [ eq : cp_b ] a lower bound for the surface mass density of @xmath40 g @xmath1 ; for @xmath30 = 1/4 , the surface density is 100(11 ) g @xmath1 . both values are consistent with those derived for the @xmath5 free mode . to more clearly illustrate the time variation in the @xmath2 shape , we show independent fits to the total @xmath2 pattern for each data set in fig . [ fig : m=2 ] . mimas longitude at epoch for each panel was obtained using the jpl ( jet propulsion laboratory ) naif ( navigation and ancillary information facility ) kernel appropriate for the epoch represented by that panel . in those fits , the parameters for the adopted @xmath5 and @xmath4 patterns from table [ tbl : elem_m=1224 ] were accounted for and fixed , but no attempt was made to remove the large radial anomalies , as this part of the discussion concerns only the qualitative character of the variation . in the figure , the amplitude variation is obvious and appears systematic , starting at about 50 km in mid-2005 , decreasing to nearly zero in late 2006 , then increasing back to about 70 km in early 2009 . moreover , the orientation of the pattern relative to mimas varies systematically , with mimas initially lagging the pattern by @xmath0 30@xmath3 , and eventually winding up leading the pattern by @xmath0 10@xmath3 . in our adopted solution , the variation in the amplitude and orientation of the total @xmath2 pattern seen in fig . [ fig : m=2 ] arises from the interference of two @xmath2 patterns of comparable amplitude moving at slightly different speeds . one mode is a response to the mimas perturbation ; the other is a free unstable mode like the @xmath5 and @xmath4 waves described in the previous two sections . we held the speed of the forced @xmath2 pattern fixed at mimas average mean motion for the interval during which the observations were acquired . according to the best available fit to mimas speed as a function of time ( r.a . jacobson , pers . comm . ) , which is dominated by its resonance with tethys , its average mean motion should have been about 381.98401@xmath3/day during that time . however , the residuals to that fit indicate that mimas true speed was actually slower than that by about 0.0005@xmath3/day . the standard deviation in mimas speed variation during that interval is about @xmath42/day . therefore , we took mimas mean speed to be 381.9835(1)@xmath3/day . for that pattern speed , the location of the mimas 2:1 ilr is @xmath43 = 117555.8(1 ) km . in the presence of collective effects and self - gravity , the half - width of the resonance zone is given by the first wavelength of the density wave generated by the resonance ( see eq . [ eq : cp_a ] ) . for the range of densities computed for the outermost portion of the ring ( sec . [ sec : m=3 ] ) , that half - width is @xmath0 60 90 km . the amplitude expected for the pattern forced by the mimas resonance is @xmath44 , where @xcite : @xmath45 taking @xmath46 , where @xmath47 km is mimas orbital semimajor axis , and noting that the expression involving the laplace coefficient @xmath48 evaluates to @xmath0 2.38 ( see tables 8.1 and 8.5 of @xcite ) , the computed amplitude is @xmath49 37.5 km , differing by about 8% from the value of 34.6 km implied by the forced eccentricity given in table [ tbl : elem_m=1224 ] . the speed of the best - fit free @xmath2 pattern in table [ tbl : elem_m=1224 ] corresponds to an ilr at an orbital radius of 117537.7(2 ) km , about 18 km interior to the mimas resonance . the beat period is @xmath50 = 5.5(1 ) yr , where @xmath51 = 0.0896(7)@xmath3/day is the difference in pattern speeds between the two components . the width of the cavity between the free @xmath2 ilr and the outer b ring edge is @xmath52 km . with @xmath30 = 1/2 , eq . [ eq : cp_b ] yields a surface mass density across the cavity of @xmath53 g @xmath1 ; for @xmath30 = 1/4 , we have @xmath18 = 84(7 ) g @xmath1 . both values are nearly identical to those computed from the @xmath5 mode , and consistent with that for the @xmath4 mode . s1 in the online supplement shows an animation of the ring edge with the radial locations of the free and forced ilrs and the mean edge overlain . in their measurements of the b - ring edge taken from cassini vims ( visual and infrared mapping spectrometer ) data sets , @xcite noticed a similar variation in the @xmath2 pattern to that in fig . [ fig : m=2 ] . however , they interpreted the variation in the context of libration near a resonance instead of as the interference of two similar patterns as we have here . they used the formalism for an isolated particle near resonance , as in @xcite , with resonant argument : @xmath54 where @xmath55 and @xmath56 are the mean longitude and periapse longitude of the particle respectively , and @xmath57 is mimas mean longitude . in that approach , the time - variable total eccentricity @xmath58 is decomposed into free and forced components , @xmath59 and @xmath60 respectively . the particle traces out a circle in a polar coordinate system with @xmath61 as the radial coordinate and the resonant argument @xmath62 as the angular coordinate , as shown in fig . [ fig : hk ] . the phase angle @xmath63 increases at a constant rate @xmath64 , yielding the explicit time variation given by @xcite ) : @xmath65 where @xmath66 is the resonance strength . the epoch @xmath67 is a constant of integration and the libration frequency @xmath68 depends on the forced eccentricity ( i.e. , it depends on the distance between the particle semimajor axis and the location of the exact resonance ) . to determine the explicit time variation in our interpretation , consider the fractional radial excursion for the composite @xmath2 pattern as a function of longitude @xmath55 in a frame rotating with mimas : @xmath69 , \\ \label{eq : bbb}\ ] ] where @xmath60 and @xmath59 are the eccentricities of the forced and free components respectively , and @xmath51 is difference between the pattern speeds of the free and forced components . the curve described by eq . [ eq : bbb ] is a quasi - sinusoid with frequency @xmath70 whose phase can be determined by finding an extremum , i.e. , by solving @xmath71 for @xmath55 . identifying @xmath72 with @xmath73 , and noting that @xmath74 at the minimum , we have : @xmath75 or @xmath76 moreover , from ( [ eq : a ] ) we have : @xmath77 since @xmath57 can be taken to be zero in the rotating frame , and @xmath74 at periapse , we have @xmath78 , or @xmath79 . the total eccentricity @xmath58 is simply the fractional amplitude ( [ eq : bbb ] ) evaluated at @xmath80 , and can be obtained by substituting ( [ eq : ddd ] ) back into ( [ eq : bbb ] ) . the time variation is therefore : @xmath81 eqs . [ eq : faa ] are identical to eqs . [ eq : fa ] , except that the frequency @xmath70 is independent of the distance between the radial locations of the streamline and the resonance , whereas in eqs . [ eq : fa ] @xmath64 is uniquely determined by @xmath60 , which is determined by that distance . indeed , our best - fit frequency of @xmath70 = 0.179(7)@xmath3/day is not consistent with the computed frequency @xmath82 = 0.128(1)@xmath3/day . moreover , we have observed two other free unstable modes near the b - ring edge . consequently , we believe that our interpretation of the observed time variation of the total @xmath2 pattern is more natural than that of @xcite . @xcite suggested that the motion of the @xmath2 pattern on the b - ring edge could be responsible for eccentric features found within various gaps in the cassini division . in particular , it was noted that a model where the oscillation frequency of the @xmath2 pattern about the saturn - mimas line was @xmath0 0.06@xmath3/day was consistent with cassini vims , rss and historical occultation data and was also equal to the difference in pattern speeds between adjacent eccentric features in the cassini division . because that frequency is slower than our result by a factor of nearly 3 , the libration model for the cassini division features may need to be modified for it to be successful . at the central epoch of our fit , jed 2454256.9 , mimas longitude was 10.1(1)@xmath3 , where the quoted uncertainty arises from the uncertainty in mimas mean speed , given above . comparison with table [ tbl : elem_m=1224 ] shows that the forced @xmath2 pattern lagged mimas by 2.9(3)@xmath3 . that lag angle , @xmath83 , is diagnostic of the rate of angular momentum exchange between the ring and the satellite , allowing us to compute the torque exerted by the ring on mimas @xcite : @xmath84 and hence the rate of orbital evolution @xmath85 assuming the torque acts only to change mimas semimajor axis and not its eccentricity . taking @xmath86 and considering the range of mean surface densities 50100 g @xmath1 computed above for the outermost zone , the resulting rate of orbital evolution @xmath87 ranges from @xmath0 0.06 to @xmath0 0.12 cm yr@xmath88 , amounting to about 30006000 km of orbital expansion if @xmath89 were to act continuously over the age of the solar system . balancing the satellite torque against the viscous torque delivering angular momentum to the ring edge , the kinematic viscosity @xmath90 is related to the lag angle by @xcite : @xmath91 which yields a kinematic viscosity @xmath92 20 @xmath93 s@xmath88 for values appropriate for the b - ring outer edge . this value for @xmath90 is on the low end of the various estimates obtained from spiral density waves in the a ring ( see for example fig . 20 of @xcite ) . note , however , that this calculation assumes that the hydrodynamical approximation applies in the outer b ring . in reality , a scalar parameter may be insufficient to represent the relation between viscous stress and deformation in this region@xcite . visual inspection of the thousands of cassini images used in our kinematical analysis revealed distinct azimuthally limited regions in the resonance zone at the outer edge in which the ring appearance was visually disturbed and not azimuthally smooth as in other regions . two of these were most prominent . the most obvious , which we have labeled region a ( see fig . [ fig : maps ] ) , has been seen in several dramatic images taken as saturn was approaching its august 11 , 2009 northern vernal equinox in which vertical structures , the tallest 3.5 km , were clearly identifiable by their shadows ( figs . [ fig : thumbs]b and [ fig : thumbs]c ) . other images taken near equinox revealed the same region , spanning about 10@xmath3 ( or 20000 km ) in azimuth , which was again betrayed by its shadows . in earlier low - sun conditions prior to the shadow - casting geometry , region a was captured in an ansa movie ( set 54 ) as a chevron - shaped structure passing around the ansa of the ring as seen from cassini ( fig . [ fig : thumbs]f ) . however , in yet earlier images taken under high sun , lit - face conditions , it was nt as easily seen , explaining its late detection . instead , region a made itself apparent in high - sun conditions by a rise in brightness , usually in concert with a small discontinuity ( or bump ) in the variation of radius with azimuth , as seen in the maps shown in fig . [ fig : maps ] . fits to positional measurements of region a revealed that it is traveling at a uniform speed ( fig . [ fig : lonplots]a ) . assuming that the fitted motion corresponds to a simple , unperturbed keplerian orbital speed , we can infer the region s precise orbital distance . the fitted speed and inferred orbital distance of region a were found to be 758.986(4 ) @xmath3/day and 117547.6(5 ) km , about 8 km interior to the 2:1 resonance , with a j2000 longitude of 95(1)@xmath3 . region a is therefore situated within the 2:1 resonance zone , in which @xmath35 1 1.2 . at these fairly high optical depths , any moderate differences in optical depths e.g. , @xmath94 are more easily distinguished when the observer is on the unlit side of the rings and the viewing angle is shallow ; the effect is enhanced when the sun angle is also low . this is a result of multiple scattering among the particles ( and therefore diffuse transmission through the rings ) becoming increasingly choked off with increasing optical path lengths ; the contrast between regions having different optical depths is greatest at the high optical depths typical of the resonance zone of the outer b - ring edge . ( compare , for example , figs . 11 and 12 in @xcite ) . hence , even when the sun was not low enough to cause 4-km structures to cast shadows , region a was noticeable on the dark side of the rings , especially at low sun angles , by its unusual chevron - shaped variations in brightness . however , in some images where region a should have been seen based on the illumination and viewing geometry and on its mean motion , it was not . apparently , its position with respect to the @xmath2 pattern is also a factor in is visibility : it appears more pronounced when it is near the steepest radial transition in the @xmath2 pattern ( i.e. , quadrature ) . in all , 17 sightings of region a have been made in cassini images spanning 3 years , from day 2006 - 247 to day 2009 - 233 . ( in a few of these cases , the region s uniform motion was used to predict where it might be at an earlier time ) . another region , labeled b , was also noticed in the resonance zone ( fig . [ fig : maps ] ) . the degree to which the ring is disturbed in region b is not nearly as extreme as in region a. it too was easier to see at low sun angles on the dark side for the same reasons as given above . sometimes , it too was seen as a slight , short - wavelength bump in the outer b - ring edge , and sometimes , when it should have been seen based on its motion and the illumination / viewing geometry , like region a it was not . over all , region b was sighted 10 times in images spanning 3.25 years , from day 2005 - 174 to day 2008 - 312 . its uniform orbital speed , determined through least - squares fitting , was found to be 758.810(4 ) @xmath3/day ; its inferred orbital radius , assuming unperturbed simple keplerian motion , is 117565.7(4 ) km , about 10 km exterior to the 2:1 resonance and very nearly coincident with the ring s outer edge ( fig . [ fig : lonplots]b ) . its j2000 longitude is 155(2)@xmath3 . the fact that these locales are moving at keplerian speeds over the course of several years , and are not traveling at the rotational speeds of any of the components in table 1 implies that they are more likely to be the sites of more - massive - than - average bodies in the rings rather than densely compressed `` shocked '' regions created by the passage of ring material through the overall non - axisymmetric pattern . we have rejected the possibility that the two observed regions are the sites of co - rotation resonances . first , there are no co - rotation resonances in the vicinity of the outer edge of the b ring with known moons big enough to have a pronounced effect . it would be more likely that the populations are in resonance with each other , perhaps similar to the behavior of janus and epimetheus , or perhaps even similar to jupiter and its trojan asteroids . in the latter case , the less massive population would be situated at either the l4 or l5 position of the more massive population ; in the former , it would be librating in a horseshoe orbit around the l3 position . however , a look at how the difference in their longitudinal positions varies with time ( fig . [ fig : lonplots]c ) ( which requires short - range extrapolation of the position of one or the other region since at any given instant , only one region was observed ) shows that libration is not likely . it is apparent from fig . [ fig : lonplots]c that the two regions coincided in azimuth in mid-2006 . while no observations of region a exist prior to this time , observations of region b do . if region b were librating , its mean motion would have changed across the conjunction . instead , its motion from 2005 through 2006 is consistent with uniformly increasing longitude with increasing time . therefore , the simplest interpretation is that these two regions are locales where one or more bodies are located , with each group orbiting independently of the other on two distinctly separate orbits . these bodies are apparently massive enough to alter the ring material around them considerably , though region a s effect is clearly far larger . despite the fact that the cassini equinox imaging captured region b , the only place around the b - ring edge where vertical structures have been seen casting shadows is region a. the shadow orientations here indicate the vertical structures are likely coincident with the bright regions and streaks lining the b - ring s outer edge ( fig . [ fig : thumbs]c ) . this vertical structure is likely due to `` splashing '' @xcite as particles are forced over one another to accommodate severe radial compression of streamlines in the already closely - packed ring . this effect may very well be far more pronounced around large bodies in the rings , where the large size of the body causes streamlines to divert and compress around it even more . bodies hundreds of meters in size in the a - ring can create three - dimensional propeller - type features in which surrounding ring particles are perturbed hundreds of meters in the vertical direction @xcite . it is plausible that , at the outer b - ring edge where the ring s density is high and resonant forcing is extreme , vertical motions surrounding bodies hundreds of meters in size can be even greater , and the extreme heights of the observed vertical structures may indicate the presence of one or more bodies of kilometer - class size . the additional possibility , that the body ( or bodies ) in region a are on inclined orbits and are driving vertical motions the way that daphnis drives vertical motions in the edge of the keeler gap @xcite , is not likely since sustained vertical motions , even for a 1-km moonlet , would almost certainly damp due to the collisions with dense ring material confined to the ringplane . thus , we speculate that at the heart of region a is a large body or bodies , hundreds of meters to perhaps kilometers in size , and massive enough to severely disturb and elevate to significant heights the material passing around it . region b is likely also the site of larger than average body or bodies , but not large enough to force large vertical motions despite its presence in the strongest resonance zone ( and therefore most extreme compressive environment ) in saturn s rings . a new object , provisionally labeled s/2009 s1 , was discovered @xcite in image n1627301569 , taken through a clear filter with an exposure of 820 ms with the spacecraft and sun on the same side of the rings ( fig . [ fig : thumbs]g ) . based on the length of the shadow @xmath0 36 km and the width of the bright feature @xmath95 1 km it is possible that we are seeing the innermost core of a propeller - shaped disturbance created by an embedded moonlet @xcite , with a height above the ringplane of @xmath0 150 meters and therefore a diameter of @xmath0 300 m. we determined its planetocentric distance ( and presumably its orbital radius ) to be 116914(17 ) km , about 650 km interior to the ring edge . no standard fiducials were available in the discovery image , so the orbital radius was estimated by scaling to two nearby features whose radii were linearly extrapolated from earlier frames in the sequence in which circular reference features were present . the uncertainty in that estimate arises primarily from the uncertainty in the radii of the fiducial features . the interpretation of s/2009 s1 as a solid body , surrounded by the core of a propeller disturbance , is supported by the presence of a shadow . if the observed feature were , say , an impact plume observed very shortly after its creation , so that differential motion had not yet sheared it out as has been observed for other impact clouds imaged with cassini ( http://www.ciclops.org/view.php?id=5782 ) , it could perhaps be optically thick enough to cast a shadow . however , no shadows have been observed accompanying any known impact plumes observed by cassini . s/2009 s1 is almost certainly long - lived . an object 300 m across could not have formed recently from accretion in the outer b ring whose present - day vertical thickness is @xmath0 5 m @xcite . and even in the past , when the ring system was likely much thicker , a solid , denser core nearly the size of the object itself would have been required to initiate accretion , especially this close to saturn ( see fig . 5 in @xcite ) . hence , we think it is safe to conclude that s/2009 s1 is predominantly a solid moonlet , perhaps with a thin rubble outer shell , that likely dates back to the origin of the ring system . the presence of this moonlet lends support to the idea put forth in the previous section that embedded moonlets exist near the edge of the b ring . in a disk of uniform surface density , objects that are too small to open a gap typically drift inwards as the torque from material inward of the body is slightly stronger than that due to material exterior to the body ( i.e , type - i migration ) . in the presence of a negative surface density gradient , @xmath96 , the net torque can be reversed resulting in outward migration @xcite . it is therefore plausible that objects of this size , and perhaps larger , migrated across the region of the outer b ring early in the history of the rings to become trapped permanently in the mimas 2:1 resonance . the alternative hypothesis , where the objects accreted from local ring material within the resonance zone @xcite , seems unlikely because the dynamical stirring in this region is more than sufficient to overcome the self - gravity of the ring particles . the escape velocity for a spherical particle is : @xmath97 where @xmath98 , @xmath99 , and @xmath100 are the mass , radius and density of the particle respectively . for a 10-m particle composed of solid ice , that yields @xmath101 cm s@xmath88 . the velocity dispersion in the resonance zone likely arises primarily from the large - amplitude radial oscillation of the ring edge : @xmath102 where @xmath103 is the peak - to - peak variation of the radial oscillation . for @xmath104 100 km , the radial velocity is @xmath0 5 m s@xmath88 . even with a coefficient of restitution as low as 0.02 ( figure 22 of @xcite ) , the resulting velocity dispersion is @xmath0 10 cm s@xmath88 , much faster than the escape velocity of the largest particles . therefore , as there s no reason to expect particles in this region to be especially sticky , we see no reason to argue for accretion in this region as opposed to other less perturbed regions in the rings . the fact that we find massive bodies within the strongest resonance zone in saturn s rings is more plausibly explained as a result of migration and resonance trapping , analogous to the process , prevalent throughout the solar system , by which satellites and even planets have migrated into , and become trapped , in resonances @xcite . the new moonlet lacks the propeller - shaped disturbance , with extensive wings , that is associated with embedded moonlets in the a ring @xcite . the absence of extensive wings may not be surprising because the ring material in a dense ring like the b ring would be expected to fill in any gaps around the moonlet more quickly than in a less dense region like the mid - a ring . alternatively , it may simply be that the propeller feature is present , but does not comprise a strong enough optical depth contrast to be observable in this optically thick part of the rings , or that it is not resolved . the presence of massive bodies orbiting just interior to the b - ring outer edge may explain the largest discrepancies between the four - component kinematical model and the measurements in fig . [ fig : plots ] . the deviations from the kinematical model ( most obvious in the @xmath4 plot ) are dominated by relatively narrow radial excursions with azimuthal widths around 2030@xmath3 . comparing the longitudes of these `` spikes '' with those of the disturbed regions , it was apparent that many of them coincide with the measured locations of regions a and b. indeed , the association is unambiguous in some cases ( e.g. , sets 2 , 3 , 54 , 64 ; see fig . [ fig : maps ] for reasons stated earlier ] ) , as those features either were initially identified based on their anomalous radial variation , or an anomalous radial excursion is an obvious part of the observed feature . moreover , most of the spikes that are not coincident with the region - a and -b sightings are coincident with their _ computed _ locations based on the the orbits derived from those measurements . therefore the kinematical modeling allows us to identify additional occurrences of those disturbances that we were not able to identify from inspection of the images . in fig . [ fig : plots ] , the spikes that are coincident with one or the other region are colored red ( region a ) or blue ( region b ) in the @xmath4 plot . in total , 13 radial spikes were associated with region a , and 5 with region b. three additional prominent spikes that coincide with neither region are colored green in the figure , and may represent one or more additional sites . indeed , the original inspection and analysis of the images that revealed regions a and b turned up a few other candidates , but none that could be linked together with a common mean motion in the way that regions a and b could . other less prominent features may also be identified as spikes , or they may be unrelated , as the residuals contain quite a bit of structure at the 10-km level . because the spikes apparently arise from a different process than that driving the normal mode patterns , those measurements were excluded from the adopted fit in table [ tbl : elem_m=1224 ] . from the elements in table [ tbl : elem_m=1224 ] , we calculate that the variation in the amplitude of the observed @xmath2 pattern reached a minimum value around mid-2006 . it is an interesting coincidence that regions a and b would have had a conjunction at about the same time , as seen in fig . [ fig : lonplots]c . moreover , as can be seen in fig . [ fig : plots ] , most of the large spikes appear to align with the @xmath4 pattern , even though we would expect no correlation between the spikes and the phase and speed of the pattern in the fitting process because those spikes were not considered in the kinematical fit . additional work will be required to understand the significance , if any , of these coincidences , but they seem to imply a complicated set of commensurabilities between the wave patterns at the edge of the ring and the motions of the embedded mass concentrations . we have confirmed conclusions based on voyager results , and cassini occultation results , that the outer edge of the b - ring is controlled only in part by the mimas 2:1 ilr . we have extended and improved earlier work using 4 years of cassini images to show that , as others recently have found @xcite , the @xmath2 distortion of the edge is undergoing large variations in amplitude and in orientation with respect to the direction to mimas . however , in addition to the forced @xmath2 component and unlike previous interpretations , our results are consistent with the presence of at least three prominent unstable _ free _ normal modes with azimuthal wavenumbers @xmath5 , @xmath2 , and @xmath4 . the forced @xmath2 component is rotating with mimas mean motion ; one of its periapses is fixed relative to the satellite , lagging by @xmath105 , and a simple linear superposition of the free and forced @xmath2 components creates the variability in the @xmath2 shape of the ring . the measured @xmath0 3@xmath3 lag angle implies that the torque exerted on mimas by the ring is sufficient to expand its orbit by as much as @xmath0 0.12 cm yr@xmath88 . with that lag angle as a constraint , forward modeling using a model along the lines of @xcite could allow for a precise estimate of the ring s viscosity ; a rough estimate gives @xmath92 20 @xmath93 s@xmath88 . we have shown that the @xmath5 normal mode , with a radial amplitude of @xmath106 km , is likely a free standing density wave trapped in a resonant cavity at the outer edge of the b - ring . this is the first observation of a long - wavelength ( @xmath107 505 km ) unforced wave in saturn s rings , and points to significant amplification of the @xmath5 wave as it reflects off the b ring s sharp outer edge . the @xmath2 and @xmath4 disturbances have smaller wavelengths of @xmath108 km and @xmath109 km , respectively , and their amplitudes are @xmath110 km and @xmath111 km . an @xmath5 free mode at the outer edge of the b ring was predicted by @xcite based on the availability of a sharp edge and the inference of viscously overstable conditions in this high surface density region . the discovery of the @xmath5 and the other free modes supports this interpretation and strongly suggests that viscous overstability is responsible for most , if not all , of the unforced , structures throughout the b ring and other dense ring regions , like the inner a ring , over a broad range of spatial scales . moreover , our results suggest that brightness variations in the outer @xmath0 250 km of the b ring may be primarily due to trapped free density waves rather than albedo variations , as suggested by @xcite . because the b - ring edge can be modeled using a simple linear superposition of modes , and because the observed value for the forced @xmath2 amplitude is nearly identical to its predicted value , we argue that any nonlinear coupling between the observed modes is weak , though interaction between the two @xmath2 modes may be enhanced during the relatively long interval when the two patterns are nearly aligned . at that time , when the total @xmath2 amplitude is the greatest , the nonlinearity parameter @xmath112 is also at its greatest and may lead to saturation of the free mode , limiting its growth . the free modes we have found here are all nodeless : i.e. , the sign of the radial displacement is constant across the resonant cavity . indeed , different modes with the same @xmath113 but with radial nodes within the resonant cavity and hence with shorter wavelengths are possible and in principle could be present . however , for the same radial displacements , these shorter - wavelength modes would have higher nonlinearity parameters @xmath114 than the nodeless modes so their interactions with the nodeless modes of the same @xmath113 would therefore be expected to be nonlinear and lead to damping ; this would be especially so for the @xmath2 and @xmath4 modes . consequently , it is plausible that the smaller amplitude modes with radial nodes , especially the @xmath115 modes , have been significantly suppressed by their more robust nodeless counterpart . nonetheless , some of these modes , if present , would have pattern speeds close enough to those measured here that they would not be detectable without a longer time baseline of observations . additional data sets will be obtained at regular intervals throughout the cassini solstice mission : during the next 7 years we hope to better than double the time baseline of this investigation . also , by using the orbit fits to regions a and b , we may now hope to unravel the historical occultation data sets dating back to the voyager era , which have been difficult to model in the past because the smoothly varying normal - mode pattern was interrupted in a heretofore unpredictable way by the spikes associated with these regions . the narrow , simply eccentric ringlets seen throughout saturn s rings and the rings of uranus are also likely the results of viscously unstable density waves of @xmath5 reflecting off the sharp edges of a resonance cavity @xcite . the @xmath116-km amplitude @xmath5 narrow maxwell ringlet in saturn s c ring is an example . however , the existence of lower- and higher - wavenumber modes are also possible within a narrow ring : the huygens ringlet , a roughly 50-km - wide ringlet a few hundred km exterior to the b - ring edge , possesses an @xmath2 mode with an amplitude of @xmath0 20 km @xcite , and possibly a 1-km @xmath117 mode @xcite , on top of a 30-km amplitude @xmath5 mode , and some of the rings of uranus exhibit @xmath118 and @xmath2 modes @xcite . the detailed analysis of the huygens and maxwell ringlets will be a topic of future work . the three free modes discovered in this work reside in zones ranging in width from 24 km to 250 km , with outer boundaries at the b - ring edge . independent estimates for the average surface mass densities over those respective regions , for a given fixed value of @xmath30 , give consistent values . we take this as support for the existence of these free modes , though if @xmath30 were strongly dependent on wavenumber , that argument would be invalid . a lower bound on the surface density was obtained by setting @xmath119 , yielding a weighted mean of @xmath0 44 g @xmath1 . the true density could be near 100 g @xmath1 , or perhaps even higher in the outermost zone . normal modes of the kind we have found at the edge of the b - ring are analogous to the free unstable modes that arise in numerical simulations of other celestial disks , such as protoplanetary disks @xcite and spiral galaxies . however , definitive identification of such free modes in real spiral disk galaxies , and their role in the creation of grand design spirals , is hindered by the enormously long periods anticipated for such modes , on the order of @xmath120 yr , which can not be directly measured . our results underscore the natural connection between saturn s rings , our solar system s most accessible celestial disk , and other disks throughout the cosmos galaxies as well as protoplanetary disks . we have found circumstantial evidence for the presence of large bodies in the resonance zone at the outer edge of the b - ring that are observable through their disturbance to the surrounding ring . these disturbances are following simple uniform motion and are in orbits separated by 18 km . they are responsible for most of the prominent short - wavelength fluctuations in radius that are observed at the ring edge ; other regions containing larger - than - average bodies are also likely present and may be responsible for other short - wavelength disturbances seen in the outer edge . the suggestion that moonlets are embedded near the outer edge of the ring is supported by the discovery of a 300-m diameter moonlet orbiting about 650 km interior to the ring edge . such bodies may have been present in the vicinity of the outer b - ring edge early in the history of saturn s rings , subsequently migrating into the resonance zone to become permanently trapped in the mimas 2:1 resonance . this result suggests that radial spikes noticed in our recent work on the a - ring outer edge @xcite could point to the presence of embedded moonlets there as well , likely trapped in the janus / epimetheus 7:6 inner lindblad resonance that controls and shapes that edge . this will be a subject of future work . we acknowledge the staff members within ciclops for their dedication and efforts in the design , execution , and processing of the ring imaging sequences used in this work . we are indebted to p. goldreich for valuable suggestions in the kinematic modeling and interpretation of our results . we also thank r. a. jacobson for sharing his fits to mimas mean motion , as well as r. greenberg and j. hahn for helpful discussions . jns and ccp were both supported by nasa and the cassini project . finally , we thank m. s. tiscareno for a thorough review that improved the quality of the text significantly . borderies , n. , goldreich , p. , tremaine , s. 1985 , icarus , 63 , 406 . borderies , n. , goldreich , p. , tremaine , s. 1982 , nature , 299 , 209 . colwell , colwell , j. e. , nicholson , p. d. tiscareno , m. s. , murray , c. d. , french , r. g. , marouf , e. a. 2009 , in saturn , ed . m. dougherty , l. esposito , and s. krimigis ( new york , ny ; springer ) , chapter 13 , 375 . colwell , j. e. , esposito , l. w. , sremcevic , m. , stewart , g. r. , mcclintock , w. e. 2007 , icarus , 190 , 127144 . cuzzi , j. n. , lissauer , j. j. , esposito , l. w. , holberg , j. b. , marouf , e. a. , tyler , g. l. , boischot , a. 1984 , in planetary rings , ed . r. greenberg & a. brahic ( tucson , az ; u of a press . ) 73 - 199 esposito , l. w. 2009 , agu fall meeting 2009 , # p51b-1130 . french , r. g. , et al . 1993 , icarus , 103 , 163214 greenberg , r. 1973 , aj , 78 , 338346 . ferraz - mello , s , beaug , c. , michtchenko , t. a. , 2003 , cel . mech , 87 , 99112 . goldreich , p. , tremaine , s. 1982 , ann . , 20 , 249283 . goldreich , p. , tremaine , s. 1978 , icarus , 34 , 240253 . hahn j. m. , spitale , j. n. , porco , c. c. 2009 , apj , 699 , 686710 . hedman , m. m , et al . 2010 , aj , 139 228251 . laughlin , g. , korchagin , v. , adams , f.c . 1997 , apj , 477 , 410 . lee , m. h. , thommes , e. w. 2009 , apj , 702 , 16621672 . marley , m. , porco , c. 1993 , icarus , 106 , 508 . murray , c. d. , dermott , s. f. 1999 . cambridge univ . press , cambridge . nicholson , p. d. , french , r. g. , tollestrup , e. , cuzzi , j. n. , harrington , j. , matthews , k. , perkovic , o. , stover , r. j. 2000 , icarus , 145 , 474 . porco , c. c. 2009 , iauc 9091 . porco , c. c. , weiss , j. w. , richardson , d. c. , dones , l. , quinn , t. , throop , h. 2008 , aj , 136 , 21722200 . porco , c. c. , thomas , p. c. , weiss , j. w. , richardson , d. c. 2007 , science , 318 , 16021607 . porco , c. c. , et al . 2005 , science , 307 , 12261247 . porco , c. c. 1990 , adv . space res . 10 , 221 - 229 . porco , c. c. , danielson , g. e. , goldreich , p. , holberg , j. b. , lane , a. l. 1984 , icarus , 60 , 1728 . spitale , j. n. , porco , c. c. 2009 , aj , 138 , 15201528 . spitale , j. n. , porco , c. c. 2006 , lpsc 37 , # 2242 . spitale , j. n. , porco , c. c. 2006 , baas 38 , 578 . tiscareno , m. s. , et al . 2010 , apjl 718/2 , l92l96 . tiscareno , m. s. , hedman , m. m. , burns , j. a. , weiss , j. w. , porco , c. c. 2009 , baas 41 , 10411042 . tiscareno , m. s. , burns , j. a. , nicholson , p. d. , hedman , m. m. , porco , c. c. 2007 . icarus 189 , 1434 . tiscareno , m. s. , burns , j. a. , hedman , m. m. , porco , c. c. , weiss , j. w. , dones , l. , richardson , d. c. , murray , c. d. 2006 . nature 440 , 648650 . ward , w. r. , 1986 . icarus 67 , 164180 . weiss , j. w. , porco , c. c. , tiscareno , m. s. 2009 , aj , 138 , 272286 . .details of data sets used in this study . the incidence , emission , and phase angles are represented by @xmath121 , @xmath122 , and @xmath123 , respectively . @xmath121 and @xmath122 are referenced to the northern ringplane normal . notes : ( a ) ansa movie . ( b ) azimuthal scan . ( c ) used for shadows only.[tbl : data ] [ cols="<,<,<,<,<,<",options="header " , ] [ fig : maps ]	voyager images and cassini occultation data have previously shown that the behavior of the outer edge of saturn s massive b ring is determined only in part by a static response to the 2:1 inner lindblad resonance with mimas . in cassini images of this region , we find , in addition to the expected wavenumber-2 forced distortion , evidence for unforced self - excited wavenumber-3 , wavenumber-2 , and wavenumber-1 normal modes . these are the first observations to suggest substantial wave amplification in saturn s broad rings . moreover , the presence of these free modes strongly implicates viscous overstability as their underlying cause and , by inference , the cause for most if not all of the unforced structures throughout the high - mass - density b ring and in other high - mass - density regions in saturn s rings . analysis of each of the inferred waves reveals a consistent lower bound on the average surface mass density of @xmath0 44 g @xmath1 for the outer 250 km of the ring , though the true surface density could be as high as 100 g @xmath1 or higher . interference between the forced and free wavenumber-2 modes yields a total wavenumber-2 pattern that varies in amplitude and orientation with a characteristic period of @xmath0 5.5 years . we also find localized disturbances , including 3.5-km - tall vertical structures , that provide circumstantial evidence for embedded massive bodies in the mimas resonance zone . the presence of such bodies is supported by the presence of a shadow - casting moonlet @xmath0 0.3 km wide near the ring s edge . _ subject headings : _ planets : rings
in simple analytic terms the abundance ratio of a short - lived ( t@xmath17 100 myr ) radioactive to a stable isotope in the material that ended up in the solar system , just after a given nucleosynthetic event and assuming that both nuclides are only produced by this type of event , can be derived using the formula @xcite : @xmath18 where @xmath19 is the production ratio of each single event , @xmath20 is the mean life ( = t@xmath21/ln 2 ) of the radionuclide , @xmath22 the timescale of the evolution of the galaxy up to the formation of the sun ( @xmath23 yr ) , and @xmath24 the recurrence time between each event . the value of @xmath24 is essentially a free parameter that may vary between 10 and 100 myr @xcite . the number k is also a free parameter that accounts for the effect of infall of low - metallicity gas , which dilutes the abundances , and the fact that a fraction of the abundances produced , particularly for stable isotopes , is locked inside stars @xcite . the value of k changes depending on whether the isotopes involved are of primary or secondary origin , i.e. , whether they are produced from the h and he abundances in the star or depend on the initial presence of cno elements , respectively . these effects are complex to evaluate analytically , but the general result is that k @xmath25 1 @xcite . et al_. @xcite did not consider the number k in their evaluation of ratios from eq . [ eq : eq1 ] , which means that effectively they used k=1 and their reported timescales represent conservative lower limits . tr14 did not use eq . [ eq : eq1 ] , but included the yields of @xmath0nb , @xmath26tc , @xmath27tc , and @xmath1sm ( and their reference isotopes @xmath0mo , @xmath27ru , and @xmath28sm ) from their snia models into full , self - consistent galactic chemical evolution ( gce ) simulations to evaluate the abundance ratios @xmath0nb/@xmath0mo , @xmath26tc/@xmath27ru , @xmath27tc/@xmath27ru , and @xmath1sm/@xmath28sm in the interstellar medium ( ism ) at the time of the birth of the sun , assuming that the production of @xmath2 nuclei only occurs in snia . these detailed models reproduce the abundances of the stable reference isotopes considered here @xcite . with eq . [ eq : eq1 ] we can recover results close to those of the detailed gce models for the four ratios considered here using as @xmath19 the average of the values given in table 1 of tr14 = 0.01 0.02 , since we are focusing on events that occurred close in time to the formation of the sun . variations in this range of @xmath29 are within 25% . when considering also @xmath29 down to 0.003 they are within a factor of 2 . the only exception is @xmath27tc/@xmath27ru , which varies by 50% in the range @xmath29 = 0.01 0.02 and by a factor of 6 when also considering @xmath29 = 0.003 . ] , t = 9200 myr from tr14 , @xmath24 = 8 myr , and k@xmath30 ( table [ tab : tab ] ) . this means that roughly ( 1/k)(t/@xmath31 600 snia @xmath2-process events contributed to the solar system abundances of the stable isotopes considered here . for the unstable isotopes , it depends on their mean life : because @xmath26tc and @xmath27tc have relatively short mean lives , the second term of the sum in eq . [ eq : eq1 ] representing the memory of all the events prior the last event counts for 26% of the total , hence , most of their ess abundances come from the last event . on the other hand , for @xmath0nb and @xmath1sm , due to their long half lives , the second term of the sum is the most important . for example , in the case of @xmath0nb it accounts for 85% of the total amount of @xmath0nb . in these cases the ratios from eq . [ eq : eq1 ] are very close to the corresponding ism steady - state ratio , i.e. , the production ratio multiplied by k@xmath20/t . although we can recover the values of the ratios produced by the full gce models using eq . [ eq : eq1 ] , we need to keep in mind some distinctions . the ratios derived by tr14 using the full gce model represent values at an absolute time 9200 myr from the birth of the galaxy , when the ism reaches solar metallicity . from these values , we can evaluate an isolation timescale ( t@xmath32 ) : the interval between the time when the material that ended up in the solar system became isolated from the ism and the time of the formation of the solar system . t@xmath32 is derived such that the ratio between the ess ratio and the ism ratio for a given radioactive and stable pair is simply given by @xmath33 . in reality , however , some mixing could have occurred . an ism mixing timescale ( t@xmath34 ) between different phases of the ism should be of the order of 10 - 100 myr . the effect of such process was analytically investigated by @xcite , from which @xcite derived that the ratio between the ess ratio and the ism ratio for a given radioactive and stable pair is given by @xmath35 , where @xmath36 t@xmath37 . in this picture , the requirement is that the composition of the star - forming region where the sun was born must have been affected by mixing with the ism , hence , t@xmath34 represents a lower limit for t@xmath32 . values derived using eq . [ eq : eq1 ] , instead , represent ratios in the matter that built up the solar system just after the last , final addition from a nucleosynthetic event . from them , we can evaluate a last - event timescale ( t@xmath38 ) : the interval between the time of the last event and the time of the formation of the solar system . t@xmath38 represents an upper limit of t@xmath32 and is derived such that the ratio between the ess ratio and the ism ratio for a given radioactive and stable pair is simply given by @xmath39 ( as for t@xmath32 ) . the more @xmath40 is lower than unity , the closer the ratio derived from eq . [ eq : eq1 ] is to the ism ratio , and the closer t@xmath38 to t@xmath32 . the main drawback of this approach is that the k and the @xmath24 values in eq . [ eq : eq1 ] are uncertain . the advantages are that there are not further complications with regards to mixing in the ism and that this description is relatively free from uncertainties related to the setting of the time of the birth of the sun . in the model of tr14 , this is defined as the time when the galactic metallicity reaches the solar value . however , stellar ages and metallicities in the galaxy are not strictly correlated @xcite . chemodynamical gce models also including the effect of stellar migration @xcite are required to obtain the observed spread , but have not been applied to the evolution of radioactive nuclei yet . we expect these models to produce a range of abundances of the radioactive nuclei at any given metallicity , but we can not predict based on first principles if these variations will be significant . finally , it should be noted that , while t@xmath32 and t@xmath34 are required to be the same for all the different types of nucleosynthetic events that contributed to the solar system matter , t@xmath38 is in principle different for each type of event . we evaluate t@xmath32 and t@xmath34 assuming that the @xmath2-process radionuclides were produced by sneia and comparing the abundance ratios obtained by tr14 to those observed in the ess ( table [ tab : tab ] ) . tr14 carefully analysed the nuclear uncertainties that affect the production of @xmath0nb and @xmath1sm in sneia . in the production of @xmath0nb , ( @xmath12,n ) reactions play the dominant role with some contribution from proton induced reactions . the nuclear uncertainties are related to the choice of the @xmath12-ray strength function and , to a lesser extent , the proton - nucleus optical potential . since the rates of some of the most important reactions are constrained experimentally , the nuclear uncertainties on the @xmath0nb production are moderate , resulting in possible variations in the ism ratio of less than a factor of two . the @xmath1sm/@xmath28sm ratio , on the other hand , is determined by ( @xmath12,n)/(@xmath12,@xmath4 ) branchings , mainly on @xmath41gd . the uncertainty range reported by tr14 is based on three choices of the @xmath41gd(@xmath12,@xmath4)@xmath28sm rate . with respect to two previous estimates @xcite , the new rate of @xcite results in a @xmath1sm/@xmath28sm ratio higher by a factor of two in snia , but lower by at least a factor of two in ccsne @xcite . realistically , the @xmath1sm/@xmath28sm ratio may have an even higher nuclear uncertainty , possibly up to one order of magnitude , owing to the lack of experimental data at the relevant low energies for the @xmath41gd(@xmath12,@xmath4)@xmath28sm rate itself and for the low energy @xmath4-nucleus optical potential required for the extrapolation @xcite . for @xmath0nb/@xmath0mo , if we compare the maximum value allowed within nuclear uncertainties of 3.12 @xmath42 to the lower limit of the ess value , we derive a maximum t@xmath32 of 5.4 myr and a maximum t@xmath34 of 3.7 myr . it is also possible to recover a solution for @xmath1sm/@xmath28sm consistent with these values , given the large uncertainties . currently , the value of the half life of @xmath1sm is debated between the two values of 68 and 103 myr @xcite . here , we have used the lower value , if we used the higher value we would obtain longer timescales than those reported in table [ tab : tab ] . we extend the study of tr14 to investigate if the origin of @xmath0nb and @xmath1sm is compatible with that of the radionuclide @xmath3mn . this nucleus provides us a strong constraint because near chandrasekhar - mass sneia are also the major producers of mn in the solar system @xcite and , together with the stable @xmath43mn , they produce @xmath3mn , whose solar abundance is well determined in the ess ( table [ tab : tab ] ) . we calculate the ism @xmath3mn/@xmath43mn ratio from eq . [ eq : eq1 ] using the same values for @xmath24 , k , and t derived above and the production ratio @xmath3mn/@xmath43mn=0.108 predicted by the same snia model used to derive the @xmath2-process abundances . this is derived from the @xmath3cr abundance given by @xcite , but , we note that different models produce very similar @xmath3mn/@xmath43mn ratios @xcite . we calculate an ism @xmath3mn/@xmath43mn ratio of 2.41 @xmath44 , which , after an isolation time of 5.4 myr , results in a ratio of 5.82 @xmath42 , one order of magnitude higher than the ess value . an isolation time of at least 19 myr is instead required to match the @xmath3mn/@xmath43mn constraint . if we make the conservative assumption that k=1 , instead of 2 , to calculate the ism @xmath3mn/@xmath43mn ratio from eq . [ eq : eq1 ] we obtain t@xmath45 15 myr . this value is strongly incompatible with the upper limit of 5.4 myr required to match the @xmath0nb/@xmath0mo , in other words , snia nucleosynthesis results in too high production of @xmath3mn relative to @xmath0nb , and to their ess abundances . a three times lower @xmath3mn/@xmath43mn ratio in snia would result from a tenfold increase of the @xmath46s(@xmath47)@xmath46p decay @xcite , in which case t@xmath32 would decrease to @xmath48 9 myr . this possibility needs to be further investigated . on the other hand , a possibly longer half life for @xmath3mn ( see discussion in @xcite ) would further increase the difference . to reconcile the ess @xmath0nb abundance with an isolation time of 15 myr , the half life of @xmath0nb should be almost a factor of three higher , which seems unrealistic . the current half life is the weighted average of two experiments that produced similar results in spite of being based on different normalizations : the first @xcite is normalized to the half life of @xmath49nb , which is not well known , and the second @xcite to an assumed value of the @xmath50nb(n,2n)@xmath0nb cross section . another option is that the @xmath0nb/@xmath0mo ratio in the ess is lower than 2.2 @xmath42 , i.e. , 21% lower than the current lower limit . it is harder to reconcile the different t@xmath34 : when using the lowest ( k=1 ) ism @xmath3mn/@xmath43mn ratio and the upper limit for the ess ratio , we obtain a lower limit for t@xmath34 of 24 myr . to reconcile the @xmath0nb/@xmath0mo ratio to this timescale an ess @xmath0nb/@xmath0mo=1.6@xmath42 is required , 43% lower than the current lower limit . a different solution is to look into another production site for @xmath0nb . as mentioned above , the @xmath12-process in ccsn models do not efficiently produce @xmath2-process isotopes in the mo - ru region , but , other ccsn nucleosynthesis components may contribute to these isotopes . pignatari _ et al_. 2013 @xcite computed ccsn models with initial mass of 15 m@xmath51 that carry in the ejecta an @xmath4-rich freezeout component . there , production of the @xmath2-process nuclei up to @xmath0mo occurs due to a combination of @xmath4 , proton , and neutron captures and their inverse reactions during the ccsn explosion in the deepest part of the ejecta , where @xmath52he is the most abundant isotope . the stellar evolution before the ccsn explosion was calculated with the code genec @xcite and the explosion simulations were based on the prescriptions for shock velocity and fallback by @xcite . the standard initial shock velocity used beyond fallback is @xmath53 cm s@xmath54 for all the masses , and we also experimented by reducing it down to 200 times lower . results are presented for ccsn models computed with two setups for the convection - enhanced neutrino - driven explosion : fast - convection ( the rapid setup ) and delayed - convection ( the delay setup ) explosions @xcite . in this framework , nuclear uncertainties are insignificant compared to model uncertainties . a post - processing code was used to calculate the nucleosynthesis before and during the explosion @xcite . the results are summarized and compared to the snia models in figure [ fig : ratios ] . in comparison to the snia models , the ccsn models produce less @xmath55mn , @xmath56fe , and @xmath57sm , and almost no @xmath27ru and @xmath58tc . on the other hand , no @xmath59fe is produced in sneia , and the cosmic origin of this isotope has to be ascribed to other types of supernovae , including ccsne . significant production of @xmath0nb and @xmath0mo occurs in ccsn models of 15 m@xmath51 , with production factors for @xmath0mo comparable to those of the snia models , and @xmath0nb/@xmath0mo ratios up to five times higher . ccsn models with initial mass larger than 15 m@xmath51 do not eject material exposed to the @xmath4-rich freezeout due to the more extended fallback . the @xmath4-rich freezeout efficiency also strongly depends on the shock velocity . when its standard value is reduced , even only by a factor of two , the amounts of @xmath0mo and @xmath0nb ejected become negligible . this is because for lower shock velocities the bulk of @xmath4-rich freezeout nucleosynthesis is shifted toward lighter elements closer to the fe group . only some @xmath12-process component is produced in these cases , which is poor in @xmath0mo . in summary , the @xmath4-rich freezeout conditions suitable to produce @xmath0mo and @xmath0nb are more likely hosted in ccsne with initial mass of 15 m@xmath51 or lower and shock velocities at least as large as those provided by @xcite . we explored the possibility that the @xmath0nb in the ess came from the @xmath4 freezeout production experienced by these 15 m@xmath51 ccsn models . since the premise of eq . [ eq : eq1 ] is that both the stable and unstable isotopes present in the ess were produced only by this type of event , under such assumption , we can apply eq . [ eq : eq1 ] only to calculate the @xmath0nb/@xmath0mo ratios to derive the time of the last such @xmath4 freezeout event , since the other isotopes are not produced in any significant abundancesnb . ] . the calculations are hampered by the error bars on the ess ratios and the large uncertainties related to the choice of the free parameters entering in eq . [ eq : eq1 ] . however , to infer further information on the birth of the sun we are particularly interested in determining the lower limit of t@xmath38 because this provides us the shortest possible timescale , within all the uncertainties above , for the life of the star forming region where the sun was born . using k=1 in eq . [ eq : eq1 ] and the upper limit of the ess @xmath0nb/@xmath0mo ratio we inferred a lower limit for t@xmath38 of 10 myr when using the delay setup ccsn model ( table [ tab : tab2 ] ) . however , we stress that k=1 represents a very conservative lower limit because the production factor @xmath0mo/@xmath0mo@xmath51 of the ccsn delay model is comparable to that of the snia model ( 140 versus 172 ) . this means that we would need a similar number of events ( which scales as 1/k@xmath24 ) to reproduce the @xmath0mo solar abundance - although the exact value would depend on the metallicity dependence of @xmath0mo/@xmath0mo@xmath51 in the ccsn models , which we did not explore . the lower limit of t@xmath38 from the rapid setup ccsn model is instead a negative value . however , the production factor @xmath0mo/@xmath0mo@xmath51 of the rapid model is five times higher than that of the delay model ( 704 versus 140 ) . this means that we would need five times less events to reach the same value of @xmath0mo in the solar system , hence , for consistency , a five times higher value of 1/k@xmath24 is more appropriate in this case . if we use , e.g. , @xmath24= 50 myr we obtain 7 myr as the convervative lower limit for t@xmath38 . our models show that @xmath1sm in the ess was produced by snia events , together with @xmath58tc . the large nuclear uncertainties on the former and the availability of upper limits only for the ess ratio for the latter two do not allow us to derive precise timescales . on the other hand , to reconcile the abundances of @xmath3mn and @xmath0nb in the ess as produced primarily by snia events , the ess @xmath0nb/@xmath0mo ratio needs to be at least 50% lower than the current nominal value , which is outside its present error bars . in this case t@xmath32 @xmath48 15 myr , which is consistent with the t@xmath38 derived for the last @xmath9 ( @xmath5 100 myr ) and @xmath11 ( @xmath5 20 myr ) process events @xcite since t@xmath38 represents an upper limit for t@xmath32 . furthermore , if all the @xmath3mn in the ess was produced by the same snia events that produced @xmath0nb , no ccsne exploding within the star - forming region where the sun was born should have contributed to the abundances of radioactive nuclei in the ess , since they would have contributed extra @xmath3mn . if correct , this will have important consequences for the origin of @xmath60al and @xmath59fe in the ess . the @xmath59fe/@xmath56fe is likely the result of the decay , after a certain isolation time , of the @xmath59fe abundance in the ism . this is consistent with the direct observational constraint from @xmath12-ray astronomy of @xmath59fe/@xmath56fe @xmath61 in the ism , and with timescales of the order of 15 myr , while @xmath60al may have originated from the winds of a massive star @xcite . another possibility is that the @xmath0nb was instead predominantly produced by ccsne that experienced the @xmath4-rich freezeout , corresponding in our models to low initial stellar masses experiencing a core - collapse with high shock velocities . in the extreme scenario where all the @xmath0nb and @xmath0mo in the ess came from such events we have derived lower limits for the timing of the last of them . these are also still consistent with a relatively long t@xmath32 of at least 7 myr , since t@xmath32 @xmath5 t@xmath62 when @xmath63 , which is verified for @xmath0nb when @xmath24=10 myr . clearly , @xmath0nb could have been produced by both snia and ccsn events . full gce models are required , including both snia , the @xmath4-rich freezeout component of ccsn considered here , and the possible production in the neutrino winds . because the @xmath4-rich freezeout ccsne events of initial mass 15 m@xmath51 produce higher amounts of @xmath0nb than snia , we expect that adding them to the balance will increase the ism @xmath0nb/@xmath0mo ratio at the time of the formation of the solar system , leading to longer t@xmath32 , and an easier match with the @xmath3mn/@xmath43mn constraint . in conclusion , a higher - precision determination of the ess @xmath0nb/@xmath0mo ratio is urgently required . this will allow us to determine if snia are the major source of light @xmath2-process nuclei in the galaxy or if ccsne also play a role ; and which fraction of the @xmath3mn in the ess originated from snia . ml is a momentum ( `` lendlet-2014 '' programme ) project leader of the hungarian academy of sciences . mp acknowledges support to nugrid from nsf grants phy 09 - 22648 ( joint institute for nuclear astrophysics , jina ) , nsf grant phy-1430152 ( jina center for the evolution of the elements ) and eu mirgct-2006 - 046520 . mp acknowledges the support from the `` lendlet-2014 '' programme of the hungarian academy of sciences ( hungary ) and from snf ( switzerland ) . c , gallino r , rauscher t , rpke fk , hillebrandt w ( 2015 ) testing the role of sne ia for galactic chemical evolution of p - nuclei with two - dimensional models and with s - process seeds at different metallicities . 799:54 . ne , borg le , hutcheon i d , jacobsen b , clayton rn ( 2014 ) samarium - neodymium chronology and rubidium - strontium systematics of an allende calcium - aluminum - rich inclusion with implications for @xmath1sm half - life . 405:1524 .	the abundances of @xmath0nb and @xmath1sm in the early solar system are determined from meteoritic analysis and their stellar production is attributed to the @xmath2 process . we investigate if their origin from thermonuclear supernovae deriving from the explosion of white dwarfs with mass above the chandrasekhar limit is in agreement with the abundance of @xmath3mn , another radionuclide present in the early solar system and produced in the same events . a consistent solution for @xmath0nb and @xmath3mn can not be found within the current uncertainties and requires that the @xmath0nb/@xmath0mo ratio in the early solar system is at least 50% lower than the current nominal value , which is outside its present error bars . a different solution is to invoke another production site for @xmath0nb , which we find in the @xmath4-rich freezeout during core - collapse supernovae from massive stars . whichever scenario we consider , we find that a relatively long time interval of at least @xmath5 10 myr must have elapsed from when the star - forming region where the sun was born was isolated from the interstellar medium and the birth of the sun . this is in agreement with results obtained from radionuclides heavier than iron produced by neutron captures and lends further support to the idea that the sun was born in a massive star - forming region together with many thousands of stellar siblings . adionuclides with half lives of the order of 1 to 100 million years ( myr ) were present in the early solar system ( ess ) and represent a clue to investigate the circumstances of the birth of the sun @xcite . the abundances of some of these nuclei in the ess , as inferred from analysis of meteorites and reported relative to the abundance of a stable nucleus , are given with error bars as low as a few percent . however , our ability to exploit radionuclides as chronometer for the events that led to the birth of the sun is undermined by the fact that our understanding of their production by nuclear reactions in stars is still relatively poor . this is because it relies on stellar models that include highly uncertain physics related to , e.g. , mixing , the mechanisms of supernova explosions , and nuclear reaction rates . here we focus on the origin of radionuclides heavier than fe . among them , @xmath0nb , @xmath6i , @xmath1sm , and @xmath7hf have abundances in the ess known with error bars lower than 20% . lugaro _ et al_. @xcite recently proposed a decoupled origin for @xmath6i and @xmath7hf : the former was produced by @xmath8 neutron captures ( @xmath9 process ) in a supernova or neutron star merger roughly 100 myr before the formation of the sun , and the latter by @xmath10 neutron captures ( @xmath11 process ) in an asymptotic giant branch star roughly 1030 myr before the formation of the sun . from these timescales it was concluded that the star forming region where the sun was born may have lived for a few tens of million years , which is possible only if such a region was very massive and gave birth to many hundreds of stellar siblings @xcite . here , we turn to the investigation of the origin the other two radionuclides heavier than fe whose abundances are well known in the ess : @xmath0nb and @xmath1sm . these nuclei are proton rich , relative to the stable nuclei of nb and sm , which means that they can not be produced by neutron captures like the vast majority of the nuclei heavier than fe . instead , their nucleosynthetic origin has been traditionally ascribed to some flavor of the so - called @xmath2 process @xcite , for example , the disintegration of heavier nuclei ( @xmath12 process ) . unfortunately , the possible astrophysical sites of origin of @xmath2-process nuclei in the universe are not well constrained . core - collapse supernova explosions ( ccsne @xcite ) deriving from the final collapse of massive stars have been considered as a possible site for the @xmath12 process , specifically , occurring in the o - ne - rich zones of the ccsn ejecta @xcite . however , models never managed to reproduce the complete @xmath2-process pattern observed in the bulk of the solar system material @xcite . for instance , ccsn models can not reproduce the relatively large abundances of @xmath13mo and @xmath14ru . taking into account nuclear uncertainties has not solved the problem @xcite , except for a possible increase of the @xmath15c+@xmath15c fusion reaction rate @xcite . another process in ccsne that can produce the light @xmath2-process nuclei up to pd - ag , including @xmath0nb , is the combination of @xmath4 , proton , neutron captures , and their reverse reactions in @xmath4-rich freezeout conditions @xcite . neutrino winds from the forming neutron star are also a possible site for the production of the light @xmath2-process nuclei @xcite , although , one of its possible components ( the so - called @xmath16p - process @xcite ) can not produce @xmath0nb because it is shielded by @xmath0mo @xcite . the same occurs in the case of the _ rp_-process in x - ray bursts @xcite . neutrino - induced reactions in ccsne ( the @xmath16-process ) can also produce some @xmath0nb @xcite , but no other @xmath2-process nuclei . thermonuclear supernovae ( sneia ) deriving from the explosion of a white dwarf that accreted material from a main sequence companion to above the chandrasekhar limit have been proposed as a site of the @xmath12 process @xcite . in these models , heavy seed nuclei are assumed to be produced by the @xmath11 process during the accretion phase and , given a certain initial @xmath11-process distribution , it is possible to reproduce the high abundances of @xmath13mo and @xmath14ru in the solar system . however , there are still many uncertainties related to the origin and the features of the neutron - capture processes activated during the white dwarf accretion leading to snia . recently , travaglio _ et al_. @xcite ( hereafter tr14 ) analyzed in detail the production of @xmath0nb and @xmath1sm in sneia using multidimensional models and concluded that such an origin is plausible for both radionuclides in the ess . here , we consider these results and combine them with new predictions of the production of @xmath2-process nuclei in @xmath4-rich freezeout conditions in ccsne to investigate the origin of @xmath0nb and @xmath1sm and to use it to further constrain the circumstances of the birth of the sun .
we start with the relevant definitions . let @xmath29 be a binary string . then , we denote by the verifier @xmath102 a quantum circuit of length @xmath103 , @xmath4 ( with local gates @xmath5 ) acting on @xmath104 qubits , which is generated uniformly from @xmath29 . the verifier takes an @xmath105 qubit quantum state @xmath7 as an input ( we will express everything in terms of @xmath1 instead of @xmath106 in the following ) , together with @xmath8 initialized ancillas , @xmath9 , applies @xmath10 , and finally measures the first qubit in the @xmath11 basis to return @xmath12 ( `` proof accepted '' ) or @xmath13 ( `` proof rejected '' ) . the acceptance probability for a proof @xmath2 is then given by @xmath17 , with @xmath107 [ sbqp1 ] let @xmath108 s.th . @xmath16 , and let @xmath32 be a verifier map with no eigenvalues between @xmath33 and @xmath34 . then , the class @xmath114 consists of all problems of the form `` compute the dimension of the space spanned by all eigenvectors of @xmath32 with eigenvalues @xmath35 '' . [ sbqp2 ] consider a verifier @xmath32 with the property that there exist subspaces @xmath115 ( @xmath116 ) such that @xmath37 for all @xmath38 , and @xmath39 for all @xmath40 , where again @xmath16 . then _ @xmath117 _ consists of all problems of the form `` compute @xmath81 '' . note that @xmath118 is well - defined : consider two decompositions @xmath119 and @xmath120 . without loss of generality , if we assume @xmath121 , it follows @xmath122 , and thus there exists a non - trivial @xmath123 , which contradicts the definition . to show that definition [ sbqp1 ] implies definition [ sbqp2 ] , let @xmath41 be spanned by the eigenvectors with eigenvalues @xmath35 . to show the converse , we use the minimax principle for eigenvalues @xcite , which states that the @xmath83th largest eigenvalue @xmath124 of a hermitian operator @xmath32 in an @xmath125-dimensional hilbert space can be obtained from either of the equivalent optimizations @xmath126 where @xmath127 is a subspace of dimension @xmath83 , and @xmath128 is a unit vector . now notice that def . [ sbqp2 ] implies that @xmath129 next , consider the minimax theorem for @xmath130 . from eq . ( [ e : maxmin ] ) we have @xmath131 now consider the case that @xmath132 . from eq . ( [ e : minmax ] ) , using the fact that @xmath133 , we have @xmath134 thus we have @xmath135 since @xmath136 . it follows that @xmath137 is the smallest eigenvalue of @xmath32 which is still larger than @xmath33 , and therefore the span of the first @xmath81 eigenvectors of @xmath32 is equal to the span of all eigenvectors with eigenvalue @xmath35 . the equivalence follows . this follows directly from the strong amplification procedure presented in @xcite , which describes a procedure to amplify any verifier map @xmath32 such that any eigenvalue above @xmath33 ( below @xmath34 ) is shifted above @xmath144 ( below @xmath145 ) at an overhead polynomial in @xmath146 . we now use the class to characterize the complexity of the density of states problem and the problem of counting the number of ground states of a local hamiltonian . we start by defining these problems , as well as the notion of local hamiltonian , and then show that both problems are -complete . [ def : klocalh ] given a set of @xmath148 quantum spins each with dimension bounded by a constant , a hamiltonian @xmath25 for the system is said to be @xmath83-local if @xmath149 is a sum of at most @xmath148 hermitian operators @xmath53 , @xmath54 , each of which acts nontrivially on at most @xmath83 spins . note that @xmath83-local does not imply any geometric locality , only that each spin interacts with at most @xmath150 other spins for any given interaction term . however , we restrict ourselves to @xmath151 so that each @xmath53 can be specified by an efficient classical description . [ def : dos ] let @xmath152 , @xmath57 , and let @xmath153 be a @xmath83-local hamiltonian such that @xmath25 has no eigenvalues in the intervals @xmath154 $ ] and @xmath155 $ ] . then , the problem dos ( density of states ) is to compute the number of orthogonal eigenstates with eigenvalues in the interval @xmath55 $ ] . let @xmath156 , with @xmath157 , and let @xmath153 be a @xmath83-local hamiltonian s.th . @xmath158 , and @xmath25 has no eigenvalues between @xmath58 and @xmath59 . then , the problem @xmath159 ( sharp local hamiltonian ) is to compute the dimension of the eigenspace with eigenvalues @xmath160 . note that # lh depends on the `` energy splitting '' @xmath161 of the low - energy subspace . in particular , for @xmath162 , @xmath163 corresponds to computing the degeneracy of the ground state subspace . as we will see in what follows , the class @xmath164 is the same for any splitting @xmath165 . we now show that # lh and dos are both -complete . we do so by giving reductions from @xmath166 to dos , from dos to , and from to @xmath167 ; this will at the same time prove the claimed independence of @xmath164 of the splitting @xmath168 . if we denote the parameters of the # lh problem by @xmath169 , then we can simply relate them to the parameters @xmath170 of a dos problem by @xmath171 , @xmath172 and @xmath173 , and the result follows directly . we start with a @xmath83-local hamiltonian @xmath25 as in def . [ def : dos ] . now define a new hamiltonian @xmath174 @xmath175 is a @xmath176-local hamiltonian ; here , @xmath177 is chosen such that each term in @xmath175 is subnormalized . any eigenvalue of @xmath25 in the interval @xmath178 $ ] translates into an eigenvalue of @xmath175 which is below @xmath179 whereas any eigenvalue outside @xmath180 $ ] translates into an eigenvalues of @xmath175 above @xmath181 the original dos problem now translates into counting the number of eigenstates of @xmath175 with negative energy , given a spectral gap in a @xmath26 sized interval @xmath182 $ ] around zero . we now use the circuit which was used in @xcite to prove that @xmath183-local hamiltonian is in ; it accepts any input state @xmath2 with probability @xmath184 where @xmath6 is the number of terms in @xmath175 . [ the idea is to randomly pick one term @xmath185 in the hamiltonian and toss a coin with probability @xmath186 . ] computing the answer to the original dos problem then translates to counting the number of states with acceptance probability @xmath187 , given that there are no eigenstates between @xmath33 and @xmath188 , and since @xmath189 , this shows that this number can be computed in . to show -hardness of # lh , we need to start with an arbitrary verifier circuit @xmath191 and construct a hamiltonian with as many ground states as the circuit has accepting inputs . by amplification , we can assume that the acceptance and rejection thresholds for the verifier are @xmath192 and @xmath193 , where we can choose @xmath194 for any polynomial @xmath110 . as before , let @xmath44 and @xmath45 be the eigenspaces of @xmath32 with eigenvalues @xmath195 and @xmath196 , respectively . define @xmath197 : = \{u{\lvert\psi\rangle}_i { \lvert\mathbf{0}\rangle}_a : { \lvert\psi\rangle}_i \in { \mathcal{r}}\}\,\end{aligned}\ ] ] and denote the projector onto this space by @xmath198}$ ] . notice that for any state @xmath199 $ ] , due to our rejection threshold @xmath200 , we have @xmath201 we now follow kitaev s original construction to encode a verifier circuit into a hamiltonian which has a `` proof history '' as its ground state for any proof @xmath202 @xcite . that is , the ground states of the hamiltonian are given by @xmath203 where the third register is used as a `` clock '' . the hamiltonian has the form * @xmath205 checks that the ancilla is property initialized , penalizing states without properly initialized ancillas ; * @xmath206 + @xmath207 + checks that the propagation from time @xmath70 to @xmath71 is correct , penalizing states with erroneous propagation ; * @xmath208}\otimes { \lvertt\rangle}{\langlet\rvert}_t$ ] causes each state @xmath209 built from an input @xmath210 ( but which nonetheless has a correctly initialized ancilla ) to receive an energy penalty . as we show in lemma [ l : blk - diag ] , the total hamiltonian @xmath25 has a spectral gap @xmath26 above the @xmath211dimensional ground state subspace . however , @xmath75 is not a local hamiltonian , but as we argue in the following , it can be replaced by the usual term @xmath212 while keeping the ground space dimension ( up to small splitting in energies ) and the @xmath26 spectral gap . as we prove in lemma [ l : spec - bound ] , @xmath213 thus , replacing @xmath75 by @xmath214 will decrease the energy of any excited state by at most @xmath215 . ( the energy of the ground states is already minimal and can not decrease . ) on the other hand , the energy of any proper proof history for @xmath25 can not increase by more than @xmath216 due to our choice of acceptance threshold , i.e. , the low energy subspace has dimension @xmath81 . we thus obtain a hamiltonian with a @xmath81 dimensional ground state subspace with energy splitting @xmath217 and a @xmath26 spectral gap above . our proof follows closely the discussion in ref . @xcite , cf . also @xcite . we can block diagonalize @xmath25 by the ( conjugate ) action of the following unitary operator , @xmath218 which maps @xmath219 . as this has no effect on the spectrum , we can work with the simpler @xmath175 from now on . let us explicitly write the effect of conjugation by @xmath220 on the terms of @xmath25 . the first term is unaffected , @xmath221 . the final term becomes @xmath222 , where @xmath223 is simply the projector onto the space @xmath45 , and the ancillas are in the correct initial state . we can now conjugate each of the terms in @xmath224 separately . for example , the first term gives @xmath225 the other terms are exactly analogous , and we find that @xmath226\,.\end{aligned}\ ] ] the total evolution hamiltonian is then block - diagonal a matrix which looks like a hopping hamiltonian in the clock register , @xmath227 where the @xmath228-by-@xmath228dimensional tri - diagonal matrix @xmath229 is given by @xmath230 to discuss the spectrum of @xmath175 , let @xmath231 and @xmath232 the orthogonal complement of @xmath233 ; i.e. , @xmath234 corresponds to evolutions starting from good proofs , @xmath235 to those starting from wrong proofs , and @xmath232 to evolutions with wrongly initialized ancillas . note that @xmath236 , thus , we can analyze the spectrum for @xmath237 separately . note that the restriction to @xmath238 does not affect @xmath239 . since we expect the ground state subspace to occur on @xmath234 , we need to compute ground state energy and gap of @xmath240 , as well as lower bound the ground state energies of @xmath241 and @xmath242 . first , @xmath243 , i.e. , the spectrum of @xmath244 equals the spectrum of @xmath229 , which can be straightforwardly determined to be @xmath245 , with @xmath246 , and eigenfunctions @xmath247 ; the ground state degeneracy is indeed @xmath248 as desired . on the other hand , @xmath249 , and @xmath250 , i.e. , in both cases the ground state energy of @xmath251 is lower bounded by the ground state energy of @xmath252 which has eigenvalues @xmath253 , with @xmath254 , and eigenfunctions @xmath255 . we begin by recalling the result due to jordan @xcite ( see ref . @xcite for a more modern treatment ) for the simultaneous canonical form of two projectors . in the subspace where @xmath260 and @xmath261 commute , both operators are diagonal in a common basis and the spectrum is either @xmath264 , or @xmath265 , and direct sums of those terms . in the subspace where they do nt commute , the problem decomposes into a direct sum of two - by - two blocks given by @xmath266 where @xmath267 for some angle @xmath268 , @xmath269 , and the subscript @xmath270 just labels a generic block . in fact , this two - by - two block form is completely general if we allow embedding our projectors into a larger space while preserving their rank . the rank - preserving condition guarantees that our bound is unchanged , since we are only appending blocks of zeros , and so we will consider this two - by - two form without loss of generality . the constraint that @xmath271 implies constraints on the values that @xmath272 can take . in particular , we can directly compute this operator norm in each block separately , and we find that for all @xmath270 @xmath273 we can also directly compute in each block that @xmath274 the spectrum of this operator is easily computed to be @xmath275 . thus , the least eigenvalue of @xmath276 is bounded from below by @xmath277 . we finally want to relate to the classical counting class . it is clear that any problem can be solved in by using a classical circuit . we will now show that conversely , can be reduced to under _ weakly parsimonious reductions _ : that is , for any function @xmath278 there exist polynomial - time computable functions @xmath89 and @xmath90 , and a function @xmath91 , such that @xmath279 . this differs from karp reductions where no postprocessing is allowed , @xmath93 , but still only requires a single call to a oracle , in contrast to turing reductions . we start from a verifier operator @xmath32 for a problem . first , we use strong error reduction to let @xmath280 and @xmath281 . it follows that @xmath282 and thus we need to compute @xmath95 to accuracy @xmath283 . this can be done using the `` path integral '' method previously used to show containments of quantum classes inside and @xcite . we rewrite @xmath284 , where the sum is over products of transition probabilities along a path , which we label @xmath285 so that @xmath286{{\lvert{i_t}\rangle\!\langle{i_t}\rvert } } u_t \cdots u_1 { \lverti_0\rangle}_i { { \lvert\bm{0}\rangle}_a}\ , . \nonumber\end{aligned}\ ] ] ( cf . fig . 1 in the main manuscript for an illustration ) . since any quantum circuit can be recast in terms of real gates at the cost of doubling the number of qubits @xcite , we can simplify the proof by assuming @xmath100 to be real . to achieve the desired accuracy it is sufficient to approximate @xmath287 up to @xmath288 digits , where @xmath289 is the number of bits in @xmath290 . now define @xmath291\ ] ] and note that @xmath292 is a positive and integer - valued function satisfying @xmath293 - \sum_\zeta f(\zeta)\bigr\rvert \le \tfrac14 \,.\ ] ] finally , by defining a boolean indicator function , @xmath294 we can write @xmath295 , and thus , @xmath296 this shows that @xmath95 can be approximated to accuracy @xmath96 , and thus @xmath81 can be determined by counting the number of satisfying assignments of a single boolean function @xmath297 that can be efficiently constructed from @xmath32 , i.e. , by a single query to a black box solving problems , together with the efficient postprocessing described by eq . ( [ eq : one - more - approx ] ) .	we study the computational difficulty of computing the ground state degeneracy and the density of states for local hamiltonians . we show that the difficulty of both problems is exactly captured by a class which we call , which is the counting version of the quantum complexity class . we show that is not harder than its classical counting counterpart , which in turn implies that computing the ground state degeneracy or the density of states for classical hamiltonians is just as hard as it is for quantum hamiltonians . understanding the physical properties of correlated quantum many - body systems is a problem of central importance in condensed matter physics . the density of states , defined as the number of energy eigenstates per energy interval , plays a particularly crucial role in this endeavor . it is a key ingredient when deriving many thermodynamic properties from microscopic models , including specific heat capacity , thermal conductivity , band structure , and ( near the fermi energy ) most electronic properties of metals . computing the density of states can be a daunting task however , as it in principle involves diagonalizing a hamiltonian acting on an exponentially large space , though other more efficient approaches which might take advantage of the structure of a given problem are not a priori ruled out . in this letter , we precisely quantify the difficulty of computing the density of states by using the powerful tools of quantum complexity theory . quantum complexity aims at generalizing the well - established field of classical complexity theory to assess the difficulty of tasks related to quantum mechanical problems , concerning both the classical difficulty of simulating quantum systems as well as the fundamental limits to the power of quantum computers . in particular , quantum complexity theory has managed to explain the difficulty of computing ground state properties of quantum spin systems in various settings , such as two - dimensional ( 2d ) lattices @xcite and even one - dimensional ( 1d ) chains @xcite , as well as fermionic systems @xcite . we will determine the computational difficulty of two problems : first , computing the density of states of a local hamiltonian , and second , counting the ground state degeneracy of a local gapped hamiltonian ; in both cases , the result holds even if the hamiltonian is restricted to act on a 2d lattice of qubits , or on a 1d chain . to this end , we will introduce the quantum counting class ( sharp bqp ) , which constitutes the natural counting version of the class ( quantum merlin arthur ) which itself captures the difficulty of computing the ground state energy of a local hamiltonian @xcite . vaguely speaking , counts the number of possible `` quantum solutions '' to a quantum problem that can be verified using a quantum computer . we show that both problems , computing the density of states and counting the ground state degeneracy , are complete problems for the class , i.e. , they are among the hardest problems in this class . having quantified the difficulty of computing the density of states and counting the number of ground states , we proceed to relate to known classical counting complexity classes , and show that equals ( under weakly parsimonious reductions ) . here , the complexity class counts the number of satisfying assignments to any efficiently computable boolean function . this can be a very hard problem which is believed to take exponential time ; in particular , it is at least as hard as deciding whether the function has at least one satisfying input , i.e. , the complexity class . examples for -complete problems ( i.e. , the hardest problems in that class ) include counting the number of colorings of a graph , or computing the permanent of a matrix with binary entries . phrased in terms of hamiltonians , what we show is that computing the density of states and counting the ground state degeneracy of a classical spin system is just as hard as solving the same problem for a quantum hamiltonian . _ _ quantum complexity classes.__let us start by introducing the relevant complexity classes . the central role in the following is taken by the _ verifier _ @xmath0 , which verifies `` quantum solutions '' ( also called _ proofs _ ) to a given problem . more formally , a verifier checking an @xmath1-qubit quantum proof ( that is , a quantum state @xmath2 ) consists of a @xmath3 length quantum circuit @xmath4 ( with local gates @xmath5 ) acting on @xmath6 qubits , which takes the @xmath1-qubit quantum state @xmath7 as an input , together with @xmath8 initialized ancillas , @xmath9 , applies @xmath10 , and finally measures the first qubit in the @xmath11 basis to return @xmath12 ( `` proof accepted '' ) or @xmath13 ( `` proof rejected '' ) . then , the class contains all problems of the form : `` decide whether there exists a @xmath2 such that @xmath14 , or whether @xmath15 for all @xmath2 , for some chosen @xmath16 , given that one is the case '' . here , the acceptance probability of a state @xmath2 is @xmath17 , with @xmath18 which we illustrate in fig . [ f : circuit ] . a verifier consists of a sequence of @xmath19 local unitary gates acting on the `` quantum proof '' @xmath2 and an ancillary register initialized to @xmath20 . the final measurement on the first qubit returns @xmath21 or @xmath22 to accept or reject the proof , respectively . transition probabilities can be computed by doing a `` path integral '' over all intermediate configurations @xmath23 . ] the idea behind this definition is that quantifies the difficulty of computing the ground state energy @xmath24 of a local hamiltonian @xmath25 up to @xmath26 accuracy . let the verifier be a circuit estimating @xmath27 ; then a black box solving problems can be used to compute @xmath24 up to @xmath26 accuracy by binary search using a single query . note also that is the quantum version of the class , where one is given an efficiently computable boolean function @xmath28 and one must figure out if there is an @xmath29 such that @xmath30 . the class has a natural counting version , known as . here , the task is to determine the _ number _ rather than the existence of satisfying inputs , i.e. , to compute @xmath31 . we will now analogously define , the counting version of . consider the verifying map @xmath32 of eq . for a problem , with the additional promise that @xmath32 does not have eigenvalues between @xmath33 and @xmath34 , @xmath16 . then the class consists of all problems of the form `` compute the dimension of the space spanned by all eigenvectors with eigenvalues @xmath35 '' . an equivalent definition for ( cf . also @xcite ) is the following : consider a verifier @xmath32 with the additional promise that there exist subspaces @xmath36 such that @xmath37 for all @xmath38 , and @xmath39 for all @xmath40 , where again @xmath16 we can think of @xmath41 and @xmath42 as containing the good and bad witnesses , respectively . ( note that there will always be `` mediocre '' witnesses the question is whether there exists a decomposition into a good and a bad witness space . ) then , consists of all problems of the form `` compute @xmath43 '' . this number is well - defined , i.e. , independent of the choice of @xmath44 and @xmath45 , and moreover , one can easily show that it is equivalent to the definition above , cf . the supplementary material . the gap promise we impose on the spectrum of @xmath32 is not present in the definition of ( though similarly restricted versions of were defined in @xcite ) . nevertheless , this promise emerges naturally when considering the counting version : captures the difficulty of determining the existence of an input state with acceptance probability above @xmath33 , up to a `` grace interval '' @xmath46 $ ] in which mistakes are tolerated ( i.e. , if the largest eigenvalue of @xmath32 is in @xmath46 $ ] , the oracle can return either outcome ) . correspondingly , captures the difficulty of counting the number of eigenvalues above @xmath33 , where eivenvalues in the grace interval @xmath46 $ ] can be miscounted . the reason why we choose to define with a gap promise rather than with a grace interval is the same as for , namely to have a unique outcome associated with any input . similarly , the idea of the hamiltonian formulation of the problem which we will discuss below is to ask for the number of eigenstates in a certain energy interval , where states which are in some small @xmath26 neighborhood of this interval may be miscounted ; again , for reasons of rigor we choose to consider only hamiltonians with no eigenstates in that interval . it should be noted , however , that all of the equivalence proofs we give equally apply if we choose to allow for miscounting of states in those grace intervals instead of requiring them to be empty , as the proofs do not make use of the gap promise itself , but rather show that all states outside those grace intervals are mapped ( and thus counted ) correctly . thus , while the actual number returned by the grace interval formulation of the counting problems might change under those mappings due to different treatment of states in the grace interval , it will still be in the correct range . the class inherits the important property from of being stable under amplification , that is , the definition of is not sensitive to the choice of @xmath33 and @xmath34 . in particular , any @xmath16 can be amplified ( by building a new poly - size @xmath47 from @xmath32 ) such that @xmath48 , @xmath49 , and keeping the eigenvalue gap between @xmath50 and @xmath51 , by using a construction called _ strong amplification _ , cf . ref . @xcite ; as shown there , strong amplification acts on all eigenvalues independently and thus also applies to . the crucial point is that strong amplification works without changing the proof itself , compared to weak amplification which takes multiple copies of the proof as an input . while this is fine for , it does change the dimension of the accepting subspace in an unpredictable way and is thus not an option for the amplification of . _ _ complexity of computing the density of states.__let us now show why the class is relevant for physical applications . in particular , we are going to show that computing the density of states of a local @xmath1-spin hamiltonian @xmath52 with few - body terms @xmath53 , @xmath54 , up to accuracy @xmath26 , is a problem which is complete for , i.e. , it is as hard as any problem in can be . the same holds true for the ( a priori weaker ) problem of counting the ground state degeneracy of a local hamiltonian , given a @xmath26 spectral gap above ( note that bravyi _ et al . _ @xcite suggested this as a definition for a quantum counting class ) . we can impose additional restrictions on the interaction structure of our hamiltonian , and as we will see , the hardness is preserved even for 2d lattices of qubits , or 1d systems . the problem dos ( density of states ) is defined as follows : given a local hamiltonian @xmath52 , compute the number of orthogonal eigenstates with eigenvalues in an interval @xmath55 $ ] with @xmath56 , where we do not allow for eigenvalues within a small grace interval of width @xmath57 centerd around @xmath58 and @xmath59 ; alternatively , we can allow for errorneous counts of eigenstates in that interval . second , the problem # lh ( sharp local hamiltonian ) corresponds to counting the number of ground states of a local hamiltonian which has a spectral gap @xmath60 above the ground state subspace , given we are told the ground state energy , and where we allow for a small splitting in the ground state energies ; again , we can alternatively allow to miscount states in the grace interval . clearly , # lh is a special instance of dos , i.e. , solving # lh can be reduced to solving dos . in order to show that dos is contained in , we can use a phase estimation circuit @xcite to estimate the energy of any given input @xmath2 and only accept if its energy @xmath27 is in the interval @xmath55 $ ] ; as the desired accuracy @xmath60 , this can be done efficiently . a detailed proof ( using a more elementary circuit ) is given in the supplementary material . let us now conversely show that # lh is a hard problem for , that is , any problem in can be reduced to counting the ground states of some gapped local hamiltonian gap has been shown in @xcite . ] . as in turn # lh can be reduced to dos , which is contained in , this proves that both # lh and dos are complete problems for , i.e. , they capture the full difficulty of this class . to this end , we need to start from an arbitrary verifier circuit @xmath4 and construct a hamiltonian which has as many ground states as the circuit has accepting inputs ( corresponding to the outcome @xmath61 on the first qubit ) . let @xmath44 and @xmath45 be the eigenspaces of @xmath32 [ eq . ] with eigenvalues @xmath62 and @xmath63 , respectively , and define @xmath64:=\{u{\lvert\psi\rangle}_i{{\lvert\bm{0}\rangle}_a}:{\lvert\psi\rangle}_i\in \mathcal r\}$ ] . we will follow kitaev s original construction for a hamiltonian encoding a verifier circuit @xcite , which for any proof @xmath65 has the `` proof history '' @xmath66 as its ground state , where the third register is used as a clock . the hamiltonian @xmath67 has three types of terms : @xmath68 makes sure the ancilla is initialized , @xmath69 ensures proper evolution from @xmath70 to @xmath71 , and @xmath72}\otimes { { \lvert{t}\rangle\!\langle{t}\rvert}}_t$ ] gives an energy penalty to states @xmath73 built from proofs @xmath74 . note that our @xmath75 differs from the usual choice @xmath76 and is in fact non - local ; as we show in the supplementary material , this does not significantly change the relevant spectral properties ( in particular , we keep the @xmath26 gap , and the ground state subspace is split at most exponentially ) . with this choice of @xmath75 , @xmath25 acts independently on the subspaces spanned by @xmath77 for any @xmath78 or @xmath79 , and @xmath80 the computational basis , and the restriction of @xmath25 to any of these subspaces describes a random walk which is characterized by the choice of @xmath7 and the number of @xmath12 s in @xmath80 . these cases can be analyzed independently ( see supplementary material ) , and it follows that @xmath25 has a @xmath81-fold degenerate ground state space with a @xmath26 gap above , proving -hardness of # lh . this shows that # lh is -hard for a hamiltonian which is a sum of @xmath82-local terms ( i.e. , each term acts on @xmath82 sites ) , as the clock register is of size @xmath82 . in order to obtain a @xmath83-body hamiltonian , kitaev suggested to use a unary encoding of the clock ( i.e. , @xmath84 is encoded as @xmath85 , with @xmath71 1 s ) , so that each hamiltonian term only acts on three qubits of the clock . however , this makes the hilbert space of the clock too big , and terms need to be added to the hamiltonian to penalize illegal clock configurations . these terms divide the hilbert space into two parts , @xmath86 . here , @xmath87 contains only legal clock states , whereas @xmath88 contains only configurations with illegal clock states @xcite . since no hamiltonian term couples these two subspaces , the hamiltonian can be analyzed independently on the two subspaces . it turns out that its restriction to @xmath88 has an at least @xmath26 higher energy , while on @xmath87 , the hamiltonian is exactly the same as before . thus , one finds that the new hamiltonian still has the right number of ground states , and a @xmath26 spectral gap . the very same argument applies in the case of 1d hamiltonians , using the -construction of ref . @xcite : again , the hamiltonian acts independently on a `` legal '' and an `` illegal '' subspace , where the latter has a polynomially larger energy , and the former reproduces the ( low - energy ) spectrum of the original hamiltonian @xcite . an alternative way to prove -hardness on restricted interaction graphs is to use so - called _ perturbation gadgets _ , which yield the hamiltonian of the kitaev construction above from a perturbative expansion ; in particular , this way one can show -hardness of pauli - type nearest - neighbor hamiltonians on a 2d square lattice of qubits @xcite . as shown in ref . @xcite , such gadgets do in fact approximately preserve the whole low - energy part of the spectrum , and thus , our hardness proof for # lh still applies to these classes of hamiltonians . it should be noted , however , that since the eigenvalues are only preserved up to an error @xmath26 , the splitting of the ground state space will now be of order @xmath26 ; however , it can still be chosen to be polynomially smaller than the spectral gap . _ _ quantum vs. classical counting complexity.__as we have seen , the quantum counting class exactly captures the difficulty of counting the degeneracy of ground states and computing the density of states of local quantum hamiltonians . in the following , we will relate to classical counting classes and prove that is equal to , counting the number of satisfying inputs to a boolean function there exist polynomial - time computable functions @xmath89 and @xmath90 , and a function @xmath91 , such that @xmath92 . this differs from karp reductions where no postprocessing is allowed , @xmath93 , but it still only requires a single call to a oracle . ] . in physical terms , this shows that counting the number of ground states or determining the density of states for a quantum hamiltonian is not harder than either problem is for a classical hamiltonian . clearly , is contained in , as the latter includes classical verifiers . it remains to be shown that any problem can be solved by computing a function . we start from a verifier operator @xmath32 , eq . , and wish to show that the dimension of its accepting subspace , i.e. , the subspace @xmath44 with eigenvalues @xmath35 , can be computed by counting satisfying inputs to some efficiently computable boolean function . using amplification , we can ensure that @xmath94 , i.e. , we need to compute @xmath95 to accuracy @xmath96 . this can be done using a `` path integral '' method , which has been used previously to show containments of quantum classes in the classical classes and ( see e.g. @xcite ) . we rewrite @xmath97 as a sum over products of transition probabilities along a path @xmath98 , where @xmath99{{\lvert{i_t}\rangle\!\langle{i_t}\rvert } } u_t \cdots u_1 { \lverti_0\rangle}_i { { \lvert\bm{0}\rangle}_a}\nonumber\end{aligned}\ ] ] ( cf . fig . [ f : circuit ] ) . since any such sum over an efficiently computable @xmath100 can be determined by counting the satisfying inputs to some boolean formula ( see the supplementary material for details ) , it follows that @xmath32 can be computed using a single query to a black box solving problems . _ _ summary and discussion.__in this work , we considered two problems : computing the density of states and computing the ground state degeneracy of a local hamiltonian of a spin system . in order to capture the computational difficulty of these problems we introduced the quantum complexity class , the counting version of the class . we proved that this complexity class exactly captures the difficulty of our two problems , even when restricting to local hamiltonians on 2d lattices of qubits or to 1d chains , since all these problems are complete problems for the class . we have further shown that is no harder than its classical counterpart . in particular this implies that computing the density of states is no harder for quantum hamiltonians than it is for classical ones . while this quantum - classical equivalence might seem surprising at the hamiltonian level , it should be noted that the classes and quite often form natural `` upper bounds '' for many quantum _ and _ classical problems . what about the problem of computing the density of states for fermionic systems , such as many - electron systems ? on the one hand , this problem will be still in and thus , since any local fermionic hamiltonian can be mapped via the jordan - wigner transform to a ( non - local ) hamiltonian on a spin system , whose energy can still be estimated efficiently by a quantum circuit @xcite . on the other hand , hardness of the problem for can be shown e.g. by using the -hardness of # lh , and encoding each spin using one fermion in two modes , similar to @xcite . thus , computing the density of states for fermionic systems is a -complete problem as well . _ _ acknowledgments.__we thank s. aaronson , d. gottesman , d. gross , y. shi , m. van den nest , j. watrous and s. zhang for helpful discussions and comments . research at pi is supported by the government of canada through industry canada and by the province of ontario through the ministry of research and innovation . ns is supported by the gordon and betty moore foundation through caltech s center for the physics of information and nsf grant no . . _ _ note added.__after completion of this work , we learned that shi and zhang @xcite have independently defined and shown its relation to @xmath101 using the same technique . 10 r. oliveira and b. m. terhal , quant . inf . comput . * 8 * , 900 ( 2009 ) , quant - ph/0504050 . d. aharonov , d. gottesman , s. irani , and j. kempe , commun . math . phys . * 287 * , 41 ( 2009 ) , arxiv:0705.4077 . n. schuch and f. verstraete , nature physics * 5 * , 732 ( 2009 ) , arxiv:0712.0483 . a. y. kitaev , a. h. shen , and m. n. vyalyi , ( american mathematical society , providence , rhode island , 2002 ) . d. aharonov and t. naveh , ( 2002 ) , quant - ph/0210077 . d. aharonov , m. ben - or , f. g. brandao , and o. sattath , ( 2008 ) , arxiv:0810.4840 . r. jain , i. kerenidis , m. santha , and s. zhang , ( 2009 ) , arxiv:0906.4425 . c. marriott and j. watrous , comput . complex . * 14 * , 122 ( 2005 ) , cs/0506068 . s. bravyi , c. moore , and a. russell , ( 2009 ) , arxiv:0907.1297 . r. cleve , a. ekert , c. macchiavello , and m. mosca , proc . r. soc . lond . * 454 * , 339 ( 1998 ) , quant - ph/9708016 . n. schuch , i. cirac , and f. verstraete , phys . rev . lett . * 100 * , 250501 ( 2008 ) , arxiv:0802.3351 . j. kempe , a. kitaev , and o. regev , siam journal of computing * 35 * , 1070 ( 2006 ) , quant - ph/0406180 . l. adleman , j. demarrais , and m .- d . huang , siam j. comput . * 26 * , 1524 ( 1997 ) . y .- k . liu , m. christandl , and f. verstraete , phys . rev . lett . * 98 * , 110503 ( 2007 ) , quant - ph/0609125 . y. shi and s. zhang , private communication ; notes available at ` http://www.cse.cuhk.edu.hk/ ` ` ~syzhang / papers / sharpbqp.pdf ` . r. bhatia , ( springer , 1997 ) . c. jordan , bulletin de la socit mathmatique de france * 3 * , 103 ( 1875 ) . p. halmos , trans . amer . math . soc . * 144 * , 381 ( 1969 ) . e. bernstein and u. vazirani , siam j. of comp . * 26 * , 1411 ( 1997 ) , quant - ph/9701001 . f.g.s.l . brandao , phd thesis ( 2008 ) , arxiv:0810.0026 . t. j. osborne , ( 2006 ) , cond - mat/0605194 .
to introduce the main theme of this paper we recall the following theorem of casselman @xcite . let @xmath0 be a non - archimedean local field whose ring of integers is @xmath9 . let @xmath10 be the maximal ideal of @xmath9 . let @xmath11 be a non - trivial additive character of @xmath0 which is normalized so that the maximal fractional ideal on which it is trivial is @xmath9 . [ thm : casselman ] let @xmath12 be an irreducible admissible infinite - dimensional representation of @xmath3 . let @xmath13 denote the central character of @xmath5 . let @xmath14 let @xmath15 @xmath16 1 . there exists a non - negative integer @xmath17 such that @xmath18 if @xmath6 denotes the least non - negative integer @xmath17 with this property then the epsilon factor @xmath19 of @xmath5 is up to a constant multiple of the form @xmath20 . ( here @xmath21 is the cardinality of the residue field of @xmath0 . ) 2 . for all @xmath22 we have @xmath23 . the assertion @xmath24 is sometimes referred to as _ multiplicity one theorem for newforms _ and the unique vector ( up to scalars ) in @xmath25 is called the _ newform _ for @xmath26 this is closely related to the classical atkin lehner theory of newforms for holomorphic cusp forms on the upper half plane @xcite . when @xmath27 we have a spherical representation and the newform is nothing but the spherical vector . newforms play an important role in the theory of automorphic forms . we cite two examples to illustrate this . first , the zeta integral corresponding to the newform is exactly the local @xmath7-factor associated to @xmath5 ( see @xcite for instance ) . in addition , newforms frequently play the role of being ` test vectors ' for interesting linear forms associated to @xmath5 . for example , the newform is a test vector for an appropriate whittaker linear functional . in showing this , explicit formulae for newforms are quite often needed . for instance , if @xmath5 is a supercuspidal representation which is realized in its kirillov model then the newform is the characteristic function of the unit group @xmath28 . this observation is implicit in casselman @xcite and is explicitly stated and proved in shimizu @xcite . since the whittaker functional on the kirillov model is given by evaluating functions at @xmath29 , we get in particular that the functional is non - zero on the newform . in a related vein @xcite and @xcite show that test vectors for trilinear forms for @xmath3 are often built from newforms . ( see also a recent expository paper of schmidt @xcite where many of these results are documented . ) in addition to casselman s theory for @xmath3 , newforms have been studied for certain other classes of groups . jacquet _ et al _ @xcite have developed a theory of newforms for _ generic _ representations of @xmath4 . in this setting , there is no satisfactory statement analogous to ( ii ) of the above theorem . however , in his recent thesis , mann @xcite obtained several results on the growth of the dimensions of spaces of fixed vectors and has a conjecture about this in general . for the group @xmath30 , @xmath31 a @xmath32-adic division algebra , prasad and raghuram @xcite have proved an analogue of casselman s theorem for irreducible principal series representations and supercuspidal representations coming via compact induction . in an unpublished work , brooks roberts has proved part of ( i ) of the above for representations of @xmath33 whose langlands parameter is induced from a two - dimensional representation of the weil deligne group of @xmath34 in a previous paper @xcite , we develop a theory of conductors and newforms for @xmath2 . in this paper we use the results of @xcite to carry out a similar program for the unramified quasi split unitary group@xmath35 . let @xmath36 crucial to our study of newforms are certain filtrations of maximal compact subgroups of @xmath37 let @xmath38 be the standard hyperspecial maximal compact subgroup of @xmath39 . let @xmath40 , where @xmath41 then @xmath42 and @xmath43 are , up to conjugacy , the two maximal compact subgroups of @xmath39 . we define filtrations of these maximal compact subgroups as follows . for @xmath17 an integer @xmath44 let @xmath45 @xmath16 let @xmath46 be an irreducible admissible infinite - dimensional representation of @xmath37 let @xmath47 denote the center of @xmath39 and let @xmath48 be the central character of @xmath49 . let @xmath50 be any character of @xmath51 such that @xmath52 on the center . let @xmath53 denote the conductor of @xmath54 for any @xmath55 @xmath50 gives a character of @xmath56 and also @xmath57 given by @xmath58 we define for @xmath59 , @xmath60 the space @xmath61 is defined analogously . we define the _ @xmath50-conductor _ @xmath62 of @xmath49 as @xmath63 we define the _ conductor @xmath64 _ of @xmath49 by @xmath65 where @xmath50 runs over characters of @xmath51 which restrict to the central character @xmath48 on @xmath47 . we deal with the following basic issues in this paper . 1 . given an irreducible representation @xmath49 , we determine its conductor @xmath66 a very easy consequence ( almost built into the definition ) is that the conductor depends only on the @xmath7-packet containing @xmath49 . we identify the conductor with other invariants associated to the representation . for instance , for @xmath2 we have shown @xcite that the conductor of a representation is same as the conductor of a minimal representation of @xmath3 determining its @xmath7-packet . we prove a similar result for @xmath1 in this paper . [ sec : comparison ] and [ sec : comparison - u11 ] . 3 . we determine the growth of the space @xmath67 as a function of @xmath17 . this question is analogous to ( ii ) of casselman s theorem quoted above . computing such dimensions is of importance in ` local level raising ' issues . see @xcite . we address the question of whether there is a multiplicity one result for newforms . it turns out that quite often @xmath68 , but this fails in general ( for principal series representations of a certain kind ) . in these exceptional cases the dimension of the space of newforms is two , but a canonical quotient of this two - dimensional space has dimension one ( see [ sec : multiplicity - one ] ) . 5 . are the newforms test vectors for whittaker functionals ? this is important in global issues related to newforms . we are grateful to benedict gross for suggesting this question to us . it turns out that our newforms are always test vectors for whittaker functionals . in the proofs we often need explicit formulae for newforms in various models for the representations . these formulas are interesting for their own sake . for example , if @xmath12 is a ramified supercuspidal representation of @xmath1 , then the newform can be taken as the characteristic function of @xmath69 where @xmath70 is regarded as a subspace of the kirillov model of a canonically associated minimal representation of @xmath3 ( cf . @xcite ) . we set up notation in [ sec : notation ] following that used in @xcite . we then briefly review the structure of @xmath7-packets for @xmath71 and @xmath35 in [ sec : packets - sl2-u11 ] . as this paper depends crucially on our previous paper @xcite on @xmath71 , we summarize the results of @xcite in [ sec : newforms ] . the heart of this paper is [ sec : unitary ] . in [ sec : defns - u11 ] we define the notion of conductor and then make some easy but technically important remarks on spaces of fixed vectors . the next two subsections deal respectively with sub - quotients of principal series representations and supercuspidal representations . in @xcite , we use kutzko s construction of supercuspidal representations of @xmath3 to obtain results for supercuspidals of @xmath2 . in this paper , we use these results , in turn , to obtain information for @xmath1 . in general , we will often reduce the proofs of statements concerning @xmath1 to those of the corresponding @xmath2 statements . in particular , we exploit the fact that @xmath2 is the derived group of @xmath1 and that @xmath72 has index two in @xmath73 in this way we avoid directly dealing with @xmath74-types and other intrinsic details for @xmath75 as much of the work has been done for @xmath2 in @xcite . finally , in [ sec : multiplicity - one ] , we briefly discuss a multiplicity one result fornewforms . we mention some further directions that arise naturally from this work . to begin with , it would be interesting to see how our theory of newforms and conductors bears upon known results about local factors for @xmath1 . in particular , are our conductors the same as ( or closely related to ) the analytic conductors appearing in epsilon factors ? also , is a zeta - integral corresponding to a newform of a representation equal to a local @xmath7-factor for the representation ? if @xmath7 is any non - archimedean local field let @xmath76 be its ring of integers and let @xmath77 be the maximal ideal of @xmath78 let @xmath79 be a uniformizer for @xmath7 , i.e. , @xmath80 let @xmath81 be the residue field of @xmath82 let @xmath32 be the characteristic of @xmath83 and let the cardinality of @xmath83 be @xmath84 which is a power of @xmath85 let @xmath86 be an element of @xmath87 . if @xmath88 is a positive integer , let @xmath89 denote the @xmath88th filtration subgroup @xmath90 of @xmath91 , and define @xmath92 . let @xmath93 denote the additive valuation on @xmath94 which takes the value @xmath95 on @xmath96 we let @xmath97 denote the normalized multiplicative valuation given by @xmath98 if @xmath99 is a character of @xmath94 we define the conductor @xmath100 to be the smallest non - negative integer @xmath88 such that @xmath99 is trivial on @xmath89 . let @xmath101 be a non - trivial additive character of @xmath7 which is assumed to be trivial on @xmath76 and non - trivial on @xmath102 for any @xmath103 the character given by sending @xmath104 to @xmath105 will be denoted as @xmath106 or simply by @xmath107 ( in all the above notations we may omit the subscript @xmath7 if there is only one field in the context . ) in the following , @xmath0 will be a fixed non - archimedean local field whose residue characteristic is odd and @xmath108 will be used to denote a quadratic extension of @xmath0 . we denote by @xmath109 the quadratic character of @xmath110 associated to @xmath111 by local class field theory . recall that the kernel of @xmath109 is @xmath112 , the norms from @xmath113 we will require the units @xmath114 and @xmath115 to be compatible in the sense that @xmath116 we let @xmath117 denote the group @xmath3 . let @xmath118 be the standard borel subgroup of upper triangular matrices in @xmath117 with levi subgroup @xmath119 and unipotent radical @xmath120 . let @xmath121 be the center of @xmath117 . let @xmath122 let @xmath123 be the standard borel subgroup of upper triangular matrices in @xmath124 with levi subgroup @xmath125 and unipotent radical @xmath120 . we set @xmath126 and @xmath127 and denote by @xmath128 and @xmath129 respectively the standard iwahori subgroups of @xmath124 and @xmath117 . suppose that @xmath111 is unramified , and let @xmath130 denote the non - trivial element of @xmath131 . we denote by @xmath39 the group @xmath35 , i.e. , the group of all @xmath132 such that @xmath133 we let @xmath134 be the standard upper triangular borel subgroup of @xmath39 with diagonal levi subgroup @xmath135 and unipotent radical @xmath120 . we note that the elements of @xmath135 are of the form @xmath136 for @xmath137 , and those of @xmath134 are of the form @xmath138 with @xmath139 and @xmath140 we let @xmath47 be the center of @xmath39 . then @xmath141 , where @xmath142 is the subgroup of norm one elements of @xmath113 denote by @xmath143 the standard iwahori subgroup of @xmath39 and by @xmath144 the standard hyperspecial maximal compact subgroup of @xmath39 . the following filtrations of maximal compact subgroups of @xmath124 will be important in our study of newforms . let @xmath145 and @xmath146 . let @xmath147 , where @xmath148 then @xmath149 and @xmath150 are , up to conjugacy , the two maximal compact subgroups of @xmath124 . for @xmath17 an integer @xmath44 @xmath151 we note that for @xmath152 the following inclusions hold up to conjugacy within @xmath124 : @xmath153 analogous results hold for the following filtration groups of @xmath39 : @xmath154 we note that the filtration subgroups for @xmath124 and @xmath39 are related by @xmath155 where @xmath156 in addition to @xmath157 , we will also make frequent use of the matrices @xmath158 @xmath159 and @xmath160 for any subsets @xmath161 we let @xmath162 = \left\ { \left ( \begin{array}{cc } a & b\\ c & d \end{array}\right ) : a \in a , b \in b , c \in c , d \in d \right\}.\ ] ] we denote @xmath163 $ ] by @xmath164 or simply by @xmath165 we let @xmath166 denote the lower triangular unipotent subgroup of @xmath124 and a similar meaning is given to @xmath167 and @xmath168 if @xmath169 is a closed subgroup of a locally compact group @xmath170 and if @xmath171 is an admissible representation of @xmath169 then @xmath172 denotes _ normalized _ induction , and @xmath173 denotes the subrepresentation of @xmath172 consisting of those functions whose support is compact mod @xmath174 the symbol @xmath175 will denote the trivial representation of the group in context . for any real number @xmath176 we let @xmath177 denote the least integer greater than or equal to @xmath176 and we let @xmath178 in this section we collect statements about the structure of @xmath7-packets for @xmath179 and @xmath180 . all the assertions made here are well - known and can be read off from a combination of labesse and langlands @xcite , gelbart and knapp @xcite and rogawski @xcite . if @xmath181 is an irreducible admissible representation of @xmath117 then its restriction to @xmath124 is a multiplicity free finite direct sum of irreducible admissible representations of @xmath124 which we often write as @xmath182 on the other hand , if @xmath5 is any irreducible admissible representation of @xmath124 then there exists an irreducible admissible representation @xmath181 of @xmath117 whose restriction to @xmath124 contains @xmath26 note that @xmath117 acts on the space of all equivalence classes of irreducible admissible representations of @xmath124 and an _ @xmath7-packet _ for @xmath124 is simply an orbit under this action . it turns out that , with the notation as above , the @xmath7-packets are precisely the sets @xmath183 appearing in the restrictions of irreducible representations @xmath181 of @xmath184 we now give some general statements concerning the @xmath7-packets for @xmath180 . the adjoint group of @xmath35 is @xmath185 , and hence @xmath186 and @xmath117 act via automorphisms on @xmath39 , hence act on the set of equivalence classes of irreducible representations of @xmath39 . rogawski ( @xcite , 11.1 ) defines an @xmath7-packet for @xmath39 to be an orbit under this action . if @xmath49 is an element of a non - trivial @xmath7-packet , then the other element of the @xmath7-packet is @xmath187 . if @xmath188 is an @xmath7-packet for @xmath39 , then the set of irreducible components of the restrictions of elements of @xmath188 to @xmath124 is an @xmath7-packet @xmath8 for @xmath124 . the direct sum @xmath189 is therefore the restriction of an irreducible admissible representation @xmath181 of @xmath184 this @xmath181 is unique up to twisting by a character . in practice , we will choose a convenient @xmath190 since @xmath191 , we obtain an action of @xmath117 on @xmath192 via the represen - tation @xmath181 . this section collects our results @xcite on conductors and newforms for @xmath2 . all these results , along with their complete proofs , can be found in @xcite . we now give our definition of the conductor of a representation of @xmath124 . the basic filtration subgroups of @xmath124 considered in this paper are @xmath194 and for @xmath152 , @xmath195 for all @xmath196 we let @xmath197 let @xmath12 be an irreducible admissible infinite - dimensional representation of @xmath198 let @xmath13 be the character of @xmath199 such that @xmath200 we let @xmath201 be any character of @xmath202 such that @xmath203 let @xmath204 denote the conductor of @xmath205 for any @xmath206 @xmath201 gives a character of @xmath207 and also @xmath208 given by @xmath209 we define @xmath210 the spaces @xmath211 are defined analogously . we note that @xmath212 for @xmath213 . we define the _ @xmath201-conductor _ @xmath214 of @xmath5 as @xmath215 we define the _ conductor @xmath6 _ of @xmath5 by @xmath216 where @xmath201 runs over characters of @xmath202 such that @xmath203 if @xmath201 is such that @xmath217 and @xmath218 ( resp . @xmath219 ) , then we call @xmath220 ( resp . @xmath221 ) a _ space of newforms _ of @xmath5 . in this case , we refer to a non - zero element of @xmath222 or @xmath223 as a _ newform _ of @xmath5 . let @xmath99 be a character of @xmath224 then @xmath99 gives a character of @xmath225 via the formula @xmath226 let @xmath227 stand for the ( unitarily ) induced representation @xmath228 it is well - known that @xmath227 is reducible if and only if @xmath99 is either @xmath229 or if @xmath99 is a quadratic character . there is an essential difference between the two kinds of reducibilities . if @xmath230 , then @xmath227 is the restriction to @xmath124 of a reducible principal series representation of @xmath117 . hence @xmath227 will have two representations in its jordan hlder series , namely the trivial representation and the steinberg representation which we will denote by @xmath231 if @xmath99 is a quadratic character , then @xmath227 is the restriction to @xmath124 of an irreducible principal series representation of @xmath117 and breaks up as a direct sum of two irreducible representations , which constitute an @xmath7-packet of @xmath124 . if @xmath232 we denote @xmath227 by @xmath233 and let @xmath234 we denote the @xmath7-packet by @xmath235 as mentioned in the introduction , one of the applications of newforms we have in mind is that they are test vectors for whittaker functionals . for principal series representations and in fact all their sub - quotients we consider the following @xmath236-whittaker functional ( see @xcite ) . for any function @xmath237 in a principal series representation @xmath227 we define @xmath238 where the haar measure is normalized such that @xmath239 ( unramified principal series representations).[prop : ps - sl2 ] @xmath16 let @xmath99 be an unramified character of @xmath110 and let @xmath227 be the corresponding principal series representation of @xmath198 we have @xmath240 2 m , & \mbox { if $ m \geq 1.$ } \end{array } \right.\ ] ] ( test vectors for unramified principal series representations)[cor : unramified - sl2-test ] @xmath16 for an unramified character @xmath99 of @xmath110 such that @xmath241 let @xmath242 be any non - zero @xmath74-fixed vector . then we have @xmath243 where @xmath244 is the standard local abelian @xmath7-factor associated to @xmath245 ( steinberg representation)[prop : stienberg - sl2 ] @xmath16 if @xmath246 is the steinberg representation of @xmath247 then the dimension of the space of fixed vectors under @xmath207 is given by @xmath248 2m-1 , & \mbox { if $ m \geq 1.$ } \end{array } \right.\ ] ] ( test vectors for the steinberg representation)[cor : steinberg - test ] @xmath16 let the steinberg representation @xmath246 be realized as the unique irreducible subrepresentation of @xmath249 the @xmath236-whittaker functional @xmath250 is non - zero on the space of newforms @xmath251 . ( ramified principal series representations)[prop : sl2-ps - ramified ] @xmath16 let @xmath99 be a ramified character of @xmath224 let @xmath252 be the corresponding principal series representation of @xmath198 let @xmath100 denote the conductor of @xmath245 1 . we have @xmath253 and further @xmath254 only for those characters @xmath201 such that @xmath255 on the group of units @xmath256 2 . if @xmath257 and @xmath258 then @xmath259 1 , & \mbox { if $ m = c(\chi),$}\\[.2pc ] 2(m - c(\chi))+1 , & \mbox { if $ m > c(\chi).$ } \end{array } \right.\ ] ] 3 . if @xmath260 and @xmath261 2 m , & \mbox { if $ m \geq 1 = c(\chi)$. } \end{array } \right.\ ] ] ( test vectors for ramified principal series representations ) [ cor : test - ram - ps]@xmath16 let @xmath99 be a ramified character of @xmath224 let @xmath252 be the corresponding principal series representation of @xmath198 assume that @xmath5 is irreducible . let @xmath262 denote the conductor of @xmath245 the space of newforms @xmath263 is one - dimensional and the whittaker functional @xmath250 is non - zero on this space of newforms . ( ramified principal series @xmath7-packets)[prop : ramified - sl2 ] @xmath16 let @xmath111 be a quadratic ramified extension . let @xmath264 be the corresponding @xmath7-packet . then we have for @xmath265@xmath266 m , & \mbox { if $ m \geq 1.$ } \end{array } \right.\ ] ] ( test vectors for ramified principal series @xmath7-packets ) [ cor : test - ramified - l - packets - sl2]@xmath16 let @xmath111 be a ramified quadratic extension and let @xmath264 be the corresponding @xmath7-packet . then one and only one of the two representations in the packet is @xmath236-generic , say , @xmath267 . then @xmath268 is @xmath269-generic . the whittaker functional @xmath250 is non - zero on the one dimensional space of newforms @xmath270 any @xmath269-whittaker functional is non - zero on the one - dimensional space of newforms for @xmath271 ( unramified principal series @xmath7-packet ) [ prop : unramfified - sl2]@xmath16 let @xmath111 be the quadratic unramified extension . let @xmath272 be the corresponding @xmath7-packet . exactly one of the two representations , say @xmath267 , has a non - zero vector fixed by @xmath149 . then the dimensions of the space of fixed vectors under @xmath207 and @xmath208 for the two representations are as follows : 1 . @xmath273 2 . @xmath274 3 . for @xmath275 , @xmath276 4 . for @xmath277 @xmath278 ( test vectors for unramified principal series @xmath7-packet ) [ cor : test - unramified - lpacket - sl2]@xmath16 let @xmath111 be the unramified quadratic extension , and let @xmath279 be the corresponding @xmath7-packet . then one and only one of the two representations in the packet is @xmath236-generic , namely @xmath267 ( using the notation of proposition [ prop : unramfified - sl2 ] ) . moreover , a @xmath236-whittaker functional is non - zero on the @xmath149-fixed vector in @xmath267 . the representation @xmath268 is not @xmath280-generic for any @xmath280 of conductor @xmath9 . it is @xmath281-generic and any @xmath281-whittaker functional is non - zero on the unique ( up to scalars ) @xmath150-fixed vector in @xmath271 we now consider supercuspidal representations of @xmath122 for this we need some preliminaries on how they are constructed . we use kutzko s construction @xcite of supercuspidal representations for @xmath117 and then moy and sally @xcite or kutzko and sally @xcite to obtain information on the supercuspidal representations ( @xmath7-packets ) for @xmath198 we begin by briefly recalling kutzko s construction of supercuspidal representations of @xmath117 via compact induction from very cuspidal representations of maximal open compact - mod - center subgroups . for @xmath282 , let @xmath283 be the principal congruence subgroup of @xmath284 of level @xmath285 . let @xmath286 let @xmath129 be the standard iwahori subgroup consisting of all elements in @xmath284 that are upper triangular modulo @xmath287 for @xmath282 , let @xmath288 \mathcal{p}^{l+1 } & 1 + \mathcal{p}^l \end{array}\right],\ ] ] and let @xmath289 we will let @xmath290 ( resp . @xmath291 ) denote either @xmath292 ( resp . @xmath284 ) or @xmath293 ( resp . @xmath129 ) . here @xmath293 is the normalizer in @xmath117 of @xmath294 in either case we let @xmath295 denote the corresponding filtration subgroup . @xcite[defn : very - cuspidal ] @xmath16 an irreducible ( and necessarily finite - dimensional ) representation @xmath296 of @xmath290 is called a very cuspidal representation of level @xmath282 if 1 . @xmath295 is contained in the kernel of @xmath297 2 . @xmath298 does not contain the trivial character of @xmath299 we say that an irreducible admissible representation @xmath181 of @xmath117 is minimal if for every character @xmath99 of @xmath110 we have @xmath300 [ thm : sc - kutzko ] there exists a bijective correspondence given by compact induction @xmath301 from very cuspidal representations @xmath298 of either maximal open compact - mod - center subgroup @xmath290 and irreducible minimal supercuspidal representations of @xmath184 moreover , every irreducible minimal supercuspidal representation of conductor @xmath302 ( resp . @xmath303 ) comes from a very cuspidal representation of @xmath292 ( resp . @xmath293 ) oflevel @xmath285 . following kutzko we use the terminology that a supercuspidal representation of @xmath117 is said to be _ unramified _ if it comes via compact induction from a representation of @xmath292 and _ ramified _ if it comes via compact induction from a representation of @xmath293 . we now take up both types of supercuspidal representations and briefly review how they break up on restriction to @xmath124 . we refer the reader to @xcite and @xcite for this . we begin with the unramified case . let @xmath298 be an irreducible very cuspidal representation of @xmath292 of level @xmath285 ( @xmath304 ) . let @xmath181 be the corresponding supercuspidal representation of @xmath184 let @xmath305 then we have @xmath306 where @xmath307 if @xmath308 or if @xmath309 and @xmath171 is irreducible , then @xmath310 is irreducible , hence so is @xmath311 . we thus have an unramified supercuspidal @xmath7-packet @xmath312 if @xmath309 and @xmath171 is reducible , then @xmath298 comes from the unique ( up to twists ) cuspidal representation of @xmath313 whose restriction to @xmath314 is reducible and breaks up into the direct sum of the two cuspidal representations of @xmath314 of dimension @xmath315 correspondingly , we have @xmath316 , and if we let @xmath317 and @xmath318 , then we obtain the unique supercuspidal @xmath7-packet @xmath319 of @xmath124 containing four elements . for the ramified case , let @xmath298 be a very cuspidal representation of @xmath293 of level @xmath285 ( @xmath320 ) and let @xmath181 be the corresponding supercuspidal representation of @xmath117 . let @xmath321 then @xmath316 for two irreducible representations @xmath322 ( @xmath323 ) of @xmath128 and @xmath324 conjugates one to the other , i.e. , @xmath325 . let @xmath326 and so @xmath327 . then the restriction of @xmath181 to @xmath124 breaks up into the direct sum of two irreducible supercuspidal representations as @xmath328 we call @xmath329 a _ ramified supercuspidal @xmath7-packet _ of @xmath198 to summarize , we have three kinds of supercuspidal @xmath7-packets for @xmath124 namely , 1 . unramified supercuspidal @xmath7-packets @xmath330 ; 2 . the unique ( unramified ) supercuspidal @xmath7-packet @xmath319 of cardinality four ; 3 . ramified supercuspidal @xmath7-packets @xmath329 . ( unramified supercuspidal @xmath7-packets of cardinality two ) [ prop : sc - unramified-2-sl2]@xmath16 consider an unramified supercuspidal @xmath7-packet @xmath330 determined by a very cuspidal representation @xmath298 of level @xmath285 of @xmath292 as above . the conductors @xmath331 are both equal to @xmath302 . the dimensions of the spaces @xmath332 and @xmath333 are as follows : 1 . for any @xmath201 such that @xmath334 we have @xmath335 2 . let @xmath334 and @xmath336 if @xmath285 is odd then for all @xmath337 .2pc ( a ) @xmath338 + \(b ) @xmath339 . 3 . let @xmath334 and @xmath336 if @xmath285 is even then for all @xmath337 .2pc ( a ) @xmath340 + \(b ) @xmath341 ( test vectors for unramified supercuspidal @xmath7-packets of cardinality two)[prop : sc - unramified-2-sl2-test ] @xmath16 let @xmath298 be a very cuspidal representation of @xmath292 which determines an unramified supercuspidal @xmath7-packet @xmath330 as above . assume that @xmath342 is realized in its kirillov model with respect to @xmath343 define two elements @xmath344 and @xmath345 in the kirillov model as follows : @xmath346 0 , & \mbox{if $ x \notin ( \mathcal{o}^{\times})^2,$ } \end{array } \right.\\[.2pc ] \phi_{\epsilon}(x ) & = \widetilde{\pi}(\gamma)\phi_1.\end{aligned}\ ] ] let @xmath347 . we have 1 . @xmath348 2 . if @xmath285 is even , then @xmath349 . in addition , @xmath5 is @xmath236-generic and any @xmath236-whittakerfunctional is non - zero on @xmath344 and vanishes on @xmath345 . furthermore , @xmath350 is not @xmath280-generic for any character @xmath280 of conductor @xmath351 . it is however @xmath352-generic , and any @xmath352-whittaker functional is non - vanishing on @xmath353 , which is a newform for @xmath354 3 . if @xmath285 is odd , then ( ii ) holds with @xmath5 and @xmath350 interchanged . ( unramified supercuspidal @xmath7-packet of cardinality four ) [ prop : sc - unramified-4-sl2]@xmath16 let @xmath298 denote a very cuspidal representation of @xmath292 of level @xmath309 such that @xmath355 let @xmath319 be the corresponding @xmath7-packet of @xmath198 then @xmath356 . moreover , 1 . let @xmath201 be any character such that @xmath357 if @xmath5 denotes any representation in the @xmath7-packet , then @xmath358 2 . let @xmath201 be any character such that @xmath359 and @xmath360 then for all @xmath361 we have .2pc ( a ) @xmath362 + \(b ) @xmath363 ( test vectors for unramified supercuspidal @xmath7-packets of cardinality four)[prop : sc - unramified-4-sl2-test ] @xmath16 with notation as above let @xmath319 be the unramified supercuspidal @xmath7-packet of cardinality four . let @xmath364 be the character of @xmath365 obtained from @xmath236 by identifying @xmath365 with @xmath366 . without loss of generality assume that @xmath367 is @xmath364-generic . then 1 . @xmath368 is @xmath236-generic , @xmath369 is @xmath352-generic , @xmath370 is @xmath269-generic , and @xmath371 is @xmath372-generic . 2 . assume that @xmath181 is realized in its @xmath236-kirillov model . the function @xmath344 of proposition [ prop : sc - unramified-2-sl2-test ] is a newform for @xmath368 . this further implies that @xmath373 is a newform for @xmath374 @xmath375 is a newform for @xmath370 and @xmath376 is a newform for @xmath377 finally , each of these newforms is a test vector for an appropriate whittaker functional comingfrom ( i ) . ( ramified supercuspidal @xmath7-packets ) [ prop : sc - ramified - sl2]@xmath16 let @xmath329 be a ramified supercuspidal @xmath7-packet of level @xmath285 as above . then @xmath378 . moreover , 1 . for any character @xmath201 of @xmath110 such that @xmath359 we have @xmath379 2 . let @xmath380 and @xmath336 for all @xmath381 we have @xmath382 ( test vectors for ramified supercuspidal @xmath7-packets ) [ prop : sc - ramified - sl2-test]@xmath16 let @xmath329 be a ramified supercuspidal @xmath7-packet coming from a very cuspidal representation @xmath298 of @xmath383 of level @xmath384 one and only one of the @xmath385 is @xmath236-generic , say @xmath369 . then @xmath371 is @xmath269-generic . let @xmath386 if @xmath344 and @xmath345 have the same meaning as in proposition [ prop : sc - unramified-2-sl2-test ] ( assuming that @xmath181 is realized in its kirillov model ) , we have 1 . @xmath387 and @xmath388 2 . any @xmath236-whittaker functional is non - zero on @xmath344 and similarly any @xmath269-whittaker functional is non - zero on @xmath345 . [ thm : conductor ] let @xmath5 be an irreducible admissible representation of @xmath122 let @xmath181 be a representation of @xmath389 whose restriction to @xmath124 contains @xmath26 assume that @xmath181 is minimal , i.e. , @xmath390 for all characters @xmath99 of @xmath224 then @xmath391 the next theorem relates the conductor of a representation @xmath5 of @xmath124 with the depth ( see @xcite ) @xmath392 of @xmath5 ( cf . @xcite ) . [ thm : conductor - depth - sl2 ] let @xmath5 be an irreducible representation of @xmath198 let @xmath392 be the depth of @xmath26 1 . if @xmath5 is any subquotient of a principal series representation @xmath227 , then @xmath393 2 . if @xmath5 is an irreducible supercuspidal representation , then @xmath394 we now define the basic filtration subgroups of @xmath39 as we did for @xmath124 in [ sec : newforms ] . let @xmath396 , @xmath397 , the standard hyperspecial subgroup of @xmath39 , and for @xmath398 , @xmath399 we let @xmath400 . let @xmath46 be an admissible representation of @xmath39 such that @xmath47 acts by scalars on @xmath70 . let @xmath50 be a character of @xmath51 such that @xmath401 ( where we have identified @xmath47with @xmath402 ) . for any such character @xmath50 and any subgroup @xmath403 of @xmath39 we define @xmath404 we define the _ @xmath50-conductor _ @xmath62 of @xmath49 to be @xmath405 we define the _ conductor _ @xmath64 of @xmath49 as @xmath406 if @xmath407 is such that @xmath408 and @xmath409 ( resp . @xmath410 ) , then we call @xmath411 ( resp . @xmath412 ) a _ space of newforms _ of @xmath413 . in this case , we refer to a non - zero element of @xmath411 or @xmath412 as a _ newform _ of @xmath413 . in this section , we will compute the dimension of @xmath414 for every irreducible admissible infinite - dimensional representation @xmath49 of @xmath39 and every character @xmath50 such that @xmath415 . we will often make use of the following fact . let @xmath5 be the restriction of @xmath49 to @xmath124 , and let @xmath416 by definition , the group @xmath207 acts on @xmath417 via the character @xmath201 , hence via @xmath50 . also , @xmath47 acts on @xmath417 via the character @xmath48 , hence via @xmath50 since @xmath418 . thus any @xmath419 acts on @xmath417 by multiplication by @xmath420 . in the light of ( [ eq : filtration ] ) , @xmath421 & = \mathcal{o}_f^{\times}/(\mathcal{o}_f^{\times})^2.\end{aligned}\ ] ] we may therefore take @xmath95 and @xmath422 as coset representatives for @xmath423 . hence if @xmath424 , then @xmath425 if and only if @xmath426 , i.e. , @xmath427 let @xmath428 be a character of @xmath429 let @xmath430 denote the principal series @xmath431 . according to @xcite , @xmath430 is irreducible except in the cases 1 . @xmath432 , 2 . @xmath433 . in case ( i ) , let @xmath434 be the character of @xmath402 defined by @xmath435 . then @xmath430 has two jordan hlder constituents , namely the one - dimensional representation @xmath436 and a square integrable representation @xmath437 . in case ( ii ) , @xmath430 is the direct sum of two irreducible representations @xmath438 and @xmath439 , which together form an @xmath7-packet of @xmath39 . we distinguish @xmath438 from @xmath439 by defining @xmath438 to be the summand that has a @xmath74-spherical vector , hence @xmath440 let @xmath441 . then the restriction of @xmath430 to @xmath124 is isomorphic to @xmath442 . it is easily seen then that the restriction to @xmath124 of any irreducible constituent of @xmath430 is itself irreducible unless @xmath99 is the character corresponding to some ramified quadratic extension @xmath443 . in this case @xmath444 decomposes as the direct sum @xmath445 . we now compute the conductors of the representations in the principal series of @xmath39 . [ thm : ps - u11 ] let @xmath428 be a character of @xmath429 . suppose that @xmath49 is an irreducible constituent of the principal series @xmath430 . 1 . if @xmath50 is a character of @xmath446 with @xmath418 , then @xmath447 if and only if @xmath448 or @xmath449 . moreover , @xmath450 1 , & \mbox{if $ \bar{\pi } = { \rm st}(\xi)$}. \end{array}\right.\end{aligned}\ ] ] 2 . suppose @xmath50 is as above . .2pc ( a ) if @xmath451 , @xmath428 is ramified , and @xmath452 , then @xmath453 m+1 , & \mbox{if $ m>0$}. \end{array } \right.\ ] ] + \(b ) for @xmath454 , we have @xmath455 \hskip -2.5pc \dim\left(\bar{\pi}^2(\bar{\chi})_{\bar{\eta}}^{\bar{k}_m}\right ) = \dim\left(\bar{\pi}^1(\bar{\chi})_{\bar{\eta}}^{\bar{k}'_m}\right ) & = \left\lceil\frac{m}{2}\right\rceil.\end{aligned}\ ] ] + \(c ) in all other cases , @xmath456 we may assume without loss of generality that @xmath428 is chosen so that @xmath49 is a subrepresentation of @xmath430 . let @xmath5 be the restriction of @xmath49 to @xmath124 . let @xmath50 be any character of @xmath51 with @xmath401 . let @xmath457 . since @xmath458 , @xmath459 we claim that @xmath460 precisely for @xmath461 or @xmath462 . the first part of ( i ) follows immediately from this claim , and the second follows from this together with the conductor calculations in [ sec : principal - sl2 ] . let @xmath463 . the only @xmath201 such that @xmath464 are @xmath465 and @xmath466 . hence we can not have @xmath460 unless @xmath50 equals @xmath467 on @xmath28 . we first prove that @xmath468 if and only if @xmath469 or @xmath449 in the case where @xmath470 and @xmath471 . since @xmath472 is contained in the restriction of @xmath430 to @xmath124 , which is isomorphic to @xmath473 , it is an easy consequence of the proofs of the statements in [ sec : principal - sl2 ] ( see @xcite ) that @xmath474 \mathbb{c}\bar{f}_w + \mathbb{c}\bar{f}_1 , & \mbox{if $ \bar{\chi}^2\|_{\mathcal{o}_f^{\times } } = { \bf 1}$ , } \end{array } \right.\end{aligned}\ ] ] where @xmath475 \bar{\chi}(t)\|t\|_e^{1/2}\eta ( d ) = \bar{\chi}(td)\|t\|_e^{1/2 } , & \mbox{if $ g= \left(\begin{array}{cc } t & * \\ 0 & ^{s}t^{-1 } \end{array}\right ) w \left(\begin{array}{cc } a & b\\ c & d\end{array}\right)$ , } \end{array } \right.\\[.2pc ] \bar{f}_1 ( g ) & = \left\ { \begin{array}{ll } 0 , & \mbox{if $ g\notin\bar{b}k_c$,}\\[.2pc ] \bar{\chi}(t)\|t\|_e^{1/2}\eta ( d ) = \bar{\chi}(td)\|t\|_e^{1/2 } , & \mbox{if $ g = \left(\begin{array}{cc } t & * \\ 0 & ^{s}t^{-1 } \end{array}\right ) \left(\begin{array}{cc } a & b\\ c & d\end{array}\right)$. } \end{array } \right.\end{aligned}\ ] ] we now determine when @xmath476 lie in @xmath477 . in the light of ( [ eq : theta ] ) , this reduces to verifying whether @xmath478 acts as the scalar @xmath479 on these vectors . it is easily checked that @xmath480 \bar{\pi}(\theta)\bar{f}_1 & = \bar{\chi}(\epsilon_e)\bar{f}_1.\end{aligned}\ ] ] hence @xmath481 if and only if @xmath482 . this is equivalent to @xmath483 since @xmath50 and @xmath428 already agree on @xmath28 and @xmath402 ( by assumption ) and since @xmath484 is a representative for the non - trivial coset in @xmath485 . similarly , if @xmath486 , then @xmath487 if and only if @xmath488 , which is equivalent to @xmath489 since @xmath50 and @xmath462 already agree on @xmath28 and @xmath402 and since the non - trivial coset in @xmath485 is represented by @xmath115 . summarizing , we have that when @xmath471 and @xmath490 if and only if @xmath491 or @xmath449 . on the other hand , if @xmath492 , note that we may exchange @xmath428 and @xmath462 in the above proof since @xmath430 and @xmath493 have the same constituents . ( of course , exchanging @xmath428 and @xmath462 may make our assumption that @xmath49 is a subrepresentation of @xmath430 false . the only case in which this matters , however , is when @xmath494 , and in this case we are already done since @xmath495 on @xmath28 . ) then carrying out the proof _ mutatis mutandis _ , we obtain again that @xmath496 if and only if @xmath461 or @xmath462 . this establishes our claim if @xmath49 is in a singleton @xmath7-packet since for all @xmath497 , @xmath498 finally , suppose that @xmath499 . by the above , @xmath500 if and only if @xmath461 or @xmath462 . also , if @xmath50 is any character of @xmath51 , then since @xmath501 and @xmath502 , we have that @xmath503 but @xmath504 by proposition [ prop : unramfified - sl2 ] since @xmath505 . thus @xmath506 so again @xmath507 precisely for @xmath461 or @xmath462 . finally , conjugating by @xmath157 as above , one easily obtains the claim in the case @xmath508 . we now compute the dimensions of @xmath414 to prove ( ii ) . since @xmath509 and @xmath510 have the same irreducible constituents , we may assume that @xmath491 . ( as above , the representations @xmath437 present no problem here since in this case @xmath511 on @xmath51 . ) if @xmath471 the proof of ( i ) shows that @xmath512 . thus @xmath513 is 1 if @xmath514 and 2 if @xmath515 . the proof also shows that @xmath516 \dim ( \bar{\pi}^2(\bar{\chi}))_{\bar{\eta}}^{\bar{k}_0 } = \dim ( \bar{\pi}^1(\bar{\chi}))_{\bar{\eta}}^{\bar{k}'_0 } & = 0.\end{aligned}\ ] ] this shows that the formulae for the dimensions are valid when @xmath517 . suppose that @xmath518 . as with theorem 5.3 of @xcite , it follows from lemma 3.2.1 and the following proofs in 3.2 of @xcite that @xmath332 is the direct sum of @xmath472 together with certain two - dimensional spaces @xmath519 of the form @xmath520 ( @xmath521 ) , where @xmath522 \bar{\chi}(t)\|t\|_e^{1/2}\eta ( d ) = \bar{\chi}(td)\|t\|_e^{1/2 } , & \mbox{if $ g = \left(\begin{array}{cc } t & * \\ 0 & ^st^{-1 } \end{array}\right ) \left(\begin{array}{cc } 1 & 0\\ \varpi^i & 1 \end{array}\right ) \left(\begin{array}{cc } a & b\\ c & d\end{array}\right)$ , } \end{array } \right.\\[.3pc ] \hskip -4pc \bar{f}_{i,\epsilon } ( g ) & = \left\ { \begin{array}{l@{\quad}l } 0 , & \mbox{if $ g\notin\bar{b } \left(\begin{array}{cc } 1 & 0\\ \varpi^m\epsilon_f & 1 \end{array}\right ) k_m$,}\\[.3pc ] \bar{\chi}(t)\|t\|_e^{1/2}\eta ( d ) = \bar{\chi}(td)\|t\|_e^{1/2 } , & \mbox{if $ g= \left(\begin{array}{cc } t & * \\ 0 & ^st^{-1 } \end{array}\right ) \left(\begin{array}{cc } 1 & 0\\ \varpi^i\epsilon_f & 1 \end{array}\right ) \left(\begin{array}{cc } a & b\\ c & d\end{array}\right)$. } \end{array } \right.\end{aligned}\ ] ] we will now verify that whenever @xmath523 , 1 . @xmath519 is @xmath56-stable , and 2 . the subspace of @xmath519 on which @xmath56 acts via the character @xmath50 is one - dimensional . if this holds , then @xmath524 and the formulae for the dimension of @xmath414 follow easily from this equation and the dimension results of [ sec : principal - sl2 ] . the dimension of @xmath525 is computed analogously . we now show ( 1 ) and ( 2 ) . by ( [ eq : theta ] ) , this reduces to showing that @xmath519 is @xmath526-stable , and that the subspace of @xmath519 on which @xmath526 acts as the scalar @xmath527 is one - dimensional . let @xmath528 be either 1 or @xmath529 . then @xmath530 if @xmath531 , this is non - zero if and only if @xmath532 , i.e. , if and only if @xmath533 . this together with the fact that @xmath534 implies that @xmath535 is a multiple of @xmath536 . the exact multiple is determined by evaluating @xmath537 thus @xmath538 similarly , @xmath539 as claimed , @xmath526 stabilizes @xmath540 . moreover , the characteristic polynomial of @xmath526 on this two - dimensional space is @xmath541 the eigenvalues of @xmath526 on @xmath542 are therefore @xmath543 . it follows that the subspace of @xmath519 on which @xmath526 acts as the scalar @xmath479 is one - dimensional.@xmath544 now suppose that @xmath49 is an irreducible representation of conductor @xmath545 in the principal series of @xmath39 and that @xmath50 is such that @xmath546 . we consider the effect of the whittaker functional @xmath547 given by ( [ eqn : whittaker - ps ] ) on @xmath548 . ( test vectors for principal series representations ) [ prop : ps - test - u11]@xmath16 suppose that @xmath49 is an irreducible representation in the principal series of @xmath39 . let @xmath50 be a character of @xmath51 with @xmath418 such that @xmath549 . let @xmath550 . 1 . if @xmath451 , @xmath428 is ramified , and @xmath551 , then @xmath49 is @xmath236-generic . moreover , the space of vectors @xmath552 on which @xmath547 vanishes is one - dimensional . 2 . if @xmath499 , then @xmath49 is @xmath236-generic and @xmath547 is non - zero on the one - dimensional space of newforms @xmath553 . 3 . if @xmath508 , then @xmath49 is not @xmath236-generic , but it is @xmath554-generic . moreover , @xmath555 is non - zero on the one - dimensional space of newforms @xmath556 . 4 . in all other cases , @xmath49 is @xmath236-generic . in addition , if @xmath557 , then @xmath558 for any newform @xmath559 in @xmath477 . let @xmath5 be the restriction of @xmath49 to @xmath124 . note that since @xmath124 and @xmath39 have borel subgroups with the same unipotent radical ( namely , @xmath120 ) , the restriction of @xmath547 to any @xmath236-generic component of @xmath5 is a non - zero @xmath236-whittaker functional on that component , while its restriction to any non-@xmath236-generic component is @xmath560 . let @xmath561 . assume we are in case ( ii ) , ( iii ) , or ( iv ) . let @xmath562 be either @xmath563 or @xmath564 , according to the case , and let @xmath565 , i.e. , @xmath7 is either @xmath566 or @xmath567 . assume that @xmath559 is a non - zero vector in @xmath568 . by theorem [ thm : ps - u11 ] , the restriction of @xmath49 to @xmath124 is irreducible of conductor @xmath545 , and @xmath569 is one - dimensional . the statements in each of these cases now follow easily from the analogous results about @xmath5 in [ sec : principal - sl2 ] . suppose now that @xmath451 with @xmath428 ramified and @xmath570 . then @xmath5 has conductor @xmath571 and @xmath572 has dimension 2 . if @xmath5 is irreducible , then @xmath5 is @xmath236-generic according to corollary [ cor : test - ram - ps ] . also , according to the proof of theorem [ thm : ps - u11 ] ( and using its notation ) , @xmath573 . it follows from corollary [ cor : test - ram - ps ] that @xmath574 . since the image of @xmath547 has dimension 1 , @xmath547 must vanish on a one - dimensional subspace of @xmath552 . if @xmath5 is reducible , then as discussed in [ sec : principal - sl2 ] , @xmath5 decomposes as the direct sum of two representations @xmath369 and @xmath371 . moreover , only one of these representations , say @xmath369 , is @xmath236-generic by corollary [ cor : test - ramified - l - packets - sl2 ] . then @xmath547 vanishes on moreover , by corollary [ cor : test - ramified - l - packets - sl2 ] , for all non - zero hence , as in the preceding paragraph , the subspace of @xmath552 on which @xmath547 vanishes is one - dimensional.@xmath544 we now consider the supercuspidal representations of @xmath39 . let @xmath49 be such a representation . it is easily deduced from analogous results on @xmath117 and @xmath124 that @xmath49 is compactly induced from an irreducible representation of @xmath144 , @xmath575 , or @xmath143 . we will call @xmath49 an _ unramified ( resp . ramified ) supercuspidal representation _ of @xmath39 if its restriction to @xmath124 contains an unramified ( resp . ramified ) supercuspidal representation of @xmath198 suppose first that @xmath49 is ramified . let @xmath5 be the restriction of @xmath49 to @xmath124 . let @xmath369 be an irreducible component of the restriction of @xmath49 to @xmath124 . then @xmath369 is a ramified supercuspidal representation of @xmath124 . we extend @xmath369 to a representation of @xmath576 via the central character @xmath48 , also denoted by @xmath369 . then @xmath49 is contained in @xmath577 , and the restriction of @xmath577 to @xmath576 is @xmath578 . but conjugation by @xmath526 and @xmath324 have the same effect on @xmath124 so , by the discussion in the beginning of [ sec : sc - sl2 ] , @xmath369 and @xmath579 comprise an @xmath7-packet for @xmath124 . since @xmath580 , @xmath577 is irreducible and hence equal to @xmath49 . thus @xmath581 , where@xmath582 . from theorem [ thm : conductor - depth - sl2 ] , we see that the conductor of both @xmath369 and @xmath371 is @xmath583 , where @xmath584 is the depth of both @xmath369 and @xmath371 . we note that the depth of a twist of @xmath49 is no less than @xmath584 . to see this , let @xmath104 be a point in the bruhat tits building of @xmath39 ( which is the same as that of @xmath124 ) and let @xmath585 be a non - negative real number . then any vector in the twist of @xmath49 that is fixed by @xmath586 is fixed by @xmath587 since @xmath588 ( see @xcite ) . it follows that the depth of the twist of @xmath49 is no less than the depth of its restriction to @xmath124 . but this restriction is @xmath5 , which has depth equal to @xmath584 . on the other hand , we may select a character @xmath99 of @xmath39 such that @xmath589 on @xmath590 ( viewed as a subgroup of @xmath47 ) . if @xmath591 , then @xmath592 is trivial on @xmath593 , and it is easily seen that @xmath594 . define @xmath595 as @xmath99 ranges over all characters of @xmath39 . then we have @xmath596 . [ thm : sc - ram - u11 ] let @xmath46 be a ramified supercuspidal representation of @xmath39 . let @xmath50 be any character of @xmath51 with @xmath418 and @xmath597 . then we have @xmath598 and @xmath599 let @xmath5 be the restriction of @xmath49 to @xmath124 . set @xmath600 and @xmath457 . as discussed above , the restriction of @xmath49 to @xmath124 is the direct sum of two ramified supercuspidal representations @xmath601 , each of conductor @xmath545 . by proposition [ prop : sc - ramified - sl2 ] , @xmath602 is non - zero if and only if @xmath603 . hence if @xmath604 , @xmath605 since @xmath606 suppose @xmath603 . as in [ sec : principal - u11 ] , we compute @xmath607 using the fact ( [ eq : theta ] ) that @xmath414 is the subspace of @xmath608 on which @xmath478 acts as the scalar @xmath609 . since @xmath610 and the conjugation action of @xmath526 and @xmath324 are the same on @xmath124 , @xmath369 and @xmath371 form an @xmath7-packet according to [ sec : sc - sl2 ] . thus @xmath611 is the restriction to @xmath124 of a minimal ramified supercuspidal representation @xmath181 of @xmath117 . in particular , we have an action of @xmath117 on @xmath70 . let @xmath612 be the one - dimensional space @xmath613 . then according to the proof @xcite of proposition [ prop : sc - ramified - sl2 ] @xmath614 now @xmath478 intertwines @xmath615 and @xmath371 and takes @xmath616 to @xmath617 . therefore , @xmath618 let @xmath619 for @xmath620 . note that @xmath621 & = \omega_{\bar{\pi}}(\epsilon_e/\,^s\epsilon_e)\bar{\pi}\left ( \left(\begin{array}{ll } \epsilon_f & 0\\ 0 & \epsilon_f^{-1 } \end{array}\right ) \right).\end{aligned}\ ] ] thus @xmath622 acts via the scalar @xmath623 on @xmath624 . it follows that @xmath478 exchanges the one - dimensional spaces @xmath625 since @xmath626 & \bar{\pi}(\theta ) ( \bar{\pi}(\theta)\widetilde{\pi}(\beta)^iw ) = \bar{\pi}(\theta)^2 ( \widetilde{\pi}(\beta)^iw ) = \bar{\eta}(^s\epsilon_e ^{-1})^2\widetilde{\pi}(\beta)^iw = \widetilde{\pi}(\beta)^iw.\end{aligned}\ ] ] in particular , each @xmath627 is stabilized by @xmath478 . moreover , since @xmath622 acts via the scalar @xmath628 on @xmath627 , the eigenspaces of @xmath478 on @xmath627 corresponding to the eigenvalues @xmath629 must each be one - dimensional . hence the subspace of @xmath630 on which @xmath478 acts via the scalar @xmath631 has dimension @xmath632 , as required.@xmath544 suppose that @xmath46 is an unramified supercuspidal representation induced from a representation @xmath633 of @xmath144 . it is easily seen that the restriction @xmath5 of @xmath49 to @xmath124 is either 1 . an irreducible unramified supercuspidal representation of @xmath124 induced from @xmath74 if the restriction of @xmath633 to @xmath74 is irreducible , or 2 . the direct sum of two irreducible unramified supercuspidal representations of @xmath124 induced from @xmath74 if the restriction of @xmath633 to @xmath74 is isomorphic to @xmath634 , where @xmath367 and @xmath635 come from the two cuspidal representations of @xmath314 of dimension @xmath636 ( as in [ sec : sc - sl2 ] ) . in case ( ii ) , we note that if @xmath5 decomposes into the direct sum of @xmath637 and @xmath638 , then @xmath639 . as discussed in the ramified case , if @xmath640 as @xmath99 ranges over all characters of @xmath39 , then the conductors of the components of @xmath5 are @xmath641 . [ thm : sc - un - u11 ] let @xmath46 be an unramified supercuspidal representation of @xmath39 that is induced from @xmath144 , and let @xmath642 . let @xmath50 be any character of @xmath51 with @xmath418 and @xmath643 . then @xmath644 . 1 . if @xmath645 is odd , then @xmath646 \hskip -1.25pc\dim ( \bar{\pi}_{\bar{\eta}}^{\bar{k}'_m } ) & = \max\left\ { \left\lceil\frac{m - c(\bar{\pi})-1}{2}\right\rceil , 0\right\ } = \dim ( ( \bar{\pi}')_{\bar{\eta}}^{\bar{k}_m}).\end{aligned}\ ] ] 2 . if @xmath645 is even , then @xmath647 \hskip -1.25pc\dim ( ( \bar{\pi}')_{\bar{\eta}}^{\bar{k}'_m } ) & = \max\left\ { \left\lceil\frac{m - c(\bar{\pi})-1}{2}\right\rceil , 0\right\ } = \dim ( \bar{\pi}_{\bar{\eta}}^{\bar{k}_m}).\end{aligned}\ ] ] we give a proof only for case ( ii ) ( @xmath648 even ) as the proof for case ( i ) is easily obtained therefrom by interchanging the representations @xmath49 and @xmath649 . moreover , we prove only the first equality of each line as the second follows by conjugating by @xmath157 . let @xmath650 be the restriction of @xmath651 to @xmath124 . set @xmath652 and @xmath653 . now @xmath350 is a direct summand of the restriction to @xmath124 of a minimal unramified supercuspidal representation @xmath654 of @xmath117 . since @xmath181 is unramified , it follows from [ sec : sc - sl2 ] that @xmath655 is isomorphic to @xmath350 and hence that @xmath656 maps @xmath70 onto @xmath70 . ( here we view @xmath70 as a subrepresentation of @xmath657 . ) as discussed above , @xmath350 is either an irreducible unramified supercuspidal representation of conductor @xmath545 or the direct sum of two such representations @xmath658 and @xmath659 , where @xmath660 . by propositions [ prop : sc - unramified-2-sl2 ] and [ prop : sc - unramified-4-sl2 ] , the level @xmath285 of the inducing data for these representations is @xmath661 . as in the ramified case , we have @xmath662 if @xmath663 . suppose @xmath603 . by ( [ eq : theta ] ) , to find @xmath664 , we compute the dimension of the subspace of @xmath665 on which @xmath666 acts as the scalar @xmath609 . let @xmath667 . since @xmath668 is odd , it follows from propositions [ prop : sc - unramified-2-sl2 ] and [ prop : sc - unramified-4-sl2 ] and their proofs @xcite that @xmath669 and @xmath670 in fact , from the proof of proposition 3.3.4 in @xcite , it follows that for a certain vector @xmath671 , @xmath672 . if @xmath350 is irreducible , then since conjugation by @xmath324 and @xmath526 have the same effect on @xmath124 , @xmath656 and @xmath666 are both elements of the one - dimensional space @xmath673 . they are therefore equal up to scalars so @xmath674 . if @xmath350 is reducible , then we may further assume that @xmath675 by proposition [ prop : sc - unramified-4-sl2-test ] . in this case , @xmath656 and @xmath666 are both elements of the one - dimensional space @xmath676 so @xmath677 as above . as in the ramified case , @xmath678 acts via the scalar @xmath628 on @xmath679 . it follows that @xmath666 exchanges the one - dimensional spaces @xmath680 since @xmath681 & \bar{\pi}'(\theta)\left(\mathbb{c}\bar{\pi}'(\theta ) \phi\right ) = \bar{\pi}'(\theta)^2\left(\mathbb{c}\phi\right ) = \bar{\eta}(^s\epsilon_e ^{-1})^2\mathbb{c}\phi = \mathbb{c}\phi.\end{aligned}\ ] ] in particular , @xmath666 stabilizes @xmath612 . again as in the ramified case , these facts imply that the eigenspaces of @xmath666 on @xmath612 corresponding to the eigenvalues @xmath629 must each be one - dimensional . the same is clearly true of @xmath682 for @xmath683 . it follows that the subspace of @xmath665 on which @xmath666 acts via the scalar @xmath684 has dimension @xmath685 as required . the computation of @xmath686 is entirely analogous.@xmath544 now suppose that @xmath49 is a supercuspidal representation of @xmath39 of conductor @xmath545 . we consider the effect of a whittaker functional @xmath547 on @xmath548 . for this we need to choose the character @xmath50 somewhat carefully . let @xmath188 be the @xmath7-packet of @xmath39 containing @xmath687 then the restriction to @xmath124 of the direct sum of representations in @xmath188 is also the restriction to @xmath124 of a minimal supercuspidal representation @xmath181 coming via kutzko s construction . we require @xmath688 ( test vectors for supercuspidal representations ) @xmath16 suppose that @xmath49 is an irreducible supercuspidal representation of @xmath39 of conductor @xmath545 . let @xmath50 be a character of @xmath51 with @xmath689 and @xmath690 ( see above ) . let @xmath691 . 1 . if @xmath49 is ramified , then @xmath49 is @xmath236-generic . moreover , @xmath558 for all non - zero @xmath559 in @xmath477 or @xmath692 . 2 . if @xmath49 is unramified and induced from @xmath144 , let @xmath693 . + .2pc ( a ) if @xmath694 is odd , then @xmath49 is @xmath236-generic and @xmath649 is @xmath554-generic . moreover , @xmath558 for all non - zero @xmath695 and @xmath696 for all non - zero@xmath697 . + \(b ) if @xmath694 is even , then @xmath649 is @xmath236-generic and @xmath49 is @xmath554-generic . moreover , @xmath558 for all non - zero @xmath698 and @xmath696 for all non - zero @xmath699 . let @xmath5 be the restriction of @xmath49 to @xmath124 . as in proposition [ prop : ps - test - u11 ] , we note that the restriction of @xmath547 to any @xmath236-generic component of @xmath5 is a @xmath236-whittaker functional on that component , while its restriction to any non-@xmath236-generic component is @xmath560 . suppose first that @xmath49 is ramified ( case ( i ) ) . then @xmath5 decomposes as the direct sum @xmath700 of irreducible ramified supercuspidal representations of conductor @xmath545 . by proposition [ prop : sc - ramified - sl2-test ] , only one summand , say @xmath369 , is @xmath236-generic and we have that @xmath250 is non - zero on @xmath701 now @xmath692 is the space of vectors in @xmath702 on which @xmath478 acts as the scalar @xmath609 . as observed in the proof of theorem [ thm : sc - ram - u11 ] , @xmath478 exchanges @xmath703 and @xmath704 . therefore , @xmath477 can not lie in either @xmath705 or @xmath704 . in particular , if @xmath706 is written as @xmath707 with @xmath708 , then @xmath709 . since @xmath710 is @xmath236-generic and @xmath371 is not , we get @xmath711 we now give a proof in case ( ii ) . we only prove ( a ) as the proof of ( b ) is obtained by interchanging @xmath49 and @xmath649 . suppose that @xmath49 is unramified and induced from @xmath144 and that @xmath712 is odd . then @xmath5 is also unramified , induced from @xmath74 , and has conductor @xmath545 . as noted in the proof of theorem [ thm : sc - un - u11 ] , @xmath713 since the level of the inducing data of @xmath5 is @xmath714 , which is even , @xmath5 is @xmath236-generic by proposition [ prop : sc - unramified-2-sl2-test ] . moreover , @xmath715 , while @xmath716 . by the proof of theorem [ thm : sc - ram - u11 ] , @xmath478 exchanges @xmath717 and @xmath718 therefore , just as in the ramified case , if @xmath719 , then @xmath720 . it follows that @xmath721 the proof of the non - vanishing of @xmath722 is entirely analogous.@xmath544 we have only considered the unitary group @xmath35 for an unramified extension @xmath723 the entire series of results go through with some minor modifications if instead we considered ramified extensions . [ thm : conductor - otherinvariants - u11 ] let @xmath49 be an irreducible admissible supercuspidal representation of @xmath37 the relation between its conductor @xmath64 and its minimal depth @xmath648 is given by @xmath724 if @xmath5 is an irreducible subrepresentation of the restriction of @xmath49 to @xmath124 then @xmath725 this follows from theorems [ thm : sc - ram - u11 ] and [ thm : sc - un - u11].@xmath544 given an irreducible representation @xmath49 of @xmath39 and a character @xmath50 of @xmath51 such that @xmath447 , one can ask if we have @xmath726 . the answer is that this is often the case but is not true in general . indeed , we have @xmath726 unless @xmath49 is the principal series representation @xmath430 , where @xmath428 is ramified and @xmath727 . for these exceptional representations , the dimension of the space of newforms is two . nevertheless , in all cases we have proved that an appropriate whittaker functional is non - vanishing on some newform . this can be used to formulate a kind of a multiplicity one result if we consider the quotient of the space of newforms by the kernel of this whittaker functional . more precisely , if @xmath728 is a non - trivial additive character of @xmath0 of conductor either @xmath9 or @xmath729 such that @xmath49 is @xmath728-generic , and @xmath730 is a @xmath728-whittaker functional for @xmath49 , then we have @xmath731 another possibility is to consider some canonical non - degenerate bilinear form on the space @xmath732 and consider the orthogonal complement of the subspace @xmath733 as a candidate for a one - dimensional space of newforms . then the multiplicity one result is formulated as @xmath734 we thank benedict gross , dipendra prasad and paul sally jr . for some helpful correspondence . we thank brooks roberts for his interest in this work and for his generous comments on new avenues of thought that might stem from it . the second author is grateful for the warm and pleasant working experience at bucknell university where a substantial part of this work was completed . we thank the referee for a very careful reading of the manuscript . kutzko p c and sally jr . p , all supercuspidal representations of @xmath737 over a @xmath32-adic field are induced , representation theory of reductive groups ( utah : park city ) ( 1982 ) pp . 185196 ; _ prog . math . _ ( boston , ma : birkhaser ) ( 1983 ) vol .	let @xmath0 be a non - archimedean local field whose residue characteristic is odd . in this paper we develop a theory of newforms for @xmath1 , building on previous work on @xmath2 . this theory is analogous to the results of casselman for @xmath3 and jacquet , piatetski - shapiro , and shalika for @xmath4 . to a representation @xmath5 of @xmath1 , we attach an integer @xmath6 called the conductor of @xmath5 , which depends only on the @xmath7-packet @xmath8 containing @xmath5 . a newform is a vector in @xmath5 which is essentially fixed by a congruence subgroup of level @xmath6 . we show that our newforms are always test vectors for some standard whittaker functionals , and , in doing so , we give various explicit formulae for newforms . = msam10 at 10pt = tibi at 10.4pt [ theore]theorem [ theore]proposition [ theore]remark [ theore]corollary [ theore]definition
super - massive black holes ( smbhs ) are ubiquitous in galactic nuclei ( @xcite ) , and binaries of these massive objects are a likely product of the hierarchical evolution of structures in the universe . after a galaxy merger , where both progenitors host a smbh , different mechanisms are responsible for the evolution of the binary orbit depending on its separation ( see review by @xcite ) . dynamical interaction with stars appears to be efficient to bring the smbhs down to parsec scales only , what is known as the `` last parsec problem '' ( @xcite ) . a possible way to overcome this barrier and merge the smbhs within a hubble time is interaction with gas . many theoretical and numerical studies have focused on the orbital evolution of a sub - parsec binary surrounded by a circumbinary disc ( e.g. @xcite ) . however , the exact mechanism that would produce such discs is still unclear ; it is necessary an efficient transport of gas from thousands or hundreds of parsecs to the central parsec . turbulence and gravitational instabilities in the interstellar medium , through the formation of clumps , allow portions of gas to travel almost unaffected by its surrounding , enhancing the probability of reaching the galactic nuclei ( @xcite ) . a possible manifestation of these events is the putative molecular cloud that resulted in the unusual distribution of young stars orbiting our galaxy s smbh . in particular , the simulation of ( * ? ? ? * bonnell & rice ( 2008 ) ) shows a spherical , turbulent cloud falling with a very low impact parameter ( @xmath20.1 pc ) onto a one million solar masses smbh . assuming that these accretion events are common in galactic nuclei , the goal of our work is to model such an event onto a binary instead of a single smbh . in particular , we are interested on the properties of the discs that will form given different relative orientations between the orbital angular momenta of the cloud and the binary . notice that this study is complementary to that shown in @xcite , as we are modeling clouds with very low orbital angular momentum . we model the interaction between the binaries and clouds using a modified version of the sph code gadget-3 ( @xcite ) . the cloud is represented using over @xmath3 gas particles with a total mass of @xmath4 , an initial turbulent velocity field and uniform density . the smbhs are modelled as two equally - massive sink particles that interact only through gravity and can accrete sph particles . the total mass of the binary is @xmath5 , and its initial orbit is keplerian and circular . the physical setup of the simulation is shown in figure [ initial ] . the initial velocity of the cloud yields a highly eccentric ( @xmath6 ) , bound orbit where the pericenter distance is @xmath7 pc , which is less than the binary radius , making the interaction between the gas and smbhs very strong . as we expect clouds approaching the binary from different directions , we model systems with three different inclinations between the cloud and binary orbits : aligned , perpendicular and counter - aligned . in this section we present the main results of the simulations with the different inclinations , in particular the discs that form around the binary and each smbhs . on the left panel of figure [ bhbxy ] we show the column density map of the simulation at different times , where we can see how the interaction develops . as the gas falls almost radially onto the binary , around 80% of the cloud is accreted by the smbhs . most of the remaining gas is pushed away due to an efficient slingshot . the bound material forms a tail that get stretched and diluted over time , feeding mini - discs " that form around each smbh . to measure the alignment between the binary orbit and the mini - discs , we compute its angular momentum on the corresponding black hole reference frame . we show the time evolution of the direction of both discs on the hammer projection of figure [ bhbxy ] . here we observe that they tend to align with the orbit of the binary , as expected , although one disc is slightly tilted with respect to the aligned position and also precesses around that position . this behavior could have distinctive electromagnetic signatures . for example , the misalignment could affect the variability of spectral lines , or each disc have different polarisation . the precession could be observed if jets are launched from the smbhs and align with the mini - discs . with this inclination , as in the previous case , around 80% of the cloud mass is added to the smbhs . however , the interaction between the gas and the binary , that we can see in figure [ bhbxz ] , is completely different respect to the aligned case . due to a less efficient slingshot , most of the remaining material stays bound to the system and it retains its original angular momentum , forming an unstable structure around the binary . the gas that reaches the smbhs also produce mini - discs , but they are less massive and more intermittent than in the aligned case . the direction of the mini - discs , shown on the right panel of figure [ bhbxz ] , shows that they tend to follow the original direction of the cloud , which makes them completely misaligned respect to the binary orbit . as well as the previous case , this could have distinctive signatures on the variability of lines or the direction of possible jets . , but for the model with perpendicular orbits . in this case the cloud moves on the x - y plane while the binary is on the x - z plane.,title="fig:",scaledwidth=60.0% ] , but for the model with perpendicular orbits . in this case the cloud moves on the x - y plane while the binary is on the x - z plane.,title="fig:",scaledwidth=40.0% ] in this case we have that the interaction of the binary with the gas produces shocks that cancel angular momentum , allowing the smbhs to accrete even more material than in the previous cases ; around 90% of the cloud is swallowed . the remaining material forms a tail that , due to the gravitational torques , produces a circumbinary ring . in this case we do not observe mini - discs during the entire length of the simulation . finally , we compute the eccentricity distribution of the gas on three different stages , as shown on the right panel of figure [ bhb - xy ] . it is interesting that , if there is star formation in the ring , the stars would have very different orbits around the binary depending on the formation time - scale . for a very rapid star formation we could have highly eccentric orbits ( solid line ) , while for a slow process the stars could be distributed on a narrow ring with nearly circular orbits ( dashed line ) . , but for the model with the counter - aligned orbits . in this case , the cloud and the binary move on the x - y plane , but the latter is rotating clockwise . right : eccentricity distribution of the gas for three different times in the simulation.,title="fig:",scaledwidth=60.0% ] , but for the model with the counter - aligned orbits . in this case , the cloud and the binary move on the x - y plane , but the latter is rotating clockwise . right : eccentricity distribution of the gas for three different times in the simulation.,title="fig:",scaledwidth=42.0% ] the preliminary results presented here show that accretion events onto binaries result in very different disc morphologies , depending on the relative inclination between the cloud and the binary orbits . our work in progress is extending this study to larger impact parameters , showing lower accretion rates , and exploring the effect of the interaction on the black holes orbit and spin evolution . these results are likely to have important implications on the multi - messenger future studies of smbh binaries and on the long - term evolution of these systems . column density maps were created with splash by @xcite . we acknowledge support from conicyt - chile through pcha / doctorado nacional , fondecyt ( 1141175 ) , basal ( pfb0609 ) , anillo ( act1101 ) , redes ( 120021 ) and exchange ( pcci130064 ) grants .	we model numerically the evolution of @xmath0 turbulent molecular clouds in near - radial infall onto @xmath1 , equal - mass super - massive black hole binaries , using a modified version of the sph code gadget-3 . we investigate the different gas structures formed depending on the relative inclination between the binary and the cloud orbits . our first results indicate that an aligned orbit produces mini - discs around each black hole , almost aligned with the binary ; a perpendicular orbit produces misaligned mini - discs ; and a counter - aligned orbit produces a circumbinary , counter - rotating ring .
the large hadron collider ( lhc ) machine will enter the collision mode at centre - of - mass energy of 10 tev in 2008 and eventually will operate at 14 tev from 2009 onwards gradually improving upon the luminosity . two general - purpose detectors , cms ( compact muon solenoid ) and atlas ( a toroidal apparatus ) , are ready for the data - taking phase with intense detector commisioning works underway . the much awaited discoveries at the lhc have to be preceeded by the _ rediscovery _ of the standard model ( sm ) , and early discovery , though possible , will be challenging . on the other hand , from early phase on , _ ie _ , starting from an integrated luminosity of mere 1 pb@xmath0 , various sm processes can be studied ; in particular , heavy quarkonia peaks of @xmath1 and @xmath2 will shine well above background in the low mass range of the invariant mass distribution of dileptons . electroweak ( ew ) processes of w and z productions , having very large rates combined with clean leptonic decay mode signatures ( @xmath3 ) , are not only the _ first physics _ at the lhc , they will be studied at all luminosities . these will be the _ standard candles _ for a large variety of lhc measurements . the w , z physics will enable , as a function of accumulated data , detector and physics - analysis tools commissioning of the experiments , precision determination of electroweak variables as well as measurements in as yet unexplored kinematic regions . these programmes constitute both the confirmation as well as the consistency checks of sm at the highest energies lhc offer . the breakthrough discoveries at the lhc will be interpreted by critically studying sm processes with experimental accuracy matched by theoretical understanding at the same level . since the production cross section and dynamics are largely controlled by qcd , events with inclusive jets , photons , dileptons , heavy quarks will be produced copiously at the lhc , as exemplified in table [ evtrate ] , providing complementary datasets which will eventually probe many aspects of strong and ew interactions with high precision . the large samples will also be useful for search of rare processes which potentially hint at as yet unknown _ new physics_. .expected triggered event yield in atlas with 1 pb@xmath0 data at 14 tev . [ cols="<,^",options="header " , ] [ wzsyst2 ] to establish the anticipated discoveries at the lhc , the standard model processes must be well understood . various aspects of z - boson production and subsequent decays will be utilised by experiments for precision determination of electroweak parameters . an early , competitive w mass measurement from the lhc is improbable . some of the di - boson productions are likely to be etsablished during the early phase of the lhc . the author wishes to thank the organisers for the kind support and local hospitality without which the participation would not have been possible . the author also gratefully acknowledges many collaborators in cms experiment for overall support in presenting the talk and preparing this report . 9 the cms collaboration , cms physics analysis summary , cms pas 2007/002 . the cms collaboration , cms physics analysis summary , cms pas ewk-08 - 005 . d.bourlikov , contribution to these proceedings . g. quast et.al . , cms note 2006/061 . m. boonekamp , acta physica olonica b , vol.38 ( 2007 ) . n. besson , m. boonekamp , atl - phys - pub-2006 - 007 , a. tricoli et.al . , hera - lhc workshop proceedings , arxiv : hep - ex/05090021 . the cms collaboration , cms physics analysis summary , cms pas ewk-08 - 003 .	the large hadron collider , lhc , though meant for discovery , will provide enough data from early phase to also perform various studies of standard model processes in as yet unexplored kinematic regions . precision measurements of the electroweak variables will be possible due to the large rates of w and z boson productions combined with clean leptonic signatures . examples of simulation results from cms and atlas collaboration studies are presented to show the wide variety of measurements possible and how various issues like background estimation , determination of systematic effects will be taken care of by the experiments .
in comparison with stationary black holes , understanding of dynamical black holes is still far from clear . although there is a lot of potentially interesting subjects such as thermodynamical properties , dynamical stability , or hawking radiation , the absence of the preferred time direction makes them intractable . also , lack of concrete exact solutions in simple systems is one of the main reasons for the relatively slow progress . in the present paper , among others , we focus on asymptotically anti - de sitter ( ads ) dynamical black holes . in the history of gravitation physics , ads black holes had been considered unrealistic and eccentric configurations for a long time . however , they stepped into the limelight by the discovery of the ads / cft duality @xcite . now ads black holes have new significance as a stage to study strongly coupled gauge theories and occupy a central position of research in string theory . the motivation of the study in the present paper is twofold . one comes from the ads / cft duality . while a static ads black hole corresponds to the field theory at the boundary which has finite temperature in equilibrium , a dynamical ads black hole would correspond to some field theory in the non - equilibrium state . actually , an asymptotically ads spacetime has been studied in a dynamical setting as a holographic dual to the bjorken flow @xcite . while the dynamical spacetime in @xcite was constructed perturbatively , exact dynamical ads black holes are desirable to derive more specific results . the second motivation comes from the recently - found dynamical instability of the ads spacetime . although the ads vacuum is known to be stable at the linear level , its nonlinear instability was numerically found with a massless klein - gordon field in arbitrary dimensions @xcite . ( see also @xcite . ) it was both numerically and analytically supported that an ads black hole forms as a result of this instability @xcite . however , there is an argument that static ads black holes are also unstable at the nonlinear level @xcite . ( see also @xcite . ) therefore , the final fate of the instability of the ads vacuum or a static ads black hole is still not clear at present . in this context , not only a static configuration but also a dynamical configuration is the candidate of the final state . an example is an oscillating or time - periodic spacetime @xcite . therefore , an exact dynamical black - hole solution might represent the final state or an intermediate stage during the time evolution and must be useful for further study . in the present paper , we consider spacetimes with spherical , plane , or hyperbolic symmetry in arbitrary dimensions . it is well - known in this system that the no - hair theorem holds for a wide class of scalar fields , which prohibits asymptotically flat black holes with non - trivial configuration of scalar fields @xcite . here one assumes staticity to prove the no - hair theorem . for a massless klein - gordon field , even a stronger result is available , namely the no - hair theorem independent of the asymptotic condition for the spacetime and the value of @xmath0 . ( see appendix a for the proof . ) as a result , all the known solutions with a non - trivial scalar field in this system contain naked singularities both for @xmath1 @xcite and @xmath2 @xcite , and the only possible static black hole is the schwarzschild(-(a)ds ) black hole with a constant scalar field or its topological generalization . therefore , in order to obtain non - trivial black - hole solutions , one has to remove the assumption of staticity . in four dimensions , a set of exact dynamical and inhomogeneous solutions has been obtained by many authors @xcite . in the present paper , we generalize this set of solutions and show that some of the solutions describe a locally ads dynamical black hole . in the case where the klein - gordon field is purely imaginary , namely ghost , an ads dynamical wormhole may be realized . in the following section , we give our spacetime ansatz and present the solutions . in sec . iii , we show that the class - i solution represents an ads dynamical black hole or wormhole . in sec . iv , we discuss the properties of other classes of solutions . concluding remarks are summarized in sec . v. the scalar no - hair theorem for a massless klein - gordon field is shown in appendix a. in appendix b , we present the counterpart of our solution in the case without a cosmological constant . in appendix c , it is shown that the class - i solution with a real scalar field does not represent a wormhole . our basic notation follows @xcite . the convention for the riemann curvature tensor is @xmath3v^\mu = { { \cal r}^\mu } _ { \nu\rho\sigma}v^\nu$ ] and @xmath4 . the minkowski metric is taken to be mostly plus sign , and greek indices run over all spacetime indices . we adopt the units such that only the @xmath5-dimensional gravitational constant @xmath6 is retained . we consider the einstein - klein - gordon-@xmath0 system in arbitrary @xmath7 dimensions . the field equations are @xmath8 and @xmath9 , where @xmath10 is defined by the @xmath5-dimensional gravitational constant @xmath6 as @xmath11 . the energy - momentum tensor for the klein - gordon field is @xmath12 in this paper , we consider an @xmath5-dimensional warped product manifold @xmath13 with the line element @xmath14 where @xmath15 is a lorentzian metric on @xmath16 and @xmath17 is a scalar on @xmath16 . @xmath18 is an @xmath19-dimensional unit space of constant curvature , where @xmath20 denotes its curvature taking the values @xmath21 , @xmath22 , and @xmath23 , and @xmath24 is the metric on @xmath18 . namely the riemann tensor on @xmath18 is given by @xmath25 where the superscript @xmath19 means the geometrical quantity on @xmath18 . the generalized misner - sharp quasi - local mass is a scalar on @xmath16 defined by @xmath26,\end{aligned}\ ] ] where @xmath27 $ ] , @xmath28 is the covariant derivative on @xmath16 and @xmath29 @xcite . @xmath30 denotes the volume of @xmath18 if it is compact and otherwise arbitrary . @xmath31 has the monotonicity and positivity properties for arbitrary ( positive ) @xmath30 and is constant in vacuum @xcite . in the asymptotically flat or ads case , that coefficient is fixed in such a way that it converges to the global mass such as the arnowitt - deser - misner mass @xcite or abbott - deser mass @xcite . in the non - static spacetime , there is no timelike killing vector to define a natural time - slicing . in such a case , the kodama vector @xmath32 defines a preferred time direction in the untrapped region , where @xmath33 and @xmath34 is a volume element of @xmath35 @xcite . the kodama vector is timelike ( spacelike ) in the untrapped ( trapped ) region defined by @xmath36 . on the trapping horizon defined by @xmath37 , the kodama vector becomes null . in the present paper , we consider solutions in the following form : @xmath38 , \label{g - sussman } \\ h(\rho)= & \left\ { \begin{array}{ll } \displaystyle{\sqrt{-{\tilde \lambda}}\sin \rho } & \mbox{[class - i~($\lambda<0$)]},\\ \displaystyle{\sqrt{-{\tilde \lambda}}\rho } & \mbox{[class - ii~($\lambda<0$)]},\\ \displaystyle{\sqrt{-{\tilde \lambda}}\sinh \rho } & \mbox{[class - iii~($\lambda<0$)]},\\ \displaystyle{\sqrt{{\tilde \lambda}}\cosh \rho } & \mbox{[class - iii~($\lambda>0$)]}. \end{array } \right . \end{aligned}\ ] ] the physical domain is the region with @xmath39 . the areal radius is given by @xmath40 , where @xmath41 is chosen such that @xmath17 is non - negative . these classes of solutions have been investigated as solutions with a stiff fluid , which is equivalent to a massless klein - gordon field if the gradient of the scalar field is timelike @xcite . they were first obtained by lake @xcite for @xmath42 and @xmath43 and independently obtained by other authors @xcite . the global structure and physical properties were investigated in @xcite . the solutions with @xmath42 and general @xmath20 were obtained by collins and lang @xcite and also in @xcite . keeping in mind this history , we call these classes of solutions the _ generalized lake solution _ in the present paper . the system reduces to the following master equation for @xmath44 : @xmath45 where a dot denotes the derivative with respect to @xmath46 and the constant @xmath47 is @xmath21 , @xmath22 , and @xmath23 for class - i , -ii , and -iii , respectively . the klein - gordon field is given as @xmath48 the scalar field is homogeneous @xmath49 in our coordinate system . the energy - momentum tensor has the form of @xmath50 , where @xmath51 is the energy density of the scalar field . important physical quantities are given as @xmath52 where a prime denotes the derivative with respect to @xmath53 . @xmath54 and @xmath55 are satisfied in the spherically symmetric class - i solution ( @xmath56 ) . using @xmath57 , the kodama vector is written as @xmath58 the master equation ( [ master ] ) is solved analytically in three and four dimensions for any @xmath20 but only for @xmath59 in higher dimensions , which will be presented later . in order to see the qualitative property of the solution , we introduce a new variable @xmath60 and write the master equation as @xmath61 this equation is integrated by parts to give @xmath62 where @xmath63 is an integration constant . this is a simple one - dimensional potential problem for the variable @xmath64 . using the following expression , @xmath65 we obtain simple expressions of @xmath66 , @xmath67 and @xmath31 : @xmath68 it is clear that the energy density of the scalar field and the quasi - local mass are positive ( negative ) for @xmath69 and then the scalar field is real ( purely imaginary , namely ghost ) . in three dimensions , the scalar field becomes trivial and we have @xmath70 and @xmath71constant . the spacetime is then locally ( a)ds . it is shown that , in the case of @xmath72 , @xmath73 corresponds to ads infinity . ( in contrast , @xmath74 can not be zero for @xmath75 in the class - iii solution . ) indeed , @xmath76 is satisfied . we actually show that the affine parameter @xmath77 blows up at @xmath78 along null geodesics . in the spacetime ( [ g - sussman ] ) , there is a conformal killing vector @xmath79 satisfying the conformal killing equation : @xmath80 along a null geodesic , with its tangent vector @xmath81 , there is a conserved quantity @xmath82 . the following expression @xmath83 is integrated to give @xmath84},\\ \displaystyle{{\tilde \lambda}\rho } & \mbox{[class - ii]},\\ \displaystyle{{\tilde \lambda}\tanh \rho } & \mbox{[class - iii~($\lambda<0$)]},\\ \displaystyle{{\tilde \lambda}/\tanh \rho } & \mbox{[class - iii~($\lambda>0$)]}. \end{array } \right . \end{aligned}\ ] ] therefore , @xmath85 for @xmath72 corresponds to @xmath86 . ads infinity is given by @xmath87 in the class - ii and -iii solutions , and by @xmath88 in the class - i solution , where @xmath89 is an integer . it is seen in eqs . ( [ mu2 ] ) and ( [ m2 ] ) that the spacetime is indeed vacuum at ads infinity ( @xmath85 ) , but @xmath31 blows up there . the quasi - local mass @xmath31 with @xmath43 converges to the abbott - deser mass in the asymptotically ads spacetime @xcite under the henneaux - teitelboim fall - off condition @xcite . ( see @xcite for the higher - dimensional version . ) its contraposition means that if @xmath31 blows up , the fall - off rate is slower than the henneaux - teitelboim condition and the spacetime is asymptotically only locally ads @xcite . there is a static solution @xmath90 of the master equation ( [ master ] ) in the case of @xmath91 in four and higher dimensions : @xmath92 where @xmath93 is constant . the energy density and the quasi - local mass are given by @xmath94 while the metric is static , the scalar field is time - dependent . in the class - iii solution with @xmath43 , the klein - gordon field is real , while it is ghost in the class - i solution with @xmath95 . we do nt present the detailed analysis for this static solution , but the penrose diagram is fig . 1(d ) for the class - i solution with @xmath95 and the solution represents a static ads wormhole . ( see @xcite for the wormhole solution without @xmath0 . ) the penrose diagram for the class - iii solution with @xmath43 is fig . 2(a ) for @xmath72 and fig . 2(f ) for @xmath75 . hereafter we do nt consider the static case . we are interested in the class - i solution because the coordinate system covers the maximally extended spacetime and describes an asymptotically locally ads black hole or wormhole . in four dimensions , @xmath44 is given by @xmath96 where @xmath97 is a constant . the energy density and the quasi - local mass are given by @xmath98 the energy density is positive ( negative ) for @xmath99 . the ads vacuum is realized for @xmath100 with @xmath101 . ( for @xmath59 , @xmath102 is not allowed since it gives @xmath103 . ) the scalar field with positive energy density is given by @xmath104},\\ \displaystyle{\sqrt{\frac{1}{2\kappa_4 ^ 2}}\ln\biggl\|\frac{1-\cos 2t}{\sin 2t}\biggl\| } & \mbox{[for $ k=0$]}. \end{array } \right . \end{aligned}\ ] ] the scalar field with negative energy density is given as @xmath105 where @xmath106 . in arbitrary dimensions , @xmath44 and @xmath66 for the class - i solution are given in closed forms only for @xmath59 : @xmath107^{2/(n-2)},\\ \pm(\phi-\phi_0)=&\sqrt{\frac{n-3}{(n-2)\kappa_n^2}}\ln\biggl\|\frac{1-\cos(n-2)t}{\sin(n-2)t}\biggl\|.\end{aligned}\ ] ] the energy density and the quasi - local mass are given by @xmath108 ^ 2},\\ m = & \frac{(n-2)v_{n-2}^{(0)}c_1^{(n-1)/2}}{2\kappa_n^2(\varepsilon h)^{n-3}[\sin(n-2)t]^{(n-3)/(n-2)}}.\end{aligned}\ ] ] in three dimensions ( @xmath109 ) , we obtain @xmath110 and @xmath71constant and the solution represents a baados - teitelboim - zanelli black hole in the non - standard coordinates @xcite . it is not difficult to understand the causal structure of the spacetime ( [ g - sussman ] ) . @xmath111 corresponds to curvature singularity , of which existence depends on the parameters . since the metric on @xmath112 in the solution ( [ g - sussman ] ) is conformally flat , a light ray runs along a 45-degree straight line in the @xmath113-plane . the penrose diagrams for this solution are presented in fig . [ adsbh ] . ( see table [ table : ads ] . ) the spacetime represents a dynamical black hole or wormhole depending on the parameters . portions of the @xmath113 plane covering the maximally extended spacetime of the class - i solution . the corresponding penrose diagrams have the same structures . a zigzag and a thick line correspond to a curvature singularity and ads infinity , respectively . bheh stands for the black - hole event horizon . figures . ( a)(c ) represent a black hole , while fig . ( d ) represents a wormhole . ] .[table : ads ] the penrose diagrams for the class - i solution with positive energy density and @xmath114 . in the case of @xmath115 with negative energy density , the penrose diagram is fig . 1(d ) . [ cols="<,^,^",options="header " , ] the corresponding penrose diagrams are drawn in fig . 2 . in those diagrams , the regular coordinate boundary @xmath116 , which consists of null hypersurfaces in fig . 2 drawn by dashed lines , is extendable . this is obvious by eq . ( [ geodesic ] ) since @xmath116 corresponds to a finite value of the affine parameter @xmath77 . furthermore , @xmath117 and @xmath118 are satisfied for @xmath116 along null geodesics . this fact indicates that a variety of @xmath119 extension is possible beyond this coordinate boundary without introducing any matter field on the junction surface . one possible extension is to attach an exact ( a)ds spacetime @xcite . although @xmath116 is regular and extendable along null geodesics , it is singular along spacelike curves with @xmath120constant , where @xmath67 blows up . therefore , although the regions ( ii ) in fig . 2(e ) and ( j ) are maximally extended , there is no regular cauchy surface . the regions ( ii ) in figs . 2(b ) and 2(e ) and ( iii ) in fig . 2(e ) contain a black - hole event horizon . the penrose diagrams for the class - ii and -iii solutions . figures ( a)(e ) [ ( f)(j ) ] correspond to the case with @xmath121 . figures ( c ) , ( d ) , ( h ) , and ( i ) correspond to the special case of the class - iii(c ) solution with @xmath122 . coordinate boundaries represented by dashed lines are extendable . the @xmath113 plane is divided by singularities and ads infinity into several portions and each of them represents a distinct spacetime . the parameters are required to give @xmath39 for the physical spacetime and the left halves of figs . ( a)(e ) are equivalent to the right halves . the region ( ii ) in fig . ( e ) is maximally extended and contains an event horizon , but there is no regular cauchy surface . ] we have presented a set of exact solutions in the einstein - klein - gordon system with a cosmological constant in arbitrary dimensions . the spacetime has spherical , plane , or hyperbolic symmetry and admits a spatially conformal killing vector . the solution is obtained in a closed form in three and four dimensions for any @xmath20 but only for @xmath59 in higher dimensions . even in the case without the explicit form , it is able to understand the qualitative properties of the solutions by analyzing the equivalent one - dimensional potential problem . in three dimensions , the solution reduces to the locally ( a)ds vacuum . for @xmath72 , the spacetime is asymptotically locally ads . the quasi - local mass blows up at ads infinity while the energy density converges to zero , which infers the slow fall - off to the ads infinity . some of the solutions admit a black - hole event horizon . in the class - i solution , the coordinate system covers the maximally extended spacetime and the solution with a real scalar field describes a dynamical ads black hole . if the scalar field is ghost , the solution represents a dynamical ads wormhole . while the solution with @xmath95 in four dimensions represents the dynamical formation of a black hole , the black hole is eternal in the case of @xmath43 in four dimensions and @xmath59 in @xmath123 dimensions . it is still not clear whether the black hole is eternal or not in the case of @xmath124 in higher dimensions . for the class - ii and -iii solutions , we have analyzed the global structures in four dimensions and in higher dimensions with @xmath59 . the regular coordinate boundary is extendable and the @xmath119 extension beyond it would be possible . there are several cases where the spacetime contains a black - hole event horizon , however , the coordinate system does not cover the maximally extended spacetime or there is no regular cauchy surface in the spacetime . in summary , the class - i solution may be a good model of a dynamical ads black hole for further investigations . thermodynamical properties , dynamical stability , and the hawking radiation are interesting subjects to study , of which results will shed light on dynamical properties of ads black holes . the author thanks julio oliva , takashi torii , kei - ichi maeda , cristin martnez , and especially fabrizio canfora for useful comments and discussions . this work has been funded by the fondecyt grants 1100328 , 1100755 ( hm ) and by the conicyt grant `` southern theoretical physics laboratory '' act-91 . this work was also partly supported by the jsps grant - in - aid for scientific research ( a ) ( 22244030 ) . the centro de estudios cientficos ( cecs ) is funded by the chilean government through the centers of excellence base financing program of conicyt . in this appendix , we present a simple proof of the no - hair theorem for a klein - gordon field ; there is no killing horizon in the spacetime represented by the metric ( [ sol2 ] ) if the spacetime is static and the klein - gordon field is static and inhomogeneous . we note that this result is independent of the value of @xmath0 and the asymptotic condition for the spacetime . we adopt the following coordinates for the static spacetime : @xmath125 we can replace @xmath17 by @xmath126 without loss of generality if @xmath17 is not constant . for @xmath127 , the klein - gordon equation @xmath9 gives @xmath128 where @xmath129 is a constant and a prime denotes here the derivative with respect to @xmath126 . if @xmath129 is zero , @xmath66 is constant . the trace of the einstein equation gives the expression of the ricci scalar @xmath130 : @xmath131 where we used eq . ( [ kg1 ] ) . a killing horizon is defined by @xmath132 with @xmath133 , where @xmath134 is its location satisfying @xmath135 . equation ( [ key ] ) shows @xmath136 for @xmath137 , namely a curvature singularity at @xmath134 , unless @xmath138 . therefore , the killing horizon is allowed only in the case where @xmath66 is constant . in this appendix , we present the counterpart of the generalized lake solution in the case without a cosmological constant . the conformal self - similarity naturally reduces to the homothetic self - similarity in the absence of a characteristic scale introduced by the cosmological constant . first let us derive the master equation for the system . introducing the double - null coordinates @xmath139 on @xmath35 and assuming that @xmath35 is flat and @xmath140 , where @xmath141 , the line element for the homothetic self - similar spacetime is given by @xmath142 the einstein tensor and the energy - momentum tensor for the scalar field are written as @xmath143,\\ g^{u}_{~{\bar z}}=&-\frac{n-2}{4u{\cal s}^2}({{\cal s}'}^2 - 2{\cal s}{\cal s}''),\quad g^{{\bar z}}_{~u}=0,\\ g^{{\bar z}}_{~{\bar z}}=&\frac{(n-2)(n-3)}{4u^2{\cal s}^2}(-2k{\cal s}-2{\cal s}{\cal s}'+{\bar z}{{\cal s}'}^2),\\ g^{i}_{~j}=&\frac{n-3}{4u^2{\cal s}^2}\biggl[(n-6){\bar z}{{\cal s}'}^2 - 2(n-4){\cal s}{\cal s } ' \nonumber \\ & -2k(n-4){\cal s}+4{\bar z}{\cal s}{\cal s}''\biggl]\delta^i_{~j}\end{aligned}\ ] ] and @xmath144 where a prime denotes here the derivative with respect to @xmath145 . then , the einstein equation gives the following master equation for @xmath146 : @xmath147 the solution for this master equation is obtained in a closed form for @xmath42 or @xmath59 . ( the solution for @xmath109 is vacuum and hence locally flat , as shown below . ) in the case of @xmath42 , @xmath148 is given by @xmath149 where @xmath150 and @xmath151 are constants . for @xmath152 , @xmath66 is real and given as @xmath153 for @xmath154 , @xmath66 is ghost and given by @xmath155 in the case of @xmath59 , the solution is obtained as @xmath156 it is noted that @xmath157 gives the minkowski spacetime . in the double - null coordinates , the metric and the generalized misner - sharp mass are written as @xmath158 for @xmath42 and @xmath159 for @xmath59 . this solution with @xmath43 and @xmath42 is the roberts solution @xcite . ( see also @xcite . ) the case with @xmath43 in arbitrary dimensions was first discussed in @xcite . this spacetime admits a homothetic killing vector @xmath160 satisfying @xmath161 . since @xmath35 is flat , @xmath162 with constant @xmath163 or @xmath164 with constant @xmath165 corresponds to null infinity . different from the generalized lake solution , the gradient of the scalar field may be spacelike in some domain of spacetime , where the solution is not equivalent to a solution with a stiff fluid . in this appendix , we show that the class - i solution with positive energy density does not represent a wormhole for @xmath114 . this result restricts the possible penrose diagram for the solution without the explicit form . here we define a wormhole spacetime by the existence of causal curves connecting one infinity to another and prove the non - existence of such curves . because the orbits of timelike curves or non - radial null curves run inside the light cone at a given spacetime point in the @xmath113 plane , to show the non - existence for radial null curves is sufficient . first we show that the areal radius @xmath166 blows up at ads infinity in the class - i solution for @xmath167 , which is characterized by @xmath168 and @xmath169 . it is obvious that @xmath170 holds if @xmath39 there . ads infinity where @xmath171 is satisfied is more subtle but it is also the case , as shown below . in the class - i solution with positive energy density ( @xmath172 ) , the master equation ( [ master - x ] ) shows us the behavior of @xmath173 near @xmath174 ( and hence @xmath171 ) : @xmath175 where @xmath176 is the time when @xmath171 . on the other hand , the metric function @xmath177 behaves as @xmath178 near ads infinity . because @xmath179 is satisfied along a radial null curve , @xmath180 is satisfied along such a light ray going to or coming from ads infinity with @xmath171 . along this curve , the areal radius @xmath17 behaves near ads infinity as @xmath181}}{\sqrt{-{\tilde\lambda}}\|t - t_0\|^{(n-3)/(n-2)}}.\end{aligned}\ ] ] therefore , the areal radius @xmath17 blows up at ads infinity . now we show that there is no radial null curve connecting two distinct ads infinity in the spacetime of the class - i solution with positive energy density . let us consider the einstein equation @xmath8 in the double null coordinates : @xmath182 the @xmath183 and @xmath184 components of the einstein equation are written as @xmath185 the null energy condition requires @xmath186 and @xmath187 . the generalized lake solution is written in the double - null coordinates by introducing @xmath165 and @xmath163 such that @xmath188 the correspondence between ( [ g - sussman ] ) and ( [ metric - double ] ) is @xmath189 if the solution represents an ads wormhole , there is a radial null ray which travels from one ads infinity ( where @xmath190 ) to the other . obviously there is at least one positive local minimum of @xmath17 along its orbit , which locally defines a wormhole throat @xcite . without loss of generality , this orbit is expressed by @xmath191 ( and @xmath192constant ) , where @xmath193 is a constant and the throat condition is then given by @xmath194 with @xmath195 . then , eq . ( [ nullbasic2 ] ) shows @xmath196 , the violation of the null energy condition at the throat . the contraposition of this result shows that the class - i solution with positive energy density does not represent an ads wormhole . maldacena , adv . * 2 * , 231 ( 1998 ) [ int . j. theor . phys . * 38 * , 1113 ( 1999 ) ] . s. kinoshita , s. mukohyama , s. nakamura , and k .- ya oda , prog . . phys . * 121 * , 121 ( 2009 ) ; + s. kinoshita , s. mukohyama , s. nakamura , and k .- ya oda , phys . lett . * 102 * , 031601 ( 2009 ) . p. bizon and a. rostworowski , phys . * 107 * , 031102 ( 2011 ) ; + j. jalmuzna , a. rostworowski , and p. bizon , phys . d * 84 * , 085021 ( 2011 ) . d. garfinkle , e - print : arxiv:1202.4025 [ gr - qc ] . dias , g.t . horowitz , and j.e . santos , e - print : arxiv:1109.1825 [ hep - th ] kitp talk by gary horowitz , + http://online.itp.ucsb.edu/online/novelnum12/horowitz/ g. holzegel and j. smulevici , e - print : arxiv:1103.3672 [ gr - qc ] ; + g. holzegel and j. smulevici , ann . henri poincar * 13 * , 991 ( 2012 ) . m. dafermos , contemp . math . * 350 * , 99 ( 2004 ) . j. d. bekenstein , _ ` no hair ' : twenty - five years after _ , chapter in _ proceedings of the second international andrei d. sakharov conference in physics _ , edited by i. m. dremin and a. m. semikhatov ( world scientific , singapore , 1997 ) . fisher , zh . fiz . * 18 * , 636 ( 1948 ) . [ gr - qc/9911008 ] ; + a.i . janis , e.t . newman , and j. winicour , phys . lett . * 20 * , 878 ( 1968 ) ; + m. wyman , phys . rev . d * 24 * , 839 ( 1981 ) ; + k.s . virbhadra , int . j. mod . a * 12 * , 4831 ( 1997 ) ; + k.a . bronnikov , j.c . fabris , and a. zhidenko , eur . j. c * 71 * , 1791 ( 2011 ) ; + b. c. xanthopoulos and t. zannias , phys . d * 40 * , 2564 ( 1989 ) ; + s. abdolrahimi and a. a. shoom , phys . d * 81 * , 024035 ( 2010 ) ; + j.d . barrow , a.b . burd , and d. lancaster , class . grav . * 3 * , 551 ( 1986 ) ; + k.s . virbhadra , pramana * 44 * , 317 ( 1995 ) ; + g. clment and a. fabbri , class . quant . * 16 * , 323 ( 1999 ) ; + v. husain , e.a . martinez and d. nez , phys . d * 50 * , 3783 ( 1994 ) . g. clment and a. fabbri , class . * 17 * , 2537 ( 2000 ) ; + s. das , j. gegenberg and v. husain , phys . rev . d * 64 * , 065027 ( 2001 ) ; + d. daghan and a.h . bilge , gen . grav . * 37 * , 1289 ( 2005 ) ; + c. vuille , gen . rel . grav . * 39 * , 621 ( 2007 ) ; + s.g . senz and c. martnez , phys . d * 85 * , 104047 ( 2012 ) . k. lake , gen . . grav . * 15 * , 357 ( 1983 ) . j. hajj - boutros , j. math . * 26 * , 771 ( 1985 ) . l. herrara and j. ponce de len , j. math * 26 * , 778 ( 1985 ) . n. van den bergh and p. wils , gen . * 17 * , 223 ( 1985 ) . collins and j.m . lang , class . * 4 * , 61 ( 1987 ) . e. shaver and k. lake , gen . relat . grav . * 20 * , 1007 ( 1988 ) . wald , _ general relativity _ ( university of chicago press , chicago , 1984 ) . misner and d.h . sharp , phys . 136 * , b571 ( 1964 ) . k. nakao , arxiv : gr - qc/9507022 . hayward , phys . d. * 53 * , 1938 ( 1996 ) . h. maeda and m. nozawa , phys . d * 77 * , 064031 ( 2008 ) . r. arnowitt , s. deser , and c. w. misner , _ gravitation , an introduction to current research _ , edited by l. witten , ( wiley , new york , 1962 ) . abbott and s. deser , nucl . phys . * b195 * , 76 ( 1982 ) . h. kodama , prog . . phys . * 63 * , 1217 ( 1980 ) . madsen , class . . grav . * 5 * , 627 ( 1988 ) . sussman , j. math . * 32 * , 223 ( 1991 ) . m. henneaux and c. teitelboim , commun . phys . * 98 * , 391 ( 1985 ) . s. hollands , a. ishibashi , and d. marolf , class . * 22 * , 2881 ( 2005 ) . k. skenderis , class . * 19 * , 5849 ( 2002 ) ; + i. papadimitriou and k. skenderis , jhep * 0508 * , 004 ( 2005 ) . h. g. ellis , j. math . * 14 * , 104 ( 1973 ) ; + m.s . morris and k.s . thorne , am . * 56 * , 395 ( 1988 ) ; + t. matos and d. nunez , class . * 23 * , 4485 ( 2006 ) ; + t. matos , gen . . grav . * 42 * , 1969 ( 2010 ) ; + t. matos , l.a . urena - lopez , and g. miranda , e - print : arxiv:1203.4801 [ gr - qc ] . roberts , gen . rel * 21 * , 907 ( 1989 ) . burko , gen . * 29 * , 259 ( 1997 ) . brady , class . . grav . * 11 * , 1255 ( 1994 ) . y. oshiro , k. nakamura , and a. tomimatsu , prog . 91 * , 1265 ( 1994 ) . hayward , class . quant . grav . * 17 * , 4021 ( 2000 ) . g. clement and s.a . hayward , class . * 18 * , 4715 ( 2001 ) . i.i . gutman and r.m . bespalko , sbornik sovrem . tbilissi , * 1 * , 201 ( 1967 ) . h. maeda , phys . d * 79 * , 024030 ( 2009 ) . frolov , class . * 16 * , 407 ( 1999 ) . d. hochberg and m. visser , phys . rev . d58 , 044021 ( 1998 ) .	we present several classes of exact solutions in the einstein - klein - gordon system with a cosmological constant . the spacetime has spherical , plane , or hyperbolic symmetry and the higher - dimensional solutions are obtained in a closed form only in the plane symmetric case . among them , the class - i solution represents an asymptotically locally anti - de sitter ( ads ) dynamical black hole or wormhole . in four and higher dimensions , the generalized misner - sharp quasi - local mass blows up at ads infinity , inferring that the spacetime is only locally ads . in three dimensions , the scalar field becomes trivial and the solution reduces to the baados - teitelboim - zanelli black hole .
x - ray novae ( xn ) , also called soft x - ray transients ( sxts ) , are binary x - ray sources that show outburst behavior @xcite . the prevailing theory for their transient behavior is based on the disk instability model ( dim ) first proposed to explain dwarf novae outbursts ( see the extensive review by @xcite ) . applying this model to xn requires significant modifications due to x - ray irradiation from the central disk and the disruption of the inner disk , presumably by an advection dominated accretion flow ( adaf ) , in quiescence . the dim with these modifications is shown to be quite successful in modeling transient outbursts similar to the canonical xn a0620@xmath100 @xcite . however , it is also clear that some sources undergo outbursts that are very difficult to reconcile with the dim . in this paper we consider one such source , xtej1118 + 480 , and develop an alternative model for the outburst based on modulated mass transfer into the outer disk , rather than a sudden increase in viscosity in a pre - existing disk as in the dim . we derive a general prescription for calculating x - ray outburst light curves based on a diffusion propagation model in the disk . we begin with the same basic diffusion equations as ( * ? ? ? * hereafter bp81 ) but we derive an analytical solution rather than solving the equations numerically . we derive a green s function from the first outburst and develop a deconvolution technique to calculate the mass input rate as a function of time for the second outburst . this allows us to derive the time evolution of the mass of the disk for the second outburst . based on the global oscillation model of ( * ? ? ? * hereafter to00 ) we expect the qpo to correlate inversely with the disk mass . this provides us with at least one observational test for the model . it is possible that this model may be applicable to outbursts observed in other sources as well . @xcite point out similarities among five xne : groj0422 + 32 , xtej1118 + 480 , x1354 - 644 , groj1719 - 24 , and gs2023 + 338 . all five have had at least one outburst in which the source remains in the low - hard state through the entire episode . in three instances a @xmath2 hz qpo drifting persistently upward in frequency is seen . these sources may be good candidates for future tests of this model ; however , in this paper we limit ourselves to xtej1118 + 480 as an example . in section [ sec - source ] we review the properties of the source , particularly the features of the 2000 outburst which are difficult to account for in the dim . in section [ sec - diff ] we describe the main features of the diffusion model . in section [ sec - solve ] we solve for the green s funtion for the specific case where @xmath3 and in the general case . in section [ sec - inverse ] we derive the deconvolution procedure to get the accretion rate into the disk and the time evolution of the disk mass during the outburst . we apply the full procedure to the data on xtej1118 + 480 in section [ sec - data ] . in section [ sec - observe ] we show that this model can explain the time evolution of the low - frequency quasi - periodic oscillation ( qpo ) seen during most of the outburst . we conclude in section [ sec - discussion ] with a discussion of the successes and limitations of this model . xtej1118 + 480 is a black hole ( bh ) transient that was observed from radio to hard x - rays during 2000 january to august . optical observations after the source returned to quiescence confirmed the black hole nature of the compact object by measuring a mass function of @xmath4 @xcite . this is among the highest mass functions measured for any transient . in addition , the combination of the small distance to the source ( @xmath5 kpc ; @xcite ) and the high galactic latitude ( @xmath6 ) results in this source having the lowest reddening of any xn . the x - ray light curve of the source is peculiar , with an unusual _ double _ outburst ( see figure [ fig - data ] ) . the first outburst appears to have a fast rise followed by an exponential decay ( fred morphology ) , but the second outburst is irregular with multiple maxima . the x - ray spectrum was essentially constant as an absorbed power law with photon spectral index of 1.73 @xcite , typical of the low / hard state of black holes . a radio counterpart at 5 ghz has been detected during outburst , although jets were not directly resolved with merlin to a limit of @xmath7 ( @xmath8/kpc ) au @xcite . an earlier paper @xcite reported detailed evolution of the 0.1 hz qpo , using x - ray data from the usa experiment and _ rxte_. the qpo frequency increased monotonically from 0.07 to 0.15 hz over @xmath9 months , failing to correlate with the x - ray luminosity which rises and falls over the same interval . in this paper , we explore using the disk mass rather than usual quantities ( such as luminosity or spectral parameters ) to correlate with the drifting qpo frequency . the low ( @xmath0 hz ) frequency suggests an origin at a large radius in the disk because dynamical timescales scale roughly with the kepler frequency . one theory that could explain such a low - frequency qpo is the global oscillation model of to00 . their model describes a large - scale coherent oscillation in the outer disk , linking qpo issues to those of overall disk structure . the model leads to a bh mass estimate for xtej1118 + 480 of @xmath10 based on the @xmath0 hz qpo frequency , in agreement with recent optical data @xcite which suggest a bh mass range of 6.07.7 @xmath11 . ( we note that there is a typographical error in equation ( 14 ) of to00 which should have @xmath12 instead of @xmath13 . ) the same qpo was seen in euv and optical wavelengths @xcite . the double outburst profile of this source is difficult to account for in the standard dim . if an outburst is triggered by a sudden increase in disk viscosity and leads to the fred - type outburst in 2000 january , how can another outburst , with five times the fluence , occur so shortly afterwards ? what mechanism turns off the first outburst ? and , why does the second outburst have such an irregular shape ? if the instability is triggered at some maximum value of the disk mass , and the amount of mass consumed in the first outburst was replenished in @xmath14 days , then the mass consumed in the second outburst should be replenished at about 0.5 years . thus , the dim models would have predicted that another outburst should have occurred by now . our alternative model is that the outbursts were fueled by sporadic mass transfer from the companion rather than from a large amount of mass stored in the disk during quiescence . it is beyond the scope of the current paper to develop the dynamics of the mass transfer which accounts for this behavior . it is instructive to look at the energetics of the outburst . the luminosity of the long outburst was about @xmath15 erg / s ( for a distance of 1.8 kpc ) which is only @xmath16 ( for an assumed black hole mass of @xmath17 ) , whereas the canonical dim outbursts approach the eddington limit at their peak . mass transfer modulation models for xn outbursts have previously been considered as viable alternatives to the dim . as the xtej1118 + 480 outbursts are not easily reconciled with the standard dim , it seems reasonable to develop a different model for this source . for example , @xcite has suggested that the long outburst of xtej1118 + 480 is a superoutburst of the type seen in the su uma class of cataclysmic variables @xcite , based on striking similarities in the x - ray and optical lightcurves and the appearance of superhumps . to account for some of the unusual aspects of xtej1118 + 480 , here we investigate an alternative model where the outburst is driven by mass transfer from the companion and the diffusive propagation of material in the disk . we now derive the main equation describing diffusive propagation of matter in the disk . using the angular momentum balance [ e.g. @xcite , see also @xcite hereafter ( to99 ) , eq . 4 ] @xmath18 and the continuity equation @xmath19 where @xmath20 and @xmath21 are the mass accretion in the disk and mass accretion input over the disk at the given radius @xmath22 and at the given time @xmath23 , and @xmath24 is the @xmath25component of a viscous stress tensor @xmath26 . we can get a relation between @xmath24 and the surface density @xmath27 using equation ( 5 ) in to99 with an assumption that the rotational frequency is equal to the keplerian one , @xmath28 : @xmath29 where @xmath30 is a half - thickness of a disk , @xmath31 is the gravitational constant , @xmath32 is the mass of the central object , @xmath33 is a viscosity , @xmath34 is the density , @xmath35 is a kinematic viscosity in the disk and @xmath36 , @xmath37 are the turbulent scale and velocity in the disk respectively . we also use the relation @xmath38 . a relation between @xmath39 and @xmath40 can be obtained using equation ( 1 ) and ( 3 ) with an assumption that @xmath28 @xmath41.\ ] ] substitution of @xmath39 from this equation into equation ( 2 ) leads to the diffusion equation @xmath42\right ) + q(r , t).\ ] ] equation ( 5 ) is identical to equation ( 2.3 ) of bp81 . it is worth noting that the derivation presented here and bp81 is general , using only the mass and angular momentum conservations and an assumption of kepler rotational velocity in the disk . it does not invoke specific assumptions regarding the structure of the disk . the viscosity value , @xmath43 implemented for calculation of @xmath24 in equation ( 3 ) are averaged over the vertical size of the disk . the product of @xmath44 is in fact an integral of the density , @xmath34 , over the @xmath45coordinate , yielding a surface density @xmath40 . in our calculation we will assume that the mass accretion input ( presumably from the companion ) occurs at the outer boundary of the disk at @xmath46 , namely @xmath47 thus , @xmath48 is the mass input rate as a function of time at the outer edge of the disk . using equations ( 2 ) and ( 6 ) we derive the differential equation for the disk mass @xmath49 : @xmath50 where mass accretion @xmath51 at the inner disk boundary can be calculated using equation ( 4 ) . we assume the reflection boundary at the outer edge of the disk ; in other words , the net mass flux @xmath52 through the outer boundary is zero : @xmath53 this is compatible with the mass input at the outer edge of the disk from eqn . ( 6 ) because the mass input is considered to be added just inside the outer radius @xmath54 where the reflection boundary condition is imposed . integration of this equation allows us to get @xmath55 as a function of time @xmath23 : @xmath56d\tau.\ ] ] we replace the function @xmath40 by the function @xmath57 . using eq . ( 4 ) the mass accretion @xmath39 is then expressed through @xmath58 as follows @xmath59 and the diffusion equation ( 5 ) can be written in the form @xmath60 we combine this equation with the boundary condition at the outer boundary @xmath61 which follows from equations ( 8 , 10 ) . we assume that at the inner boundary @xmath62 , @xmath63 , which is equivalent to @xmath64 introduction of a new variable @xmath65 reduces the main equation ( 11 ) to : @xmath66 with the boundary conditions @xmath67 and @xmath68 one can check by direct substitution that the solution of this boundary problem for the inhomogeneous equation ( 14 ) can presented through the following convolution @xmath69 where @xmath70 satisfies the homogeneous equation @xmath71 with the initial condition @xmath72 and boundary conditions ( 15 - 16 ) . because our ultimate goal is to find the x - ray luminosity , which is presumably related to the energy release of the accreting matter at the inner disk edge , we will calculate @xmath73 at @xmath74 ( or at @xmath75 ) . we assume that the mass accretion rate at the inner edge is converted with efficiency @xmath76 into the x - ray luminosity , i.e. @xmath77 . thus @xmath78 the solution @xmath79 of equation ( 18 ) with the initial condition ( 19 ) at @xmath80 and boundary conditions ( 15 - 16 ) can been presented using separation of variables as a series @xmath81 where @xmath82 and @xmath83 are eigenfunctions and eigenvalues which can be found from the inhomogeneous ordinary differential equation : @xmath84 combined with the boundary conditions @xmath85 @xmath86 @xmath87 is the norm of the eigenfunction , which is calculated through the integral ( for example , see the derivation of this formula in @xcite ) @xmath88 where @xmath89 is the weight function . here , we consider a general class of problems where @xmath90 . in the following sections , we solve the problem for one specific value of @xmath91 , then present the general solution . when the turbulent scale @xmath36 is proportional to the characteristic scale @xmath22 and the turbulent velocity @xmath37 is independent of radius , @xmath92 ( i.e. @xmath93 ) . in this case , @xmath94,\ ] ] and the eigenvalues @xmath95 are found from the equation @xmath96=0\ ] ] where the solutions of this equation are @xmath97 for this particular case the solution is expressed by harmonics ( eqs . [ 21 ] and [ 26 ] ) @xmath98\sin[{{\pi(2n-1)x}/2x_0 } ] \sin[{{(2n-1)\pi}/2}]\ ] ] where @xmath99 . @xmath100 is the viscous timescale and determines both the rise and fall time of the response function . we define the green s function @xmath101 as the flux through the inner edge of the disk at time @xmath23 from delta - function injection at the outer edge at time @xmath80 . thus , it is proportional to @xmath102 : @xmath103(2n-1)\sin[{{\pi(2n-1)}/2}].\ ] ] by definition , @xmath101 should be normalized to unity , which can be done analytically in the case @xmath3 . the integral of the series over @xmath23 in this equation is @xmath104 . the asymptotic form of the series ( 29 ) for @xmath105 can be presented as the integral @xmath106\sin(\lambda x ) \sin(\lambda x_0 ) d\lambda,\ ] ] where @xmath107 and consequently @xmath108 . transformation of the product of @xmath109 into the difference @xmath110/2 $ ] followed by integration over @xmath111 leads us to the formula @xmath112-\exp[-(x+x_0)^2/4x_0 ^ 2(t / t_0)]\right\}.\ ] ] after differentiation we have @xmath113^{3/2}}}\exp(-t_0/4 t ) \quad\textrm{for}\quad t\ll 4t_0/\pi^2.\ ] ] at large times , we can use the series ( 30 ) to get @xmath114 \quad\textrm{for}\quad t\gg 4t_0/\pi^2.\ ] ] from a combination of the two asymptotic forms ( eq . [ 31,34 ] ) we construct the simple analytical approximation of the green s function . this combination is a product of the combined exponentials with a linear superposition of the factors before the exponents with weights 1 and 2 respectively : @xmath115 \exp[-(t_0/4t+\pi^2 t/4t_0)],\ ] ] where c is a normalization constant to make the integral of @xmath101 equal to unity . we have chosen the weights of the two terms in this approximation to make it more accurately reproduce the series result . in this section we investigate the asymptotic behavior of the green s function for @xmath116 and @xmath117 in the case of arbitrary kinematic viscosity as a function of @xmath22 ( or @xmath118 ) . we have already demonstrated for the @xmath3 case that the asymptotic for @xmath116 is determined by the contribution of terms of the series with @xmath119 ( see eq . [ 30 ] ) . the calculation of the series is reduced to an integral over @xmath111 ( eq . [ 31 ] ) where the contribution of terms with @xmath120 of order few is negligible . on the other hand , for @xmath121 the term with @xmath122 is dominant and the contribution of terms with @xmath123 is negligible . it is worth noting that the integrated function in equation ( 31 ) is a product of an exponent and harmonics . the similar asymptotic form can be obtained in a general case of the kinematic viscosity . one can check by direct substitution that the eigenfunctions @xmath82 defined as nontrivial solutions of equations ( 2224 ) have the harmonic asymptotic form for large @xmath120 @xmath124\ ] ] with @xmath125 when @xmath126dx\simeq \mu(x_0)/2.\ ] ] with an assumption that the kinematic viscosity @xmath127 is a power law function of the radial coordinate @xmath128 , namely @xmath129 , one can find that @xmath130 is expressed through the bessel functions @xmath131\ ] ] with @xmath132 . among the two bessel functions @xmath133 and @xmath134 , only the former one allows @xmath82 to satisfy the inner boundary condition @xmath135 for @xmath136 . we now investigate the behavior of the solution @xmath130 in the limits of large and small arguments @xmath137 . using the asymptotic form of the bessel function for large arguments @xcite @xmath138^{1/2}\cos[\lambda_n\mu(x)-\pi(6-\psi)/4(4-\psi)].\ ] ] comparing this formula with equation ( 36 ) we see that @xmath139^{1/2}$ ] . for small values of the argument , the eigenfunction @xmath130 can be presented as series expansion around 0 , @xmath140^{1/2 } x^{1/2}\left[\mu(x)\lambda_n/2\right]^{1/(4-\psi ) } \sum_{k=0}^{\infty } { { [ -(\lambda_n\mu(x))^2/4]^k}\over{k!\gamma[(5-\psi)/(4-\psi)+k]}}.\ ] ] in fact , it follows from this equation that @xmath141 when @xmath142 . the eigenvalues @xmath83 are determined by the equation @xmath143 for small @xmath144 the eigenvalues are best determined using the series ( 41 ) , whereas for large @xmath120 the asymptotic form ( 40 ) is appropriate . fortunately , even @xmath145 calculated using the two different methods produces results which differ from each other by less than one percent . for example , the first three terms in series ( 41 ) provide an accuracy of @xmath145 better than 0.1% . the use of these three terms leads us to a quadratic equation for @xmath146 where @xmath147 : @xmath148 the first root of this equation @xmath149^{1/2}\ ] ] allows us to calculate the first eigenvalue . for example , for @xmath150 , @xmath151 , whereas using the second method gives @xmath152 ( see below ) . for calculation of the eigenvalues using the eigenfunction form ( 40 ) we have the transcendental equation : @xmath153=-(2-\psi)/2z(4-\psi).\ ] ] the solution of this equation can be written in an analytical form : @xmath154,\ ] ] where @xmath155}}.\ ] ] we apply equations ( 47 - 48 ) for calculations of the eigenvalues for any @xmath120 because even for the first eigenvalue we have almost the same results using these two methods ( compare with eq . the green s function @xmath101 can be written using ( 41 ) as follows : @xmath156 where @xmath157 and @xmath158.\ ] ] using the asymptotic form of @xmath130 at @xmath159 ( see eq . 40 ) we can rewrite equation ( 49 ) as follows : @xmath160 where @xmath161 and @xmath162 . as in the case with @xmath3 , the viscous time is @xmath100 which in the general case is @xmath163 thus for @xmath164 or @xmath165 @xmath166\exp(-z_1 ^ 2t / t_0).\ ] ] for @xmath167 one should take into account all terms in the series ( eq . [ 51 ] ) . as we have already demonstrated in 3 , the summation of such a series can be reduced to an integration over @xmath111 : @xmath168 where @xmath169 . now we proceed with calculation of the integral from equation ( 54 ) , @xmath170 where @xmath171 and @xmath172 expansion of cosine and sine in series and integration over @xmath111 leads us to ( see also prudnikov , bruchkov & marichev 1981 ) @xmath173m[(\gamma+1)/2,1/2 , -b^2/4t]\ ] ] and @xmath174m[(\gamma+2)/2,3/2 , -b^2/4t].\ ] ] the asymptotic form of the degenerate hypergeometric function @xmath175 for large arguments @xmath176 ( abramowitz & stegun 1968 ) gives us the asymptotic form for integrals @xmath177 and @xmath178 : @xmath179 finally we have @xmath180 we construct the analytical form of the green s function @xmath101 using the asymptotic forms eq . ( 53 ) for @xmath181 and eq . ( 61 ) for @xmath182 , similar to what was done for @xmath3 in section 4.1 : @xmath183\ } \exp(-t_0/4t - z_1 ^ 2t / t_0),\ ] ] where c is a normalization constant to make the integral of @xmath101 equal to unity . the factor 2.5 in the first term was empirically determined to make this combination of asymptotic forms a better match to the series formulation . this form is primarily a computational convenience . note that this form breaks down for @xmath184 so we restrict @xmath91 to be less than 3 . a graphical comparison of the series and analytical forms for several values of @xmath91 can be found in figure [ fig - kgreen ] . an important consequence of the previous derivation is that the fred outburst shape is a natural consequence of the diffusion model . the exponential decay is due to diffusion in a bounded medium , while the fast rise is due to the fact that the probe ( x - ray emission from the inner edge of the disk ) is located at a distance from the source ( accreting material added to the outer edge of the disk ) . thus we have derived the outburst shape for a large class of viscosity structures when the viscosity is a function of only the radius . in this section , we demonstrate a deconvolution procedure to derive @xmath48 , the mass input rate at the outer edge of the disk , and then use @xmath48 and @xmath185 to derive the mass of the disk as a function of time . now we proceed with general case of the input mass supply in the disk @xmath21 which is proportional to @xmath48 ( see eq . furthermore , we assume that the mass accretion rate at the inner disk edge is converted with efficiency @xmath76 into the x - ray luminosity , i.e. @xmath186 . thus using the light curve of @xmath185 we can restore information regarding the input function @xmath48 by the deconvolution of the following equation ( see eq . 20 ) : @xmath187 to do this we also assume that the variability time scale for a(t ) is much longer than the viscous time scale @xmath100 . in other words the support t of the function @xmath48 is much larger than that for @xmath101 . then using the steepest descent method we get from equation ( 63 ) that @xmath188 or @xmath189 where @xmath190 is the time point for the maximum of @xmath191 in the integration interval . the green s function @xmath101 has a sharp maximum at @xmath192 ( which depends on @xmath91 and @xmath100 ) and thus @xmath193 for @xmath194 , @xmath195 for @xmath196 , and @xmath197 for @xmath198 . with the steepest descent approximation of the convolution ( 64 - 65 ) one can resolve that equation with respect to the input function @xmath48 : @xmath199 where @xmath200 , and @xmath201 where @xmath202 $ ] . notice that @xmath203 . to calculate the mass of the disk as a function of time , we make the assumption that the only terms that matter are mass input at the outer boundary of the disk ( proportional to @xmath48 ) and the mass leaving the disk into the black hole ( proportional to @xmath204 ) . hence the instantaneous change of the disk mass @xmath205 is determined by the difference between the input function @xmath48 and the output function . if we assume that the output mass accretion rate @xmath206 is proportional to the observed luminosity @xmath185 in accordance with equation ( 20 ) and the input function @xmath48 is found as a result of deconvolution with the operator @xmath207 defined by eqs . ( 66 - 67 ) , we can calculate the disk mass as a function of time as follows : @xmath208-l_x(t^{\prime})\}dt^{\prime}.\ ] ] it is seen from eqs . ( 66 - 67 ) that @xmath207 is a composite operator which shifts and differentially expands the input function @xmath185 . we now apply this formalism to the 2000 january august outburst of xtej1118 + 480 . figure [ fig - data ] shows a merged 2 - 12 kev light curve from the asm on rxte and the usa experiment on argos , which together cover the full double - peaked outburst . only the asm viewed the initial ( fred ) outburst because the notification trigger did not come until the second outburst . during the second outburst intensive observation of the source commenced . the greater sensitivity of usa is better able to cover the final decline of the source than rxte . in order to proceed with the fitting , we need to select a value for the @xmath91 parameter that defines the radial dependence of the viscosity in the disk . we have no strong physical reason to prefer one value of @xmath91 over another , so we have chosen @xmath209 primarily because it seems to work relatively well . we found that @xmath91 close to 3 worked better than @xmath91 of 1 or 2 , so we chose a @xmath91 slightly less than 3 so that equation ( 62 ) would still be valid . in the future , as this method is applied to other sources it may become clear that there is a preferred value for @xmath91 which is valid for a class of sources . figure [ fig - green ] shows the fit of the green s function ( eq . [ 62 ] , with @xmath209 ) to the first outburst . the fit has only two parameters : the viscous timescale @xmath100 , and the initial time , @xmath210 . a normalization amplitude is also fitted in figure [ fig - green ] , but subsequently discarded when the function is normalized to unity . hence the sole nontrivial parameter ( beyond our choice of @xmath91 ) in the fit is @xmath100 . the integral of this measured function is found to approximate the shape of the initial rise of the second outburst . the integral has the shape expected from a step function turn on ; evidently the second outburst begins with something close to a step function ( see figure [ fig - rise ] ) . this result is a consistency check of the convolution model for the second outburst . in fact , for any input function @xmath48 whose support is much larger than that of the green s function @xmath101 , the shape of the convolution ( eq . [ 63 ] ) begins with the integral of @xmath101 multiplied by the initial value of @xmath48 . this property of the convolution is a useful tool to distinguish the convolution model of the outburst from other possible models . equations ( 66 - 67 ) are used to extract the form of the input function @xmath48 . figure [ fig - reconv ] verifies that this is the solution by re - folding the solution with the green s function to derive the x - ray luminosity . the solution is smoothed because the convolution with the green s function acts as a low - pass filter . figure [ fig - m ] displays the inferred mass @xmath205 and its reciprocal @xmath211 on a plot of the qpo evolution . the interesting result is that the total disk mass is a nearly monotonically decreasing function during the interval when the millihertz qpo frequency is increasing . despite the fact that the computed evolution of the disk mass looks very similar to the evolution of the x - ray luminosity , the two curves are not identical . the disk mass evolution is obtained from the light curve by the transformation which is effectively a shift transformation . while figure [ fig - m ] is not yet a complete model of the qpo evolution , we argue that it represents a very substantial advance over what was previously available . to date there has been no model which successfully explains the evolution of this low - frequency qpo . if the qpo is some large scale phenomenon in the disk then it should correlate with large scale disk properties . it fails completely to correlate with x - ray luminosity in the sense that @xmath204 rises and falls while qpo rises monotonically . the function @xmath211 , while not perfect , moves the turnover earlier than that seen for @xmath204 and hence closer to that observed in the qpo frequency history . we believe that this result , combined with a physical model for the qpo evolution represents a significant improvement in our understanding of the observed qpo . note that our model included several simplifying assumptions which when corrected may account for the remaining differences . we may now reconsider the global oscillation model ( to00 , ) in which a monotonic dependence of @xmath212 on the qpo frequency @xmath213 is derived , giving @xmath214 ( see eqs . [ 4 , 7 ] there ) . the observed dependence is flatter than the predicted one ; that discrepancy could be a result of our assumption of constant conversion efficiency @xmath215 of the mass accretion rate ( in fact , the surface density ) at the inner disk edge into x - ray luminosity . it is possible that the inner radius of the disk changes along with the outer disk radius , causing the variation of @xmath215 . the global oscillation is a vertical displacement of a large portion of the disk . the exact mechanism whereby this modulates x - ray ( as well as optical and euv ) emission to produce a qpo remains unclear . one possibility is that the modulation is geometrical in nature , _ e.g _ from varying obscuration of the hot inner portions of the disk . this would require the system to be at a fairly high inclination , as found by @xcite for this source . in addition , for this idea be viable the corona would have to be somewhat limited in extent . thus if the relevant disk oscillations were at @xmath216 ( where @xmath217 is the schwarzschild radius ) and the corona were confined to less than @xmath218 the geometrical model could work . other scenarios could include having the disk brightness respond in some way to the global oscillation . the best - fit parameter @xmath219 d in the green s function from equation ( 62 ) ( see fig . [ fig - green ] ) allows us to derive the kinematic viscosity value because @xmath220 from equation ( 49 ) . thus the kinematic viscosity is @xmath221 @xmath222 s@xmath223 since @xmath224 cm ( from eq . [ 10 ] in to00 for @xmath225 and using @xmath226 and the orbital period 4.1 hours ) . this estimate leads to @xmath227 cm if we assume that @xmath37 is of order of a thermal velocity at the outer edge of the disk , i.e. @xmath228 cm s@xmath223 . here we have presented a detailed mathematical analysis of the diffusion propagation model for x - ray outbursts . we investigated the intrinsic properties of the disk density evolution equation ( 5 ) in a general case . we have analyzed the diffusion models determined by the disk kinematic viscosity dependence on the radius . we have demonstrated that the green s ( response ) function is characterized by the fast rising and exponential decay parts ( fred shape ) and that it is an intrinsic property for a wide class of diffusion models . similar results were obtained by sunyaev & titarchuk ( 1980 ) for photon diffusion models . particularly , the light curve of the instant turn - on of a photon source placed in the center of a spherical cloud ( or the center disk plane ) has a typical fred type shape ( figs . 1 - 2 there ) . the success in fitting the diffusion model to the observed light curve is very promising . we have examined the case where the viscosity is a power law function of position in the disk . we note that a dependence of viscosity on surface density is also possible , but this will make the problem nonlinear ( see lyubarskii & shakura 1987 and lyubarskii 1997 for details ) . we showed in [ sec - inverse ] that the steepest descent method can be employed to solve the inverse problem ( a deconvolution ) and extract the input mass accretion function @xmath48 using the observed x - ray luminosity evolution . such a deconvolution is possible because the inner mass accretion rate and the x - ray rate production are related to each other and the timescale for viscous diffusion is short enough compared to the timescale of the outburst . using the derived @xmath48 as the source term and @xmath204 as a measure of the mass lost from the disk into the black hole , we derived the mass of the disk as a function of time . this appears to inversely correlate with the observed qpo frequency as would be expected from the global disk oscillation model , but the correlation is not perfect . there are several factors that have been ignored for simplicity in this initial treatment which could modify the derived disk mass as a function of time : ( 1 ) a changing efficiency of the conversion of mass to @xmath204 at the inner edge of the disk , ( 2 ) an effective viscous time ( @xmath100 ) which varies with time , and ( 3 ) an additional sink for disk matter such as a jet . we realize that the model is imperfect but we think that the initial application is encouraging and that further study and application to other sources is warranted . we conclude by summarizing the main results of the diffusion model . ( 1 ) the asymptotic form ( eq . [ 62 ] ) is characterized by the exponential rise when the escape time is less than the diffusion timescale from the location of mass injection . the same asymptotic form would be seen for a semi - infinite medium and thus is determined by the escape boundary condition ( in our case at @xmath74 ) and the location of the mass source . this asymptotic form is insensitive to the size of the medium . the second asymptotic form is characterized by an exponential decay with the viscous timescale @xmath100 which is a signature of the bounded medium ( in our case , a disk ) rather than any specific model of the diffusion . ( 2 ) the `` fred '' shape of the first outburst of xte j1118 + 480 is in good agreement with the diffusion green s function form ( eq . [ 62 ] ) , for which only one free parameter @xmath100 is important . the variation of the viscosity index @xmath91 leads to a change of @xmath100 but the `` fred '' shape remains . this particular result is very important because it is closely related to the intrinsic property of the diffusion green s function . ( 3 ) the value of the parameter @xmath100 has a physically plausible value for the viscous timescale of the disk and ultimately can be applied to the calculation of the disk turbulent scale using certain assumptions regarding the disk size and the turbulent velocity ( 4 ) . the rise of the second outburst follows the form expected for a step function input , the integral of the green s function . it is worthwhile to emphasize that the shape of this rise is precisely the integral of the green s function derived from the first outburst . ( 5 ) the derived mass varies approximately inversely with the qpo frequency . ( 6 ) similar x - ray luminosity and qpo evolution would be expected in other sources with similar behavior , and in fact is seen in x1354@xmath1644 and groj1719@xmath124 ( brocksopp et al . 2000 ) . in our model , the rising qpo frequency is a signature of the accretion disk mass decreasing as it disappears into the black hole . if qpo frequency varies inversely with disk mass throughout the outburst there is a possibility that early detection before the mass reaches its maximum might find the qpo when its frequency is still decreasing . both the relatively low brightness in the earliest stages of outbursts and the difficulties of prompt response would appear to have selected against detection of this effect to date . we would like to acknowledge the many helpful comments of an anonymous referee . basic research in x - ray astronomy at the naval research laboratory is supported by the office of naval research . this work was performed while rmb held a national research council research associateship award at nrl . this paper made use of quick - look results provided by the asm/_rxte _ team ( see http://xte.mit.edu ) .	we present a linear diffusion model for the evolution of the double - peaked outburst in the transient source xtej1118 + 480 . the model treats the two outbursts as episodic mass deposition at the outer radius of the disk followed by evolution of disk structure according to a diffusion process . we demonstrate that light curves with fast - rise , exponential decay profile are a general consequence of the diffusion process . deconvolution of the light curve proves to be feasible and gives an input function specifying mass deposition at the outer disk edge as well as the total mass of the disk , both as functions of time . the derived evolution of total disk mass can be correlated with the observed evolution of the @xmath0 hz qpo in the source reported in @xcite .
in this paper , we study the problem of discrete polynomial blending , a discretization of the problem of curve blending . in curve blending , one approximates a bivariate function by interpolating to a network of curves extracted from the graph of the function . discrete blending involves a second level of discretization , whereby the blended curves are interpolated at finite sets of points . this is illustrated in fig . [ fblend1 ] . in the figure , the blended interpolant would interpolate to the 9 curves ( 4 horizontal and 5 vertical ) in the network , while the discrete blended surface interpolates at @xmath0 grid points . in our work , the term `` interpolation '' can be taken loosely to mean interpolation with respect to a given set of functionals , and may not necessarily imply point evaluation . the grid in fig . [ fblend1 ] is not uniform due to gaps between some grid points . if we filled in these holes we would have a uniform grid with @xmath1 points that can be interpolated using tensor product polynomials . hence , we can interpolate at just @xmath0 points rather than @xmath2 , and in some cases do so with the same ( or nearly the same ) rate of approximation , as we shall show in this paper . in discrete polynomial blending , the approximating spaces are not generally tensor product polynomial spaces ( although that is a special case ) . as it turns out , our approximating spaces are the `` sum of tensor product polynomial spaces '' . hence , our study is one of approximation from the sum of tensor product polynomial spaces . what is known about discrete blending comes mainly from the literature on boolean sum interpolation , sparse grid methods , lower set interpolation and finite elements . the topic was perhaps first studied by biermann @xcite who constructed polynomial interpolants using the bivariate lagrange basis . the book @xcite is an excellent summary of boolean sum methods , including an analysis of biermann interpolation . in @xcite , a construction was given that generalized biermann interpolation to interpolation with respect to more general sequences of functionals , much like we will do in this paper . biermann interpolation was generalized to higher dimensions in @xcite , under the title `` d - variate boolean interpolation '' , and more recently to arbitrary `` lower sets '' in @xcite ( which reduces to the results in @xcite for total degree interpolation ) . in this paper we construct a new discrete blended quasi - interpolant based on the bernstein basis . to do so , we will bring in some techniques on `` dual basis in subspaces '' that the author has studied in @xcite . from this we construct dual bases for the space of discrete blending , and compute approximation estimates . one of the main contributions is to show our quasi - interpolants achieve rates of approximation comparable or the same as that of tensor product interpolation on a larger grid , but with much fewer data points . this leads us to the construction of a quasi - interpolant analogous to the serendipity elements in the finite element method . the results presented in this paper originate from a talk by the author at the conference on curves and surfaces in oslo , norway , in 2012 . the remainder of the paper is organized as follows . * the approximating space . * quasi - uniform grids . * the bernstein basis and univariate quasi - interpolant . * discretely blended bernstein - bzier quasi - interpolants . * order of approximation . * serendipity elements . * examples . in discrete polynomial blending , the approximating space is the algebraic sum of tensor product polynomial spaces . let @xmath3 $ ] and @xmath4 $ ] be sequences in @xmath5 , the space of @xmath6-tuples of non - negative integers . let @xmath7 be the tensor product of the spaces @xmath8 and @xmath9 of polynomials of degrees at most @xmath10 and @xmath11 , respectively . then , we define our approximating spaces as @xmath12 we assume that both @xmath13 and @xmath14 are strictly increasing sequences , an assumption that we justify by the following lemma . let @xmath15 be defined as in ( [ e1 ] ) for some @xmath13 and @xmath14 in @xmath5 . there exists strictly increasing sequences @xmath16 and @xmath17 in @xmath18 for some @xmath19 such that @xmath20 . to begin , let @xmath21 and let @xmath22 and @xmath23 . to prove this result , we will rearrange and truncate @xmath16 and @xmath17 until they are strictly increasing . suppose @xmath24 for some @xmath25 and @xmath26 . then either @xmath27 or @xmath28 . in the former case , @xmath29 can be removed from the representation in ( [ e1 ] ) without changing @xmath30 , and so we remove @xmath31 and @xmath32 from the sequences @xmath33 and @xmath34 . in the latter case , we remove @xmath35 and @xmath36 . after removing these unnecessary terms , the terms left in the revised sequence @xmath33 are distinct . that is , @xmath37 for all @xmath25 and @xmath26 . further , since @xmath38 is not affected by the ordering of the tensor product terms , we can rearrange the pairs @xmath39 in @xmath40 so that @xmath16 is strictly increasing . thus , assume that @xmath16 is strictly increasing . now , if @xmath34 is not strictly increasing , then there exists an index @xmath25 such that @xmath41 and @xmath42 , in which case @xmath43 , and so @xmath29 can be removed from the sum without changing @xmath30 . after trimming away all such terms in the sequence , we are left with @xmath34 strictly increasing . hence , we are left with strictly increasing sequences @xmath33 and @xmath34 in @xmath44 for some @xmath45 such that @xmath46 . further , since we are not adding new terms , @xmath19 . a polynomial @xmath47 can be represented @xmath48 with @xmath49 . each @xmath50 can be expressed @xmath51 for some coefficients @xmath52 . therefore , the power basis for the space @xmath15 is the union @xmath53 however , this union is not disjoint . this basis can visualized by the dots in a _ lower grid _ , which is defined to be the graph of the _ lower set _ \times [ 0 , \ldots , n_{r - k}].\ ] ] the power basis for @xmath15 is therefore @xmath55 , and the dimension of @xmath15 is the number of dots in the lower grid . by counting the distinct dots in the lower grid , we arrive at the following : [ prop0 ] the space @xmath15 is a vector space of dimension @xmath56 with @xmath57 . in tbl . [ f1a ] , the dimension of the approximating space is given for a few choices of @xmath13 and @xmath14 . in fig . [ f1b ] , the corresponding lower grids are plotted . . approximating spaces for : @xmath58 $ ] , @xmath59 $ ] , @xmath60 $ ] and @xmath61 $ ] . [ cols="<,<,<,<",options="header " , ] [ 6b ] in this paper we have constructed a new quasi - interpolant for discrete blended surface approximation based on dual basis functions in the bernstein basis , and we have established error estimates comparable to approximation on full tensor product grids , much like the serendipity elements in the finite element literature . throughout this paper we assumed @xmath62 and @xmath63 . if we do not have this , we can still apply our construction by first using degree elevation . for example , if @xmath64 $ ] we would degree elevation to get @xmath65 $ ] , and then proceed as before . the framework for this originates from the talk `` dual bases on subspaces and the approximation from sums of polynomial and spline spaces '' , given by the author at the conference `` mathematical methods for curves and surfaces '' , held in oslo norway , june , 2012 . the talk included constructions for hermite and spline blended elements similar to the construction in this paper . our plan is to publish the details of these results in forthcoming papers .	in this paper we study `` discrete polynomial blending , '' a term used to define a certain discretized version of curve blending whereby one approximates from the `` sum of tensor product polynomial spaces '' over certain grids . our strategy is to combine the theory of boolean sum methods with dual bases connected to the bernstein basis to construct a new quasi - interpolant for discrete blending . our blended element has geometric properties similar to that of the bernstein - bzier tensor product surface patch , and rates of approximation that are comparable with those obtained in tensor product polynomial approximation . [ multiblock footnote omitted ]
given an @xmath6 hermitian matrix @xmath1 , the _ spectral density _ , also commonly referred to as the _ density of states _ ( dos ) , is formally defined as @xmath7 here @xmath8 is the dirac distribution commonly referred to as the dirac @xmath8-``function '' ( see e.g. @xcite ) , and the @xmath9 s are the eigenvalues of @xmath1 , assumed here to be labeled non - decreasingly . the dos is an important quantity in many physics problems , in particular in quantum physics , and a large volume of numerical methods were developed by physicists and chemists @xcite for this purpose . besides being used as a qualitative visualization tool for understanding spectral characteristics of the matrix , the dos can also be used as to quantitatively compute the trace of a matrix function , as given in the formal formulation below @xmath10 = \sum_{i=1}^{n } f(\lambda_{i } ) \equiv n \int_{-\infty}^{\infty } f(t)\phi(t ) { \,\mathrm{d}}t . \label{eqn : tracefa}\ ] ] here @xmath11 is a smooth function , and the formal integral in eq . should be interpreted in the sense of distribution . if one had access to all the eigenvalues of @xmath1 , the task of computing the dos would become a trivial one . however , in many applications , the dimension of @xmath1 is large . the computation of its entire spectrum is prohibitively expensive , and a procedure that relies entirely on multiplications of @xmath1 with vectors is the only viable approach . fortunately , in many applications @xmath1 only has @xmath5 nonzero entries , and therefore the cost of matrix - vector multiplication , denoted by @xmath12 , is @xmath5 . in some other cases the matrix is a dense matrix but fast matrix - vector multiplication method still exists with @xmath13 cost , where @xmath14 is a integer that is not too large . this is the case when the matrix - vector multiplication can be carried out effectively with fast algorithms , such as the fast fourier transform ( fft ) , the fast multipole method ( fmm ) @xcite , the hierarchical matrix @xcite , and the fast butterfly algorithm @xcite , to name a few . rigorously speaking , the dos is a distribution and can not be directly approximated by smooth functions . in order to assess the accuracy of a given numerical scheme for estimating the dos , the dos must be properly _ regularized_. the basic idea for estimating the dos is to first expand the regularized dos using simple functions such as polynomials . then it can be shown that the estimation of the dos can be obtained by computing the trace of a polynomial of @xmath1 , which can then be estimated by repeatedly applying @xmath1 to a set of random vectors . this procedure has been discovered more or less independently by statisticians @xcite and by physicists and chemists @xcite , and will be referred to as hutchinson s method in the following . in physics such method is often referred to as the kernel polynomial method ( kpm ) @xcite with a few different variants . a recent review on the choice of regularization and different numerical methods for estimating the dos is given in @xcite . there are also a variety of randomized estimators that can be used in hutchinson s method , and the quality of different estimators is analyzed in @xcite . * contribution . * to the extent of our knowledge , all randomized methods so far for estimating the dos are based on different variants of hutchinson s method . these methods estimate the dos by averaging the information obtained from @xmath3 random vectors directly . the numerical error , when properly defined , decays asymptotically as @xmath15 . as a result , high accuracy is difficult to achieve : every extra digit of accuracy requires increasing the number of random vectors by @xmath16 fold . in this work , we demonstrate that the accuracy for estimating the regularized dos can be significantly improved by making use of the _ correlated information _ obtained among different random vectors . we use the fact that each point of the dos can be evaluated as the trace of a numerically low rank matrix , and such trace can be evaluated by repeatedly applying @xmath1 to a small number of random vectors , and by taking certain linear combination of the resulting vectors . if different set of random vectors were needed for different points on the spectrum the method will be prohibitively expensive . however , we demonstrate that it is possible to use _ the same set of random vectors _ to `` sweep through '' in principle the entire spectrum . therefore we call our method a `` spectrum sweeping method '' . our numerical results indicate that the spectrum sweeping method can significantly outperform hutchinson type methods in terms of accuracy , as the number of random vectors @xmath3 becomes large . however , the computational cost and the storage cost can still be large when the dos needs to be evaluated at a large number of points . furthermore , the accuracy of the spectrum sweeping method may be compromised when the right number of randomized vectors is not known _ a priori_. we develop a robust and efficient implementation of the spectrum sweeping method to overcome these two problems . under certain assumption on the distribution of eigenvalues of the matrix @xmath1 , and the cost of the matrix - vector multiplication is @xmath5 , we demonstrate that the computational cost of the new method scales as @xmath4 and the storage cost scales as @xmath5 for increasingly large matrix dimension @xmath17 . we also demonstrate that the new method for evaluating the dos can be useful for accurate trace estimation as in eq . . * other related works . * the spectrum sweeping method is not to be confused with another set of methods under the name of `` spectrum slicing '' methods @xcite . the idea of the spectrum slicing methods is still to obtain a partial diagonalization of the matrix @xmath1 . the main advantage of spectrum slicing methods is enhanced parallelism compared to conventional diagonalization methods . due to the natural orthogonality of eigenvectors corresponding to distinct eigenvalues of a hermitian matrix , the computational cost for each set of processors handling different parts of the spectrum can be reduced compared to direct diagonalization methods . however , the overall scaling for spectrum slicing methods is still @xmath18 when a large number of eigenvalues and eigenvectors are to be computed . * notation . * in linear algebra notation , a vector @xmath19 is always treated as a column vector , and its conjugate transpose is denoted by @xmath20 . for a randomized matrix @xmath21 , its entry - wise expectation value is denoted by @xmath22 $ ] and its entry - wise variance is denoted by @xmath23 $ ] . we call @xmath24 a ( real ) random gaussian matrix , if each entry of @xmath25 is real , and follows independently the normal distribution @xmath26 . in the case when @xmath27 , @xmath25 is called a ( real ) random gaussian vector . the imaginary unit is denoted by @xmath28 . the paper is organized as follows . in section [ sec : dos ] we introduce the dos estimation problem . we also demonstrate the delta - gauss - chebyshev ( dgc ) method , which is a variant of the kernel polynomial method , to estimate the dos . we develop in section [ sec : doslowrank ] the spectrum sweeping method based on the randomized estimation of the trace of numerically low rank matrices , and demonstrate a robust and efficient implementation of the spectrum sweeping method in section [ sec : robustdoslowrank ] . we show how the dos estimation method can be used to effective compute the trace of a matrix function in section [ sec : traceest ] . following the numerical results in section [ sec : numer ] , we conclude and discuss the future work in section [ sec : conclusion ] . without loss of generality , we shall assume that the spectrum of @xmath1 is contained in the interval @xmath29 . for a general matrix with spectrum contained in the interval @xmath30 , we can apply a spectral transformation @xmath31 the spectrum of @xmath32 is contained in @xmath29 , and all the discussion below can be applied to @xmath32 . the fact that the spectral density @xmath33 is defined in terms of dirac @xmath8-functions suggests that no smooth function can converge to the spectral density in the limit of high resolution , in the usual @xmath34 norms ( @xmath35 ) @xcite . in order to compare different numerical approximations to the spectral density in a meaningful way , the dos should be regularized . one simple method is to employ a gaussian regularization @xmath36 . \label{}\ ] ] here @xmath37 is a gaussian function . in the following our goal is to compute the smeared dos @xmath38 . the key of computing the dos is the estimation of the trace of a matrix function without diagonalizing the matrix . to the extent of our knowledge , randomized methods developed so far are based on hutchinson s method or its variants @xcite . the following simple and yet useful theorem is a simple variant of hutchinson s method and explains how the method works . [ thm : evartrace ] let @xmath0 be a hermitian matrix , and @xmath39 be a random gaussian vector , then @xmath40 = \operatorname{tr}[a ] , \quad \mathrm{var } [ w^ { * } a w ] = 2\sum_{i\ne j } ( { \mathrm{re}}a_{ij})^2 . \label{eqn : evartrace}\ ] ] first @xmath40 = \mathbb{e}\left[\sum_{ij } w_{i } w_{j } a_{ij}\right ] = \sum_{i } a_{ii } = \operatorname{tr}[a].\ ] ] here we used that @xmath41= 0 , \quad \mathbb{e } [ w w^ { * } ] = i$ ] and @xmath42 is a real vector . this follows directly from that each entry of @xmath42 is independently distributed and follows the gaussian distribution @xmath26 . second , @xmath43 = & \mathbb{e } \left[(w^ { * } a w)^2 - ( \operatorname{tr}[a])^2\right ] = \mathbb{e } \left[\sum_{ijkl } w_{i } w_{j } w_{k } w_{l } a_{ij } a_{kl } - ( \operatorname{tr}[a])^2\right]\\ = & \sum_{ik } a_{ii } a_{kk } + 2 \sum_{i\ne j } ( { \mathrm{re}}a_{ij})^2 - ( \sum_{i } a_{ii})^2 = 2 \sum_{i\ne j } ( { \mathrm{re}}a_{ij})^2 . \end{split}\ ] ] using theorem [ thm : evartrace ] , if we choose @xmath44 to be a random gaussian matrix , then @xmath45 \approx \frac{1}{n_{v}}\operatorname{tr}[w^{*}g_{\sigma}(ti - a ) w].\ ] ] in practice in order to compute @xmath46 , we can expand @xmath47 into polynomials of @xmath1 for each @xmath48 , and then evaluate the trace of polynomial of @xmath1 . a stable and efficient implementation can be obtained by using chebyshev polynomials . other choices of polynomials such as legendre polynomials can be constructed similarly . using the chebyshev polynomial , @xmath49 is approximated by a polynomial of degree @xmath50 as @xmath51 the coefficients @xmath52 need to be evaluated for each @xmath48 . since the dirac @xmath8-function is regularized using a gaussian function , following the notion in @xcite we refer to eq . as the `` delta - gauss - chebyshev '' ( dgc ) expansion . since @xmath53 is a smooth function on @xmath29 , the coefficient @xmath54 in the dgc expansion can be computed as @xmath55 here @xmath56 is the kronecker @xmath8 symbol . with change of coordinate @xmath57 , and use the fact that @xmath58 , we have @xmath59 eq . can be evaluated by discretizing the interval @xmath60 $ ] using a uniform grid , and the resulting quadrature can be efficiently computed by the fast fourier transform ( fft ) . this procedure is given in alg . [ alg : dgcexpansion ] , and this procedure is usually inexpensive . [ cols= " < , < " , ] for large hermitian matrices that the only affordable operation is matrix - vector multiplication , randomized algorithms can be an effective way for obtaining a rough estimate the dos . however , so far randomized algorithms are based on hutchinson s method , which does not use the correlated information among different random vectors . the accuracy is inherently limited to @xmath2 where @xmath3 is the number of random vectors . we demonstrate that randomized low rank decomposition can be used as a different mechanism to estimate the dos . by properly taking into account the correlated information among the random vectors , we develop a spectrum sweeping ( ss ) method that can sweep through the spectrum with a reasonably small number of random vectors and the accuracy can be substantially improved compared to @xmath2 . however , for spectrally uniformly distributed matrices with a large number of points to evaluate the dos , the direct implementation of the spectrum sweeping method can have @xmath18 complexity . we also present a robust and efficient implementation of the spectrum sweeping method ( ress ) . for spectrally uniformly distributed matrices , the complexity for obtaining the dos can be improved to @xmath4 , and the extra robustness comes from a hybridization with hutchinson s method for estimating the residual . we demonstrate how the regularized dos can be used for estimating the trace of a smooth matrix function . this is based on a careful balance between the smoothness of the function and that of the dos . such balance is implemented through a deconvolution procedure . numerical results indicate that this allows the accurate estimate of the trace with again @xmath4 scaling . the current implementation of the spectrum sweeping method is based on chebyshev polynomials . motivated from the discussion in @xcite , lanczos method would be more efficient than chebyshev polynomials for estimating the dos , and it would be interesting to extend the idea of spectrum sweeping to lanczos method and compare with chebyshev polynomials . we also remark that the spectrum sweeping method effectively builds a low rank decomposition near each point on the spectrum for which the dos is to be evaluated . combining the deconvolution procedure as in the trace estimation and the low rank decomposition could be potentially useful in some other applications to directly estimate the whole or part of a matrix function . for instance , in electronic structure calculation , the diagonal entries of the fermi - dirac function is needed to evaluate the electron density . these directions will be explored in the future . this work is partially supported by the scientific discovery through advanced computing ( scidac ) program and the center for applied mathematics for energy research applications ( camera ) funded by u.s . department of energy , office of science , advanced scientific computing research and basic energy sciences , and by the alfred p. sloan fellowship . we thank the anonymous referees for their comments that greatly improved the paper . g. a. parker , w. zhu , y. huang , d. hoffman , d. j. kouri , matrix pseudo - spectroscopy : iterative calculation of matrix eigenvalues and eigenvectors of large matrices using a polynomial expansion of the dirac delta function , comput . 96 ( 1996 ) 2735 . h. m. aktulga , l. lin , c. haine , e. g. ng , c. yang , parallel eigenvalue calculation based on multiple shift invert lanczos and contour integral based spectral projection method , parallel comput . 40 ( 2014 ) 195212 .	for a large hermitian matrix @xmath0 , it is often the case that the only affordable operation is matrix - vector multiplication . in such case , randomized method is a powerful way to estimate the spectral density ( or density of states ) of @xmath1 . however , randomized methods developed so far for estimating spectral densities only extract information from different random vectors independently , and the accuracy is therefore inherently limited to @xmath2 where @xmath3 is the number of random vectors . in this paper we demonstrate that the `` @xmath2 barrier '' can be overcome by taking advantage of the correlated information of random vectors when properly filtered by polynomials of @xmath1 . our method uses the fact that the estimation of the spectral density essentially requires the computation of the trace of a series of matrix functions that are numerically low rank . by repeatedly applying @xmath1 to the same set of random vectors and taking different linear combination of the results , we can sweep through the entire spectrum of @xmath1 by building such low rank decomposition at different parts of the spectrum . under some assumptions , we demonstrate that a robust and efficient implementation of such spectrum sweeping method can compute the spectral density accurately with @xmath4 computational cost and @xmath5 memory cost . numerical results indicate that the new method can significantly outperform existing randomized methods in terms of accuracy . as an application , we demonstrate a way to accurately compute a trace of a smooth matrix function , by carefully balancing the smoothness of the integrand and the regularized density of states using a deconvolution procedure .
the cdf @xcite and d0 @xcite collaborations have recently announced strong evidence for the existence of the top quark , the isospin partner to the @xmath0 quark required in the standard model ( sm ) , using @xmath1 and @xmath2 data samples respectively of @xmath3 collisions at @xmath4 tev . a signal consistent with @xmath5 has been observed , exceeding the background prediction by @xmath6 @xcite and @xmath7 @xcite . in refs . @xcite , @xcite the branching ratio br ( @xmath8 ) is taken to be @xmath9 , and this is a valid assumption in the minimal sm ( i.e. one higgs doublet ) ; the decays @xmath10 , @xmath11 are absent at tree level due to the gim mechanism , and the charged current processes @xmath12 , @xmath13 are negligible due to heavy ckm matrix suppression ( @xmath14 is usually predicted in such models and could be quite large due to higgs bosons coupling in proportion to mass . in some extended models ( although not all ) , @xmath15 is still allowed by current electroweak precision tests , and if such a light @xmath16 exists then on shell @xmath17 decays will occur . this would provide an alternative decay channel for the top quark and is an option not considered in refs . it is the aim of this work to examine whether the presence of a light @xmath16 is compatible or not with the cdf data . we shall consider in particular the case of @xmath18 gev i.e. @xmath16 within the discovery range of lep2 . the paper is organised as follows . in section 2 we briefly review the various extended higgs models that may contain a light @xmath16 . section 3 examines how significant the channel @xmath17 can be , while section 4 studies how one would search for @xmath16 . in section 5 we apply the analyses of sections 3 and 4 to the current data sample from the tevatron , while section 6 considers prospects at an upgraded tevatron and at the large hadron collider ( lhc ) . finally section 7 contains our conclusions . the minimal sm consists of one higgs doublet ( @xmath19 , @xmath20 ) , although extended sectors can be considered and have received substantial attention in the literature . for a general review there are two main constraints on such models : * there must be an absence of flavour changing neutral currents ( fcnc ) . * the rho parameter , @xmath21 , must be very close to one . condition ( i ) is satisfied by constraining the yukawa couplings to the fermions @xcite . condition ( ii ) requires models with only doublets , to which any number of singlets ( @xmath22 , @xmath23 ) can be added . models with triplets ( @xmath24 ) can also be considered , although obtaining @xmath25 is achieved in a less natural way than for cases with only doublets . the theoretical structure of the two higgs doublet model ( 2hdm ) is well known @xcite , while the general multi higgs doublet model ( mhdm ) @xcite has received substantially less attention . in the mhdm it is conventional to assume that one of the charged scalars is much lighter than the others and thus dominates the low energy phenomenology doublets there exists @xmath26 @xmath16s . ] . the relevant part of the lagrangian for the 2hdm and mhdm can be written as @xcite : @xmath27 here @xmath28 , @xmath29 ( @xmath30 , @xmath31 ) denote left and right handed up ( down ) type quark fields , @xmath32 is the left handed neutrino field , and @xmath33 the right handed charged lepton field . @xmath34 , @xmath35 , @xmath36 are the diagonal mass matrices of the down type quarks , up type quarks and charged leptons respectively . @xmath37 is the ckm matrix , and @xmath38 , @xmath39 and @xmath40 are coupling constants ( see below ) . the cp conserving 2hdm which is usually considered in the literature @xcite contains an important parameter @xmath41 with @xmath42 and @xmath43 being real vacuum expectation values ( vevs ) of the two higgs doublets and @xmath44 . there are 4 variants of the 2hdm depending on how the doublets are coupled to the fermions . their coupling constants are given in table 1 @xcite . ii@xmath45 + @xmath38 & @xmath46 & @xmath46 & @xmath47 & @xmath47 + @xmath39 & @xmath48 & @xmath48 & @xmath48 & @xmath48 + @xmath40 & @xmath46 & @xmath47 & @xmath47 & @xmath46 + in the mhdm @xmath38 , @xmath39 and @xmath40 are _ arbitrary _ complex numbers . it follows that combinations of parameters like @xmath49 have different values depending on the model under consideration , thus leading to phenomenological differences . this has important consequences , particularly when one calculates loop diagrams involving @xmath16 . one such decay , @xmath50 , is sensitive to charged scalars and has recently been observed for the first time by the cleo collaboration . the value for the branching ratio was measured to be @xcite @xmath51 which corresponds to @xmath52 the theoretical calculation of the branching ratio appears in refs . @xcite , @xcite , @xcite , @xcite . from this it can be shown @xcite , @xcite , @xcite that the above bound constrains @xmath53 from the 2hdm ( model ii and ii@xmath45 ) to be @xmath54 gev , while no bound can be obtained on @xmath53 from the 2hdm ( model i and i@xmath45 ) and the mhdm . hence it is possible that an on shell @xmath16 from these latter models contributes to top decay , and may even be light enough to be detectable at lep2 @xcite , @xcite . also , we note that the most popular model with higgs isospin triplets ( htm ) @xcite , @xcite , predicts a charged scalar @xmath55 which has exactly the same couplings as @xmath16 ( 2hdm , model i ) , and thus may also contribute to top decay and/or be detectable at lep2 @xcite . for all these @xmath16s there exists a lower bound from lep @xcite of @xmath56 gev . an important constraint on @xmath48 is obtained from precision measurements of the @xmath57 vertex . charged scalars with a tree level @xmath58 coupling contribute to this decay , and ref . @xcite shows that @xmath59 this bound is for @xmath60 gev and @xmath61 gev , and provides a stronger constraint on @xmath62 than can be obtained from the decay @xmath50 . in order to search for @xmath16 one needs to consider its decays . the various branching ratios ( brs ) are shown in table 2 , with @xmath63 referring to either an electron or a muon . @xmath64 + @xmath16 ( 2hdm , model i ) & @xmath65 & @xmath66 & @xmath67 + @xmath16 ( 2hdm , model i@xmath45 ) & @xmath68 & @xmath69 & @xmath67 + @xmath16 ( mhdm ) & @xmath70 & @xmath70 & @xmath67 + @xmath71 & @xmath65 & @xmath72 & @xmath73 + for the 2hdm ( model i ) the brs are independent of @xmath47 , while in the 2hdm ( model i@xmath45 ) the constraint of eq . ( 5 ) creates the inequalities in table 2 . in the mhdm the brs are dependent on the arbitrary parameters @xmath38 , @xmath39 and @xmath40 , and thus span the full range @xmath74 . in this section we evaluate how competitive the decay channel @xmath75 can be compared to the conventional @xmath76 . the feynman rule for the @xmath58 vertex is given by @xcite @xmath77\;.\ ] ] here @xmath78 is a ckm matrix element and @xmath79 is the usual su(2 ) gauge coupling . the 2hdm ( models i and i@xmath45 ) requires the replacements @xmath80 and @xmath81 . from eq . ( 6 ) one can evaluate the partial width and thus br @xmath82 @xcite : @xmath83\cot^2\beta \over m^2_w(m^2_t+m^2_b-2m^2_w)+(m^2_t - m^2_b)^2}\;.\ ] ] here @xmath84 and @xmath85 refer to the magnitude of the three momentum of the @xmath86 and @xmath87 measured in the rest frame of the top quark . we take @xmath88 gev , @xmath89 gev and @xmath90 gev . in figure [ fig : top6 ] we plot br @xmath82 as a function of @xmath47 for @xmath91 gev ( comfortably in the lep2 range ) , for @xmath92 ( at the edge of the lep2 range ) , and for @xmath93 gev ( out of lep2 range ) . we recall that eq . ( 5 ) states @xmath94 or equivalently @xmath95 . using this bound we see from figure 1 that large values of br @xmath82 are still possible e.g. for @xmath91 gev and @xmath96 , br @xmath97 . as @xmath47 increases br @xmath82 falls , dropping to @xmath98 at @xmath99 . hence for large enough @xmath47 and/or @xmath53 the top decay to charged higgs will be difficult to detect experimentally . we shall be particularly interested in the case of @xmath100 i.e. for @xmath16 in the discovery range of lep2 . in the mhdm , br @xmath101 ) depends not on @xmath47 but instead on @xmath102 and @xmath103 . however , it is easily verified from eq . ( 7 ) ( unless @xmath102 is unnaturally large see below ) that the strongly dominant part is proportional to @xmath104 which has the same constraint as @xmath48 ( eq . 5 ) . therefore figure 1 can be used for the mhdm , with the replacement @xmath105 on the ordinate axis . we note that there exists an unnatural region of parameter space , @xmath106 and @xmath107 , coming from the measurement of @xmath50 . @xcite uses @xmath108 for @xmath18 gev , although with the new measurement in eq . ( 3 ) , @xmath109 is more accurate . ] which would allow larger brs in the mhdm for a given @xmath53 than is ever possible in the 2hdm . we shall not consider this possibility . we shall consider two ways with which to detect @xmath16 from top decays . the first method is to search in the lepton channel . we see from table 2 that the decays of @xmath16 violate lepton universality due to a preference to couple to the heaviest lepton ( @xmath110 ) . this is in contrast to the lepton decays of @xmath71 . the second method is to search for the quark decays of @xmath16 by reconstructing the invariant masses of the jets . a significant peak separate from that of @xmath71 would be distinctive . a search using the di lepton channel @xmath111 ( via @xmath112 ) has been performed @xcite although we shall not consider this method here . we shall focus first on the @xmath113 decays of @xmath16 . a previous search in this channel has been performed at the tevatron using a @xmath114 data sample @xcite . here br ( @xmath17)=100% was assumed which we know now to be untrue since @xmath115 , and so on shell @xmath76 decays are definitely present . in ref . @xcite the process @xmath116 was searched for with the trigger being missing @xmath117 gev . from figure 1 we see that br ( @xmath118 can be as low as @xmath119 , and in such cases the above detection method is certainly not relevant for future searches . one must rely on the process @xmath120 , with a hard isolated lepton ( @xmath121 or @xmath122 ) to act as a trigger @xcite , @xcite . we also require a @xmath110 jet to which various cuts will be made @xcite . non @xmath123 backgrounds , although sizeable , are substantially reduced when the above requirements are applied . the main background is from @xmath124 which fakes the signal from @xmath125 and is usually the more common top decay process . fortunately this background can be reduced by making use of the polarisation of the @xmath126 . to be specific we shall focus on @xmath127 and @xmath128 , thus following the notation of ref . @xcite . weak vector bosons couple to left handed fermions and right handed antifermions , and thus @xmath129 . however this is not so for the charged higgs ( @xmath128 ) which couples to right handed fermions via @xmath130 . this is a consequence of the yukawa couplings . it has been known for some time that the momentum distribution of the decay products from @xmath126 depend on its polarisation @xcite . @xcite analyses these differences for the hadronic decays @xmath131 . it concludes that the energetic particles coming from @xmath127 are primarily @xmath132 and @xmath133 , with the subscript @xmath134 referring to a transversely polarised meson . in comparison , the energetic particles coming from @xmath128 are @xmath135 , @xmath136 , and @xmath137 with @xmath138 referring to a longitudinally polarised meson . the case of the energetic @xmath137 from @xmath128 can easily be understood in the following way . due to polarisation of the parent @xmath139 , spin conservation favours emission of the @xmath137 in the direction of motion of the @xmath139 . this results in the @xmath140 spectrum in the context of @xmath123 production of the @xmath137 being boosted compared to those pions coming from @xmath127 decay ( in the latter case the @xmath137 prefers to be emitted _ opposite _ to the direction of the parent @xmath139 ) . therefore the fragmentation function of @xmath137 from @xmath141 is harder than that from @xmath142 , and so a hard pion is a signal of @xmath128 decay . the subsequent decays of @xmath143 and @xmath144 , @xmath145 have different energy distributions depending on the polarisation of the parent meson . ref . @xcite shows that for the @xmath135 the final state pions will be much more asymmetric in energy than those originating from @xmath133 . for the case of the @xmath132 the three pions tend to share the energy equally , while for the @xmath136 one or two of them are soft with the rest hard . hence the signature of the @xmath128 is a final state of one or more energetic pions . it is straightforward to evaluate the theoretical prediction for the number of energetic pions from top decay . we take first the case of @xmath146 . one of the @xmath147s is decayed to @xmath64 to act as a trigger for the event , while the other is decayed to @xmath113 . the theoretical prediction is : @xmath148 here @xmath149 is a shorthand for br ( @xmath17 ) and @xmath150 is the number of @xmath123 pairs produced . the parameter @xmath151 is an overall efficiency for detecting an energetic pion from the above event , incorporating any kinematical cuts . it can be written as @xmath152 the summation is over the three possible decays of @xmath139 to pions : @xmath153 , @xmath154 are extra pions originating from @xmath155 , and @xmath156 respectively , while @xmath157 . for the case of @xmath158 theory predicts @xmath159 with @xmath160 being different to @xmath151 . this latter difference arises due to the polarisation of the parent @xmath110 and thus @xmath161 . there are two ways of searching for violation of lepton universality . the first method ( applied in ref . @xcite to the ssc ) requires an isolation cut on the energetic pion . this eliminates virtually all of the qcd background , although also has the disadvantage of removing most of the hard pions originating from @xmath162 and @xmath163 . a second method ( suggested in ref . @xcite ) aims to include these latter decays by rejecting the isolation cut , and instead making a cut on @xmath164 carried by the neutral pions . the signal favours large @xmath165 . this method keeps more of the signal than would be kept by making the isolation cut , although the qcd background is larger . the lepton trigger efficiency , @xmath166 and @xmath167 , depends only on @xmath168 and is therefore the same for both events . we will ignore the case of both top quarks decaying to a charged higgs ( @xmath169 ) for the following reasons : * br @xmath170 could well be very small ( see figure 1 ) . * the trigger here would be @xmath171 . this process would be suppressed in models with lower br ( @xmath172 ) i.e. 2hdm ( model i ) , htm and various mhdm , coupled with the fact that br @xmath173 . * the trigger decay in ( ii ) predicts @xmath63 with less @xmath140 than for the @xmath63 originating from @xmath174 . hence it is less likely to pass the lepton trigger cut . reasons ( ii ) and ( iii ) also allow us to ignore events @xmath175 which fake @xmath176 . we can quantify the amount of deviation from universality that would be caused by a non zero @xmath177 : @xmath178 with @xmath179 being the number of standard deviations by which the observed number of hard , isolated pions exceeds that predicted from universality . this concludes our account of the search for @xmath172 . the second method is to consider the @xmath180 decay and reconstruct the invariant masses of the jets . this method would be needed for the case of a ` leptophobic ' higgs i.e. br @xmath181 and br @xmath182 , which is possible in the mhdm for large @xmath103 and small @xmath102 , @xmath183 . the trigger will again be a hard isolated lepton originating from @xmath184 . three jets need to be reconstructed in the opposite hemisphere to the lepton ( i.e. the jets originating from @xmath185 ) , with one of them a tagged @xmath0 jet ( efficiency @xmath186 ) . @xcite deals with this method applied to the ssc , and the analysis is also relevant for the tevatron with a few minor modifications e.g. the @xmath140 cuts given in ref . @xcite are larger than those required at the tevatron due to @xmath187 at the ssc being much greater . adding the invariant masses of the non @xmath0 jets will result in a peak centered on @xmath53 . however , one must be careful here because in the mhdm it is possible to have a large br @xmath188 @xcite ; in this scenario there is a chance of the wrong @xmath0 jet being tagged and therefore the wrong jets would be used to reconstruct the invariant mass of @xmath16 . however , br @xmath189 would be needed to cause significant problems here , and ref . @xcite shows that the parameter space for this large branching ratio is quite small . requiring that the three jet invariant mass is centered on @xmath168 reduces the background further . a clear peak would be strong evidence for @xmath16 . @xcite suggests the following simple rule for this method to provide a significant peak at @xmath53 : @xmath190 this result was obtained by using the isajet monte carlo , comparing the size of the @xmath16 peak to the @xmath71 background for different values of @xmath47 and @xmath53 . it is only valid for a large event sample ( at the ssc @xmath191 ) , and so will be more relevant at an upgraded tevatron and/or lhc . if eq . ( 12 ) is satisfied , the statistical significance of the peak at @xmath53 is greater with a larger event sample , i.e. at the lhc . we note that eq . ( 12 ) was derived using ssc detection efficiencies ; however it will still be valid at the tevatron to a good approximation . contributes to the relative size of the peaks at @xmath192 and @xmath53 . this is the only factor which may invalidate eq . ( 12 ) for use at the tevatron . ] in this section we have studied two direct ways with which one may search for @xmath16 . in the next section we shall apply these techniques to the current data sample from the tevatron to see if a light ( @xmath193 gev ) charged higgs boson could have been observed . the current cdf data sample at the tevatron [ 1 ] contains @xmath194 . for @xmath195 we shall use the mrs(a ) partons @xcite which suggests a value of 4.5 pb if @xmath88 gev pb . ] @xcite . @xcite estimates that the theoretical error in @xmath195 is @xmath196 . we can now evaluate the expected number of @xmath123 pairs , and then use eqs . ( @xmath197 ) to predict the number of hard pions for various values of br @xmath170 . we shall take @xmath198 ; from table 2 one has br @xmath199 and br @xmath200 . if there is no charged higgs boson i.e. br @xmath201 then one has from eq . ( 8) : @xmath202 the efficiency @xmath151 is given by eq . ( 9 ) , and we shall discuss its value below . the prediction for @xmath177 is given by eq . we shall take br @xmath203 for illustrative purposes i.e. the maximum br for @xmath91 gev ( see figure 1 ) . thus we obtain : @xmath204 we shall consider the search strategy using an isolation cut on the charged pion . as mentioned before this eliminates most of the qcd background . the charged pion is required to contain at least @xmath205 of the energy in a cone of size @xmath206 , as well as possessing @xmath207 gev . this removes most of the @xmath147 background contribution from @xmath208 and @xmath209 and most of the signal contribution from @xmath209 @xcite . therefore we may use the following approximations in eq . ( 9 ) : @xmath210 the efficiency @xmath160 will be larger than @xmath151 for the following reasons . most importantly , @xmath211 i.e. pions originating from @xmath16 are more energetic ( see section 4 ) and are more likely to pass an energy cut than those from @xmath71 . using the figures in ref . @xcite we can approximate @xmath212 . also we want to maximise the signal for illustrative purposes and so take @xmath213 the @xmath214 gev cut has little effect on @xmath215 and @xmath216 ; the isolation cut hardly affects the @xmath217 contribution but removes part of the @xmath208 contribution . hence in reality @xmath218 . for heavier @xmath16 the pion is on average more energetic and so @xmath219 will increase . the trigger efficiencies for an isolated , @xmath220 gev lepton ( @xmath166 , @xmath167 ) are close to 100% @xcite . combining the above efficiencies , it is clear that @xmath221 . from table 2 we can obtain the value of br @xmath222 for the various higgs models . the maximum signal is obtained when br @xmath223 ) for @xmath224 and in the mhdm . however , in the calculation we also use br @xmath203 , and both these brs can only be satisfied simultaneously in the mhdm for @xmath225 and @xmath183 large . ] and we will use this in our calculation . the remaining parameter values needed are br ( @xmath226 and br ( @xmath227 . using the above values for efficiencies and brs in eqs . ( 13 ) and ( 14 ) we find @xmath228 using eq . ( 11 ) we see that to obtain a @xmath229 signal @xmath230 is required and this is far from the prediction of 11.35 . also , we used the most favourable values for br @xmath231 , br @xmath170 and @xmath160 . lower values for these parameters would decrease @xmath177 further e.g. for @xmath91 gev , @xmath232 and br @xmath233 ( model i ) we find @xmath234 . therefore we conclude that , over the vast majority of parameter ( @xmath53 , @xmath47 ) space , @xmath16 would not provide a statistically significant signal given the current data at the tevatron . the other method available to us exploits the quark decays of @xmath16 . from eq . ( 10 ) , with br @xmath235 replacing br @xmath236 ) , we again see that too few events are produced . the conclusion of the above analyses is that a direct signature of @xmath16 can not be obtained using the current data sample from the tevatron . however an indirect signature might be possible . the top quark search [ 1 ] relied on an excess of @xmath237 jet events over the qcd background . the events searched for consisted of one @xmath147 providing the hard lepton trigger , while the other decays to two quarks . a light @xmath16 can mimic this signal although it would suppress the expected number of events in this channel ( @xmath238 , and so increase the experimentally measured cross section for @xmath123 production . we shall now quantify this for various values of br ( @xmath76 ) . the experimentally measured cross section ( @xmath239 ) can be written as @xmath240 with @xmath241 ( @xmath242 ) being a detection efficiency for the @xmath147 plus 4 jets channel , and @xmath243 is a parameter depending on branching ratios and machine luminosity . to a good approximation the efficiency @xmath241 remains the same whether we input a charged higgs or not . is independent of @xmath53 while @xmath244 depends relatively weakly on @xmath53 , especially in the mass range 50 gev @xmath245 80 gev which gives larger br ( @xmath17 ) . ] however , the presence of @xmath16 will change @xmath243 while @xmath246 is a fixed experimental measurement irrespective of which theory we are considering . in the following analysis we shall consider model i which has a larger br @xmath247 jets ) than model i@xmath45 ( see table 2 ) . therefore the latter model would suppress @xmath246 more . in the mhdm br ( @xmath248 jets ) may take any value between @xmath67 and @xmath9 . table 3 shows how @xmath243 ( for @xmath249 ) varies with different values of br ( @xmath76 ) . 100% & 19.75 + 99% & 19.55 + 95% & 18.76 + 90% & 17.76 + 62% & 12.24 + ref . [ 1 ] assumes br @xmath201 and @xmath250 is measured to be @xmath251 pb . from eq . ( 18 ) one can see that if br @xmath252 with @xmath253 the cdf measurement would be scaled by the ratio @xmath254 . the lower error bar on @xmath250 currently lies within the region of @xmath255 @xmath256 pb with @xmath88 gev ) , and the inclusion of @xmath17 decays will shift the @xmath250 data point to higher values . if br ( @xmath257 then the lower error bar on @xmath250 lies just outside the maximum value of @xmath255 ( 5.85 pb ) . therefore lower branching ratios than this are more desirable . from figure 1 we see that a considerable parameter space exists for br ( @xmath258 ; in such cases @xmath250 would only be increased by @xmath259 and so a sizeable portion of the lower error bar would still lie within the region of @xmath255 . therefore we conclude that top quark decays to charged higgs scalars are consistent with the current cdf data over a very large parameter space of @xmath47 and @xmath53 . therefore @xmath16 may lie in the energy range of lep2 . the tevatron will continue to operate until the end of 1995 , eventually collecting 140 @xmath260 of data . from eqs . ( 13 ) and ( 14 ) one sees that this twofold increase would only provide a statistically significant signal for a minute part of parameter space . the most favourable choice of parameters ( used previously ) gives @xmath261 . higher luminosity colliders are required and these are studied in the next section . there is a proposal to increase the luminosity of the tevatron by an order of magnitude and it is thought that 2 @xmath262 might be possible by the year 2000 . further proposals are for future luminosities of 20 @xmath262 and/or @xmath263 tev . finally , by the year 2010 the large hadron collider ( lhc ) should be operating . table 4 @xcite summarises the properties of these colliders . @xmath195 @xmath264 & @xmath138 @xmath265 & @xmath266 + & 1.8 & 4.5 & 2 & 2000 + * & 1.8 & 4.5 & 20 & @xmath267 + & 4 & 26 & @xmath268 & @xmath267 + & 14 & 430 & 100 & 2010 + we shall now consider the detection prospects of @xmath16 via top quark decay at each of these colliders . in this section we shall concentrate on studying the mass range @xmath269 gev @xmath270 ; this is because lep2 will search for @xmath271 gev @xcite , @xcite before any of the above colliders will operate . an analysis of the detection prospects of the minimal susy @xmath16 at the upgraded tevatron appears in ref . the higgs models we are studying differ noticeably from the susy @xmath16 with respect to br ( @xmath17 ) . the difference arises because susy requires model ii couplings @xcite and thus br ( @xmath17 ) has a sinusoidal dependence on @xmath47 with a minimum at @xmath272 . in contrast we are studying model i and i@xmath45 type couplings which cause br ( @xmath17 ) to fall with increasing @xmath47 ( figure 1 ) . therefore @xmath16 in these latter models will always be hidden for large enough @xmath47 . starting with an upgraded tevatron providing 2 @xmath262 , we again make use of eqns . ( 8) and ( 10 ) using the same values for the efficiencies as in section 5 . the number of @xmath123 pairs ( @xmath150 ) will now be taken to be @xmath273 although there is still a theoretical error here of @xmath196 . in figures [ fig : top1 ] and [ fig : top ] we plot eq . ( 11 ) which shows the excess ( quantified as a number of standard deviations above that predicted from universality ) of high energy pions caused by @xmath16 decay . figure [ fig : top1 ] ( for model i@xmath45 ) shows that for @xmath274 gev , @xmath275 is maintained until @xmath276 while @xmath179 @xmath277 is managed until @xmath278 . for heavier @xmath53 the signal is obviously weaker ; for @xmath93 gev , @xmath279 ( @xmath280 ) is maintained only until @xmath281 . for model i ( figure [ fig : top ] ) the situation is worse due to the inferior br @xmath282 . in the mhdm , br @xmath282 can take any value from @xmath283 . however the case of br @xmath284 approximates to figure [ fig : top1 ] as @xmath47 increases . br @xmath285 for @xmath286 . ] prospects are improved at the tevatron * ( 20 fb@xmath287 , di tevatron ( 2 fb@xmath287 and di tevatron ( 20 fb@xmath288 ) ; these would scale @xmath179 by 3.2 , 2.4 and 7.6 respectively . the latter collider would maintain a @xmath229 signal in model i@xmath45 ( model i ) for @xmath93 gev if @xmath289 ( 8) . the other detection method available is to reconstruct the invariant masses of the quark decays of @xmath16 . certainly this method will be needed in the case of a leptophobic higgs ( see section 4 ) which is possible in the mhdm . figure [ fig : top5 ] shows that for such a higgs a noticeable peak at @xmath53 can be obtained if @xmath274 gev ( 130 gev ) for @xmath290 ( 2 ) . figure [ fig : top2 ] is the analogous plot for model i and i@xmath45 ; it is clear that in model i reconstructing the @xmath291 jets channel is competitive with the previously considered @xmath172 decay ( if @xmath292 fb@xmath288 ) . a noticeable peak would be present for @xmath274 gev ( 130 gev ) if @xmath293 ( 1.7 ) . in comparison from figure [ fig : top ] we see that the cases of @xmath274 gev , @xmath294 and @xmath93 gev , @xmath295 both give @xmath296 . hence the detection methods are competitive . at the tevatron * and di tevatron the increased @xmath138 and/or @xmath187 enhances @xmath179 and so the @xmath172 channel offers the best prospects for model i. in model i@xmath45 this is the case even when @xmath292 fb@xmath288 . finally we shall consider the lhc . using the same values for the various efficiencies as was used at the tevatron , figure [ fig : top4 ] shows that for @xmath93 gev in model i@xmath45 ( model i ) @xmath297 is maintained until @xmath298 ( 15 ) . however , using similar efficiencies to those used in ref . @xcite ( which are for the ssc ) we find that the above values of @xmath298 and 15 must be replaced by 8 and 5 . a @xmath0tag is included here and @xmath299 is taken , while at the tevatron @xmath300 and no @xmath0tag is required . also , ref . @xcite does not make use of @xmath301 . despite this reduced coverage , it is clear that the lhc will probe a significant region of @xmath47 and @xmath53 parameter space , especially if the efficiencies used in ref . @xcite are improved . we have studied the prospects for detecting charged higgs scalars @xmath302 contributing to top quark decay ( @xmath17 ) in the context of the non minimal standard model . considered were @xmath16s from the 2hdm ( models i and i@xmath45 ) , htm and mhdm , all of which escape the mass bounds from @xmath50 and thus may be light enough to contribute to top quark decay . two detection methods were presented : * searching for an excess of energetic pions ( @xmath303 ) over that predicted from lepton universality . * reconstructing the quark decays of @xmath16 . neither method is able to extract a statistically significant signature of @xmath16 with the current cdf data sample ( 67 pb@xmath288 ) at the tevatron , and even the expected 140 pb@xmath288 ( available by the end of 1995 ) will prove insufficient over most of the parameter ( @xmath53 , @xmath47 ) space . thus the existence of a light @xmath16 ( 50 gev @xmath304 ) is fully consistent with current data , and may be searched for at lep2 . prospects are much better at proposed higher energy / luminosity colliders . at an upgraded tevatron of @xmath292 fb@xmath288 method ( i ) will provide ( in model i@xmath45 for @xmath274 gev ) an energetic pion excess of @xmath229 ( @xmath305 ) if @xmath306 ( 8) . for @xmath93 gev an excess of @xmath229 ( @xmath305 ) is maintained until @xmath232 ( 5 ) . for model i the situation is worse due to the inferior br @xmath282 . a di tevatron with @xmath263 tev and @xmath307 fb@xmath288 would provide a @xmath229 signal for @xmath308 ( 8) in model i@xmath45 ( model i ) for @xmath93 gev . the lhc would improve the coverage to 25 and 15 respectively . method ( ii ) is necessary for the case of a leptophobic higgs with br ( @xmath309 which is possible in the mhdm . a noticeable peak at @xmath53 can be obtained for @xmath274 gev ( 130 gev ) if @xmath290 ( 2 ) . for larger @xmath47 the detection of a leptophobic @xmath16 appears unlikely in top quark decay and other production methods must be considered . for model i method ( ii ) is competitive with method ( i ) at the tevatron ( @xmath292 fb@xmath288 ) , but in all other cases method ( i ) offers the best detection prospects . i wish to thank w.j . stirling and d.p . roy for useful comments . this work has been supported by the uk epsrc research council .	charged higgs scalars light enough to contribute to top quark decays are possible in various non minimal higgs models . we show that such a decay would be consistent with the current tevatron data , and will remain hidden until a larger luminosity can be achieved . dtp/95/70 + august 1995 + 1.cm * hidden top quark decay to charged higgs scalars at the tevatron + [ 3 mm ] * 1.cm a.g . akeroyd .4 cm _ department of mathematical sciences , centre for particle theory , + university of durham , + durham dh1 3le , england . _ +
it is a basic problem in @xmath3-dimensional topology to classify smooth structures on @xmath3-manifolds . constructions of exotic smooth structures on @xmath3-manifolds with small euler characteristics are currently in rapid progress ( see , for example , park @xcite , stipsicz - szab @xcite , fintushel - stern @xcite , park - stipsicz - szab @xcite and akhmedov - park @xcite ) . however , it is still unknown whether or not @xmath4 and @xmath5 admit an exotic smooth structure . if such a structure exists , then each handle decomposition of it has at least either a @xmath1- or @xmath2-handle ( see proposition [ s^4 ] ) . to the contrary , many classical simply connected closed smooth @xmath3-manifolds are known to admit neither @xmath1- nor @xmath2-handles in their handle decompositions ( cf . gompf - stipsicz @xcite ) . problem 4.18 in kirby s problem list @xcite is the following : `` does every simply connected , closed @xmath3-manifold have a handlebody decomposition without @xmath1-handles ? without @xmath1- and @xmath2-handles ? '' the elliptic surfaces @xmath6 are candidates of counterexamples to problem 4.18 . it is not known whether or not the simply connected closed smooth @xmath3-manifold @xmath7 admits a handle decomposition without @xmath1-handles ( cf . gompf @xcite and gompf - stipsicz @xcite ) . in particular , harer , kas and kirby have conjectured in @xcite that every handle decomposition of @xmath0 requires at least a @xmath1-handle . note that by considering dual handle decompositions , their conjecture is equivalent to the assertion that @xmath0 requires both @xmath1- and @xmath2-handles . in this article we construct the following smooth @xmath3-manifolds by using rational blow - downs and kirby calculus . for @xmath8 , there exists a smooth @xmath3-manifold @xmath9 with the following properties : + @xmath9 is homeomorphic to @xmath10 ; + @xmath9 has the same seiberg - witten invariant as @xmath10 ; + @xmath9 admits a handle decomposition without @xmath1-handles , namely , @xmath11 there exists a smooth @xmath3-manifold @xmath12 with the following properties : + @xmath12 is homeomorphic to @xmath0 ; + @xmath12 has the same seiberg - witten invariant as @xmath0 ; + @xmath12 admits a handle decomposition without @xmath1- and @xmath2-handles , namely , @xmath13 as far as the author knows , @xmath9 and @xmath12 are the first examples in the following sense : if @xmath9 ( resp . @xmath12 ) is diffeomorphic to @xmath10 ( resp . @xmath0 ) , then the above handle decomposition of @xmath10 ( @xmath14 [ resp . @xmath12 ] ) is the first example which has no @xmath1-handles . otherwise , i.e. , if @xmath9 ( resp . @xmath12 ) is not diffeomorphic to @xmath10 ( resp . @xmath0 ) , then @xmath9 ( resp . @xmath12 ) and @xmath10 ( resp . @xmath0 ) are the first homeomorphic but non - diffeomorphic examples which are simply connected closed smooth @xmath3-manifolds with the same non - vanishing seiberg - witten invariants . an affirmative solution to the harer - kas - kirby conjecture implies that both @xmath16 and @xmath12 are not diffeomorphic to @xmath0 , though these three have the same seiberg - witten invariants . in this case , the minimal number of @xmath1-handles in handle decompositions does detect the difference of their smooth structures . our construction is inspired by rational blow - down constructions of exotic smooth structures on @xmath17 by park @xcite , stipsicz - szab @xcite , fintushel - stern @xcite and park - stipsicz - szab @xcite . our method is different from theirs since , firstly , we use kirby calculus to perform rational blow - downs , whereas they used elliptic fibrations on @xmath18 ( and knot surgeries ) , secondly , they did not examine handle decompositions . the author wishes to express his deeply gratitude to his adviser , professor hisaaki endo , for encouragement and many useful suggestions . he would like to thank professors selman akbulut , kazunori kikuchi , ronald j. stern and yuichi yamada for helpful comments and discussions . kikuchi s theorem ( * ? ? ? * theorem 4 ) partially gave him the idea of the construction . yamada gave him interesting questions ( cf . remark [ yamada ] ) . in this section we review the rational blow - down introduced by fintushel - stern @xcite . for details , see also gompf - stipsicz @xcite . let @xmath19 and @xmath20 be the smooth @xmath3-manifolds defined by kirby diagrams in figure [ c_p ] , and @xmath21 elements of @xmath22 given by corresponding @xmath23-handles in the figure such that @xmath24 @xmath25 . the boundary @xmath26 of @xmath19 is diffeomorphic to the lens space @xmath27 and to the boundary @xmath28 of @xmath20 . the following lemma is well known . @xmath29 @xmath30 , @xmath31 and @xmath32 . + @xmath33 @xmath34 and @xmath35 suppose that @xmath19 embeds in a smooth @xmath3-manifold @xmath36 . the smooth @xmath3-manifold @xmath37 is called the rational blow - down of @xmath36 along @xmath19 . note that @xmath38 is uniquely determined up to diffeomorphism by a fixed pair @xmath39 . this operation preserves @xmath40 , decreases @xmath41 , may create torsions in the first homology group , and has the following relation with the logarithmic transformation . [ th:2.1 ] suppose that a smooth @xmath3-manifold @xmath36 contains a cusp neighborhood , that is , a @xmath42-handle with a @xmath23-handle attached along a @xmath42-framed right trefoil knot . let @xmath43 be the smooth @xmath3-manifold obtained from @xmath36 by performing a logarithmic transformation of multiplicity @xmath44 in the cusp neighborhood . then there exists a copy of @xmath19 in @xmath45 such that the rational blow - down of @xmath45 along the copy of @xmath19 is diffeomorphic to @xmath46 . let @xmath47 be the simply connected elliptic surface with euler characteristic @xmath48 and with no multiple fibers , and @xmath49 the elliptic surface obtained from @xmath47 by performing logarithmic transformations of multiplicities @xmath50 . we denote @xmath51 as a canonical orthogonal basis of @xmath52 such that @xmath53 and @xmath54 . since there is a diffeomorphism @xmath55 which maps the class of a regular fiber of @xmath56 to @xmath57 ( cf . etg - park @xcite , gompf - stipsicz @xcite ) , theorem [ th:2.1 ] gives us the following corollary . [ cor:2.2 ] for each natural number @xmath44 and @xmath58 , the elliptic surface @xmath59 is obtained from @xmath60 by rationally blowing down along a certain copy @xmath61 of @xmath62 such that @xmath63 satisfy @xmath64 as elements of @xmath65 . @xmath59 is homeomorphic but non - diffeomorphic to @xmath18 , in the case @xmath66 and @xmath67 ( cf . gompf - stipsicz @xcite ) . in this section we construct @xmath16 , @xmath68 and @xmath69 , and prove theorem 1.1.(1)(a)(c ) and ( 2)(a)(c ) . in kirby diagrams , we write the second homology classes given by @xmath23-handles , instead of usual framings . note that the square of the homology class given by a 2-handle is equal to the usual framing . we do not draw ( whole ) kirby diagrams of @xmath70 and the other manifolds appeared in the following construction . however , one can easily draw whole diagrams . we begin with a construction of a cusp neighborhood in @xmath71 such that its embedding into @xmath71 has the same homological properties as that of the regular neighborhood of a cusp fiber of @xmath72 . we do not know if these embeddings into @xmath71 are the same up to diffeomorphism . [ lem:3.1 ] @xmath71 admits the handle decomposition drawn in figure [ cusp ] . here @xmath73 denotes @xmath74 . , width=192 ] we firstly create two @xmath23-handles with framings @xmath75 and @xmath76 in a kirby diagram of @xmath5 . figure [ 4.1 ] is a basic kirby diagram of @xmath5 . introducing a @xmath23-handle/@xmath2-handle pair gives figure [ 4.2 ] . handle slides and isotopies yield figure [ 4.5](pairs of bold lines in figures denote ` bands ' ) : @xmath77 creating a @xmath23-handle/@xmath2-handle pair gives figure [ 4.6 ] . handle slides produce figure [ 4.10 ] : @xmath78 we secondly blow up @xmath5 nine times : @xmath79 we lastly make a handle addition @xmath80 . this leads to figure [ 4.14 ] , and an isotopy gives figure [ cusp ] . [ prop:3.2 ] @xmath81 admits the handle decomposition drawn in figure [ 5.1 ] . in particular @xmath81 contains the copy of @xmath82 drawn in the figure . the elements @xmath83 given by this copy of @xmath82 are the same as that given by @xmath84 . ] + @xmath85 admits the handle decomposition drawn in figure [ 6.1 ] . in particular @xmath85 contains the copy of @xmath86 drawn in the figure . the elements @xmath87 given by this copy of @xmath86 are the same as that given by @xmath88 . ] firstly we give a proof for ( 1 ) . blowing up in figure [ cusp ] yields figure [ 5.2 ] . the handle slide drawn in figure [ ex - handleslide ] gives figure [ 5.3 ] . an additional blow - up yields figure [ 5.4 ] , and an isotopy gives figure [ 5.1 ] . secondly we give a proof for ( 2 ) . handle slides , isotopies and blow - ups in figure [ 5.4 ] yield figure [ 6.1 ] : @xmath89 [ prop:3.3 ] @xmath85 admits the handle decomposition drawn in figure [ 8.1 ] . in particular @xmath85 contains the copy of @xmath86 drawn in the figure . ] recall the construction in the proof of lemma [ lem:3.1 ] . in this construction , we created a @xmath23-handle/@xmath2-handle pair twice . instead of introducing a @xmath23-handle/@xmath2-handle pair twice , blowing up twice yields figure [ 8.7 ] : @xmath90 handle slides and blow - ups as in proofs of lemma [ lem:3.1 ] and proposition [ prop:3.2 ] gives figure [ 8.8 ] . repeating handle slides drawn in figure [ ex - handleslide ] yields figure [ 8.9 ] . an additional blow - up gives figure [ 8.10 ] , and an isotopy gives figure [ 8.1 ] . let @xmath9 be the smooth @xmath3-manifold obtained from @xmath60 by rationally blowing down along the copy of @xmath62 obtained in proposition [ prop:3.2 ] , for @xmath8 . let @xmath12 be the smooth @xmath3-manifold obtained from @xmath85 by rationally blowing down along the copy of @xmath86 obtained in proposition [ prop:3.3 ] . it is not known whether or not there exists a copy of @xmath86 in @xmath85 such that the rational blow - down is diffeomorphic to @xmath0 . in @xcite we will construct more examples of exotic @xmath71 without @xmath1- and @xmath2-handles , by improving the construction of @xmath69 . the author does not know if these examples have the same seiberg - witten invariants as the elliptic surfaces @xmath59 . we prepare the following lemma . [ without - handle]suppose that a simply connected closed smooth @xmath3-manifold @xmath36 has the handle decomposition drawn in figure [ without ] . here @xmath91 is an arbitrary integer , @xmath92 and @xmath93 are arbitrary natural numbers . note that we write usual framings instead of homology classes in the figure . let @xmath38 be the rational blow - down of @xmath36 along the copy of @xmath19 drawn in figure [ without ] . then @xmath38 admits a handle decomposition @xmath94 in particular @xmath38 admits a handle decomposition without @xmath1-handles . ] draw a kirby diagram of @xmath38 , following the procedure introduced in ( * ? ? ? * section 8.5 ) ( see also ( * ? ? ? * solution of exercise 8.5.1.(a ) ) ) . then the @xmath91-framed unknot drawn in figure [ without ] changes into a meridian of a unique dotted circle which naturally appears in this procedure . thus we can cancel the @xmath1-handle/@xmath23-handle pair . note that this procedure does not produce new @xmath2-handles . the following proposition gives theorem 1.1.(1)(a)(c ) and ( 2)(a)(c ) . for @xmath8 , the manifold @xmath9 is homeomorphic to @xmath10 and admits a handle decomposition without @xmath1-handles , namely,@xmath11 @xmath69 is homeomorphic to @xmath0 and admits a handle decomposition without @xmath1- and @xmath2-handles , namely , @xmath95 lemma [ without - handle ] shows the above properties of @xmath9 and @xmath69 about handle decompositions . thus @xmath9 and @xmath69 are simply connected . since @xmath9 is obtained from @xmath60 by rationally blowing down along a copy of @xmath62 , we have @xmath96 similarly we have @xmath97 and @xmath98 . therefore freedman s theorem together with rochlin s theorem shows that @xmath9 and @xmath69 are homeomorphic to @xmath71 . thus @xmath9 and @xmath69 are homeomorphic to @xmath10 . in this section , we briefly review facts about the seiberg - witten invariants with @xmath99 . for details and examples of computations , see fintushel - stern @xcite , @xcite , @xcite , stern @xcite , park @xcite , @xcite , ozsvth - szab @xcite , stipsicz - szab @xcite and park - stipsicz - szab @xcite . suppose that @xmath36 is a simply connected closed smooth @xmath3-manifold with @xmath100 . let @xmath101 be the set of characteristic elements of @xmath102 . fix a homology orientation on @xmath36 , that is , orient @xmath103 . then the ( small - perturbation ) seiberg - witten invariant @xmath104 is defined for every positively oriented element @xmath105 and every element @xmath106 such that @xmath107 . let @xmath108 and @xmath109 be the euler characteristic and the signature of @xmath36 , respectively , and @xmath110 the even integer defined by @xmath111 for @xmath106 . it is known that if @xmath112 for some @xmath105 , then @xmath113 . the wall - crossing formula tells us the dependence of @xmath114 on @xmath115 : if @xmath116 and @xmath106 satisfy @xmath117 and @xmath113 , then @xmath118 note that these facts imply that @xmath114 is independent of @xmath115 in the case @xmath119 , in other words , the seiberg - witten invariant @xmath120 is well - defined . we recall the change of the seiberg - witten invariants by rationally blowing down . assume that @xmath36 contains a copy of @xmath19 . let @xmath38 be the rational blow - down of @xmath36 along the copy of @xmath19 . suppose that @xmath38 is simply connected . the following theorems are known . [ thm:4.1 ] for every element @xmath121 , there exists an element @xmath122 such that @xmath123 and @xmath124 . we call such an element @xmath122 a of @xmath125 . [ nthm:4.1 ] if an element @xmath122 is a lift of some element @xmath121 , then @xmath126 for every positively oriented element @xmath105 which is orthogonal to the subspace @xmath127 of @xmath128 . note that we view @xmath115 as a positively oriented element of @xmath129 . [ thm:4.2 ] if an element @xmath122 satisfies that @xmath130 and @xmath131 with @xmath132 , then there exists an element @xmath121 such that @xmath133 is a lift of @xmath125 . [ cor:4.3 ] if an element @xmath122 satisfies @xmath134 and @xmath135 , then @xmath133 is a lift of some element @xmath121 . in this section we complete the proof of theorem 1.1 . we prepare the following lemma here . [ lem:5.1 ] let @xmath36 be a simply connected closed smooth @xmath3-manifold which contains a copy of @xmath19 , and @xmath136 the inclusion @xmath137 . let @xmath138 be the orthogonal complement of the subspace spanned by @xmath139 , that is , @xmath140 suppose that there exists an element @xmath141 such that @xmath142 and @xmath143 . then @xmath144 firstly we give a proof for ( 1 ) . since every element of @xmath145 is represented by a surface , it is clear that @xmath146 . let @xmath147 be the inclusion @xmath148 . mayer - vietoris exact sequence of @xmath149 is as follows : @xmath150 since @xmath19 is negative definite and @xmath151 , we have @xmath152 . we determine @xmath153 here . there clearly exists an element @xmath154 such that @xmath155 . the above exact sequence ensures the existence of elements @xmath156 and @xmath157 such that @xmath158 . the element @xmath159 satisfies @xmath160 @xmath161 and @xmath162 @xmath163 . since @xmath164 is a basis of @xmath165 , we can easily see @xmath166 by using the intersection form of @xmath19 . hence @xmath153 is a generator of @xmath167 . suppose that some element @xmath168 satisfies @xmath169 . since @xmath153 is a generator of @xmath167 , there exists an element @xmath170 with @xmath171 such that @xmath172 . applying the above argument about @xmath173 to @xmath174 , we get @xmath175 . this is a contradiction . thus we obtain @xmath176 . therefore @xmath177 . thus it is easy to see @xmath178 . secondly we give a proof for ( 2 ) . since the above @xmath179 is onto , we can easily show by using mayer - vietoris exact sequence . \(1 ) since @xmath180 is injective , the above lemma allows us to identify @xmath181 with @xmath138 . + ( 2 ) under the same assumption as that in lemma [ lem:5.1 ] , we can also show @xmath182 . here @xmath38 denotes the rational blow - down of @xmath36 along the copy of @xmath19 . it is not known whether or not the fundamental groups of @xmath183 and @xmath38 vanish . the following proposition gives us theorem 1.1.(1)(b ) . in the rest of this section , we denote the symbol @xmath184 as @xmath185 . [ prop:5.2 ] @xmath9 has the same seiberg - witten invariant as @xmath10 , that is , there exists a homeomorphism between @xmath9 and @xmath10 which preserves the orientations , the homology orientations and the seiberg - witten invariants , for @xmath8 . we give a proof for @xmath186 , firstly . let @xmath187 be the elements defined by @xmath188 we can view @xmath189 as elements of @xmath190 by lemma [ lem:5.1].(1 ) , corollary [ cor:2.2 ] and the following natural identification : @xmath191 this identification preserves cup products . therefore the elements @xmath192 of @xmath190 satisfy @xmath193 recall that the intersection form of @xmath0 is @xmath194 ( this notation of the intersection form is the same as that in gompf - stipsicz ( * ? ? ? * section 1.2 ) . ) . this implies that either the matrix @xmath195 or @xmath196 represents the symmetric bilinear form on @xmath197 . we here denote the symbol @xmath197 as the orthogonal complement of the subspace spanned by @xmath198 . since @xmath199 and @xmath200 are elements of @xmath197 , it is easy to check that the matrix @xmath195 represents the symmetric bilinear form on @xmath197 . we can easily see that there exists an element @xmath201 such that @xmath202 , by using a basis of @xmath197 . note that @xmath203 is a basis of @xmath190 . proposition 3.2.(1 ) allows us to apply the above argument to @xmath16 . thus we get a basis @xmath204 of @xmath205 which is corresponding to the basis @xmath206 of @xmath190 . let @xmath207 be the isomorphism defined by @xmath208 @xmath209 . here @xmath210 denotes poincar dual . the isomorphism @xmath211 preserves the intersection forms and the homology orientations of @xmath0 and @xmath16 . proposition [ thm:4.1 ] gives us a lift @xmath212 of @xmath125 for every @xmath213 . lemma [ lem:5.1].(2 ) together with the universal coefficient theorem implies that @xmath214 and @xmath215 are uniquely determined by their values on @xmath216 . since @xmath217 @xmath218 and @xmath219 as elements of @xmath220 , it is easy to check that @xmath133 is also a lift of the element @xmath221 . thus theorem [ nthm:4.1 ] shows @xmath222 . hence the isomorphism @xmath211 preserves the seiberg - witten invariants of @xmath0 and @xmath16 . freedman s theorem gives us a required homeomorphism @xmath223 which preserves the orientations and satisfies @xmath224 . we briefly give a proof for @xmath225 , secondly . let @xmath226 be the elements defined by @xmath227 applying the above argument to @xmath68 , we obtain a proof . to prove theorem 1.1.(2)(b ) , we compute the seiberg - witten invariant of @xmath0 . [ lem:5.3 ] let @xmath228 be the element defined by @xmath229 . here @xmath230 denote the elements of @xmath190 defined in the proof of proposition [ prop:5.2 ] . then @xmath231 satisfies @xmath232 and is the unique element of @xmath233 up to sign for which @xmath234 is nonzero . let @xmath235 and @xmath236 be the elements defined by @xmath237 and @xmath238 . note that @xmath115 is orthogonal to the subspace @xmath239 of @xmath240 . it is well known that @xmath241 for every @xmath242 and every @xmath243 . applying the wall - crossing formula to @xmath244 and @xmath245 , we have @xmath246 . corollary [ cor:4.3 ] shows that @xmath247 is a lift of some element @xmath228 . thus theorem [ nthm:4.1 ] gives @xmath232 . since @xmath247 is a lift of @xmath231 , the element @xmath231 satisfies @xmath248 @xmath218 and @xmath249 . hence the values of @xmath231 are as follows : @xmath250 , @xmath251 and @xmath252 . therefore we get @xmath229 . suppose that an element @xmath253 satisfies @xmath254 . proposition [ thm:4.1 ] ensures the existence of a lift @xmath255 of @xmath256 such that @xmath257 . we put @xmath258 . since @xmath256 is characteristic and @xmath259 , the integer @xmath260 is odd and @xmath261 . in the case @xmath262 , cauchy - schwartz inequality ( @xmath263 for @xmath264 ) and @xmath265 show @xmath266 since @xmath254 and @xmath262 , the wall - crossing formula shows @xmath267 . therefore we get @xmath268 . this together with @xmath267 shows @xmath269 @xmath270 . we thus have @xmath271 . similarly we have @xmath272 in the case @xmath273 . hence @xmath274 . the following proposition completes the proof of theorem 1.1 . [ prop:5.4 ] @xmath69 has the same seiberg - witten invariant as @xmath0 , that is , there exists a homeomorphism between @xmath69 and @xmath0 which preserves the orientations , the homology orientations and the seiberg - witten invariants . let @xmath275 and @xmath276 be the elements defined by @xmath277 and @xmath278 . note that @xmath279 is orthogonal to the subspace @xmath280 of @xmath281 . applying the wall - crossing formula to @xmath282 and @xmath245 , we get @xmath283 . corollary [ cor:4.3 ] shows that @xmath284 is a lift of some element @xmath285 . thus theorem [ nthm:4.1 ] gives @xmath286 . the same argument as that in the proof of lemma [ lem:5.3 ] shows that @xmath287 is the unique element up to sign for which @xmath288 is nonzero . let @xmath289 be the element defined by @xmath290 . lemma [ lem:5.1].(1 ) allows us to view @xmath291 as an element of @xmath292 . we set @xmath293 by @xmath294 . the element @xmath295 is a characteristic element of @xmath296 and satisfies @xmath297 and @xmath298 , because @xmath299 , @xmath300 and @xmath301 . we here denote the symbol @xmath296 as the orthogonal complement of the subspace spanned by @xmath302 . since the symmetric bilinear form on @xmath296 is @xmath303 , the following lemma together with the above property of @xmath295 gives us an orthogonal basis @xmath304 of @xmath305 such that @xmath306 , @xmath307 and @xmath308 . [ lem:5.5 ] let @xmath309 be a free @xmath310-module equipped with a symmetric bilinear form @xmath303 . if a characteristic element @xmath125 of @xmath309 satisfies @xmath311 , then there exists an automorphism of @xmath309 which preserves the symmetric bilinear form on @xmath309 and maps @xmath125 to @xmath312 . here @xmath313 denotes an arbitrary orthogonal basis of @xmath309 such that @xmath306 and @xmath314 . similarly the above lemma together with lemma [ lem:5.3 ] gives us an orthogonal basis @xmath315 of @xmath316 such that @xmath317 , @xmath318 and @xmath319 . let @xmath320 be the isomorphism defined by @xmath321 . the isomorphism @xmath322 preserves the intersection forms and the seiberg - witten invariants . let @xmath323 be the element defined in the proof of lemma [ lem:5.3 ] . recall that we can view @xmath115 and @xmath279 as positively oriented elements of @xmath324 and @xmath325 , respectively . note that @xmath326 and @xmath327 we thus have @xmath328 and @xmath329 . these two inequalities together with the lemma below show @xmath330 . hence @xmath331 preserves the homology orientations . freedman s theorem gives us a required homeomorphism @xmath332 which preserves the orientations and satisfies @xmath333 . [ lem:5.6 ] let @xmath334 be a vector space of rank @xmath91 over @xmath335 equipped with a symmetric bilinear form such that @xmath336 and @xmath337 . here @xmath40 and @xmath41 are the same notation as that in gompf - stipsicz @xcite . if elements @xmath338 satisfy @xmath339 , @xmath340 , @xmath341 , @xmath342 and @xmath343 , then @xmath344 . let @xmath345 be the subspace spanned by @xmath159 , and @xmath346 the orthogonal complement of @xmath345 . the subspace @xmath346 is negative definite , because @xmath336 and @xmath339 . since @xmath347 , there exist elements @xmath348 , @xmath349 such that @xmath350 and @xmath351 . cauchy - schwartz inequality implies @xmath352 . inequalities @xmath340 and @xmath341 give us @xmath353 and @xmath354 . these three inequalities together with @xmath343 show @xmath355 . this inequality and @xmath342 give us @xmath356 . hence @xmath344 . this completes the proof of proposition [ prop:5.4 ] . we conclude this article by making some remarks . in figure [ 4.8]@xmath357[4.10 ] , we used the peculiar bands , that is , bands not in local positions to prove lemma [ lem:3.1 ] . note that standard bands , that is , bands in local positions are also enough to prove lemma [ lem:3.1 ] . however , the peculiar bands are the key of our construction of exotic @xmath358 ( see @xcite ) . in the proof of lemma [ lem:3.1 ] , we used two @xmath23-handles with framings @xmath359 . instead of these two @xmath23-handles , we can use two @xmath23-handles with framings both @xmath360 and @xmath361 to prove lemma [ lem:3.1 ] . we can also use a @xmath23-handle with a framing @xmath362 to construct figure [ further - construction ] . in this construction , we can decrease the number of @xmath2-handles of @xmath16 . precisely @xmath16 admits a handle decomposition @xmath363 [ yamada ] yamada asked the author if a topologically trivial but smoothly non - trivial _ h_-cobordism between @xmath9 and @xmath10 exists . following the argument in gompf - stipsicz ( * ? ? ? * example 9.2.15 ) , we can prove that such an _ h_-cobordism exists . note that the same argument also shows that a topologically trivial but smoothly non - trivial _ h_-cobordism between @xmath10 and itself exists . let @xmath36 be a simply connected closed smooth @xmath3-manifold which contains a copy of @xmath19 , and @xmath38 the rational blow - down of @xmath36 along the copy of @xmath19 . suppose that @xmath38 is simply connected . do the following two conditions , @xmath36 and the homomorphism @xmath364 induced by the copy of @xmath19 , suffice to determine the ( small perturbation ) seiberg - witten invariant of @xmath38 ? [ s^4 ] if a smooth @xmath3-manifold is homeomorphic to @xmath4 ( resp . @xmath5 ) and admits neither @xmath1- nor @xmath2-handles in a handle decomposition , then the @xmath3-manifold is diffeomorphic to @xmath4 ( resp . @xmath5 ) . note that if a simply connected closed smooth @xmath3-manifold has neither @xmath1- nor @xmath2-handles in a handle decomposition , then the number of @xmath23-handles appeared in the handle decomposition is equal to the rank of the second homology group of the @xmath3-manifold . suppose that a smooth @xmath3-manifold is homeomorphic to @xmath4 and has neither @xmath1- nor @xmath2-handles in a handle decomposition . then this handle body consists of a @xmath42-handle and a @xmath3-handle . since attaching a @xmath3-handle is unique ( see gompf - stipsicz @xcite ) , the @xmath3-manifold is diffeomorphic to @xmath4 . suppose that a smooth @xmath3-manifold is homeomorphic to @xmath5 and has neither @xmath1- nor @xmath2-handles in a handle decomposition . then this handle body consists of a @xmath42-handle , a @xmath23-handle and a @xmath3-handle . thus the attaching circle of the @xmath23-handle produces @xmath365 by a dehn surgery with coefficient @xmath366 . since such a knot is unknot ( see gordon - luecke @xcite ) , the @xmath3-manifold is diffeomorphic to @xmath5 . contrary to the above proposition , many simply connected closed topological @xmath3-manifolds are known to admit at least two different smooth structures without @xmath1- and @xmath2-handles ( cf . gompf - stipsicz @xcite ) . as far as the author knows , @xmath4 and @xmath5 are the only known exceptions . thus the following problem is natural . this article is based on the author s announcement @xcite . in @xcite , we will give the rest of examples announced in @xcite . in addition to these examples , we will construct a smooth @xmath3-manifold which has the same seiberg - witten invariant as @xmath0 and admits no @xmath1-handles as follows : we ` naturally ' construct figure [ further - construction ] and perform a logarithmic transformation of multiplicity @xmath2 in the cusp neighborhood . in @xcite , we will construct examples of exotic @xmath358 by using rational blow - downs and kirby calculus . we also prove that our examples admit a handle decomposition without @xmath1- and @xmath2-handles in the case @xmath367 . 25 a. akhmedov and b.d . exotic smooth structures on small 4-manifolds _ , arxiv : math/0701664 . t. etg and b.d . symplectic tori in rational elliptic surfaces _ , math . ann . * 334 * ( 2006 ) , no . 3 , 679691 . r. fintushel and r.j . stern , _ rational blowdowns of smooth @xmath3manifolds _ , j. differential geom . * 46 * ( 1997 ) , no . 2 , 181235 . r. fintushel and r.j . double node neighborhoods and families of simply connected @xmath3manifolds with @xmath368 _ , j. amer . math . * 19 * ( 2006 ) , no . 1 , 171180 . r. fintushel and r.j . stern , _ six lectures on four 4-manifolds _ , arxiv : math/0610700 . gompf , _ nuclei of elliptic surfaces _ , topology * 30 * ( 1991 ) , no . 3 , 479511 . gompf and a.i . stipsicz , _ @xmath3-manifolds and kirby calculus _ , graduate studies in mathematics , * 20. american mathematical society , 1999 . gordon and j. luecke , _ knots are determined by their complements _ , j. amer . 2 * ( 1989 ) , no . 2 , 371415 . j. harer , a. kas and r. kirby , _ handlebody decompositions of complex surfaces _ , mem . * 62 * ( 1986 ) , no . 350 . k. kikuchi , _ positive @xmath23-spheres in @xmath3-manifolds of signature @xmath369 _ , pacific j. math . * 160 * ( 1993 ) , no . 2 , 245258 . r. kirby , _ problems in low - dimensional topology _ , in geometric topology ( w. kazez ed . ) , ams / ip stud . adv math . vol . 2.2 , amer . 1997 , 35473 . p. ozsvth and z. szab , _ on park s exotic smooth four - manifolds _ , geometry and topology of manifolds , fields inst . 47 , amer . providence , ri , 2005 , 253260 , j. park , _ seiberg - witten invariants of generalised rational blow - downs _ , bull . austral . * 56 * ( 1997 ) , no . 3 , 363384 . j. park , _ simply connected symplectic 4-manifolds with @xmath370 and @xmath371 _ , invent . math . * 159 * ( 2005 ) , no . 3 , 657667 . j. park , a.i . stipsicz and z. szab , _ exotic smooth structures on @xmath372 _ , math . * 12 * ( 2005 ) , no . 5 - 6 , 701712 . stern , _ will we ever classify simply - connected smooth 4-manifilds ? _ , floer homology , gauge theory , and low - dimensional topology , ( d. ellwood , _ et . eds . ) , cmi / ams publication , 2006 , 225239 . stipsicz and z. szab , _ an exotic smooth structure on @xmath373 _ , geom . topol . * 9 * ( 2005 ) , 813832 . wall , _ on the orthogonal groups of unimodular quadratic forms . , j. reine angew . math . * 213 * ( 1963/1964 ) , 122136 . k. yasui , _ an exotic rational elliptic surfaces without @xmath1- or @xmath2-handles _ , to appear in the proceedings of the international conference , intelligence of low dimensional topology 2006 " , world scientific publishing co. in the knots and everything book series .	harer , kas and kirby have conjectured that every handle decomposition of the elliptic surface @xmath0 requires both @xmath1- and @xmath2-handles . in this article , we construct a smooth @xmath3-manifold which has the same seiberg - witten invariant as @xmath0 and admits neither @xmath1- nor @xmath2-handles , by using rational blow - downs and kirby calculus . our manifold gives the first example of either a counterexample to the harer - kas - kirby conjecture or a homeomorphic but non - diffeomorphic pair of simply connected closed smooth @xmath3-manifolds with the same non - vanishing seiberg - witten invariants .
detecting the scintillation light from liquid noble elements ( `` noble liquids '' ) is an important component of several large experimental programs engaged in efforts to directly detect dark matter@xcite as well as neutrinos from both the sun@xcite and accelerator beam - lines@xcite . liquid noble elements have a number of characteristics that make their use as scintillators very attractive . they have good stopping power compared to other liquid scintillators , with densities between 1.2 and 3.1 g/@xmath1@xcite . their scintillation yield is also very high , typically tens of photons per kev@xcite ( similar to that of nai@xcite ) . the time structure of scintillation light from the noble elements also offers excellent particle identification capability , a property that will be discussed in section [ sec : nescint ] . lastly , noble liquids are essentially transparent to their own scintillation light , eliminating the need to dissolve fluors directly into the scintillation volume , as is the case with most organic liquid scintillators@xcite . the primary difficulty in using noble elements as a scintillator comes not from the cryogenics required to keep them in a liquid state , but from the challenges of efficiently detecting the short wavelength scintillation light . noble elements scintillate in the extreme ultraviolet ( euv ) wavelength range . most current interest in the community involves xenon and argon , which have scintillation emission spectra ranging from roughly 120 to 200 nm . as such , the bulk of the work presented in this article will focus on this wavelength band . there is also interest in using neon@xcite and helium@xcite as scintillators , both of which emit scintillation light at much shorter wavelengths , from 60 to 100 nm . one other difficulty associated with the use of noble elements as scintillators is purification . very small quantities of nitrogen , oxygen or water can strongly quench scintillation light@xcite . purification is largely a solved problem in terms of methodology . implementation of the necessary purification regime can however be somewhat technically challenging and should not be ignored . the detection of euv photons is a very challenging problem . their wavelength is short enough that they are strongly absorbed by nearly all materials used for visible optics , such as the quartz or glass windows of photomultiplier tubes , but they are not energetic enough to be treated calorimetrically like x rays or @xmath2 rays . this obstacle can be side - stepped by down - scattering euv photons into the visible wavelength band with a fluor . one very common technique is to coat a surface in contact with the noble element with a thin film of the organic compound tetra - phenyl butadiene ( tpb ) . this article will explore the efficiency with which a tpb film converts euv to visible photons as well as the spectrum of the re - emitted photons as a function of the wavelength of the absorbed euv photons . we will begin section [ sec : thepast ] with a discussion of noble element scintillation . we will present a summary of previous measurements involving tpb films , including both the fluorescence efficiency and the visible re - emission spectrum . we will also discuss how previous noble element scintillation detectors avoided the problem of the large uncertainty on both of these quantities . section [ sec : hardware ] will discuss the experimental apparatus used for this study , with special attention paid to the improvements made over previous measurements . section [ sec : visspec ] will detail our measurements of the tpb visible re - emission spectrum as a function of input wavelength , and section [ sec : fleff ] will present our fluorescence efficiency measurements . last , we will draw some conclusions from this work and discuss our plans for future measurements in section [ sec : concl ] . as discussed above , noble elements are transparent to their own scintillation light , unlike organic scintillators . this is due to the somewhat unusual way in which scintillation occurs in noble elements . this is discussed in detail in @xcite . rather than simply exciting an atomic or molecular state which then emits light as it relaxes to the ground state , scintillation in noble elements is actually produced by the de - excitation of dimer molecules formed from combinations of excited or ionized atoms and a ground state atom . this scintillation light has two components arising from two different states : a short - lived , singlet state , and a long - lived , triplet state . the lifetime difference comes from whether the noble atom in question is ionized or simply excited when it forms the dimer molecule . dimers formed from an excited atom can decay very quickly with a typical lifetime of @xmath3 ns . ionized dimers , on the other hand , must overcome a spin flip before the dimer state can decay . this dramatically increases the lifetime of the dimer to microseconds or longer . because the scintillation light corresponds to a transition in a temporary dimer state , there is no corresponding transition in the surrounding monoatomic medium to absorb photons once produced . as a result , these scintillation photons have a very long path length in the bulk liquid , making noble elements essentially transparent at their scintillation wavelengths . figure [ fig : scintspec ] shows the scintillation spectra of all five noble elements along with the percent transmittance of several common optical window materials . the data presented in figure [ fig : scintspec ] are reproduced from references @xcite . each spectrum has been normalized to unit area to facilitate comparison of their shapes . several measurements of the visible re - emission spectrum of tpb films have been made@xcite . however , the tpb was excited with uv sources at 253.7 nm @xcite and 185 nm @xcite , which are both longer wavelength than the scintillation spectrum of argon , neon , and helium . furthermore , the spectra measured in these references are quite different , leading to the possibility of a significant dependence on the input euv wavelength or the details of the tpb deposition method . rather than use one of these existing measurements of the re - emission spectrum as an input to the fluorescence efficiency measurement , we directly measured the re - emission spectrum of our tpb samples using several excitation wavelengths from 128 nm to 250 nm . there have also been several measurements of the fluorescence efficiency of tpb films . reference @xcite measured this efficiency relative to sodium salicylate , for several tpb coating thicknesses , illuminating the films with euv light from about 100 to above 300 nm . to convert this to an absolute efficiency , one requires the absolute efficiency of sodium salicylate . reference @xcite refers its readers to reference ( * ? ? ? * table 7.1 ) for these efficiency numbers . unfortunately , the absolute efficiency numbers in ( * ? ? ? * table 7.1 ) differ by more than a factor of three . reference @xcite also measures the fluorescence efficiency of tpb along with several other fluors , but does so at much shorter wavelengths , ranging from 0.989 to 6.76 nm . the uncertainties of the efficiencies measured in reference @xcite are quite good compared to most others , at roughly 25% . tpb represents a significant improvement over other fluors ( like sodium salicylate ) because it can be applied to a substrate via vacuum deposition rather than by wet - dipping a slide . this dramatically enhances film reproducibility as allowing for cleaner application of the wavelength shifter . this is particularly important in low - background applications like direct dark matter detection , where the low levels of naturally occurring uranium and thorium can serve to obscure the signal of interest . our experimental apparatus consists of three stages : an euv light source , a filter wheel with several 2.5 cm diameter acrylic sample disks , and then one of two photon sensors . a diagram of the experimental apparatus can be found in figure [ fig : expapp ] . window , and the tpb film was evaporated onto the side of the sample disc facing the monochromator . ] in the first configuration , light was observed by a uv / visible spectrometer , and in the second configuration , a calibrated , silicon photodiode cell was installed . in order to minimize the effect of attenuation on our sensitivity , the entire space from the exit window of the deuterium lamp to the photon sensor was kept at a very low pressure , never exceeding @xmath4 mbar . the vacuum space also included the entire filter wheel assembly , and the photon sensor . additionally , in the first configuration , the spectrometer was coupled to the vacuum space through a collimating lens and fiber feedthrough . the light source used in our measurement was a model 632 deuterium light source , combined with a model 234/302vm 0.2-meter euv monochromator from mcpherson , inc . this light source has strong peaks at 128 nm and 160 nm , with a long high - wavelength tail that extends up to approximately 250 nm . the mgf@xmath5 window separating the deuterium arc lamp from the rest of the light source means that the intensity of the lamp cuts off at 110 nm ( see figure [ fig : scintspec ] for the transmittance of mgf@xmath5 as a function of wavelength ) . the monochromator was built around a 1200 g / mm holographic diffraction grating , which is rotated with respect to the instrument s entrance and exit slits to select specific wavelengths . both slits are 10 mm high , and have adjustable widths from 0.01 mm to 3 mm . we took data for all measurements with the entrance and exit slits set to 3 mm to maximize the light output of our system . there is a range from 205 nm to 245 nm where our visible spectrometer is sensitive to the direct light from the lamp ( see figure [ fig : directlamp ] ) . in this range , we measure the full - width , half - maximum resolution of the monochromator to be @xmath6 nm for our slit configuration , after subtracting the resolution of the spectrometer itself . we use this for our estimate of the monochromator resolution at all wavelengths . after leaving the exit slit of the light source , the monochromatic light entered a model 648 vacuum filter wheel , also from mcpherson . this allowed for selection between an open port , a fully absorbing metal shutter or one of several tpb - coated acrylic disks . in the open port configuration , the light source intensity could be observed directly by the installed photosensor . the metal shutter allowed a dark reading to be made with each photosensor and subtracted . the ability to make both of these measurements without breaking the vacuum seal was critical to minimizing systematic uncertainties from time variation of the light source and the photosensors . the light source was observed to drop in intensity by 3.6% per hour during data collection . as a result , the intensity of the lamp was always measured with the filter wheel in the open position no more than a few minutes before the measurement of the re - emitted light with a tpb disk in position . the residual time variation is taken as an uncertainty . the sample tpb film was evaporated onto the face of the acrylic disc facing the light source , because the acrylic is itself opaque to euv ( see figure [ fig : transmitsuvt ] ) . the tpb film was deposited on circular disks 2.5 cm in diameter and 0.47 cm in thickness . to allow a wide range of re - emission wavelengths to pass through the disk to the photosensors , the disks were made of solacryl suvt acrylic , manufactured by spartech polycast . we measured the transmittance to be greater than 80% for wavelengths longer than 290 nm , as shown in figure [ fig : transmitsuvt ] . the index of refraction of the acrylic is listed by the manufacturer s data sheet to be 1.49@xcite . the tpb was scintillation grade , purchased from aldrich , and the film was created on the acrylic disks via vacuum deposition . for this process , tpb powder was placed in a metal boat inside a vacuum bell jar , which then pumped down to a pressure of @xmath7 mbar . the tungsten chimney boat was then heated for approximately fifteen to twenty minutes , evaporating tpb onto the acrylic sample to a film thickness of 0.22 mg/@xmath8 . this film thickness was measured directly to be @xmath9 @xmath10 m using reflectometry . the film thickness was chosen to maximize light yield after a series of optimization measurements done with prototype detectors and test benches . the film tested in this work here is much thinner than previous studies outlined in @xcite . all tests described in this article were performed on the same sample . it is important to emphasize the fact that tpb films prepared via vacuum deposition will behave rather differently than those prepared via other methods ( such as wet dipping or some kind of process involving a solvent or paint ) . while films prepared via vacuum deposition are generally viewed as more uniform and repeatable than those prepared via some other methods , no attempt was made to check the film thickness for variability across the disk . the reflectometry measurement was however performed at the same spot on the disc illuminated by the exit slit . it is also true that serial testing of multiple tpb films , prepared under ostensibly identical circumstances would aid in understanding `` film to film '' variability . the visible re - emission spectra were taken with a qe65000 spectrometer from ocean optics . in order to direct light from the vacuum chamber containing the light source and tpb samples , a collimating lens was used to focus light onto a fiber optic vacuum feedthrough out to a 200 @xmath10 m diameter fiber leading to the spectrometer input port . spectra were dark - subtracted in the data acquisition software and analyzed with the root analysis framework@xcite . a broadband xenon lamp was used to calibrate the relative transmittance of the optical train leading to the spectrometer , shown in figure [ fig : specopticstrans ] . the xenon lamp was observed through a fiber with a transmittance function calibrated by the manufacturer . then the spectrum of the same lamp was observed through the complete optical train , divided by the raw lamp spectrum , and rescaled to produce the relative transmittance function . m fiber leading to the spectrometer . the function has been normalized so that the highest value is 1 . ] the absolute intensity measurements were made with an axuv100 g photodiode from international radiation detectors , inc ( ird)@xcite . this device has a one centimeter by one centimeter active photodiode area . it is a windowless silicon photodiode cell , with a calibrated sensitivity from 80 nm to 1100 nm . the device is completely passive , so it was read out with a keithley 6487 picoammeter . the calibration , provided by nist for incident light from 80 to 254 nm and ird for light from 200 to 1100 nm , scales a measured current into a radiant flux as a function of wavelength . figure [ fig : axuvcalibration ] shows the wavelength dependence of the photodiode response . we captured visible re - emission spectra for input euv wavelengths of : 128 , 160 , 175 , and 250 nm . these wavelengths correspond to : the peak of the argon scintillation spectrum , a bright peak in the emission spectrum of our light source , the peak of the xenon scintillation spectrum , and the peak of the emission spectrum of most ultraviolet light emitting diodes . the normalized visible re - emission spectra captured for all four input wavelengths are presented in figure [ fig : visspec ] . upon examination of figure [ fig : visspec ] , we find no strong incident wavelength dependence in the shape of the visible re - emission spectra . all four spectra show a strong cut - off at short wavelength at roughly 400 nm . the only slight difference appears in the 128 nm spectrum , which shows a slight excess above the others at just over 500 nm . we use the 128 nm re - emission spectra to calculate the fluorescence efficiency for all incident wavelengths in section [ sec : fleff ] . we define the fluorescence efficiency of the tpb layer as a function of incident wavelength , @xmath11 , to be the ratio of the number of re - emitted visible photons to the number of uv photons incident on the tpb layer . the efficiency at a given wavelength , @xmath12 , is computed from the current produced by the photodiode when directly viewing the light source , @xmath13 , the current when viewing the opaque shutter , @xmath14 , and the current when viewing the acrylic disk with tpb film on it , @xmath15 . the dark - subtracted current ratio is converted to an efficiency using the equation @xmath16 where @xmath17 is a geometric efficiency based on the configuration of the apparatus , @xmath18 is the photodiode response function shown in figure [ fig : axuvcalibration ] , @xmath19 is the gaussian wavelength distribution of the monochromator centered around @xmath12 , and @xmath20 is the re - emission spectrum of tpb shown in figure [ fig : visspec ] . the geometric efficiency , @xmath17 , is the ratio of the number of uv photons reaching the photodiode when the filter wheel is in the open position to the number of visible photons observed by the photodiode when a tpb sample is in position , assuming unit fluorescence efficiency . this constant is independent of the wavelength of the incident euv light and the re - emission spectrum of visible light . in general , @xmath17 is different from 1 due to the different angular distributions of incident and reemitted light , the effects of refraction at the acrylic - vacuum interface of the tpb sample disk , and the limited solid angle of the photodiode . we evaluate @xmath17 using a simple photon - tracing monte carlo simulation that initiates photons at the diffraction grating of the monochromator , propagates them through the exit aperture to the 2.3 cm diameter hole in the filter wheel where they can interact with a tpb layer deposited on an acrylic disk before arriving at the photodiode . fresnel reflection and refraction at the tpb and acrylic surfaces is included in the monte carlo , but reflection at the various interior black walls of the apparatus are neglected as second - order effects . not only does the geometric efficiency depend on the physical location and size of objects in the measurement apparatus , but it also depends on the intrinsic angular distribution of re - emitted light from the tpb film . the angular distribution of re - emission from vapor - deposited thin films of tpb has not been measured yet , so we would like to report an efficiency measure that is independent of this unknown angular distribution . therefore , we define the forward efficiency , @xmath21 , to be @xmath22 where @xmath23 is the ratio of the number of uv photons reaching the photodiode when the filter wheel in the open position to the number of uv which reach the tpb surface . the arrangement of our apparatus is such that @xmath23 is exactly 1 . by defining forward efficiency in this manner , the total efficiency can be calculated given an angular distribution of tpb re - emission , @xmath24 where @xmath25 is the acceptance fraction of a 1 cm by 1 cm square illuminated by the tpb angular distribution from a distance of 2.1 cm away after passing through 7 mm of suvt acrylic . the forward efficiency of our tpb film is presented in figure [ fig : forwardefficiency ] . there is a general trend below 190 nm toward increasing fluorescence efficiency with decreasing uv wavelength . to illustrate the relative sizes of the components of the overall uncertainty , table [ tab : uncert ] shows the fractional uncertainties for the tpb efficiency with 130 nm incident light . the statistical uncertainty in all current measurements was estimated by considering the rms variation in the current values taken with the metal shutter closed throughout the data collection process . this uncertainty was propagated through dark - subtracted current ratio from equation [ eq : forward_efficiency ] . the systematic uncertainty caused by short term variation in the lamp was estimated by observing the lamp intensity at 120 nm , where the effect was most significant , at different times and dividing this difference over the time between measurements at the other wavelengths . the final two systematic uncertainties account for the uncertainty in the response of the photodiode integrated over the ultraviolet and visible bands . this includes not only the uncertainty in the calibration , but also the uncertainty in the spectrum of the light source and tpb re - emission . by far , the dominant uncertainty in the overall measurement is statistical , caused by relatively large fluctuations in the current measurement relative to the low intensity of the light source at 130 nm . number of visible photons observed at photodiode sensor per incident euv photon a function of input euv photon wavelength . ] .[tab : uncert]uncertainties on forward efficiency per unit solid angle at 130 nm . [ cols="<,>",options="header " , ] to illustrate what the total fluorescence efficiency of tpb could be given a reasonable assumption about the angular distribution of re - emission , we can look to a similar fluor : sodium salicylate . for film thicknesses that maximize fluorescence efficiency , the angular distribution of re - emitted photons from sodium salicylate follows a cosine law ( or lambertian ) distribution in both the forward and backward directions@xcite . we can apply this same distribution to tpb in order to estimate the total fluorescence efficiency . figure [ fig : totalefficiency ] shows the total efficiency of tpb fluorescence computed from our measurement under this assumption . the acceptance fraction , @xmath25 , is 0.0179 in this case . under the assumption of lambertian re - emission , we find variation an average value of the total fluorescence efficiency between 0.7 and 1.35 visible photons emitted for each euv photon absorbed . this reaches a minimum value of approximately 0.7 at 190 nm . there is a strong upturn in the fluorescence efficiency at short wavelengths , leading to a value of approximately 1.2 at 128 nm . we have presented measurements of both the visible re - emission spectrum and fluorescence efficiency of thin tpb films evaporated onto acrylic . the visible spectrum we measure here is somewhat closer to that presented in reference @xcite than it is to the one from @xcite , although our spectrum seems cuts off at a somewhat shorter wavelength than seems apparent in @xcite . the details of the work presented in this article are somewhat specific to the experiment in terms of the preparation of the tpb film . the next step in tpb characterization is to measure the angular distribution of re - emission in order to test the assumption of a lambertian distribution in both the forward and backward directions . as a future study , it also would be interesting to vary the manner in which the tpb film is prepared . additionally , our lamp intensity cuts off at approximately 110 nm because of the transmittance of its mgf@xmath5 window ( see figure [ fig : scintspec ] ) . clearly , it would be advantageous to extend these efficiency measurements down to 7080 nm so that we could characterize these films near the scintillation wavelengths of neon and helium as well . it is also possible that we could examine other fluor molecules as candidates for use in noble element scintillation detectors . all of these efforts will help to continue to make these noble element scintillation detectors a viable technology in the coming years , and will continue to push forward the optimization of this radiation detection technique . this work was supported by los alamos national laboratory s directed research and development program , the department of energy s na-22 office of nonproliferation research & development , national science foundation grant number phy-0758120 , and a research excellence development grant from the university of south dakota . the authors thank victor h. gehman , jr . and stephen h. jaditz for their careful reading and constructive comments . the authors would also like to thank hugh lippincott , james nikkel and dan mckinsey for the preparation of the sample disks and valuable discussions regarding detection of euv scintillation light with tpb as well as andrew m. dattelbaum and anatoly v. efimov for assistance with reflectometry measurements of our tpb film thickness . 00 e. aprile , a. e. bolotnikov , a. i. bolozdynya , and t. doke . noble gas detectors . wiley - vch verlag gmbh & co. kgaa . weinheim , germany , ( 2006 ) . r. brunetti , _ et al_. warp liquid argon detector for dark matter survey . _ new astron . _ , * 49 * , 265 ( 2005 ) . m. haranczyk , _ et al_. the ardm experiment . _ acta physica polonica b _ , * 41 * , 1441 ( 2010 ) . d.n . mckinsey and the miniclean collaboration . the mini - clean experiment . _ , * 173 * , 152 . mckinsey and k.j . coakley . neutrino detection with clean . _ , * 22 * , 355 ( 2005 ) . icarus collaboration . a second - generation proton decay experiment and neutrino observatory at the gran sasso laboratory . _ lngs - p28/2001 _ , ( 2001 ) . m. soderberg and the microboone collaboration . microboone : a new liquid argon time projection chamber experiment . _ aip conf . _ , * 1189 * , 83 ( 2009 ) . v. barger , _ et al_. report of the us long baseline neutrino experiment study . _ fermilab-0801-ad - e _ , ( 2007 ) . michniak , r. alleaume , d.n . mckinsey , and j.m . doyle . alpha and beta particle induced scintillations in liquid and solid neon . _ nuclear instruments and methods in physics research a _ , * 482 * , 387 ( 2002 ) . glenn f. knoll . radiation detection and measurements , third edition . john wiley and sons , inc . hoboken , nj . ( 2000 ) j.b . scintillation counters . mcgraw - hill book co , new york , ( 1953 ) . the nedm experiment ( m. d. cooper and b. w. filippone , spokespersons ) . http : p25ext.lanl.gov / edm / edm.html r. golub and s.k . lamoreaux , _ physics reports _ , * 237 * , 1 ( 1994 ) r. acciarri , _ et al_. effects of nitrogen and oxygen contamination in liquid argon . _ nuclear physics b - proceedings supplements _ , * 197 * , 70 ( 2009 ) . doi : 10.1016/j.nuclphysbps.2009.10.037 . arxiv:0804.1217v1 [ nucl - ex ] w.h . lippincott . direct detection of dark matter with liquid argon and neon . phd thesis . yale university ( 2010 ) . m. stockton , j.w . keto and w.a . ultraviolet emission spectrum of electron - bombarded superfluid helium . _ physical review letters _ , * 24 * , 654 ( 1970 ) . packard , f. reif and c.m . ultraviolet emission spectra of electron - excited solid and liquid neon . _ physical review letters _ , * 25 * , 1435 ( 1970 ) . k. saito , _ et al_. absolute number of scintillation photons emitted by alpha - particles in rare gases . _ ieee transactions on nuclear science _ , * 49 * , 1674 ( 2002 ) . hamamatsu photonics . photomultiplier tubes : basics and applications , third edition . http://sales.hamamatsu.com/assets/applications/etd/ + pmt_handbook_complete.pdf ( 2006 ) c.h . et al_. uv quantum efficiencies of organic fluors . _ nuclear instruments and methods in physics research b _ , * 117 * , 421 ( 1996 ) . burton and b.a . fluorescence of tetraphenyl - butadiene in the vacuum ultraviolet . _ applied optics _ , * 12 * , 87 ( 1973 ) . j.a.r techniques of vacuum ultraviolet spectroscopy . john wiley & sons , inc . regan , _ et al_. measured conversion efficiencies of p45 , paraterphenyl , tetraphenyl butadiene , and sodium salicylate phosphors in the soft - x - ray wavelength range . _ applied optics _ , * 33 * , 3595 ( 1994 ) . r. brun and f. rademakers . root - an object oriented data analysis framework . proceedings aihenp96 workshop , lausanne , sep . 1996 , _ nuclear instruments and methods in physics research a _ , * 389 * 81 - 86 ( 1997 ) . canfield , _ et al_. absolute silicon photodiodes for 160 nm to 254 nm photons . _ metrologia _ , * 35 * , 329 ( 1998 ) e.m . gullikson , _ et al_. stable silicon photodiodes for absolute intensity measurements in the vuv and soft x - ray regions . _ journal of electron spectroscopy and related phenomena _ , * 80 * , 313 ( 1996 ) spartech polycast . solacryl stabilized uv transmitting acrylic datasheet . r. allison , j. burns , and a. j. tuzzolino . absolute fluorescent quantum efficiency of sodium salicylate . _ , * 54 * , 747750 ( 1964 ) .	a large number of current and future experiments in neutrino and dark matter detection use the scintillation light from noble elements as a mechanism for measuring energy deposition . the scintillation light from these elements is produced in the extreme ultraviolet ( euv ) range , from 60200 nm . currently , the most practical technique for observing light at these wavelengths is to surround the scintillation volume with a thin film of tetraphenyl butadiene ( tpb ) to act as a fluor . the tpb film absorbs euv photons and reemits visible photons , detectable with a variety of commercial photosensors . here we present a measurement of the re - emission spectrum of tpb films when illuminated with 128 , 160 , 175 , and 250 nm light . we also measure the fluorescence efficiency as a function of incident wavelength from 120 to 250 nm . # 1#2@xmath0#2 noble gasses ; scintillation light ; ultraviolet photons ; dark matter ; neutrinos
the concept of quantum repeater @xcitewas introduced for long - distance quantum communication in order to overcome the problems caused by inevitable photon loss in the transmission channel . generating distant entanglement is a crucial ingredient of a quantum repeater protocol . in 2001 , duan , lukin , cirac and zoller ( dlcz ) proposed a original scheme @xcite to use atomic - ensembles and linear optics in which robustly generating entanglement over long distances can be achieved . motivated by the dlcz protocol , much experimental effort@xcite has been made in the last few years . however , it has been shown that dlcz protocol requires severe phase stability since entanglement generation and entanglement swapping in the protocol depend on single - photon mach - zehnder type interferences@xcite , and this problem make its experimental realization extremely difficult by the current technology . hence a novel scheme@xcite based on dlcz scheme was proposed to use phase - insensitive two - photon quantum interference which dramatically relax this phase stability requirements . a latest experiment@xcite has primarily demonstrated the scheme over a distance of 300 meters . meanwhile , a fast and robust approach @xcite was proposed and also can solve the phase stability problem as long as the entanglement generation is performed locally , and they gave a detailed comparisons between the schemes . in addition , there are another novel schemes @xcite proposed to improve the efficiency of quantum repeaters . these protocols greatly stimulate experimental implementations of quantum repeater , but there are still some problems need to be solved , such as reliable transmission of photon s polarization states over noisy channel . since the photon interferences rely on the polarization states , the ability to maintain photonic polarizations is indispensable in the process of distant entanglement generation or swapping . most of the time , optical fibers are used as photon transmission channel . due to the fiber birefringence , the photonic polarizations will be changed randomly @xcite . experimentally , active feedback compensation could be applied to solve this problem@xcite , but it is efficient only when the thermal and mechanical fluctuations are rather slow . furthermore , even though polarization compensations can be used efficiently , imperfect shared reference frame ( srf ) for polarization orientation may cause some errors . it is difficult to correct this kind of errors since establishing a perfect srf requires infinite communication@xcite . due to these reasons , it would be better to have a quantum repeater scheme with inherent polarization insensitivity . in this paper , we propose an alternative approach to create distant entanglement between atomic - ensembles . in our scheme neither the phase stabilizing nor the polarization calibrating is needed for photons transmitted over long distance . through introducing time - bin photonic states and using a new two - photon interference configuration , we make only the unchanged part of initial polarization states contribute to the desired results . combined with local entanglement swapping for entanglement connection , our scheme can be used to implement a robust quantum repeater . here optical thick atomic ensemble , which includes @xmath0 atoms with @xmath1 level structure ( see inset of fig.[fig:1 ] ) , is used as quantum memory . each atomic ensemble is illuminated by a short , off - resonant write pulse that induces a spontaneous raman process . this process will produce a forward - scattered stokes light and a collective atomic excitation state@xcite . the photon - atom system can be described as(neglecting the higher - order terms ) @xmath2\|vac\rangle,\end{aligned}\ ] ] where @xmath3 denotes that all the ensemble atoms are in the ground state @xmath4 and the stokes light in the vacuum state , @xmath5 is the creation operator of the stokes light , and the collective atomic excitation is defined by @xmath6 . the excitation probability @xmath7 can be achieved by manipulating the write laser pulse . at each communication node , two atomic ensembles are used to encode a memory qubit(see the fig.[fig:1 ] ) . the two ensembles are excited simultaneously by write laser pulses , and the stokes photons generated from them have orthogonal polarization states , i.e. @xmath8 and @xmath9 , which denote horizontal and vertical linear polarization , respectively . the state @xmath8 propagates through a short path ( @xmath10 ) and the state @xmath9 goes through a long path ( @xmath11 ) , and are combined at a polarization beam splitter ( @xmath12 ) which transmits @xmath8 and reflects @xmath9 . therefore photon wave packets with different polarizations correspond to different time bins . as long as the path difference is less than the photons coherence length ( about @xmath13 or more for photons generated from atomic ensembles@xcite ) , after the pbs the memory qubit is effectively entangled with the polarization and the time - bin states of the emitted stokes photons . the state of the atom - photon system can be written as @xmath14\}\|vac\rangle , \end{split}\end{aligned}\ ] ] where the subscripts @xmath15 and @xmath16 are used to distinguish the two ensembles , and @xmath17(@xmath18 ) denotes the creation operator of the stokes photon with vertical ( horizontal ) polarization passing through the long ( short ) path . assume two neighboring communication nodes , denoted by @xmath19 and @xmath20 , are connected through certain transmission channel such as optical fiber with a distance of @xmath21 . at the middle point between the two nodes , the stokes photons are firstly directed to pbs . due to channel noise or imperfect srf , the polarization states of photons arrived at the middle point have been changed , and as long as the time intervals between the subsequent transmitted photon wave packets are small , it is reasonable to assume that the polarization changes are the same to them . after the pbs , there are also a long path and a short one whose difference is the same as the one at each node . the polarization transformation during this process are usually considered as a unitary transformation@xcite , so the evolution of the photonic polarization components can be described as : @xmath22 @xmath23 where @xmath24 and @xmath25 are random noise parameters . obviously , there are four time bins @xmath26 and @xmath27 for each stokes photon , and @xmath28 are the same time bin . now the atom - photon system can be described as : @xmath29 with @xmath30\}\|vac\rangle , \end{split}\end{aligned}\ ] ] @xmath31\}\|vac\rangle , \end{split}\end{aligned}\]]@xmath32\}\|vac\rangle , \end{split}\end{aligned}\ ] ] here for simplicity the @xmath25 and the polarization subscripts are not visibly expressed in the states @xmath33 and @xmath34 without any influences on the subsequent analysis . the setup for photon interferences at the middle point ( see fig.[fig:1 ] ) is that the photons after the @xmath12 are directed into beam splitters ( @xmath35 ) followed by single photon detectors which are turned on only at the time @xmath36(@xmath37 ) . therefore the terms only including time bins @xmath38 and @xmath27 will not have any contributions to the detection results and can be safely neglected . to generate entanglement between the nodes @xmath19 and @xmath20 , laser pulses excite the ensembles in both nodes simultaneously , and the whole system is described by the state @xmath39 , where @xmath40 and @xmath41 are given by equation ( [ eq : state at m ] ) with all the operators and states distinguished by the subscript @xmath19 and @xmath20 . a coincidence count between the detectors at either detection zone , e.g. @xmath42 and @xmath43 , will project the neighboring memory qubits into a complex state with contributions from second - order excitations . note that in second order in @xmath44 , the states @xmath45 can just trigger one detector at either detection zone at the time @xmath36(@xmath37 ) and consequently have no contribution to the coincidence . thus a coincidence count between detectors , for instance @xmath42 and @xmath43 , projects the two memory quits into @xmath46\|vac\rangle\ , \end{split}\end{aligned}\ ] ] where @xmath24(@xmath47 ) and @xmath48(@xmath49 ) are the noise parameter and the polarization - independent phase that the photons acquires during the transmission from the node @xmath19(@xmath20 ) to the middle point , respectively . the first part of the state is the desired maximally entangled state . the second part is unwanted two - excitation state and can be effectively eliminated by entanglement swapping . the success probability is on the order of @xmath50 , where @xmath51 is the detection efficiency and @xmath52 is the channel attenuation length . until now , we have generated the state similar to the one ( eq.(s3 ) in supplementary information of ref.@xcite ) which was created in the recent experimental demonstration of quantum repeater over a distance of 300 meters . it is obvious that the polarization noises only influence the success probability but have no effect on the fidelity of the desired state , and the phases @xmath48 and @xmath49 only lead to a trivial global factor @xmath53 on the desired state . the entanglement swapping setup is depicted in fig.[fig : enw ] , which is the same as the one in the schemes@xcite . note that there is no path difference at this situation . consider three communication nodes @xmath19 , @xmath20 and @xmath54 , and assume that we have created the complex entangled states ( given by eq.([eq : final state ] ) ) @xmath55 and @xmath56 both in ( @xmath57 ) and in ( @xmath58 ) , respectively . the memory qubits @xmath59 and @xmath60 at node @xmath20 are illuminated simultaneously by retrieval laser pulses . the retrieved anti - stokes photons are subject to bell - state measurement ( @xmath61 ) which is used to eliminate the two - excitation terms since the arrangement of the @xmath12s is to identify @xmath62at @xmath63 polarization basis , where @xmath64 . therefore the two - photon states generated from the unwanted two - excitation terms are directed into the same detection zone and will not induce any coincidences . in addition , the capability of distinguishing photon numbers is technically demanding , and the retrieve efficiency is determined by the optical depth of the atomic ensembles@xcite . taking into account of these imperfections , the coincidence counts in @xmath61 actually prepare the memory qubits into a mixed entangled state of the form @xmath65 , where the coefficients @xmath66 , @xmath67 and @xmath68 are determined by the retrieval efficiency and detection efficiency@xcite . here @xmath69 is the maximal entangled state , where @xmath70 , @xmath71 is the vacuum state that all the atomic ensembles are in the ground states , and @xmath72 is a maximally mixed state in which only one of the four atomic ensembles has one excitation . although @xmath73 still includes unwanted terms , it can be projected automatically to maximally entangled states in the entanglement - based quantum cryptography schemes . when implementing quantum cryptograph via ekert protocol@xcite , only the coincidence counts between the detectors at two remote nodes are registered and used for quantum cryptography . therefore only the maximally entangled state in @xmath73 will contribute to the experimental results . in this sense , @xmath73 is equivalent to the bell state @xmath74 . to implement a quantum repeater protocol , further entanglement swapping is required . since the noise parameters only change the coefficients of the obtained states , the analysis of the entanglement connection in ref.@xcite can be directly used here , that is , as long as the excitation probability @xmath44 is small enough , the contributions from the higher - order excitations can be safely neglected . the probability to find an desired entangled pair in the remaining memory qubits is almost a constant and will not decrease significantly during the entanglement connection process . in the existing quantum repeater protocols , the ability to reliably transfer of photon s polarization is indispensable , but it is not easy to meet the prerequisite in practice . for this reason , we have proposed an alternative approach with inherent polarization insensitivity to generate entanglement between distant communication nodes . in our scheme , neither the phase stabilizing nor the polarization calibrating is needed for photons transmitted over long distance . through introducing path difference , we make photon wave packets with different polarizations correspond to different time - bins , and use the two - photon interference with different configuration to generate entanglement between remote communication nodes . hence only the unchanged part of initial polarization states can induce the coincidences between detectors , so that the polarization noise only have influence on the success probability but the fidelity of the desired states . consider the function of atom ensembles as quantum memory for entanglement of a storage time up to milliseconds@xcite , the capacity of creating high - fidelity entangled states may be preferred even if at the cost of some efficiencies . combined with local entanglement swapping for entanglement connection , our scheme can be used to implement a robust quantum repeater . comparing with the dlcz scheme and the scheme of ref.@xcite , the generated entangled states in our scheme are of higher fidelity . finally , it is pointed out that our scheme can be used to perform remote entanglement swapping after locally generating entanglement , which would make the quantum repeater scheme more efficient@xcite . for instance , the local entangled state ( eq.(10 ) in ref.@xcite ) generated via the setup in fig.[fig : enw ] may be robustly connected to longer communication distance by use of our scheme . we thank cheng - xi yang , lin yang and xian - min jin for useful discussions . x.b.wang was supported in part by the national basic research program of china grant no . 2007cb907900 , and 2007cb807901 , nsfc grant no . 60725416 and china hi - tech program grant no . liang was supported by national funds of natural science grant no .	we propose an alternative scheme for quantum repeater without phase stabilization and polarization calibration of photons transmitted over long - distance channel . we introduce time - bin photonic states and use a new two - photon interference configuration to robustly generate entanglement between distant atomic - ensemble - based memory qubits . our scheme can be performed with current experimental setups through making some simple adjustments .
finite element analysis ( fea ) is the most widely used analysing tool in computer aided engineering ( cae ) . one key factor to achieve an accurate fea is the layout of the finite element mesh , including both mesh density and element shape @xcite . regions containing complex boundaries , rapid transitions between geometric features or singularities require finer discretisation @xcite . this leads to the development of adaptive meshing techniques that assure the solution accuracy without sacrificing the computational efficiency @xcite . the construction of a high quality mesh , in general , takes the most of the analysis time @xcite . the recent rapid development of the isogeometric analysis @xcite , which suppressed the meshing process , has emphasised the significance of mesh automation in engineering design and analysis . quadtree in fea is a kind of hierarchical tree - based techniques for adaptive meshing of a 2d geometry @xcite . it discretises the geometry into a number of square cells of different size . the process is illustrated in fig.[fig : qtreerep1 ] using a circular domain . the geometry is first covered with a single square cell , also known as the root cell of the quadtree ( fig.[fig : qtreerep1]a ) . as shown in fig.[fig : qtreerep1]b , the root cell is subdivided into 4 equal - sized square cells and each of the cells is recursively subdivided to refine the mesh until certain termination criteria are reached . in this example , a cell is subdivided to better represent the boundary of the circle and the subdivision stops when the predefined maximum number of division is reached . the final mesh is obtained after deleting all the cells outside the domain ( fig.[fig : qtreerep1]c ) . the cell information is stored in a tree - type data structure , in which the root cell is at the highest level . it is common practice to limit the maximum difference of the division levels between two adjacent cells to one @xcite . this is referred to the @xmath0 rule and the resulting mesh is called a balanced @xcite or restricted quadtree mesh @xcite . a balanced quadtree mesh not only ensures there is no large size difference between adjacent cells , but also reduces the types of quadtree cells in a mesh to the 6 shown in fig.[fig : qtreecell ] . owing to its simplicity and large degree of flexibility , the quadtree mesh is also recognised in large - scale flood / tsunami simulations @xcite and image processing @xcite . generation of quadtree mesh on a circular domain . ( a ) cover the domain with a square root cell ( b ) subdivide the square cells ( c ) select the cells based on the domain boundary ] 6 main types of master quadtree cells with @xmath0 rule enforced ] it is , however , not straightforward to integrate a quadtree mesh in a fea . the two major issues are illustrated by fig.[fig : qtreerep ] , which shows the quadtree mesh of the top - right quadrant of the circular domain in fig.[fig : qtreerep1 ] . 1 . _ hanging nodes _ middle nodes , shown as solid dots in fig.[fig : qtreerep ] , exist at the common edges between the adjacent cells with different division levels . when conventional quadrilateral finite elements are used , a middle node is connected to the two smaller elements ( lower level ) but not to the larger element ( higher level ) . this leads to incompatible displacement along the edges and the middle nodes are called the hanging nodes @xcite . fitting of curved boundary _ quadtree cells are composed of horizontal and vertical lines only . as shown in fig.[fig : qtreerep ] , the quadtree cells intersected with the curved boundary have to be further divided into smaller ones to improve the fitting of the boundary . generally , the mesh has to be refined in the area surrounding the boundary . despite this , the boundary may still not be smooth ( fig.[fig : qtreerep1]c ) and may result in unrealistically high stresses . additional procedure is required to conform the mesh to the boundary . there exist a number of different approaches to ensure displacement compatibility when hanging nodes are present @xcite . three typical approaches among all are briefly discussed here . the first one is to subdivide the higher level quadtree cells next to a hanging node into smaller triangular elements @xcite as shown in fig[fig : qtreerep ] . additional nodes may be added to improve the mesh quality and/or reduce the number of element types . these techniques lead to a final mesh that only contains conforming triangular elements . a similar approach was adpoted by @xcite , in which the quadtree mesh was subdivided into a conforming mesh dominated by quadrilateral elements . the second approach introduces special conforming shape functions @xcite to ensure the displacement compatibility . an early work by gupta @xcite reported the development of a transition element that had additional node along its side . a conforming set of shape functions was derived based on the shape functions of the bilinear quadrilateral elements . owing to its simplicity and applicability , gupta s work was further extended by @xcite and @xcite to hexahedral elements . fries et al . @xcite investigated two approaches to handle the hanging nodes within the framework of the extended finite element method ( xfem ) . they were different in whether the enriched degrees - of - freedom ( dofs ) were assigned to the hanging node . a similar work was reported by legrain et al . @xcite , in which the selected dofs are enriched and properly constrained to ensure the continuity of the field . quadtree mesh of the top - right quadrant of a circular domain . demonstration of subdivision ( dashed lines ) is given in two quadtree cells with hanging nodes on their sides . ] the third approach is to model the quadtree cells as _ _ n-__sided polygon elements by treating hanging nodes as vertices of the polygon . this approach generally requires a set of polygonal basis function . special techniques are usually required to integrate the resulting equations over arbitrary polygon domain @xcite . this development was initiated by @xcite who showed the use of rational basis functions for elements with arbitrary number of sides . tabarraei and sukumar @xcite in their work adapted their polygon element @xcite to quadtree mesh . the set of polygonal basis functions was derived using laplace interpolant . by using an affine map on the reference polygon , the conforming shape functions of a quadtree cell with the same number of vertices ( including the hanging nodes ) were obtained . they also reported a fast technique for computing the global stiffness matrix , making use of the quadtree structure by defining parent elements @xcite . in this way , the elemental stiffness matrix has to be computed only 15 times ( 4-node cell not included ) for a balanced quadtree mesh ( when @xmath0 rule is enforced ) . further development of their work with xfem was also reported in @xcite . as mentioned in a recent paper @xcite , the development of high - order polygon elements received relatively less attention . mibradt and pick @xcite devised high order basis functions for polygons based on the natural element coordinates . those basis functions , however , are not complete polynomials . rand et al . @xcite developed a quadratic serendipity element for arbitrary convex polygons based on generalised barycentric coordinates . the potential of using their approach for higher order serendipity elements on convex polygons was also reported . based on the same approach , sukumar @xcite recently developed the quadratic serendipity shape functions that were applicable for convex and nonconvex polygons and were complete quadratic polynomials . the shape functions were obtained through solving an optimisation problem , which was derived from the maximum - entropy principle . besides dealing with the hanging nodes in a quadtree mesh , fitting complex boundaries is another challenging part in the mesh generation . @xcite proposed trimming the quadtree cells , intersected with the boundary , into polygons before a further subdivision process into triangles or quadrilaterals . alternatively those cells were first subdivided and some of the vertices were repositioned based on their projections onto the boundary @xcite . in @xcite and @xcite , after the subdivision , a buffer zone was introduced between the boundary and the internal quadtree cells . a compatible mesh was then constructed to fill up this zone . all these techniques require an additional optimisation step to ensure the final mesh quality . within the framework of xfem , quadtree cells intersected with the boundary were not modified in pre - processing stage @xcite . however , when constructing the stiffness matrix , the domain boundary is still required to identify the portion of the cell within the domain for numerical integration . in the integration process , that portion of the cell within the domain is either subdivided into geometric sub - cells @xcite or treated as a polygon @xcite . the scaled boundary finite element method ( sbfem ) provides an attractive alternate technique to construct polygon elements ( scaled boundary polygon ) @xcite ( fig.[fig : sbfempolygon ] ) . it is a semi - analytical procedure developed by song and wolf to solve boundary value problems @xcite . the only requirement for a scaled boundary polygon is that its entire boundary is visible from the _ scaling centre _ only the edges of the polygon are discretised into line elements . the number of line elements on an edge can be as many as required . any type of displacement - based line elements , including high - order spectral elements , can be used . the domain of the scaled boundary polygon is constructed by scaling from its scaling centre to its boundary , and the solution within the polygon is expressed semi - analytically @xcite . a salient feature of the scaled boundary polygons is that stress singularities occurring at crack and notch tips , formed by one or several materials , can be accurately modelled without resorting to asymptotic enrichment and local mesh refinement . its high accuracy and flexibility in mesh generation lead to simple remeshing procedures when modelling crack propagation @xcite . scaled boundary representation of a polygon ] scaled boundary representation of quadtree cells ] this paper presents a technique for the stress and fracture analysis by integrating the scaled boundary finite element method ( sbfem ) with quadtree mesh of high - order elements . this integrated technique possesses the following features : 1 . hanging nodes are treated without cell subdivision . each quadtree cell is modelled as a scaled boundary polygon as shown in fig.[fig : sbfequadtreerep ] . the edges of a quadtree cell can be divided into more than one line element to ensure displacement compatibility with the adjacent smaller cells . hanging - nodes are thus treated the same as other nodes . owing to the sbfe formulations @xcite , no additional procedure is required to compute the shape functions for the quadtree cells . high - order elements can also be used within each quadtree cell directly . the entire quadtree meshing process is simple and automatic . the boundary of the problem domain is defined using signed distance functions @xcite . only seed points @xcite are required to be predefined to control the mesh density . owing to the ability of the sbfem in constructing polygon elements of , practically , arbitrary shape and order , the quadtree cells trimmed by curved boundaries are simply treated as a non - square scaled boundary polygon . high - order elements can be used to fit curved boundaries closely . the resulting mesh conforms to the boundary without excessive mesh refinement ( see fig.[fig : sbfequadtreerep ] ) . no local mesh refinement or asymptotic enrichment is required for a quadtree cell containing a crack tip to accurately model the stress singularity . the present paper is organised as follows . the summary of the sbfem and its application to quadtree cells are first presented in the next section . it is followed by the developed algorithm of quadtree mesh generation in section[sec : quadtree - mesh - generation ] . five examples are given in section[sec : numerical - examples ] with detailed discussion on accuracy and convergence . finally , conclusions of the present work are stated in section[sec : conclusion ] . this section summarises the scaled boundary finite element method for 2d stress and fracture analysis . only the key equations that are related to its use with a quadtree mesh are listed . a detailed derivation of the method based on a virtual work approach is given in @xcite . the sbfem can be formulated on quadtree cells by treating each cell as a polygon with arbitrary number of sides ( fig . [ fig : sbfequadtreerep ] ) . in each cell , a local coordinate system @xmath1 is defined at a point called the scaling centre from which the entire boundary is visible . @xmath2 is the radial coordinate with @xmath3 at the scaling centre and @xmath4 at the cell boundary . the edges of each cell are discretised using one - dimensional finite elements with a local coordinate @xmath5 having an interval of @xmath6 . it is noted that the hanging nodes appearing in the quadtree structure do not require any special treatment in the sbfem formulation . they are simply used as end nodes of the 1d elements . the coordinate transformation between the cartesian @xmath7 and the local @xmath1 coordinate systems are given by the scaled boundary transformation equations @xcite : @xmath8 where @xmath9^{\mathrm{t}}$ ] is the cartesian coordinates of a point in the cell , @xmath10 is the shape function matrix and @xmath11^{\mathrm{t}}$ ] is the vector of nodal coordinates of a cell with @xmath12 nodes . the displacement field in each cell @xmath13 is interpolated as @xmath14 where @xmath15 are radial displacement functions and are obtained by solving the scaled boundary finite element equation in displacement @xcite : @xmath16 with coefficient matrices @xmath17 where @xmath18 is the material constitutive matrix , @xmath19 and @xmath20 are the sbfem strain - displacement matrices and @xmath21 is the jacobian on the boundary required for coordinate transformation . . is solved by introducing the variable @xmath22 @xcite @xmath23^{\mathrm{t}}\label{eq : xksi}\end{aligned}\ ] ] where @xmath24 so that eq . is transformed into a first order ordinary differential equation with twice the number of unknowns : @xmath25 with the hamiltonian matrix @xmath26 @xcite @xmath27\label{eq : hamilton}\end{aligned}\ ] ] an eigenvalue decomposition of the @xmath26 results in @xmath28= & \left[\begin{array}{cc } \boldsymbol{\phi}_{\mathrm{u}}^{\mathrm{(n ) } } & \boldsymbol{\phi}_{\mathrm{u}}^{\mathrm{(p)}}\\ \boldsymbol{\phi}_{\mathrm{q}}^{\mathrm{(n ) } } & \boldsymbol{\phi}_{\mathrm{q}}^{\mathrm{(p ) } } \end{array}\right]\left[\begin{array}{cc } \boldsymbol{\lambda}^{\mathrm{(n ) } } & 0\\ 0 & \boldsymbol{\lambda}^{\mathrm{(p ) } } \end{array}\right]\label{eq : eigendecomp}\end{aligned}\ ] ] where @xmath29 and @xmath30 are the eigenvalue matrices with real parts satisfying @xmath31 and @xmath32 , respectively . @xmath33 and @xmath34 are the corresponding eigenvectors of @xmath29 whereas @xmath35 and @xmath36 are the eigenvectors corresponding to @xmath30 . for bounded domains such as those considered in this paper , only the eigenvalues satisfying @xmath31 lead to finite displacements at the scaling centre . using eq . and , the solutions for @xmath15 and @xmath37 are @xmath38 the integration constants @xmath39 in eq . and eq . are obtained from the nodal displacements at the cell boundary @xmath40 as @xmath41 the stiffness matrix of each quadtree cell is formulated as @xcite @xmath42 substituting eq . into eq . , the displacement field in a cell is @xmath43 using the hooke s law and the strain - displacement relationship , the stress at a point in a cell is @xcite @xmath44 where @xmath45^{\mathrm{t}}$ ] is the stress mode @xmath46 fig.[fig : crackrep ] shows how a crack is modelled with a quadtree cell . the crack tip is chosen as the scaling centre . the crack surfaces are not discretised . the line elements discretising the cell boundary do not form a closed loop . modelling of a crack with the scaled boundary finite element method . ] when a crack is modelled by the sbfem , two eigenvalues , @xmath47 , @xmath48 satisfying @xmath49 appear in @xmath29 . from eq.[eq : stresfield ] , it can be discovered that these eigenvalues lead to a stress singularity as @xmath50 . using the two modes corresponding to these two eigenvalues , the singular stresses are expressed as @xmath51 where @xmath52\label{eq : singeigenval}\end{aligned}\ ] ] and @xmath53 are the integration constants corresponding to @xmath54 . the singular singular stress modes @xmath55^{\mathrm{t}}$ ] is written as @xmath56 where @xmath57 are the modal displacements in @xmath33 corresponding to @xmath54 . the stress intensity factors can be computed directly from their definitions . for a crack that is aligned with the cartesian coordinate system as shown in fig.[fig : crackrep ] , the stress intensity factors are defined as @xmath58 substituting the stress components in eq . into eq . and using the relation @xmath59 ( @xmath60 is the distance from the scaling centre to the boundary along the direction of the crack , see fig.[fig : crackrep ] ) at @xmath61 leads to @xmath62 this section presents the developed algorithm for quadtree mesh generation . fig.[flowchart ] shows the flow chart of the overall process . the entire generation process is automatic with minimal number of inputs required from the user , which include * maximum allowed number of seed points in a cell @xmath63 , * seed points on each boundary @xmath64 and region of interest @xmath65 , * maximum difference between the division levels of adjacent cells @xmath66 , which is equal to 1 for a balanced quadtree mesh . flow chart of the quadtree mesh generation . ] this section is organised based on fig[flowchart ] . it first presents defining geometry using signed distance function , and assigning seed points on the boundary and the regions of interest . detailed explanations of the meshing steps , which include generating the initial quadtree grid , trimming the boundary quadtree cells into polygons and merging cells surrounding a crack tip , are then followed . to facilitate the description of the meshing steps , fig.[qtreedes0 ] shows a square plate with a circular hole and two local refinement features to be used as an example throughout this section . an efficient computation of the global stiffness matrix , by taking advantage on the quadtree mesh , is described at the end of this section . example to illustrate the quadtree mesh generation process : a square plate with a circular hole . an additional circle and an inclined line ( dashed lines ) are included to control local mesh density . ] the geometry is defined by using the signed distance function @xcite . it provides all the essential information of a geometry and can be operated with simple boolean operations to build up more complex geometries @xcite . the signed distance function of a point @xmath67 associated with a domain @xmath68 , which is a subset of @xmath69 , is given as @xmath70 where @xmath71 represents the boundary of the domain and @xmath72 is the _ euclidean norm _ in @xmath69 with @xmath73 . the sign function @xmath74 is equal @xmath75 when @xmath76 lies inside the domain and is equal 1 otherwise . this definition of the signed distance function is visualised in fig.[dispfunc ] . a number of distance functions in matlab for simple geometries are given in @xcite , including their boolean operations . signed distance function of the points inside the domain ( @xmath77 ) , on the boundary ( @xmath78 ) and outside the domain ( @xmath79 and @xmath80 ) ] for each boundary and region of interest , a set of pre - defined seed points @xcite is introduced to control the quadtree mesh density . there require four sets of predefined seed points for the example in fig.[qtreedes0 ] . two sets are for the square and the circular hole representing the actual domain boundary . the number of seed points directly controls the local density of the quadtree cells and the quality of fitting the boundary . this is further discussed in section[sub : polygon - boundary - cells ] . the other two sets are for the large circle and inclined line controlling local mesh density only . the meshing process starts with covering the problem domain with a single square cell ( the root cell ) . the dimension of the root cell is based on the larger one between the maximum vertical and maximum horizontal dimension of the geometry . the developed algorithm will check the number of seed points in the cell . if the number is larger than the predefined maximum allowed number , the cell will be divided into 4 equal - sized cells . this generation process is applied recursively until all the cells have seed points no more than the predefined value . for each recursive loop , the maximum difference between the division levels of adjacent cells @xmath66 is enforced . for cells that have division level difference with the adjacent cells larger than @xmath81 , the higher level cell is subdivided into 4 equal - sized cells . fig.[qtreedes1 ] shows the initial quadtree grid of the example in fig.[qtreedes0 ] . initial quadtree grid of the example in fig . [ qtreedes0 ] . vertices with solid square markers are on the boundary , with square box markers are inside the domain , and without any markers are outside the domain . ] the initial quadtree grid shown in fig.[qtreedes1 ] does not conform to the boundary . those cells that have edges intersected with the boundary need to be identified and trimmed . by using the signed distance function , the locations of the vertices ( inside the domain , on the domain boundary or outside the domain as shown in fig.[qtreedes1 ] ) are identified based on the sign and value of the function . for edges containing two vertices with opposite signs , they are identified as the edges intersected with the boundary . for each of those edges , the intersection point with the boundary is computed . some quadtree cells could have vertices very close to the boundary in comparison with the lengths of their edges . after trimming , poorly shaped polygon cells with some edges much shorter than the others could be generated and may adversely affect the mesh quality . to avoid this situation , the vertices that are within a threshold distance away from the boundary are identified and then moved to their closest points on the boundary . in the present work , @xmath82 of the length of the cell edge ( based on the smallest cell attaching to the vertex ) is used as the threshold value . the edges connecting to these vertices will no longer be cut by the boundary . the trade - off of this process is the presence of additional non - square cells that lead to additional computation of the stiffness matrix . this is discussed in section[sub : an - efficient - assembly ] . model curved boundary by quadtree refinement or using high - order elements . nodes are represented with small circles along the cell edges . ] at the end of the trimming process , the edges of a cell cut by the boundary are updated with the intersection points and the enclosed segment of boundary is added to the cell . this will result in polygon cells . after trimming the quadtree in fig.[qtreedes1 ] , the polygon cells around the hole of the example problem is shown in fig.[qtreedes2 ] . it is clear from fig.[qtreedes2 ] that the circular boundary is not represented accurately if a single linear element is used on the edge of the cell . in order to represent the curved boundary more accurately , two alternates are available in the developed algorithm . the first is to reduce the element size ( @xmath83-refinement ) . this is achieved by increasing the number of seed points on the curved boundary . fig.[qtreedes3 ] shows the initial quadtree layout of the example problem after increasing the seed points around the hole by 4 times . it can be seen by comparing fig.[qtreedes1 ] with fig . [ qtreedes3 ] that the refinement is limited to a small region around the hole . the refined quadtree ( fig[qtreedes2 ] ) demonstrates the improvement of capturing the circular boundary . quadtree mesh after refinement ] the second option to improve the modelling of curved boundaries is to utilise high - order elements ( @xmath84-refinement ) . fig.[qtreedes2 ] shows the example problem with each line segment on the circular boundary modelled with a 4th order element . with this approach , curved boundaries can be captured more accurately using fewer elements . both options to improve the modelling of the boundaries can be applied simultaneously without conflicts . the numerical accuracy of both approaches is discussed through numerical examples given in section[sec : numerical - examples ] . owing to the capability of the sbfem for fracture analysis @xcite , the domain containing a crack tip is modelled with a single cell . in the stress solution , the variation along the radial direction , including the stress singularity , is given analytically and the variation along the circumference of the cell is represented numerically by the line elements on boundary . to obtain accurate results , sufficient nodes have to present on the boundary of the cell to cover the angular variation of the solution . in the developed algorithm , the size of a cell containing a crack tip is controlled , as shown in fig.[qtreedes8 ] with an inclined crack , by a predefined set of seed points on a circle . quadtree mesh for a crack problem before and after merging cells . the two crack tips are marked with a cross . the two circles are to control the size of quadtree cells covering the crack tips . ] for problems with cracks , only one additional step is required after the initial mesh is generated . the cells surrounding the crack tip are refined to the same division level and then merged into a single cell as shown in fig.[qtreedes8 ] . this step avoids having a crack tip too close to the edges of the cell , which could affect the mesh quality and the solution accuracy @xcite . after the cells are merged , the intersection point between the edge of the resulting cell and the crack is computed to define the two crack mouth points . the other cells on the crack path are split by the crack into two cells . the splitting process is similar to the trimming of cells by the boundary , but two vertices are created at every intersection point between the cell edge and the crack to split the original cells . the global stiffness matrix is simply the assembly of the stiffness matrices of each master quadtree and polygon cell . when the @xmath0 rule is enforced to the mesh , only 6 main types of master quadtree cells are present as given in fig.[fig : qtreecell ] . by rotating the geometry of the master cells orthogonally , the maximum number of types of these master quadtree cells are 24 . for isotropic homogeneous materials , rotation does not have effect on 4-node or 8-node cells and only two 2 rotations are required for the first type of 6-node cell ( the top one in fig.[fig : qtreecell ] ) . the maximum number of master quadtree cells that require stiffness matrix calculation reduces to 16 ( only 15 in @xcite as 4-node cell is excluded ) . after the mesh generation , the algorithm will check which master cells out of the 16 appear in the mesh . their stiffness matrices are then computed and stored . during the stiffness assembling process , the stiffness matrix of each regular quadtree cell is directly extracted from those computed stiffness matrices . for the polygon cells and those irregular quadtree cells ( with their vertices moved to fit the boundaries ) , individual stiffness matrix calculation is required . this approach clearly improves the computational efficiency of constructing the global stiffness matrix , especially for large scale problems that contain a significant number of cells . with the use of high - order elements in the quadtree mesh , this assembling approach becomes even more economical . this section presents five numerical examples to highlight the capability and the performance of the proposed technique . in the first example , an infinite plate with a circular hole is modelled and the results are compared with the analytical solution . the proposed technique is then used to analyse a square plate with multiple holes to highlight the automatic meshing capability in handling transition between geometric features . in the third example , a square plate with a central hole and multiple cracks is studied to demonstrate the performance of the proposed technique in handling complicated geometries with singularities . thereafter , a square plate with two cracks cross each other is analysed . it is aimed to emphasise the automation and simplicity of the mesh generation in the proposed technique . in the first four examples , the same material properties , with young s modulus @xmath85 and poisson s ratio @xmath86 , are used . the final example is a cracked nuclear reactor under internal pressure . it is aimed to show the simplicity of the present technique in modelling practical non - regular structures . the computation time reported in this section is based on a desktop pc with intel(r ) core(tm ) i7 3.40ghz cpu and 16 gb of memory . the proposed technique is implemented in matlab and the computation time is extracted in interactive mode of matlab . an infinite plate containing a circular hole with radius @xmath87 at its centre is considered in this example . the plate is subject to a uniaxial tensile load as shown in fig.[openhole ] . the analytical solution of the stresses in polar coordinates @xmath88 is given by @xcite : @xmath89 the displacement solutions are : @xmath90\nonumber \\ u_{2}(r,\theta ) & = \frac{a}{8\mu}\left[\frac{r}{a}(\kappa-3)\sin\theta+\frac{2a}{r}\left((1-\kappa)\sin\theta+\sin3\theta\right)-\frac{2a^{3}}{r^{3}}\sin3\theta\right],\label{eq : exohdisp}\end{aligned}\ ] ] where @xmath91 is the shear modulus and @xmath92 is the kolosov constant for plane stress condition . the problem is solved by analysing a finite dimension of the plate with a dimension of @xmath93 ( see fig.[openhole ] ) . analytical traction ( eq.[eq : exohstr ] ) is applied at the four edges of this finite plate . fig.[ohm1 ] shows the quadtree mesh of the plate for @xmath94 . each edge on a quadtree cell is discretised with 1st order line elements . the @xmath0 rule is enforced . based on the proposed technique , the curved boundary is handled as shown in fig.[ohm2 ] with polygon cells . convergence study is conducted based on the @xmath95refinement . three different element orders @xmath96 are investigated . fig.[openholecon ] shows the present results of the relative error in the displacement norm @xmath97 , with @xmath98 the analytical solution given in eq . and @xmath99 the solution computed by the proposed technique . the results show that all three types of elements have monotonic convergence . for higher order elements , more accurate results with similar number of dof are obtained and the convergence rate is also faster . is the element order and @xmath100 is the slope of the fitted line ] there are 37 out of 100 cells calculated for the stiffness matrices . among those 37 cells , 9 are regular quadtree cells and 28 are polygon cells surrounding the hole . for the remaining cells , their stiffness matrices are simply extracted from those 9 regular quadtree cells . to further demonstrate the accuracy of the proposed technique , @xmath101 along @xmath102 ( see fig.[openhole ] ) is plotted in fig.[openholesigmatheta ] using the mesh given in fig.[ohm1 ] with 4th order elements . it can be seen that the results of the proposed technique agree well with the analytical solution , which has @xmath103 at @xmath104 ( @xmath105 ) . for points away from @xmath104 , @xmath101 approaches 1 . the same infinite plate can be approximated by increasing the @xmath106 ratio . the application of quadtree mesh facilitates such a study . only the left and right sides of the plate are subjected to uniaxial in - plane tension stress @xmath107 . the element order used in this study is @xmath108 . the same mesh given in fig.[ohm1 ] is used for @xmath94 . the adaptive capability of quadtree mesh leads to the same mesh pattern for all @xmath106 ratios . fig.[ohm3 ] shows the cells around the hole for @xmath109 and it is exactly the same as the one shown in fig.[ohm2 ] . for @xmath109 , although there are 316 cells in total , only 37 cells are calculated for the stiffness matrices , which is the same as the previous study . the results of @xmath101 at @xmath104 with varying @xmath106 ratio are given in table[openholesigmathetavsratio ] . it is seen that the analytical solution ( @xmath103 ) is quickly approached when increasing the @xmath106 ratio . ] .normalised stress ( @xmath101 ) at @xmath104 of the thin square plate with a circular hole [ cols="^,^,^",options="header " , ] [ nreacttab1 ] this paper has presented a numerical technique to automate stress and fracture analysis using the sbfem and quadtree mesh of high - order elements . owing to the nature of the sbfem , the proposed technique has no specific requirement , such as deriving conforming shape functions or sub - triangulation , to handle quadtree cells with hanging nodes . high - order elements are used within each quadtree cell directly . the quadtree mesh generation is fully automatic and involves minimal number of user inputs and operation steps . boundaries are modelled with scaled boundary polygons and this allows the proposed technique to conform the boundary without excessive mesh refinement . the meshing algorithm is also applicable for problems with singularities . the use of quadtree mesh leads to an efficient approach to compute the global stiffness matrix . this facilitates the analysis that requires a significant number of cells using high - order elements . five numerical examples are presented to highlight the functionality and performance of the proposed technique . the present results show excellent agreement with analytical solutions and those computed by the fem . [ [ section ] ]	this paper presents a technique for stress and fracture analysis by using the scaled boundary finite element method ( sbfem ) with quadtree mesh of high - order elements . the cells of the quadtree mesh are modelled as scaled boundary polygons that can have any number of edges , be of any high orders and represent the stress singularity around a crack tip accurately without asymptotic enrichment or other special techniques . owing to these features , a simple and automatic meshing algorithm is devised . no special treatment is required for the hanging nodes and no displacement incompatibility occurs . curved boundaries and cracks are modelled without excessive local refinement . five numerical examples are presented to demonstrate the simplicity and applicability of the proposed technique . scaled boundary finite - element method ; quadtree mesh ; high order elements ; polygon elements
bose - einstein condensation and superfluidity for cold nonrelativistic atoms can be experimentally investigated in systems of various dimensions @xcite . two dimensional systems can be achieved by building asymmetric traps , resulting in different characteristic sizes for one `` transverse extension '' @xmath5 and two `` longitudinal extensions '' @xmath0 of the atom cloud @xcite . for @xmath6 the system behaves effectively two - dimensional for all modes with momenta @xmath7 . from the two - dimensional point of view , @xmath5 sets the length scale for microphysics it may be as small as a characteristic molecular scale . on the other hand , the effective size of the probe @xmath0 sets the scale for macrophysics , in particular for the thermodynamic observables . two - dimensional superfluidity shows particular features . in the vacuum , the interaction strength @xmath1 is dimensionless such that the scale dependence of @xmath1 is logarithmic @xcite . the bogoliubov theory with a fixed small @xmath1 predicts at zero temperature a divergence of the occupation numbers for small @xmath8 , @xmath9 @xcite . in the infinite volume limit , a nonvanishing condensate @xmath10 is allowed only for @xmath11 , while it must vanish for @xmath12 due to the mermin - wagner theorem @xcite . on the other hand , one expects a critical temperature @xmath2 where the superfluid density @xmath13 jumps by a finite amount according to the behavior for a kosterlitz - thouless phase transition @xcite . we will see that @xmath3 ( with @xmath14 the atom - density ) vanishes in the infinite volume limit @xmath15 . experimentally , however , a bose - einstein condensate can be observed for temperatures below a nonvanishing critical temperature @xmath2 at first sight in contradiction to the theoretical predictions for the infinite volume limit . a resolution of these puzzles is related to the simple observation that for all practical purposes the macroscopic size @xmath0 remains finite . typically , there will be a dependence of the characteristic dimensionless quantities as @xmath16 , @xmath3 or @xmath1 on the scale @xmath0 . this dependence is only logarithmic . while @xmath17 , @xmath18 , @xmath19 , in accordance with general theorems , even a large finite @xmath0 still leads to nonzero values of these quantities , as observed in experiment . an appropriate tool for studying the scale dependence of various quantities is the functional renormalization for the average action @xcite . within a two - dimensional context , this starts with a given microphysical or classical action at some ultraviolet momentum scale @xmath20 . then a scale dependent effective action @xmath21 is defined by integrating out successively the modes with momenta @xmath22 larger than some infrared cutoff scale @xmath23 . the flow of @xmath21 with @xmath23 obeys an exact renormalization group equation , which we solve approximately by truncation of the most general form of @xmath21 . when @xmath23 reaches a scale @xmath24 , all fluctuations are included since no larger wavelength are present in a finite size system . the experimentally relevant quantities and the dependence on @xmath0 can be obtained from @xmath25 . this paper investigates the functional renormalization group for a nonrelativistic bose gas in two spatial dimensions . a priori , this method is not restricted to a perturbative treatment for a small coupling @xmath1 and can therefore be used to investigate the effects of fluctuations beyond the bogoliubov approximation . we assume that the interaction can be approximated by a contact term . this should be appropriate especially for dilute gases where the details of the interaction are not relevant . more explicitly our microscopic action reads @xmath26=\int_x \,{\big \{}\phi^\,(\partial_\tau-\delta-\mu)\,\phi\,+\,\frac{1}{2}\lambda(\phi^\phi)^2{\big \}}. \label{microscopicaction}\ ] ] we work with the functional integral within the matsubara formalism . the `` euclidean time '' @xmath27 parameterizes a torus with circumference @xmath28 , and the bosonic fields @xmath29 are complex , depending on the two - dimensional coordinates @xmath30 and @xmath27 , @xmath31 . in addition to the temperature @xmath32 and the chemical potential @xmath33 the action involves the dimensionless coupling @xmath1 as a parameter . we use @xmath34 and units of energy where the mass of the atoms is set to @xmath35 , @xmath36 . all macroscopic properties following from the microscopic model can be obtained from the full effective action @xmath37 $ ] . for example , the thermodynamic properties can be inferred from the grand canonical partition function @xmath38 or the corresponding grand potential @xmath39 . the latter is related to the effective action by @xmath40=\phi_g / t,\ ] ] where the expectation value @xmath41 is obtained as a solution of the field equation @xmath42=0.\ ] ] since the effective action is the generating functional of the one particle irreducible ( 1pi ) correlation functions we can also derive dynamical properties from @xmath37 $ ] . for a homogeneous field @xmath41 the full propagator @xmath43 is obtained from @xmath44{\bigg \|}_{\phi=\phi_\text{eq}}\\ & = & ( g^{-1})_{ij}(p_1)\delta(p_1-p_2).\end{aligned}\ ] ] here we employ a basis with two real fields , @xmath45 , @xmath46 , @xmath47 . the calculation of the functional @xmath37 $ ] is not an easy task it includes the summation of infinitely many diagrams in perturbation theory . our method determines @xmath37 $ ] by using an exact flow equation for the scale dependent effective action @xmath48$] the average action . due to the presence of an infrared cutoff at the scale @xmath23 only the quantum and thermal fluctuations with momenta @xmath49 are included . in the infinite volume limit @xmath21 interpolates between the microscopic action and the full effective action @xmath50 & \rightarrow s[\phi ] & \quad ( k\rightarrow \lambda),\\ \gamma_{k}[\phi ] & \rightarrow \gamma[\phi ] & \quad ( k\rightarrow 0).\end{aligned}\ ] ] for a system with finite size @xmath0 we are interested in @xmath25 , @xmath51 . if statistical quantities for finite size systems depend only weakly on @xmath0 , they can be evaluated from @xmath25 in the same way as their thermodynamic infinite volume limit follows from @xmath52 . details of the geometry etc . essentially concern the appropriate factor between @xmath53 and @xmath54 . the average action obeys the exact flow equation @xcite @xmath55=\frac{1}{2}\text{tr}(\gamma_k^{(2)}[\phi]+r_k)^{-1}\partial_k r_k . \label{eqflowequation}\ ] ] despite its simple one loop structure it includes all orders in perturbation theory as well as nonperturbative effects . nevertheless , in praxis this equation is not solvable analytically since it is a differential equation for a functional . approximate solutions can be found by truncating the space of possible functionals @xmath48 $ ] to a manageable size . such a truncation does not have to rely on an expansion in a small parameter , as it it the case for perturbation theory . because of this , the method is nonperturbative and also applicable for large interaction strength . in this work we use a very simple truncation which includes terms with up to two derivatives @xmath56 here the effective potential @xmath57 contains no derivatives and is a function of @xmath58 . at the end we will evaluate all quantities for a fixed chemical potential @xmath59 the term @xmath60 will be employed for an easy evaluation of @xmath61 , as needed for the density . in addition to the couplings parametrizing @xmath57 ( see below ) our truncation contains three further @xmath23-dependent ( `` running '' ) couplings @xmath62 , @xmath63 and @xmath64 . the particular form of our truncation is motivated by a more systematic derivative expansion and an analysis of symmetry constraints ( ward identities ) in ref . ( this reference may also be consulted for details of the formalism and the flow equations . ) a truncation similar to eq . has been investigated at zero temperature in refs . a simpler truncation was used in ref . @xcite . in terms of renormalized fields @xmath65 , @xmath66 , renormalized kinetic coefficients @xmath67 , @xmath68 and effective potential @xmath69 , eq . reads @xmath70 for the effective potential , we use an expansion around the @xmath23-dependent location of the minimum @xmath71 , and the @xmath23-independent value of the chemical potential @xmath59 that corresponds to the physical particle number density @xmath14 . we determine @xmath71 and @xmath59 by the requirements @xmath72 more explicitly , we take a truncation for @xmath73 of the form @xmath74(\rho-\rho_0)\\ & & + \frac{1}{2}\left[\lambda+\beta(\mu-\mu_0)\right](\rho-\rho_0)^2.\end{aligned}\ ] ] in the symmetric phase we have @xmath75 , while in the phase with spontaneous symmetry breaking , we have @xmath76 . the atom density @xmath77 corresponds to @xmath78 in the limit @xmath79 . in total , we have at fixed chemical potential @xmath80 and fixed temperature @xmath32 four running renormalized couplings @xmath71 , @xmath81 , @xmath82 , and @xmath83 . ( in the symmetric phase @xmath13 is replaced by @xmath84 . ) in addition , we need the anomalous dimension @xmath85 . to determine the flow of the density @xmath78 , we also need the dependence of the effective potential on @xmath86 and we therefore include the couplings @xmath87 , @xmath88 . the precise procedure how the flow equation for the average action is projected to flow equations for the running couplings is explained in ref . the flow equation for the effective potential @xmath89 and the kinetic coefficients @xmath90 and @xmath91 as well as the anomalous dimension @xmath92 are given there for arbitrary dimension @xmath93 . in this paper we apply these flow equations for @xmath94 . it is well known that in two dimensions the scattering properties can not be determined by a scattering length as it is the case in three dimensions . in experiments where a tightly confining harmonic potential restricts the dynamics of a bose gas to two dimensions , the interaction strength has a logarithmic energy dependence in the two - dimensional regime @xcite . for low energies and in the limit of vanishing momentum the scattering amplitude vanishes . in our formalism this is reflected by the logarithmic running of the interaction strength @xmath81 in the vacuum , where both temperature and density vanish . in general , the flow equations in vacuum describe the physics of few particles like for example the scattering properties or binding energies . following the calculation of ref . @xcite , we find the flow equations for the interaction strength ( @xmath95 ) @xmath96 since in vacuum the propagator is not renormalized , @xmath97 , we set @xmath98 and @xmath99 on the right - hand side of eq . . the vacuum corresponds to @xmath76 @xcite and we obtain the flow equation @xmath100 the vacuum flow is purely driven by quantum fluctuations . it will be modified by the thermal fluctuations for @xmath101 and for nonzero density @xmath14 . the solution of eq . @xmath102 goes to zero logarithmically for @xmath103 , @xmath104 . in contrast to the three - dimensional system , the flow of the interaction strength @xmath1 does not stop in two dimensions . to relate the microscopic parameter @xmath105 to experiments exploring the scattering properties , we have to choose a momentum scale @xmath106 , where experiments are performed . to a good approximation the relevant interaction strength can be computed from eq . by setting @xmath107 . if not specified otherwise , we will use a renormalized coupling @xmath108 . for our calculation we also have to use a microscopic scale @xmath109 below which our approximation of an effectively two dimensional theory with pointlike interaction becomes valid . our two - dimensional computation only includes the effect of fluctuations with momenta smaller then @xmath109 . in experiments @xmath110 is usually given either by the range of the van der waals interaction or by the length scale of the potential that confines the system to two dimensions . we choose in the following @xmath111 at this stage the momentum or length units are arbitrary , but we will later often choose the density to be @xmath112 , so that we measure length effectively in units of the interparticle spacing @xmath113 . for typical experiments with ultracold bosonic alkali atoms one has @xmath114 . the flow of @xmath81 for different initial values @xmath105 is shown in fig . [ figflowoflambda ] . at zero temperature and density for different initial values @xmath115 , @xmath116 , @xmath117 , @xmath118 , @xmath119 , and @xmath120 ( from top to bottom ) . ] following the flow from @xmath109 to @xmath53 yields the dependence of @xmath108 on @xmath105 as displayed in fig . [ figinteraction ] . at the macroscopic scale @xmath121 in dependence on the microscopic interaction strength @xmath105 at @xmath122 ( solid ) . the upper bound @xmath123 is also shown ( dashed ) . ] it follows from eq . that for positive initial values @xmath105 the interaction strength @xmath1 is bounded by @xmath124 the last equation holds for our choice of @xmath109 and @xmath53 . we emphasize that our bound holds only if the interactions are approximately pointlike for all momenta below @xmath109 . close to a feshbach resonance this may not be true and our formalism would need to be extended by considering nonlocal interactions or introducing an additional field for the exchanged two - atom state in the `` closed channel '' . in other words , close to a feshbach resonance the effective cutoff @xmath109 for the validity of a two - dimensional model with pointlike interactions may be substantially lower , and the upper bound on @xmath1 correspondingly higher . in this section we investigate the many body problem for a nonvanishing density @xmath14 at zero temperature @xmath11 . a crucial new ingredient as compared to the vacuum is the nonzero superfluid density @xmath125 for interacting bosons at zero temperature the density @xmath14 and the superfluid density @xmath126 are equal . in contrast , the condensate density is given by the unrenormalized order parameter @xmath127 due to the repulsive interaction , @xmath128 may be smaller than @xmath14 , the difference @xmath129 being the condensate depletion . ( in the limit @xmath130 we have @xmath131 . ) to obtain a nonzero density at temperature @xmath11 we have to go to positive chemical potential @xmath132 . at the microscopic scale @xmath133 the minimum of the effective potential @xmath134 is then at @xmath135 . the superfluid density @xmath13 is connected to a nonvanishing `` renormalized order parameter '' @xmath136 , with @xmath137 . it is responsible for an effective spontaneous breaking of the u(1)-symmetry . indeed , the expectation value @xmath136 points out a direction in the complex plane so that the global u(1)-symmetry of phase rotations is broken by the ground state of the system . goldstones theorem implies the presence of a gapless goldstone mode , and the associated linear dispersion relation @xmath138 accounts for superfluidity . the goldstone physics is best described by using a real basis in field space by decomposing the complex field @xmath139 . without loss of generality we can choose the expectation value @xmath136 to be real . the real fields @xmath140 and @xmath141 describe then the radial and goldstone mode . respectively . for @xmath80 the inverse propagator reads in our truncation @xmath142 here @xmath143 is the momentum of the collective excitation , and for @xmath11 the frequency obeys @xmath144 . in the regime with spontaneous symmetry breaking , @xmath145 , the propagator for @xmath146 has @xmath147 , @xmath148 , giving rise to the linear dispersion relation characteristic for superfluidity . this strongly modifies the flow equations as compared to the vacuum flow equations once @xmath149 . for @xmath150 the flow is typically in the regime with @xmath151 . in practice , we have to adapt the initial value @xmath152 such that the flow ends at a given density @xmath153 . for @xmath154 one finds that @xmath155 depends only very little on @xmath53 . as mentioned above , we will often choose the density to be unity such that effectively all length scales are measured in units of the interparticle distance @xmath113 . in contrast to the vacuum with @xmath156 , the flow of the propagator is nontrivial in the phase with @xmath157 and spontaneous u(1 ) symmetry breaking . in fig . [ figflowkinetic ] we show the flow of the kinetic coefficients @xmath62 , @xmath91 , @xmath90 for a renormalized or macroscopic interaction strength @xmath158 . ( solid ) , @xmath90 ( dashed ) , and @xmath91 ( dashed - dotted ) at zero temperature @xmath11 , density @xmath112 , and vacuum interaction strength @xmath158 . ] the wavefunction renormalization @xmath62 increases only a little at scales where @xmath159 and saturates then to a value @xmath160 . as will be explained below , we can directly infer the condensate depletion from the value of @xmath62 at macroscopic scales . the coefficient of the linear @xmath27-derivative @xmath90 goes to zero for @xmath161 . the frequency dependence is then governed by the quadratic @xmath27-derivative with coefficient @xmath91 , which is generated by the flow and saturates to a finite value for @xmath161 . the flow of the interaction strength @xmath81 for different values of @xmath108 is shown in fig . [ figflowlambda ] . at zero temperature @xmath11 , density @xmath112 , for different initial values @xmath105 . the dotted lines are the corresponding graphs in the vacuum @xmath162 . the vertical line labels our choice of @xmath53 . the lower plot shows @xmath163 for the same parameters , demonstrating that @xmath164 for small @xmath23 . ] while the decrease with the scale @xmath23 is only logarithmic in vacuum , it becomes now linear @xmath164 for @xmath165 . it is interesting that the ratio @xmath163 reaches larger values for smaller values of @xmath105 . as @xmath23 is lowered from @xmath109 to @xmath53 , the renormalized order parameter or the superfluid density @xmath13 increases first and then saturates to @xmath166 this is expected since the superfluid density equals the total density at zero temperature . in contrast , the bare order parameter @xmath167 flows to a smaller value @xmath168 . as argued in @xcite , the bare order parameter is just the condensate density , such that @xmath169 is the condensate depletion . its dependence on the interaction strength @xmath1 is shown in fig . [ figdepletion ] . as a function of the vacuum interaction strength @xmath1 . the dashed line is the bogoliubov result @xmath170 for reference . ] for small interaction strength @xmath1 the condensate depletion follows roughly the bogoliubov form @xmath171 however , we find small deviations due to the running of @xmath1 which is absent in bogoliubov theory . for large interaction strength @xmath172 the deviation from the bogoliubov result is quite substantial , since the running of @xmath1 with the scale @xmath23 is more important . we also investigate the dispersion relation at zero temperature . the dispersion relation @xmath173 follows from the condition @xmath174 where @xmath175 is the inverse propagator after analytic continuation to real time @xmath176 . as was shown in @xcite the generation of the kinetic coefficient @xmath91 by the flow leads to the emergence of a second branch of solutions of eq . . in our truncation the dispersion relation for the two branches @xmath177 and @xmath178 are @xmath179 in the limit @xmath180 , @xmath181 we find that the lower branch approaches the bogoliubov result @xmath182 while the upper branch diverges @xmath183 and thus disappears from the spectrum . the lower branch is dominated by phase changes ( goldstone mode ) , while the upper branch reflects waves in the size of @xmath13 ( radial mode ) . in principle , the coupling constants on the right - hand side of eq . also depend on the momentum @xmath184 . since an external momentum provides an infrared cutoff of order @xmath185 we can approximate the @xmath186-dependence by using on the right - hand side of eq . the @xmath23-dependent couplings with the identification @xmath187 . our result for the lower branch of the dispersion relation is shown in fig . [ figdispersionlinear ] . at temperature @xmath11 and for the vacuum interaction strength @xmath158 ( solid curve ) , @xmath188 ( upper dashed curve ) , and @xmath189 ( lower dashed curve ) . the units are set by the density @xmath112 . we also show the bogoliubov result for @xmath158 ( upper dotted curve ) and @xmath188 ( lower dotted curve ) . for @xmath189 the bogoliubov result is identical to our result within the plot resolution . ] we also plot the bogoliubov result @xmath190 for comparison . for small @xmath1 our result is in agreement with the bogoliubov result , while we find substantial deviations for large @xmath1 . both branches @xmath191 and @xmath192 are shown in fig . [ figdispersion ] on a logarithmic scale . , @xmath193 at temperature @xmath11 and for vacuum interaction strength @xmath158 ( solid ) , @xmath188 ( long dashed ) , and @xmath189 ( short dashed ) . the units are set by the density @xmath112 . ] since we start with @xmath99 at the microscopic scale @xmath109 we find @xmath194 for @xmath195 . the sound velocity @xmath196 can be extracted from the dispersion relation . more precisely , we compute the microscopic sound velocity for the lower branch @xmath178 as @xmath197 at @xmath198 . in our truncation we find @xcite @xmath199 our result for @xmath196 at @xmath11 is shown in fig . [ figsoundvelocity ] as a function of the interaction strength @xmath1 . as a function of the vacuum interaction strength ( solid ) . we also show the bogoliubov result @xmath200 for reference ( dashed ) . ] for a large range of small @xmath1 we find good agreement with the bogoliubov result @xmath201 . however , for large @xmath1 or result for @xmath196 exceeds the bogoliubov result by a factor up to 2 . at nonzero temperature and for infinite volume , long range order is forbidden in two spatial dimensions by the mermin - wagner theorem . because of that , no proper bose - einstein condensation is possible in a two - dimensional homogeneous bose gas at nonvanishing temperature . however , even if the order parameter vanishes in the thermodynamic limit of infinite volume , one still finds a nonzero superfluid density for low enough temperature . the superfluid density can be considered as the square of a renormalized order parameter @xmath202 and the particular features of the low - temperature phase can be well understood by the physics of the goldstone boson for a phase with effective spontaneous symmetry breaking @xcite . the renormalized order parameter @xmath136 is related to the expectation value of the bosonic field @xmath203 and therefore to the condensate density @xmath204 by a wave function renormalization , defined by the behavior of the bare propagator @xmath205 at zero frequency for vanishing momentum @xmath206 while the renormalized order parameter @xmath71 remains nonzero for @xmath161 if @xmath207 , the condensate density @xmath167 vanishes since @xmath62 diverges with the anomalous dimension , @xmath208 . after restoring dimensions the relation @xmath209 is the josephson relation @xcite . the strict distinction between a zero bose - einstein condensate @xmath210 and a nonzero superfluid density @xmath157 for nonzero temperature @xmath211 is valid only in the infinite volume limit of a homogeneous system . for a finite size of the system , as atoms in a trap , the running of @xmath212 is effectively stopped at some scale @xmath53 . there are simply no collective modes with wavelength larger than the size of the system , whose fluctuations would be responsible for a further increase of @xmath62 . with a finite @xmath213 both @xmath214 and @xmath13 are nonzero for @xmath207 , and the distinction between a bose - einstein condensate and superfluidity is no longer relevant in practice . for large systems @xmath215 can be large , however , such that the condensate density can be suppressed substantially as compared to the superfluid density . in any case , there is only one critical temperature @xmath2 , defined by @xmath216 . the flow equations permit a straightforward computation of @xmath217 for arbitrary @xmath32 , once the interaction strength of the system has been fixed at zero temperature and density . we have extracted the critical temperature as a function of @xmath108 for different values of @xmath53 . the behavior for small @xmath1 , @xmath218 is compatible with the free theory where @xmath2 vanishes for @xmath219 and with the perturbative analysis in ref . we find that the value of @xmath220 depends on the choice of @xmath53 . for @xmath121 we find @xmath221 , while @xmath222 corresponds to @xmath223 and @xmath224 to @xmath225 . in fig . [ figtcoflambda ] we show or result for @xmath3 as a function of @xmath1 for these choices . we also plot the curve in eq . with the monte - carlo result @xmath226 from ref . @xcite . as a function of the interaction strength @xmath1 . we choose here @xmath121 ( circles ) , @xmath222 ( boxes ) and @xmath224 ( diamonds ) . for the last case the bound on the scattering length is @xmath227 . we also show the curve @xmath228 ( dashed ) with the monte - carlo result @xmath226 @xcite for reference . ] we find that @xmath2 vanishes for @xmath229 in the interacting theory as well . this is due to the increase of @xmath220 and , for a fixed microscopical interaction , to the decrease of @xmath230 . since the vanishing of @xmath3 is only logarithmic in @xmath53 , a phase transition can be observed in practice . we find agreement with monte - carlo results @xcite for small @xmath1 if @xmath231 . the dependence of @xmath3 on the size of the system @xmath232 remains to be established for the monte - carlo computations . the critical behavior of the system is governed by a kosterlitz - thouless phase transition . usually this is described by considering the thermodynamics of vortices . in refs . @xcite it was shown that functional renormalization group can account for this `` nonperturbative '' physics without explicitly taking vortices into account . the correlation length in the low - temperature phase is infinite . in our picture , this arises due to the presence of a goldstone mode if @xmath157 . the system is superfluid for @xmath207 . the powerlike decay of the correlation function at zero frequency @xmath233 is directly related to the running of @xmath62 . as long as @xmath234 the bare propagator obeys approximately @xmath235 once @xmath236 , the effective infrared cutoff is given by @xmath237 instead of @xmath238 , and therefore @xmath212 gets replaced by @xmath239 , turning eq . into eq . . for large @xmath13 the anomalous dimension depends on @xmath13 and @xmath32 , @xmath240 . another characteristic feature of the kosterlitz - thouless phase transition is a jump in the superfluid density at the critical temperature . however , a true discontinuity arises only in the thermodynamic limit of infinite volume ( @xmath241 ) , while for finite systems ( @xmath242 ) the transition is smoothened . in order to see the jump , as well as essential scaling for @xmath32 approaching @xmath2 from above , our truncation is insufficient . these features become visible only in extended truncations that we will briefly describe next . for very small scales @xmath243 , the contribution of matsubara modes with frequency @xmath244 , @xmath245 , is suppressed since nonzero matsubara frequencies act as an infrared cutoff . in this limit a dimensionally reduced theory becomes valid . the long distance physics is dominated by classical two - dimensional statistics , and the time dimension parametrized by @xmath27 no longer plays a role . the flow equations simplify considerably if only the zero matsubara frequency is included , and one can use more involved truncations . such an improved truncation is indeed needed to account for the jump in the superfluid density . in ref . @xcite the next to leading order in a systematic derivative expansion was investigated . it was found that for @xmath246 the flow equation for @xmath13 can be well approximated by @xmath247 we switch from the flow equation in our more simple truncation to the improved flow equation for scales @xmath23 with @xmath248 . we keep all other flow equations unchanged . a similar procedure was also used in ref . @xcite . in fig . [ figflowofnrho ] we show the flow of the density @xmath14 , the superfluid density @xmath13 and the condensate density @xmath214 for different temperatures . ( solid ) , the superfluid density @xmath13 ( dashed ) , and the condensate density @xmath214 ( dotted ) for chemical potential @xmath249 , vacuum interaction strength @xmath188 and temperatures @xmath11 ( top ) , @xmath250 ( middle ) and @xmath251 ( bottom ) . the vertical line marks our choice of @xmath53 . we recall @xmath252 for @xmath11 such that the upper dashed and solid lines coincide . ] in fig . [ figrhodnoftemperature ] we plot our result for the superfluid fraction of the density as a function of the temperature for different scales @xmath53 . as a function of the dimensionless temperature @xmath253 for interaction strength @xmath188 at different macroscopic scales @xmath254 ( upper curve ) , @xmath255 , @xmath256 , @xmath257 , @xmath121 , @xmath258 ( bottom curve ) . we plot the result obtained with the improved truncation for small scales ( solid ) as well as the result obtained with our more simple truncation ( dotted ) . ( the truncations differ only for the three lowest lines . ) ] one can see that with the improved truncation the jump in the superfluid density is indeed found in the limit @xmath241 . [ figcondensatedensity ] as a function of the dimensionless temperature @xmath253 for interaction strength @xmath188 at macroscopic scale @xmath121 ( solid curve ) and @xmath222 ( dashed curve ) . for comparison , we also plot the superfluid density @xmath259 at @xmath121 ( dotted ) . these results are obtained with the improved truncation . ] shows the condensate fraction @xmath16 and the superfluid density fraction @xmath259 as a function of @xmath253 . we observe the substantial @xmath53 dependence of the condensate fraction , as well as an effective jump at @xmath2 for small @xmath53 . we recall that the infinite volume limit @xmath260 amounts to @xmath210 for @xmath12 . the kosterlitz - thouless description is only valid if the zero matsubara frequency mode ( @xmath162 ) dominates . for a given nonzero @xmath32 this is always the case if the the characteristic length scale goes to infinity . in the infinite volume limit the characteristic length scale is given by the correlation length @xmath261 . the description in terms of a classical two dimensional system with u(1 ) symmetry is the key ingredient of the kosterlitz - thouless description and holds for @xmath262 . in the infinite volume limit this always holds for @xmath207 or near the phase transition , where @xmath261 diverges or is very large . for a finite size system the relevant length scale becomes @xmath232 if this is smaller than @xmath261 . thus the kosterlitz - thouless picture holds only for @xmath263 . for very small temperatures @xmath264 one expects a crossover to the characteristic behavior near a quantum critical phase transition , governed by the quantum critical fixed point . the crossover between the different characteristic behaviors for @xmath263 and @xmath264 can be observed in several quantities . as an example we may take fig . [ figflowofnrho ] and compare the flow of @xmath13 and @xmath214 for low @xmath32 ( close to the @xmath11 curve ) or large @xmath32 ( other curves ) . in conclusion , we have demonstrated that functional renormalization for the average action yields a unified picture for ultracold nonrelativistic bosons in arbitrary dimension @xmath93 . we have employed in this paper the same simple truncation for @xmath94 as in ref . @xcite for @xmath265 . while for @xmath265 the perturbative bogoliubov treatment is reliable for a wide range of parameters , we find for @xmath94 substantial deviations if the coupling @xmath1 is large , or for the regime close to the phase transition . the kosterlitz - thouless type of the transition is nonperturbative in @xmath1 and can not be seen within the bogoliubov approximation . 1 . an upper bound on the interaction strength @xmath1 in vacuum arises if the characteristic momentum of the scattering particle @xmath184 is sufficiently below the ultraviolet scale @xmath109 below which a two - dimensional description with pointlike interactions becomes valid , @xmath266 the dependence of @xmath267 on @xmath268 is logarithmic , @xmath269 . the condensate depletion at @xmath11 is reasonably well approximated by the bogoliubov result if @xmath1 is small , @xmath270 . logarithmic corrections arise due to the running of @xmath1 . for large @xmath1 the repulsive interaction suppresses the condensate and the condensate depletion can exceed the bogoliubov result by a large factor . 3 . for the dispersion relation at @xmath11 we observe a substantial deviation from the bogoliubov result @xmath271 for large @xmath1 . we also find a second branch of `` radial excitations '' . the lower branch is dominated by phase fluctuations and shows the characteristic superfluid behavior @xmath272 for @xmath273 . the upper branch corresponds to excitations of the local superfluid density @xmath13 . the sound velocity at @xmath11 exceeds the bogoliubov result for large @xmath1 by up to a factor @xmath274 , while the perturbative result is valid for small @xmath1 . 5 . the system shows a second order phase transition . the critical temperature @xmath2 depends logarithmically on @xmath1 , @xmath275 . the coefficient @xmath220 depends logarithmically on the size of the system @xmath276 . for @xmath277 we find @xmath278 and agreement with monte - carlo simulations . the @xmath0-dependence of @xmath3 still needs to be established by numerical simulations . the phase transition is found to be of the kosterlitz - thouless type , with a jump of the superfluid density @xmath279 . for a finite size system this jump is smoothened . below @xmath2 , the correlation length is infinite . the correlation function shows a powerlike decay , with an anomalous dimension @xmath92 , @xmath280 . the anomalous dimension depends on the superfluid density @xmath13 and on the temperature @xmath240 . we have computed the superfluid density @xmath13 and the condensate density @xmath281 as a function of @xmath32 . the condensate density vanishes in the infinite volume limit @xmath15 , if @xmath12 . for finite @xmath0 a nonvanishing condensate can be observed . in this paper we have employed a rather simple truncation which is sufficient to show all the characteristic features of two - dimensional cold and dilute nonrelativistic bosons . one exception is the jump in the superfluid density which needs an improved truncation in the range of very small @xmath282 . extensions of the truncation are straightforward and should improve the quantitative accuracy for large @xmath1 . with increasing experimental precision the verification of a quantitatively precise theoretical computation will be a challenge . c. pethick and h. smith , _ bose - einstein condensation in dilute gases _ ( cambridge university press , cambridge , 2002 ) ; o. morsch and m. oberthaler , rev . * 78 * , 179 ( 2006 ) ; i. bloch , j. dalibard , and w. zwerger , _ ibid . _ * 80 * , 885 ( 2008 ) . a. grlitz _ et al . _ , phys . * 87 * , 130402 ( 2001 ) ; d. rychtarik , b. engeser , h. c. ngerl and r. grimm , phys . lett . * 92 * 173003 ( 2004 ) ; n. l. smith _ et al . _ , j. phys . b * 38 * , 223 ( 2005 ) ; s. burger _ et al . _ , europhys . 57 * , 1 ( 2002 ) ; m. khl _ et al . _ , j. low temp . phys . * 138 * , 635 ( 2005 ) ; c. orzel _ et al . _ , science * 291 * 2386 ( 2001 ) ; i. b. spielman , w. d. phillips , and j. v. porto , phys . * 98 * , 080404 ( 2007 ) ; z. hadzibabic _ et al . _ , * 93 * , 180403 ( 2004 ) ; z. hadzibabic _ et al . _ , nature ( london ) * 441 * , 1118 ( 2006 ) . v. l. berezinskii , sov . jetp * 32 * , 493 ( 1971 ) ; * 34 * , 610 ( 1972 ) ; j. m. kosterlitz and d. j. thouless , j. phys . c * 6 * , 1181 ( 1973 ) ; d. r. nelson and j. m. kosterlitz , phys . rev . * 39 * , 1201 ( 1977 ) . c. wetterich , phys . b * 301 * , 90 ( 1993 ) . j. berges , n. tetradis and c. wetterich , phys . rep . * 363 * , 223 ( 2002 ) ; c. bagnuls and c. bervillier , phys . rep . * 348 * , 91 ( 2001 ) ; b. delamotte , e - print arxiv : cond - mat/0702365 ; m. salmhofer and c. honerkamp , prog . 105 * , 1 ( 2001 ) . v. n. popov , _ functional integrals in quantum field theory and statistical physics _ ( reidel , dordrecht , 1983 ) ; d. s. fisher and p. c. hohenberg , phys . b * 37 * , 4936 , ( 1988 ) ; m. holzmann , g. baym , j. p. blaizot , and f. lalo , proc . * 104 * , 1476 ( 2007 ) . n. prokofev , o. ruebenacker and b. svistunov , phys . lett . * 87 * 270402 ( 2001 ) ; n. prokofev and b. svistunov , phys . rev . a * 66 * , 043608 ( 2002 ) ; s. sachdev and e. demler , phys . b * 69 * , 144504 ( 2004 ) ; s. sachdev , phys . b * 59 * , 14054 ( 1999 ) .	we investigate bose - einstein condensation for ultracold bosonic atoms in two - dimensional systems . the functional renormalization group for the average action allows us to follow the effective interactions from molecular scales ( microphysics ) to the characteristic extension of the probe @xmath0 ( macrophysics ) . in two dimensions the scale dependence of the dimensionless interaction strength @xmath1 is logarithmic . furthermore , for large @xmath0 the frequency dependence of the inverse propagator becomes quadratic . we find an upper bound for @xmath1 , and for large @xmath1 substantial deviations from the bogoliubov results for the condensate depletion , the dispersion relation and the sound velocity . the melting of the condensate above the critical temperature @xmath2 is associated to a phase transition of the kosterlitz - thouless type . the critical temperature in units of the density , @xmath3 , vanishes for @xmath4 logarithmically .
in this work we address the question of chaos in temperature in the sherringhton - kirkpatrick ( sk ) model . according to the parisi solution @xcite the sk model has many equilibrium states at a given temperature . these states are correlated : the overlap between different states can take a whole range of values from zero to the self - overlap @xmath1 . a natural and simple question is : what is the overlap between states at different temperatures ? according to the chaos in temperature hypothesis the overlap between any couples of equilibrium states at different temperatures is zero . consequently , the probability distribution of the overlap of configurations equilibrated at different temperatures @xmath2 should be a delta function centered at @xmath3 . one can be more quantitative considering the convergence to a delta function of the function @xmath4 corresponding to a finite system size @xmath5 . standard arguments tell us that for large @xmath5 this quantity is given by the large - deviations formula @xmath6 \label{pnq}\ ] ] the quantity @xmath7 is the free - energy cost , _ i.e. _ the difference between the free energy of a system of two copies at different temperatures of a spin - glass with the same realization of the disorder constrained to have a given value @xmath8 of the overlap , and the sum of the free energies at temperatures @xmath9 and @xmath10 ( actually here and in the following we call free energy " the opposite of the logarithm of the partition function , without the usual @xmath11 factors ) . consequently we can talk of a small or large chaos effect depending on the function @xmath12 ; if @xmath12 is small the convergence with @xmath5 of the @xmath4 will be slower and the chaos effect will be seen only when considering large - size systems . in particular if @xmath13 turns out to be zero for a finite range of values of the constraint there is no extensive chaos ; furthermore a null @xmath14 is expected if the @xmath2 is definitely different from a delta centered on zero . in the following we shall compute the quantity @xmath13 . in a recent paper @xcite , one of us reexamined the problem through a perturbative expansion in the replica framework , showing that chaos , if at all present , should be a very small effect . actually , this is a quite old problem @xcite , and has gained new attention in connection with the experimental effects of memory and rejuvenation ( see @xcite and references therein ) . recently some interesting results have been found for other models . sales and yoshino @xcite have convincingly shown that there is chaos in the directed polymer in random media ( dprm ) in 1 + 1 dimensions due to the interplay between energy and entropy , a mechanism which had been postulated in the droplet theory for finite dimensional spin - glass @xcite . on the other hand the existence of high correlations between systems at different temperatures has been proved in a well known class of mean - field spin - glass models @xcite . some evidence against chaos in the naive tap equations has been found in @xcite , a result which has been questioned in @xcite . powerful numerical efforts have been devoted by billoire and marinari ( bm ) to investigate chaos in temperature in the sk model @xcite ; even though they do nt actually see the effect , there is some evidence that it may be extremely weak , thus observable only on very large lattices . in the replica framework one obtains a variational expression for the free - energy functional @xmath15 of two systems with a constraint @xmath16 @xcite . we have expressed @xmath15 in terms of a parisi differential equation . the variational parameters are three @xmath17 matrices @xmath18 , @xmath19 , and @xmath20 : @xmath21 the function @xmath22 is defined through the following differential equation : @xmath23 with initial condition : @xmath24 + 2 e^{-\delta}\cosh[\beta_1y_1-\beta_2 y_2 ] ] \label{inif}\ ] ] @xmath25 notice that the initial condition depends on the difference @xmath26 . in order to maximize the free energy functional lagrange multipliers @xmath27 are used much as in the standard single system problem @xcite , then introducing the following quantities @xmath28 and the operator @xmath29 the saddle - point ( sp ) equations for the constrained systems read : @xmath30 @xmath31 @xmath32 the initial conditions for the functions @xmath33 , @xmath34 can be obtained from ( [ inif ] ) and ( [ defm ] ) ; for @xmath27 we find @xmath35 the presence of a constraint @xmath16 leads to the following equation @xmath36-e^{-\delta}\cosh[\beta_1y_1-\beta_2 y_2 ] } { e^{\delta}\cosh [ \beta_1 y_1+\beta_2 y_2]+ e^{-\delta}\cosh[\beta_1y_1-\beta_2 y_2]}\ .\label{eqq_c}\ ] ] if @xmath37 the initial condition ( [ inif ] ) separates and eq . ( [ eqq_c ] ) implies @xmath38 . finally , we recall that the derivative of the free energy with respect to the constraint can be expressed as @xmath39 when @xmath41 there are two possible situations depending on the value of the constraint @xmath16 . for all values of @xmath16 inside the support of the overlap distribution function @xmath2 we have solutions with @xmath37 and therefore @xmath42 . as a consequence for @xmath43 we have @xmath44 as it is to be expected since the function @xmath2 is non trivial . the solutions of the sp equations satisfy @xmath45 and are : @xmath46 these solutions were first proposed in @xcite for the truncated model . in @xcite it has been proven that they exist in any models with replica - symmetry breaking ( rsb ) by noticing that they are a permutation of the standard parisi solution . when @xmath16 lies outside the support of the function @xmath2 there are off - equilibrium solutions of the sp equations with @xmath47 and @xmath48 @xcite . these solutions can be obtained by solving the sp equations ( [ sp1],[sp2],[sp3 ] ) perturbatively in @xmath49 and @xmath50 . this has been done at the lower orders in @xcite where it was found that near the critical temperature the free - energy difference grows with @xmath51 in the off - equilibrium region . we pushed the computation to very high orders and we can safely claim that there are no @xmath52 terms in the expansion . recently this result was confirmed numerically @xcite . the off - equilibrium phase can be described also by considering what happens for @xmath53 rather then for @xmath54 . in the following section we will discuss this limit which can also be used to obtain another series expansion . in this case the only possible couples of configurations are those in which a configuration is coupled with itself . it is easy to realize that the thermodynamics of such a constrained system is equal to that of a single system at temperature @xmath56 . it is also natural that @xmath48 . by looking at equation ( [ eqq_c ] ) we see that @xmath55 requires @xmath57 . in this case the initial conditions on @xmath58 and @xmath59 read @xmath60\ ] ] by making the change of variables @xmath61 the sp equations ( [ sp1],[sp2],[sp3 ] ) reduce to those of a single system at temperature @xmath62 , as expected . the case @xmath41 has been recently discussed in @xcite . it is interesting to notice that the constrained model with @xmath55 has a critical temperature whose value is exactly twice as the critical temperature of the free model . therefore if we consider a couple of constrained systems such that @xmath63 , @xmath64 but @xmath65 we shall have a phase transition from a paramagnetic replica - symmetric ( rs ) phase to a rsb phase by changing the value of the constraint @xmath16 . for completeness we report the paramagnetic rs solution of the sp equations : @xmath66 notice that @xmath67 is divergent for @xmath68 ; therefore , according to ( [ pepsi ] ) , the derivative of the free - energy is divergent too . this result is valid also in the rsb phase . for @xmath69 , @xmath70 the critical value of the constraint @xmath71 turns out to be @xmath72 . for @xmath16 near @xmath73 the order parameter is small and we solved the equations perturbatively in @xmath74 . we also obtained a perturbative expansion for @xmath75 , @xmath76 and @xmath53 . indeed , in this region the three functions @xmath77 , @xmath78 and @xmath79 are small since the value @xmath55 corresponds to a free system at a temperature near @xmath80 . region is much smaller than in the figure . ] in @xcite an analytical picture for the absence of chaos in spin - glasses was proposed based on the possible existence of specific solutions of the sp equations ( [ sp1],[sp2],[sp3 ] ) . the structure of these solutions is depicted in figure ( [ figure1 ] ) : in the small-@xmath82 region the three functions @xmath83 and @xmath79 are all different till they reach the point @xmath84 where @xmath85 ; then for @xmath82 greater than @xmath84 @xmath79 remains constant while @xmath86 and @xmath87 after an intermediate plateau are connected continuously to the corresponding free parisi solutions . these solutions may exist for values of the constraint from zero to a maximum value where the two plateaus of the function at the higher temperature merge ( @xmath88 in figure ( [ figure1 ] ) ) . these solutions reduce to eq . ( [ soliso ] ) in the limit @xmath89 and are intrinsically non - chaotic due to the relation @xmath37 @xcite . however , their existence must be checked explicitly by attempting to solve the sp equations ( [ sp1],[sp2],[sp3 ] ) with the ansatz @xmath37 . in @xcite this has been checked up to the fifth order in a perturbative expansion near the critical temperature , _ i.e. _ it has been found that @xmath14 is a quantity smaller than the fifth order . recently it has been proven that these solutions exist in the spherical spin - glass models with multi-@xmath90 spin interaction @xcite . in that context it has been shown that they imply ultrametricity between the equilibrium states at different temperatures so that the parisi tree is essentially the same at all temperatures . furthermore , arguments have been advanced in order to show that the reference free energies of the clusters of states do not change with temperature , much as in the generalized random - energy model . perfect ultrametricity is a special property of those models due to the fact that the functions @xmath77 , @xmath78 and @xmath79 are all equal in the small-@xmath82 region . in general , however , we believe that the physical interpretation of the solutions proposed in @xcite is similar . indeed , we think that the non - chaotic solutions proposed in @xcite are the mathematical implementation of the old idea of a unique parisi tree which bifurcates when lowering the temperature plus the additional hypothesis that the ordering of the reference free energies of the clusters of states does not change when lowering the temperature . it would be interesting to check whether the logical equivalence between these solutions and the previous picture can be established explicitly through the cavity method as was done for the parisi solution @xcite . in order to deal with the sp equations ( [ sp1],[sp2],[sp3 ] ) of the sk model we resorted to a high - order computer - assisted perturbative expansion using the methods we recently applied to the parisi solution @xcite . the procedure is conceptually straightforward : one tries to solve the sp equations with the non - chaotic ansatz of figure ( [ figure1 ] ) assuming that near the critical temperature the functions @xmath77 , @xmath78 , @xmath79 are small and the @xmath82 regions where they vary are also small . therefore one expands all the quantities of interest in series of @xmath91 , @xmath82 , @xmath92 , @xmath93 and @xmath16 , ( @xmath94 ) ; the first is of order @xmath95 , the others are of order 1 ; the function @xmath96 , which is singular , can be treated considering its @xmath91 cumulants . this is essentially the same computation that was carried on in @xcite by hand up to the fifth order of the free - energy . it was found that up to that relatively high order the sp equations admit of a non - chaotic solution with @xmath37 for values of @xmath16 between @xmath97 and some maximum value near the self - overlap of the states at the higher temperature . instead we have found that at ninth order in the free energy it is impossible to find a solution of the sp equation ( [ sp1],[sp2],[sp3 ] ) with @xmath37 : _ the non - chaotic solutions do nt exist in the sk model_. accordingly there must be a free - energy cost in imposing the constraint , which one expects to be of ninth order . since off - equilibrium solutions must have @xmath47 , things become very complicated ; indeed , as explained in @xcite , when @xmath47 the three functions @xmath77 , @xmath78 and @xmath79 are coupled in the whole small-@xmath82 region , _ i.e. _ if one of them is varying in some interval of @xmath82 the others must vary too . we guess that off - equilibrium solutions have the same structure of the non chaotic solutions of figure ( [ figure1 ] ) plus small corrections of order @xmath98 , which is expected to be of the seventh order . therefore we have tried to solve the sp equations ( [ sp1],[sp2],[sp3 ] ) with a complicated ansatz made up of five @xmath82 regions on each of whom the functions @xmath77 , @xmath78 and @xmath79 have a different expansion . unfortunately , in spite of the very high number of parameters of this ansatz , we have not been able to solve the sp equations ( [ sp1],[sp2],[sp3 ] ) explicitly . therefore , in order to estimate the free - energy cost we plugged the same ansatz directly into equations ( [ f ] , [ eqf ] ) and extremized the free energy with respect to the variational parameters . we obtained the following estimate for the free - energy cost of constraining two copies of a sk spin - glass at temperature @xmath99 and @xmath100 to have a fixed overlap @xmath16 : @xmath101 as expected , this quantity turns out to be of the ninth order . let us comment that although in principle this must be considered an upper bound for the real value , it is most likely exact . indeed , in @xcite we observed that the order of the function @xmath102 relevant for the exact value of the free energy is much lower than the one expected by power counting arguments ; for instance if we maximize the free energy at order @xmath103 , we are able to determine the order parameter at an order near @xmath104 rather than the order @xmath105 obtained by power counting . therefore even if we know that our ansatz is unable to reproduce the correct solution at the seventh order , most likely this has no influence on the value of the free energy at the ninth order . , @xmath78 and @xmath79 obtained numerically for @xmath106 , @xmath107 and @xmath108 . the dotted lines represent the corresponding free solutions . within the numerical precision they coincide with those of figure ( [ figure1 ] ) ; the accuracy could be raised by the methods discussed in @xcite ] . [ figcris ] we also integrated the sp equations numerically _ via _ a pseudo - spectral code as we did in @xcite for the standard parisi solution . since the fields have three variables , time and memory required by the computation grow and the obtainable precision decreases , preventing us from seeing the chaos effect . the advantage of this method , however , is that we can investigate the form of the solutions in a non - perturbative way . to obtain the three functions @xmath77 , @xmath78 and @xmath79 , we applied the iterative scheme of @xcite to the sp equations ( [ sp1],[sp2],[sp3 ] ) ; we imposed at each iteration @xmath109 and we set by hand @xmath37 ; indeed @xmath98 is expected to be much smaller than the attainable precision . consistently with this assumption , the free - energy cost computed integrating ( [ f ] ) and ( [ eqf ] ) turned out to be zero within the precision . remarkably the three functions @xmath77 , @xmath78 and @xmath79 have precisely the structure proposed in @xcite ( cfr . figures ( [ figure1 ] ) and ( [ figcris ] ) ) . the chaos effect determines small corrections which should be seen at higher precision . when @xmath16 is greater than some @xmath110 the plateau of the function @xmath79 stops at @xmath110 irrespective of @xmath16 , and in order to obtain higher values of @xmath79 we should impose a finite value of @xmath98 . in the isothermal case this happens when one goes out of equilibrium for @xmath111 ; when @xmath112 this effect marks a change in the behaviour of @xmath14 which is much larger for @xmath113 ( see discussion in the conclusions ) . we believe that the chaos effect is beyond the attainable numerical precision of our code at the present , while improvements for the functions @xmath77 , @xmath78 and @xmath79 are at hand . we settled the longstanding problem of chaos in temperature in the sk model : _ chaos in temperature is present but it is an exceedingly small effect : @xmath14 is a ninth order quantity_. the analytical picture for absence of chaos in temperature proposed in @xcite does nt apply to the sk model , while it holds in the multi-@xmath90 spin spherical models @xcite ; this is an important difference between these two models whose behaviour has been often assumed to be similar . the effect is so weak that it ca nt be observed in present numerical simulations . indeed , if we make an estimate for the function @xmath4 through the equations ( [ pnq ] ) and ( [ df ] ) with parameters @xmath114 , @xmath115 and @xmath116 , which are the temperatures and size investigated in the latest numerical work by billoire and marinari ( bm ) @xcite , we find a completely flat function ; therefore we do nt expect to see anything similar to the delta peak corresponding to the @xmath117 limit . we notice that other corrections must be considered in order to explain those data ; indeed , according to our results , a formula like @xmath118 $ ] in any case would give a @xmath4 with an absolute maximum in @xmath3 . although we have not been able to determine exactly the off - equilibrium solutions corresponding to a non - zero constraint , we believe that they are similar to the non - chaotic solutions proposed in @xcite plus small correction of order @xmath119 ; this is supported by the numerical integration of the sp equations ( see fig . [ figcris ] ) . as a consequence some of the features of the non - chaotic models discussed in @xcite are valid in the sk model at finite @xmath5 . for instance , we expect that the support of the function @xmath4 has a maximum @xmath110 located near the self - overlap @xmath120 of the system at the higher temperature . in the spherical models this happens because @xmath44 for @xmath121 and @xmath122 for @xmath123 ; instead in the sk model @xmath110 marks a crossover in the behaviour of the function @xmath14 which is ninth order for @xmath124 while for @xmath125 is much larger , most likely third order as in the isothermal case . this is clearly seen in the numerical simulations of bm and also confirms their hypothesis that the approach to a delta function does nt proceed by a splitting of the peak at @xmath110 but rather by the emergence of a third peak in @xmath3 .	we prove the existence of chaos in temperature in the sherringhton - kirkpatrick model . the effect is exceedingly small , namely of the ninth order in perturbation theory . the equations describing two systems at different temperatures constrained to have a fixed overlap are studied analytically and numerically , yielding information about the behaviour of the overlap distribution function @xmath0 in finite - size systems .
one of the most important problem in quantum information theory is to determine the quantum capacity of a noisy quantum channel . the capacity is defined as the transmission rate optimized over all possible quantum error correcting codes such that decoding errors vanish in the limit of asymptotically many uses of the channel . mathematically , a quantum channel can be described by the trace preserving completely positive ( tpcp ) map from the input system to the output system . by using the stinespring dilation of the tpcp map , we can naturally define a complementary channel to an environment system , and we can regard the noisy quantum channel as a wire - tap channel @xcite from the sender to the legitimate receiver and the eavesdropper who can observe the environment system of the channel ( eg . see @xcite ) . then we can define the private capacity of the noisy quantum channel as the transmission rate optimized over all possible wire - tap codes such that decoding errors and information leakage vanish in the limit of asymptotically many uses of the channel . the private capacity and quantum capacity of noisy quantum channels were established in @xcite . however unlike the capacity formula of a classical noisy channel or the private capacity formula of a classical wire - tap channel , the private capacity and quantum capacity formulae are not single letter characterized , i.e. , the formulae involve the limit with respect to the number of channel uses , and they are not computable . indeed , some numerical evidences clarified that the expressions in the capacity formulae are not additive @xcite , and the single letter characterization is not possible in general at least by using the same expressions . a quantum channel is called degradable if there exists another degrading channel such that the conjunction of the channel to the legitimate receiver and the degrading channel coincide with the complementary channel to the eavesdropper . in such a case , the single letter characterizations of the private capacity and quantum capacity were established @xcite . a quantum channel is called conjugate degradable if there exists another degrading channel such that the conjunction of the channel to the legitimate receiver and the degrading channel coincide with the complementary channel to the eavesdropper up to complex conjugation . in such a case , the single letter characterizations were also established @xcite . to date , all quantum channel whose capacities are single letter characterized are degradable or conjugate degradable , and it is important to clarify a broader class of quantum channels such that the single letter characterizations are possible . aside from the possibility of the single letter characterizations , there is also another interesting problem . in the quantum information theory , the private information transmission and the quantum information transmission are closely related @xcite , and the possibility of the latter implies the possibility of the former . however , the private information transmission and the quantum information transmission are not exactly equivalent . indeed , although the private capacity and quantum capacity coincide for degradable quantum channels @xcite , the former can be strictly larger than the latter in general . especially the private capacity can be positive even if the quantum capacity is zero @xcite . thus it is important to clarify a condition on quantum channels such that the private capacity and quantum capacity coincide or not . to shed light on the above mentioned two problems , we introduce two new classes of quantum channels , which we call _ more capable _ and _ less noisy_. the more capable class consists of channels such that the quantum capacities of the complementary channels are zero . the less noisy class consists of channels such that the private capacities of the complementary channels are zero . later , these definitions turn out to be natural analogies of the partial orderings , more capable and less noisy , between classical channels @xcite . the inclusive relation of the degradable , the conjugate degradable , the less noisy , and the more capable classes are summarized in fig . [ fig : quantum - channel ] . in this paper , we show that the private capacity and quantum capacity coincide for channels in the more capable class . furthermore , we also show that the private capacity and quantum capacity can be single letter characterized for channels in the less noisy class . these results provide partial solutions to the above mentioned two problems . the rest of the paper is organized as follows . in section [ section : preliminaries ] , we review some known results on the private capacity and quantum capacity of quantum channels . in section [ section : main ] , the more capable and less noisy classes are introduced , and we state our main results . in section [ section : proof ] , we summarize certain properties implied by more capable and less noisy , and show proofs of our main results . we finalize the paper with conclusion in section [ section : proof ] . let @xmath0 be a quantum channel from an input system @xmath1 to an output system @xmath2 . by using the stinespring dilation ( eg . see @xcite ) , there exist an environment system @xmath3 and an isometry @xmath4 from @xmath1 to the joint system @xmath5 such that @xmath6\end{aligned}\ ] ] for every input @xmath7 , where @xmath8 is the partial trace with respect to the environment system . by using this representation , we can naturally define another channel @xmath9,\end{aligned}\ ] ] which is usually called the complementary channel of @xmath0 . although the stinespring dilation is not unique , the following arguments do not depend on the choice of the dilation because two dilations can be converted to each other by applying a local unitary to the environment systems . throughout the paper , we basically follow the notations from @xcite . the von neumann entropy of a density matrix @xmath7 is defined by @xmath10 , and the quantum relative entropy between @xmath7 and @xmath11 is defined by @xmath12 . for input state @xmath13 to the channel @xmath0 , the coherent information is defined by @xmath14 . when the input state is clear from the context , we omit the subscript and denote @xmath15 . the quantum mutual information of @xmath16 on the joint system is defined by @xmath17 . especially , when @xmath16 is classical with respect to @xmath18 , i.e. , @xmath16 is of the form @xmath19 then the quantum mutual information can be written as @xmath20 when the legitimate receiver can observe the output of @xmath0 and the eavesdropper can observe the output of @xmath21 , the private capacity is characterized by @xcite @xmath22 where @xmath23,\end{aligned}\ ] ] where @xmath24 are states ( not necessarily pure states ) on @xmath1 indexed by @xmath25 , and @xmath26 is a probability distribution on a finite set @xmath27 . on the other hand , when the sender want to transmit quantum information to the receiver through the channel @xmath0 , the quantum capacity is characterized by @xcite @xmath28 where @xmath29 both eqs . ( [ eq : private - capacity - formula ] ) and ( [ eq : quantum - capacity - formula ] ) can not be single letter characterized , i.e. , @xmath30 and @xmath31 in general @xcite . furthermore , the private capacity can be strictly larger than the quantum capacity , i.e. , @xmath32 for some channels @xcite . the channel @xmath0 is said to be degradable if there exists a tpcp map @xmath33 such that @xmath34 . this is a quantum analogue of degraded broadcast channel @xcite . when @xmath0 is degradable , then it is known @xcite that the single letter formulae hold , i.e. , @xmath35 and @xmath36 . furthermore , it is also known that @xmath37 for degradable channel @xmath0 @xcite . let @xmath38 denote entry - wise complex conjugation with respect to a fixed basis of @xmath3 . then , the channel @xmath0 is said to be conjugate degradable if there exists a tpcp map @xmath33 such that @xmath39 . when @xmath0 is conjugate degradable , it is known that @xmath36 @xcite . later , it will turn out that @xmath40 also holds . in this section , we introduce two new classes of quantum channels , and show our main results . the quantum channel @xmath0 is said to be _ more capable _ than the environment , or just more capable , if the quantum capacity of the complementary channel to the environment is zero , i.e. , @xmath41 the quantum channel @xmath0 is said to be _ less noisy _ than the environment , or just less noisy , if the private capacity of the complementary channel to the environment is zero , i.e. , @xmath42 since @xmath43 , less noisy implies more capable . by using the eigenvalue decomposition @xmath44 of @xmath45 , we can rewrite the coherent information as @xmath46 where we set @xmath47 . thus , by noting eq . ( [ eq : quantum - capacity - formula ] ) , the quantum channel @xmath0 being more capable can be also described as @xmath48 holds for every @xmath49 . furthermore , by noting eq . ( [ eq : private - capacity - formula ] ) , the quantum channel @xmath0 being less noisy can be also described as @xmath50 holds for every @xmath49 . ( [ eq : q - more - capable - alternative ] ) and ( [ eq : q - less - noisy - alternative ] ) resemble the definitions of more capable and less noisy for classical channels @xcite , and it is justified to call quantum channels satisfying eqs . ( [ eq : q - more - capable ] ) or ( [ eq : q - less - noisy ] ) more capable or less noisy . in @xcite , alternative description of less noisy , less divergence contracting , was introduced , and we can also extend such a description to the quantum channel . the quantum channel @xmath0 is said to be less divergence contracting if @xmath51 holds for every @xmath49 . later , we will show that the quantum channel is less noisy if and only if less divergence contracting ( proposition [ proposition : q - equivalence - less - noisy - divergence ] ) . this alternative description plays a crucial role when we prove theorem [ theorem : less - noisy ] . the followings are our main results . [ theorem : more - capable ] suppose that the quantum channel @xmath0 is more capable . then , we have @xmath52 [ theorem : less - noisy ] suppose that the quantum channel @xmath0 is less noisy . then , we have @xmath53 when @xmath0 is conjugate degradable , we can show that @xmath54 as follows . suppose that the sender sends a state @xmath55 that corresponds to the message @xmath56 , and the eavesdropper . ] uses a povm @xmath57 . then , for the entry - wise complex conjugate povm @xmath58 , we have @xmath59 & = & { \mathrm{tr}}[\bar{m}_i { \cal c}^{\otimes n } \circ { \cal n}_e^{\otimes n}(\rho_i ) ] \\ & = & { \mathrm{tr}}[m_i { \cal n}_e^{\otimes n}(\rho_i)],\end{aligned}\ ] ] where the last equality follows because @xmath60 and @xmath61 . thus , the legitimate receiver can get exactly the same information as the eavesdropper and the private information transmission to the eavesdropper is impossible . from this argument , conjugate degradable implies less noisy . when the quantum capacity of the channel is @xmath62 but it can be used to share bound entanglement , then the channel is called a binding entanlement channel @xcite . especially when the channel produces a positive partial transpose ( ppt ) bound entanglement , the channel is called ppt binding entanglement channel . if a complementary channel is a binding entanglement channel , then the main channel obviously belongs to the more capable class . since the complementary channel of the conjugate degradable channel can only produce ppt bipartite state @xcite , if there exists a conjectured negative partial transpose ( npt ) binding entanglement channel , the complementary of such a channel belongs to the more capable class but not to the conjugate degradable class . it is known that there exists a channel such that the quantum capacity is zero ( ppt binding entanglement channel ) but the private capacity is strictly positive @xcite . let the complementary channel @xmath21 be such a channel . then the channel @xmath0 is more capable but not less noisy , then the complementary channel @xmath21 is more capable . then , theorem [ theorem : more - capable ] implies that @xmath63 , which contradict with the fact @xmath64 . ] . thus , the more capable class is strictly broader than the less noisy class . however , it is not yet clear whether the less noisy class is strictly broader than the degradable or conjugate degradable classes . in this section , we summarize the properties of @xmath65 and @xmath66 when eqs . ( [ eq : q - more - capable - alternative ] ) or ( [ eq : q - less - noisy - alternative ] ) hold for @xmath67 . for @xmath68 , we can also show similar properties of @xmath69 and @xmath70 when eqs . ( [ eq : q - more - capable - alternative ] ) or ( [ eq : q - less - noisy - alternative ] ) hold for each @xmath71 by considering @xmath71 times extension of @xmath0 . the following properties can be regarded as quantum extensions of the properties shown for classical channels in the literatures @xcite [ proposition : q - more - capable - no - preprocessing ] suppose that eq . ( [ eq : q - more - capable - alternative ] ) holds for @xmath67 . then we have @xmath72 for any @xmath26 and @xmath24 , let @xmath73 be an eigenvalue decomposition . let @xmath74 be the random variable on @xmath75 such that @xmath76 then , we have @xmath77 } \nonumber \\ & & - [ i(\tilde{x};b\|u ) - i(\tilde{x } ; e\|u ) ] . \label{eq : q - more - capable - no - preprocessing - proof-1}\end{aligned}\ ] ] since eq . ( [ eq : q - more - capable - alternative ] ) holds for @xmath67 , we have @xmath78 for every @xmath79 , which means that the second bracket of eq . ( [ eq : q - more - capable - no - preprocessing - proof-1 ] ) is non - negative . furthermore , by noting that @xmath80 are pure , we have @xmath81 where @xmath82 since @xmath26 and @xmath83 are arbitrary , we have @xmath84 the opposite inequality is obvious from the definitions of @xmath65 , @xmath66 , and eq . ( [ eq : alternative - coherent ] ) . [ proposition : q - channel - preprocessing ] suppose that eq . ( [ eq : q - more - capable - alternative ] ) does not hold for @xmath67 , and the density operator @xmath85 maximizing @xmath15 is full rank . then , we have @xmath86 especially when @xmath87 and @xmath88 , the sufficient and required condition on @xmath89 is that eq . ( [ eq : q - more - capable - alternative ] ) holds for @xmath67 . next , we show the latter statement of the proposition . the sufficient condition follows from proposition [ proposition : q - more - capable - no - preprocessing ] . suppose that @xmath106 holds . since @xmath88 , there exists @xmath26 and @xmath83 such that @xmath107 which implies the required condition . next , we consider the case such that eq . ( [ eq : proof - q - strict - improvement-1 ] ) does not hold . in this case , we have @xmath108 then , since @xmath87 , @xmath92 must be full rank . thus , the required condition follows from the former statement of the proposition . [ proposition : concavity ] eq . ( [ eq : q - less - noisy - alternative ] ) holds for @xmath67 if and only if the coherent information is concave is degradable ( * ? ? ? * eq . ( 9.89 ) ) . ] , i.e. , @xmath109 where @xmath110 . let @xmath111 be an eigenvalue decomposition . then , let @xmath112 , @xmath113 , @xmath74 be the random variable on @xmath114 such that @xmath115 then , we have @xmath116 } \nonumber \\ & & - [ i(\tilde{x } ; b\|u ) - i(\tilde{x } ; e\|u ) ] . \label{eq : q - less - noisy - concavity - proof}\end{aligned}\ ] ] we also have @xmath117 and @xmath118 thus , from eq . ( [ eq : q - less - noisy - concavity - proof ] ) , eq . ( [ eq : q - less - noisy - alternative ] ) holds for @xmath67 if and only if the coherent information @xmath119 is concave . [ proposition : q - equivalence - less - noisy - divergence ] the following two conditions are equivalent . 1 . [ q - condition1 ] eq . ( [ eq : q - less - noisy - alternative ] ) holds for @xmath67 . [ q - condition2 ] eq . ( [ eq : q - less - divergence - contracting ] ) holds for @xmath67 . we first show that ( [ q - condition1 ] ) implies ( [ q - condition2 ] ) . for any @xmath13 , @xmath90 , and @xmath120 , let @xmath121 , @xmath122 , @xmath123 , @xmath124 , and @xmath125 . then , let @xmath126 where @xmath127 by elementary calculation ( cf . * ex . 1.4 ) ) , we have @xmath128 obviously , we have @xmath129 . since eq . ( [ eq : q - less - noisy - alternative ] ) holds for @xmath67 , @xmath130 for any @xmath131 . thus , @xmath132 must be non - negative , which means that eq . ( [ eq : q - less - divergence - contracting ] ) holds for @xmath67 . next , we show that ( [ q - condition2 ] ) implies ( [ q - condition1 ] ) . for any @xmath26 and @xmath83 , we have @xmath133 where @xmath134 since eq . ( [ eq : q - less - divergence - contracting ] ) holds for @xmath67 , we have @xmath135 it is a straight forward consequence of proposition [ proposition : q - more - capable - no - preprocessing ] . since @xmath0 is more capable , eq . ( [ eq : q - more - capable - alternative ] ) holds for every @xmath49 . thus , we have @xmath136 for every @xmath49 , and the statement of the theorem follows from eqs . ( [ eq : private - capacity - formula ] ) and ( [ eq : quantum - capacity - formula ] ) . @xmath137 since less noisy implies more capable , by theorem [ theorem : more - capable ] , it suffice to show @xmath36 . for this purpose , we will show that @xmath138 holds for every @xmath49 . for any input state @xmath139 on @xmath140 , let @xmath141 and @xmath142 be the partial traces . then , we have @xmath143 since eq . ( [ eq : q - less - noisy - alternative ] ) holds for @xmath144 , by ( @xmath71 times extension of ) proposition [ proposition : q - equivalence - less - noisy - divergence ] , eq . ( [ eq : q - less - divergence - contracting ] ) also holds for @xmath144 , which implies @xmath145 thus , eq . ( [ eq : goal - main - theorem ] ) can be proved by induction . @xmath137 in this paper , we introduced two new classes of quantum channels , which we call more capable and less noisy . for the more capable class , we showed that the private capacity and quantum capacity coincide . for the less noisy class , we showed that the private capacity and quantum capacity can be single letter characterized . our results shed light on further understanding of the private and quantum capacities of quantum channels . however , the conditions such that a certain channel belongs to the more capable class or the less noisy class are hard to verify in general , and we do not yet know whether there exists a channel that belongs to the less noisy class but not to the degradable or conjugate degradable classes , which is an important future research agenda . this research is partly supported by grand - in - aid for young scientists(b):2376033700 and grant - in - aid for scientific research(a):2324607101 . the author also would like to thank an anonymous reviewer for helpful comments .	two new classes of quantum channels , which we call _ more capable _ and _ less noisy _ , are introduced . the more capable class consists of channels such that the quantum capacities of the complementary channels to the environments are zero . the less noisy class consists of channels such that the private capacities of the complementary channels to the environment are zero . for the more capable class , it is clarified that the private capacity and quantum capacity coincide . for the less noisy class , it is clarified that the private capacity and quantum capacity can be single letter characterized .
higher - order unification is the problem given an equation @xmath0 containing free variables is there a solution substitution @xmath1 such that @xmath2 and @xmath3 have the same normal form ? the terms @xmath4 and @xmath5 are from the simply typed lambda calculus and the same normal form is with respect to @xmath6-equivalence . higher - order matching is the particular instance when the term @xmath5 is closed ; can @xmath4 be pattern matched to @xmath5 ? although higher - order unification is undecidable ( even if free variables are only second - order @xcite ) , higher - order matching was conjectured to be decidable by huet @xcite . if matching is decidable then it is known to have non - elementary complexity @xcite . decidability has been proved for the general problem up to order @xmath7 ( by showing decidability of observational equivalence of lambda terms ) and for various special cases @xcite . comon and jurski define tree automata that characterise all solutions to a @xmath7th - order problem , thereby , showing that they form a regular set @xcite . loader showed that matching is undecidable for the variant definition of the same normal form that only uses @xmath8-equivalence by encoding lambda definability as matching @xcite : see @xcite for a proof that uses the halting problem . an excellent source of information about unification and matching is @xcite . in this paper , we confirm huet s conjecture that higher - order matching is decidable . the proof first appeals to padovani s and schubert s reduction of matching to the ( dual ) interpolation problem @xcite and is then in the tradition described by dowek @xcite : `` these [ decidability ] proofs are rather technical @xmath9 because they all proceed by transforming potential solutions into smaller ones cutting and pasting term pieces '' . the proof method is partly inspired by model - checking games ( such as in @xcite ) where a model , a transition graph , is traversed relative to a property and players make choices at appropriate positions . given a ( dual ) interpolation problem @xmath10 , we define a game where the model is a closed lambda term @xmath4 in @xmath11-long normal form that is a potential solution to @xmath10 ; game - playing moves around the term @xmath4 ( dependent on @xmath10 ) . the game captures the dynamics of @xmath8-reduction on @xmath4 without changing it ( using substitution ) . unlike model - checking games , play may arbitrarily jump around a term because of binding . the principal virtue of the game is that small pieces of a solution term can be understood in terms of sequences of positions in the game and how they , thereby , contribute to solving the problem @xmath10 . we identify regions of a term , `` tiles '' , and classify them according to these intervals of play . two transformations that preserve solution terms are introduced . with these , we show that @xmath12rd - order matching is decidable using the small model property : if there is a solution to a problem then there is a small solution to it . in @xcite , we introduced the game and two more transformations that uniformly solved known subcases of matching ( including @xmath7th - order ) . the key observation for decidability at @xmath12rd - order is the tree - model property : each play descends a branch of a solution term because of the paucity of binding . for higher - orders , the idea is to induce as far as possible the tree - model property , to tame play jumping within a term due to binding . the mechanism for doing this involves unfolding a lambda term with respect to game playing which is analogous to unravelling a transition system in modal logic . unfolding involves `` tile lowering '' , copying regions of a term down its branches . the proof of decidability at higher - order uses unfolding and from its combinatorial properties the small model property follows . in section [ section2 ] we introduce higher - order matching and ( dual ) interpolation and in section [ prelim ] we define some basic ingredients of the problem . in section [ treecheck ] we describe the tree - checking game that characterises ( dual ) interpolation and in section [ props ] some properties of the resulting game are highlighted . in section [ tiles ] we identify tiles as regions of terms and define their plays and in section [ third ] we show that @xmath12rd - order matching is decidable using a tree model property of game playing ; the first step in this argument is to define a partition of each play in a game . the definition of partition is extended to all orders in section [ unfold ] . this then forms the basis for the notion of tile unfolding that is described in section [ fifth ] and how it leads to decidability of matching via the small model property . the complexity analysis is the size of a smallest term , if there is one , that solves a problem . however , the bounds are extremely coarse . finally , we conclude with general remarks and ideas for future work . simple types are generated from a single base type @xmath13 using the binary @xmath14 operator . for simplicity , we assume only one base type : everything that is to follow can be extended to arbitrary many base types . a type is @xmath13 or @xmath15 where @xmath16 and @xmath17 are types . if @xmath18 then it has the form @xmath19 , assuming @xmath14 associates to the right , which we abbreviate to @xmath20 following ong @xcite . a standard definition of _ order _ is : the order of @xmath13 is @xmath21 and the order of @xmath22 is @xmath23 where @xmath24 is the maximum of the orders of the @xmath25s . terms of the simply typed lambda calculus are built from a countable set of typed variables @xmath26 and constants @xmath27 ( so , each variable and constant has a unique type ) . the set of simply typed lambda terms is the smallest set @xmath28 such that 1 . if @xmath29 ( @xmath30 ) has type @xmath16 then @xmath31 @xmath32 , 2 . if @xmath33 and @xmath31 then @xmath34 , 3 . if @xmath35 and @xmath36 then @xmath37 . the _ order _ of a typed term @xmath38 is the order of its type @xmath16 . in a sequence of unparenthesised applications , we adopt the usual convention that application associates to the left ; so @xmath39 is @xmath40 . the usual definitions of free and bound variable occurrences and when a typed term is closed are assumed . moreover , we assume the standard definitions and properties of @xmath41-equivalence , @xmath8-reduction , @xmath11-reduction and @xmath42-equivalence , @xmath43 , such as strong normalisation of @xmath8-reduction : see for instance , barendregt @xcite . a _ matching problem _ is an equation @xmath44 where @xmath45 and @xmath5 is closed . the _ order _ of the problem is the maximum of the orders of the free variables @xmath46 in @xmath47 . solution _ is a sequence of terms @xmath48 such that @xmath49 where @xmath50 is the simultaneous substitution of @xmath51 for each free occurrence of @xmath52 in @xmath47 for each @xmath53 . the decision question is : given a matching problem , does it have a solution ? it suffices to consider the case when @xmath54 : a problem @xmath44 of type @xmath55 reduces to the equivalent problem @xmath56 of type @xmath13 where each @xmath57 is a fresh constant which is not allowed in solution terms . equivalent variants of matching include the `` range question '' and the `` pattern matching problem '' @xcite . as described by huet @xcite , every simply typed lambda calculus term is @xmath6-equivalent to a unique term in _ @xmath11-long normal form _ , 1 . if @xmath58 then it is @xmath59 where @xmath5 is a constant or a variable , or @xmath60 where @xmath61 is a constant or a variable and each @xmath62 is in @xmath11-long normal form , 2 . if @xmath63 then @xmath4 is @xmath64 where each @xmath65 and @xmath66 is in @xmath11-long normal form . throughout , we write @xmath67 for @xmath68 . a term is _ well - named _ if each occurrence of a variable @xmath69 within a lambda abstraction is unique . in the following , we assume that a term in _ normal form _ is always in @xmath11-long normal form ; consequently , @xmath8-equality and @xmath70-equality coincide ( for instance , see @xcite ) . assume @xmath59 and each @xmath71 , @xmath72 , is a closed term in normal form and @xmath73 . 1 . @xmath74 is an _ interpolation equation_. 2 . @xmath75 is an _ interpolation disequation_. 3 . a finite family of interpolation equations @xmath76 , where @xmath77 , with the same free variable @xmath29 , is an _ interpolation problem _ @xmath10 . 4 . a finite family of interpolation equations and disequations @xmath78 , when @xmath77 , with the same free variable @xmath29 and where each @xmath79 , is a _ dual interpolation problem _ @xmath10 . the type of problem @xmath10 is that of @xmath29 and the order of @xmath10 is the order of @xmath29 . a _ solution _ of @xmath10 of type @xmath16 is a closed term @xmath38 in normal form such that for each equation @xmath80 and , in the case of dual interpolation , for each disequation @xmath81 . we write @xmath82 if @xmath4 is a solution of @xmath10 . conceptually , ( dual ) interpolation is simpler than matching because there is a single variable @xmath29 that appears at the head of each ( dis)equation . however , schubert shows that a matching problem of order @xmath83 reduces to an interpolation problem of order at most @xmath84 and padovani shows it reduces to a dual interpolation problem of order @xmath83 , @xcite . consequently , the higher - order matching problem reduces to the following decision question . * decision question * given a ( dual ) interpolation problem @xmath10 , is there a term @xmath82 ? it is this question that is solved positively in the rest of the paper . throughout , we assume a fixed dual interpolation problem @xmath10 of type @xmath16 whose order is greater than @xmath21 ( as an order @xmath21 problem is easily decided ) . a problem @xmath10 has the form @xmath85 , @xmath86 , where the normal form terms @xmath87 and @xmath88 are well - named and no pair share bound variables . we start with some examples of interpolation problems . [ examp1 ] the following is a @xmath7th - order problem @xmath89 with @xmath90 and @xmath91 . [ examp2 ] the problem @xmath92 has order @xmath12 where @xmath29 has type @xmath93 and @xmath94 assuming @xmath95 . [ examp3 ] the next equation , due to luke ong , is @xmath96th - order @xmath97 with @xmath98 and @xmath99 . a right term @xmath5 of an interpolation ( dis)equation @xmath100 may contain bound variables , such as @xmath101 of example [ examp2 ] . let @xmath102 be the set of bound variables in @xmath5 and let @xmath103 be a fresh set of constants where each @xmath104 has the same type as @xmath52 . the _ ground closure _ of a closed term @xmath105 , whose bound variables belong to @xmath106 , with respect to @xmath107 , written cl@xmath108 , is defined inductively : 1 . if @xmath109 then cl@xmath110 , 2 . if @xmath111 then cl@xmath112 , 3 . if @xmath113 then cl@xmath114 . if @xmath115 then its ground closure with respect to @xmath116 is the set of terms @xmath117 . the ground closure of @xmath118 of example [ examp3 ] with respect to the empty set is its subterms @xmath119 . an element of a ground closure always has base type ; in the case of a right term @xmath5 of an interpolation ( dis)equation , its ground closure with respect to a set of constants consists of all subterms of type @xmath13 when free variables @xmath52 are replaced with constants @xmath104 of the same type . we also identify subterms of left terms @xmath120 of a ( dis)equation @xmath121 relative to the finite set @xmath107 of constants for @xmath5 . such a subterm may contain free variables and may have a higher - order type . the _ subterms _ of @xmath105 relative to @xmath107 , written sub@xmath122 , is defined inductively using an auxiliary set sub@xmath123 : 1 . if @xmath105 is a variable or a constant then sub@xmath122 @xmath124 sub@xmath123 @xmath124 @xmath125 , 2 . if @xmath105 is @xmath126 then sub@xmath122 @xmath124 sub@xmath123 @xmath124 @xmath127 , 3 . if @xmath105 is @xmath128 then sub@xmath122 @xmath124 sub@xmath123 @xmath124 @xmath129 , 4 . if @xmath105 is @xmath130 then sub@xmath122 @xmath124 @xmath131 and sub@xmath123 @xmath124 + @xmath132 . the subterms of @xmath133 relative to @xmath134 of example [ examp2 ] is @xmath135 . if instead of @xmath133 the left term of this example is @xmath136 , then sub@xmath137 is @xmath138 : bound variables directly beneath a constant are replaced in their body by constants in @xmath107 with the same type . given the problem @xmath10 with ( dis)equations @xmath139 , @xmath77 , for each @xmath140 let @xmath141 be the ( possibly empty ) set of bound variables in @xmath88 and @xmath142 be a corresponding set of new constants ( that do not occur in @xmath10 ) , the _ forbidden _ constants . we are interested in closed terms @xmath4 in normal form where @xmath82 and @xmath4 does not contain forbidden constants . [ def3 ] assume @xmath10 is the fixed dual interpolation problem of type @xmath16 . 1 . @xmath143 is the set of _ subtypes _ of @xmath16 including @xmath16 and the subtypes of subterms of @xmath88 . 2 . for each @xmath140 , right subterms _ are @xmath144 @xmath124 cl@xmath145 and @xmath146 . 3 . for each @xmath140 , the _ left subterms _ are @xmath147 @xmath124 @xmath148 and @xmath149 . the _ arity _ of @xmath10 , @xmath41 , is the largest width @xmath24 of any type @xmath150 . clearly , the sets @xmath143 , @xmath151 and @xmath152 are each finite and @xmath41 is bounded with respect to a given problem @xmath10 . in example [ examp1 ] , the set of forbidden constants @xmath153 is empty and @xmath151 is @xmath154 , @xmath155 is @xmath156 and its arity is @xmath157 . the _ right size _ , @xmath158 , of a right term @xmath5 relative to its constants @xmath107 is defined inductively : 1 . if @xmath159 then @xmath160 , 2 . if @xmath161 then @xmath162 , 3 . if @xmath163 , then @xmath164 . [ delta ] the _ right size _ for @xmath10 , @xmath165 , is @xmath166 of its right terms . for instance , @xmath167 and @xmath168 . if the right size @xmath165 for @xmath10 is @xmath169 , then each ( dis)equation in @xmath10 has restricted form @xmath170 where @xmath171 . padovani proved that dual interpolation is decidable for this special case , the _ atoms _ case , by showing decidability of observational equivalence within minimal models @xcite . we introduce a game - theoretic characterisation of dual interpolation inspired by model - checking games ( such as in @xcite ) where a model , a transition graph , is traversed relative to a property and players make choices at appropriate positions . similarly , in the following game the model is a putative solution term @xmath4 that is traversed relative to the dual interpolation problem . the central motivation is to model the dynamics , @xmath8-reduction , without changing @xmath4 by substituting into it . because of binding play may jump around @xmath4 . a potential solution term @xmath4 for @xmath10 has the right type , is in normal form , is well - named ( with variables that are disjoint from variables in @xmath10 ) and does not contain forbidden constants . term @xmath4 is represented as a tree , tree@xmath172 . if @xmath4 is @xmath173 or @xmath174 then tree@xmath172 is the single node labelled with @xmath4 . in the case of @xmath175 when @xmath5 is a variable or a constant , we assume that a dummy lambda with the empty sequence of variables is placed directly above any subterm @xmath176 in its tree representation . with this understanding , if @xmath4 is @xmath175 then tree@xmath172 consists of the root node labelled @xmath5 and @xmath24-successors , tree@xmath177,@xmath9,tree@xmath178 . we use the notation @xmath179 to represent that tree @xmath180 is the @xmath140th successor of the root node of @xmath4 . we also use the standard abbreviation @xmath181 for @xmath182 for some @xmath183 , so @xmath184 is possibly the empty sequence of variables . if @xmath4 is @xmath185 then tree@xmath172 consists of the root node labelled @xmath181 and a single successor tree@xmath186 , so @xmath187 . for ease of exposition , we allow @xmath4 to range over lambda terms , their trees and their root nodes : the context will make it clear which is meant . the introduction of dummy lambdas is a slight extension to @xmath11-long normal form ; they make term trees more homogeneous , allow for an easier analysis of game playing and , as we shall see in later sections , they are useful for individuating regions of a term and for defining region transformations . [ examp41 ] a solution @xmath4 from @xcite for the problem of example [ examp1 ] is the following term @xmath188 . for instance , if @xmath189 then the normal form of @xmath190 is @xmath191 . @xmath192 the tree for @xmath4 ( without indices on edges ) is depicted in figure [ ex1 ] . for instance , in this tree @xmath193 and @xmath194 . each node ( which we have uniquely identified with a natural number ) is labelled with a @xmath195 , a variable or a constant . a node labelled with a constant or variable of type @xmath13 is a leaf of the tree ( such as node @xmath196 ) . a node labelled with a higher - order constant or variable of type @xmath197 has precisely @xmath24-successor nodes each labelled with a @xmath181 ( which may be dummy ) . a node labelled with a @xmath195 has a single successor which is labelled with a constant or a variable . therefore , every even level of the tree ( when the root is at level @xmath169 ) is a lambda node . innocent game semantics following ong in @xcite provides a possible game - theoretic foundation for ( dual ) interpolation . given a potential solution term @xmath4 and a ( dis)equation @xmath198 there is the game board in figure [ ex32 ] . player opponent chooses a branch of @xmath88 . then , there is a finite play that starts at the root of @xmath4 and may repeatedly jump in and out of @xmath4 and in and out of the @xmath87 s . at a constant @xmath199 play ends . at other constants @xmath30 , player proponent tries to match opponent s choice of branch . proponent wins , when the play finishes , if the sequence of constants encountered matches the branch chosen by opponent ( assuming a mechanism for forbidden constants ) . play , for example , may reach @xmath200 in @xmath4 and then jump to @xmath87 , as it is this subtree that is applied to @xmath181 at the root of @xmath4 , and then when at @xmath201 in @xmath87 play may return to @xmath4 to an immediate successor of @xmath200 labelled @xmath195 ; play then may proceed to @xmath202 and return to a successor of @xmath201 in @xmath87 , and so on . game semantics models @xmath8-reduction on the fixed structure of figure [ ex32 ] without changing it using substitution . this is the rationale for the tree - checking game . however , the game that we now define starts from the assumption that only @xmath4 is the common structure for the problem @xmath10 . moreover , in later sections , we shall define transformations on @xmath4 justified by game playing which introduces an asymmetry between it and the argument terms @xmath87 of @xmath10 which are fixed . consequently , we insist that play is always in the term @xmath4 . jumping in and out of the @xmath87 s is coded using states , as play traverses @xmath4 . the game also avoids the justification pointers of game semantics by appealing to iteratively defined look - up tables . the tree - checking game @xmath203 is played by one participant , player @xmath204 , the _ refuter _ who attempts to show that @xmath4 is not a solution of @xmath10 . it appeals to a finite set of states involving elements of @xmath152 and @xmath151 from definition [ def3 ] : @xmath152 are the subterms of the @xmath87 s and @xmath151 are those of the @xmath88 s , both modulo the forbidden constants . there are four kinds of state , as follows . 1 . an _ argument _ state has the form @xmath205 $ ] where each @xmath206 ( and @xmath24 can be @xmath169 ) and @xmath207 . such a state will only occur at a node in @xmath4 labelled @xmath208 where each @xmath209 has the same type as @xmath210 : @xmath211 are the subterms that are applied to the subterm rooted at @xmath212 . a state @xmath213 $ ] occurs at a node of @xmath4 labelled with a dummy lambda . 2 . a _ value _ state has the form @xmath214 $ ] where @xmath215 and @xmath207 . this state can only occur at a node of @xmath4 labelled with a variable @xmath69 which has the same type as @xmath216 : @xmath216 is the subterm of some @xmath87 that play at @xmath69 would jump to in game semantics . an _ empty _ state has the form @xmath217 $ ] where @xmath207 and can only occur at a node of @xmath4 labelled with a higher - order constant @xmath218 when @xmath219 has the form @xmath220 . a _ final _ state is either @xmath221 $ ] , winning for the refuter , or @xmath222 $ ] , losing for the refuter . as play traverses @xmath4 , there are two kinds of free variables : those in the current subtree of @xmath4 ( such as @xmath200 in figure [ ex32 ] ) and those in the left terms of a current argument or value state @xmath223 $ ] or @xmath214 $ ] ( such as @xmath201 in figure [ ex32 ] ) . a free variable in a subtree of @xmath4 is associated with a single left subterm ( an element of @xmath152 ) and a free variable in a left subterm @xmath216 of an argument or value state is associated with a unique subtree of @xmath4 . so , the game appeals to look - up tables or nested environments @xmath224 and @xmath225 at a position @xmath226 that are defined iteratively . [ defi42 ] the sets of partial mappings @xmath227 and @xmath228 are defined iteratively as follows . 1 . @xmath229 and @xmath230 where both @xmath231 and @xmath232 are empty ( that is , have no entries ) . 2 . for @xmath233 , @xmath224 iff @xmath1 is a partial map from variables that are labels of nodes in @xmath4 to triples @xmath234 where @xmath215 , @xmath235 and @xmath236 . 3 . for @xmath233 , @xmath225 iff @xmath237 is a partial map from variables that can occur in terms of @xmath152 to triples @xmath238 where @xmath180 is a subtree of @xmath4 , @xmath235 and @xmath239 . a variable @xmath69 in @xmath4 may be associated with a left subterm @xmath215 which contains free variables : hence , the need for @xmath240 to be a triple @xmath234 as @xmath237 records the values of the free variables in @xmath216 as determined at the earlier position @xmath241 : we include the position @xmath241 which is crucial to the understanding of game playing in later sections , when relationships between positions are analysed . similarly , a variable @xmath242 in a left subterm may be associated with a subtree of @xmath4 which contains free variables ; so , @xmath243 where @xmath1 has entries for the free variables in @xmath180 at the earlier position @xmath241 . initially , at the beginning of play when there are no free variables and no previous moves , @xmath244 and @xmath245 are both empty . the look - up tables play the same role as environments for abstract machines of the lambda calculus ( such as a krivine machine ) . so the game will simulate the evaluation of a branch of the normal form in the same way abstract machines compute closures . however , unlike these abstract machines , the game here essentially depends on @xmath11-long normal forms . a _ play _ of @xmath203 is a sequence of positions @xmath246 where 1 . each @xmath51 is a node of @xmath4 and @xmath247 is the root node of @xmath4 , 2 . each @xmath248 is a state , @xmath249 is a final state and @xmath250 is decided as follows : @xmath251 chooses a ( dis)equation @xmath252 in @xmath10 and @xmath253 $ ] , 3 . for each @xmath140 , @xmath254 and @xmath255 , 4 . position @xmath256 , @xmath257 , is determined by a single move in figure [ game ] from position @xmath258 according to the label at @xmath259 . a. @xmath259 is labelled @xmath260 and @xmath261 . assume @xmath262 $ ] . + then , @xmath263 such that @xmath264 , @xmath265 and @xmath266 , @xmath267 are defined by cases on the label at @xmath268 . 1 . @xmath199 . then , @xmath269 . if @xmath270 then @xmath271 $ ] else @xmath272 $ ] . 2 . @xmath273 . then , @xmath269 . if @xmath274 then @xmath275 $ ] else @xmath276 $ ] . 3 . @xmath277 . if @xmath278 , then @xmath279 and @xmath280 $ ] . b. @xmath259 is labelled @xmath281 . assume @xmath282 $ ] . 1 . then , @xmath283 , @xmath269 and @xmath251 chooses @xmath284 and @xmath263 such that @xmath285 and @xmath266 is by cases on @xmath286 . + @xmath287 . then @xmath288 $ ] . + @xmath289 . then @xmath266 @xmath124 @xmath290 $ ] . c. @xmath259 is labelled @xmath69 . assume @xmath291 $ ] . + if @xmath292 then @xmath269 else @xmath293 and @xmath294 , @xmath295 , and @xmath296 . elements @xmath268 , @xmath266 and @xmath297 are by cases on @xmath216 . 1 . @xmath298 or @xmath299 where @xmath199 . then , @xmath300 and @xmath283 . + if @xmath301 then @xmath302 $ ] else @xmath303 $ ] . . then @xmath283 . + if @xmath305 then @xmath300 and @xmath306 $ ] else @xmath307 and @xmath251 chooses @xmath308 , and @xmath263 such that @xmath309 and @xmath266 is by cases on @xmath286 . + @xmath287 . then @xmath288 $ ] . . then @xmath266 @xmath124 @xmath290 $ ] . 3 . @xmath311 or @xmath312 . then @xmath300 and @xmath313 . + if @xmath314 , then @xmath306 $ ] else @xmath315 and @xmath251 chooses @xmath284 and @xmath266 is by cases on @xmath286 . + @xmath287 . then @xmath316 $ ] . + @xmath317 and @xmath318 . then @xmath266 @xmath124 @xmath319 $ ] . 4 . @xmath320 or @xmath321 where @xmath322 . + if @xmath323 then @xmath263 , @xmath324 and @xmath325 $ ] . the initial position of a play of @xmath203 is @xmath326 \theta_{1 } \xi_{1}$ ] where @xmath247 is the initial node of @xmath4 labelled @xmath182 for some @xmath327 and @xmath328 is a ( dis)equation in @xmath10 . at this position we are interested whether @xmath329 where the look - up tables are viewed as substitutions : initially , they are empty because there are no free variables in the terms @xmath330 . the initial position is the same as an initial position in game semantics , except the terms @xmath209 and @xmath219 in the game here are part of the state ( and the choice of branch in @xmath219 will take place as play proceeds ) . the possible moves from position @xmath331 to @xmath332 in a play are listed in figure [ game ] and are divided into three groups that depend on the label at @xmath259 of position @xmath331 . group a covers the case when it is a @xmath181 , group b a constant @xmath30 ( whose type is not @xmath13 ) and group c a variable @xmath69 . for look - up tables @xmath297 and @xmath267 , we assume that the substitution notation also stands for function updating : @xmath333 means that the entries in @xmath334 are the same as in @xmath335 except for the @xmath336 s , as @xmath337 for each @xmath336 , @xmath338 . this notation is permitted even if @xmath339 . consider group a moves when position @xmath331 is at the node @xmath259 labelled @xmath260 , @xmath261 , so @xmath340 has the form @xmath341 $ ] where each @xmath342 has the same type as @xmath336 . ( if @xmath343 then the position is at a dummy lambda . ) the `` interpretation '' of position @xmath331 in terms of @xmath8-reduction ( see the proof of theorem [ imp ] ) is whether @xmath344 @xmath345 . node @xmath259 has a single successor @xmath268 and so play descends to it . however , the subtree at @xmath268 may contain free occurrences of the @xmath336 s ( when @xmath346 ) : the interpretation of each such occurrence is @xmath347 because @xmath348 and so @xmath297 is an updated version of @xmath349 reflecting this association of the @xmath336 s with the @xmath342 s and the look - up table @xmath350 at position @xmath331 . move a1 is when @xmath268 is labelled with @xmath199 . we can now immediately decide whether @xmath351 just by comparing @xmath219 and @xmath298 ; so position @xmath332 is final ( and won by the refuter if @xmath219 is different from @xmath298 ) . for move a2 , if @xmath268 is labelled with @xmath218 and @xmath219 does not have the form @xmath352 then @xmath303 $ ] as we now know that @xmath344 @xmath353 . if @xmath274 then position @xmath332 is @xmath354 \theta_{m+1 } \xi_{m+1}$ ] where @xmath269 . move a3 is when @xmath268 is labelled with a variable @xmath69 . thus , @xmath69 is a free variable occurrence in the tree @xmath268 whose interpretation is @xmath278 decided at the earlier position @xmath355 . so , @xmath267 is set to @xmath237 ( as it interprets the free variables in @xmath216 in @xmath356 $ ] ) . the @xmath17 move covers the case when @xmath259 is labelled with a constant @xmath218 ; because of move a2 , we only need to consider it when the state is @xmath357 $ ] for some @xmath358 . the `` interpretation '' of such a position @xmath331 is whether @xmath359 . in which case @xmath360 so , @xmath359 if , and only if , for each @xmath308 , @xmath361 . the refuter @xmath251 chooses such a @xmath362 . the delicacy is that @xmath286 may be of higher type , of the form @xmath363 and so , therefore , @xmath364 is labelled @xmath365 for some @xmath366 because it has the same type : position @xmath332 is then @xmath367 \theta_{m+1 } \xi_{m+1}$ ] where @xmath283 , and @xmath269 : here we are making use of the forbidden constants @xmath368 . if @xmath286 has ground type then @xmath288 $ ] ( as the label at @xmath364 is a dummy lambda ) . group c moves cover the case where @xmath259 is labelled with a variable @xmath69 . the `` interpretation '' of position @xmath331 , @xmath369 \theta_{m } \xi_{m}$ ] , is whether @xmath370 where for @xmath261 , @xmath294 when @xmath295 . if @xmath346 then @xmath293 for some @xmath371 and the free occurrences of @xmath372 in @xmath105 are associated with @xmath373 at position @xmath331 ; so , @xmath267 is an updated version of @xmath350 reflecting this association . when @xmath374 , @xmath375 and the interpretation is whether @xmath376 . for both cases @xmath293 and @xmath377 play proceeds by examining the `` head '' of @xmath105 . move c1 covers the case where it is @xmath199 ( which is possibly a forbidden constant ) . we can now immediately decide whether @xmath370 just by comparing @xmath216 and @xmath219 . move c2 covers the case where @xmath216 is a forbidden constant @xmath378 at higher type @xmath379 . if @xmath380 then we know that @xmath381 and , so @xmath382 $ ] . if @xmath383 then @xmath384 if , and only if , for each @xmath362 , @xmath385 . as with b1 , the refuter @xmath251 chooses such a @xmath362 . again we need to specially deal with the case that @xmath286 has higher type . move c3 deals with the case that @xmath105 is @xmath311 . again , if @xmath219 does not have the form @xmath352 then @xmath386 $ ] as we now know that @xmath387 . if @xmath274 then @xmath370 if , and only if , for each @xmath362 , @xmath388 . therefore , @xmath251 chooses such a @xmath362 . for the next position , again we need to deal with the possibility that @xmath286 is of higher type : if @xmath289 then @xmath389 which has the same type must be of the form @xmath390 for some @xmath391 ; so we must substitute the same forbidden constant for each @xmath392 and @xmath200 ; so @xmath266 is @xmath319 $ ] . if @xmath287 then the next position @xmath332 is @xmath393 \theta_{m+1 } \xi_{m+1}$ ] where @xmath394 and @xmath395 . move c4 covers the case when @xmath396 , @xmath322 . therefore , @xmath29 is a free variable in @xmath216 whose interpretation @xmath323 which was determined at the earlier position @xmath355 ; so , @xmath29 is associated with the subtree @xmath180 of @xmath4 and play therefore jumps to it . @xmath397 & \\ & ( 2 ) y \ar[d ] & \\ & ( 3 ) \lambda \ar[d ] & \\ & ( 4 ) y \ar[d ] & \\ & ( 5 ) \lambda \ar[d ] & \\ & \ar[dl ] ( 6 ) f \ar[dr ] & \\ \ar[d ] ( 7 ) \lambda x z_{1}z_{2 } & & ( 13 ) \lambda \ar[d ] \\ \ar[d ] ( 8) x & & ( 14 ) y \ar[d ] \\ \ar[d ] ( 9 ) \lambda & & ( 15 ) \lambda \ar[d ] \\ \ar[d ] ( 10 ) y & & ( 16 ) a \\ \ar[d ] ( 11 ) \lambda & & \\ ( 12 ) z_{2 } & & } \ ] ] @xmath251 can exercise choice , by carving out a branch of a right term of a ( dis)equation modulo forbidden constants , with moves b1 , c2 and c3 . the look - up tables are used in earnest with moves a3 and c4 to interpret the two kinds of free variable . move c4 allows play to jump elsewhere in the term tree ( always to a node labelled with a lambda ) : it also opens the possibility that play can repeatedly be at the same node of @xmath4 . with moves a1a3 , b1 and c2 ( unless play finishes ) control passes down the term tree while it remains stationary in the case of c1 and c3 . if @xmath398 is a play of @xmath203 then player @xmath251 _ wins the play _ if the final state is @xmath399 $ ] and she _ loses _ it otherwise ( if the final state is @xmath400 $ ] ) . [ examp42 ] let @xmath10 be the problem @xmath401 where @xmath402 of example [ examp2 ] . the bound variables in the right term @xmath5 are @xmath403 ; so , assume corresponding forbidden constants @xmath404 . let @xmath405 : tree@xmath172 is depicted in figure [ picture2 ] . @xmath406 consists of two plays that descend @xmath4 . ( for @xmath12rd - order problems @xmath10 play can not jump around a term tree , as we observe in section [ third ] . ) both plays start as follows where we have supplied which move is applied to produce the next position : the initial component of each position is a node of @xmath4 . @xmath407\ , \theta_{1 } \ , \xi_{1 } & & & \\ ( 2 ) \ , q [ \lambda z.z , u ] \ , \theta_{2 } \xi_{2 } & \theta_{2 } = \theta_{1 } { \ { ( \lambda z.z ) \xi_{1 } 1/y \ } } & \xi_{2 } = \xi_{1 } & a3\\ ( 3 ) \ , q [ ( \ ) , u ] \ , \theta_{3 } \xi_{3 } & \theta_{3 } = \theta_{2 } & \xi_{3 } = \xi_{1 } { \ { ( 3)\theta_{2 } 2/z \ } } & c4\\ ( 4 ) \ , q[\lambda z. z , u ] \ , \theta_{4 } \xi_{4 } & \theta_{4 } = \theta_{2 } & \xi_{4 } = \xi_{1 } & a3\\ ( 5 ) \ , q [ ( \ ) , u ] \ , \theta_{5 } \ , \xi_{5 } & \theta_{5 } = \theta_{2 } & \xi_{5 } = \xi_{1 } { \ { ( 5 ) \theta_{2 } 4/z \ } } & c4\\ ( 6 ) \ , q[- , u ] \ , \theta_{6 } \ , \xi_{6 } & \theta_{6 } = \theta_{2 } & \xi_{6 } = \xi_{5 } & a2 \end{array}\ ] ] the initial state is an argument state @xmath408 $ ] and control is at node ( 1 ) . play descends from node ( 1 ) to ( 2 ) calling the value @xmath133 by a3 . next , by c4 , because @xmath242 is the head variable in the body of @xmath133 , has no arguments and is associated with node ( 3 ) , the next state is the argument state @xmath409 $ ] and control is at ( 3 ) . play descends from ( 3 ) to ( 4 ) calling the value @xmath133 by a3 . again by c4 , @xmath242 is the head variable in the body of @xmath133 , has no arguments and is now associated with ( 5 ) , the next state is the argument state @xmath410 $ ] and control is at ( 5 ) . a2 is then applied ( because the right term @xmath5 in the state has @xmath30 as head constant ) and control passes from ( 5 ) to ( 6 ) . move b1 is now applied and there is a @xmath251 choice as to which branch of @xmath5 to take . if direction @xmath21 is chosen then play continues as follows . @xmath411 \ , \theta_{7 } \ , \xi_{7 } & \theta_{7 } = \theta_{2 } & \xi_{7 } = \xi_{5 } & b1\\ ( 8) \ , q[c_{1 } , c_{1 } c_{3 } ] \ , \theta_{8 } \ , \xi_{8 } & \theta_{8 } = \theta_{2 } { \ { c_{1}\xi_{5 } 7\!/\!x , c_{2 } \xi_{5 } 7\!/\!z_{1 } , c_{3}\xi_{5 } 7\!/\!z_{2 } \ } } & \xi_{8 } = \xi_{5 } & a3\\ ( 9 ) \ , q [ ( \ ) , c_{3 } ] \ , \theta_{9 } \ , \xi_{9 } & \theta_{9 } = \theta_{8 } & \xi_{9 } = \xi_{5 } & c2 \\ ( 10 ) \ , q[\lambda z. z , c_{3 } ] \ , \theta_{10 } \xi_{10 } & \theta_{10 } = \theta_{8 } & \xi_{10 } = \xi_{5 } & a3\\ ( 11 ) \ , q [ ( \ ) , c_{3 } ] \ , \theta_{11 } \ , \xi_{11 } & \theta_{11 } = \theta_{8 } & \xi_{11 } = \xi_{5 } { \ { ( 11 ) \theta_{8 } 10\!/\!z \ } } & c4\\ ( 12 ) \ , q [ c_{3},c_{3 } ] \ , \theta_{12 } \ , \xi_{12 } & \theta_{12 } = \theta_{8 } & \xi_{12 } = \xi_{5 } & a3 \\ ( 12 ) \ , q [ \ { \exists}\ ] \ , \theta_{13 } \ , \xi_{13 } & \theta_{13 } = \theta_{8 } & \xi_{13 } = \xi_{5 } & c1 \end{array}\ ] ] forbidden constants are introduced for replacing @xmath412 in the body of @xmath5 to give the right term @xmath413 and as arguments @xmath414 for the variables @xmath415 bound at ( 7 ) ; see the updated look - up table @xmath416 . at ( 8) the value @xmath417 is called using move a3 and then by c2 , control proceeds to ( 9 ) and the right term becomes the argument @xmath418 of @xmath417 . at ( 10 ) the value @xmath133 is called again by a3 and by c4 as @xmath242 is the head variable and is associated with node ( 11 ) control passes to it ( with the empty sequence of arguments ) . finally , at ( 12 ) , the value @xmath418 is called by a3 , and then by c1 , @xmath251 loses the play . she also loses if direction @xmath157 is chosen at position @xmath419 as the reader can verify . [ examp43 ] let @xmath10 be the first equation of example [ examp1 ] @xmath420 where @xmath421 ( which is a @xmath7th - order problem ) . the tree @xmath4 of figure [ ex1 ] solves @xmath10 as was demonstrated in example [ examp41 ] through @xmath8-reduction . the single play in @xmath203 is presented in figure [ ex2 ] . @xmath422\ , \theta_{1 } \ , \xi_{1 } & & & \\ ( 2 ) \ , q [ v , fa ] \ , \theta_{2 } \xi_{2 } & \theta_{2 } = \theta_{1 } { \ { v \xi_{1 } 1/z \ } } & \xi_{2 } = \xi_{1 } & a3\\ ( 3 ) \ , q[(y_{2 } ) , fa ] \ , \theta_{3 } \xi_{3 } & \theta_{3 } = \theta_{2 } & \xi_{3 } = \xi_{1 } { \ { ( 3)\theta_{2 } 2/y_{1},(11)\theta_{2 } 2/y_{2 } \ } } & c4\\ ( 4 ) \ , q[- , fa ] \ , \theta_{4 } \xi_{4 } & \theta_{4 } = \theta_{2 } { \ { y_{2}\xi_{3 } 3/x \ } } & \xi_{4 } = \xi_{3 } & a2\\ ( 5 ) \ , q [ ( \ ) , a ] \ , \theta_{5 } \ , \xi_{5 } & \theta_{5 } = \theta_{4 } & \xi_{5 } = \xi_{4 } & b1\\ ( 6 ) \ , q[v , a ] \ , \theta_{6 } \ , \xi_{6 } & \theta_{6 } = \theta_{4 } & \xi_{6 } = \xi_{1 } & a3\\ ( 7 ) , q[(y_{2 } ) , a ] \ , \theta_{7 } \ , \xi_{7 } & \theta_{7 } = \theta_{4 } & \xi_{7 } = \xi_{1 } { \ { ( 7)\theta_{4 } 6/y_{1},(9)\theta_{4 } 6/y_{2 } \ } } & c4\\ ( 8) \ , q[y_{2 } , a ] \ , \theta_{8 } \ , \xi_{8 } & \theta_{8 } = \theta_{4 } { \ { y_{2 } \xi_{7}7/u \ } } & \xi_{8 } = \xi_{3 } & a3\\ ( 11 ) \ , q [ ( \ ) , a ] \ , \theta_{9 } \ , \xi_{9 } & \theta_{9 } = \theta_{2 } & \xi_{9 } = \xi_{3 } & c4 \\ ( 12 ) \ , q [ v , a ] \ , \theta_{10 } \xi_{10 } & \theta_{10 } = \theta_{2 } & \xi_{10 } = \xi_{1 } & a3\\ ( 13 ) \ , q[(y_{2 } ) , a ] \ , \theta_{11 } \xi_{11 } & \theta_{11 } = \theta_{2 } & \xi_{11 } = \xi_{1 } { \ { ( 13)\theta_{2 } 10/y_{1},(19)\theta_{2 } 10/y_{2 } \ } } & c4\\ ( 14 ) \ , q[v , a ] \ , \theta_{12 } \xi_{12 } & \theta_{12 } = \theta_{2 } { \ { y_{2}\xi_{11 } 11/y \ } } & \xi_{12 } = \xi_{1 } & a3\\ ( 15 ) \ , q[(y_{2 } ) , a ] \ , \theta_{13 } \ , \xi_{13 } & \theta_{13 } = \theta_{12 } & \xi_{13 } = \xi_{1 } { \ { ( 15)\theta_{12 } 12/y_{1 } , ( 17)\theta_{12 } 12/y_{2 } \ } } & c4\\ ( 16 ) \ , q[y_{2 } , a ] \ , \theta_{14 } \ , \xi_{14 } & \theta_{14 } = \theta_{12 } { \ { y_{2}\xi_{13 } 13/s \ } } & \xi_{14 } = \xi_{13 } & a3\\ ( 17 ) \ , q [ ( \ ) , a ] \ , \theta_{15 } \ , \xi_{15 } & \theta_{15 } = \theta_{12 } & \xi_{15 } = \xi_{13 } & c4\\ ( 18 ) \ , q[y_{2 } , a ] \ , \theta_{16 } \ , \xi_{16 } & \theta_{16 } = \theta_{12 } & \xi_{16 } = \xi_{11 } & a3\\ ( 19 ) \ , q [ ( \ ) , a ] \ , \theta_{17 } \ , \xi_{17 } & \theta_{17 } = \theta_{2 } & \xi_{17 } = \xi_{11 } & c4\\ ( 20 ) \ , q [ \ { \exists}\ ] \ , \theta_{18 } \ , \xi_{18 } & \theta_{18 } = \theta_{2 } & \xi_{18 } = \xi_{11 } & a1\\ \end{array}\ ] ] the initial state is @xmath423 $ ] at the root of @xmath4 . by move a3 , play descends to node @xmath424 calling the value @xmath47 . next by c4 because @xmath425 is the head variable in the body of @xmath47 , is associated with the subtree at @xmath426 and has argument @xmath427 play moves to @xmath426 with argument state @xmath428 $ ] . by a2 play descends to node @xmath429 labelled @xmath30 and then move b1 is applied without a choice for @xmath251 because the type of @xmath430 has arity one , and so the play descends to node @xmath431 with a change in the right term of the state from @xmath432 to @xmath298 . by move a3 play descends to node ( 6 ) calling the value @xmath47 again . by c4 because the head variable @xmath425 is associated with the subtree at @xmath433 and has argument @xmath427 play moves to @xmath433 . by move a3 , play descends to @xmath434 . the entry @xmath435 is @xmath436 , so the state is @xmath437 $ ] . by move c4 because @xmath438 , that is , @xmath427 is associated with the subtree rooted at @xmath439 , play jumps from node @xmath434 to node @xmath439 . if node @xmath434 were labelled @xmath5 then because the entry @xmath440 the state would again be @xmath437 $ ] ; play would then jump to node @xmath441 because @xmath442 . play descends from node @xmath439 to @xmath443 , @xmath444 , @xmath445 , @xmath446 and @xmath447 and jumps to node @xmath448 and descends to @xmath449 and then jumps to node @xmath450 before descending to node @xmath451 by move a1 , where the refuter loses the play . @xmath452 & \\ & & \ar[dl](2 ) z \ar[dr ] & \\ & ( 3 ) \ar[d ] \lambda z_{1 } & & \ar[d ] ( 17 ) \lambda z_{2 } & \\ & \ar[dl ] ( 4 ) z \ar[dr ] & & \ar[d ] ( 18 ) h \\ \ar[d ] ( 5 ) \lambda x_{1 } & & \ar[d ] ( 13 ) \lambda x_{2 } & \ar[d ] ( 19 ) \lambda \\ \ar[d](6 ) z_{1 } & & \ar[d ] ( 14 ) g & ( 20 ) z_{2 } \\ \ar[d ] ( 7 ) \lambda & & \ar[d ] ( 15 ) \lambda & \\ \ar[d ] ( 8) x_{1 } & & ( 16 ) x_{2 } & \\ \ar[d ] ( 9 ) \lambda & & & \\ \ar[d ] ( 10 ) z_{1 } & & & \\ \ar[d ] ( 11 ) \lambda & & & \\ ( 12 ) a & & & } \ ] ] [ examp44 ] we now examine the equation of example [ examp3 ] which illustrates play jumping in more detail and how the game moves , especially a3 and c4 , essentially depend on @xmath11-long normal forms let @xmath10 be the equation @xmath453 and let @xmath4 be @xmath454 which is a solution ; @xmath4 as a tree is depicted in figure [ ex14 ] . the single play for @xmath203 is presented in figure [ ex15 ] where the following abbreviations for left and right subterms are employed . @xmath455 @xmath456\ , \theta_{1 } \ , \xi_{1 } & & & \\ ( 2 ) \ , \ , q [ v , u ] \ , \theta_{2 } \xi_{2 } & \theta_{2 } = \theta_{1 } { \ { v \xi_{1}1/z \ } } & \xi_{2 } = \xi_{1 } & a3\\ ( 3 ) \ , \ , q[(v_{1 } ) , u ] \ , \theta_{3 } \xi_{3 } & \theta_{3 } = \theta_{2 } & \xi_{3 } = \xi_{1 } { \ { ( 3)\theta_{2 } 2/y_{1},(17)\theta_{2 } 2/y_{2 } \ } } & c4\\ ( 4 ) \ , \ , q[v , u ] \ , \theta_{4 } \xi_{4 } & \theta_{4 } = \theta_{2 } { \ { v_{1}\xi_{3 } 3/z_{1 } \ } } & \xi_{4 } = \xi_{1 } & a3\\ ( 5 ) \ , \ , q[(v_{1 } ) , u ] \ , \theta_{5 } \ , \xi_{5 } & \theta_{5 } = \theta_{4 } & \xi_{5 } = \xi_{1 } { \ { ( 5)\theta_{4 } 4/y_{1 } , ( 13)\theta_{4 } 4/y_{2 } \ } } & c4\\ ( 6 ) \ , \ , q[v_{1 } , u ] \ , \theta_{6 } \ , \xi_{6 } & \theta_{6 } = \theta_{4 } { \ { v_{1}\xi_{5 } 5/x_{1 } \ } } & \xi_{6 } = \xi_{3}&a3 \\ ( 17 ) \ , q[(v_{2 } ) , u ] \ , \theta_{7 } \ , \xi_{7 } & \theta_{7 } = \theta_{2 } & \xi_{7 } = \xi_{3 } { \ { ( 7)\theta_{6 } 6/y_{3 } \ } } & c4\\ ( 18 ) \ , q[- , u ] \ , \theta_{8 } \ , \xi_{8 } & \theta_{8 } = \theta_{2 } { \ { v_{2 } \xi_{7 } 7/z_{2 } \ } } & \xi_{8 } = \xi_{7 } & a2\\ ( 19 ) \ , q [ ( \ ) , u_{1 } ] \ , \theta_{9 } \ , \xi_{9 } & \theta_{9 } = \theta_{8 } & \xi_{9 } = \xi_{7 } & b1\\ ( 20 ) \ , q [ v_{2 } , u_{1 } ] \ , \theta_{10 } \xi_{10 } & \theta_{10 } = \theta_{8 } & \xi_{10 } = \xi_{7 } & a3\\ ( 3 ) \ , \ , q[(v_{3 } ) , u_{1 } ] \ , \theta_{11 } \xi_{11 } & \theta_{11 } = \theta_{2 } & \xi_{11 } = \xi_{7 } & c4\\ ( 4 ) \ , \ , q[v , u_{1 } ] \ , \theta_{12 } \xi_{12 } & \theta_{12 } = \theta_{2 } { \ { v_{3 } \xi_{7 } 11/z_{1 } \ } } & \xi_{12 } = \xi_{1 } & a3\\ ( 5 ) \,\ , q[(v_{1 } ) , u_{1 } ] \ , \theta_{13 } \ , \xi_{13 } & \theta_{13 } = \theta_{12 } & \xi_{13 } = \xi_{1 } { \ { ( 5)\theta_{12 } 12/y_{1 } , ( 13)\theta_{12 } 12/y_{2 } \ } } & c4\\ ( 6 ) \ , \ , q[v_{3 } , u_{1 } ] \ , \theta_{14 } \ , \xi_{14 } & \theta_{14 } = \theta_{12 } { \ { v_{1}\xi_{13 } 13/x_{1 } \ } } & \xi_{14 } = \xi_{7 } & a3 \\ ( 7 ) \ , \ , q [ ( \ ) , u_{1 } ] \ , \theta_{15 } \ , \xi_{15 } & \theta_{15 } = \theta_{6 } & \xi_{15 } = \xi_{14 } { \ { ( 7)\theta_{14 } 14/y_{4 } \ } } & c4\\ ( 8) \ , \ , q[v_{1 } , u_{1 } ] \ , \theta_{16 } \ , \xi_{16 } & \theta_{16 } = \theta_{6 } & \xi_{16 } = \xi_{5 } & a3\\ ( 13 ) \ , q[(v_{2 } ) , u_{1 } ] \ , \theta_{17 } \ , \xi_{17 } & \theta_{17 } = \theta_{4 } & \xi_{17 } = \xi_{5 } { \ { ( 9)\theta_{6 } 16/y_{3 } \ } } & c4\\ ( 14 ) \ , q[- , u_{1 } ] \ , \theta_{18 } \ , \xi_{18 } & \theta_{18 } = \theta_{4 } { \ { v_{2 } \xi_{17 } 17/x_{2 } \ } } & \xi_{18 } = \xi_{17 } & a2\\ ( 15 ) \ , q [ ( \ , ) , u_{2 } ] \ , \theta_{19 } \ , \xi_{19 } & \theta_{19 } = \theta_{18 } & \xi_{19 } = \xi_{17 } & b1\\ ( 16 ) \ , q[v_{2 } , u_{2 } ] \ , \theta_{20 } \ , \xi_{20 } & \theta_{20 } = \theta_{18 } & \xi_{20 } = \xi_{17 } & a3\\ ( 5 ) \ , \ , q[(v_{3 } ) , u_{2 } ] \ , \theta_{21 } \ , \xi_{21 } & \theta_{21 } = \theta_{4 } & \xi_{21 } = \xi_{17 } & c4\\ ( 6 ) \ , \ , q[v_{1 } , u_{2 } ] \ , \theta_{22 } \ , \xi_{22 } & \theta_{22 } = \theta_{4 } { \ { v_{3}\xi_{17 } 21/x_{1 } \ } } & \xi_{22 } = \xi_{3 } & a3\\ ( 17 ) \ , q[(v_{2 } ) , u_{2 } ] \ , \theta_{23 } \ , \xi_{23 } & \theta_{23 } = \theta_{2 } & \xi_{23 } = \xi_{3 } { \ { ( 7)\theta_{22 } 22/y_{3 } \ } } & c4\\ ( 18 ) \ , q[- , u_{2 } ] \ , \theta_{24 } \ , \xi_{24 } & \theta_{24 } = \theta_{2 } { \ { v_{2 } \xi_{23 } 23/z_{2 } \ } } & \xi_{24 } = \xi_{23 } & a2\\ ( 19 ) \ , q [ ( \ ) , h a ] \ , \theta_{25 } \ , \xi_{25 } & \theta_{25 } = \theta_{24 } & \xi_{25 } = \xi_{23 } & b1\\ ( 20 ) \ , q [ v_{2 } , h a ] \ , \theta_{26 } \xi_{26 } & \theta_{26 } = \theta_{24 } & \xi_{26 } = \xi_{23 } & a3\\ ( 3 ) \ , \ , q[(v_{3 } ) , h a ] \ , \theta_{27 } \xi_{27 } & \theta_{27 } = \theta_{2 } & \xi_{27 } = \xi_{23 } & c4\\ ( 4 ) \ , \ , q[v , h a ] \ , \theta_{28 } \xi_{28 } & \theta_{28 } = \theta_{2 } { \ { v_{3 } \xi_{23 } 27/z_{1 } \ } } & \xi_{28 } = \xi_{1 } & a3\\ ( 5 ) \,\ , q[(v_{1 } ) , h a ] \ , \theta_{29 } \ , \xi_{29 } & \theta_{29 } = \theta_{28 } & \xi_{29 } = \xi_{1 } { \ { ( 5)\theta_{28 } 28/y_{1 } , ( 13)\theta_{28 } 28/y_{2 } \ } } & c4\\ ( 6 ) \ , \ , q[v_{3 } , h a ] \ , \theta_{30 } \ , \xi_{30 } & \theta_{30 } = \theta_{28 } { \ { v_{1}\xi_{29 } 29/x_{1 } \ } } & \xi_{30 } = \xi_{23 } & a3\\ ( 7 ) \ , \ , q [ ( \ ) , h a ] \ , \theta_{31 } \ , \xi_{31 } & \theta_{31 } = \theta_{22 } & \xi_{31 } = \xi_{23 } { \ { ( 7)\theta_{30 } 30/y_{4 } \ } } & c4\\ ( 8) \ , \ , q[v_{3 } , h a ] \ , \theta_{32 } \ , \xi_{32 } & \theta_{32 } = \theta_{22 } & \xi_{32 } = \xi_{17 } & a3\\ ( 9 ) \ , \ , q [ ( \ ) , h a ] \ , \theta_{33 } \ , \xi_{33 } & \theta_{33 } = \theta_{6 } & \xi_{33 } = \xi_{32 } { \ { ( 9)\theta_{22 } 32/y_{4 } \ } } & c4 \\ ( 10 ) \ , q[v_{1 } , h a ] \ , \theta_{34 } \ , \xi_{34 } & \theta_{34 } = \theta_{6 } & \xi_{34 } = \xi_{3 } & a3\\ ( 17 ) \ , q[(v_{2 } ) , h a ] \ , \theta_{35 } \ , \xi_{35 } & \theta_{35 } = \theta_{2 } & \xi_{35 } = \xi_{3 } { \ { ( 11)\theta_{6 } 34/y_{3 } \ } } & c4\\ ( 18 ) \ , q[- , h a ] \ , \theta_{36 } \ , \xi_{36 } & \theta_{36 } = \theta_{2 } { \ { v_{2 } \xi_{35 } 35/z_{2 } \ } } & \xi_{36 } = \xi_{35 } & a2\\ ( 19 ) \ , q [ ( \ ) , a ] \ , \theta_{37 } \ , \xi_{37 } & \theta_{37 } = \theta_{36 } & \xi_{37 } = \xi_{35 } & b1\\ ( 20 ) \ , q [ v_{2 } , a ] \ , \theta_{38 } \xi_{38 } & \theta_{38 } = \theta_{36 } & \xi_{38 } = \xi_{35 } & a3\\ ( 3 ) \ , \ , q[(v_{3 } ) , a ] \ , \theta_{39 } \xi_{39 } & \theta_{39 } = \theta_{2 } & \xi_{39 } = \xi_{35 } & c4\\ ( 4 ) \ , \ , q[v , a ] \ , \theta_{40 } \xi_{40 } & \theta_{40 } = \theta_{2 } { \ { v_{3 } \xi_{35 } 39/z_{1 } \ } } & \xi_{40 } = \xi_{1 } & a3\\ ( 5 ) \,\ , q[(v_{1 } ) , a ] \ , \theta_{41 } \ , \xi_{41 } & \theta_{41 } = \theta_{40 } & \xi_{41 } = \xi_{1 } { \ { ( 5)\theta_{40 } 40/y_{1 } , ( 13)\theta_{40 } 40/y_{2 } \ } } & c4\\ ( 6 ) \ , \ , q[v_{3 } , a ] \ , \theta_{42 } \ , \xi_{42 } & \theta_{42 } = \theta_{40 } { \ { v_{1}\xi_{41 } 41/x_{1 } \ } } & \xi_{42 } = \xi_{35 } & a3 \\ ( 11 ) \ , q [ ( \ ) , a ] \ , \theta_{43 } \ , \xi_{43 } & \theta_{43 } = \theta_{6 } & \xi_{43 } = \xi_{35 } { \ { ( 7)\theta_{42}42/y_{4 } \ } } & c4 \\ ( 12 ) \ , q [ \ { \exists}\ ] \ , \theta_{44 } \ , \xi_{44 } & \theta_{44 } = \theta_{6 } & \xi_{44 } = \xi_{43 } & a1\\ \end{array}\ ] ] play starts at node @xmath457 with state @xmath458 $ ] and by a3 descends to @xmath424 . the head variable of the body of @xmath47 is @xmath425 which has argument @xmath459 and so play moves to @xmath426 by c4 with state @xmath460 $ ] . similarly , it descends to @xmath429 by a3 and then to @xmath431 by c4 with state @xmath460 $ ] at which point it moves to @xmath461 which is labelled @xmath462 ; @xmath463 and the head variable of the body of @xmath459 is @xmath427 and its argument is @xmath464 ; so , play jumps to @xmath448 with state @xmath465 $ ] . it then descends to @xmath449 , @xmath450 and @xmath451 with state @xmath466 $ ] ; the head variable of @xmath464 is @xmath425 and it has argument @xmath467 ; so play returns to @xmath426 with state @xmath468 $ ] . it then descends to @xmath429 and @xmath431 and moves to @xmath461 ; the head variable of the body of @xmath467 is @xmath469 which is associated with node @xmath433 , as @xmath470 . play proceeds to ( 8) , jumps to ( 13 ) , reaches ( 16 ) and then returns to ( 5 ) and so on . player @xmath251 eventually loses when play reaches ( 12 ) as the reader can check . [ def45 ] if @xmath10 is a ( dual ) interpolation problem then @xmath251 _ loses the game _ @xmath203 if and only if 1 . for every equation in @xmath10 , @xmath251 loses every play whose initial state is given from it , 2 . for each disequation in @xmath10 , @xmath251 wins some play whose initial state is given from it . the game _ characterises _ dual interpolation . [ thm1 ] @xmath251 loses @xmath203 if , and only if , @xmath82 . the simply typed @xmath471-calculus is strongly normalising and is church - rosser modulo @xmath41-equivalence . for every term @xmath4 there is an @xmath331 such that @xmath4 reduces to normal form using at most @xmath331 @xmath8-reductions ( whatever the reduction strategy ) . therefore , for any position @xmath472 of a play of @xmath473 we say that it @xmath331-holds ( @xmath331-fails ) if @xmath474 $ ] ( @xmath475 $ ] ) and when @xmath248 is not final , by cases on @xmath51 and @xmath248 ( and look - up tables become delayed substitutions ) 1 . if @xmath476 , @xmath477 $ ] and @xmath180 is @xmath478 then @xmath479 ( @xmath480 ) and @xmath180 reduces to normal form with at most @xmath331 @xmath8-reductions , 2 . if @xmath481 , @xmath482 $ ] and @xmath180 is @xmath483 then @xmath484 ( @xmath485 ) and @xmath180 reduces to normal form with at most m @xmath8-reductions , 3 . if @xmath486 , @xmath487 $ ] and @xmath488 , @xmath489 for @xmath322 , and @xmath180 is @xmath490 then @xmath484 ( @xmath485 ) and @xmath180 reduces to normal form with at most @xmath331 @xmath8-reductions . the proof is done by invoking as a measure a pair of integers , first the size of right term in @xmath248 and second the maximal number of @xmath8-reductions needed for @xmath180 to reduce to normal form ; the pair is ordered with the lexicographic ordering . the following properties are easy to show by case analysis . 1 . if @xmath472 @xmath331-holds then @xmath474 $ ] or for any next position @xmath491 there is an @xmath492 such that it @xmath492 holds and either @xmath493 or the size of the right term in @xmath494 is strictly smaller than in @xmath248 . 2 . if @xmath472 @xmath331-fails then @xmath475 $ ] or there is an @xmath492 and a next position @xmath491 that @xmath492-fails and either @xmath493 or the size of the right term in @xmath494 is strictly smaller than in @xmath248 . for instance , assume @xmath472 @xmath331-holds , @xmath495 , @xmath496 , @xmath497 for @xmath498 and @xmath499 $ ] . therefore , @xmath500 , @xmath501 $ ] and @xmath502 when @xmath503 . so , @xmath504 and by assumption @xmath505 . with @xmath24 @xmath8-reductions we obtain @xmath506 and position @xmath507 , therefore , @xmath508-holds . next , assume @xmath472 @xmath331-holds , @xmath481 , @xmath509 $ ] and @xmath488 for @xmath489 . by assumption , so , @xmath511 . consider any choice of next position . if @xmath512 then @xmath513 $ ] , @xmath514 and @xmath515 . therefore , @xmath516 and so this next position @xmath492-holds for some @xmath492 and @xmath517 is strictly smaller than @xmath352 . alternatively , @xmath518 . therefore , @xmath519 and @xmath520 @xmath492-holds for some @xmath492 provided that the @xmath368 s are new ( which is guaranteed as they are forbidden constants ) . so the next position @xmath492-holds and the right term of @xmath521 $ ] is strictly smaller than @xmath352 . assume @xmath472 @xmath331-holds and @xmath522 , @xmath487 $ ] , @xmath523 , @xmath524 , @xmath488 for @xmath489 and @xmath525 ; so , @xmath526 and @xmath527 . by assumption , @xmath528 . with @xmath24 @xmath8-reductions the left term in this equation becomes @xmath529 and so the next position @xmath508-holds . all other cases of ( 1 ) are close to one of these three , and the proof of ( 2 ) is also very similar . the only cases where the measure , size of right term in state and the number of @xmath8-reductions to normal form , does not decrease are applications of a3 when @xmath259 is labelled with a dummy lambda and c4 when @xmath377 of figure [ game ] . as a supplementary argument we show that there can not be an indefinite sequence of such applications of a3 followed by c4 by examining the index @xmath241 that is called at these positions ; namely , @xmath530 in the case of a3 and @xmath531 in the case of c4 ; this index must be strictly decreasing in such repeated sequence of applications of a3 followed by c4 . the result now follows : if @xmath82 then for each initial position that starts from an equation there is an @xmath331 such that it @xmath331-holds and for each disequation there is an @xmath331 such that it @xmath331-fails . conversely , if @xmath532 then there is an initial position for some equation that @xmath331-fails or for some disequation there is an @xmath331 such that it @xmath331-holds . the game is analogous to a model - checking game in the sense that deciding a possibly complex temporal property of a transition graph can be formulated as a game whose arena is the graph and where the moves are locally small steps that traverse it ; similarly , the complex property whether @xmath4 solves @xmath10 is here formalised as a game whose arena is @xmath4 involving locally small steps and local moves . in both cases , play proceeds until one definitely knows an outcome . in the following we let @xmath533 range over plays in a game @xmath473 . the total number of different plays is at most the sum of the number of branches in the right terms @xmath5 of @xmath10 . for instance , in the case of example [ examp44 ] whose right term is @xmath534 there is a single play . we now examine some properties of plays and introduce relationships between play positions that uses the play indices in the look - up tables . [ rem1 ] theorem [ thm1 ] allows one to restrict the set of constants that can appear in a potential solution term @xmath4 for @xmath10 . let @xmath535 be a new constant that does not occur in any right term @xmath5 of a ( dis)equation in @xmath10 ( and which is also not a forbidden constant ) . without loss of generality , we can assume that any potential solution term @xmath4 to @xmath10 only contains the constant @xmath362 and constants that occur in the right terms of @xmath10 : a similar observation is made in @xcite . the justification appeals to moves a1 and a2 of figure [ game ] . assume @xmath82 . control in a play associated with an interpolation equation in @xmath10 can never be at a node in @xmath4 labelled with a constant that does not occur in a right term ( as @xmath251 would win the play ) . if control in a play associated with an interpolation disequation in @xmath10 is at a node @xmath180 labelled with a constant that does not occur in a right term then replacing @xmath180 in @xmath4 with a single node labelled @xmath362 preserves @xmath251 s win . 1 . the _ length _ of @xmath536 , @xmath537 , is the number of positions @xmath538 in @xmath536 . 2 . the _ @xmath140th position _ of @xmath536 is @xmath539 where @xmath540 . the sequence of positions @xmath541 , @xmath542 , is written @xmath543 . we write @xmath544 , @xmath545 , @xmath546 and @xmath547 when @xmath548 and @xmath549 means that @xmath550 and @xmath551 . we shall describe a sequence of positions @xmath543 as an _ interval_. the _ right term _ of state @xmath223 $ ] , @xmath552 $ ] or @xmath214 $ ] is the term @xmath219 . each @xmath342 is a _ left term _ of @xmath223 $ ] and @xmath216 is the _ left term _ of @xmath214 $ ] . [ ri ] [ defi48 ] the interval @xmath543 is _ ri _ , _ right term invariant _ , if @xmath545 and @xmath553 share the same right term @xmath219 . it is _ nri _ if it is not ri and @xmath553 is not a final state . clearly , if @xmath543 is ri then _ every _ state at every position in this interval shares the same right term . for instance , when @xmath536 is the play of figure [ ex2 ] , the interval @xmath554 is ri as all its states share the right term @xmath298 ; also , each position in @xmath555 is the result of moves a3 or c4 . the outcome of the other moves in figure [ game ] , a1 , a2 , b1 , c1 , c2 and c3 , depend on the right term of the state . [ fact55 ] if @xmath543 is ri and @xmath556 is labelled @xmath181 for some @xmath184 then each position in @xmath557 is the result of move a3 or c4 of figure [ game ] . intervals that are ri do not directly contribute to the solution of @xmath10 . [ factri ] if @xmath558 is a sequence of positions that is ri , @xmath559 is labelled @xmath181 and @xmath560 is state q with right term @xmath561 instead of @xmath219 , then @xmath562 is also a sequence of positions that is ri . consider a position @xmath563 of a play . if there is a free occurrence of @xmath69 in the subtree @xmath4 then @xmath240 is defined ; similarly , if there is a free occurrence of a variable @xmath242 in a left term of @xmath564 then @xmath565 is defined . in contrast , if there is a bound occurrence of @xmath69 , a node labelled @xmath566 with @xmath567 , in the subtree @xmath4 then @xmath240 is not defined and similarly , if there is a bound occurrence of @xmath242 in a left term of @xmath564 then @xmath565 is not defined . [ imp ] assume @xmath568 . 1 . if @xmath566 labels a node in the subtree rooted at @xmath4 then for each @xmath569 , @xmath570 is undefined . if @xmath69 occurs free in the subtree rooted at @xmath4 then @xmath240 is defined . 3 . if @xmath571 $ ] or @xmath223 $ ] and @xmath572 occurs in @xmath216 or in some @xmath209 , @xmath573 , then for any @xmath574 , @xmath575 is not defined . if @xmath571 $ ] or @xmath223 $ ] and @xmath242 occurs free in @xmath216 or in some @xmath209 , @xmath569 , then @xmath565 is defined . we prove this by induction on the position @xmath140 in a play @xmath536 . we also show by induction on @xmath140 that if @xmath576 @xmath124 @xmath577 then the following additional four properties hold . 1 . for any @xmath69 , if @xmath578 and @xmath572 occurs in @xmath216 , then for any @xmath574 , @xmath579 is not defined . 2 . for any @xmath69 , if @xmath578 and @xmath242 occurs free in @xmath216 , then @xmath580 is defined . 3 . if @xmath581 and @xmath566 labels a node in the subtree rooted at @xmath582 then for @xmath583 , @xmath584 is undefined . 4 . if @xmath581 and @xmath69 occurs free in the subtree rooted at @xmath582 then @xmath585 is defined . for the base case , consider an initial position @xmath586 @xmath124 @xmath587 \ , \theta_{1 } \xi_{1}$ ] where @xmath247 is the root node of @xmath4 labelled @xmath67 . there are no free variables in @xmath4 or in the @xmath87 s . moreover , both @xmath231 and @xmath232 are empty ; therefore ( 1 ) to ( 4 ) hold . now we need to show the additional properties for @xmath588 : by definition @xmath589 and @xmath590 . therefore , these properties hold . consider next the general case for position @xmath539 . if @xmath539 is the result of moves a1-a3 of figure [ game ] applied to @xmath591 then @xmath457 and @xmath424 follow because they are true at @xmath591 and @xmath546 is an update of @xmath592 with respect to the ( potentially ) free variables @xmath593 such that @xmath594 is labelled @xmath260 . parts @xmath426 and @xmath429 trivially hold for a1 and a2 . in the case of a3 , if @xmath595 is labelled @xmath596 when @xmath594 is labelled @xmath260 then @xmath426 and @xmath429 follow from the induction hypothesis that they hold at @xmath591 ; otherwise they follow from the induction hypothesis for the first two additional properties at @xmath591 . if @xmath539 is the result of move b1 of figure [ game ] to @xmath591 then as @xmath1 , @xmath237 are unchanged @xmath597 remain true . finally , we examine the case when @xmath539 is the result of moves c1-c4 to @xmath591 . cases @xmath426 and @xmath429 hold because @xmath547 is a simple updating of @xmath598 where they hold . cases @xmath457 and @xmath424 hold for c1-c3 because @xmath546 and @xmath592 are the same . in the case of move c4 , @xmath457 and @xmath424 either follow from the induction hypothesis that they hold at @xmath591 or from the induction hypothesis for the final two additional properties at @xmath591 . using that @xmath597 are true at @xmath539 , the argument is similar for showing that the four additional properties hold at @xmath599 . now we examine some simple relationships between look - up tables . we allow @xmath335 , @xmath600 to range over both kinds of look - up tables . two look - up tables @xmath335 , @xmath334 are equal , @xmath601 , if , and only if , @xmath602 and for all @xmath603 , @xmath604 ; that is , if @xmath605 and @xmath606 then @xmath607 , @xmath608 and @xmath609 . this is well defined because the definitions of look - up tables @xmath224 and @xmath225 in definition [ defi42 ] are well - founded with respect to the embedding of look - up tables . a look - up table @xmath335 _ extends _ @xmath334 if for all @xmath610 , @xmath604 . assume @xmath536 is the play in figure [ ex15 ] that operates on the term tree in figure [ ex14 ] . the look - up table @xmath611 consists of three entries @xmath612 ; it therefore extends @xmath613 which is just the single entry @xmath614 . on the other hand , although @xmath615 extends @xmath616 it does not extend @xmath1 ; it consists of the entries @xmath617 . in similar fashion , @xmath618 which is @xmath619 extends @xmath620 which only contains the first two of these entries . let @xmath548 and let @xmath621 be a later position . if @xmath180 is a subtree of @xmath4 and @xmath616 extends @xmath1 then the free variable occurrences that are common to both @xmath4 and @xmath180 have the same interpretation ; their meaning is preserved at position @xmath241 . similarly , if @xmath622 extends @xmath11 then the free variable occurrences that are common to the left terms of @xmath564 and @xmath623 have the same interpretation . both look - up tables of positions that are the result of moves a1 , a2 , b1 , c1-c3 of figure [ game ] extend ( or are equal to ) those of the previous position . in the case of a3 the @xmath1 look - up table extends the one from the previous position but this is not true , in general , for the @xmath237 table . dually , in the case of c4 the @xmath237 table extends the one of the previous position but this may not hold for the @xmath1 look - up table . we want to restore when _ both _ look - up tables are extensions of an earlier position . for this we introduce a similar notion to that in game semantics that later positions are justified by earlier positions @xcite . we define when a later position is a child of an earlier position . it is at this point that we appeal to the third component of an entry in a look - up table . [ def410 ] position @xmath624 is a _ child _ of position @xmath539 if @xmath625 and the following by cases of which move @xmath626 is the result of 1 . a2 , b1 , c2 or c3 : then @xmath627 , 2 . a3 : then @xmath4 is labelled @xmath69 and @xmath628 , 3 . c4 : then @xmath214 \in \pi(j-1)$ ] , the head variable in @xmath216 is @xmath29 and @xmath629 . assume @xmath630 is the child of @xmath631 ; if @xmath626 is the result of a3 then @xmath180 is the binder of @xmath4 and if it is the result of c4 then @xmath4 is a successor of @xmath180 . [ fact411 ] assume @xmath626 is a child of @xmath539 . 1 . if @xmath632 \ , \theta ' \xi'$ ] is the result of a3 and @xmath633 \theta \ , \xi$ ] , then @xmath4 binds @xmath180 , @xmath634 and for some @xmath635 , @xmath636 . if @xmath637 \theta ' \xi'$ ] is the result of c4 , @xmath638 \in \pi(j-1)$ ] and @xmath639 \theta \xi$ ] then @xmath640 , @xmath641 and for some @xmath642 , @xmath643 and the head variable of @xmath644 is @xmath645 . [ child ] if @xmath646 then there is a unique @xmath647 such that @xmath626 is a child of @xmath539 . for this reason , we also say that @xmath539 is _ the parent _ of @xmath626 instead of @xmath626 is a child of @xmath539 . let @xmath536 be the play in figure [ ex15 ] which is on the tree in figure [ ex14 ] . every position that occurs at nodes @xmath424 and @xmath429 labelled with @xmath242 is a child of @xmath586 ; examples include @xmath648 , @xmath649 and @xmath650 . not every position that occurs at the nodes @xmath461 and @xmath196 labelled @xmath462 is a child of @xmath651 ; positions @xmath652 , @xmath653 and @xmath654 are whereas @xmath655 , @xmath656 and @xmath657 are not . nodes @xmath431 and @xmath461 are the successors of @xmath429 ; children of position @xmath658 at @xmath429 are @xmath659 , @xmath660 and @xmath661 while @xmath662 , @xmath663 and @xmath664 are not . position @xmath665 is a child of @xmath666 through a2 and @xmath667 is a child of @xmath668 through b1 . if @xmath626 is a child of @xmath539 , @xmath544 and @xmath669 then @xmath670 or there is a path of successors from @xmath4 to @xmath180 . the following is a critical consequence of the definition of a child position that _ both _ its look - up tables extend those of its parent . [ child1 ] assume @xmath626 is a child of @xmath539 . 1 . @xmath671 extends @xmath546 , 2 . @xmath672 extends @xmath673 . assume that @xmath626 is a child of @xmath539 and @xmath674 and @xmath675 . if @xmath626 is a child of @xmath539 as a result of b1 or c2 of figure [ game ] then the result is true because the look - up tables of @xmath539 and @xmath626 are the same . in the case of c3 , @xmath640 and @xmath676 or @xmath677 for some @xmath678 , @xmath372 , @xmath338 . therefore , @xmath679 extends @xmath237 as the @xmath372 s are not defined in @xmath237 by proposition [ imp ] . a similar argument applies when @xmath626 is the result of a2 ; now @xmath634 and @xmath640 or @xmath680 for some @xmath681 and @xmath336 , @xmath338 , and @xmath616 extends @xmath1 as the @xmath336 s are not defined in @xmath1 using proposition [ imp ] . next we examine the cases for a3 and c4 . by definition of a3 , the look - up table @xmath634 and by definition of c4 , @xmath640 . therefore , we just need to prove the result for the other look - up tables . in both cases the proof proceeds by case analysis of @xmath682 . the initial case is when @xmath683 . therefore , for a3 this means that @xmath669 is labelled with a variable @xmath69 which is bound by @xmath181 which labels @xmath684 ; so @xmath616 @xmath124 @xmath685 for some @xmath331 and @xmath686 , @xmath687 . consequently , @xmath616 extends @xmath1 using proposition [ imp ] ( since @xmath1 does not have an entry for @xmath69 ) . similarly , in the case of c4 when @xmath683 , @xmath688 \in \pi(i)$ ] for some @xmath322 and @xmath679 @xmath124 @xmath689 and @xmath237 does not have entries for the @xmath372 s using proposition [ imp ] . for the general case for a3 , we examine the branch between @xmath544 and @xmath669 where @xmath4 is labelled @xmath181 and @xmath180 is labelled @xmath69 ; which is the sequence of nodes with labels @xmath690 where each @xmath88 is a constant or variable and @xmath183 . consider position @xmath691 . clearly , @xmath616 extends @xmath692 by a similar argument to the base case . position @xmath693 is a child of @xmath694 for some @xmath695 with @xmath696 which is labelled @xmath697 by c4 , b1 or c2 : in all cases @xmath698 extends @xmath699 by definition of these moves . the argument continues for position @xmath700 . so , we reach a position @xmath701 with @xmath702 labelled @xmath703 . by assumption , @xmath704 extends @xmath705 : this means they have the same entry for the @xmath336 s in @xmath706 when they are bound by @xmath181 which is the label of @xmath544 . consider the relationship between @xmath701 and @xmath539 . clearly , it can not be the case that @xmath707 because this would contradict the entries in @xmath708 for the variables in @xmath184 . moreover , by definition of the moves in a , @xmath709 has the same entries for the @xmath69 s in @xmath184 as @xmath701 and @xmath710 . a small argument shows that @xmath701 must , therefore , be @xmath711 : otherwise , @xmath712 would not be a child of @xmath701 . consequently , @xmath616 extends @xmath1 . we now examine the general case for c4 . one possibility is that @xmath626 is the result of a sequence of c3 moves followed by c4 : clearly , in this case @xmath679 extends @xmath237 . otherwise , @xmath713 \in \pi(i)$ ] and @xmath714 is @xmath715 for some @xmath331 ; so @xmath716 extends @xmath237 . therefore , @xmath717 is labelled with some @xmath718 and @xmath719 has the form @xmath720 $ ] where @xmath216 is @xmath721 or @xmath722 for some @xmath24 : we know that @xmath679 extends @xmath723 . there may be a sequence of positions @xmath724 where each @xmath725 , @xmath726 is the result of c3 ( and , therefore , @xmath679 extends @xmath727 ) . otherwise @xmath728 . position @xmath694 is a child of a unique position @xmath729 labelled @xmath730 by a3 and so @xmath731 extends @xmath732 . this argument is now repeated : consider position @xmath733 labelled with @xmath734 . again , there may be a sequence @xmath735 where the moves are the result of c3 and either @xmath736 or @xmath737 is a child of @xmath738 . eventually , for some @xmath24 , @xmath739 as each @xmath731 has the entries @xmath740 for the @xmath202 s . [ def416 ] the binary relation _ is a descendent of _ on positions is the reflexive and transitive closure of `` is a child of '' . in this section , we connect the static structure , regions of a potential solution term of a ( dual ) interpolation problem , with the dynamics of game playing . to this end , partial subtrees of a term tree are introduced . assume @xmath741 . 1 . node @xmath180 labelled @xmath471 is an _ atomic leaf _ of type @xmath13 . 2 . node @xmath180 labelled @xmath742 is an _ atomic leaf _ of type @xmath17 when each @xmath743 3 . if @xmath180 is labelled @xmath744 then @xmath180 is a _ simple tile_. 4 . if node @xmath180 is labelled @xmath745 and each node @xmath746 is an atomic leaf of type @xmath747 and @xmath748 , @xmath489 , then @xmath749 is a _ simple tile_. a potential solution tree without its initial lambda is a tree of simple tiles . for instance , the region @xmath750 of figure [ ex1 ] is a simple tile labelled @xmath751 with atomic leaves @xmath426 and @xmath439 labelled @xmath752 and @xmath471 ; the region @xmath753 is labelled @xmath754 ; the region @xmath755 is a simple tile labelled @xmath756 . single nodes such as ( 16 ) and ( 20 ) are also simple tiles but without atomic leaves . the definition precludes node @xmath424 by itself or @xmath757 as simple tiles . in the following , as it makes the presentation cleaner , we describe tiles directly through their labelling . for example , @xmath751 identifies @xmath750 of figure [ ex1 ] . when there is ambiguity , such as with @xmath758 of figure [ ex14 ] , we disambiguate by describing the root node ; @xmath758 at @xmath461 versus @xmath758 at @xmath196 . tiles can be composed to form composite tiles . a ( possibly composite ) tile is a partial tree which can be extended at any of its atomic leaves . if @xmath759 is a tile with leaf @xmath760 and @xmath761 is a simple tile , then @xmath762 is the composite tile that is the result of placing @xmath761 directly beneath @xmath760 in @xmath763 . for instance , we can compose the tile @xmath754 of figure [ ex1 ] with the tile @xmath764 beneath @xmath765 and produce the composite tile @xmath766 which has three atomic leaves : in figure [ ex1 ] this tile is the region @xmath767 . we write @xmath768 if @xmath763 is a ( composite ) tile with atomic leaves @xmath769 . a tile @xmath768 is a multi - holed context . it is also a subregion of a term and we assume that the usual definitions of free and bound variable occurrences apply : for instance , the free variables in @xmath770 are the two occurrences of @xmath242 . later we shall manipulate tiles and , therefore , we have given them an independent existence . [ basic ] a ( composite ) tile @xmath763 is said to be _ basic _ if @xmath763 contains 1 . exactly one occurrence of a free variable and no occurrences of constants , or 2 . exactly one occurrence of a constant and no occurrences of free variables . by definition , simple tiles are basic . the single occurrence of the free variable or constant in a basic tile is its `` head '' element . particular contiguous regions of a term tree are basic tiles . in figure [ ex1 ] the region @xmath771 is a basic tile rooted at ( 14 ) with the single atomic leaf @xmath471 . however , if we also included node ( 18 ) then it would be a composite tile @xmath772 without atomic leaves , but not a basic tile . [ defi53 ] assume @xmath763 and @xmath773 are basic tiles in a tree . 1 . @xmath773 is _ @xmath768 if there is a path of successors from @xmath774 to @xmath773 . 2 . @xmath773 is _ below _ @xmath763 ( or @xmath763 is _ above _ @xmath773 ) if @xmath773 is @xmath241-below @xmath763 for some @xmath241 . in figure [ ex14 ] , the tile @xmath775 , rooted at ( 8) , is @xmath21-below @xmath776 and @xmath777 at ( 20 ) is @xmath157-below the same tile . [ defdepend ] assume @xmath763 and @xmath773 are basic tiles in a tree . 1 . @xmath773 is an _ immediate @xmath241-dependent _ of tile @xmath763 if @xmath773 is @xmath241-below @xmath763 and @xmath773 contains a free variable that is bound in @xmath763 . 2 . @xmath773 is a _ @xmath241-dependent _ of @xmath763 if it is an immediate @xmath241-dependent of @xmath763 or there is a @xmath761 that is an immediate @xmath241-dependent of @xmath763 and @xmath773 is a @xmath24-dependent of @xmath761 for some @xmath24 . @xmath773 is a ( _ immediate _ ) _ dependent _ of @xmath763 if @xmath773 is a ( immediate ) @xmath241-dependent of @xmath763 for some @xmath241 . the tile @xmath758 rooted at ( 6 ) of figure [ ex14 ] is an immediate @xmath21-dependent of @xmath776 rooted at ( 2 ) and @xmath777 rooted at ( 20 ) is a @xmath157-dependent . given a tile @xmath763 , its set of dependents are all the tiles below it whose free variables are either bound within @xmath763 or are bound within a dependent of @xmath763 . for instance , the dependents of @xmath778 of figure [ ex14 ] are @xmath775 rooted at ( 8) and @xmath779 at ( 16 ) . [ family ] tiles @xmath763 and @xmath773 belong to the _ same family _ in a tree if one is a dependent of the other , or there is a @xmath761 such that both are dependents of @xmath761 . the _ family _ of tiles associated with @xmath763 in a tree consists of @xmath763 and each tile @xmath773 that belongs to the same family as @xmath763 . in figure [ ex14 ] the family associated with @xmath779 at ( 16 ) is the set of tiles containing @xmath778 at ( 4 ) , @xmath775 at ( 8) and @xmath779 . assume basic tiles @xmath763 and @xmath773 . if position @xmath626 is the result of move a3 of figure [ game ] and is a child of @xmath539 and @xmath669 is in @xmath773 and @xmath544 is in @xmath763 then @xmath773 is a dependent of @xmath763 . if @xmath626 is the result of move c4 and @xmath669 is in @xmath773 and @xmath780 is in @xmath763 , then these two tiles belong to the same family : this important property is proved in section [ unfold ] . in the play of figure [ ex15 ] , play at @xmath781 at position 16 jumps to @xmath782 of @xmath778 and play at @xmath779 at position 20 jumps to @xmath783 . [ defi54 ] assume @xmath763 and @xmath773 are basic tiles that each contain an occurrence of a free variable . 1 . @xmath763 and @xmath773 are _ equivalent _ , written @xmath784 , if they are @xmath41-equivalent ; that is , they are the same basic tiles with the same number and type of atomic leaves and with the same single free variable occurrence @xmath69 . if @xmath784 and @xmath247 is a node of @xmath763 and @xmath785 is the corresponding equivalent node of @xmath773 then we write @xmath786 . if @xmath784 and @xmath773 is below @xmath763 then @xmath773 is said to be an _ embedded _ tile . the reader can verify that `` tile equivalence '' is indeed an equivalence relation . in figure [ ex1 ] , each pair of tiles @xmath751 , @xmath787 , @xmath754 and @xmath764 is equivalent : nodes such as those with labels @xmath788 and @xmath789 correspond . each of these tiles except the first is also an embedded tile . the tile @xmath758 rooted at ( 6 ) is equivalent to @xmath758 rooted at ( 10 ) of figure [ ex14 ] : however , neither is equivalent to @xmath775 rooted at ( 8) . [ static ] assume @xmath790 is a basic tile . 1 . @xmath763 is a _ top _ tile if it contains a free variable occurrence that is is bound by the initial lambda of the term tree . 2 . @xmath763 is _ @xmath241-end _ if @xmath763 has no immediate @xmath241-dependents . it is an _ end _ tile if it is @xmath241-end for _ all _ @xmath791 . 3 . @xmath763 is a _ constant _ tile if it contains an occurrence of a constant or it is a dependent of a constant tile . the tile @xmath792 rooted at ( 2 ) in figure [ ex1 ] is a top tile as @xmath242 is bound by the initial lambda at node ( 1 ) . it is also @xmath157-end because no tile beneath node ( 11 ) contains a free variable occurrence that is bound within it : however , it is not @xmath21-end because of the tile occurrence @xmath29 at node ( 8) . tile @xmath787 rooted at ( 6 ) is a top tile and also an end tile . tiles @xmath793 rooted at ( 6 ) , @xmath794 at ( 8) and @xmath777 at ( 12 ) in figure [ picture2 ] are all constant tiles . the previous definitions provide a classification of basic tiles within a tree that only appeals to the static structure of the tree . tiles can also be categorised in terms of dynamic properties of game playing . the interval @xmath543 is a _ play _ on the simple tile @xmath795 if @xmath796 , @xmath797 for some @xmath798 and @xmath626 is a child of @xmath539 . m - play _ if @xmath797 . a simple tile has the form @xmath799 or @xmath800 . a play @xmath543 on such a tile starts at the head of the tile and ends at one of its atomic leaves ; importantly , @xmath626 must be a child of @xmath539 . a play on a simple constant tile @xmath801 is a consecutive pair of positions @xmath802 with @xmath796 and @xmath803 for some @xmath331 ( by moves b1 or c2 of figure [ game ] ) . [ fact57 ] if @xmath543 is a play on a simple constant tile then @xmath683 . for instance , if @xmath536 be the play in example [ examp42 ] whose term tree is depicted in figure [ picture2 ] then @xmath804 is a play on the simple constant tile @xmath805 and @xmath806 is a play on the simple constant tile @xmath794 . a play @xmath543 on a simple non - constant tile @xmath799 can have arbitrary length . it starts at the node labelled with @xmath69 and finishes at a node labelling an atomic leaf @xmath807 . in between , control can be almost anywhere in the tree ( including the node labelling @xmath69 ) . however , because @xmath626 is a child of @xmath539 the look - up tables of @xmath626 extend those of @xmath539 as shown in proposition [ child1 ] . let @xmath536 be the play in figure [ ex15 ] on the tree in figure [ ex14 ] . the simple tile @xmath808 is rooted at ( 2 ) . there are various plays on @xmath763 : @xmath21-plays include @xmath809 and @xmath810 ; @xmath811 and @xmath812 are @xmath157-plays . if @xmath813 rooted at ( 4 ) then @xmath814 and @xmath815 are @xmath21-plays on @xmath773 : however , the interval @xmath816 is not a play on @xmath773 because @xmath662 is a child of @xmath649 . if @xmath763 is the end tile @xmath775 rooted at ( 8) then there is just one play @xmath817 on it ; the interval @xmath818 is not a play on @xmath763 . the definition of play on a simple tile can be extended to arbitrary composite tiles by composing consecutive plays on the simple tiles from its root to one of its atomic leaves . [ def612 ] the interval @xmath543 is a play on the composite tile @xmath819 if there is a path of successor nodes with labels @xmath820 from the root of @xmath763 to an atomic leaf @xmath821 such that @xmath822 and @xmath823 is a play on the simple tile @xmath824 with @xmath825 for @xmath826 . it is an _ @xmath331-play _ if @xmath797 . if @xmath827 , @xmath828 , and @xmath4 is labelled with a lambda then there is a unique partition of @xmath829 into plays on the simple tiles that occur on the branch from the node directly beneath the initial @xmath471 of the tree to @xmath4 . the partition also preserves children . [ prop511 ] assume @xmath827 and @xmath4 is labelled with @xmath830 and @xmath831 is the labels of the sequence of nodes from the root of the tree to @xmath4 . then there is a unique partition of @xmath832 such that for @xmath833 1 . @xmath834 and @xmath835 is a play on the simple tile @xmath836 with @xmath837 , 2 . if @xmath838 is a variable bound by @xmath839 then @xmath840 is a child of @xmath841 . assume @xmath827 and @xmath4 is labelled with @xmath830 and @xmath842 are the labels of the sequence of nodes from the root of the term tree to @xmath830 . assume @xmath843 . let @xmath844 be the play on the simple tile @xmath845 such that @xmath694 is a child of @xmath846 : move @xmath694 is the result of b1 , c2 or c4 of figure [ game ] ( and , therefore , @xmath846 is uniquely defined from definition [ def410 ] ) . let @xmath847 , the argument is now repeated for @xmath848 as a play on @xmath849 and so on for subsequent tiles in the branch from the node labelled @xmath830 to the root . clearly , this will define a partition of @xmath850 ) into @xmath851 with @xmath852 : hence , we can add @xmath853 with @xmath854 at the beginning . next assume that @xmath838 is a variable bound by @xmath839 . it is straightforward to show that the look - up tables in @xmath840 extend those in @xmath855 , which , therefore , implies that @xmath840 is a child of @xmath841 ; this follows from repeated application of proposition [ child1 ] . [ def614 ] if @xmath827 and @xmath4 is labelled with @xmath830 and @xmath831 is the labels of the sequence of nodes from the root of the tree to @xmath4 then the unique partition @xmath832 such that for @xmath833 where each @xmath835 is the play on @xmath856 of proposition [ prop511 ] is called the _ b - partition for position @xmath626_. [ example512 ] let @xmath536 be the play in figure [ ex15 ] on the tree in figure [ ex14 ] . consider the two positions @xmath857 and @xmath858 where @xmath471 is at node ( 7 ) . below we illustrate the b - partitions for @xmath859 and @xmath860 . the branch from the root to @xmath471 is presented horizontally . @xmath861 the b - partitions capture children along their branches : @xmath648 and @xmath658 are children of @xmath586 ; @xmath651 is a child of @xmath648 ; @xmath659 and @xmath661 are children of @xmath658 ; @xmath652 and @xmath653 are children of @xmath651 . plays within these b - partitions may jump around the term tree : for instance , the play @xmath862 on @xmath758 rooted at ( 10 ) includes a sequence of moves that descends the branch below ( 17 ) . we now exhibit some useful uniformities of plays on tiles . given a play @xmath536 and tile @xmath763 we examine three aspects . first , there can be multiple plays @xmath863 , @xmath9 , @xmath864 on @xmath763 from the same initial position . second , there can be multiple plays from different starting positions @xmath865 , @xmath9 , @xmath866 on @xmath763 . a third consideration is the relationship between plays on @xmath761 and @xmath763 when these tiles are equivalent . we start with a pertinent property of a simple non - constant tile . [ plays ] if @xmath867 and @xmath868 are plays on the simple tile @xmath869 and @xmath870 then there is a position @xmath871 , @xmath872 , that is a child of @xmath873 . assume @xmath874 \ , \theta \ , \xi$ ] , @xmath4 is labelled @xmath69 and @xmath867 , @xmath868 , @xmath875 , are plays on the simple tile @xmath869 ; so , @xmath876 for some @xmath241 . the look - up table @xmath714 is @xmath877 and the @xmath237 look - up tables in @xmath878 and @xmath879 both extend @xmath716 because @xmath873 , @xmath880 are children of @xmath539 . clearly , no look - up table @xmath881 , @xmath882 , has entries @xmath883 because for any @xmath242 , if @xmath884 then @xmath885 . consider the first position @xmath886 after @xmath873 that is at a node labelled with a variable @xmath425 . clearly , this node is below the node labelled @xmath774 in the tree . if @xmath425 is bound by @xmath774 then @xmath886 is a child of @xmath873 and the result is proved . otherwise @xmath886 is a child of an earlier position @xmath887 . either @xmath888 or @xmath889 . assume the former ; so , via move a3 the look - up table @xmath881 can not have an entry for any @xmath242 of the form @xmath890 because @xmath891 . play may then jump anywhere in the term tree using move c4 . if there is not a play @xmath892 on the simple tile whose head node is labelled @xmath425 then for all later positions @xmath893 , @xmath894 , @xmath895 can not include an entry for any @xmath242 of the form @xmath890 ; this is a contradiction because of the position @xmath880 . therefore , play must reach a position @xmath896 that is a child of @xmath886 ; so @xmath897 is a successor of the node labelled with @xmath425 and is therefore below ( the node labelled ) @xmath774 . next assume that @xmath898 so @xmath425 is bound by a @xmath181 that is below ( the node labelled ) @xmath774 . but then @xmath425 is bound to a leaf of a constant tile that occurs between the node labelled @xmath774 and the node labelled @xmath425 and so move c2 must apply and so play descends to a successor with position @xmath896 that is a child of @xmath886 where @xmath899 . this argument is now repeated for the next position after @xmath896 that is at a node labelled with a variable @xmath900 : this node must be below the node labelled @xmath774 . the argument proceeds as above , except there is the new case that @xmath893 is a child of @xmath896 . however , by move a3 , @xmath895 can not include an entry for some @xmath242 of the form @xmath901 . therefore , play must eventually reach a child of @xmath873 . by iteration , this property extends to basic tiles . [ plays1 ] if @xmath867 and @xmath868 are plays on the basic tile @xmath768 , @xmath870 , and @xmath902 where @xmath903 is the path of ( labels of ) nodes from the root of @xmath763 to its atomic leaf @xmath904 and @xmath823 is a play on the simple tile @xmath905 with @xmath906 for @xmath826 then there is a position @xmath492 such that @xmath907 and @xmath871 is a child of @xmath908 for some @xmath644 . one consequence of corollary [ plays1 ] is that there can not be more than one play on a basic end tile that starts from the same position . more precisely , we have the following . [ end1 ] if @xmath867 is a @xmath241-play on the basic tile @xmath763 which is @xmath241-end , @xmath868 is a play on @xmath763 and @xmath909 then @xmath910 . assume @xmath911 is @xmath241-end . so @xmath763 has no dependents below the node labelled @xmath774 ; therefore , the binder of every free variable that occurs in the subtree below this node occurs above the tile @xmath763 in the tree . assume @xmath867 is a @xmath241-play on @xmath763 and @xmath868 , @xmath875 , is also a play on @xmath763 . the play @xmath867 can be partitioned into plays on the simple tiles of @xmath763 from its root to its atomic leaf @xmath876 . therefore , by corollary [ plays1 ] there is a position @xmath871 , @xmath907 , that is a child of one of the final positions of a simple tile in the partition of @xmath867 . but then , by the definition of child , see fact [ fact411 ] , there must be a free variable below the node labelled @xmath774 that is bound within @xmath763 which contradicts the assumption that @xmath763 is @xmath241-end . special restricted plays of basic tiles are defined . [ defi516 ] assume @xmath543 is a play on basic tile @xmath763 . 1 . shortest play _ on @xmath763 if no proper prefix @xmath912 , @xmath913 , is also a play on @xmath763 . 2 . it is a _ shortest @xmath331-play _ on @xmath763 if @xmath543 is a @xmath331-play on @xmath763 and no proper prefix @xmath912 , @xmath914 , is also an @xmath331-play on @xmath763 . it is also an _ play ( @xmath331-play ) on @xmath763 provided that for any @xmath915 , if @xmath916 then @xmath4 is a node of @xmath763 . it is also an _ ri _ play ( @xmath331-play ) on @xmath763 if @xmath543 is ri ( see , definition [ ri ] ) . let @xmath536 be the play in figure [ ex2 ] . the interval @xmath809 is a shortest play on tile @xmath751 rooted at ( 2 ) of figure [ ex1 ] : this play is also ri , internal and a shortest @xmath21-play . although @xmath917 is a shortest @xmath157-play , it is neither a shortest play nor an internal play . next , we define a uniformity condition concerning multiple plays on a basic tile . assume @xmath763 is a basic tile whose root node is @xmath180 . tile @xmath763 is _ @xmath241-directed with respect to the interval @xmath918 _ if 1 . @xmath919 for all @xmath920 , or 2 . @xmath873 is the first position @xmath921 with @xmath922 and there is a shortest @xmath241-play @xmath923 on @xmath763 , @xmath923 is ri and @xmath763 is @xmath241-directed with respect to @xmath924 . [ directed ] the basic tile @xmath763 is _ @xmath241-directed _ with respect to the game @xmath925 if @xmath763 is @xmath241-directed with respect to @xmath926 for every play @xmath927 . if @xmath763 is @xmath241-directed with respect to a game then every play @xmath536 contains a ( unique ) sequence of ri intervals @xmath928 , for some @xmath322 , which are shortest @xmath241-plays on @xmath763 as follows ( assuming @xmath180 is the root of @xmath763 and @xmath774 is its @xmath241th atomic leaf ) . @xmath929 by definition , @xmath180 can not occur in a position that is outside of these intervals @xmath928 . a tile @xmath763 can be @xmath241-directed with respect to a game for multiple @xmath241 . if @xmath536 is the play of figure [ ex15 ] on the tree in figure [ ex14 ] then tile @xmath778 is @xmath21-directed with respect to @xmath926 because of the following sequence : @xmath930 however , it is not @xmath157-directed with respect to @xmath926 because @xmath931 is nri . recall that a basic top tile , see definition [ static ] , has a single variable occurrence that is bound by the initial lambda of the term tree . equivalent basic top tiles are subject to the following local uniformity properties . [ prop515 ] assume @xmath763 , @xmath773 are basic top tiles , @xmath784 and @xmath932 is a shortest play on @xmath763 . 1 . @xmath932 is an internal play on @xmath763 . if @xmath932 is ri and @xmath933 is a shortest play on @xmath773 then @xmath933 is ri , @xmath934 and for all @xmath935 , @xmath936 implies @xmath937 where @xmath786 . if @xmath932 is ri and a @xmath241-play on @xmath763 then @xmath763 and @xmath773 are @xmath241-directed with respect to @xmath926 . if @xmath932 is nri and @xmath938 where @xmath4 is the root node of @xmath773 then there is an interval @xmath939 which is internal to @xmath773 that is either a shortest play on @xmath773 and nri or @xmath940 . let @xmath4 be the root node of top tile @xmath763 and assume @xmath544 . the state @xmath545 has the form @xmath941 $ ] where @xmath47 is a closed left term , a @xmath942 from a ( dis)equation of @xmath10 . therefore , a shortest play @xmath943 on @xmath763 is internal ( as a jump outside @xmath763 requires there to be a free variable in @xmath47 via move c4 of figure [ game ] ) which shows ( 1 ) . if @xmath784 , @xmath180 is the root of @xmath773 and @xmath944 then the state @xmath945 has the same left term @xmath946 $ ] as @xmath941 \in \pi(i)$ ] . if the play @xmath932 is ri then there is a corresponding ri play @xmath947 on @xmath773 consisting of the same sequence of corresponding positions in @xmath763 and states ( except for their right terms @xmath219 and @xmath561 ) which shows ( 2 ) . furthermore , tile @xmath763 is @xmath241-directed with respect to @xmath536 when @xmath948 is a @xmath241-play on @xmath763 because each time play is at @xmath4 there is the same repeated shortest play on it ( and similarly for @xmath773 ) . for ( 4 ) , if the shortest play @xmath932 on @xmath763 is nri and @xmath944 then either there is a shortest play on @xmath773 that is nri ( but may involve different @xmath251-choices ) or play remains within @xmath773 and the final state is reached . tiles are equivalent to themselves , @xmath949 ; so proposition [ prop515 ] also applies to repeated shortest plays on a single top tile @xmath763 . consider the tree in figure [ ex14 ] and the play in figure [ ex15 ] . the equivalent tiles @xmath950 are top tiles . shortest @xmath536 plays on the first of these tiles @xmath814 , @xmath815 , @xmath951 and @xmath952 are @xmath21-plays that are ri and each corresponds to the single shortest play @xmath809 on @xmath776 . top basic tiles are distinguished because their only free variable is bound by the initial lambda . we now show that there are also play uniformities on other equivalent tiles , in the case of _ embedded _ tiles ( as in definition [ defi54 ] ) . if @xmath784 and @xmath773 is below @xmath763 then shortest plays on the embedded tile @xmath773 are constrained by earlier plays on @xmath763 . [ prop516 ] assume @xmath763 , @xmath773 are basic tiles , @xmath784 and @xmath773 is below @xmath763 . 1 . if @xmath953 is a shortest play on @xmath773 that is ri and a @xmath24-play then there is a shortest play @xmath932 on @xmath763 , @xmath954 , that is ri and a @xmath24-play . if @xmath955 is a shortest play on @xmath773 that is nri then there is a shortest play @xmath956 on @xmath763 , @xmath954 , that is nri . assume @xmath763 and @xmath773 are basic equivalent tiles , @xmath784 , @xmath773 is below @xmath763 and @xmath955 is a shortest play on @xmath773 that is ri and a @xmath24-play . consider the unique b - partition of position @xmath957 from the root of the term tree to the atomic leaf of @xmath773 , see definition [ def614 ] . because @xmath773 is below @xmath763 , this b - partition contains a play @xmath958 on @xmath763 . because @xmath763 and @xmath773 are equivalent and @xmath955 is ri , these tiles contain the same single free variable @xmath69 at their root ; therefore , positions @xmath539 and @xmath626 share a common parent . therefore , @xmath547 is the same look - up table as @xmath959 , @xmath704 extends @xmath546 and if @xmath720 \in \pi(j)$ ] then @xmath214 \in \pi(i)$ ] for some @xmath219 . now , it is an easy argument that @xmath932 is ri and a @xmath24-play on @xmath763 . if instead @xmath955 is nri then some prefix of @xmath960 is a shortest nri play on @xmath763 that may involve different @xmath251 choices than in @xmath955 . in the case of the tree in figure [ ex14 ] and the play in figure [ ex15 ] , tile @xmath758 rooted at ( 10 ) is an embedded ( end ) tile : its single play @xmath961 which is nri corresponds to the earlier play @xmath862 on @xmath758 rooted at ( 6 ) . there is an even stronger property of embedded end tiles : an embedded end tile is either @xmath241-directed with respect to a game for some @xmath241 , has at least one nri play or play finishes within it . [ end ] assume @xmath763 , @xmath773 are basic end tiles in a term tree @xmath4 , @xmath773 is @xmath241-below @xmath763 , @xmath784 and @xmath962 . then either 1 . @xmath773 is @xmath241-directed with respect to @xmath926 , or 2 . there exist @xmath963 such that @xmath964 , @xmath923 is a nri @xmath241-play on @xmath763 , @xmath965 is the root node of @xmath773 and @xmath966 is a nri play on @xmath773 or @xmath967 and no prefix of @xmath966 is a play on @xmath773 . assume @xmath763 , @xmath773 are basic end tiles in term tree @xmath4 , @xmath784 and and @xmath773 is @xmath241-below @xmath763 . we follow the argument in the proof of proposition [ prop516 ] . both @xmath763 and @xmath773 have the same head variable @xmath69 that is bound to the same @xmath181 above @xmath763 in @xmath4 . let @xmath968 . consider the first position @xmath922 where @xmath180 is the root of @xmath773 and the b - partition for @xmath878 . this partition must contain a @xmath241-play on @xmath763 , @xmath969 ( because @xmath773 is @xmath241-below @xmath763 ) ; positions @xmath539 and @xmath873 share a common parent . if @xmath970 is play on @xmath773 then the b - partition for @xmath871 is a simple extension of that for @xmath878 ; the play on @xmath763 is still @xmath969 . by proposition [ end1 ] , @xmath969 is a @xmath241-play and so it is a corresponding play to @xmath970 by proposition [ prop516 ] ; so , @xmath970 is ri and is a @xmath241-play on @xmath773 . this argument is repeated for the next position , @xmath971 , such that @xmath972 or until the @xmath241-play on @xmath763 in the b - partition is nri ; in which case , there is either an nri play on @xmath773 or a final state is reached . given an interpolation problem @xmath10 , the aim is to prove decidability of higher - order matching by establishing a small model property : if @xmath82 then there is a _ small _ term @xmath973 ( whose size is determined from @xmath10 ) . as we saw in section [ props ] , the number of plays in the game @xmath203 is bounded by the number of branches in the right terms @xmath5 of the ( dis)equations of @xmath10 . however , there is no upper bound on the length of a play that is independent of the size and order of @xmath4 . nevertheless , a long play must contain significant ri intervals that do not directly contribute to solving @xmath10 : the number of times the right term in a state changes within all plays is bounded by @xmath165 , the right size of @xmath10 , definition [ delta ] . game - theoretically , what will underpin the conversion of an arbitrary solution term @xmath4 to a small solution term @xmath180 is manipulation of tiles and their ri plays ( using omission , repetition and permutation ) . the proof for the general case is quite intricate . so , we proceed in stages . in this section we provide two proofs of decidability of @xmath12rd - order matching , one using a tree model property of game playing and the other using properties of embedded end tiles . both proofs appeal to omission of tiles and their ri intervals . as a first step , we introduce two transformations on term trees ( for terms of any order ) . a transformation @xmath974 converts a term tree @xmath4 into a term tree @xmath180 , written @xmath975t@xmath976 . assume @xmath180 is a subtree of @xmath4 whose root is labelled with a variable @xmath69 or a constant @xmath977 . the game @xmath473 _ avoids _ @xmath180 if for every play @xmath968 , @xmath978 for all @xmath979 . assume @xmath180 and @xmath582 are trees whose roots are labelled with a constant or a variable . let @xmath980 $ ] be the result of replacing the subtree @xmath180 of @xmath4 with the tree @xmath582 . the first transformation is straightforward : if no play enters a subtree of @xmath4 then it can be replaced with the single node labelled with constant @xmath981 ( introduced in remark [ rem1 ] ) . 1 . if @xmath473 avoids @xmath180 then transform @xmath4 to @xmath982 $ ] the second transformation removes inner tiles from @xmath4 : if a basic tile is both @xmath241-end , definition [ static ] , and @xmath241-directed with respect to the game @xmath203 , definition [ directed ] , then it is redundant . 1 . if @xmath768 is a @xmath241-end basic tile and @xmath241-directed with respect to @xmath473 , @xmath180 is the subtree of @xmath4 rooted at @xmath763 and @xmath746 is the subtree directly beneath @xmath774 of @xmath763 then transform @xmath4 to @xmath983 $ ] . an application of * t2 * not only removes the tile @xmath768 from @xmath4 but also all subtrees that occur directly beneath any atomic leaf @xmath984 , @xmath985 , of @xmath763 . because @xmath763 has no @xmath241-dependents , all free variables that occur in the subtree @xmath746 directly below @xmath774 are bound above @xmath763 in @xmath4 ; therefore , the result of applying * t2 * is still a closed term ( in normal form with the right type ) . if @xmath763 is @xmath241-directed with respect to a game then each play involves a ( unique ) sequence of ri intervals which are shortest @xmath241-plays on @xmath763 , as described in the previous section . game - theoretically underpinning the correctness of * t2 * is _ omission _ of these inessential intervals from each play . if @xmath986 , @xmath987ti@xmath976 and @xmath82 then @xmath988 . this is clear in the case of @xmath989 . assume @xmath82 , @xmath768 is @xmath241-end and @xmath241-directed with respect to the game @xmath473 , @xmath582 is the subtree at the root of @xmath763 and @xmath746 is the subtree directly beneath @xmath774 . let @xmath990 $ ] . we show that @xmath988 . we convert each @xmath968 into a play @xmath991 that ends with the same final state . because @xmath763 is @xmath241-directed with respect to each play , @xmath536 can be split uniquely into the following regions @xmath992 where each @xmath993 is a ( shortest ) @xmath241-play on @xmath763 and is ri ; by definition of @xmath241-directed , node @xmath582 can not occur outside of these intervals . therefore , @xmath994 is a child of @xmath995 for each @xmath24 consequently , the play @xmath996 is just the outer intervals @xmath997 ( modulo the changes to the entries in the look - up tables ) because for each @xmath216 , @xmath998 is a child of @xmath999 . we show , that if @xmath1000 is a node in @xmath763 or is @xmath331-below an atomic leaf @xmath807 , @xmath1001 , of @xmath763 then @xmath1000 can not occur in any outer interval of @xmath536 . if @xmath1000 were to appear in such a position then move c4 must have applied : there is then a variable @xmath69 and a position in an outer region @xmath1002 and @xmath1003 and @xmath1004 and there is a free variable @xmath242 in @xmath216 such that @xmath1005 . however , this is impossible . consider @xmath1006 : clearly , there is no free variable in the subtree rooted at @xmath582 with this property . when play reaches @xmath886 because @xmath763 is a @xmath241-end tile and because the look - up tables in @xmath886 extend those in @xmath701 there can not be a free variable in the subtree @xmath746 with this property either . this argument is now repeated for subsequent positions @xmath995 and @xmath994 . the two transformations are also _ reversible _ : we can add gratuitous subtrees and intersperse redundant @xmath241-end tiles with arbitrary subtrees beneath their other atomic leaves in any solution term . [ examp61 ] consider the @xmath7th - order problem @xmath1007 where @xmath1008 and @xmath1009 from example [ examp1 ] . a solution term is in figure [ ex1 ] . there are two plays , @xmath536 in figure [ ex2 ] for the first equation and @xmath1010 in figure [ nex2 ] for the second . @xmath1011\ , \theta_{1 } \ , \xi_{1 } & & & \\ ( 2 ) \ , q [ w , f(fa ) ] \ , \theta_{2 } \xi_{2 } & \ ! \theta_{2 } = \theta_{1 } { \ { w \xi_{1}1\!/\!z \ } } & \ ! \xi_{2 } = \xi_{1 } & \ ! a3\\ ( 3 ) \ , q[(y_{3}y_{4 } ) , f(f a ) ] \ , \theta_{3 } \xi_{3 } & \ ! \theta_{3 } = \theta_{2 } & \ ! \xi_{3 } = \xi_{1 } { \ { ( 3)\theta_{2}2\!/\!y_{3},(11)\theta_{2}2\!/\!y_{4 } \ } } & \ ! c4\\ ( 4 ) \ , q[- , f(f a ) ] \ , \theta_{4 } \xi_{4 } & \ ! \theta_{4 } = \theta_{2 } { \ { ( y_{3 } y_{4 } ) \xi_{3 } 3\!/\!x \ } } & \ ! \xi_{4 } = \xi_{3 } & \ ! a2\\ ( 5 ) \ , q [ ( \ ) , f a ] \ , \theta_{5 } \ , \xi_{5 } & \ ! \theta_{5 } = \theta_{4 } & \ ! \xi_{5 } = \xi_{3 } & \ ! b1\\ ( 6 ) \ , q[w , fa ] \ , \theta_{6 } \ , \xi_{6 } & \ ! \theta_{6 } = \theta_{4 } & \ ! \xi_{6 } = \xi_{1 } & \ ! a3\\ ( 7 ) \ , q[(y_{3}y_{4 } ) , f a ] \ , \theta_{7 } \ , \xi_{7 } & \ ! \theta_{7 } = \theta_{4 } & \ ! \xi_{7 } = \xi_{1 } { \ { ( 7)\theta_{4}6\!/\!y_{3},(9)\theta_{4}6\!/\!y_{4 } \ } } & \ ! c4\\ ( 8) \ , q[y_{3 } y_{4 } , fa ] \ , \theta_{8 } \ , \xi_{8 } & \ ! \theta_{8 } = \theta_{4 } { \ { y_{2 } \xi_{7 } 7\!/\!u \ } } & \ ! \xi_{8 } = \xi_{3 } & \ ! a3\\ ( 3 ) \ , q[(y_{4 } ) , f a ] \ , \theta_{9 } \xi_{9 } & \ ! \theta_{9 } = \theta_{2 } & \ ! \xi_{9 } = \xi_{3 } & \ ! c4\\ ( 4 ) \ , q[- , , \theta_{10 } \xi_{10 } & \ ! \theta_{10 } = \theta_{2 } { \ { y_{4}\xi_{3}9\!/\!x \ } } & \ ! \xi_{10 } = \xi_{3 } & \ ! a2\\ ( 5 ) \ , q [ ( \ ) , a ] \ , \theta_{11 } \ , \xi_{11 } & \ ! \theta_{11 } = \theta_{10 } & \ ! \xi_{11 } = \xi_{3 } & \ ! b1\\ ( 6 ) \ , q[w , a ] \ , \theta_{12 } \ , \xi_{12 } & \ ! \theta_{12 } = \theta_{10 } & \ ! \xi_{12 } = \xi_{1 } & \ ! a3\\ ( 7 ) \ , q[(y_{3 } y_{4 } ) , a ] \ , \theta_{13 } \ , \xi_{13 } & \ ! \theta_{13 } = \theta_{10 } & \ ! \xi_{13 } = \xi_{1 } { \ { ( 7)\theta_{10}12 \!/\!y_{3},(9)\theta_{10 } 12\!/\!y_{4 } \ } } & \ ! c4\\ ( 8) \ , q[y_{4 } , a ] \ , \theta_{14 } \ , \xi_{14 } & \ ! \theta_{14 } = \theta_{10 } { \ { y_{4 } \xi_{13 } 13\!/\!u \ } } & \ ! \xi_{14 } = \xi_{3 } & \ ! a3\\ ( 11 ) \ , q [ ( \ ) , a ] \ , \theta_{15 } \ , \xi_{15 } & \ ! \theta_{15 } = \theta_{2 } & \ ! \xi_{15 } = \xi_{3 } & \ q [ w , a ] \ , \theta_{16 } \xi_{16 } & \ ! \theta_{16 } = \theta_{2 } & \ ! \xi_{16 } = \xi_{1 } & \ ! a3\\ ( 13 ) \ , q[(y_{3 } y_{4 } ) , a ] \ , \theta_{17 } \xi_{17 } & \ ! \theta_{17 } = \theta_{2 } & \ ! \xi_{17 } = \xi_{1 } { \ { ( 13)\theta_{2 } 16\!/\!y_{3},(19)\theta_{2}16\!/\!y_{4 } \ } } & \ ! c4\\ ( 14 ) \ , q[w , a ] \ , \theta_{18 } \xi_{18 } & \ ! \theta_{18 } = \theta_{2 } { \ { ( y_{3}y_{4})\xi_{17}17\!/\!y \ } } & \ ! \xi_{18 } = \xi_{1 } & \ ! a3\\ ( 15 ) \ , q[(y_{3}y_{4 } ) , a ] \ , \theta_{19 } \ , \xi_{19 } & \ ! \theta_{19 } = \theta_{18 } & \ ! \xi_{19 } = \xi_{1 } { \ { ( 15)\theta_{18 } 18\!/\!y_{3 } , ( 17)\theta_{18 } 18\!/\!y_{4 } \ } } & \ ! c4\\ ( 16 ) \ , q[y_{3}y_{4 } , a ] \ , \theta_{20 } \ , \xi_{20 } & \ ! \theta_{20 } = \theta_{18 } { \ { ( y_{3}y_{4})\xi_{19 } 19\!/\!s \ } } & \ ! \xi_{20 } = \xi_{19 } & \ ! a3\\ ( 15 ) \ , q[(y_{3}y_{4 } ) , a ] \ , \theta_{21 } \ , \xi_{21 } & \ ! \theta_{21 } = \theta_{20 } & \ ! \xi_{21 } = \xi_{19 } { \ { ( 15)\theta_{20 } 20\!/\!y_{3 } , ( 17)\theta_{20 } 20\!/\!y_{4 } \ } } & \ ! c4\\ ( 16 ) \ , q[y_{4 } , a ] \ , \theta_{22 } \ , \xi_{22 } & \ ! \theta_{22 } = \theta_{20 } { \ { y_{4}\xi_{21 } 21\!/\!s \ } } & \ ! \xi_{22 } = \xi_{21 } & \ ! a3\\ ( 17 ) \ , q [ ( \ ) , a ] \ , \theta_{23 } \ , \xi_{23 } & \ ! \theta_{23 } = \theta_{22 } & \ ! \xi_{23 } = \xi_{21 } & \ ! c4\\ ( 18 ) \ , q[y_{3}y_{4 } , a ] \ , \theta_{24 } \ , \xi_{24 } & \ ! \theta_{24 } = \theta_{22 } & \ ! \xi_{24 } = \xi_{17 } & \ ! a3\\ ( 13 ) \ , q[(y_{4 } ) , a ] \ , \theta_{25 } \xi_{25 } & \ ! \theta_{25 } = \theta_{2 } & \ ! \xi_{25 } = \xi_{17 } & \ ! c4\\ ( 14 ) \ , q[w , a ] \ , \theta_{26 } \xi_{26 } & \ ! \theta_{26 } = \theta_{2 } { \ { ( y_{3 } y_{4})\xi_{17}25\!/\!y \ } } & \ ! \xi_{26 } = \xi_{1 } & \ ! a3\\ ( 15 ) \ , q[(y_{3}y_{4 } ) , a ] \ , \theta_{27 } \ , \xi_{27 } & \ ! \theta_{27 } = \theta_{26 } & \ ! \xi_{27 } = \xi_{1 } { \ { ( 15)\theta_{26 } 26\!/\!y_{3 } , ( 17)\theta_{26 } 26\!/\!y_{4 } \ } } & \ ! c4\\ ( 16 ) \ , q[y_{3}y_{4 } , a ] \ , \theta_{28 } \ , \xi_{28 } & \ ! \theta_{28 } = \theta_{26 } { \ { ( y_{3}y_{4})\xi_{27 } 27\!/\!s \ } } & \ ! \xi_{28 } = \xi_{27 } & \ ! a3\\ ( 15 ) \ , q[(y_{4 } ) , a ] \ , \theta_{29 } \ , \xi_{29 } & \ ! \theta_{29 } = \theta_{28 } & \ ! \xi_{29 } = \xi_{27 } { \ { ( 15)\theta_{28 } 28\!/\!y_{3 } , ( 17)\theta_{28 } 28\!/\!y_{4 } \ } } & \ ! c4\\ ( 16 ) \ , q[y_{4 } , a ] \ , \theta_{30 } \ , \xi_{30 } & \ ! \theta_{30 } = \theta_{28 } { \ { y_{4}\xi_{29 } 29\!/\!s \ } } & \ ! \xi_{30 } = \xi_{29 } & \ ! a3\\ ( 17 ) \ , q [ ( \ ) , a ] \ , \theta_{31 } \ , \xi_{31 } & \ ! \theta_{31 } = \theta_{30 } & \ ! \xi_{31 } = \xi_{29 } & \ ! c4\\ ( 18 ) \ , q[y_{4 } , a ] \ , \theta_{32 } \ , \xi_{32 } & \ ! \theta_{32 } = \theta_{30 } & \ ! \xi_{32 } = \xi_{17 } & \ ! a3\\ ( 19 ) \ , q [ ( \ ) , a ] \ , \theta_{33 } \ , \xi_{33 } & \theta_{33 } = \theta_{2 } & \ ! \xi_{33 } = \xi_{17 } & \ ! c4\\ ( 20 ) \ , q [ \ { \exists}\ ] \ , \theta_{34 } \ , \xi_{34 } & \ ! \theta_{34 } = \theta_{2 } & \ ! \xi_{34 } = \xi_{17 } & \ ! a1\\ \end{array}\ ] ] we examine applications of * t2 * to the term . the simple tile @xmath1012 rooted at ( 6 ) is @xmath21-end and @xmath21-directed with respect to the game : there are three @xmath21-plays on it which are all ri , @xmath804 , @xmath1013 and @xmath1014 . transformation * t2 * allows us to remove this tile , so the leaf node ( 8) is directly beneath node ( 5 ) . basic _ tile @xmath771 rooted at ( 14 ) is @xmath21-end and @xmath21-directed with respect to the game : the only plays @xmath1015 , @xmath1016 and @xmath1017 are ri . a second application of * t2 * removes it ; therefore , node ( 18 ) is directly beneath node ( 13 ) . consequently , the basic tile @xmath1018 rooted at ( 12 ) is also @xmath21-end and @xmath21-directed with respect to the `` reduced '' game ; the plays @xmath1019 and @xmath1020 become the sequences @xmath1021 and @xmath1022 ( modulo changes to the look - up tables ) . the starting term in figure [ ex1 ] is , therefore , reduced to the smaller solution term @xmath1023 . assume @xmath4 is a @xmath12rd - order term . if we inspect it top - down , starting beneath the initial lambda then it consists of simple tiles , each of which is either a constant tile or a top tile @xmath1024 , @xmath322 , where each atomic leaf is labelled with a dummy @xmath471 because @xmath69 has order at most @xmath157 ; therefore , it is also an end tile . [ fact75 ] if @xmath4 has order @xmath12 and @xmath763 is a simple tile in @xmath4 then either @xmath763 is a constant tile or an end tile which is also a top tile . for instance , the tree in figure [ picture2 ] consists of four simple top tiles @xmath1025 that are also end tiles rooted at @xmath424 , @xmath429 , @xmath196 and @xmath445 and the simple constant tiles @xmath793 , @xmath794 , @xmath777 and @xmath298 . therefore , by repeated application of fact [ fact57 ] and proposition [ end1 ] each play @xmath1026 when @xmath4 is @xmath1027-order merely descends a branch of @xmath4 until it reaches a final state . we now examine this _ tree model property _ of plays in more detail and show how it leads to a very straightforward proof of decidability of @xmath12rd - order matching . assume @xmath10 is @xmath1027-order . we define a ( unique ) partition of any play @xmath968 in stages ; at each stage we identify a simple tile , a subpart of @xmath4 , and the interval at that stage . we call this iteratively defined notion of partition a p - partition ( a `` play partition '' ) to distinguish it from definition [ def614 ] of the b - partition for a position . ( at @xmath12rd - order , these partitions are intimately related as we shall note ; in the next section we extend p - partitions to all orders and its definition uses b - partitions . ) [ def76 ] assume @xmath10 is @xmath12rd - order and @xmath1026 . p - partition _ of @xmath536 is defined in stages @xmath1028 for some @xmath83 as @xmath1029 where @xmath854 . at each stage @xmath24 there is 1 . the p - partition up to stage @xmath1030 , @xmath1031 ; 2 . the simple tile @xmath1032 which occurs in @xmath4 directly beneath node @xmath1033 ; 3 . the composite tile @xmath1034 of @xmath4 consisting of all the nodes in the tiles @xmath1035 ; 4 . the position @xmath995 with @xmath1036 which is the root node of @xmath1032 ; 5 . the interval @xmath1037 determined as follows : @xmath1038 is the least @xmath1039 such that 1 . @xmath1040 is an atomic leaf of @xmath1032 , or 2 . @xmath1041 . the idea of a p - partition of @xmath968 is to structurally relate parts of @xmath536 to parts of @xmath4 . at stage @xmath21 , @xmath1042 is the simple tile directly beneath the initial lambda of @xmath4 which is either a constant or a top tile . the subpart of @xmath4 at this stage , @xmath1043 is just @xmath1042 . assume @xmath247 is the root of @xmath1042 ; therefore , @xmath1044 ( because @xmath1045 ) . consider the interval @xmath1046 : @xmath712 is the first position such that either it is at an atomic leaf of @xmath1042 or it is the final position of the play . in the first case , @xmath1046 is a shortest play on @xmath1042 ; the tile @xmath1047 is the simple tile directly beneath the atomic leaf @xmath1048 of @xmath1042 and @xmath1049 is @xmath1042 and @xmath1047 . the p - partition thereby continues : at stage @xmath24 , the interval @xmath1037 is either a shortest play on @xmath1032 or @xmath1050 and then there are no further stages . the following is an easy consequence of fact [ fact57 ] , proposition [ prop515 ] and the definition of b - partition . assume the _ p - partition _ of @xmath968 is @xmath1029 and @xmath1032 is the simple tile at stage @xmath24 . 1 . for @xmath1051 , @xmath1037 is a shortest play on @xmath1032 . 2 . for @xmath1052 , @xmath1037 is internal to @xmath1032 . 3 . for @xmath1053 the b - partition of @xmath1054 . [ ex63 ] if @xmath536 is the play of example [ examp42 ] on the tree of figure [ picture2 ] then its p - partition is depicted below linearly . @xmath1055 the simple tiles at each stage are ; @xmath1056 , @xmath1057 rooted at @xmath429 , @xmath1058 , @xmath1059 , @xmath1060 rooted at @xmath196 and @xmath1061 . for the other play @xmath1010 in this example , there is the following p - partition . @xmath1062 the two plays share the first three simple tiles , but play is at different atomic leaves of @xmath1063 after stage @xmath12 . consider the p - partitions of all plays in @xmath473 . we slightly abuse notation : we let @xmath1037 , @xmath1064 be their intervals at stage @xmath24 even when they have different ranges . instead of a branch of simple tiles there is a tree of simple tiles : as each play shares the same simple tile @xmath1042 at stage @xmath21 of its p - partition . the simple tile @xmath763 in @xmath4 is _ special _ if it obeys one of the following three conditions 1 . @xmath1065 for @xmath536 at stage @xmath24 and @xmath1037 is nri ( see definition [ defi48 ] ) , 2 . @xmath1066 for @xmath536 at stage @xmath24 and @xmath1050 , 3 . @xmath1066 for @xmath536 , @xmath1010 at stage @xmath24 and @xmath1067 when @xmath1068 , @xmath1069 . the first kind of special tile explicitly contributes to solving @xmath10 . the second kind identifies where a play finishes . the third kind separates plays ; each p - partition after stage @xmath21 that ends at the same atomic leaf of @xmath1042 shares @xmath1047 at stage @xmath157 and so on . therefore , branching in the tree of simple tiles will occur at @xmath1032 if there are plays @xmath536 , @xmath1010 that end at different atomic leaves of @xmath1032 at stage @xmath24 ( and agree on atomic leaves at all earlier stages ) . in example [ ex63 ] , @xmath793 separates the plays @xmath536 and @xmath1010 . the other special tiles in these plays are @xmath794 because of the nri interval @xmath806 and @xmath777 and @xmath298 as plays finish within them . any simple tile in @xmath4 with at least one atomic leaf which is not special is superfluous . either every play avoids it ( so , * t1 * applies ) or every play on it is ri and ends at the same atomic leaf @xmath774 for some @xmath241 ( so , is both @xmath241-end and @xmath241-directed with respect to the game and * t2 * applies ) : four instance , all four simple tiles @xmath1025 in example [ ex63 ] are both @xmath21-end and @xmath21-directed with respect to the game . there is an upper bound ( relative to the problem @xmath10 ) on the number of special tiles that can occur in a term @xmath4 as follows 1 . at most @xmath165 ( @xmath124 the right size for @xmath10 , definition [ delta ] ) tiles that involve nri intervals ; 2 . at most @xmath1070 ( @xmath124 the number of plays is bounded by the number of branches in the right terms of @xmath10 . ] ) tiles where a play ends ; 3 . at most @xmath1071 tiles that are play separators . decidability of @xmath12rd - order matching , via the small model property , is , therefore , a simple consequence of the p - partitions and the identification of special simple tiles . for example [ ex63 ] , the term of figure [ picture2 ] can be reduced to the smaller solution term @xmath1072 . [ def64 ] @xmath1073 is the number of simple tiles with atomic leaves in a longest branch of @xmath4 and @xmath1074 is the total number of simple tiles with atomic leaves in @xmath4 . [ bounds ] if @xmath4 is a smallest solution to @xmath12rd - order @xmath10 and @xmath1070 is the number of plays in @xmath473 , then @xmath1075 . an alternative , and equally simple , proof of decidability of @xmath12rd - order matching that does not appeal directly to the tree model property uses proposition [ end ] and transformations * t1 * and * t2. in a large solution term , there must be embedded end tiles that are redundant . dowek observes that solutions with embeddings @xmath1076 can be reduced to smaller solutions @xmath1077 in his proof of decidability of @xmath12rd - order matching @xcite . assume @xmath4 is a _ smallest _ solution term ( with respect to @xmath1078 ) for @xmath10 . it contains at most @xmath1079 constant tiles with atomic leaves ( otherwise , t1 * would apply and produce a smaller solution ) . let @xmath378 be the number of constant tiles with atomic leaves in @xmath1080 . therefore , by proposition [ end ] , @xmath4 contains at most @xmath1081 embedded simple top tiles ( otherwise , * t2 * would apply and produce a smaller solution ) . if @xmath41 is the arity of @xmath10 , definition [ def3 ] , then there are at most @xmath41 inequivalent top simple tiles : as soon as a branch of @xmath4 contains @xmath1082 simple top tiles , there must be at least one embedded end tile . consequently , no branch of @xmath4 can contain more than @xmath1083 simple top tiles . because @xmath1084 , @xmath1085 and @xmath12rd - order matching is decidable . the question is how to extend these straightforward proofs to all higher - orders . with a @xmath7th or @xmath96th - order tree there are two levels of simple non - constant tiles : top tiles @xmath869 and end tiles @xmath1086 where @xmath242 is bound by a @xmath774 . the number of levels increases with order : at @xmath1087th or @xmath1088th - order there are four levels . as soon as there is more than one level , game playing may jump around the tree as the examples in figures [ ex2 ] and [ ex15 ] illustrate . for any order , if the tree - model property holds , so each play can be p - partitioned into internal plays on the simple tiles from the root to a tile where the play ends , then decidability is assured by iteration : redundant simple end tiles are first removed which causes further tiles to be end and so on . schubert defines an _ unsophisticated _ lambda term in @xcite : @xmath1089 is unsophisticated if each occurrence of @xmath1090 within @xmath4 has the property that no @xmath1091 contains a free variable @xmath1092 , @xmath1093 . a @xmath96th - order ( dual ) interpolation problem is unsophisticated if in each ( dis)equation @xmath1094 the left terms @xmath1095 are unsophisticated . schubert proves decidability of @xmath96th - order unsophisticated dual interpolation . this restriction implies the tree model property . when play is at an end tile with head variable @xmath69 , the state has the form @xmath1096 $ ] where @xmath1097 is closed : consequently , play can not jump back to the tile which binds @xmath69 . ( decidability can be extended to all orders by defining _ hereditary _ unsophisticated terms where each occurrence of @xmath1098 within it is unsophisticated . ) for arbitrary order , if @xmath4 is a smallest solution term for @xmath10 and it contains @xmath378 simple constant tiles with atomic leaves then @xmath1084 . if @xmath4 is large then it must contain embedded tiles . with @xmath7th - order there must be embedded top tiles and with @xmath96th - order there can also be embedded end tiles . however , proposition [ end ] implies , for any order , that there can not be more than @xmath1099 embedded end tiles . consequently , if @xmath10 is a @xmath96th - order problem and @xmath4 contains a bounded number @xmath24 of top simple tiles then the small model property holds . [ fact66 ] if @xmath4 is a smallest solution to @xmath96th - order @xmath10 that contains at most @xmath24 top simple tiles , then @xmath1100 fact [ fact66 ] generalises to all orders : if @xmath4 is a smallest solution to @xmath10 that contains at most @xmath24 simple tiles that are neither end nor constant tiles then it has a bounded size . this result slightly extends some cases examined in @xcite where there are restrictions on the number of free variables that a @xmath181 can bind within a solution term . a family of tiles in a tree consists of a top tile together with all its dependents ( see definition [ family ] ) . the problem case for @xmath7th or @xmath96th - order is a solution term with arbitrary many top tiles ( and , therefore , arbitrary many families ) . if a tree is large , then it must contain families of tiles all of whose plays are ri and , therefore , do not contribute to solving @xmath10 . for even higher - orders , for the same reason , whole subfamilies of tiles are redundant . the difficulty is how to extract and remove redundant families and subfamilies of tiles . they may be entangled , occurring anywhere in a tree . what we would like to do is disentangle them thereby restoring , as far as possible , the tree model property . a key ingredient is _ tile lowering _ where we generalise to _ basic _ tiles . consider the left tree in figure [ tl ] and assume @xmath1032 is a basic tile and @xmath1101 is a simple tile whose head variable is bound within @xmath1032 ( and no variable on the branch between @xmath1102 and @xmath1101 is bound within @xmath1032 ) . what we would like to do is to transform the left tree into the right tree where @xmath1032 ( and all the subtrees beneath its atomic leaves other than @xmath1102 ) is copied immediately above @xmath1101 ( thereby producing a larger basic tile ) with the understanding that it is the lower occurrence of @xmath1032 that binds @xmath1101 and any free variables beneath @xmath1101 that are bound within @xmath1032 in the left tree . in @xcite , we introduced an explicit local transformation that has this effect for @xmath7th - order matching and for the atoms case at all orders . the virtue of the transformation is that families of tiles are disentangled as the lower occurrence of @xmath1032 is brought next to its dependent tile @xmath1101 ( and the upper occurrence of @xmath1032 in the right tree is `` closer '' to being an end tile , as ( some ) binding is lost below ) . game - theoretically , tile lowering will be justified in terms of permutations , omissions and repetitions of ri plays on tiles . the analysis now shifts from how single tiles to how families of tiles in a term contribute to solving a problem . we extend the definition of the p - partition of a play from the previous section to all orders . we start by examining simple properties of game playing that involve families of tiles : the discussion uses the notion of when a position is a descendent of another position , definition [ def416 ] , when one tile is a dependent of another , definition [ defdepend ] , and when two tiles belong to the same family , definition [ family ] . [ c4jump ] assume @xmath763 with root node @xmath4 and @xmath1103 with root node @xmath180 are simple tiles . 1 . if @xmath544 , @xmath669 , @xmath626 is the result of move a3 of figure [ game ] and @xmath626 is a descendent of @xmath539 then @xmath773 is a dependent of @xmath763 . if @xmath544 and @xmath1104 and @xmath576 is the result of move c4 then @xmath763 and @xmath773 belong to the same family of tiles . \(1 ) assume that @xmath544 , @xmath669 and @xmath626 is a descendent of @xmath539 . therefore , because the descendent relation is the reflexive and transitive closure of the child relation , there is a subsequence of positions @xmath1105 with @xmath1106 , @xmath1107 , for all @xmath83 , @xmath1108 is a child of @xmath846 . assume @xmath1109 for each @xmath1110 ; so , @xmath1111 and @xmath1112 . because @xmath626 is the result of a3 , @xmath180 is labelled with a variable , that is bound by @xmath1113 : therefore , @xmath773 is a dependent of the simple tile rooted at @xmath1114 . this argument is now repeated because @xmath1115 must in turn be the result of a3 and so the simple tile rooted at @xmath1114 is a dependent of the simple tile rooted at @xmath1116 , and so on ; consequently , as the dependency relation is transitive closed , @xmath773 is a dependent of @xmath763 . ( 2 ) consider the following subsequences of positions @xmath1117 where @xmath1118 and @xmath1119 ; each position @xmath1120 is the result of c4 on @xmath1121 which is the result of @xmath169 or more applications of c3 on @xmath1122 , @xmath1123 ; furthermore , for @xmath1124 , @xmath1122 is the result of a3 and is a child of @xmath1125 and @xmath712 is the result of a3 and a child of @xmath586 . now by a routine induction on @xmath24 , it follows that for every @xmath798 , @xmath1126 is an atomic leaf of some simple tile rooted at @xmath1127 for some @xmath1128 and that all such tiles belong to the same family . in the previous section we alluded to the notion of tile level . [ level ] the _ level _ of a non - constant tile is defined inductively . 1 . if @xmath763 is a top tile then @xmath763 has level @xmath21 . if @xmath763 is an immediate dependent of @xmath773 and @xmath773 has level @xmath331 then @xmath763 has level @xmath332 . the number of possible levels increases with order . in a @xmath7th or @xmath96th - order term there are at most two levels of tile : top tiles and end tiles as illustrated in figures [ ex1 ] and [ ex14 ] . with a @xmath1087th or @xmath1088th - order term , there are at most four levels of tile and so on . the presence of dummy lambda in a term tree does not affect the notion of level ( because a dummy lambda can not be a binder ) . if there is more than one level of tile in a term , game playing may pass repeatedly through the same sequences of nodes of a tree . the following definition captures when two such intervals are said to correspond . [ defsimilar ] the intervals @xmath543 , @xmath1129 _ correspond _ , written @xmath1130 , provided that @xmath1131 , @xmath1132 and for all @xmath1133 , 1 . if @xmath1134 then @xmath1135 . 2 . if @xmath214 \in \pi(i+k)$ ] then for some @xmath561 , @xmath720 \in \pi(i'+k)$ ] 3 . if @xmath1136 \in \pi(i+k)$ ] then for some @xmath561 , @xmath1137 \in \pi(i'+k)$ ] . 4 . if @xmath552 \in \pi(i+k)$ ] then for some @xmath561 , @xmath1138 \in \pi(i'+k)$ ] . the relation @xmath1139 on intervals is an equivalence relation . although this definition abstracts from the look - up tables , it requires agreement on the sequences of nodes of the tree and on the left terms of states . for instance , intervals that are shortest ri plays on a top tile correspond , as shown in the proof of proposition [ prop515 ] . when @xmath1010 is the play in figure [ nex2 ] on the tree in figure [ ex1 ] , the nri intervals @xmath1140 and @xmath1141 correspond . in section [ tiles ] we described some uniformity properties of game playing for top and embedded tiles ( especially for embedded end tiles ) . now our aim is to understand uniformities of play for tiles of arbitrary level . given a position @xmath626 at a lambda node @xmath180 , there is an associated ( unique ) b - partition of @xmath1142 into intervals that are plays on the simple tiles between the root of the term tree and node @xmath180 by definition [ def614 ] . if more than one position is at a lambda node @xmath180 then their associated b - partitions must differ in their plays at an earliest tile with the same starting positions but different end positions . [ prop73a ] assume @xmath669 , @xmath1143 , @xmath1144 and @xmath180 is labelled @xmath760 for some @xmath1145 . if @xmath1146 is the b - partition for @xmath626 and @xmath1147 is the b - partition for @xmath725 then there is a @xmath1148 such that @xmath1149 and @xmath1150 and for all @xmath1151 , @xmath1152 and @xmath1153 . assume that @xmath1154 is the branch from the root of the term tree to @xmath180 labelled @xmath1155 , @xmath1156 and @xmath1143 for @xmath1157 . let @xmath1146 be the b - partition for @xmath626 and @xmath1158 be the b - partition for @xmath725 . because @xmath1159 , @xmath1160 and @xmath1161 , there must be a least @xmath24 such that @xmath1149 and @xmath1150 . the b - partitions of example [ example512 ] illustrate this proposition ; node @xmath433 of figure [ ex14 ] labelled @xmath471 belongs to both @xmath859 and @xmath860 of the play in figure [ ex15 ] . these b - partitions agree on the play @xmath809 on the initial tile @xmath776 that ends at @xmath1162 and then they have different @xmath21-plays @xmath814 and @xmath1163 on the next tile @xmath778 . [ defi744 ] assume @xmath669 , @xmath1143 , @xmath1144 , @xmath180 is labelled @xmath760 for some @xmath1145 and @xmath1164 is the branch from the root of the term tree to @xmath180 . let @xmath1146 be the b - partition for @xmath626 and let @xmath1147 be the b - partition for @xmath725 and let @xmath1165 be the first simple tile for @xmath226 such that @xmath1149 and @xmath1150 . the positions @xmath1166 are then said to _ vary at @xmath1167 with ( simple tile ) @xmath763_. two positions vary at @xmath1167 with @xmath763 if they are at the same lambda node and @xmath763 is the first simple tile in their b - partitions where there is a difference in play ; there are two distinct @xmath331-plays on @xmath763 for some @xmath331 , @xmath1037 and @xmath1168 . in the case of example [ example512 ] discussed previously , the positions @xmath859 , @xmath860 vary at @xmath1169 with @xmath778 . given two positions @xmath626 and @xmath725 at the same node labelled @xmath760 we are interested in defining when intervals @xmath1170 and @xmath1171 correspond in the sense of definition [ defsimilar ] . the simplest case is when @xmath1166 vary at @xmath1166 with @xmath1172 ; their b - partitions agree except on the last simple tile in the branch from the root to @xmath760 . the two intervals @xmath1173 and @xmath1174 are , therefore , both @xmath24-plays on @xmath763 for some @xmath24 ; both positions @xmath626 and @xmath725 are children of @xmath846 and , therefore , must be the results of move c4 of figure [ game ] . this means that the look - up tables @xmath1175 and @xmath1176 only differ in their entries for the variables in @xmath1145 : therefore , for each @xmath1123 their continuations @xmath1177 , @xmath1178 correspond as long as play does not reach children of @xmath626 and @xmath725 ; or positions where different @xmath251 choices are exercised ; or a position with a final state . consider next the general situation when @xmath1166 vary at @xmath1167 with @xmath1165 ; their b - partitions differ at a simple tile that is earlier in the branch from the root to @xmath760 . what we want to capture is the following uniformity : if the plays @xmath835 and @xmath1179 on tiles @xmath1180 that do not involve positions that are children of @xmath841 and @xmath1181 correspond , then the continuations from @xmath1182 and @xmath1183 will also correspond as long as the positions are not descendents of @xmath841 or @xmath1181 or the result of a different @xmath251-choice , or one has a final state . the issue is how to formally capture this correspondence in terms of the relationship between the look - up tables @xmath1184 and @xmath1185 . this is the motivation for the following bisimulation like definition . [ defi75 ] assume @xmath763 is a simple non - constant tile . 1 . two look - up tables @xmath335 , @xmath334 are _ @xmath83-similar except for @xmath763 _ , @xmath1186 , which is defined iteratively , for @xmath183 . 1 . @xmath1187 iff @xmath601 ; 2 . @xmath1188 iff ( 1 ) for all @xmath69 . @xmath240 is defined iff @xmath1189 is defined and ( 2 ) if @xmath1190 is not a dependent of @xmath763 and @xmath1191 then @xmath1192 and @xmath1193 ; 3 . @xmath1194 iff ( 1 ) for all @xmath242 . @xmath565 is defined iff @xmath1195 is defined and ( 2 ) if @xmath1196 then @xmath1197 and @xmath1198 . two look - up tables @xmath335 , @xmath334 are _ similar except for @xmath763 _ , @xmath1199 , if there is an @xmath183 such that @xmath1200 . in the following assume @xmath335 , @xmath334 are look - up tables ( of the same kind ) . 1 . @xmath1201 . 2 . if @xmath1186 then @xmath1202 . 3 . if @xmath1186 and @xmath1203 then @xmath1204 . 4 . if @xmath1205 and @xmath875 then @xmath1206 . the key point with the definition of @xmath1207 is that the entries of @xmath1 and @xmath616 should be very similar except in the case that they are labels of nodes of simple tiles that are dependents of @xmath763 . [ fact78 ] if @xmath543 and @xmath1208 are @xmath24-plays on the simple tile @xmath763 , @xmath1184 and @xmath1185 then @xmath1207 . we now come to a critical uniformity property . [ prop79 ] assume @xmath626,@xmath725 vary at @xmath1167 with @xmath763 , @xmath1184 , @xmath1185 and @xmath1207 . if @xmath1209 is ri , @xmath1210 is labelled @xmath760 for some @xmath1145 and no @xmath1211 , @xmath1093 , is a descendent of @xmath841 then @xmath1212 . assume that @xmath1154 is the branch from the root of term tree @xmath4 to @xmath1155 , @xmath1213 and @xmath1214 for @xmath1215 . let @xmath1216 be the b - partition for @xmath626 and @xmath1147 be the b - partition for @xmath725 . assume that @xmath626 , @xmath725 vary at @xmath841 , @xmath1181 with @xmath1217 and that @xmath1207 when @xmath1184 and @xmath1185 . assume that @xmath1177 is ri , @xmath1210 is labelled @xmath760 , no @xmath1218 within this interval is at a dependent of @xmath763 . by a routine induction on @xmath331 , @xmath1219 . the interval @xmath1177 only involves moves a3 and c4 of figure [ game ] because it is ri and ends at @xmath1210 by fact [ fact55 ] . initially , for some simple non - constant tile @xmath1220 , @xmath1221 and @xmath1222 . therefore , because @xmath1182 is not a descendent of @xmath841 , @xmath761 is not a dependent of @xmath763 ; by a3 of figure [ game ] position @xmath1223 \theta \xi_{j+1}$ ] and position @xmath1224 \theta ' \xi_{j'+1}$ ] where @xmath1225 . the next position is a result of c4 : if @xmath1226 then @xmath1227 and @xmath1228 . consequently , the result follows . [ rem811 ] proposition [ prop79 ] can be strengthened to include corresponding nri plays that involve the same @xmath251-choices . for instance , in example [ example512 ] when @xmath536 is in figure [ ex15 ] the positions @xmath1169 vary at @xmath1169 with @xmath778 . position @xmath1229 is the first position after @xmath659 that is a child of @xmath659 ; therefore , the continuations @xmath1230 and @xmath862 correspond even though they pass through the constant tile @xmath1231 . we shall now extend the definition of the p - partition of a play @xmath1232 , definition [ def76 ] , from @xmath12rd - order @xmath10 to all higher orders . again , it is defined in stages using simple tiles from @xmath4 ; so @xmath1233 for some @xmath83 and there is an associated sequence of simple tiles @xmath1234 from @xmath4 such that for each @xmath1235 , @xmath1236 where @xmath180 is the root of @xmath1237 and @xmath1237 occurs directly below @xmath1238 in @xmath4 . however , the same simple tile may occur more than once in this sequence of tiles ; so we adopt the following notation . [ def81 ] assume that @xmath1239 is a sequence of simple tiles associated with a play @xmath1026 . we write @xmath1240 to identify the @xmath24th tile @xmath1032 in this sequence and we use the notation @xmath1241 for node @xmath180 of @xmath1240 . after stage @xmath24 , the p - partition for @xmath536 consists of the sequence of tiles @xmath1242 ; the composite tile @xmath1034 consists of these tiles . however , we also provide a more graphical representation of the composite tile @xmath1034 by including a labelled edge for each @xmath1243 of the form @xmath1244 where @xmath1245 and @xmath180 is an atomic leaf of @xmath1246 . the relation @xmath1247 between a tile at stage @xmath331 and an atomic leaf of a tile at an earlier stage @xmath216 captures control structure in the p - partition that @xmath1248 . this linearisation of the p - partition for each @xmath536 will be useful in the decidability proof ; we can , of course , reconstitute the subtree @xmath1034 from its linear presentation after stage @xmath24 just by viewing @xmath1247 as the subtree relation . we assume this extra intensionality is also reflected in the @xmath237 look - up tables for @xmath536 ; in the case of the c moves of figure [ game ] if @xmath293 and @xmath1249 then @xmath1250 . consequently , at an application of move c4 , if @xmath29 is the head variable of the left term of the state and @xmath1251 then the next position is at @xmath1252 . [ partit ] assume @xmath968 . p - partition _ of @xmath536 is defined in stages @xmath1028 for some @xmath83 as @xmath1029 where @xmath854 . at each stage @xmath24 there is 1 . the p - partition up to stage @xmath1030 , @xmath1031 ; 2 . the composite tile @xmath1253 consisting of the tiles @xmath1254 with edges @xmath1247 ; 3 . the simple tile @xmath1240 which occurs in @xmath4 directly beneath node @xmath1255 and @xmath1256 in @xmath1257 ; 4 . the position @xmath995 with @xmath1258 where @xmath582 is the root node of @xmath1032 ; 5 . the interval @xmath1037 determined as follows : 1 . set @xmath1259 2 . @xmath1260 while @xmath1261 and @xmath1262 is not labelled with a lambda do j = j+1 ; 1 . if @xmath1041 or @xmath1263 then @xmath1264 ; 2 . find a largest @xmath1265 , if there is one , such that there is a @xmath1266 , @xmath1166 vary at @xmath1267 with @xmath1268 in the same family as @xmath1240 and @xmath1269 where @xmath1184 , @xmath1185 , and 1 . no position in @xmath1270 is a descendent of @xmath1271 ; 2 . if @xmath1272 then @xmath1273 ; 3 . if there is such a @xmath1265 set @xmath1274 and goto @xmath1260 else @xmath1264 . as with definition [ def76 ] , if @xmath1240 is a top or a constant tile then @xmath1037 is internal to it and either ends at one of its atomic leaves , or a final state is reached : via clause @xmath1260 , @xmath1038 will be the least @xmath1039 such that @xmath669 is an atomic leaf of @xmath1240 or @xmath1041 . the new case is when @xmath1275 is neither a constant tile nor a top tile . position @xmath995 is at the root of @xmath1240 ; the first position @xmath626 , if there is one , after @xmath995 that is at a lambda node @xmath180 need not be internal to @xmath1240 ; however @xmath180 must be an atomic leaf of some tile @xmath1276 , @xmath1277 , which belongs to the same family as @xmath1240 by proposition [ c4jump ] . if @xmath1278 then @xmath1264 and @xmath1037 finishes at @xmath180 . otherwise @xmath1279 and @xmath1280 : one checks whether there are previous positions @xmath725 such that @xmath1166 vary at @xmath1267 with @xmath1268 in the same family as @xmath1240 ( and , therefore , also , @xmath1276 ) and whether @xmath1281 when @xmath1184 and @xmath1185 . if there are no such positions , for instance if @xmath180 is an atomic leaf of the composite tile @xmath1034 , then @xmath1264 and @xmath1037 finishes at @xmath180 . otherwise there are such positions ; we then look for a longest continuation from @xmath1182 that corresponds to a previous interval from such a @xmath1183 . if there is no such continuation , so @xmath1282 , then @xmath1182 is at a descendent of @xmath887 ; so control is at the root of a tile in @xmath1034 in the same family as @xmath1240 ; and the loop starts again . if there is such a continuation then @xmath1283 and the loop starts again from @xmath1284 ; if @xmath1037 is ri then @xmath1284 will also be at the root of a tile in @xmath1034 in the same family as @xmath1240 as this position must be a descendent of @xmath887 . consequently , as we shall prove , if @xmath1037 is ri then it is guaranteed to finish at an atomic leaf of a tile in the same family as @xmath1240 . [ ex713 ] we describe the p - partition of @xmath1010 in figure [ nex2 ] for the term tree in figure [ ex1 ] ( omitting the initial move ) . we present tile @xmath1285 , without its superscript @xmath1010 , the play @xmath1064 and the edge relation @xmath1286 in figure [ ex78 ] . @xmath1287 the intervals for the first three stages are plays on top and constant tiles . tile @xmath1288 : play jumps at position @xmath1289 to @xmath752 of @xmath1042 ; the positions @xmath1290,@xmath1291 vary at @xmath1292 with @xmath1042 and @xmath1293 when @xmath1294 and @xmath1295 ; there are the maximal corresponding intervals @xmath1140 and @xmath1141 . play returns to @xmath29 and jumps to @xmath1296 which completes stage @xmath7 as @xmath1297 is the first position at this atomic leaf . the p - partition continues with plays on the top tiles @xmath1298 and @xmath1299 . after stage @xmath1300 , the subtree @xmath1301 of figure [ ex1 ] associated with the partition consists of tiles @xmath1302 plus the edges already described ( which has atomic leaves @xmath1303 , @xmath1304 , @xmath1305 and @xmath1306 ) . tile @xmath1307 and @xmath1308 vary at @xmath1308 with @xmath1299 and @xmath1309 when @xmath1310 , @xmath1311 ; however , there are no corresponding continuations because @xmath1312 and @xmath1313 are children of @xmath1314 and @xmath1315 . the interval at stage @xmath419 is , therefore , @xmath1316 that finishes at @xmath1317 . tile @xmath1318 and the play at this stage jumps to @xmath765 of @xmath1298 ; positions @xmath1319 , @xmath1320 vary at @xmath1319 , @xmath1320 with @xmath1298 and @xmath1321 where @xmath1322 and @xmath1323 ; @xmath1017 corresponds to @xmath1016 and then play returns to @xmath1324 and jumps to @xmath1325 ; so , @xmath1326 is the interval at stage @xmath1087 . finally , @xmath1327 and stage @xmath1088 is @xmath1328 . a reason that the same tile may be repeated in a p - partition starts with different @xmath251-choices . for instance , consider a situation where @xmath1329 and @xmath1330 as follows where @xmath1032 is a constant tile with arity @xmath157 , @xmath1331 . consider figure [ ex834 ] : the p - partition for @xmath536 has a play on @xmath1332 , then @xmath1333 and then @xmath1032 choosing the left branch of @xmath30 . tile @xmath1334 is a dependent of @xmath1333 and play jumps to its atomic leaf above @xmath1032 and so this position with the earlier one will vary at themselves at @xmath1333 ; play thus proceeds to @xmath1032 ; but now there is a different @xmath251-choice ; so at stage @xmath1335 , play finishes at the other leaf of @xmath1032 . play now proceeds through @xmath1237 to @xmath1336 which is a dependent of @xmath1332 ; it jumps to the atomic leaf of @xmath1332 above @xmath1333 and again it varies with the earlier position at themselves at @xmath1332 : instead of the corresponding play which passes through @xmath1032 twice , play descends to @xmath1032 and now the @xmath251-choice is the right branch ; therefore , play after stage @xmath1337 also finishes at @xmath1338 of the second branch . these different @xmath251-choices involve stages of a p - partition that have nri intervals ; if an interval is ri then it is well - behaved as proposition [ vip ] shows . [ def91 ] assume @xmath1339 is the p - partition of @xmath1026 and @xmath1340 is the associated sequence of tiles . @xmath536-path for @xmath1240 _ is the sequence of tiles @xmath1341 such that @xmath1342 , @xmath1343 and @xmath1344 for @xmath1345 . [ vip ] assume @xmath1339 is the p - partition of @xmath1026 and @xmath1340 is the associated sequence of tiles . 1 . if @xmath257 and @xmath1268 is a top or constant tile then @xmath1346 for some @xmath180 . if @xmath835 is ri and @xmath1347 then @xmath1240 belongs to the same family as @xmath1268 . 3 . if @xmath835 is ri and @xmath1268 is an embedded tile in @xmath4 then @xmath1346 for some @xmath180 . if @xmath835 is ri and @xmath1348 then @xmath1349 is not in the @xmath536-path of @xmath1237 . \(1 ) is clear from definition [ partit ] . for ( 2 ) assume @xmath1350 and @xmath835 is ri . consider the first position @xmath626 in this interval at a lambda node . by proposition [ c4jump ] , it is an atomic leaf of a tile in the same family as @xmath1268 . either this position is @xmath1351 and so the result follows , or there are earlier positions @xmath725 and @xmath1143 and @xmath1166 vary at @xmath1352 at @xmath761 in the same family as @xmath1268 and @xmath1353 when @xmath1184 and @xmath1185 . one chooses the @xmath1354 that allows longest corresponding intervals @xmath1355 and @xmath1356 for @xmath1357 . using proposition [ prop79 ] , it follows that @xmath1284 is a descendent of @xmath887 and so is at a tile @xmath1358 in the same family as @xmath1240 ; the same argument is now repeated . for ( 3 ) , consider the b - partition for @xmath1359 ; because @xmath1268 is an embedded tile , there is a tile @xmath1240 such that @xmath1360 and @xmath1268 is below @xmath1240 in @xmath4 and in the b - partition there is a play on @xmath1240 , @xmath1361 ; consider the shortest interval @xmath1362 that is a play on @xmath1240 . using proposition [ prop516 ] and definition [ partit ] it follows that @xmath1363 as @xmath1364 is an atomic leaf of @xmath1268 . in the case of @xmath429 assume @xmath835 is ri and @xmath1365 and @xmath1252 is in the @xmath536-path of @xmath1268 ; so , @xmath1240 and @xmath1268 belong to the same family by ( 2 ) because @xmath835 is ri . also , there must be a tile @xmath1246 , @xmath1366 such that @xmath1367 and either @xmath1368 or @xmath1246 is on the @xmath536-path for @xmath1268 . if @xmath1240 is a top tile then consider the b - partition for @xmath1359 and its interval that is a play on @xmath1240 that ends at @xmath1252 at position @xmath725 ; clearly , if position @xmath626 in @xmath835 is at @xmath1252 then @xmath626 , @xmath725 vary at @xmath626 , @xmath725 at @xmath1240 which means that @xmath1369 and so it is impossible that @xmath1370 . otherwise , @xmath1240 is a dependent tile ; the argument is now similar but more general ; if position @xmath626 in @xmath835 is at @xmath1371 for @xmath1246 that is @xmath1240 or a tile that @xmath1240 is a dependent of and @xmath1371 is in the @xmath536-path for @xmath1268 then there is a tile @xmath1372 such that @xmath1372 is a dependent of @xmath1240 or of some tile that @xmath1240 is a dependent of and the b - partition for @xmath1373 which involves a play on @xmath1246 is such that it finishes at @xmath582 at position @xmath725 and @xmath626 , @xmath725 vary at @xmath712 , @xmath1374 at @xmath1375 in the same family as @xmath1268 ; therefore , @xmath626 can not be a final position for @xmath835 . [ ex715 ] the p - partition of @xmath536 in figure [ ex15 ] where the term tree is in figure [ ex14 ] is presented in figure [ ex79 ] ( omitting the initial move ) ; @xmath1376 where we provide the tile @xmath1240 , omitting @xmath536 , the interval @xmath1037 and the edge relation at each stage @xmath24 . play proceeds through the top tiles @xmath1042 and @xmath1047 . tile @xmath1377 and play at this stage is @xmath804 that ends at @xmath1378 . the next tile is a constant tile . tile @xmath1298 is @xmath777 and the pair @xmath1379 , @xmath651 vary at @xmath1380 with @xmath1042 and @xmath1293 for @xmath1381 and @xmath1382 ; so the interval at stage @xmath96 is @xmath1383 that ends at @xmath471 of @xmath1063 as there is no earlier position where control is at this leaf . the tile @xmath1384 and @xmath1385 is the interval at this stage . play then proceeds through a constant tile at stage @xmath419 . tile @xmath1386 and the interval is @xmath1387 ; positions @xmath1388 vary at @xmath1388 with @xmath1047 and @xmath1389 for @xmath1390 and @xmath1391 ; the intervals @xmath1230 and @xmath862 correspond as described in remark [ rem811 ] . tile @xmath1392 is @xmath1393 ; positions @xmath1394 vary at @xmath1395 with @xmath1042 and @xmath1293 when @xmath1396 and @xmath1397 ; the intervals @xmath1398 and @xmath1399 correspond . with the next position @xmath1400 and @xmath1401 vary at @xmath1401 with @xmath1042 and @xmath1293 for @xmath1402 and @xmath1403 . the intervals and @xmath952 , @xmath815 correspond . therefore at stage @xmath1088 , the interval is built from two separate subintervals . finally , stage @xmath1404 is the constant tile @xmath298 . we now prove decidability of higher - order matching at all orders , by showing the small model property ; if @xmath82 then there is a small term @xmath988 . the proof starts with the tree of tiles that captures the p - partitions of all plays in a game @xmath473 and then extends it to a tree of basic tiles . we then define unfolding on such trees which underpins the small model property . as with the @xmath12rd - order case in section [ third ] , we examine the p - partitions of all plays in @xmath203 . we maintain abuse of notation : if @xmath536 and @xmath1010 are two plays we let @xmath1037 , @xmath1064 be their intervals at stage @xmath24 irrespective of their ranges . instead of a sequence of simple tiles there is a tree of simple tiles that is associated with the p - partitions as each p - partition shares the initial tile @xmath1042 of @xmath4 . as a representation for this tree of tiles , we let its root be @xmath1405 when @xmath1406 is the set of all plays in @xmath473 ; any other node of this tree has the form @xmath1407 which represents that for each @xmath1408 , @xmath1240 is its tile at stage @xmath24 and for all earlier stages @xmath1151 , every play in @xmath1409 also shares the same tile at stage @xmath331 . thus , the tree has the form depicted in figure [ ex81 ] . however , we also assume the induced edge relations @xmath1247 within this tree : @xmath1410 if @xmath1411 and @xmath1412 . let @xmath1413 be the tree of tiles for the p - partitions of all plays in @xmath473 . we drop the superscript @xmath1409 from tiles @xmath1407 whenever the context allows . we assume the definition of @xmath536-path , definition [ def91 ] , which picks out the sequence of tiles in @xmath4 from its root to @xmath1407 when @xmath1408 . also , we shall assume that definitions [ defi53 ] , [ defdepend ] and [ defi54 ] of @xmath241-below , below , immediate @xmath241-dependent , dependent and embedded apply to tiles in a tree @xmath1413 by examining @xmath536-paths and bindings : for instance , @xmath1237 is an immediate @xmath241-dependent of @xmath1032 if @xmath1032 is in the @xmath536-path for @xmath1237 , the free variable @xmath69 at the head of @xmath1237 is bound in @xmath1032 and @xmath1237 is @xmath241-below @xmath1032 relative to this @xmath536-path . however , because of the linear representation of a p - partition there is also the idea that a tile @xmath1237 is _ later _ than @xmath1032 in @xmath28 if there is a play @xmath536 and @xmath1414 and @xmath1415 and @xmath1416 ; we also say that @xmath1032 is _ earlier _ than @xmath1237 . as with the @xmath12rd - order case in section [ third ] , we identify _ special _ tiles in the tree . [ def92 ] the tile @xmath1417 is _ special _ if it obeys one of the following three conditions ; 1 . for some @xmath1418 , @xmath1037 is nri , 2 . for some @xmath1418 , @xmath1419 is a final state , 3 . for some @xmath536 , @xmath1420 , @xmath1067 when @xmath1068 and @xmath1069 . a tile is _ x - special _ if it obeys @xmath457 or @xmath424 of these conditions . a special tile that is not x - special is a play separator . there is the same upper bounds on the number of special tiles in @xmath1413 as in the @xmath12rd - order case . [ fact92 ] assume @xmath1413 is tree of tiles associated with @xmath203 . within @xmath1413 there are 1 . at most @xmath165 ( @xmath124 the right size for @xmath10 , definition [ delta ] ) tiles that involve nri intervals ; 2 . at most @xmath1070 ( @xmath124 the number of plays in @xmath203 ) tiles where play ends ; 3 . therefore , at most @xmath1079 are x - special tiles ; 4 . at most @xmath1071 tiles that are play separators . the proof of the small model property for the @xmath12rd - order case is straightforward : use transformation @xmath1421 of section [ third ] to remove any tile that is not special from @xmath1413 and update edges . with higher orders we can not just omit a tile that is not special . it may have dependents , so its removal would result in a tree that is no longer a closed term . or an associated interval may finish at an atomic leaf of some other tile in the tree , so its removal may not preserve game playing . instead , we introduce tile _ unfolding _ as a transformation on a tree @xmath1413 . we need to generalise the notion of tree to that of a _ tree of basic _ tiles . we will be interested in a tile @xmath1422 that is an immediate dependent of a top or embedded tile @xmath1032 which is not x - special and which also does not have x - special later tiles belonging to the same family . therefore , as we shall see , @xmath1237 can be replaced in the tree by a basic tile that is constructed from @xmath1032 and @xmath1237 ; this may require revision of later edges in the tree . before developing the full account , we shall now briefly illustrate it . [ ex91 ] consider the p - partition in figure [ ex78 ] of example [ ex713 ] for the play @xmath1010 in figure [ nex2 ] on the term tree in figure [ ex1 ] . the tree of tiles with the edge relation @xmath1286 , with @xmath1010 omitted , is pictured in the top diagram in figure [ ex92 ] . there are edges , for instance , from different atomic leaves of @xmath1042 to @xmath1047 and @xmath1298 . the special ( and x - special ) tiles are @xmath1047 , @xmath1423 and @xmath1392 ; intervals on the first two of these are nri and the third is where the play @xmath1010 finishes . transformation @xmath1421 of section [ third ] would allow us to remove the tile @xmath1063 with the effect that the edge from @xmath1047 would then be to @xmath1423 : dynamically , in terms of play , this means that the ri interval @xmath1424 , a play on @xmath1063 , is omitted and the interval @xmath1425 reduced ( by omitting its ri play on @xmath1063 ) . tile @xmath1298 is a top tile with immediate dependent @xmath1324 ; neither of these tiles is x - special . however , because of this binding and the fact that it has two outgoing edges from different atomic leaves , @xmath1298 can not be omitted ( like @xmath1063 ) ; nor can we remove @xmath1324 because @xmath1426 ends at an atomic leaf of @xmath1298 . instead we can unfold @xmath1298 at @xmath1324 : we introduce the basic tile @xmath1427 which prefaces @xmath1298 to @xmath1324 ; the new interval @xmath1426 , in effect , includes an extra interval that corresponds to @xmath1428 as a prefix and then omits the ri plays on @xmath1299 and @xmath1429 from the old interval @xmath1426 . play in the new interval @xmath1426 which is still ri now finishes at an atomic leaf of @xmath1430 ; so edges may need to be changed ; here , @xmath1430 now has an edge to @xmath1392 . the effect of this unfold is pictured in the lower diagram of figure [ ex92 ] . as a consequence , both @xmath1298 and @xmath1430 can now be removed by transformation @xmath1421 . as the reader can verify , tile @xmath1299 can also be unfolded at @xmath1429 . we call the process `` unfolding '' @xmath763 at @xmath761 ( where @xmath761 is an immediate dependent of @xmath763 ) because it is analogous to unfolding or unravelling a transition system in modal logic ; here , there is the extra dimension of binding . as with unravelling , the purpose of unfolding in @xmath1413 is also to approximate the tree model property . technically , from a game - theoretic point of view , what will justify unfolding is permuting , repeating and omitting corresponding ri intervals : for instance , with the replacement of @xmath1324 by @xmath1430 in example [ ex91 ] there is a repetition and a permutation of intervals that correspond to the earlier ri interval on @xmath1298 within @xmath1430 , and then omission of the ri intervals that correspond to @xmath1431 , @xmath1432 within @xmath1426 . although initially unfolding increases the size of a tree , its point is to reduce tile levels . tile @xmath1324 is level @xmath157 whereas its replacement @xmath1430 is a level @xmath21 ( basic end ) tile . unfolding is not defined as a transformation in the sense of section [ third ] because it is not local ; edges to later tiles may be revised in its application . first , we extend the notion of a tree to that of a tree of basic tiles with associated plays that are p - partitioned . [ defitree ] the tree of basic tiles @xmath1413 has associated plays @xmath1406 if @xmath1406 is the set of plays down the branches of @xmath1413 and for each @xmath1418 , @xmath1233 for some @xmath83 such that 1 . if @xmath1433 then @xmath835 is an interval on the tiles @xmath1434 in @xmath1413 that starts at the root of @xmath1237 , 2 . if @xmath257 and @xmath1435 then @xmath180 is an atomic leaf of @xmath1240 and @xmath1436 , 3 . if @xmath257 and @xmath1268 is a top or constant tile then @xmath1346 for some @xmath180 , 4 . if @xmath835 is ri and @xmath1347 then @xmath1240 belongs to the same family as @xmath1268 , 5 . if @xmath835 is ri and @xmath1268 is an embedded tile in @xmath4 then @xmath1346 for some @xmath180 , 6 . if @xmath835 is ri and @xmath1437 then @xmath1438 is not in the @xmath536-path of @xmath1237 . initially , when @xmath1413 is the tree of simple tiles constructed from the p - partitions of the plays in @xmath203 then @xmath1413 has associated plays @xmath925 ; parts ( 1 ) and ( 2 ) of definition [ defitree ] follow from definition [ partit ] of p - partition , the remainder from proposition [ vip ] . we assume that the definitions of b - partition , dependent , special , x - special , @xmath536-path and so on are extended to basic tiles in a tree of basic tiles . [ def93 ] assume @xmath1413 is a tree of basic tiles with associated plays . tile @xmath1032 _ is unfoldable at _ @xmath1237 if 1 . @xmath1032 is a top or embedded tile , 2 . @xmath1237 is the first tile in @xmath1439 that is a dependent of @xmath1032 , 3 . @xmath1032 and no later tile that is in the same family as @xmath1032 is x - special . if @xmath1032 is unfoldable at @xmath1237 then we define the unfolding of @xmath1032 at @xmath1237 in @xmath1413 as the tree @xmath1440 with the same nodes as @xmath1413 except that @xmath1237 is replaced by a basic tile @xmath1441 that is a composition of @xmath1032 and @xmath1237 . ( edges in @xmath1413 may also be changed in @xmath1440 . ) in the following we define the associated plays on @xmath1440 from those on @xmath1413 : the definition uses the notion of corresponding positions as defined ( for intervals ) in definition [ defsimilar ] . however , because @xmath1440 is different from @xmath1413 , the notion of correspondence is slightly weakened in specific circumstances to allow that a sequence of positions may correspond to a single position . [ def94 ] assume @xmath1442 is unfoldable at @xmath1443 in @xmath1413 and for each @xmath1408 , @xmath1444 . the _ unfolding of @xmath1032 at @xmath1237 _ in @xmath1413 is the tree @xmath1440 where @xmath1440 has the same tiles as @xmath1413 except that @xmath1237 is replaced by @xmath1445 . for each play @xmath1233 on @xmath1413 there is a play @xmath1446 on @xmath1440 , defined as follows in stages and top down : 1 . if @xmath1447 or @xmath1408 and @xmath1245 then @xmath1448 and @xmath1449 iff @xmath1450 ; 2 . if @xmath1408 , @xmath1451 , @xmath1452 and @xmath1453 then @xmath1454 is the continuation from the head of @xmath1455 in @xmath1456 with edges @xmath1457 . any position @xmath1458 in @xmath1454 corresponds to @xmath1459 in @xmath823 in the sense of definition [ defsimilar ] except in the following circumstances where the notion of correspondence is weakened : 1 . _ the positions are at @xmath1441 and @xmath1237 _ : if @xmath1460 and @xmath1461 then @xmath1462 corresponds to @xmath1459 where @xmath1463 is a shortest play on @xmath1032 in @xmath1441 that ends at @xmath1464 , 2 . _ there is a jump into @xmath1441 and @xmath1032 _ : as a result of move c4 of figure [ game ] , @xmath1465 and @xmath1466 ; then @xmath1467 corresponds to @xmath1468 if these intervals are internal to @xmath1441 and @xmath1032 , 3 . _ the positions are at @xmath1469 and @xmath1470 _ : @xmath1471 , @xmath1472 and @xmath1473 is the first later position such that @xmath1474 and @xmath1475 is the play on @xmath1032 in the b - partition for @xmath1473 ; then @xmath1458 corresponds to @xmath1476 . + for the edges : if @xmath1455 is @xmath1237 or is below @xmath1237 and @xmath1477 then @xmath1478 ; otherwise , @xmath1479 if @xmath1480 . we say that @xmath1481 on @xmath1440 is _ the companion _ of @xmath536 on @xmath1413 and that @xmath1440 is _ an unfolding _ of @xmath1413 . in example [ ex91 ] , @xmath1298 is unfolded at @xmath1324 ; @xmath1413 is the upper and @xmath1440 the lower tree in figure [ ex92 ] . the companion play @xmath1463 on @xmath1440 of @xmath1010 on @xmath1413 has the same intervals up to and including stage @xmath419 . in defining the weakened correspondence , play in @xmath1463 is at @xmath1430 in @xmath1440 and in @xmath1010 at @xmath1324 in @xmath1413 : so @xmath1482 where @xmath1483 is the initial play on @xmath1298 in @xmath1430 now corresponds to @xmath1484 ; next there is a jump into @xmath1430 and @xmath1298 by move c4 which is to @xmath1485 and @xmath1486 and so this @xmath1463 position corresponds to the interval that is from @xmath1487 to the lambda node directly above @xmath1324 ; so then the next positions will again correspond . in both these cases , where an interval corresponds to a position , the interval must be ri . [ prop95 ] assume @xmath1413 has associated plays @xmath1406 . if @xmath1440 is an unfolding of @xmath1413 then @xmath1440 has associated plays @xmath1488 . assume that @xmath1413 has associated plays @xmath1406 and @xmath1440 is the unfolding of @xmath1032 at @xmath1489 in @xmath1413 . by definition , if @xmath1447 and @xmath1481 is its companion then @xmath1490 is a play on a branch of @xmath1440 as required . otherwise , assume @xmath1442 , @xmath1444 for each @xmath1408 , @xmath1491 and @xmath1445 . we now show that we can find corresponding positions @xmath1458 in @xmath1454 and @xmath1459 in @xmath823 as described in definition [ def94 ] for each @xmath1492 . for @xmath1245 , this holds because @xmath1493 and @xmath1494 iff @xmath1495 . consider next the case that corresponding positions are at @xmath1441 and @xmath1237 : @xmath1496 and @xmath1497 . we examine the b - partitions for @xmath1498 and @xmath1499 and their component plays @xmath1463 , @xmath1010 on @xmath1032 ; we show that @xmath1463 and @xmath1010 correspond , that they are ri and that they are shortest plays on @xmath1032 ( that end at the leaf @xmath1500 ) . we prove this by induction on corresponding positions @xmath1498 , @xmath1499 , @xmath1501 . the base case is when @xmath1502 . at this point the b - partitions for @xmath1498 and @xmath1499 are the same . let @xmath1463 be the component interval in this b - partition on @xmath1032 . that @xmath1463 is ri and is a shortest play on @xmath1032 that ends at @xmath760 follows from the fact that @xmath1237 is the first tile that is a dependent of @xmath1032 in the sequence @xmath1439 , that @xmath1032 is a top or embedded tile and that @xmath1032 and all later tiles in the same family are not x - special ; consequently , any tile between @xmath1032 and @xmath1237 in the same family as @xmath1032 has an associated ri interval and , therefore , if play in such an interval were at a different atomic leaf of @xmath1032 than @xmath760 , then @xmath1237 would not be a dependent of @xmath1032 . for the inductive step the argument is similar after noting the following property : if corresponding positions are at @xmath1441 and @xmath1237 then it is not possible that play was in a dependent tile @xmath1333 of @xmath1032 that is below @xmath1237 before jumping back into @xmath1032 , then to @xmath1503 and then proceeding to @xmath1237 because in @xmath1481 the simulating position would be at @xmath1237 in @xmath1441 ( because in @xmath1440 , @xmath1333 is a dependent of @xmath1441 and edges are updated in @xmath1440 ) . so this property holds . therefore , returning to the main argument , the continuation from @xmath1498 consists first of a sequence of moves @xmath1483 that corresponds to @xmath1463 except it is on @xmath1032 within @xmath1441 ; therefore , @xmath1504 weakly corresponds to @xmath1459 . consider next corresponding positions such that at the next positions they jump into @xmath1441 and @xmath1032 ; @xmath1505 and @xmath1466 . then the intervals @xmath1468 , @xmath1467 that are internal to these tiles correspond except that they take place in @xmath1032 and @xmath1441 . the final circumstance to examine is that corresponding positions are @xmath1458 , @xmath1459 such that @xmath1506 and @xmath1507 . we show that there is a first position such that @xmath1508 and for some @xmath216 , @xmath1475 is the play on @xmath1032 in the b - partition for @xmath1473 and the b - partition for @xmath1458 contains the ri interval @xmath1509 which corresponds to @xmath1476 . positions @xmath1458 and @xmath1459 must be the result of a c4 move . however , the look - up table where this entry is defined must be at a position within a tile @xmath1333 that is a dependent of @xmath1032 in @xmath1413 and of @xmath1441 in @xmath1440 that is , below @xmath1237 . now the result follows as @xmath1032 and all later tiles in the same family are not x - special . clearly , @xmath1440 has no other plays than the companions of @xmath1406 . moreover , each companion play @xmath1481 obeys the six conditions in definition [ defitree ] given that they hold for each @xmath536 on @xmath1413 . to prove the small model property , assume a smallest term @xmath4 such that @xmath1510 and let @xmath1413 be its tree of simple tiles that captures the p - partitions of every @xmath1026 . first , we describe the proof for a particular case of @xmath1413 , a _ general atoms case _ , that obeys the following condition : if @xmath1032 is not a constant tile then it is not x - special . what this means is that every top and embedded tile is unfoldable at a first dependent even though some of these tiles may be x - special . ] . the decidability proof now reduces to the @xmath12rd - order case as the tree - model property is regained . initially , restricting further , assume @xmath10 is @xmath96th - order : there are , therefore , only two levels of non - constant tiles , top and end tiles . starting top down with @xmath1511 , at each stage @xmath1512 , a top tile @xmath763 that is closest to the root and that has dependents is unfolded at a first dependent @xmath761 . the construction finishes at some stage @xmath83 , when @xmath1513 has no unfoldable tiles ; that is , when it has no top tiles with dependents . this means that @xmath1513 only consists of constant tiles and basic top tiles that are also end tiles . once the tree is in this form , only the special tiles ( those that are constant tiles and play separators ) need to be kept : the remainder are redundant using transformation * t2. to obtain a small term , the initial @xmath181 of @xmath4 is placed at its top and the constant @xmath535 is placed below every leaf @xmath195 . the bound on the size of @xmath4 is larger than in fact [ bounds ] because the units are now _ basic _ tiles instead of simple tiles . in the worst case , each whole tile consists of one simple top tile and at most @xmath41 simple end tiles in any branch ( where @xmath41 is the arity of @xmath10 ) . therefore , using this construction , we obtain the following bound where the measures are all from @xmath10 : @xmath1514 . for higher - orders , the bounds become even larger . initially , all top and embedded tiles with dependents are unfoldable . now unfolding is iterated . at each stage @xmath1512 , a top or embedded tile @xmath763 with dependents which has the _ greatest _ level is unfolded ; if there is more than one such tile then one that is closest to the root is chosen to be @xmath763 and it is unfolded at a first dependent to give @xmath1515 . consider what may happen when @xmath10 is @xmath419th - order . there are now three levels of non - constant tiles : top , middle and end . unfolding reconciles embedded middle tiles with their immediate end dependents , which may in turn create larger embedded middle tiles or end tiles . for instance , assume the following branch of tiles in @xmath1516 @xmath1517 where @xmath763 is a top tile , @xmath1042 , @xmath1047 , @xmath1063 are dependents of @xmath763 , @xmath1047 and @xmath1063 are both embedded middle tiles because of @xmath1042 and @xmath1518 is a dependent of @xmath1047 and @xmath1519 of @xmath1063 . first , @xmath1047 is unfolded at @xmath1518 and then @xmath1063 is unfolded at @xmath1519 to give the following sequence . @xmath1520 the situation has reduced to the @xmath96th - order case , as there are now only two levels of tiles . so , the complete unfolding is the following sequence of top tiles that are also end tiles . @xmath1521 therefore , using t2 * we can remove any non - special tiles from the unfolded tree and any embedded ( basic ) end tiles within special tiles . the following is a very crude size bound on a smallest term that solves the problem @xmath10 of order @xmath1522 in this restricted case : @xmath1523 where @xmath1524 and @xmath1525 . in the general atoms case there is a bounded size solution term with a simple form that is a _ transferring _ term @xcite . for instance , consider the definition of transferring in @xcite ; for every subterm of @xmath1526 of the form @xmath1527 the free variables of any @xmath517 belong to @xmath1528 . let us return to the unrestricted case . we show that there is a bounded size term that _ almost _ has the tree model property . not all top or embedded tiles with dependents can be unfolded because of their contribution to solving @xmath10 ; for instance , in example [ ex91 ] , tile @xmath1042 can not be unfolded at @xmath1423 . there is also a further issue that does not occur with the general atoms case . after unfolding , we need to extract a term from the unfolded tree . the intention is that the edge relation @xmath1247 should be the subtree relation . however , there can be multiple edges of the form @xmath1529 and @xmath1530 : so , we need to guarantee that the `` subterms '' rooted at @xmath1268 and @xmath1531 are compatible . [ defi816 ] the tree @xmath1413 has the _ subterm property _ if there is a smallest equivalence relation @xmath1532 on its tiles such that whenever @xmath1533 , 1 . they are syntactically the same tile , 2 . if @xmath1534 and @xmath1535 then @xmath1536 . if @xmath1413 has the subterm property then its _ extraction _ is defined top - down . for any edge @xmath1537 the tile @xmath1237 is moved to be directly below @xmath1538 . the initial @xmath181 of @xmath4 is placed at the top and the constant @xmath535 is placed below every leaf @xmath195 . next we compute the smallest equivalence relation @xmath1532 starting with the identity , for each tile @xmath1539 and then closing up inductively under the conditions in definition [ defi816 ] : tiles @xmath1533 are identified . [ lem818 ] if @xmath1413 has has the subterm property then its extraction is a term . to obtain the small model property , we show that given the initial tree there is a way of unfolding such that the resulting tree has the subterm property . we then examine the extraction and apply the transformation * t2 * to remove any redundant tiles . the upper bound is very crude . [ thm93 ] if @xmath4 is a smallest solution of @xmath10 of order at most @xmath1540 , then @xmath1541 where @xmath41 is the arity of @xmath10 , @xmath1542 , @xmath1524 and @xmath1525 . assume @xmath4 is a smallest solution to @xmath10 of order @xmath1543 . let @xmath1516 be the tree of simple tiles for the p - partitions of each play in @xmath203 . by fact [ fact92 ] there are at most @xmath1544 special tiles in the tree . the issue is to define a largest sequence of unfolds such that afterwards the resulting tree has the subterm property . if every top and embedded tile is unfoldable then at each stage @xmath1545 those top or embedded tiles with the largest level that have dependents are chosen ; one of them that is closest to the root is then unfolded at a first dependent ; so @xmath1515 is an unfolding of @xmath1512 . unfolding is continued until @xmath1513 for some @xmath83 when there are no more unfoldable tiles ; an easy argument shows that there is such an @xmath83 as the unfolding reduces tile level . by proposition [ prop95 ] , it follows that companion plays preserve final states . after unfolding , every tile in the resulting tree is a basic top tile or a constant tile and , consequently , obeys the subterm property . so , its extraction @xmath180 is a term . except for the special tiles in @xmath180 , the remainder are redundant using transformation * t2. therefore , this produces a term @xmath582 where @xmath1546 . because @xmath4 is a smallest term that solves @xmath10 , @xmath4 obeys this bound . otherwise , some top or embedded tiles must be excluded from being unfolded . in the worst case there are at most @xmath1079 x - special tiles in @xmath1516 and , therefore , at most @xmath1547 distinct tiles in @xmath1516 that are x - special or dependents of x - special tiles and that are also not embedded tiles . moreover , there are at most a further @xmath1548 tiles that are special . we now proceed with a sequence of unfoldings as follows : find a highest level unfoldable tile that is closest to the root and unfold it at a first dependent ; and keep repeating this until there are no more unfoldable tiles . if the resulting tree @xmath1513 fails to have the subterm property then at some earlier stage @xmath241 there are tiles @xmath1032 and @xmath1455 in @xmath1549 that are syntactically the same and one of them , say @xmath1032 is later unfolded whereas @xmath1455 is not and both have edges to the same atomic leaf : a failure of condition ( 2 ) of definition [ defi816 ] . both these tiles are dependents of the same tile in @xmath1549 . therefore , we exclude the tile that @xmath1032 is a dependent of from being unfolded and then examine a complete sequence of unfoldings without it . by repeating this argument , after the sequence of unfoldings the resulting tree has the subterm property ; the number of distinct tiles that are excluded from being unfolded is at most @xmath1547 . by proposition [ prop95 ] its extraction @xmath180 solves @xmath10 . to begin with there are @xmath1550 tiles that are either special or dependents of x - special tiles . again , we wish to apply the transformation t2 * to remove redundant end tiles from @xmath180 . the question is how many extra unfolded tiles are also special because they have become separators . this depends on the number of plays @xmath1070 . if @xmath1551 then no extra tiles are needed . if @xmath1552 then for each of the @xmath1553 tiles , there could be at most @xmath1071 new play separators in the full unfolding . therefore , the result follows as each tile has bounded length @xmath1554 . although we have shown that higher - order matching is decidable , the upper size bound is very coarse and more work needs to be done to make it more accurate . although our complexity analysis is in terms of term size , it crudely agrees with the known non - elementary complexity lower bound based on @xcite . as order of a problem increases , so does its level and , therefore , the size of a smallest solution term according to our analysis increases ( exponentially ) . implicit in the analysis are positive sensible algorithms for solving dual interpolation problems . the game - theoretic characterisation of dual interpolation allows us to examine incomplete terms . as a first step , given a problem @xmath10 its component set of simple tiles can be defined from the subtypes and the constants ( including forbidden constants and the new constant @xmath362 of section [ treecheck ] ) . from these , we can then define varieties of basic tiles that have no embedded end tiles . we can then proceed to construct a term tree in stages , first by seeing if there are basic top tiles that separate plays , and , thereby , continue recursively . otherwise , we need to check if there must be non - top basic tiles , and so on . in the worse case , we need to examine all possible terms whose size is bounded by theorem [ thm93 ] . it may be worth investing effort to implement a tool that builds such terms . an open question is whether the set of all solutions of an interpolation problem is independently characterisable . for instance , comon and jurski define tree automata that characterise all solutions to a @xmath7th - order problem @xcite . in section 8 of their paper they describe two problems with extending their automata beyond order @xmath7 . the first is that states of an automaton are constructed out of the observational equivalence classes of terms . up to a 4th - order problem , one only needs to consider finitely many terms . with @xmath96th and higher orders , this is no longer true and one needs to quotient the potentially infinite terms into their respective observational equivalence classes in order to define only finitely many states : however as padovani shows this procedure is , in fact , equivalent to the matching problem itself @xcite . the second problem is the term trees that the automata recognise . for a @xmath7th - order problem , there is an automaton that recognises its full set of solutions ( up to @xmath41-equivalence ) even though the syntax may be infinite . comon and jurski define a special kind of automata , @xmath1555-automata , to achieve this @xcite . the occurrence of a leaf @xmath1555 in a term tree represents any ( syntactically correct ) subtree . a @xmath1555 can not contribute to the solution of a matching problem . this is no longer true at @xmath96th - order , as illustrated by their example @xmath1556 . solutions of this problem include the terms @xmath1557 where for some @xmath1558 , every @xmath1559 , @xmath1560 , is the identity and @xmath1561 is the constant function @xmath298 . because @xmath24 is arbitrarily large , one can not use a bounded number of variables to capture all these terms @xcite . in the general case , by iterated unfolding , any solution term can be transformed into another solution term that only uses ( and reuses ) boundedly many variables . whether this is capturable using some kind of automaton ( such as a transducer ) is open . in subsequent work @xcite , we are able to overcome the first problem of comon and jurski at higher orders but not the second . we provide a tree - automata characterisation relative to a _ finite _ alphabet : given a problem @xmath10 , a finite set of variables and constants the ( possibly infinite ) set of terms that are built from those components and that solve @xmath10 is regular . the states of the automaton are built from abstractions of sequences of moves in a ( variant version of the ) game which works for all orders . although there is active research extending automata on words and trees to infinite alphabets which preserve `` good '' properties , such as decidability of non - emptiness , see @xcite for a recent survey , the results do not yet apply to the case caused by higher - order binding . as we briefly mentioned in section [ treecheck ] , ong has shown that the tree - checking game can be presented using game - semantics and innocent strategies . another question is whether this framework provides an alternative basis for understanding higher - order matching . i am indebted to luke ong for many extremely helpful discussions about matching . i would also like to thank the lmcs referees who engaged with the paper and proposed productive improvements . 999 barendregt , h. lambda calculi with types . in _ handbook of logic in computer science , vol 2 _ , ed . abramsky , s. , gabbay , d. and maibaum , t. , oxford university press , 118 - 309 , ( 1992 ) . comon , h. and jurski , y. higher - order matching and tree automata . _ lecture notes in computer science _ , * 1414 * , 157 - 176 , ( 1997 ) . dougherty , d. and wierzbicki , t. a decidable variant of higher order matching . _ lecture notes in computer science _ , * 2378 * , 340 - 351 , ( 2002 ) . huet , g. _ rsolution dquations dans les langages dordre @xmath21 , @xmath157 , @xmath9 @xmath1562_. thse de doctorat dtat , universit paris vii , ( 1976 ) . joly , t. on @xmath471-definability i : the fixed model problem and generalizations of the matching problem . _ fundamenta informaticae _ , * 65 * , 135 - 151 , ( 2005 ) . ong , c .- h . l. on model - checking trees generated by higher - order recursion schemes . _ proceedings of 21st annual ieee symposium on logic in computer science _ , ( lics ) , 81 - 90 , ( 2006 ) . ( longer version available from ong s web page , 55 pages preprint . ) schmidt - schau@xmath1563 , m. decidability of arity - bounded higher - order matching . _ lecture notes in artificial intelligence _ , * 2741 * , 488 - 502 , ( 2003 ) . segoufin , l. automata and logics for words and trees over an infinite alphabet . _ lecture notes in computer science _ , * 4207 * , 41 - 57 , ( 2006 ) .	we show that the higher - order matching problem is decidable using a game - theoretic argument .
searches for gravitational microlensing events that monitor millions of stars located in the galactic bulge and large or small magellanic clouds are being carried out by several groups ( macho : alcock et al . 1997 ; eros : ansari et al . 1996 ; ogle : udalski et al . 1997 ; duo : alard & guibert 1997 ) . through their efforts more than 300 candidate events have been detected . the light curve of a single - lens event is represented by @xmath0^{1/2 } , \eqno(1.1)\ ] ] where @xmath1 is the lens - source separation in units of the einstein ring radius @xmath2 , @xmath3 is the einstein ring radius crossing time ( the einstein time scale ) , @xmath4 is the impact parameter , and @xmath5 is the time of maximum amplification . the lensing parameters @xmath6 of the event are obtained by fitting the observed light curve to the theoretical curves given by equation ( 1.1 ) . among these parameters , the einstein time scale provides information about the lens because it is related to the physical lens parameters by @xmath7 where @xmath8 is the lens - source transverse motion , @xmath9 is the lens mass , and @xmath10 , @xmath11 , and @xmath12 are the separations between the observer , lens , and source star . on the other hand , when the lens is binary , the light curve deviates from that of the single - lens event in equation ( 1.1 ) . the most distinctive feature of binary - lens event light curve occurs as a source star crosses a lens caustic ( see 2 ) . whenever a source star crosses a caustic , an extra pair of micro images appears ( or disappears ) , producing a sharp spike in the light curve , giving rise to `` strong '' binary - lens events . the optimal condition for a strong binary - lens event is that the separation between the two binary components is comparable to the angular einstein ring radius @xmath13 , corresponding to the combined mass of the binary ( combined angular einstein ring radius ) . on the other hand , when the binary separation is too small or too large , the event has little chance to involve caustic crossings , resulting in a light curve with a relatively small deviation from that of a single - lens event ( i.e. a `` weak '' binary - lens event ; mao & paczyski 1991 ) . currently a total of 13 candidate binary - lens events have been reported . these include macho lmc#-1pt 1 ( dominik & hirshfeld 1994 , 1996 ; rhie & bennett 1996 ) , ogle#-1pt 6 ( mao & di stefano 1995 ) , ogle#-1pt 7 ( udalski et al . 1994 ) , duo#-1pt 2 ( alard , mao , & guibert 1995 ) , macho#-1pt 9 ( bennett et al . 1996 ) , 95-blg-12 , 96-blg-3 , 97-blg-1 , 97-blg-28 , 97-blg-41 , 98-blg-12 , 98-blg-42 , and 98-smc-1 ( http://darkstar.astro.washington.edu ) . detecting binary - lens events is important due to their astronomical applications . first , one can obtain physical information about binaries such as mass ratios , @xmath14 , and projected separations , @xmath15 . second , a strong binary - lens event provides an opportunity to measure how long it takes for the caustic line to transit the face of the source star . by using the measured caustic crossing time along with an independent determination of the source star size , one can measure the lens proper motion relative to the observer - source line - of - sight , and thus can estimate the location of the lens ( gould 1994 ; nemiroff & wickramasinghe 1994 ; witt & mao 1994 ; peng 1997 ) . caustic - crossing events can also be used to determine the radial surface brightness profiles of source stars ( albrow et al . finally , detection of fine structure in the light curve is an excellent method to discover extra - solar planetary systems ( mao & paczyski 1991 ; gould & loeb 1992 ; griest & safizadeh 1998 ) . to extract useful information from binary - lens events , it is essential to precisely determine the binary - lens parameters ( see 2.1 ) . however , the surface of @xmath16 behaves in a very complicated manner over the multi - parameter space ( albrow et al . as a result , multiple local minima exist , causing ambiguity in the binary - lens fit . recently , routine astrometric followup observations of microlensing events with high precision instruments such as the _ space interferometry mission _ ( hereafter sim , http://sim.jpl.nasa.gov ) have been discussed as a method to measure the distance and mass of machos ( h-1pt g , novikov , & polnarev 1995 ; paczyski 1998 ; boden , shao , & van buren 1998 ; han & chang 1999 ) . when a microlensing event is due to a single - point lens , the observed source star image is split into two , and the location of the center of light ( centroid ) between the two separate images with respect to the position of the source star traces out an ellipse ( walker 1995 ; jeong , han , & park 1999 ) . however , if the lens is a binary , both the number and locations of the images differ from those of a single - lens event , resulting in a centroid shift trajectory that deviates from an ellipse . in this paper , we study the astrometric properties of gravitational microlensing events caused by binary lenses . by investigating the centroid shifts for various types of binary - lens events , we find that the deviations of the centroid shift trajectories from the ellipses of single - lens events are characterized by distortions , twistings , and big jumps . we study the conditions of binary - lens system configurations and source star trajectories for individual types of deviations . we also find dramatic differences in the astrometric centroid shifts for binary - lens events that would be degenerate had their parameters been determined photometrically . therefore , when astrometric observations of binary - lens events are available in addition to a light curve , one can resolve the ambiguity of the binary - lens fit , and thus determine the binary - lens parameters uniquely . when lengths are normalized to the combined einstein ring radius , the lens equation of a binary - lens event in complex notation is given by @xmath17 where @xmath18 and @xmath19 are the mass fractions of individual lenses , @xmath20 and @xmath21 are the positions of the lenses , @xmath22 and @xmath23 are the positions of the source and images , and @xmath24 denotes the complex conjugate of @xmath25 ( witt 1990 ) . the amplification of each image , @xmath26 , is given by the jacobian of the transformation ( 2.1 ) evaluated at the image s position , i.e. , @xmath27 the images and source positions with infinite amplifications , i.e. @xmath28 , form closed curves , called critical curves and caustics , respectively . the total amplification of a source star is given by the sum of the individual amplifications , i.e. @xmath29 , with @xmath26 given by equation ( 2.2 ) . the position of the source star image centroid is the amplification - weighted average of the positions of individual images , i.e. @xmath30 , each of which is obtained by solving the lens equation ( 2.1 ) numerically . to characterize a single - band light curve of a binary - lens event , at least seven parameters are required . these include the binary separation @xmath15 and the mass ratio @xmath14 , the closest separation between the source and the center of the mass of the binary system @xmath4 ( impact parameter of a binary - lens event ) , the time of the closest approach to the center of mass @xmath5 , the angle that the source star trajectory makes with the projected binary axis @xmath31 ( approaching angle ) , the combined einstein ring radius crossing time scale @xmath3 , and finally , the baseline source star flux @xmath32 . in actual experiments in which observations are performed toward very dense star fields one should include an additional parameter for the background flux @xmath33 to account for blending . due to the combination of a large number of binary - lens parameters and the complicated behavior of the @xmath16 surface over parameter space , fitting the light curve of a binary - lens event is a formidable task ( albrow et al . the most serious problem is that there can be multiple local minima , but verifying that one of the local minima is a global solution is very difficult . the ambiguity in binary - lens fits is demonstrated well by dominik ( 1999 ) . he showed that several models having a large variety of parameters agree well with the observed binary - lens events ogle#-1pt 7 and duo#-1pt 2 . in figure 1 , we reproduce four different model light curves based on the solutions to the binary - lens parameters determined by dominik ( 1999 ) and the measured @xmath34-band data points of the binary - lens event ogle#-1pt 7 ( udalski et al . 1994 ) . in figure 2 , we present the configurations of the binary - lens system and the source star trajectories for the individual model fits corresponding to the model light curves in figure 1 . in figure 2 , the projected locations of the lenses are marked by ` x ' and the caustics and source star trajectories are represented by the thick curves and thin straight lines respectively . in each panel of figure 2 , we also mark the values of the binary - lens parameters @xmath15 , @xmath14 , and @xmath3 . one finds that despite the similarity of the light curves , the resulting binary - lens parameters for individual model fits are significantly different from each other . when a source star is microlensed by a single point - mass lens , its image is split into two . the typical separation between the two images for a galactic bulge event caused by a typical stellar mass lens is on the order of a milliarcsecond , and thus one can not resolve the individual images . however , the source star image centroid shifts caused by microlensing can be measured with high - precision interferometers such as the sim that will have a positional accuracy of @xmath35 . the centroid shift for a single - lens event is related to the lensing parameters by @xmath36;\qquad \theta_{\rm e } = \left ( { 4gm\over c^2}{d_{ls}\over d_{ol}d_{os } } \right)^{1/2 } , \eqno(2.2.1)\ ] ] where @xmath37 is the angular einstein ring radius , the @xmath38 and @xmath39 axes represent the directions which are parallel and normal to the lens - source transverse motion and @xmath40 and @xmath41 are the unit vectors for these corresponding directions . if we let @xmath42 and @xmath43 , where @xmath44 , the coordinates are related by @xmath45 where @xmath46 and @xmath47 . therefore , the trajectory of the apparent source star image centroids traces out an ellipse during the event , known as the ` astrometric ellipse ' ( walker 1995 ; jeong et al . 1999 ) . if a lens is binary , on the other hand , the number and positions of the source star images differ from those of a single - lens event(safizadeh , dalal , & griest 1998 ) . as a result , the centroid shift trajectory of the binary lens event deviates from an ellipse . to see the pattern of the deviations , we compute the centroid shifts for events caused by binaries with various separations and axis ratios and show their trajectories in figure 3 . in figure 4 , we also present the lens system configurations and source star trajectories which are responsible for the centroid shifts in figure 3 . in figure 4 , the caustics and positions of the lenses are represented in the same manner as in figure 2 and the source star trajectories are represented by straight lines with line types chosen to match those of the centroid shift trajectories in figure 3 . the impact parameters of the individual source star trajectories are @xmath48 ( solid lines ) , 0.5 ( dotted lines ) , and 0.7 ( dashed lines ) , and the approaching angle @xmath31 is @xmath49 for all trajectories . from figure 3 , one finds that the diversity of the centroid shift trajectories expected from binary - lens events is very large , similar to photometric light curves . however , despite the large variations , one finds that the deviations of the centroid shift trajectories from the astrometric ellipses of single - lens events can be classified by having _ distortions _ , _ twistings _ , and _ big jumps_. first , distortions in the trajectories of centroid shifts occur for non - caustic - crossing binary - lens events in which binary separations are relatively small compared to their impact parameters . for an event with a very close binary separation , i.e. @xmath50 , the centroid shift trajectory is well approximated by the astrometric ellipse of a single - lens event with mass equal to the total mass of the binary and located at the center of mass of the binary . however , we note that for the same binary - lens event the fractional deviation of the astrometric centroid shifts from those of the corresponding single - lens event is significantly larger than the fractional deviation of the photometric amplifications , making astrometrically observing microlensing events a useful method to detect very close binaries ( chang & han 1999 ) . the amount of deviation increases with an increasing ratio of @xmath51 . secondly , when the binary separation is equivalent to the impact parameter , i.e. @xmath52 , the trajectory of the centroid shift becomes twisted , developing a loop near the closest approach . for a given lens system configuration , the size of the loop is bigger for events with smaller impact parameters . we note that similar distortions and twistings occur due to the effect of finite source size ( mao & witt 1998 ) . however , while the source star trajectories deviated by the finite source effect are symmetric with respect to the line connecting the position of the unlensed source star and the centroid at the time of maximum amplification , the centroid shift trajectories caused by binary lenses are , in general , asymmetric . therefore , one can distinguish the trajectories of astrometric centroid shifts caused by binary lenses from those affected by the finite source effect . finally , whenever the source star crosses a caustic , the centroid of the source star images shifts by a large amount , up to about @xmath53 - @xmath54 ( witt & mao 1996 ) , producing a big jump in the centroid shift trajectory . in the previous subsection , we showed that , depending on the lens - system configurations and source star trajectories , i.e. binary - lens parameters , the trajectories of the centroid shifts of binary - lens events take various forms . thus , one can determine the binary - lens parameters from the centroid shift trajectory in a manner similar to the determination of parameters from the light curve . to fit a centroid shift trajectory , however , the number of parameters is almost the same as the number to fit a light curve . because the centroid shifts are not only affected by the flux of the blended star , but also by its location . however , one can correct for the effect of nearly all types of blending , and thus both the values of @xmath55 and @xmath33 can be determined by using the diffraction - limited resolution of space - borne observations and the high positional accuracy of the sim ( han & kim 1999 ) . therefore , the effective number of parameters necessary to describe the centroid shift trajectory is one less than the number of parameters needed to characterize the light curve . ] the result is that the astrometric determination of the binary - lens parameters suffers from the same degree of ambiguity as the photometric determination . therefore , there is no advantage to determining the binary - lens parameters from the centroid shifts alone . however , one can completely resolve the ambiguity in finding the parameters by using both the photometric light curve and the astrometric centroid shift trajectory . this is because the astrometric centroid shifts expected from the photometrically determined binary - lens parameters are significantly different from each other . to demonstrate this , in figure 5 we present the trajectories of the centroid shifts that are expected for the four degenerate lens - parameter solutions corresponding to the model light curves in figure 1 . from figure 5 one finds that despite the very similar the light curves , the trajectories of the centroid shifts differ dramatically from each other . to isolate a global solution , the measured centroid shift trajectory is simply compared to the solutions producing the photometric fit . the findings from our investigation of the astrometric properties of binary - lens gravitational microlensing events are summarized as follows . 1 . the trajectories of centroid shifts expected from binary - lens events take various forms depending on the lens - system configurations and source star trajectories . 2 . despite the large diversity of centroid shift trajectories , the deviations from the elliptical trajectories of single - lens events can be categorized into distortions , twistings , and large jumps . 3 . the centroid shift trajectories expected from different degenerate solutions of the photometrically determined binary - lens parameters are dramatically different from each other . therefore , with the additional information provided by the astrometrically determined centroid shifts , one can completely resolve the ambiguity of the photometric binary - lens fit , and thus can uniquely determine the binary - lens parameters . we would like to thank m. everett for careful reading of the manuscript . m. s. chun was supported by the financial grant from the korea research foundation ( 1998 - 015-d00287 ) in the program year of 1999 . alard , c. , mao , s. , & guibert , j. 1995 , , 300 , l17 alard , c. , & guibert , j. 1997 , , 326 , 1 albrow , m. d. , et al . 1999a , preprint ( astro - ph/9903008 ) albrow , m. d. , et al . 1999b , , submitted ( astro - ph/9811479 ) alcock , c. , et al . 1997 , , 486 , 697 ansari , r. , et al . 1996 , , 314 , 94 bennett , d. p. , et al . 1996 , nucl . , 51b , 152 boden , a. f. , shao , m. , & van buren , d. 1998 , , 502 , 538 chang , k. , & han , c. 1999 , , in press dominik , m. 1999 , , 341 , 943 dominik , m. , & hirshfeld , a. c. 1994 , , 289 , l31 dominik , m. , & hirshfeld , a. c. 1996 , , 313 , 841 gould , a. 1994 , , 421 , l71 gould , a. , & loeb , a. 1992 , , 396 , 104 griest , k. , & safizadeh , n. 1998 , , 500 , 37 han , c. , & chang , k. 1999 , , 304 , 845 han , c. , & kim , t .- w . 1999 , , 305 , 795 h-1pt g , e. , novikov , i. d. , & polnarev , a. g. 1995 , , 294 , 287 jeong , y. , han , c. , & park , s .- h . 1999 , , 511 , 569 mao , s. , & di stefano , r. 1995 , , 440 , 22 mao , s. , & paczyski , b. 1991 , , 374 , l37 mao , s. , & witt , h. j. 1998 , , 300 , 1041 nemiroff , r. j. , & wickramasinghe , w. a. d. t. 1994 , , 424 , l21 paczyski , b. 1998 , 404 , l23 peng , e. 1997 , , 475 , 43 rhie , s. h. , bennett , d. p. 1996 , nucl . . proc . , 51b , 86 safizadeh , n. , dalal , n. , & griest , k. 1998 , preprint ( astro - ph/9811233 ) udalski , a. , et al . 1994 , , 436 , l103 udalski , a. , et al . 1997 , acta . astron . , 47 , 169 walker , m. a. 1995 , , 453 , 37 witt , h. j. , & mao , s. 1994 , , 430 , 505 witt , h. j. , & mao , s. 1996 , in iau symp . 173 , astrophysical applications of gravitational lensing , ed . c. kochanek & j. hewitt ( dordrecht : kluwer ) , 233 the measured @xmath34-band data points of the binary - lens event ogle#-1pt 7 and the four different model light curves ( solid lines ) . the designations of individual model light curves ( bl2 , bl0 , bl1 , and bl3 ) are borrowed from dominik ( 1999 ) who originally determined the binary - lens parameters for the model light curves the configurations of the binary - lens systems and source star trajectories ( thin solid lines ) that correspond to the model light curves in figure 1 . for each configuration , the lens locations and the caustics are marked by ` x ' and thick solid lines respectively . also marked in each panel are the values of the binary - lens parameters @xmath15 , @xmath14 , and @xmath3 . one finds that despite the similarity of the light curves ( in figure 1 ) , the difference in the binary - lens parameters are significant . the lens system configurations and source trajectories ( thin solid lines ) which are responsible for the centroid shift trajectories in figure 3 . the caustics and the positions of lenses are represented in the same way as in figure 2 . the trajectories have the closest separations of @xmath48 ( solid lines ) , 0.5 ( dotted lines ) , and 0.9 ( dashed lines ) and have the same approaching angle of @xmath56 . the line type of each source star trajectory is chosen so that it matches with the line type of the centroid shift trajectory in figure 3 . the centroid shift trajectories that are expected for the four lens parameters solutions corresponding to the model light curves in figure 1 . one finds that despite the similarity in the light curves , the centroid shift trajectories are dramatically different from one another .	in this paper , we study the astrometric properties of gravitational microlensing events caused by binary lenses . by investigating the centroid shifts for various types of binary - lens events , we find that the deviations of the centroid shift trajectories from the elliptical ones of single - lens events are characterized by distortions , twistings , and big jumps . we study the conditions of binary - lens system configurations and source star trajectories for individual types of deviations . we find dramatic differences in the astrometric centroid shifts for binary - lens microlensing events that would be degenerate had their parameters been determined photometrically . therefore , when additional astrometric observations of a binary - lens event are available , one can resolve the ambiguity of the binary - lens fit , and uniquely determine the binary - lens parameters . 6.5 in & submitted to _ the astrophysical journal _ : apr 13 , 1999 preprint : cnu - a&ss-04/99
there are several approaches to calculate the @xmath4 transition form factors , such as the lattice qcd technique , the qcd light - cone sum rules ( lcsrs ) and the perturbative qcd ( pqcd ) approach . the pqcd calculation is more reliable when the involved energy scale is hard , i.e. in the large recoil regions ; the lattice qcd results of the @xmath5 transition form factors are available only for soft regions ; while , the qcd lcsrs can involve both the hard and the soft contributions below @xmath6 ( @xmath7 is a typical hadronic scale of roughly @xmath8 mev ) and can be extrapolated to higher @xmath9 regions . therefore , the results from the pqcd approach , the lattice qcd approach and the qcd lcsrs are complementary to each other , and by combining the results from these three methods , one may obtain a full understanding of the @xmath5 transition form factors in its whole physical region . in refs.@xcite , we have done a consistent analysis of the @xmath10 transition form factor in the whole physical region . similarly , one can obtain a deep understanding of the @xmath0 transition form factor in the physical energy regions by combining the qcd lcsr results with the pqcd results and by properly taking the @xmath1 breaking effects into account . the @xmath0 transition form factors are defined as follows : @xmath11 where the momentum transfer @xmath12 . if we confine ourselves to discuss the semi - leptonic decays @xmath13 , it is found that the form factors @xmath14 is irrelevant for light leptons ( @xmath15 ) and only @xmath16 matters , i.e. @xmath17 where @xmath18 is the usual phase - space factor . so in the following , we shall concentrate our attention on @xmath16 . the @xmath0 transition form factor has been analyzed by several groups under the qcd lcsr approach @xcite , where some extra treatments to the correlation function either from the b - meson side or from the kaonic side are adopted to improve their lcsr estimations . it is found that the main uncertainties in estimation of the @xmath0 transition form factor come from the different twist structures of the kaon wave functions . it has been found that by choosing proper chiral currents in the lcsr approach , the contributions from the pseudo - scalars twist-3 structures to the form factor can be eliminated @xcite . in the present paper , we calculate the @xmath0 form factor with chiral current in the lcsr approach to eliminate the most uncertain twist-3 light - cone functions contributions . and more accurately , we calculate the @xmath19 corrections to the kaonic twist-2 terms . the @xmath1-breaking effects from the twist-2 and twist-4 kaon wave functions shall also be discussed . in ref@xcite , we have calculated the @xmath0 transition form factor up to @xmath3 in the large recoil region within the pqcd approach @xcite , where the b - meson wave functions @xmath20 and @xmath21 that include the three - fock states contributions are adopted and the transverse momentum dependence for both the hard scattering part and the non - perturbative wave function , the sudakov effects and the threshold effects are included to regulate the endpoint singularity and to derive a more reliable pqcd result . further more , the contributions from different twist structures of the kaon wave function , including its @xmath1-breaking effects , are discussed . so we shall adopt the pqcd results of ref.@xcite to do our discussion , i.e. to give a consistent analysis of the @xmath0 transition form factor in the large and intermediate energy regions with the help of the lcsr and the pqcd results . the paper is organized as follows . in sec.ii , we present the results for the @xmath0 transition form factor within the qcd lcsr approach . in sec.iii , we discuss the kaonic das with @xmath1 breaking effect being considered . especially , we construct a model for the kaonic twist-2 wave function based on the two gegenbauer moments @xmath22 and @xmath23 . numerical results is given in sec.iv , where the uncertainties of the lcsr results and a consistent analysis of the @xmath0 transition form factor in the large and intermediate energy regions by combining the qcd lcsr result with the pqcd result is presented . the final section is reserved for a summary . the sum rule for @xmath16 by including the perturbative @xmath19 corrections to the kaonic twist-2 terms can be schematically written as @xcite @xmath24 ds \;,\ ] ] where @xmath25 is the contribution from the twist-2 da and @xmath26 is for twist-4 da , @xmath27 is the b - meson decay constant . the borel parameter @xmath28 and the continuum threshold @xmath29 are determined such that the resulting form factor does not depend too much on the precise values of these parameters ; in addition the continuum contribution , that is the part of the dispersive integral from @xmath29 to @xmath30 that has been subtracted from both sides of the equation , should not be too large , e.g. less than @xmath31 of the total dispersive integral . the functions @xmath25 and @xmath26 can be obtained by calculating the following correlation function with chiral current @xmath32p_{\mu}+\pi_-[q^2,(p+q)^2]q_{\mu } \;. \label{chiral}\end{aligned}\ ] ] the calculated procedure is the same as that of @xmath10 form factor that has been done in refs.@xcite . so for simplicity , we only list the main results for @xmath0 and highlight the parts that are different from the case of @xmath10 , and the interesting reader may turn to refs.@xcite for more detailed calculation technology . as for @xmath25 , it can be further written as @xmath33 where @xmath34 is the renormalized hard scattering amplitude , @xmath35 stands for the b - quark pole mass @xcite . defining the dimensionless variables @xmath36 , @xmath37 and @xmath38 $ ] , up to order @xmath39 , we have @xmath40\nonumber\\ & & + \theta(\rho-1)\left [ 8 \left . \frac{\ln(\rho-1)}{\rho-1 } \right\|_{+ } + 2 \frac{1}{r_2-\rho } \left ( \frac{1}{\rho } -\frac{1}{r_2}\right)\right.\nonumber\\ & & \left.+2 \left ( \ln r_2 + \frac{1}{r_2 } -2 -2 \ln(r_2 - 1 ) + \ln \frac{m_b^{2}}{\mu^2 } \right ) \left . \frac{1}{\rho-1 } \right\|_{+ } + \frac{1-\rho}{\rho^2 } \right . + \nonumber\\ & & \frac{2 ( 1-r_1)}{(r_1-r_2)(r_2-\rho ) } \left ( \ln \frac{\rho } { r_2 } -2 \ln \frac{\rho-1}{r_2 - 1 } \right ) -\frac{4 \ln \rho}{\rho-1 } -\nonumber\\ & & \left . \frac{2 ( r_2 - 1)}{(r_1-r_2)(\rho - r_1)}\left ( \ln \rho -2 \ln(\rho-1 ) + 1 - \ln \frac{m_b^{2}}{\mu^2 } \right ) \right ] + \nonumber\\ & & \theta(1-\rho ) \left [ 2 \left ( \ln r_2 + \frac{1}{r_2 } -2 \ln(r_2 - 1 ) -\ln \frac{m_b^{2}}{\mu^2 } \right ) \left . \frac{1}{\rho-1 } \right\|_{+ } - \right . \nonumber\\ & & \left . \frac{2(1-r_2)}{r_2(r_2-\rho ) } \right . -\frac{2(1-r_1)}{(r_1-r_2)(r_2-\rho ) } \left ( 1+\ln \frac{r_2}{(r_2 - 1)^2}-\ln \frac{m_b^{2}}{\mu^2 } \right ) \right ] \right\},\end{aligned}\ ] ] for the case of @xmath41 and @xmath42 . as for the coefficients of @xmath43 , the higher power suppressed terms of order @xmath44 have been neglected due to its smallness . the dilogarithm function @xmath45 and the operation @xmath46 is defined by @xmath47\frac{1}{1-\rho}.\ ] ] in the calculation , both the ultraviolet and the collinear divergences are regularized by dimensional regularization and are renormalized in the @xmath48 scheme with the totally anti - commuting @xmath49 . and similar to ref.@xcite , to calculate the renormalized hard scattering amplitude @xmath50 , the current mass effects of @xmath51-quark are not considered due to their smallness . by setting @xmath52 , it returns to the case of @xmath10 and it can be found that the coefficients of @xmath53 and @xmath54 agree with those of refs.@xcite , while the coefficients of @xmath43 confirm that of ref.@xcite and differ from that of ref.@xcite . the present results can be checked with the help of the kernel of the brodsky - lepage evolution equation @xcite , since the @xmath55-dependences of the hard scattering amplitude and of the wave function should be compensate to each other . as for the sub - leading twist-4 contribution @xmath56 , we calculate it only in the zeroth order in @xmath39 , i.e. @xmath57}{u^3 m^4 } \right ) \nonumber \right . \\ & & + \int_{0}^{1 } dv \int d\alpha_i\frac { \theta(\alpha_1+v \alpha_3 -\delta)}{(\alpha_1+v \alpha_3)^2 m^2 } e^{-\frac { m_b^2-(1-\alpha_1- v \alpha_3 ) ( q^2-(\alpha_1 + v \alpha_3 ) m_k^2)}{m^2 ( \alpha_1+v \alpha_3)}}\times \nonumber\\ & & \left.\left ( 2 \varphi_\perp(\alpha_i)+2 \widetilde\varphi_\perp(\alpha_i ) -\varphi_\parallel(\alpha_i)-\widetilde\varphi_\parallel(\alpha_i ) \right ) \right\ } , \end{aligned}\ ] ] where @xmath58 , @xmath59 , @xmath60 and @xmath61 are three - particle twist-4 das respectively , and @xmath62 and @xmath63 are two - particle twist-4 wave functions . here , @xmath64 , @xmath65 and @xmath29 denotes the subtraction of the continuum from the spectral integral . by setting @xmath52 ( the lower integration range of @xmath66 should be changed to be @xmath67 for the case ) , we return to the results of @xmath10 @xcite . generally , the leading twist-2 da @xmath68 can be expanded as gegenbauer polynomials : @xmath69 . \label{phikgen}\ ] ] in the literature , only @xmath70 is determined with more confidence level and the higher gegenbauer moments are still with large uncertainty and are determined with large errors . alterative determinations of gegenbauer moments rely on the analysis of experimental data . the first gegenbauer moment @xmath71 has been studied by the light - front quark model @xcite , the lcsr approach @xcite and the lattice calculation @xcite and etc . in ref.@xcite , the qcd sum rule for the diagonal correlation function of local and nonlocal axial - vector currents is used , in which the contributions of condensates up to dimension six and the @xmath72-corrections to the quark - condensate term are taken into account . the moments derived there are close to that of the lattice calculation @xcite , so we shall take @xmath73 to do our discussion . at the scale @xmath74 gev , @xmath75 with the help of the qcd evolution . the higher gegenbauer moments , such as @xmath23 , are still determined with large uncertainty and are determined with large errors @xcite . for example , ref.@xcite shows that the value of @xmath23 is very close to the asymptotic distribution amplitude , i.e. @xmath76 ; while refs.@xcite gives larger values for @xmath23 , i.e. @xmath77 @xcite , @xmath78 @xcite and @xmath79 @xcite . it should be noted that the value of @xmath23 affects not only the twist-2 structure s contribution but also the twist-4 structures contributions , since the @xmath1-breaking twist-4 das also depend on @xmath23 due to the correlations among the twist-2 and twist-4 das as will be shown in the next subsection . since the value of @xmath23 can not be definitely known , we take its center value to be a smaller one , i.e. @xmath80 , for easily comparing with the results of ref.@xcite . further more , to study the uncertainties caused by the second gegenbauer moment @xmath23 , we shall vary @xmath23 within a broader region , e.g. @xmath81 $ ] , so as to see which value is more favorable for @xmath23 by comparing with the pqcd results . before doing the numerical calculation , we need to know the detail forms for the kaon twist-2 da and the twist-4 das . as for the twist-2 da , we do not adopt the gegenbauer expansion ( [ phikgen ] ) , since its higher gegenbauer moments are still determined with large errors whose contributions may not be too small , i.e. their contributions are comparable to that of the higher twist structures . for example , by taking a typical value @xmath82 @xcite , our numerical calculation shows that its absolute contributions to the form factor is around @xmath83 in the whole allowable energy region , which is comparable to the twist-4 structures contributions . recently , a reasonable phenomenological model for the kaon wave function has been suggested in ref.@xcite , which is determined by its first gegenbauer moment @xmath22 , by the constraint over the average value of the transverse momentum square , @xmath84 @xcite , and by its overall normalization condition . with the help of such model , a more reliable pqcd calculation on the @xmath0 transition form factors up to @xmath3 have been finished . in the following , we construct a kaon twist-2 wave function following the same arguments as that of ref.@xcite but with slight change to include the second gegenbauer moment @xmath23 s effect , i.e. @xmath85\frac{a_k}{x(1-x ) } \exp \left[-\beta_k^2 \left(\frac{k_\perp^2+m_q^2}{x}+ \frac{k_\perp^2+m_s^2 } { 1-x}\right)\right],\ ] ] where @xmath86 , @xmath87 are the gegenbauer polynomial . the constitute quark masses are set to be : @xmath88 and @xmath89 . the four parameters @xmath90 , @xmath91 , @xmath92 and @xmath93 can be determined by the first two gegenbauer moments @xmath22 and @xmath23 , the constraint @xmath94 @xcite and the normalization condition @xmath95 . for example , we have @xmath96 , @xmath97 , @xmath98 and @xmath99 for the case of @xmath100 and @xmath80 . quantitatively , it can be found that @xmath91 , @xmath92 and @xmath93 decreases with the increment of @xmath22 ; @xmath93 decreases with the increment of @xmath23 , while @xmath91 and @xmath92 increase with the increment of @xmath23 . under such model , the uncertainty of the twist-2 da mainly comes from @xmath22 and @xmath23 . it can be found that the @xmath1 symmetry is broken by a non - zero @xmath91 and by the mass difference between the @xmath51 quark and @xmath66 ( or @xmath101 ) quark in the exponential factor . the @xmath1 symmetry breaking effect of the leading twist kaon distribution amplitude has been studied in refs.@xcite and references therein . the @xmath1 symmetry breaking in the lepton decays of heavy pseudoscalar mesons and in the semileptonic decays of mesons have been studied in ref.@xcite . after doing the integration over the transverse momentum dependence , we obtain the twist-2 kaon da , @xmath102\nonumber\\ & & \times\exp \left[-\beta_k^2 \left(\frac{m_q^2}{x}+ \frac{m_s^2 } { 1-x}\right)\right]\left[1-\exp\left(-\frac{\beta_k^2\mu^2_0 } { x(1-x)}\right ) \right ] , \label{phik}\end{aligned}\ ] ] where @xmath103 for the present case . then , the gegenbauer moments @xmath104 can be defined as @xmath105 ^ 2 } \ ; , \ ] ] where @xmath106 other than @xmath107 is adopted to compare the moments with those defined in the literature , e.g. @xcite , since in these references @xmath108 stands for the momentum fraction of @xmath51-quark in the kaon ( @xmath109 ) , while in the present paper we take @xmath108 as the momentum fraction of the light @xmath110-(anti)quark in the kaon ( @xmath111 ) . the twist-3 contribution is eliminated by taking proper chiral currents under the lcsr approach , so we only need to calculate the subleading twist-4 contributions . the needed four three - particle twist-4 das that are defined in ref.@xcite can be expressed as @xcite @xmath112 , \nonumber\\ \widetilde{\varphi}_{\perp}(\alpha_i ) & = & -30 \alpha_3 ^ 2\left [ h_{00}(1-\alpha_3)+h_{01}\big[\alpha_3(1-\alpha_3)-6\alpha_1\alpha_2\big ] + h_{10}\big[\alpha_3(1-\alpha_3)-\frac{3}{2}(\alpha_1 ^ 2 + \alpha_2 ^ 2)\big]\right],\nonumber\\ { \varphi}_{\parallel}(\alpha_i ) & = & 120 \alpha_1\alpha_2\alpha_3 \left [ a_{10 } ( \alpha_1-\alpha_2)\right],\nonumber\\ \tilde{\varphi}_{\parallel}(\alpha_i ) & = & 120 \alpha_1\alpha_2 \alpha_3 \left [ v_{00 } + v_{10 } ( 3\alpha_3 - 1)\right],\end{aligned}\ ] ] where @xmath113 and @xmath114 with @xmath115 , @xmath116 and @xmath117 and @xmath118 @xcite . with the help of qcd evolution , we obtain @xmath119 and @xmath120 . it can be found that the dominant meson - mass effect are proportional to @xmath23 and @xmath121 , so if setting @xmath52 or the value of @xmath23 is quite small , then we return to the results of ref.@xcite . for the remaining two - particle twist-4 wave functions , their contributions are quite small in comparison to the leading twist contribution and even to compare with those of the three - particle twist-4 wave functions . and by taking the leading meson - mass effect into consideration only , they can be related to the three - particle twist-4 wave functions through the following way : @xmath122\ ] ] and @xmath123,\ ] ] which lead to @xmath124 + \nonumber\\ & & a_{10 } { \bar{u}}^3 ( 10 - 15\bar{u}+6\bar{u}^2)\ln \bar{u } + a_{10 } u^3 ( 10 - 15u+6u^2 ) \ln u , \\ g_2(u ) & = & \frac{5 \bar{u } u ( u -\bar{u})}{2 } \left [ 4 h_{00 } + 8 a_{10 } \bar{u } u - h_{10 } ( 1 + 5 \bar{u } u ) + 2 h_{01 } ( 1 - \bar{u}u)\right ] .\end{aligned}\ ] ] similarly , it can be found that when setting @xmath125 , the above expressions of @xmath62 and @xmath63 return to those of ref.@xcite . here by adopting the relations @xmath126 $ ] and @xmath127 , one can conveniently obtain the higher mass - correction terms for @xmath62 and @xmath63 on the basis of @xmath128 and @xmath129 derived in refs.@xcite , and numerically , it can be found that these terms contributions are indeed small . in the numerical calculations , we use @xmath130 next , let us choose the input parameters entering into the qcd sum rule . in general , the value of the continuum threshold @xmath29 might be different from the phenomenological value of the first radial excitation mass . here we set the threshold value of @xmath29 to be smaller than @xmath131 , whose root is slightly bigger than the mass of the b - meson first radial excitation predicted by the potential model @xcite . the pole quark mass @xmath35 is taken as @xmath132 . another important input is the decay constant of b meson @xmath133 . to keep consistently with the next - to - leading order calculation of twist-2 contribution , we need to calculate the two - point sum rule for @xmath133 up to the corrections of order @xmath39 . and in doing the numerical calculation , we shall adopt the nlo @xmath133 to calculate the nlo twist-2 contribution and lo @xmath133 for the lo twist-4 contributions for consistence . as a function of borel parameter @xmath28 at @xmath134 , where @xmath135 , @xmath136 , @xmath137 , @xmath138 . the solid line stands for the total contributions , the dashed line is for nlo result of the twist-2 kaonic wave function and the dotted line is for the lo result of twist-4 kaonic wave functions.,scaledwidth=50.0% ] the reasonable range for the borel parameter @xmath28 is determined by the requirement that the contributions of twist-4 wave functions do not exceed @xmath139 and those of the continuum states are not too large , i.e. less than @xmath31 of the total dispersive integration . at a typical @xmath134 , we draw @xmath140 versus @xmath28 in fig.([varym2 ] ) . it can be found that the contribution from the kaonic twist-2 wave function slightly increases with the increment of @xmath28 while the contributions from the kaonic twist-4 wave functions decreases with the increment of @xmath28 , as a result , there is a platform for @xmath140 as a function of the borel parameter @xmath28 for the range of @xmath141 . for convenience , we shall always take @xmath142 to do our following discussions . in the following we discuss the main uncertainties caused by the present lcsr approach with the chiral current . the present adopted chiral current approach has a striking advantage that the twist-3 light - cone functions which are not known as well as the twist-2 light - cone functions are eliminated , and then it is supposed to provide results with less uncertainties . in fact , it has been pointed out that the twist-3 contributions can contribute @xmath143 to the total contribution @xcite by using the standard weak current in the correlator , e.g. @xmath144 if the twist-3 wave functions are not known well , then the uncertainties shall be large . so in the literature , two ways are adopted to improve the qcd sum rule estimation on the twist-3 contribution : one is to calculate the above correlator by including one - loop radiative corrections to the twist-3 contribution together with the updated twist-3 wave functions @xcite ; the other is to introduce proper chiral current into the correlator , cf . eq.([chiral ] ) , so as to eliminate the twist-3 contribution exactly , which is what we have adopted . we shall make a comparison of these two approaches in the following . for such purpose , we adopt the following form for the qcd sum rule of ref.@xcite , which splits the form factor into contributions from different gegenbauer moments : @xmath145 where @xmath146 contains the contributions to the form factor from the asymptotic da and all higher - twist effects from three - particle quark - quark - gluon matrix elements , @xmath147 contains the contribution from the higher gegenbauer term of da that is proportional to @xmath22 , @xmath23 and @xmath148 respectively . the explicit expressions of @xmath149 can be found in table v and table ix of ref.@xcite . and in doing the comparison , we shall take the same da moments for both methods , especially the value of @xmath150 is determined from eq.([moments ] ) . for @xmath151 $ ] , @xmath152 $ ] and @xmath153gev$ ] . the solid line is obtained with @xmath154 , @xmath155 and @xmath156 ; the dashed line is obtained with @xmath157 , @xmath158 and @xmath159 , which set the upper and the lower ranges of @xmath160 respectively . as a comparison , the shaded band shows the result of ref.@xcite together with its @xmath161 theoretical uncertainty.,scaledwidth=50.0% ] we show a comparison of our result of @xmath160 with that of eq.([sumruleapp ] ) in fig.([compare ] ) by varying @xmath151 $ ] , @xmath152 $ ] and @xmath162gev$ ] . in fig.([compare ] ) the solid line is obtained with @xmath154 , @xmath155 and @xmath163 ; the dashed line is obtained with @xmath157 , @xmath158 and @xmath159 , which set the upper and the lower ranges of @xmath160 respectively . the shaded band in the figure shows the result of eq.([sumruleapp ] ) within the same @xmath22 and @xmath23 region and with its @xmath161 theoretical uncertainty @xcite . it can be found that our present lcsr results are consistent with those of ref.@xcite within large energy region @xmath164 $ ] . in another words these two treatments on the most uncertain twist-3 contributions are equivalent to each other , while the chiral current approach is simpler due to the elimination of the twist-3 contributions . one may also observe that in the lower @xmath9 region , different from ref.@xcite where @xmath160 increases with the increment of both @xmath22 and @xmath23 , the predicted @xmath165 will increase with the increment of @xmath23 but with the decrement of @xmath22 . this difference is caused by the fact that we adopt the model wave function ( [ model ] ) to do our discussion , whose parameters are determined by the combined effects of @xmath22 and @xmath23 ; while in ref.@xcite , @xmath22 and @xmath23 are varied independently and then their contributions are changed separately . .parameters for @xmath133 , where @xmath166 and @xmath133 are given in @xmath167 , @xmath29 and @xmath28 in @xmath168 . [ cols="^,^,^,^,^,^,^",options="header " , ] next we discuss the main uncertainties caused by the present lcsr approach with the chiral current . firstly , we discuss the uncertainties of @xmath140 caused by the effective quark mass @xmath35 by fixing @xmath169 and @xmath170 . under such case , the value of @xmath29 , the lo and nlo vales of @xmath133 should be varied accordingly and be determined by using the two - point sum rule with the chiral currents , e.g. to calculate the following two - point correlator : @xmath171 the sum rule for @xmath133 up to nlo can be obtained from ref.@xcite through a proper combination of the scalar and pseudo - scalar results shown there ,- quark mass to the present case of @xmath172-quark mass and we take @xmath173 @xcite and @xmath174 @xcite to do the numerical calculation . ] which can be schematically written as @xmath175 where the spectral density @xmath176 can be read from ref.@xcite . the borel parameter @xmath28 and the continuum threshold @xmath29 are determined such that the resulting form factor does not depend too much on the precise values of these parameters ; in addition , 1 ) the continuum contribution , that is the part of the dispersive integral from @xmath29 to @xmath30 , should not be too large , e.g. less than @xmath31 of the total dispersive integral ; 2 ) the contributions from the dimension - six condensate terms shall not exceed @xmath177 for @xmath133 . further more , we adopt an extra criteria as suggested in ref.@xcite to derive @xmath133 : i.e. the derivative of the logarithm of eq.([sumfb ] ) with respect to @xmath178 gives the b - meson mass @xmath179 , @xmath180 and we require its value to be full - filled with high accuracy @xmath181 . these criteria define a set of parameters for each value of @xmath182 . some typical values of @xmath133 are shown in tab.[tabfb ] , where @xmath133 is taken as the extremum within the reasonable region of @xmath183 and the value of @xmath35 is taken as @xcite : @xmath184 . @xmath133 decreases with the increment of @xmath182 . the nlo result agrees with the first direct measurement of this quantity by belle experiment @xmath185 mev from the measurement of the decay @xmath186 @xcite . as a function of @xmath9 with varying @xmath35 . the solid , dashed line and the dash - dot line are for @xmath138 , @xmath187 and @xmath188 respectively , where @xmath169 and @xmath170.,scaledwidth=50.0% ] the value of @xmath140 for three typical values of @xmath35 , i.e. @xmath138 , @xmath187 and @xmath188 respectively , are shown in fig.([varymb ] ) . @xmath140 increases with the increment of @xmath182 . it can be found that the uncertainty of the form factor caused by @xmath189 $ ] is @xmath190 at @xmath191 and increases to @xmath192 at @xmath193 . by taking a more accurate @xmath35 , e.g. @xmath194 as suggested by ref.@xcite , the uncertainties can be reduced to @xmath195 at @xmath191 and @xmath190 at @xmath193 . as a function of @xmath9 with varying @xmath196 , where @xmath80 . the solid line , the dashed line and the dash - dot line are for @xmath154 , @xmath197 and @xmath198 respectively . , scaledwidth=50.0% ] as a function of @xmath9 with varying @xmath199 , where @xmath100 . the solid line , the dashed line and the dash - dot line are for @xmath158 , @xmath200 and @xmath201 respectively . , scaledwidth=50.0% ] secondly , we discuss the uncertainties of @xmath140 caused by the twist-2 wave function @xmath202 , i.e. the two gegenbauer moments @xmath196 and @xmath199 . for such purpose , we fix @xmath135 and @xmath138 . to discuss the uncertainties caused by @xmath196 , we take @xmath80 . @xmath203 for three typical @xmath196 , i.e. @xmath154 , @xmath197 and @xmath198 respectively , are shown in fig.([varya1k ] ) . @xmath140 decreases with the increment of @xmath22 . it can be found that the uncertainty of form factor caused by @xmath204 $ ] is small , i.e. it is about @xmath205 at @xmath191 and becomes even smaller for larger @xmath9 . similarly , to discuss the uncertainties caused by @xmath199 , we fix @xmath206 . since the value of @xmath23 is less certain than @xmath22 , so we take three typical values of @xmath199 with broader separation to calculate @xmath140 , i.e. @xmath100 , @xmath200 and @xmath201 respectively . the results are shown in fig.([varya2k ] ) . it can be found that the uncertainty of the form factor caused by @xmath207 $ ] is also small , i.e. it is about @xmath208 at @xmath191 and becomes smaller for larger @xmath9 . @xmath140 increases with the increment of @xmath23 in the lower energy region @xmath209 and decreases with the increment of @xmath23 in the higher energy region @xmath210 . as a summary , a more accurate values for @xmath35 , @xmath22 and @xmath23 shall be helpful to derive a more accurate result for the form factor . our results favor a smaller @xmath23 to compare with the form factor in the literature , e.g. @xmath211 . and under such region , the uncertainties from @xmath23 is small , i.e. its uncertainty is less than @xmath208 for @xmath212 $ ] . it can be found that by varying @xmath151 $ ] and @xmath152 $ ] , the kaonic twist-4 wave functions contribution is about @xmath213 of the total contribution at @xmath191 . the uncertainties of @xmath22 shows that the @xmath1-breaking effect is small but it is comparable to that of the higher twist structures contribution . so the @xmath1 breaking effect and the higher twist s contributions should be treated on the equal footing . using the chiral current in the correlator , as shown in eq.([chiral ] ) , the theoretical uncertainty can be remarkably reduced . and our present lcsr results are consistent with those of ref.@xcite within large energy region @xmath164 $ ] , which is calculated with the correlator ( [ ballsr ] ) and includes one - loop radiative corrections to twist-2 and twist-3 contributions together with the updated twist-3 wave functions . in another words these two approaches are equivalent to each other in some sense , while the chiral current approach is simpler due to the elimination of the more or less uncertain twist-3 contributions . for higher energy region @xmath214 , the lcsr approach is no longer reliable . therefore the lattice calculations , would be extremely useful to derive a more reliable estimation on the high energy behaviors of the form factors . and @xmath215 , where @xmath138 , @xmath135 , @xmath216 , @xmath217 , @xmath100 and @xmath218 . the solid line and the dashed line are for @xmath140 and @xmath219 respectively . , scaledwidth=50.0% ] to have an overall estimation of the @xmath1 breaking effect , we make a comparison of the @xmath10 and @xmath0 form factors : @xmath220 and @xmath160 . the formulae for @xmath220 can be conveniently obtained from that of @xmath160 by taking the limit @xmath52 . in doing the calculation for @xmath220 , we directly use the gegenbauer expansion for pion twist-2 da , because different to the kaonic case , now the higher gegenbauer terms contributions are quite small even in comparison to the twist-4 contributions , e.g. by taking @xmath221 @xcite , our numerical calculation shows that its absolute contributions to the form factor is less than @xmath222 in the whole allowable energy region . we show a comparison of @xmath140 and @xmath219 in fig.([pik ] ) with the parameters taken to be @xmath138 , @xmath135 , @xmath216 , @xmath217 , @xmath100 and @xmath218 . secondly , by varying @xmath153gev$ ] , @xmath223 $ ] and @xmath224 $ ] , we obtain @xmath225 $ ] and @xmath226 $ ] . then we obtain @xmath2 , which favors a small @xmath1 breaking effect and is consistent with the pqcd estimation @xmath227 @xcite , the qcd sum rule estimations , e.g. @xmath228\approx 1.16 $ ] @xcite$ ] from ref.@xcite , we take @xmath229 . ] , @xmath230 @xcite and @xmath231 @xcite respectively , and a recently relativistic treatment that is based on the study of the dyson - schwinger equations in qcd , i.e. @xmath232=1.23 $ ] @xcite . recently , ref@xcite gives a calculation of the @xmath0 transition form factor up to @xmath3 in the large recoil region within the pqcd approach @xcite , where the b - meson wave functions @xmath20 and @xmath21 that include the three - fock states contributions are adopted and the transverse momentum dependence for both the hard scattering part and the non - perturbative wave function , the sudakov effects and the threshold effects are included to regulate the endpoint singularity and to derive a more reliable pqcd result . further more , the uncertainties for the pqcd calculation of the @xmath0 transition form factor has been carefully studied in ref.@xcite . so we shall adopt the pqcd results of ref.@xcite to do our discussion . only we need to change the twist-2 kaon wave function @xmath202 used there to the present one as shown in eq.([model ] ) . . the solid line is for our lcsr result , the dash - dot line is for the lcsr result of ref.@xcite with @xmath157 and @xmath158 . the shaded band is the pqcd result with @xmath233 $ ] and @xmath234 $ ] , where the dashed line is for the center values @xmath235 and @xmath236 , the upper edge of the band is for @xmath237 and @xmath238 and the lower edge of the band is for @xmath239 and @xmath240 . , scaledwidth=50.0% ] we show the lcsr results together with the pqcd results in fig.([lcsrpqcd ] ) . in drawing the figure , we take @xmath157 , @xmath158 and @xmath241 . and the uncertainties of these parameters cause about @xmath242 errors for the lcsr calculation . while for the pqcd results , we should also consider the uncertainties from the b - meson wave functions , i.e. the values of the two typical parameters @xmath243 and @xmath244 , and we take @xmath233 $ ] and @xmath234 $ ] @xcite . it can be found that the pqcd results can match with the lcsr results for small @xmath9 region , e.g. @xmath209 . then by combining the pqcd results with the lcsr results , we can obtain a consistent analysis of the form factor within the large and the intermediate energy regions . inversely , if the pqcd approach must be consistent with the lcsr approach , then we can obtain some constraints to the undetermined parameters within both approaches . for example , according to the qcd lcsr calculation , the form factor @xmath140 increases with the increment of b - quark mass , then the value of @xmath166 can not be too large or too small , is from the experimental value @xcite on @xmath133 . ] i.e. if allowing the discrepancy between the lcsr result and the pqcd results to be less than @xmath177 , then @xmath35 should be around the value of @xmath245 . in the paper , we have calculated the @xmath0 transition form factor by using the chiral current approach under the lcsr framework , where the @xmath1 breaking effects have been considered and the twist-2 contribution is calculated up to next - to - leading order . it is found that our present lcsr results are consistent with those of ref.@xcite within large energy region @xmath164 $ ] , which is calculated with the conventional correlator ( [ ballsr ] ) and includes one - loop radiative corrections to twist-2 and twist-3 contributions together with the updated twist-3 wave functions . and our present adopted lcsr approach with the chiral current is simpler due to the elimination of the more or less uncertain twist-3 contributions . the uncertainties of the lcsr approach have been discussed , especially we have found that the second gegenbauer moment @xmath23 prefers asymptotic - like smaller values . by varying the parameters within the reasonable regions : @xmath153gev$ ] , @xmath223 $ ] and @xmath246 $ ] , we obtain @xmath247 and @xmath248 , which are consistent with the pqcd and the qcd sum rule estimations in the literature . consequently , we obtain @xmath2 , which favors a small @xmath1 breaking effect . also , it has been shown that one can do a consistent analysis of the @xmath0 transition form factor in the large and intermediate energy regions by combining the qcd lcsr result with the pqcd result . the pqcd approach can be applied to calculate the @xmath0 transition form factor in the large recoil regions ; while the qcd lcsr can be applied to intermediate energy regions . combining the pqcd results with the qcd lcsr , we can give a reasonable explanation for the form factor in the low and intermediate energy regions . further more , the lattice estimation shall help to understand the form factors behaviors in even higher momentum transfer regions , e.g. @xmath214 . so , we suggest such a lattice calculation can be helpful . then by comparing the results of these three approaches , the @xmath0 transition form factor can be determined in the whole kinematic regions . this work was supported in part by the natural science foundation of china ( nsfc ) and by the grant from chongqing university . this work was also partly supported by the national basic research programme of china under grant no . the authors would like to thank z.h . wang and f.zuo for helpful discussions on the determination of @xmath133 .	we present an improved calculation of the @xmath0 transition form factor with chiral current in the qcd light - cone sum rule ( lcsr ) approach . under the present approach , the most uncertain twist-3 contribution is eliminated . and the contributions from the twist-2 and the twist-4 structures of the kaon wave function are discussed , including the @xmath1-breaking effects . one - loop radiative corrections to the kaonic twist-2 contribution together with the leading - order twist-4 corrections are studied . the @xmath1 breaking effect is obtained , @xmath2 . by combining the lcsr results with the newly obtained perturbative qcd results that have been calculated up to @xmath3 in ref.@xcite , we present a consistent analysis of the @xmath0 transition form factor in the large and intermediate energy regions . + * pacs numbers : * 12.38.aw , 12.38.lg , 13.20.he , 14.40.aq # 1#1 # 1 _
it is commonly believed that the blue - shifted galaxies are relatively nearby ones whose peculiar motion overcomes the hubble flow . all of the most distant galaxies ( and indeed the overwhelming majority of all galaxies ) are red - shifted . according to the conventional definition , the redshift of galaxies is the sum of two terms : the isotropic cosmic expansion velocity and the peculiar velocity owning to gravitational attraction by the surrounding matter . in practice , determining the peculiar velocity of a galaxy requires knowledge of both its observable radial velocity relative to some reference system and the distance to the galaxy determined independently of the radial velocity . according to the linear theory of gravitational instability , the peculiar velocities of galaxies are related to fluctuations in the mass ( peebles 1980 ) . burbidge & demoulin ( 1969 ) first observed ic 3258 with a blueshift of @xmath1490 km@xmath0s@xmath2 . they give three possible interpretations of their observations . first , ic 3258 is a member of the virgo cluster and has a very high velocity relative to the average for the cluster . second , ic 3258 is a field galaxy closer to the virgo cluster and its large velocity is just a random motion . third , ic 3258 has velocity because it has been ejected in an outburst involving one of the radio galaxies in the virgo cluster . several other blue - shifted galaxies appear in the direction of the virgo cluster . by measuring the distance @xmath5 of a galaxy , one can obtain the peculiar velocity of a galaxy @xmath6 , here @xmath7 is hubble expansion velocity , @xmath8 is observed velocity of the galaxy . since the hubble expansion velocity is small for nearby galaxies , the peculiar velocity could be negative . negative peculiar velocities are seen all over the region around the virgo cluster and this have long been seen as a reflex of the pull of the cluster on us ( aaronson et al . we live in the local supercluster , which is overdense part of the universe . so there is possibly the a local retardation of the cosmic expansion or a net infall within this region . in another example , an observer living on the outskirts of a large concentration is also pulled towards the overdense part of the clusters . when the radiation propagates inside the collapsing body it is blue - shifted . if this blueshift is greater than the redshift caused by the propagation of the radiation through expanding universe , distant observer can detect the gravitational blueshift from the collapsing object . also the agns have blue - shifted spectrum . bian et al . ( 2005 ) studied the radial velocity difference between the narrow emission - line components and of [ o iii ] @xmath9 and h@xmath10 in a sample of 150 sdss narrow - line seyfert@xmath01 galaxies . they found seven ` blue outliers ' with [ o iii ] blueshifted by more than 250 km@xmath0s@xmath2 . they interpreted the blueshift as possible result of the outflowing gas from the nucleus and the obscuration of the receding part of the flow by an optically thick accretion disk and on the viewing angle . spatial orientation of angular momentum of blue - shifted galaxies ( sdss ) has not been studied , so we are interested to carry out the spatial orientation of blue - shifted galaxies . an idea of the origin of angular momentum of galaxies is very important to understand the evolution of large scale structures of the universe . this paper is organized as follows : in sect . 2 we describe the sample used and the method of data reduction . in sect . 3 we describe the methods , statistical tools and the selection effects . finally , a discussion of the statistical results and the conclusions are presented in sects . 4 and 5 . we compiled a database of @xmath11 blue - shifted galaxies from the sloan digital sky survey seventh data release ( sdss dr7 ) . all sky distribution of blue - shifted galaxies is shown in fig . 1a . the inhomogeneous distribution of galaxies is because of the nature of the survey . the distribution of blue - shifted galaxies is shown in fig . we found a linear relationship between the blue - shift and logarithm of the number of galaxies ( @xmath12@xmath13log(@xmath14 ) ) . + we have retained only those galaxies that have blue - shift ( @xmath12 ) data at 95% level of significance . this removed 569 galaxies form the original data . since blue - shift is found to be decreases lineally with number ( fig . 1b ) , the remaining @xmath15 galaxies were classified into three bins by considering the bin size of @xmath16 . this resulted in three bins with number of galaxies roughly in the ratio of @xmath17 in the largest , medium and the smallest bins respectively . in the binning process , galaxies that have very low and high blue - shift values were also removed . since our galaxies are blue - shifted , their apparent magnitude increases with time . in order to check the effect of blue - shift on preferred alignments , we have chosen two extreme filters : infrared ( @xmath18 ) and ultraviolet ( @xmath19 ) . the wavelengths of sdss @xmath18 and @xmath19 filters are 7625 @xmath20 and 3543 @xmath20 , respectively . the true magnitudes of @xmath18 filter lies in the far - infrared and @xmath19 in the visual bands . the study of far - infrared and optical activity in the galaxy gives information regarding the early star formation activity and the hii region , respectively . 1c , d shows the magnitude distribution of near infrared and ultraviolet galaxies . for both , gaussian distribution fits well with the observed distribution . in order to find angular momentum vectors ( or spin vectors , sv hereafter ) , the diameters , position angles and positions of galaxies should be known . we have compiled the database of diameters and position angle of galaxies using sdss survey . we follow the method suggested by flin & godlowski ( 1986 ) to convert two dimensional parameters ( positions , diameters , position angles of the dr7 sdss blue shifted galaxies ) into three dimensional parameters ( galaxy rotation axes in spherical polar coordinates ) . the expected isotropic distribution for angular momentum vectors or svs of galaxies are determined by using the method proposed by aryal & saurer ( 2000 ) . the observed and expected distributions are compared with the help of three statistical tests namely chi - square , auto - correlation and the fourier . ( polar angle between the sv of the galaxy and the reference plane ) and @xmath21 ( azimuthal angle between the projection of sv and the x - axis of the reference plane ) . the galactic longitude ( @xmath22 ) and latitude ( @xmath23 ) are shown . for details : flin & godlowski ( 1986 ) and aryal et al . ( 2008).,height=245 ] the two - dimensional sdss parameters ( positions , position angles and diameters ) are converted into polar ( @xmath24 ) and azimuthal ( @xmath21 ) angles of galaxies using flin & godlowski ( 1986 ) . our blue - shifted galaxies have negative radial velocities probably due to their larger peculiar velocity . these galaxies are mostly nearby galaxies . thus , it is convenient to use galactic coordinate system as a physical reference plane . the formulae to obtain @xmath24 and @xmath21 in are as follows ( flin & godlowski , 1986 ) : @xmath25 @xmath26 where @xmath22 , @xmath23 and @xmath27 are the galactic longitude , latitude and position angle , respectively . the @xmath18 represents the inclination angle , obtained using holmberg s ( 1946 ) formula : cos@xmath28 = [ @xmath29@xmath30@xmath31/(1@xmath30@xmath32 ) where @xmath33 is the measured axial ratio and @xmath30 is the intrinsic flatness of disk galaxies . the method of determination of intrinsic flatness of galaxies is the same as in aryal et al . ( 2013 ) . the above formulae show that there are two possible solutions for a given galaxy . the normals ( @xmath34 , @xmath35 , @xmath36 , @xmath37 ) shown in fig . 2 can not be determined unambiguously , because we do not know the side of the galaxy which is nearer / far to us , and the direction of rotation . thus , there are four solutions of the sv orientation for a galaxy . we count all four possibilities independently in our analysis . aryal & saurer ( 2000 ) studied the effects of various types of selections in the database and concluded that such selections may cause severe changes in the shapes of the expected isotropic distribution curves in the galaxy orientation study . their method has been applied by several authors in galaxy orientation studies ( hu et al . 2006 , aryal et al . 2013 and the references therein ) . two kinds of selection effects are noticed in our database : ( 1 ) inhomogeneous distribution of positions of galaxies , and ( 3 ) less number of high inclination ( edge - on ) galaxies . these selection effects are removed and the expected isotropic distribution curves ( @xmath24 and @xmath21 ) are determined using the numerical simulation method as proposed by aryal & saurer ( 2000 ) . for this , a true spatial distribution of s of galaxies is assumed to be isotropic . then , due to the projection effects , @xmath18 can be distributed @xmath13 @xmath38 , latitude can be distributed @xmath13 @xmath39 , the variables longitude ( @xmath22 ) and pa can be distributed randomly , and formulae ( 1 ) and ( 2 ) can be used to simulate ( numerically ) the corresponding distribution of @xmath24 and @xmath21 . the isotropic distribution curves are based on simulations including 10@xmath40 virtual galaxies . the simulation procedure is described in aryal & saurer ( 2004 ) . we perform numerical simulation with respect to galactic coordinate system systems . these expected isotropic distribution curves are compared with the observed distribution . our observed @xmath24 and @xmath21-distributions are compared with expected isotropic distribution curves . for this comparison we applied chi - square , autocorrelation and the fourier ( godlowski 1993 ) tests . these tests are described in the appendix of aryal et al . these statistical tests are a proper method in our case , because @xmath24 and @xmath21 are independent data . the significance level is chosen to be 95% , the null hypothesis is established to be an equidistribution for the @xmath24 and @xmath21 . we have classified our database into six subsamples on the basis of their redshift values and @xmath19@xmath0&@xmath0@xmath18-magnitudes . here we discuss the distribution of the polar @xmath41 and azimuthal @xmath42 angles of galaxy rotation axes in each subsamples . we study the spatial orientation of svs of galaxies with respect to the galactic coordinate system . any deviation from the expected isotropic distribution will be tested using four statistical parameters , namely the chi - square probability ( @xmath43p@xmath44 ) , auto - correlation coefficient ( @xmath43c / c@xmath45 ) , first order fourier coefficient ( @xmath46 ) , and first order fourier probability ( @xmath43p@xmath47 ) . the conditions for anisotropy are the following : p@xmath48 @xmath49 0.050 , c / c(@xmath50 ) and @xmath51/@xmath50(@xmath51 ) @xmath52 1 , and p(@xmath53 ) @xmath49 0.150 . these statistical limits were proposed by godlowski ( 1993 , 1994 ) . in the @xmath24-distribution , a positive ( negative ) @xmath51 suggests that the svs of galaxies tend to orient parallel ( perpendicular ) with respect to plane of the milky way . in the @xmath21-distribution , a positive ( negative ) @xmath51 suggests that the sv projections of galaxies tend to point radially ( tangentially ) with respect to center of the milky way . any ` humps ' or ` dips ' in the histogram will be discussed . here ` hump ' and ` dip ' are defined as having more or less observed solutions than expected respectively . [ cols="^,^,^,^,^,^,^,^",options="header " , ] all three statistical tests suggest isotropy in the @xmath24 distribution of subsample @xmath1801 ( table 1 ) . the galaxies in this subsample are nearby blue - shifted ( rv @xmath49 @xmath128.5 km s@xmath2 ) having @xmath18-magnitude in the range 12.45 and 32.08 . in the histogram , no significant humps or dips can be seen ( fig . thus , a random orientation of angular momentum vectors of galaxies is found , suggesting hierarchy model of the structure formation as suggested by peebles ( 1969 ) . ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the cosine distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] the galaxies in the subsample @xmath1802 are moderately blue shifted ( rv : 28.5 - 58.5 km s@xmath2 and m@xmath55 : 10.46 - 36.10 ) . the chi - square and the correlation tests show anisotropy ( table 1 ) . however , the first order fourier probability and the first order fourier coefficient suggest isotropy . since the observed distribution follow the expected , we regard the fourier test as more reliable . there are small dips at 5@xmath56 , @xmath57 65@xmath56 , and a hump at 15@xmath56 ( fig.3b ) . these are possibly due to the binning effect . strong blue - shifted ( rv : @xmath5858.5 km s@xmath2 ) sdss galaxies ( m@xmath55 : 10.82 - 42.23 ) are grouped in the subsample @xmath1803 . in the histogram , a very good agreement between the observed and expected distributions can be seen . similar to the subsample @xmath1803 , chi - square probability and correlation coefficient show anisotropy , whereas first - order fourier probability and the first - order fourier coefficient suggests isotropy ( table 1 ) . several humps ( 0@xmath56 , 45@xmath56 , 60@xmath56 ) and dips ( 30@xmath56 , 35@xmath56 , 60@xmath56 ) suggests the local effect . thus , the strong blue - shifted infrared galaxies are in the process of grouping or subclustering because of the tidal or gravitational shearing effect . all three statistical tests support isotropy in @xmath24 distribution of @xmath19-magnitude low blue - shifted ( rv @xmath49 58.5 km s@xmath2 ) galaxies . in fig . 3d , no significant humps or dips are found . a small hump at 20@xmath56 and a dip at 60@xmath56 are because of the binning effects . these effects do not change the statistics of the subsample @xmath1901 . hence , no preferred alignments of svs of galaxies in subsample @xmath1901 is noticed . in subsample @xmath1902 , all three statistics show isotropy in polar angle ( @xmath24 ) distribution . a small dip at 10@xmath56 is because of the binning effect ( fig . a random orientation of svs of galaxies is noticed in this subsample . all statistics except correlation coefficient showed isotropy in the polar angle distribution of high blue - shifted ( rv @xmath58 58.5 km s@xmath2 ) @xmath19-magnitude galaxies ( subsample @xmath1903 ) . the correlation coefficient is found to be 1.8 ( @xmath521.5@xmath50 level ) . this is because of the humps at 10@xmath56 , 45@xmath56 , 55@xmath56 and dips at 15@xmath56 , 20@xmath56 . these humps and dips suggests the possibility of grouping or subclustering because of the tidal or gravitational shearing effects between comoving galaxies . to sum up , the spatial orientation of blue - shifted galaxies is found to be random , supporting hierarchy model ( peebles 1969 ) of structure formation . it is found that the rapidly moving blue shifted galaxies are in the process of large structure ( galaxy groups , subclusters , clusters , etc ) formation . in addition , magnitude is found to be independent of the preferred alignments . all three statistics support isotropy in the azimuthal angle distribution of low blue - shifted high magnitude infrared ( subsample @xmath1801 ) galaxies ( table 2 ) . there is a hump at @xmath5965@xmath56 and dips at @xmath15@xmath56 and @xmath155@xmath56 ( fig.4a ) . these are due to the binning effect and hence do not change the statistics . thus , no preference is noticed for spin vector projections of low blue - shifted infrared galaxies . all statistics except @xmath60-test suggest anisotropy in the @xmath21 distribution of the subsample @xmath1802 . in the histograms for the @xmath21-distribution of the subsample @xmath1802 ( fig.4b ) , humps at @xmath160@xmath56 , @xmath5930@xmath56 and @xmath5990@xmath56 , and dips at @xmath150@xmath56 to @xmath5920@xmath56 can be seen . these humps and dips lead this subsample to show anisotropy . the spin vector projection of @xmath1802 galaxies is found to be directed tangentially outwards with respect to the center of the milky way . ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=160 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=160 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] ) angle distributions of blue shifted sdss galaxies in 6 subsamples . the solid circles with @xmath54 1@xmath50 error bars represent the observed distribution . the solid line represents the expected isotropic distributions . the average distributions ( dashed ) are shown for the comparison.,title="fig:",height=162 ] fig . 4c shows the distribution of the spin vector projections of high blue - shifted galaxies ( subsample @xmath1803 ) . similar to the subsample @xmath1802 , all statistical parameters except p@xmath48 show anisotropy . there is a significant hump at @xmath5990@xmath56 and dips at @xmath110@xmath56 and 0@xmath56 . the sv projections of high blue - shifted galaxies is found to be oriented tangentially with respect to the center of the milky way . all statistics except c / c(@xmath50 ) show anisotropy in the @xmath21-distribution for subsample @xmath1901 . in the histogram , several humps and dips can be seen : humps at @xmath190@xmath56 and @xmath5980@xmath56 , and dips in the region from @xmath150@xmath56 to @xmath5950@xmath56 ( fig . this subsample showed isotropy in the polar and anisotropy in the azimuthal angle distribution . this inconsistency is because of the inappropriate reference coordinate system . a physical reference system should be identified in the future . fourier test suggests isotropy in the subsample @xmath1902 . there are dips at @xmath140@xmath56 , @xmath5920@xmath56 and a hump at @xmath5990@xmath56 ( fig.4e ) . we observe weak anisotropy in the spin vector projection of @xmath1902 galaxies with respect to galactic coordinate system . interestingly , all statistics show isotropy in the distribution of sv projections of high blue - shifted galaxies ( @xmath1903 ) . there are small dips at 70@xmath56 and 40@xmath56 , not significant to alter the statistics of the subsample . there is a very good agreement between the observed and expected distributions ( fig . 4f ) thus , a random orientation of spin vector projections is found . in the numerous past literatures authors have studied the spatial orientation of red - shifted galaxies in the field , clusters and superclusters and found mixed result : ( 1 ) noticed anisotropy supporting either ` pancake model ' ( flin & godlowski 1986 ; godlowski 1993 , 1994 ; godlowski , baier & macgillivray 1998 , flin 2001 ) as suggested by doroshkevich ( 1973 ) or ` primordial vorticity model ' ( baier , godlowski & macgillivray 2003 ) as proposed by ozernoy ( 1978)(2 ) found isotropy suggesting a random orientation of spin vectors of galaxies supporting ` hierarchy model ' ( bukhari & cram 2003 ; aryal & saurer 2005a , aryal et al . 2006 , 2012 , 2013 ) as recommended by peebles ( 1969 ) . in addition to these scenarios a bimodal tendency ( kashikawa & okamura 1992 ) , local anisotropy ( flin 1995 , djorgovski 1983 , aryal & saurer 2004 , 2005b ) , global anisotropy ( parnovsky , karachentsev & karachentseva 1994 ) are noticed . godlowski & ostrowski ( 1999 ) noticed a strong systematic effect , generated by the process of deprojection of a galactic axis from its optical image . in isolated abell clusters , only brightest galaxies are preferentially aligned ( trevese , cirimele & flin 1992 ) . panko et al . ( 2013 ) , godlowski ( 2012 ) and godlowski et al . ( 2010 ) noticed a dependence of alignment with respect to cluster richness . the anisotropy is found mostly for those samples which is taken from a limited region of the sky ( e.g. , clusters , sub - clusters , groups , etc ) . for the field galaxies and superclusters , i.e. , the database taken from the large scale structure , authors notice a random orientation supporting hierarchy model . in both the cases ( isotropy or anisotropy ) , a local effect is noticed . this effect might arises due to the tidal connections between the neighboring galaxies or because of the gravitational shearing effect . interestingly , blue - shifted galaxies support hierarchy model . thus , it can be concluded that the blue - shift is not different cosmological effect , it is because of the peculiar velocity as suggested by aaronson et al . we studied the spatial orientation of spin vectors of 5@xmath0987 blue - shifted galaxies ( @xmath187.6 to @xmath10.3 km@xmath0s@xmath2 ) observed through @xmath18- and @xmath19-filter . these database were taken by sdss ( 7@xmath61 data release ) . we classified our data into 6 subsamples based upon their blue - shift . we have used the ` pa - inclination ' method proposed by flin & godlowski ( 1986 ) to convert two dimensional observed parameters to three dimensional galaxy rotation axes ( polar and azimuthal angles ) and carried out random simulation by creating 5@xmath310@xmath4 virtual galaxies in order to remove selection effects from the database ( aryal & saurer 2000 ) . to check for anisotropy or isotropy , we have carried out three statistical tests : chi - square , auto - correlation and the fourier . since our observed distributions do not vary significantly from the expected distribution , we have regarded the fourier test as more reliable . the local effects are examined by the correlation test . we have taken @xmath60 test in order to check the binning effect . in general , we found isotropic distribution of the spin vector of galaxies in all six subsamples with respect to the galactic coordinate system . there are humps and dips in the polar angle distributions , and these humps and dips alter the correlation test showing anisotropy in subsamples @xmath1802 , @xmath1803 , and @xmath1903 . hence , local effect was observed in these subsamples suggesting a local tidal connection between the rotation axes of neighboring galaxies or the gravitational shearing effect . however , in the rest of the subsamples ( @xmath1801 , @xmath1901 , and @xmath1902 ) small humps and dips do not alter the statistics of the total subsample , so we suppose these as binning effects . in general , we observe that there is no preferred alignment of spin vectors of galaxies . our results of @xmath24-distribution support the hierarchical clustering scenario ( peebles 1969 ) , which predicts the random orientation of the directions of the spin vectors of galaxies . on the other hand , we found interesting results for the @xmath21-distribution . as mentioned above , we have regarded the fourier test as more reliable . out of six subsamples only two subsamples ( @xmath1801 and @xmath1903 ) show isotropy in all tests . rest of the subsamples ( @xmath1802 , @xmath1803 , @xmath1901 , and @xmath1902 ) show anisotropic distribution of the spin vector projections of galaxies . also , all subsamples ( except @xmath1903 ) have negative value of the first order fourier coefficient @xmath51 . therefore , in general the spin vector projection of the galaxies tend to be orientated perpendicular to the equatorial plane . however , this result needs to be examined by using other reference system such as supergalactic coordinate system in the future . this research has made use of the sloan digital sky survey ( sdss ) database ( www.sdss.org ) which is funded by alfred p. sloan foundation and the participating institutions . we acknowledge dr . rajendra parajuli , amrit campus , kathmandu for his fruitful comments on the manuscript . we thank mr . r. k. bachchan for helping in the data compilation and mr . s. dangi for the data reduction . one of the authors ( sny ) acknowledges central department of physics , tribhuvan university , nepal for all kinds of support for his ph.d . aaronson , m. , huchra j. , mould , j. , schechter p. l. , tully r. b. : _ astrophys . j. _ * 258 * , 64 ( 1982 ) bian , w. , yuan , q. , zhao , y. , _ monthly notices royal astron . soc . _ * 364 * 187 ( 2005 ) aryal , b. , bhattarai , h. , dhakal , s. , rajbahak , c. , saurer , w. : _ monthly notices royal astron . soc . _ * 434 * , 1939 ( 2013 ) aryal , b. , paudel , r. r. ; saurer , w. : _ astrophys . space sci . _ * 337 * , 313 ( 2012 ) aryal , b. , kafle , p. , saurer , w. : _ monthly notices royal astron . soc . _ * 389 * , 741 ( 2008 ) aryal , b. , paudel , s. , saurer , w. : _ monthly notices royal astron . soc . _ * 379 * , 1011 ( 2007 ) aryal , b. , saurer , w. : _ monthly notices royal astron . soc . _ * 366 * , 438 ( 2006 ) aryal , b. , saurer , w. : _ astronom . astrophys . _ * 432 * , 841 ( 2005b ) aryal , b. , saurer , w. : _ astronom . _ * 432 * , 431 ( 2005a ) aryal , b. , saurer , w. : _ astronom . astrophys . _ * 425 * , 871 ( 2004 ) aryal , b. , saurer , w. : _ astronom . lett . _ * 364 * , l97 ( 2000 ) baier , f.w . , godlowski , w. , macgillivray , h.t . : _ astroph . space sci . _ * 403 * , 847 ( 2003 ) burbidge , e. m. , demoulin , m. h. : _ astrophys . j. _ * 157 * , 155 ( 1969 ) bukhari , f.a . , cram , l.e . : _ astrophys . space sci . _ * 283 * , 173 ( 2003 ) doroshkevich , a.g . : _ astrophys . j. _ * 14 * , l11 ( 1973 ) flin , p. : _ comments astrophys . _ overseas publishers association , amsterdam b.v . * 18 * , no . 2 , p.81 ( 1995 ) flin , p. , godlowski , w. : _ monthly notices royal astron . soc . _ * 222 * , 525 ( 1986 ) flin , p. : _ monthly notices royal astron . soc . _ * 325 * , 49 ( 2001 ) godlowski , w. : _ astrophys . j. _ * 747 * , 1 , 7 ( 2012 ) godlowski , w. , piwowarska , p. , panko , e. , flin , p. : _ astrophys . j. _ * 723 * , 985 ( 2010 ) godlowski , w. , ostrowski , m. : _ monthly notices royal astron . _ , * 303 * , 50 ( 1999 ) godlowski , w. , baier , f.w . , macgillivray , h.t : _ astronom . _ * 339 * , 709 ( 1998 ) godlowski , w. : _ monthly notices royal astron . soc . _ * 271 * , 19 ( 1994 ) godlowski , w. : _ monthly notices royal astron . * 265 * , 874 ( 1993 ) holmberg , e. : _ medd . _ , ser vi , no . 117 ( 1946 ) kashikawa , n. , okamura , s. : _ pasj _ , * 44 * , 493 ( 1992 ) ozernoy , l.m . : in : longair m.s . , einasto j. , eds , proc . 79 , _ the large scale structure of the universe . _ reidel , dordrecht , p.427 ( 1978 ) panko , e. , piwowarska , p. , godlowska , j. , godlowski , w. , flin , p. : _ astrophysics _ * 56 * , 3 , 322 ( 2013 ) parnovsky , s.l . , karachentsev , i.d . , karachentseva , v.e . : _ monthly notices royal astron . soc . _ * 268 * , 665 ( 1994 ) peebles , p.j.e . : _ astrophys . j. _ * 155 * , 393 ( 1969 ) trevese , d. , cirimele , g. , flin , p. : _ astronom . j. _ * 104 * , 935 ( 1992 )	we present the analysis of the spin vector orientation of 5@xmath0987 sdss galaxies having negative redshift from @xmath187.6 to @xmath10.3 km@xmath0s@xmath2 . two dimensional observed parameters are used to compute three dimensional galaxy rotation axes by applying ` position angle inclination ' method . we aim to examine the non - random effects in the spatial orientation of blue - shifted galaxies . we generate 5@xmath310@xmath4 virtual galaxies to find expected isotropic distributions by performing numerical simulations . we have written matlab program to facilitate the simulation process and eliminate the manual errors in the process . chi - square , auto - correlation , and the fourier tests are used to examine non - random effects in the polar and azimuthal angle distributions of the galaxy rotation axes . in general , blue - shifted galaxies show no preferred alignments of galaxy rotation axes . our results support hierarchy model , which suggests a random orientation of angular momentum vectors of galaxies . however , local effects are noted suggesting gravitational tidal interaction between neighboring galaxies . 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
cytoskeletal molecular motors are a large class of proteins that generate movement and force in biological cells by cycling between states bound and unbound from a cytoskeletal filament @xcite . in general , they can be classified as processive or non - processive motors . processive motors like kinesin , dynein or myosin v have a duty ratio ( fraction of time of the motor cycle spent on the filament ) close to unity and therefore are particularly suited for persistent transport of cellular cargo , such as vesicles , small organelles or viruses . using small groups of processive motors increases the walk length and the efficiency of transport compared to the single motor @xcite . a theoretical treatment with a one - step master equation showed that the effective unbinding rate decreases exponentially with the size of the motor ensemble @xcite . moreover groups of motors can also work against larger load than the single motor @xcite . if motors of different directionality on the substrate are attached to the same cargo , bidirectional movement can ensue @xcite , as often observed in cargo transport . a similar tug - of - war setup has been used earlier to explain mitotic spindle oscillations @xcite . non - processive motors such as have a duty ratio significantly smaller than unity . therefore , non - processive motors have to operate in groups in order to generate appreciable levels of force . similar to processive motors , the duty ratio of a group of non - processive motors increases with the size of the group and can become large enough that the group effectively behaves like a processive motor . this is certainly true for the sarcomeres in skeletal muscle , where typically hundreds of work together as one group . combining structural investigations of skeletal muscle with modeling has led to the swinging cross - bridge model for single @xcite . a statistical treatment then has allowed to accurately model the dynamics of the motor ensemble in muscle sarcomeres @xcite . groups of motors also play a crucial role for the mechanics and adhesion of non - muscle tissue cells . cytoskeletal assembles into bipolar minifilaments consisting of @xmath1-@xmath2 motors @xcite . they interact with an actin cytoskeleton which is much less ordered than in muscle , mainly in the actin cortex as well as in the contractile actin networks and bundles associated with cell adhesion and migration @xcite . recently it has been shown that the activity of minifilaments contributes to the sorting of actin filament orientation because of the asymmetric elasticity of actin filaments @xcite . the based forces generated in the actin cytoskeleton are transmitted to the extracellular environment via adhesion sites , which have been shown to harbor different mechanosensitive processes @xcite . in particular , mature focal adhesions are often connected to actin stress fibers consisting of parallel bundles of actin filaments with alternating polarity enabling minifilaments to contract the bundles and thus mechanically load the adhesion sites . to apply these forces effectively , the extracellular matrix underlying the adhesion sites must not be too soft . therefore , cells are sensitive to the elasticity of the substrate and adhere preferentially to stiffer substrates @xcite . if the environment is much stiffer than the cell , it essentially deforms itself and becomes insensitive to the environmental stiffness @xcite . therefore cellular stiffness sets the scale for the sensitivity of rigidity sensing @xcite . due to the complex interplay of many components in a cell , it is difficult to identify the exact contribution of to the rigidity response of cells . one promising experimental route is the reconstruction of _ in vitro _ systems of motors and filaments @xcite , which in the future might allow us to probe these relations in more quantitative detail . with the focus on the description of large assemblies of motors in the muscle sarcomere , theoretical progress has been made mainly through mean - field models @xcite or computer simulations @xcite . for ensembles consisting of a large number of motors , details about internal motor states are less important and experimentally accessible . instead , collective quantities such as velocity , walk length and number of bound motor are of large interest . for example , generic two - state ratchet models have been used to study the behavior of mechanically coupled motors @xcite . here we aim at understanding minifilaments with few molecules for which molecular details and stochastic effects are expected to be more important . in this context , cross - bridge models are appropriate , which have been studied before mainly with computer simulations @xcite . however , this approach is numerically costly , in particular for extensions to systems with multiple minifilaments . recently the parallel cluster model ( ) based on the cross - bridge cycle has been introduced as an efficient yet detailed model for stochastic effects in small ensembles @xcite . in this manuscript , we extend the to minifilaments by modeling them as two ensembles of motors working against each other by walking along two actin tracks with opposing polarity . this situation can be considered as a tug - of - war of the two ensembles of non - processive motors , in analogy to a tug - of - war of processive motors @xcite . in contrast to those studies , however , we do not use a phenomenological force - velocity relation , but rather a cross - bridge model to explicitly include the molecular details of the motor cycle of . in particular , we account for the catch bond character of unbinding ( dissociation rate decreases under load , in contrast to the classical case of a slip bond ) and for the detailed kinetics of the power stroke . from our model definition , it becomes clear that the mechanical situation in bipolar minifilaments is very complex , with an effective spring constant that depends on internal mechanics , external mechanics and the exact state of the motor ensembles . our main result is that minifilaments show a kind of mechanosensitivity that is reminiscent of the way cells respond to environmental stiffness . we show that this effect not only results from the molecular catch - bonding property , but also from the inability to perform the power stroke in a stiff environment with sufficiently large force . we also find that catch - bonding of on stiff substrates leads to frequent switches of direction of the ensemble movement and therefore to an effective diffusion constant which decreases with increasing ensemble size , in marked contrast to a tug - of - war of processive motors with slip bonds . in the parallel cluster model ( ) , individual motors are described by a cross - bridge model with three discrete states and stochastic transitions between them @xcite . here we generalize this model for minifilaments and discuss it with parameter values originally introduced for modeling skeletal muscle @xcite . for cytoskeletal , these values depend on the exact isoform one is considering . the parameter values used here result in a duty ratio of @xmath3 , which lies in the range of duty ratios reported for cytoskeletal b @xcite . as shown schematically in ( a ) , a motor comprises three mechanical elements . the motor head binds to the substrate and contains the hydrolysis site , which binds or its hydrolysis products and . the rigid lever arm is hinged to the motor head and alternates between stretched and primed conformation , thus amplifying conformational changes in the motor head . the linear elastic neck linker with spring constant @xmath4 anchors the lever arm to the rigid motor filament . in the unbound ( ) state , the motor head is loaded with and and the lever arm is primed . the motor reversibly transitions to the weakly - bound ( ) state with on - rate @xmath5 and off - rate @xmath6 . with release of , the lever arm swings to the stretched conformation and the motor enters the post - power - stroke ( ) state . the power stroke is reversible with forward rate @xmath7 and reverse rate @xmath8 but is driven towards the state by the free energy bias @xmath9 . this energy is stored in the primed conformation of the lever arm as part of the energy released in hydrolysis . replacing by , unbinding and hydrolysis brings from the to the state , thus completing the motor cycle . these events are subsumed in a single reaction with unloaded rate @xmath10 which is assumed to be irreversible due of the hydrolysis of . dynamics is characterized by the load dependence of power stroke and unbinding from state : the power stroke moves the lever arm by the power - stroke length @xmath11 ; unbinding from the state requires an additional movement of the lever arm by a distance @xmath12 . thus , both reactions become slower under load . the reduced rate of unbinding under load makes the state of a catch bond rather than a slip bond in the range of forces considered here . motors in the ensemble moving to the right and @xmath13 motors in the ensemble moving to the left . the state of the ensemble is described by the number @xmath14 of bound motors and the number @xmath15 of motors in the post - power - stroke state . the configuration in ( b ) corresponds to @xmath16 in the ensemble and @xmath17 in the ensemble . the displacement of the ensembles on the actin filaments is denoted by @xmath18 and @xmath19 . bound motors within each ensemble are arranged in parallel ; the two ensembles are arranged in series . ( c ) the parallel cluster model ( ) applied to a bipolar minifilament treats the motor ensembles as two adhesions clusters of parallel bonds coupled in series with external springs . all bound motors in equivalent mechano - chemical states have the same strain . the strain of weakly - bound ( ) motors is denoted by @xmath20 and @xmath21 . the power stroke shortens the rest length of the neck linker by the power - stroke length @xmath22 so that post - power - stroke ( ) motors have the strain @xmath23.[fig : setup ] ] the arrangement of motors in a bipolar minifilament is depicted schematically in ( b ) . the minifilament consists of two ensembles of motors working in opposite direction . the motors are anchored to the rigid motor filament joining the two ensembles . within each ensemble , motors are arranged in parallel whereas the two ensembles operate in series . the total number of motors in the ensemble working in direction ( towards the right in ( b ) ) is denoted by @xmath24 ; the number of motors in the ensemble working in direction ( towards the left in ( b ) ) is denoted by @xmath25 . the ensembles move on actin filaments of opposite polarity which are attached to linear elastic elements with spring constant @xmath26 and @xmath27 to represent the effective elasticity of the environment . the opposite polarity allows the minifilament to slide the actin filaments relative to each other and thereby to stretch the springs . the first approximation of the is to assume that motors in equivalent mechano - chemical states exert equal forces ( equal load sharing ) . this mean - field approximation is justified by the small duty ratio of non - processive motors . the elongation or strain of motors in and ensemble is denoted by @xmath20 and @xmath21 , respectively , so that motors exert the force @xmath28 . the strain is positive , when the neck linker is stretched against the moving direction of the ensemble ( inwards in ( b ) ) . with the assumption of equal load sharing , motor ensembles are mechanically equivalent to adhesion clusters of parallel bonds as depicted in ( c ) @xcite . in contrast to the adhesion cluster , however , the rest length of the bond is not fixed , but is reduced by a length @xmath22 due to the power stroke . thus , motors in the state have the strain @xmath23 and exert the force @xmath29 . the strain of as well as motors is determined by the offset @xmath30 between the bound motor head on the substrate and the anchor in the motor filament . the state @xmath31 of a minifilament is described by the numbers @xmath32 and @xmath33 of bound motors and @xmath34 and @xmath35 of motors in both ensembles . the number of motors is @xmath36 and that of motors @xmath37 . with @xmath15 motors and @xmath37 motors , the ensemble exerts the force @xmath38 $ ] . to determine the offset of and motors , we assume that a minifilament is always in mechanical equilibrium . for the arrangement of motor ensembles and external springs in series as in ( c ) this requires that the ensemble forces @xmath39 balance the force @xmath40 exerted by the external springs . the latter is @xmath41 $ ] where @xmath42 is the effective external spring constant and @xmath43 the displacement of the ensemble from its origin on the actin filament . the extension of the external springs is @xmath44 assuming they are relaxed when @xmath45 and all bound motors are in the state with @xmath46 . solving the balance of forces , @xmath47 , yields @xmath30 as function of minifilament state @xmath31 and contraction @xmath48 , @xmath49 the force @xmath50 as function of minifilament state and contraction then reads @xmath51\,,\ ] ] where the total spring constant is defined as @xmath52 and therefore varies dynamically , in contrast to @xmath53 . when all bound motors in the ensemble are in the state ( @xmath54 ) , the offset is positive , @xmath55 , and the neck linkers are stretched against the direction . for growing contraction @xmath56 , the offset @xmath20 increases because @xmath40 increases . a growing number @xmath57 of motors in the ensemble increases @xmath20 further , because transitions to the state shorten the minifilament and increase @xmath40 . on the other hand , @xmath20 decreases and can become negative for a growing number @xmath58 of motors in the ensemble , although @xmath40 increases further . for @xmath59 , motors contribute to the external load which is carried by the motors whose strain is always positive , @xmath60 . the assumption of equal load sharing of equivalent motors defines a four dimensional network of minifilament states @xmath31 . the second approximation of the reduces this network further by assuming that bound states are in local thermal equilibrium ( ) . this is justified by the strong separation of time scales between fast power - stroke and slow binding kinetics @xcite . in , the conditional probability that @xmath61 motors are in the state when @xmath62 motors are bound is the boltzmann distribution @xmath63 where @xmath64 is the partition sum . the energy @xmath65 of a minifilament is the sum of the elastic energy in the external springs , @xmath66 ^ 2 / 2 $ ] , the elastic energy @xmath67 in the neck linkers , where @xmath68 / 2 $ ] , and the free energy bias @xmath69 towards the state . inserting for @xmath30 yields @xmath70 as function of minifilament state @xmath31 and contraction @xmath71 , @xmath72 ^ 2 + \frac{\kl d^2}{2 } \left[\frac{\jp\left(\ip-\jp\right)}{\ip } + \frac{\jm\left(\im-\jm\right)}{\im } \right ] + \left(\jp + \jm\right ) \epp\,.\ ] ] the elastic energy is split into two contributions : the first is due to overall stretching of external springs and neck linkers . for @xmath56 , it increases with increasing contraction and number of bound motors in state . the second contribution is due to internal tension caused by motors in different bound states . it vanishes when all bound motors in an ensemble are either in state or in state , that is , for @xmath73 or @xmath74 , and is positive for intermediate states with @xmath75 . the assumption for the bound states leaves the numbers @xmath76 and @xmath77 of bound motors as the only remaining variables . binding and unbinding changes the state @xmath62 by @xmath78 and @xmath79 . binding proceeds only to the state and with constant on - rate @xmath80 . because @xmath36 motors can bind independently the effective forward rate for the transition @xmath81 in the ensemble is @xmath82 the forward rate is independent of @xmath15 and depends only on @xmath14 in the respective ensemble . unbinding of motors proceeds either with constant off - rate @xmath83 from the state or with load dependent off - rate @xmath84 from the state . thus , the reverse rate for the transition @xmath85 in state @xmath31 is @xmath86 the effective reverse rate for transitions in state @xmath62 is obtained by averaging over @xmath15 with the distribution from , @xmath87 we use a kramers type load dependence for the off - rate from the state , @xmath88 . the off - rate decreases exponentially with increasing load @xmath89 on a motor to describe the catch bond character of the state , where @xmath90 sets the unbinding force scale . inserting effective forward and reverse rate , a two dimensional master equation for the probability @xmath91 that @xmath62 motors are bound can be formulated as @xmath92 { p_{\ip,\im}\xspace}\ , . \end{split}\ ] ] the probability for a specific state @xmath31 is the product of the coarse - grained probability distribution @xmath91 with the conditional probability distribution @xmath93 . the master equation can not be separated in two one - dimensional equations because the reverse rates @xmath94 and @xmath95 depend on @xmath76 and @xmath77 in both ensembles . because the effective reverse rates @xmath96 depend on the contraction @xmath71 , the master equation for binding dynamics has to be solved together with rules for the displacement of the ensembles upon binding and unbinding @xcite . we define the position @xmath43 of an ensemble as the average position of bound motor heads on the substrate . the position of the motor filament is then given by @xmath97 , where @xmath30 is the offset of the motors . new motors are assumed to bind with vanishing offset , thus shifting the ensemble position by @xmath98 . note that the offset @xmath30 is negative for ensembles subjected to small forces and with the majority of motors in the state . to implement the rules for ensemble movement with the offset needs to be averaged with the distribution . this gives @xmath99 in state @xmath62 and the position change @xmath100 upon binding of a motor . unbinding of a motor does not change ensemble position because all bound motors are assumed to be at the same position @xmath43 . complete detachment of one ensemble relaxes the external springs and the still attached ensemble moves freely with offset @xmath101 . reattachment therefore places the ensemble at @xmath102 . because @xmath18 and @xmath19 are defined in opposite directions , the sum @xmath103 describes the contraction of the actin substrates . with these definitions , the for minifilaments is completely defined . the distribution @xmath93 is shaped by the three parts of the minifilament energy @xmath70 in . the second part elastic energy due to internal tension of opposing motors in both ensembles is symmetric against exchanging @xmath15 and @xmath37 . it vanishes when all bound motors are either in state ( @xmath74 ) or in state ( @xmath73 ) . between @xmath73 and @xmath104 ( or @xmath74 to @xmath105 ) the energy increases by @xmath106 so that the relative occupancy is @xmath107 . this implies that intermediate states with @xmath75 are hardly occupied and @xmath93 is close to a binary distribution , in which either none or all of the bound motors in an ensemble perform the power stroke . only the four states @xmath108 , @xmath109 , @xmath110 and @xmath62 , which are local minima of @xmath70 , can be appreciably occupied . the third part of @xmath70 in decreases by the gain @xmath69 of conformational energy for each of the @xmath111 motors . this conformational energy bias is opposed by the first contribution from the elastic energy to @xmath70 in which increases with @xmath15 . this elastic energy bias increases with contraction @xmath71 ( if @xmath112 and @xmath113 ) and total spring constant @xmath114 and eventually exceeds the conformational bias towards the state . ( see ) for a minifilament with @xmath115 bound motors as function of contraction @xmath103 for external spring stiffness @xmath116 , that is , @xmath117 and @xmath118 . the power - stroke probability is shown for @xmath119 , @xmath110 and @xmath120 ; @xmath121 is smaller than @xmath122 . ( b ) effective reverse rates @xmath94 and @xmath95 ( see ) as function of @xmath71 for the minifilament from ( a).[fig : psprob ] ] \(a ) shows the power - stroke probability @xmath93 for a minifilament with @xmath115 bound motors as function of contraction @xmath71 . for small @xmath71 , the gain of conformational energy in the power stroke exceeds the increase of elastic energy and the minifilament is in state @xmath123 with probability @xmath124 . at an intermediate value of @xmath71 , the minifilament switches to @xmath125 in a sharp transition . this means that above this threshold , only the ensemble with larger number of bound motors ( @xmath126 ) is able to perform the power stroke . above a second threshold , neither ensemble performs the power stroke and the minifilament switches to @xmath108 . the plot confirms that the distribution can be almost neglected for intermediate states with @xmath127 and ensembles are said to be either in or state according to the dominant state of the bound motors . the thresholds for the transitions can be determined by comparing the energy in the four states with @xmath73 or @xmath74 . for @xmath128 the transition from @xmath119 to @xmath110 occurs at @xmath129 and from @xmath130 to @xmath120 at @xmath131 . in the symmetric case @xmath132 , the states @xmath133 and @xmath109 are degenerate and the minifilament occupies both with equal probability . it is important to note that the thresholds contain an additional dependence on @xmath62 via @xmath114 . for a given value of @xmath71 , fluctuations of @xmath62 will therefore induce transitions to the state for small @xmath14 and to the state for large @xmath14 . \(b ) plots the effective reverse rates @xmath96 in and ensemble as function of @xmath71 for the same minifilament setup as in ( a ) . for small @xmath71 the minifilament is in state @xmath134 and both rates decrease exponentially with @xmath71 due to catch bonding of motors . note , however , that the relation between @xmath71 and force depends on power - stroke probability : it is @xmath135 for @xmath119 , @xmath136 for @xmath133 and @xmath137 for @xmath108 . at the transition to @xmath138 , the bound motors in the ensemble can no longer perform the power stroke and the effective reverse rate of the ensemble drops to the value @xmath139 determined by the small off - rate @xmath140 from of the state . because the transition to the state decreases the force on the @xmath141 bound motors in the ensemble , the effective reverse rate @xmath142 increases during the transition . with the transition to @xmath108 the bound motors in the ensemble enter the state and the effective reverse rate drops to @xmath143 . the two dimensional master equation describes a non - equilibrium process without detailed balance . moreover , there is a nonlinear feedback between minifilament displacement and binding dynamics so that the master equation can not be solved analytically . instead , we analyze minifilament dynamics numerically using the direct method of the gillespie algorithm . shows typical stochastic trajectories of symmetric minifilaments with varying size and external spring stiffness . the lower panel of each plot shows the number of bound motors @xmath62 in the two ensembles . the upper panel plots the minifilament force @xmath50 from for a given @xmath62 weighted with the appropriate conditional probability over @xmath61 . due to the binary nature of the power - stroke probability , @xmath61 is dominated by the states with @xmath144 or @xmath74 . ( a ) shows a trajectory of a small minifilament with @xmath145 motors for @xmath146 , that is , @xmath147 . due to the soft external springs , the power stroke in both ensembles does not increase the load on the motors appreciably so that the effective reverse rate remains close to its large intrinsic value . moreover , the threshold value of @xmath71 for the transition from to state is large and at least one of the ensembles typically detaches before reaching the threshold . therefore , trajectories are characterized by frequent detachment of ensembles . and @xmath148 ) with varying ensemble size and varying external spring stiffnesses . in each plot , the bottom panel shows the number of bound motors in ensemble ( @xmath76 , red ) and ensemble ( @xmath77 , blue ) . the top panel shows the force on the ensembles , that is , @xmath50 ( see ) weighted with @xmath93 . ( a ) @xmath145 and @xmath149 ( @xmath147 ) , ( b ) @xmath145 and @xmath150 ( @xmath151 ) and ( c ) @xmath152 and @xmath153 ( @xmath154).[fig : trajectories ] ] \(b ) shows a trajectory of a minifilament with the same size as in ( a ) but for stiffer external springs with @xmath155 and @xmath151 . detachment is much less frequent than in ( a ) and the series of short force peaks is no longer observed . instead , initial attachment is followed by a gradual increase of force towards a state with strongly fluctuating but on average constant force . this is the result of two effects . first , the force generated by the power stroke at @xmath156 decreases the off - rate of motors appreciably so that the time to detachment of the minifilament is increased . second , the threshold for the transition from to state is lowered and in combination with smaller off - rate is more likely to be reached before detachment . to stabilize attachment , it is sufficient that the power stroke can not be performed for @xmath157 so that the last motor in each ensemble unbinds slowly from the state . for @xmath157 the transition from @xmath133 to @xmath120 occurs at a force @xmath158 below and @xmath159 above the transition . the fast fluctuations of force upon binding and unbinding in ( b ) thus indicate transitions between and state , while the slower variations are due to variations of @xmath71 following multiple binding events . fluctuations to small values of @xmath71 allow both ensembles to perform the power stroke and the increased reverse rate can lead to detachment of the minifilament . \(c ) shows a trajectory of a larger minifilament with @xmath152 motors for soft external springs @xmath154 as in ( a ) . initially , the miniflament detaches repeatedly as observed for the smaller ensemble . due to its larger size and detachment time , the minifilament eventually reaches the threshold @xmath160 above which the last motors in both ensembles unbind from the state . although the minifilament is stalled for @xmath161 , it continues to builds up larger force because the typical number of bound motors is larger . for forces above @xmath162 , the trajectory of the number of bound motors shows that it becomes unlikely to find a single bound motor . this force is large enough to keep the last two bound motors in the state with low unbinding rate and to make detachment of any of the two ensembles unlikely . the dynamics of force can be understood considering the sequence of to transitions in . for small @xmath71 , both ensembles perform the power stroke and @xmath71 increases quickly through the activity of both ensembles . the increase of @xmath71 is terminated by the detachment of one of the ensembles . if the minifilament remains attached sufficiently long , @xmath71 reaches the threshold above which the ensemble with smaller number of bound motors enters the state . in this case , the ensemble moves forward upon binding of motors as long as @xmath163 . the ensemble , on the other hand , will step backward upon binding because @xmath164 without motors . the contraction @xmath71 continues to increase as long as the forward movement of the ensemble is faster than the backward movement of the ensemble . as @xmath71 reaches the threshold above which both ensembles enter the state , @xmath71 can only decrease because @xmath165 for @xmath166 so that both ensembles step backwards upon binding of motors . since the threshold is reached first for small @xmath14 , detachment of the minifilament becomes very unlikely . on the other hand , for an increasing number of such states the minifilament enters an isometric state in which contraction @xmath71 and force @xmath40 fluctuate with a constant average , because forward movement at large @xmath14 is balanced by backward stepping at small @xmath14 . the trajectories in are for symmetric minifilaments with the same number @xmath167 of motors in and ensemble and equal external spring constants , @xmath148 . differences of @xmath168 do not affect results , because @xmath169 enters the dynamic description only via the effective external spring constant @xmath53 . differences of ensemble sizes , @xmath170 , do affect minifilament dynamics but trajectories are qualitatively similar as long as the difference is not too large . most importantly , asymmetric minifilaments display a net movement in the direction of the larger ensemble . when both ensembles are attached and perform the power stroke , fewer motors are bound on average in the smaller ensemble . these are subject to larger force so that position steps are smaller or even negative ; catch bonding of motors reduces this difference . moreover , the smaller ensemble detaches more frequently allowing the larger ensemble to move freely . finally , the smaller ensemble is more likely to transition from to state and form a passive anchor for the larger ensemble . the frequency of detachment of asymmetric minifilaments and the probability to reach large forces , though , is determined by the smaller ensemble . the stochastic trajectories reveal a switch in minifilament dynamics from transient attachment without sustained force ( see ( a ) ) to near permanent attachment ( see ( b - c ) ) in response to increasing external spring stiffness . to elucidate this mechanosensitivity further , ( a ) plots the mean force @xmath171 obtained by averaging in the steady state generated by symmetric minifilaments with varying ensemble size @xmath167 as function of external spring stiffness @xmath172 . ( a ) reveals a strongly nonlinear increase of @xmath171 with @xmath169 , which is caused by catch bonding of motors in combination with to transitions in the power - stroke probability . generated by symmetric minifilaments with varying number @xmath167 of motors as function of external spring constant @xmath173 . @xmath171 is obtained by time averaging over stochastic trajectories . ( a ) mean force @xmath171 for @xmath174 , @xmath175 , @xmath176 and @xmath177 . ( b ) mean force @xmath171 for @xmath178 for several model variants in comparison with the full model : @xmath179 no state , that is , fixed power - stroke probability @xmath180 , @xmath181 no catch bonding of motors which unbind with constant off - rate @xmath182 , @xmath183 increased off - rate @xmath184 from the state and @xmath185 single ensemble working against an external spring with @xmath24 and @xmath26 as for the minifilament.[fig : meanforce ] ] for small stiffness , @xmath171 increases linearly with @xmath169 . here , the contraction @xmath71 reached before one of the ensembles detaches is on the order of the power - stroke length @xmath22 . for small values of @xmath53 , the corresponding force is too small to increase the time to detachment significantly , so that the typical contraction @xmath71 hardly increases with @xmath53 . in the limit of small @xmath186 , it is @xmath187 so that the force @xmath188 is proportional to @xmath53 . the mean force along a trajectory is proportional to the duty ratio of a minifilament , that is , the fraction of time both ensembles are attached . the duty ratio is hardly affected by small forces but increases with minifilament size so that the slope in the linear regime increases with @xmath189 . catch bonding of motors eventually increases detachment time and duty ratio so that @xmath71 increases beyond the force free case . the mutual positive feedback between force and duty ratio leads to a rapid increase of @xmath171 which is reinforced by transitions from to state . on the other hand , these transitions limit the increase of force , because motors are required for forward movement of the ensembles and an increase of @xmath71 . once minifilaments are almost permanently attached , @xmath171 increases slowly with further increasing @xmath169 . this reflects the increase of the average number of bound motors in the isometric state , which is caused by increased force fluctuations for large @xmath53 and the nonlinear dependence of the off - rate on force . the increasing number of bound motors allows the minifilament to reach larger values of @xmath71 although the to transition occurs for increasing values of @xmath14 . the linear regime is most prominent for small @xmath189 where the intermediate , super linear regime can not be discerned . with increasing @xmath189 , the linear regime shrinks and the super linear growth in the intermediate regime approaches a step increase of @xmath171 , because detachment time depends sensitively on changes of the off - rate for large ensembles @xcite . \(b ) plots the mean force generated by minifilaments with @xmath190 as function of @xmath169 for model variants in which components of the motor cycle are omitted in order to elucidate their contribution to force generation . @xmath179 without the transition from to state ( no state ) , that is , for fixed power - stroke probability @xmath180 , the super linear increase of @xmath171 at intermediate @xmath169 is comparable in steepness to the full model but is shifted to larger values of @xmath169 . without stabilization through the to transition , the increase of duty ratio and mean force is due to catch bonding of motors alone . on the other hand , @xmath171 reaches significantly larger values and continues to increase over the whole range of @xmath169 because @xmath71 is no longer limited by the transition to the state of both ensembles . instead , @xmath171 is limited by the stall force at which the offset @xmath30 in both ensembles vanishes . thus the transition of the power - stroke probability is required to increase the sensitivity of the response and to generate a switch - like behavior with a plateau at large forces . without this transition , the model would become unphysical because then the power stroke would require more energy than provided by the hydrolysis . @xmath181 without catch bonding , that is , with constant off - rate @xmath191 from the state , the super linear increase of @xmath171 at intermediate @xmath169 is present but occurs for larger @xmath169 and is much weaker than in the full model . also the overall level of @xmath171 is strongly reduced . trajectories show that minifilaments do reach the transition from to state , but continue to detach frequently . thus , catch bonding of motors does not only provide the feedback between force and duty ratio needed for the steep increase of @xmath171 at intermediate @xmath169 , but also stabilizes the isometric state by increasing the average number of bound motors . @xmath183 with a large off - rate @xmath192 of motors , transitions from to state increase the effective reverse rates ( see ) . hence , to transitions induce detachment and neutralize the effect of catch bonding . the level of @xmath171 is smaller than in the case without catch bonding and a steep intermediate regime is not observed . thus a low off - rate from the state is required to conserve the large level of force in the isometric state and to generate the observed switch - like response . @xmath185 a single ensemble working against an external spring ( representing one half of a minifilament ) , generates slightly larger @xmath171 than minifilaments because the frequency of detachment is reduced . due to the larger total spring constant , this difference increases as @xmath169 approaches @xmath193 . for very large @xmath169 , on the other hand , the mean force generated by a single ensemble collapses because the power stroke can no longer be performed at @xmath156 . this does not occur for minifilaments because @xmath114 is limited by @xmath193 . thus the interplay between external and internal mechanics is essential for the functioning of minifilaments . the contraction @xmath71 is confined within a narrow range around a stable fixed point for attached minifilaments . fluctuations to large @xmath71 are limited by the transition of both ensembles to the state . fluctuations to small @xmath71 induce transitions to the state in both ensembles so that forward movement of both ensembles increases @xmath71 rapidly . the range of @xmath71 narrows with increasing @xmath53 . to facilitate analysis of this situation , we replace the external springs by rigid anchorage . the total spring constant reduces to @xmath194 and the contraction @xmath71 is identical to the sum of the offset @xmath30 in the two ensembles , @xmath195 . due to the large stiffness of the neck linkers , at most one ensembles can be in state at @xmath156 , while the other is in state . the threshold for the transition to state @xmath196 , at which the minifilament is stalled , is small and reached within few binding steps . as a consequence a stationary state is established quickly and the time dependent solution @xmath91 of the master equation can be replaced by the stationary limit @xmath197 . results of stochastic simulations for @xmath30 reveal two clearly separated peaks at a negative and a positive value of @xmath20 ( data not shown ) . the peak at negative @xmath20 corresponds to the ensemble in state with force @xmath198 per motor and the peak at positive @xmath20 to the ensemble in state with force @xmath199 per motor . the separation of the peaks equals the power - stroke length , @xmath200 , which is the expected difference of @xmath30 for @xmath201 ( see ) . finite width and asymmetry of the observed peaks are due to fluctuation of @xmath14 and @xmath71 for the case @xmath202 . the same results applies to the offset @xmath21 in the ensemble . stochastic simulations confirm that the ensemble with larger number of bound motors usually performs the power stroke while the other ensemble is in state ( see ) . for @xmath203 , the minifilament is in state @xmath133 or @xmath109 with equal probability . therefore , the power - stroke probability from can be replaced by @xmath204 where @xmath205 is the heaviside step function with @xmath206 and @xmath207 the kronecker delta . as a consequence of the mechanical coupling of ensembles and the mechanosensitivity of , the numbers @xmath76 and @xmath77 of bound motors in and ensemble are synchronized . the correlation increases with @xmath189 but is reduced for minifilaments with soft external springs . synchronization is due to the transition to the state of the ensemble with fewer bound motors . the small off - rate of relative to motors ( see ( b ) ) tends to equalize the number of bound motors . soft external springs weaken the mechanical coupling of ensembles so that motors in one ensemble are less sensitive to variations of the number of bound motors in the other . because minifilaments move in the direction of the ensemble , synchronization of the number of bound motors tends to reverse the direction of motion of minifilaments and prevents long , persistent runs . this is different from a tug - of - war of processive motors , which are usually described as slip bonds which favor large differences of the number of bound motors . to derive an approximation for the stationary distribution @xmath208 we replace the continuous distribution of @xmath30 by a discrete one with two @xmath209-peaks . assuming constant contraction @xmath71 during a typical binding and unbinding cycle through states @xmath210 allows to estimate the negative offset in the ensemble as @xmath211 . the offset in the ensemble is @xmath212 . for @xmath213 , excellent agreement of this constant strain approximation with stochastic simulations is observed . the load dependent off - rate of motors becomes independent of @xmath62 . the strain of motors in the leading ensemble is @xmath214 and their off - rate @xmath215 . together with the binary approximation of for the power - stroke probability , the off - rate of motors in the ensemble ( see ) reduces to @xmath216\,.\ ] ] the first term describes unbinding from the state , the second from the state and the third is for vanishing force in the case of a minifilament with detached ensemble . an analogous expression holds for @xmath95 . for symmetric minifilaments with @xmath190 . ( a ) results of stochastic simulations for @xmath208 as function of @xmath76 and @xmath77 . solid and dashed line indicate slices of @xmath208 for constant @xmath217 and @xmath218 . ( b ) comparison of simulation results ( lines ) and analytical results ( symbols ) from the constant offset approximation for @xmath208 as function of @xmath77 for constant @xmath217 and @xmath218.[fig : probability ] ] the constant offset approximation yields a two dimensional network of states with constant transition rates per motor . analytical solutions for the stationary distribution are found by solving the corresponding linear system of equations . ( a ) shows numerical results for @xmath208 from the exact model for a minifilament in a tug - of - war with @xmath190 . the distribution is symmetric with respect to exchanging @xmath76 and @xmath77 and strongly peaked at @xmath219 . it is centered along the diagonal with @xmath220 which expresses the effect of synchronization of the ensembles . ( b ) shows the stationary probability for fixed @xmath76 as function of @xmath77 and compares the numerical solution to the analytical solution obtained via the constant offset approximation . considering the approximations entering the analytical solution the excellent agreement is quite remarkable . of a minifilament as function of ensemble size @xmath167 for symmetric minifilaments . numerical results ( symbols ) are compared to results from the constant strain approximation ( see ) . the inset shows a logarithmic plot of the data for a wider range of @xmath24.[fig : diffusion ] ] the absolute position of the minifilament can be defined as the mean position @xmath221 of and ensemble . symmetric minifilaments perform an unbounded random walk with vanishing mean which is characterized by the diffusion coefficient @xmath222 . due to the limited range of the contraction @xmath71 , the diffusion coefficient of @xmath223 will be identical to that of @xmath18 and @xmath19 in the long time limit , @xmath224 . we calculate @xmath225 via the limit of the second jump moment per time in the limit for a vanishing time step as @xmath226 @xcite . for @xmath227 , binding in the ensemble yields @xmath228 where @xmath229 . detachment of the ensemble releases the strain of both ensembles and yields @xmath230 for @xmath231 . movement of the detached ensemble through stepping of the ensemble contributes by @xmath232 for @xmath233 and @xmath234 . the average diffusion constant of the minifilament position @xmath223 is @xmath235 { p_{\ip,\im}(\infty)\xspace}\,.\ ] ] within the constant offset approximation , the offset is @xmath236 and @xmath237 when both ensembles are attached and @xmath238 if one of the ensembles is detached . using the approximate expression for @xmath208 from , the contributions to the diffusion constant reduce to @xmath239 \quad\text{for}\quad \ip > 0\\ d_{+}^{\rm detach } & = ( \xneg+\xpos)^2 \left[\kzns \theta(\ip-\im ) + \ken \theta(\im-\ip ) \right]\quad\text{for}\quad \ip = 1\\ d_{+}^{\rm drag } & = \gmin(\im ) \frac{d^2}{(\im+1)^2 } \quad\text{for}\quad \ip = 0 , \im > 0\,.\end{aligned}\ ] ] plots the diffusion constant @xmath222 of a minifilament as function of ensemble size @xmath167 . the diffusion coefficient decreases with increasing ensemble size . simulation results show that for large @xmath240 , where detachment of ensembles is negligible , the diffusion coefficient scales as the inverse of ensemble size , @xmath241 . for smaller @xmath189 , ensemble detachment leads to a stronger dependence on @xmath189 . in the approximation of , only @xmath242 contributes to @xmath222 for @xmath240 . for a given @xmath243 , the squared step @xmath244 scales as @xmath245 and the forward rate as @xmath246 . using @xmath247 , this yields the scaling expression @xmath248 which fits well to the observed dependence @xmath249 . the deviations between simulation and analytical results are probably due to correlations in the fluctuations of the offset @xmath250 . the decrease of @xmath222 is in contrast to the dependence of @xmath222 on ensemble size in a tug - of - war of processive motors . processive motors are usually described as slip bonds so that the time to switch between directions increases exponentially with ensembles size , which induces an exponential increase of the diffusion coefficient @xcite . the synchronization of the number of bound motors as a consequence of the catch bond character of leads to a weak dependence of the switching time on ensemble size and prevents the increase of the diffusion coefficient . here we have introduced a new model for the stochastic dynamics of minifilaments . analyzing the stochastic trajectories and the mean force generation predicted by our model revealed a switch of the dynamic behavior of the minifilaments in response to changes of the environmental elasticity as a consequence of two independent mechano - sensitive processes . first catch bonding of motors depends directly on the load on the motors . a gradual increase of load decreases the reverse rate and increases the number of bound motors to stabilize ensemble attachment . second the transition from to state provides another type of catch bonding because unbinding from state is slower than from state , @xmath140 . there are two important differences : @xmath179 to transitions reduce the reverse rates @xmath96 abruptly and @xmath181 the transition is sensitive not only to minifilament load but also to the elasticity of the environment . for single ensembles , to transitions only occur in elastic environments , but not for ensembles working against constant external load @xcite . in a miniflament , the elastic neck linkers provide an elastic environment for the ensembles even with rigid anchorage . therefore our model demonstrates the importance of considering both the internal and external mechanics when investigating the function of a minifilament . both mechano - sensitive processes cooperate in stabilizing minifilament attachment . while permanent attachment can be reached with catch bonding of motors alone , the transition from to state shifts the switch to smaller external stiffness and steepens the transition . on the other hand , the transition to the state alone is not sufficient for stable attachment . the two mechano - sensitive processes have contrasting effects on minifilament dynamics . catch bonding of motors allows to distribute load on a growing number of bound motors . this increases the velocity of an ensemble at a given load and increases the stall force . in the context of sarcomeric and cytoskeletal ensembles working against a constant external load , it was shown that catch bonding of motors leads to the characteristic upward convex force - velocity relation @xcite . for minifilaments , catch bonding of motors increases the mean force exerted in an elastic environment . transitions from to state , on the other hand , limit the increase of force because motors are required for an increase of contraction , so that minifilaments adjust themselves to an isometric state . in single ensembles working against a linear external load , the force generated by the ensembles breaks down for very stiff external load because the power stroke can no longer be performed at @xmath156 . in minifilaments , the effective stiffness is limited by the elasticity of the neck linkers and this breakdown is prevented . therefore , minifilaments in the isometric state generate a mean level of force , which depends mainly on the number of motors available for binding , but is robust to variations of elastic properties of the environment . transitions to the state limit the mean force as long as they occur before the stall force of an ensemble in state is reached . this is the case for the parameters in our model ; for significantly smaller neck linker stiffness as used in ref.@xcite with comparable power - stroke length , the stall force is reduced strongly and could be reached before the transition to the state . for the smaller stall force , however , catch bonding of motors would not be sufficient to decrease the reverse rate sufficiently to achieve stable attachment . the mechanosensitivity of minifilaments has important implications for the structure and function of acto - myosin networks . in the actin cortex or reconstituted mixtures of actin with minifilaments , the apparent stiffness of an actin filament depends largely on the level of crosslinking . due to their mechanosensitivity , minifilaments will most efficiently form crosslinks with and exert force to filaments which are already firmly linked to the network . loose filaments , on the other hand , which do not contribute to overall network tension , could not be integrated by minifilaments . this could help to create densely linked networks in which minifilaments provide active crosslinks maintaining a constant level of tension , which is restored after deformations . stretching a network would initially reinforce attachment and prevent rupture of the network . on longer time scales , the increased tension would be released by the flow of the network as a consequence of backward stepping of minifilaments . compression , on the other hand , would reduce the load on the minifilament crosslinks . subsequent forward movement of ensembles then helps to restore the initial tension in the network . persistent compression , however , could lead to detachment and network disintegration . in an actin network of random polarity , mechanosensitivity of minifilaments in combination with the asymmetric elasticity of actin filaments could lead to a selection of filaments according to their orientation , because large forces can be exerted when filaments are stretched , but small forces on filaments buckling under compression @xcite . this could also contribute to the observed contraction rather than expansion of actin networks of random orientation . unbalanced preferential attachment to strongly linked filaments could cause aggregation of minifilaments and actin , as it is observed in reconstituted assays @xcite . in adherent cells elasticity of the extracellular matrix contributes to the effective stiffness experienced by minifilaments . actin stress fibers tend to form and reinforce at focal contacts on sufficiently stiff substrates @xcite . mechanosensitivity of minifilaments could contribute to discrimination of substrate stiffness because they are not able to crosslink and to build up forces in soft environments . although this possible relation is rather speculative at this stage and the exact role of myosin minifilaments in the rigidity response of cells is hard to pin down , it is worth mentioning that activity is an integral part of mechanosensing , which is completely disrupted by inhibiting activity , e.g. with the pharmacological inhibitor blebbistatin @xcite . according to our model , the dynamics of minifilaments is strongly determined by the existence of at least two load bearing bound states of motors . for small load , the free energy bias allows the motors to perform the power stroke and complete the motor cycle ; for large load , however , the increase of the elastic energy exceeds the free energy bias and the power stroke can no longer be performed . because the free energy bias is limited by the energy gained in hydrolysis , this transition is inevitable , independent of the parameters . the exact value of the transition thresholds , however , depends on the stiffness @xmath193 of the neck linkers . this also means that the value of @xmath193 determines the range of external spring stiffness @xmath169 to which minifilaments are sensitive . the values for @xmath193 used in the literature range over an order of magnitude , @xmath251 @xcite . although the effect of varying @xmath193 is reduced by combining small @xmath193 with large values of the power - stroke length @xmath22 , small values as in ref . @xcite could lead to stalling of minifilaments before the transition of the power - stroke probability is reached . unless , however , the unbinding force scale @xmath252 is decreased as well , this stall force would not be enough to achieve stable attachment . another set of parameters determining the dynamics of minifilaments in our model are the on- and off - rates of motors , which determine the duty ratio . the load - free duty ratio ( @xmath253 ) of with the transition rates used here is comparable to values reported for non - muscle isoform b but significantly larger than for non - muscle isoform a or smooth muscle . changes of the transition rates will not change the outcome of the model qualitatively as long as the off - rate from the state is small compared to the off - rate from the state under load . smaller values of the duty ratio could be compensated by larger ensemble sizes in order to achieve the stabilization of minifilament attachment observed here . we have further shown with an analytical treatment for the case of minifilaments on rigid substrates that the effective diffusion constant decreases with ensemble size , in marked contrast to the case of the tug - of - war of processive motors , where the diffusion coefficient increases exponentially with the number of motors @xcite . processive motors are usually described as slip bonds favoring strongly asymmetric states , in which one of the ensembles is completely detached . with increasing number of motors , a tug - of - war of processive motors generates persistent movement in either direction , with exceedingly rare reversal of the direction of motion . this dependence of the diffusion coefficient on ensemble size is in line with the function of processive motors for transport , which becomes more efficient with increasing number of motors . non - processive motors , on the other hand , serve as active , force generating crosslinks which are not required to move through the cytoskeleton . an experimental realization of a tug - of - war of two ensembles of non - processive motors , e.g. using a solid substrate covered by actin filaments with nematic order but random polarity or bipolar actin constructs on a surface covered with motors , could be used to investigate the characteristic difference between processive , slip - bond motors and non - processive , catch - bond motors . in general , we envision that biomimetic assays provide rewarding avenues to experimentally test our theoretical predictions .	tissue cells are in a state of permanent mechanical tension that is maintained mainly by minifilaments , which are bipolar assemblies of tens of molecular motors contracting actin networks and bundles . here we introduce a stochastic model for minifilaments as two small motor ensembles engaging in a stochastic tug - of - war . each of the two ensembles is described by the parallel cluster model that allows us to use exact stochastic simulations and at the same time to keep important molecular details of the cross - bridge cycle . our simulation and analytical results reveal a strong dependence of minifilament dynamics on environmental stiffness that is reminiscent of the cellular response to substrate stiffness . for small stiffness , minifilaments form transient crosslinks exerting short spikes of force with negligible mean . for large stiffness , minifilaments form near permanent crosslinks exerting a mean force which hardly depends on environmental elasticity . this functional switch arises because dissociation after the power stroke is suppressed by force ( catch bonding ) and because ensembles can no longer perform the power stroke at large forces . symmetric minifilaments perform a random walk with an effective diffusion constant which decreases with increasing ensemble size , as demonstrated for rigid substrates with an analytical treatment . + @xmath0correspondence : [email protected]
the one - dimensional poisson equation , @xmath0 with dirichlet boundary conditions , @xmath1 plays an important role in many branches of science . particularly , the poisson equation is essential in self - consistent calculations in solid state physics @xcite . in general , we have to solve it numerically many times . therefore , is vital to have the fastest and the most accurate numerical scheme to solve it . in this article , we present a very efficient direct method , based on a numerov @xcite sixth order numerical scheme , to solve the poisson equation numerically . because of its efficiency and simplicity , this new method can be used as a canonical numerical scheme to accurately solve the one - dimensional poisson equation . this article is organized as follows . our numerical scheme is presented in section [ sec : numerov ] . its linearization , together with a few discussions , are presented in section [ sec : discus ] . our conclusions are presented in section [ sec : conclus ] . let @xmath2 represents the solution of at the @xmath3-th point , @xmath4 , of an equally spaced net of step @xmath5 and dimension @xmath6 . let also @xmath7 represents the @xmath8-th derivative evaluated at the same point @xmath9 . then we can evaluate the solution @xmath10 at the nearest neighborhood points @xmath11 of @xmath9 using taylor series @xcite , @xmath12 the basic idea in the numerov approach is to eliminate the fourth order derivative in the expression @xmath13 where @xmath14 to obtain the sixth order three - point numerical scheme @xmath15 where we chose @xmath16 and , consequently , @xmath17 . in a similar way , we can eliminate the third order derivative from @xmath18 where @xmath19 to obtain the fifth order three - point numerical scheme @xmath20 for the first derivative of @xmath10 , where we chose @xmath21 and , consequently , @xmath22 . so far , the three - point numerical scheme is an iterative method , i.e. , given two informations , @xmath23 and @xmath24 , we can calculate @xmath25 . one difficulty of this iterative method is related with the dirichlet boundary conditions : they are known only at end - points @xmath26 and @xmath27 . thus , we can not initiate our iterative scheme . fortunately , the recurrence relation in is linear with constant coefficients . these two features imply we can find an unique solution to it , @xmath28 where @xmath29 and @xmath30 must be expressed in terms of @xmath31 ( the dirichlet boundary conditions ) , @xmath32 now we have an analytical sixth order numerical scheme to solve accurately the poisson equation with the dirichlet boundary conditions . it should be mentioned that the analytical third order numerical scheme presented by hu and oconnell @xcite , making use of tridiagonal matrices , can also be derived by the present approach restricted to the third order , @xmath33 where @xmath34 although we have found a very accurate analytical direct method to solve the one - dimensional poisson equation with dirichlet boundary conditions , namely , the sixth order numerov scheme , it has one undesirable feature : its execution time is proportional to the square of the grid dimension . fortunately it can be linearized . first , we create a vector @xmath35 , whose components are the partial sums @xmath36 ( @xmath37 ) . next , we create a second vector @xmath38 with @xmath39 and @xmath40 . we also need a third vector @xmath41 with @xmath42 and a fourth vector @xmath43 with the complete sums @xmath44 . using these new vectors , our sixth order numerov scheme can be rewritten as follows , @xmath45.\ ] ] this numerical scheme has now a linear execution time proportional to five times the grid dimension @xmath6 . let us use a gaussian density , @xmath46 to verify the accuracy and the efficiency of the non - linear numerical scheme , as well as the linear numerical scheme . the solution for the poisson equation , along with the boundary conditions @xmath47 and @xmath48 , is @xmath49 where @xmath50 is the error function , @xmath51 figure [ f1 ] shows the execution time as a function of the grid dimension @xmath6 for three cases . in one case ( the dotted line ) , the numerical solution was computed by the non - linear third order numerical scheme . in the second case ( the dashed line ) , the numerical solution was computed by the non - linear sixth order numerical scheme . in the last case ( the solid line ) , the numerical solution was computed by the linear sixth order numerical scheme . at @xmath52 , the execution time of the non - linear third ( sixth ) order numerical scheme is approximately 145 ( 51 ) times the execution time of the linear sixth order numerical scheme . clearly , we can see that the linearization process described above plays an essential role in the present numerov scheme . in order to measure the accuracy of the present numerov scheme , we can compute the euclidean norm @xmath53^{2}}\ ] ] where @xmath54 stands for the exact solution and @xmath55 stands for the numerical solution . figure [ f2 ] shows ( right vertical axis ) a comparasion between two euclidean norms : one ( dashed line ) using the third - order numerical scheme and the other ( solid line ) using the sixth - order numerical scheme . note that , at @xmath56 , the exact euclidean norm of the third - order scheme is approximately four orders of magnitude above the exact euclidean norm of the sixth - order scheme . naturally , we can see that the sixth - order numerical scheme is much more accurate and efficient than the third - order numerical scheme . of course , we do nt know the exact solution in practical applications . in that case , the best we can do is to compute the mean euclidean norm of the numerical solution @xmath55 , @xmath57 this mean euclidean norm can be used as a convergency criterion , as shown in figure [ f2 ] ( left vertical axis ) . we have applied the numerov method to derive a sixth - order numerical scheme to solve the one - dimensional poisson equation with dirichlet boundary conditions . the resulting recurrence relations were exactly solved and the corresponding execution time was linearized [ see ] in such way to avoid the handling of a dense matrix . therefore , the numerical scheme is both accurate and efficient as illustrated in figure [ fig : tempos ] . moreover , it is extremely ease to implement in any numerical or algebraic computer language . as pointed by j. m. blatt @xcite , the numerov method is both a three - point method , which implies it is stable , and of highest order , which implies it is accurate . all these features make the numerical scheme the canonical method of choice for the integration of the poisson equation . the author wish to thank rafael casalverini for useful discussions and fapesp for financial supports .	in this article , we present an analytical direct method , based on a numerov three - point scheme , which is sixth order accurate and has a linear execution time on the grid dimension , to solve the discrete one - dimensional poisson equation with dirichlet boundary conditions . our results should improve numerical codes used mainly in self - consistent calculations in solid state physics .
understanding the source of quantum advantage in quantum computation is a long - standing issue in quantum information science . previous researches have shown that certain quantum computation is ` classical ' , for the reason that it is efficiently simulateable by classical computers . one example is any computation performed just by local operations and classical communication ( locc ) @xcite without using any entangled resources . all models of quantum computation outperforming classical counterparts use entanglement resources ( such as measurement - based quantum computation @xcite ) or some kind of non - locc operation . non - locc operations are called ` global ' operations . the source of quantum speedup must be due to the properties of the global operations . in this paper , we refer to the properties exclusive to global operations as _ globalness _ of quantum operations . it is also known that not all global operations result in quantum speedup for quantum computation . there must be a specific globalness that differentiates the quantum operations leading to quantum speedup from those do not . the difference may be due to more than one kind of globalness , but even this is not clear at this point . for this reason , having a good understanding of the globalness of quantum operations is important . in this paper , we try to understand the simplest case of the global operations , namely , bipartite unitary operations . to investigate globalness of unitary operations , it is important to clarify what kind of states is given as inputs of the unitary operations . we want to evaluate the globalness that does not depend on a choice of a particular input state . by introducing the concept of _ pieces of quantum information _ , we analyze characterizations of unitary operations for two pieces of quantum information represented by arbitrary unknown states , in terms of _ delocalization power _ @xcite and _ entanglement cost _ @xcite . we compare these characterizations with another characterization , _ entangling power _ of global operations @xcite , which characterizes the globalness of unitary operations acting on a set of known states . then we extend our analysis of globalness in terms of the delocalization power in two ways by introducing new locc tasks . one of the tasks is _ locc one - piece relocalization _ for _ one piece _ of delocalized quantum information that corresponds to the case when a part of input state is unknown and arbitrary but the other part can be chosen from a set of known state . the other task is _ locc one - piece relocation _ for two pieces of delocalized quantum information , which evaluates the ability of the unitary operation to relocate one of the two pieces of quantum information from one hilbert space to another by adding locc to the unitary operation . the rest of the paper is organized as following . in section [ overview ] , we introduce the concept of pieces of quantum information and present an overview on the three characterizations . we summarize the comparison of different aspects of the globalness of bipartite unitary operations presented in the previous works in section [ comparison ] . we extend the analysis of the delocalization power in sections [ fixed_input ] and [ relocation ] . in section [ fixed_input ] , we show the result on locc one - piece relocalization for one piece of delocalized quantum information . in section [ relocation ] , we analyze locc one - piece relocation of two pieces of quantum information . finally , in section [ conclusion ] , we present our conclusion . first , we define _ a piece of quantum information _ for a @xmath0-dimensional quantum system , or _ qudit _ , whose hilbert space is denoted by @xmath1 . if a pure quantum state of @xmath2 qudits @xmath3 is given by @xmath4 where @xmath5 is a fixed set of normalized and mutually orthogonal states in @xmath6 and the coefficients @xmath7 are arbitrary and unknown except for the normalization @xmath8 , the unknown state @xmath9 is said to represent _ one piece _ of quantum information for a qudit . in the formalism presented above , a piece of quantum information for a single qudit can be stored in an @xmath2-qudit system using an arbitrary set of orthonormal states , @xmath5 . any such set of states would form a _ logical _ qudit space , but in a special case satisfying @xmath10 for all @xmath11 , where the set of states @xmath12 forms an orthonormal basis of @xmath13 and @xmath14 is independent of @xmath15 , the piece of quantum information is stored in a _ physical _ qudit . hence it is possible to _ assign _ one physical qudit for each piece of quantum information . using this formalism , now we provide the formal definition of one piece of _ localized _ quantum information for a qudit . we label the qudits of an @xmath2-qudit system from @xmath16 to @xmath17 and denote the hilbert space of qudit @xmath18 by @xmath19 . the hilbert space of @xmath17 qudits _ excluding _ a certain qudit @xmath18 will be denoted by @xmath20 . we will also assume that two different pieces of quantum information in the same system are assigned to different physical qudits . for @xmath21 , a piece of quantum information represented by an unknown @xmath2-qudit state @xmath9 is said to be _ localized _ at _ an assigned _ hilbert space @xmath19 , or simply localized when there is no fear of confusion , if it is represented in the form @xmath22 where @xmath23 is any basis of the hilbert space of the assigned qudit ( _ i.e. _ , @xmath19 ) , @xmath24 is an @xmath25-qudit state determined independently of the set of coefficients @xmath26 , and @xmath26 are arbitrary coefficients satisfying the normalization condition @xmath8 . note that the global phase factor of the coefficients is not a physical quantity , so we take the global phase equivalence . there are @xmath27 complex degrees of freedom in total for one piece of quantum information . for @xmath28 , since @xmath29 is the minimal hilbert space to store one piece of quantum information for a qudit , one piece of quantum information has to be localized in @xmath30 . we define the concept of _ delocalized _ quantum information , which is the complement of localized quantum information , and also the concept of _ delocalization _ of quantum information . if a piece of quantum information is _ not _ localized , then it is said to be _ delocalized_. the task of delocalizing quantum information is called _ delocalization_. next , we consider two - qudit states , where the state of each qudit represents one piece of localized quantum information . we denote the two hilbert spaces of the qudits by @xmath31 and @xmath32 . the two pieces of localized quantum information can be represented by a tensor product state @xmath33 , where the subscripts of the kets denote the assignment of the hilbert spaces of the qudits , @xmath34 and @xmath35 . we define delocalization for two pieces of quantum information as the following . two pieces of quantum information are said to be _ delocalized _ , if the state representing the two pieces of quantum information _ can not _ be written by @xmath36 , where @xmath37 on @xmath38 and @xmath39 on @xmath40 are arbitrary local unitary operations but independent of @xmath41 and @xmath42 . we again note that we have already assigned a hilbert space for each piece of quantum information , so the state @xmath43 represents delocalized two pieces of quantum information out from the assigned hilbert spaces . now we investigate the effects of a global unitary operation @xmath44 applied on two pieces of localized quantum information @xmath33 . if the unitary operation @xmath44 is not a tensor product of two local unitary operations on @xmath38 and @xmath40 , @xmath44 always transforms each piece of localized quantum information to delocalized quantum information . in this paper , we say that the unitary operations have _ delocalization power _ , which in a sense is the ` strength ' of delocalization of quantum information due to the unitary operations . how pieces of quantum information are delocalized is determined only by the set of orthonormal states representing the quantum information , which in turn is determined by the unitary operation used for the delocalization . therefore , the globalness of unitary operations can be studied by understanding how a unitary operation delocalizes pieces of quantum information . later , we argue that certain pieces of quantum information are ` more ' delocalized than others . the difference in the level of delocalization can only have come from the difference in the globalness of the unitary operation , namely , the delocalization power . hence , we can classify the delocalization power by analyzing the level of the delocalization that each unitary operation achieves . to define and classify the level of delocalization , we introduce the following locc task , _ locc one - piece relocalization _ , that aims to localize just one of the two pieces of delocalized quantum information by sacrificing the other piece of quantum information in @xmath45 . we denote the set of density operators on the hilbert space @xmath13 by @xmath46 . _ locc one - piece relocalization _ of qudit @xmath47 for two pieces of quantum information delocalized by a global unitary operation @xmath44 is a task to find an locc - implementable completely positive trace preserving ( cptp ) map @xmath48 satisfying @xmath49 = { { \left \vert}\psi_b { \right \rangle}}_b { { \left \langle}\psi_b { \right \vert}}\ ] ] for any @xmath34 and @xmath35 . we characterize the delocalization power of global unitary operations in terms of their ability to allow locc one - piece relocalization on two pieces of quantum information delocalized by the global unitary operations . we define the order of the delocalization power of two global unitary operations @xmath44 and @xmath50 on two pieces of quantum information by the following . if locc one - piece relocalization of two pieces of delocalized quantum information is possible for a unitary operation @xmath44 , but not possible for another unitary operation @xmath50 , the order of the delocalization power of @xmath44 is defined to be smaller than that of @xmath50 in terms of locc one - piece relocalization . another way to quantify the globalness of a unitary operation applied on quantum information is to evaluate how much extra global resource is required on top of locc operations to implement the unitary operation on two pieces of quantum information . the minimum amount of entanglement required to deterministically implement a given global operation is unique , based on the fact that entanglement can not be generated by locc . we define an locc task , _ entanglement assisted deterministic locc implementation _ and then define _ entanglement cost _ of the unitary operation on quantum information in terms of this locc task . _ entanglement assisted deterministic locc implementation _ of a global unitary operation @xmath44 on two pieces of localized quantum information @xmath51 using a fixed bipartite resource state @xmath52 is a task of finding an locc - implementable cptp map @xmath53 satisfying @xmath54 for any @xmath55 and @xmath56 . _ entanglement cost _ of a unitary operation @xmath44 on two pieces of quantum information is given by the minimum amount of entanglement of the resource state @xmath57 required to perform entanglement assisted deterministic locc implementation of @xmath44 on two pieces of localized quantum information . the entanglement cost of unitary operations can be regarded as the minimum entanglement cost for delocalizing two pieces of quantum information . thus this is another way to characterize the globalness of unitary operations applied on quantum information . we can also quantify the globalness of a unitary operation by evaluating its ability of entanglement generation in place of entanglement cost , similarly to the pair of the definitions for evaluating the globalness of quantum states , namely , distillable entanglement and entanglement cost . however , the amount of entanglement generated by a unitary operation strongly depends on the choice of input states , therefore it is difficult to define a quantity in terms of quantum information , namely , unknown states . instead , _ entangling power _ of a global operation @xcite is defined by optimizing the generated amount of entanglement over a set of known input states . the entangling power of a bipartite unitary operation @xmath44 ( denoted by @xmath58 ) is defined as the maximum amount of entanglement generated between the bipartite cut by applying @xmath44 on a known state , _ i.e. _ , @xmath59 where @xmath60 is an entanglement measure of choice and @xmath61 is chosen among the given set of states . in the previous section , we introduced three different characterizations for the globalness of bipartite unitary operations : delocalization power , entanglement cost and entangling power . in this section , we summarize known results on the three characterizations and investigate whether the globalness characterized by each method is same to or different from that of by the others . [ theorem : qq1pr ] locc one - piece relocalization for two pieces of quantum information of qudits delocalized by a unitary operation @xmath44 is possible if and only if @xmath44 is a locally unitary equivalent to a controlled - unitary operation @xmath62 , where @xmath63 forms an orthonormal basis for one of the subsystems and @xmath64 is a set of unitary operations on the other subsystem . @xcite the characterization of globalness based on the delocalization power reveals that there are two classes of globalness for bipartite unitary operations , one class is a local unitary equivalent of a controlled - unitary operations , and the other class is all the rest of global unitary operations . for any given two - qubit controlled - unitary operation , its entanglement cost for entanglement assisted deterministic locc implementation is 1 ebit when the schmidt number , the number of non - zero schmidt coefficients , of the resource state is 2 . @xcite for other two - qubit unitary operations , the entanglement cost of the swap operation @xmath65 , of which action is given by @xmath66 for any @xmath67 and @xmath68 , is easily shown to be 2 ebit by considering the situation where the two input qubits are parts of maximally entangled states . however , for more general operations , it is not easy to evaluate the minimum entanglement cost of entanglement assisted locc implementation , therefore , it is still unknown . the formulation of entangling power depends on the set of allowed input states and the measure of entanglement . entangling power is usually difficult to calculate because it involves two optimizations . one is the maximization over all possible input states ( usually taken to be separable or product states ) . the other is the calculation of the amount of generated entanglement according to the chosen entanglement measure . even when the quantum operation is restricted to bipartite unitary operations , the exact value is obtained for only limited cases @xcite . for example , it is known that the cnot operation @xmath69 ( @xmath70 where @xmath71 and @xmath72 is given by the pauli matrix @xmath73 ) has the entangling power of 1 ebit and the swap operation @xmath65 has the entangling power of 2 ebit when we allow to use ancilla qubits . nevertheless , we can make a relatively generic statement about entangling power if the entanglement measure is continuous . the statement is as follows . the identity operation clearly generates no entanglement at all , hence its entangling power should be zero . invoking a continuity argument , there should be a set of operations in the neighborhood of the identity operation such that their entangling power is arbitrarily small . however , when we evaluate the globalness in terms of the delocalization power and entanglement cost , a fundamental difference arrises . by using these two characterizations , all two - qubit controlled - unitary operations of the form @xmath74 where @xmath75 is a single qubit unitary operation , belong to the same class of globalness irrelevant to their entangling power . thus , two - qubit controlled - unitary operations and their local unitary equivalent operations belong to a distinct class from the the class of identity operation , even when the local unitary operations are close to the identity ( @xmath76 ) , in contrast to the characterization in terms of entangling power . on the other hand , for general _ known _ bipartite pure qudit states , the locc convertibility condition between two states is known . [ majorizationtheorem ] a bipartite state @xmath77 can be transformed to another state @xmath78 by using only locc , if and only if the schmidt coefficients of the @xmath77 is majorized by those of @xmath78 @xcite . since the locc conversion protocol depends on the choice of states @xmath77 and @xmath78 ( when locc conversion from @xmath77 to @xmath78 is possible ) , it is essential that these states are known . by taking @xmath79 and @xmath80 and using theorem [ majorizationtheorem ] , we can see that if we have a resource state of which schmidt coefficients are equal to the those of @xmath78 , it is possible to obtain @xmath81 by locc . for two - qubit unitary operations , which can only create an entangled state with schmidt number 2 , the majorization condition is equivalent to the comparison of the amount of entanglement . thus , we have the following corollary . the entanglement cost of the resource state for entanglement assisted deterministic locc implementation of unitary operations on a _ given known _ state @xmath33 is given by the the amount of entanglement of @xmath82 . this corollary gives a justification to define the entanglement cost for an entanglement assisted deterministic locc implementation on a set of known states by finding the largest minimum entanglement cost to perform @xmath44 over the set of input states . in this case , the entanglement cost for a set of known states coincides to the entangling power of @xmath44 . the results in this section indicate that there are several aspects of globalness in quantum operations . it is particularly important to clarify the types of input states , known states or unknown states representing pieces of quantum information , for analyzing globalness of unitary operations , since they lead to a fundamental difference . in the classification of globalness of unitary operations in terms of the delocalization power presented in the previous sections , we analyzed global properties of two pieces of delocalized quantum information . that is , we analyzed the globalness of unitary operations totally independent of input states . on the other hand , in section [ delocalizationpwer ] , we also defined _ one piece _ of delocalized quantum information . this situation corresponds to the case where one of the two input qudits is in an arbitrary and unknown state , but that of the other qudit is in a _ known _ state , and we can choose the most suitable state for performing tasks . in this section , we extend our analysis on globalness of unitary operations in terms of delocalization power to the case for one piece of quantum information . we define the task of locc one - piece relocalization of _ one piece _ of delocalized quantum information . _ locc one - piece relocalization _ of the qudit @xmath47 for _ one piece _ of quantum information delocalized by a global unitary operation @xmath44 is a task to find an locc - implementable cptp map @xmath48 and a state @xmath83 satisfying @xmath84 = { { \left \vert}\psi_b { \right \rangle}}_b { { \left \langle}\psi_b { \right \vert}}\ ] ] for any @xmath35 . we show the following lemma . the global unitary operations that allow locc one - piece relocalization for one piece of delocalized quantum information is in a strictly wider class of global unitary operations than that allows locc one - piece relocalization for two pieces of delocalized quantum information . to prove this lemma , we present an example of two - qubit unitary operations , @xmath85 , where locc one - piece relocalization is impossible for delocalized two pieces of quantum information , but it becomes possible if one of the qubits is promised to be in a particular pure state . let us take the computational basis , which is an orthonormal basis of the composite hilbert space @xmath45 given by @xmath86 , where @xmath87 and @xmath88 are orthonormal base for @xmath38 and @xmath40 , respectively . the matrix representation of @xmath89 in the computational basis is given by @xmath90 first , we show that locc one - piece relocalization of the qubit @xmath47 for one piece of quantum information delocalized by @xmath85 is possible by presenting that @xmath85 can be simulated by a locally unitary equivalent operation to a controlled - unitary operation if qubit @xmath91 is set to a particular state . we set the state @xmath92 to be @xmath93 . it is easy to check that for an arbitrary @xmath42 , @xmath94 where @xmath69 is a controlled - not operation and @xmath95 denotes the single - qubit hadamard operation represented in the computational basis by @xmath96 ( the same calculation can be done using the stabilizer formalism @xcite by exploiting the fact that @xmath89 is a clifford operation . ) thus the action of @xmath85 can be simulated by @xmath97 , a locally unitary equivalent operation to the controlled - not operation , when we fix one of the qubits to be in the state @xmath98 . from theorem [ theorem : qq1pr ] , any operation which is locally unitary equivalent to a controlled - unitary operation is locc one - piece relocalizable for two pieces of delocalized quantum information . note that , if an locc protocol relocalizes two pieces of delocalized quantum information , the same protocol must also relocalize one piece of delocalized quantum information . therefore , @xmath85 is locc one - piece relocalizable for one piece of delocalized quantum information . next , we show that @xmath89 itself is not locally unitary equivalent to controlled - unitary operations , therefore it is _ not _ locc one - piece relocalizable for two pieces of delocalized quantum information . to show this , we analyze the cartan numbers for two - qubit unitary operations . it is known that any two - qubit unitary operator on @xmath45 has the following cartan decomposition @xcite , @xmath99 \cdot v_a \otimes v_b,\ ] ] by taking appropriate local unitary operations @xmath37 , @xmath100 on @xmath38 and @xmath39 , @xmath101 on @xmath40 , coefficients @xmath102 ( @xmath103 ) , where @xmath73 , @xmath104 and @xmath105 denote the pauli matrices and each subscript of the pauli matrix indicates the corresponding hilbert space . in this decomposition , the nonlocal component of the unitary operation is represented by the set of coefficients @xmath106 . in this paper , we refer @xmath107 to be a cartan coefficient , and the number of non - zero cartan coefficients to be the cartan number . the cartan number of a unitary operation can not be changed by local unitary operations , and two unitary operations with different cartan numbers can not be locally unitary equivalent to each other @xcite . the cartan decomposition of @xmath89 is given by @xmath108 \cdot v_a \otimes v_b\ ] ] by using appropriate local unitary operators @xmath37 , @xmath100 , @xmath39 and @xmath101 @xcite . thus the cartan number of @xmath89 is 2 . on the other hand , the cartan decomposition of a controlled phase operation @xmath109 where the single - qubit phase operation @xmath110 is defined by @xmath111 is given by @xmath112.\ ] ] thus , the cartan number of the controlled - phase operation is 1 . it is also known that any controlled - unitary operations @xmath113 is locally unitary equivalent to @xmath114 , therefore , the cartan number of the operations that are locally unitary equivalent to controlled - unitary operations is also 1 . therefore , @xmath89 can not be locally unitary equivalent to the controlled - unitary operations . note that some unitary operations remain locc one - piece _ _ un__relocalizable even for one piece of delocalized quantum information . such an example is the swap operation @xmath115 . by performing @xmath115 , even if one of the input qubit states is fixed to a particular known state , a piece of quantum information represented by the other qubit s unknown input state completely moves out from the original hilbert space and stored in the other hilbert space . this phenomena is an example of what we call a _ relocation _ of a piece of quantum information . once this relocation happens , it is not possible to relocalize the piece of quantum information back to the original hilbert space by locc alone . it requires 1 ebit of entanglement to relocalize the one piece of relocated quantum information on top of locc by using quantum teleportation @xcite . in the previous section , we briefly introduced the concept of _ relocation _ of a piece of quantum information . but actually , @xmath115 provides relocation of both two pieces of quantum information . @xmath115 is the only unitary operation that has the delocalization power strong enough to relocate two pieces of quantum information simultaneously without any additional operation or resource . a wider class of unitary operations , namely , the local unitary equivalents of @xmath115 , also relocates two pieces of quantum information , if local operations are allowed as an extra operation . to define and classify delocalization power of unitary operations in terms of relocation , we further relax the condition of the additional operations to locc and investigate locc relocatability of one of the two pieces of quantum information delocalized by unitary operations . _ locc one - piece relocation _ from @xmath38 to @xmath40 for two pieces of quantum information delocalized by a unitary operation @xmath44 is the task to find an locc - implementable cptp map @xmath116 satisfying @xmath117 = { { \left \vert}\psi_a { \right \rangle}}_b { { \left \langle}\psi_a { \right \vert}}\ ] ] for any @xmath118 and @xmath119 . this is a task similar to locc one - piece relocalization for two pieces of quantum information , in the sense by sacrificing one of two pieces of quantum information , we obtain one piece of localized quantum information . the difference between these tasks is the location of the piece of localized quantum information . we define the order of the delocalization power of two global unitary operations @xmath44 and @xmath50 on two pieces of quantum information in terms of locc one - piece relocation by the following . if locc one - piece relocation of two pieces of delocalized quantum information is possible for a unitary operation @xmath44 , but not possible for another unitary operation @xmath50 , the order of the delocalization power of @xmath44 is defined to be larger than that of @xmath50 in terms of locc one - piece relocation . note that for locc one - piece relocation , feasibility of the task implies more delocalization power , whereas for locc one - piece relocalization , feasibility of the task implies less delocalization power . as the first step to classify the delocalization power of global unitary operations in terms of locc one - piece relocation , we show that where the locally unitary equivalent class of controlled - unitary operations lies in this classification . if two pieces of quantum information are delocalized by an operation locally unitary equivalent to controlled - unitary operations , locc one - piece relocation is not possible . to prove this lemma , we employ the formulation of locc using accumulated operators @xcite . in the following subsections , we first summarize the formulation , and then show the proof by contradiction . we adopt the standard formulation of locc @xcite . in a two - party scenario , alice and bob perform one local measurement operation in turns while exchanging the outcome of each measurement operation by classical communication . the measurement operation at a particular turn is chosen according to all the outcomes by the both parties up to that turn , where the choice is made following a protocol agreed beforehand by the parties . strictly speaking , we may consider locc protocols which can not be expressed in this form . these protocols , however , can always be substituted by the protocols in this standard form . each local quantum operation can be described as a generalized measurement , which is represented by a set of operators @xmath120 satisfying the completeness relation @xmath121 there exists one operator for each outcome in the measurement , which is denoted by the superscript @xmath122 . we add a subscript to the outcome index , for example @xmath123 , to specify to which measurement operation the index belongs . in this notation , @xmath123 belongs to the @xmath18-th measurement operation in the sequence . we use @xmath124 to denote the set of measurement outcomes of the first @xmath18 measurement operations in the sequence . the @xmath125-th measurement operation is a function of @xmath126 and we denote the set of operators describing this measurement operation by @xmath127 let us denote alice s measurement operations by @xmath128 and bob s by @xmath129 . we set @xmath130 and @xmath131 note that @xmath128 is an operator on @xmath38 and @xmath129 is on @xmath40 . when @xmath2-th turn is alice s turn then @xmath132-th turn is bob s turn , which implies that alice does not perform any operation during this @xmath132-th turn . in this case , we set alice s measurement operation to the identity operation , _ i.e. _ , @xmath133 if this @xmath132-th turn happens to be bob s , then his measurement operation is set to the identity operation . the effect of the measurement operations accumulates as an locc protocol proceeds . the accumulated effect up to a particular turn is expressed by the product of all the measurement operators corresponding to all the measurement outcomes obtained up to that point . given a particular sequence of measurement outcomes @xmath134 , we represent the accumulated effect corresponding to this sequence by an _ accumulated operator _ @xmath135 defined by @xmath136 bob s accumulated operator will be denoted by @xmath137 defined by a similar way of @xmath135 . let us focus on two - qubit controlled - unitary operations for simplicity . the following argument can be extended to arbitrary two-_qudit _ controlled - unitary operations . we prove by contradiction that locc one - piece relocation of two pieces of quantum information delocalized by any controlled - unitary operation is impossible . now , consider the following scenario where alice has an extra ancilla qubit , whose hilbert space is denoted by @xmath138 . let @xmath139 denote a maximally entangled state between alice s input qubit and the ancilla qubit defined by @xmath140 suppose that there is an locc one - piece relocation protocol for the given controlled - unitary operation @xmath113 defined by eq . ( [ c - u ] ) . let alice set her input qubit and the ancilla in the state of @xmath139 while bob s input remains arbitrary . alice and bob perform @xmath113 and the locc protocol to complete the relocation of one piece of quantum information from @xmath38 to @xmath40 . then alice s ancilla qubit and bob s input qubit are in the state of @xmath141 which implies that @xmath142 holds for an arbitrary @xmath143 . using the accumulated operator representation of @xmath144 , we have @xmath145 = { { \left \vert}\phi { \right \rangle}}_{ab}{{\left \langle}\phi { \right \vert}}.\ ] ] we modify the locc protocol @xmath144 by adding an extra measurement operation by alice described by @xmath146 , just after alice s final measurement . we denote this modified protocol by @xmath147 . direct substitution reveals that @xmath148\\ + { \mathrm{tr}}_a [ \sum_{{\vec{r}_{n } } } ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}1 { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n } } } ) ( c_u { { \left \vert}\phi { \right \rangle}}_{aa}{{\left \langle}\phi { \right \vert } } \otimes { { \left \vert}\psi_b { \right \rangle}}{{\left \langle}\psi_b { \right \vert } } c_u^\dag ) ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}1 { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n}}})^\dag ] . \end{gathered}\ ] ] since the partial trace @xmath149 is taken and the additional measurement introduced for the protocol @xmath147 acts only on @xmath38 , we have @xmath150 thus we obtain @xmath151 because the right hand side is a pure state , it must be true that @xmath152\\ = p^{{\vec{r}_{n}},k,\psi_b } { { \left \vert}\phi { \right \rangle}}_{ab}{{\left \langle}\phi { \right \vert}}\end{gathered}\ ] ] for all @xmath134 and @xmath153 , where @xmath154 is a positive coefficient normalized by @xmath155 since eq . ( [ relocation_perbranch ] ) holds for any @xmath143 , we can replace @xmath68 by a completely mixed state @xmath156 , and obtain @xmath157\\ = \frac{1}{2 } { \mathrm{tr}}_a [ ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}k { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n } } } ) ( c_u { { \left \vert}\phi { \right \rangle}}_{aa}{{\left \langle}\phi { \right \vert } } \otimes { { \left \vert}0 { \right \rangle}}{{\left \langle}0 { \right \vert } } c_u^\dag ) ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}k { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n}}})^\dag]\\ \qquad + \frac{1}{2 } { \mathrm{tr}}_a [ ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}k { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n } } } ) ( c_u { { \left \vert}\phi { \right \rangle}}_{aa}{{\left \langle}\phi { \right \vert } } \otimes { { \left \vert}1 { \right \rangle}}{{\left \langle}1 { \right \vert } } c_u^\dag ) ( { { \left \vert}0 { \right \rangle}}_a{{\left \langle}k { \right \vert}}a^{{\vec{r}_{n } } } \otimes b^{{\vec{r}_{n}}})^\dag]\\ = ( p^{{\vec{r}_{n}},k,0 } + p^{{\vec{r}_{n}},k,1 } ) { { \left \vert}\phi { \right \rangle}}_{ab}{{\left \langle}\phi { \right \vert}}.\end{gathered}\ ] ] note that @xmath158 acts only on alice s input qubit . taking the partial trace over alice s _ ancilla _ qubit @xmath159 , eq . ( [ relocation_oni ] ) gives @xmath160 = \frac{(p^{{\vec{r}_{n}},k,0 } + p^{{\vec{r}_{n}},k,1})}{2 } { \mathbb{i}},\ ] ] where we have used the relation @xmath161 noting that the identity operator commutes with any unitary operators , after performing the partial trace @xmath149 , we have @xmath162 this equation guarantees that bob s accumulated operator @xmath137 for each sequence of measurement outcomes @xmath134 is proportional to a unitary operator , _ i.e. _ , @xmath163 where the coefficient @xmath164 is set to satisfy @xmath165 for any linear operator @xmath166 on @xmath38 and the maximally entangled state given by @xmath139 , @xmath167 , where @xmath168 denotes the transpose of @xmath166 in the computational basis , holds . let @xmath169 denote a set of operators forming a basis for the operators on @xmath170 ( where @xmath171 ) . that is , for any @xmath166 on @xmath170 , there exists a set of complex numbers @xmath172 such that @xmath173 ( an example of such a basis is the set of pauli operators and the identity operator , if the hilbert space in question has the dimension of 2 . ) with this basis , @xmath113 on @xmath174 can be expressed as a linear combination of @xmath175 , namely , @xmath176 where @xmath177 denotes the coefficient of @xmath175 . let us choose @xmath178 to satisfy @xmath179 in the computational basis and define @xmath180 on @xmath181 by @xmath182 under these conventions , we have @xmath183 comparing this equation to eq . ( [ relocation_perbranch ] ) , it must be that @xmath184 which is equivalent to @xmath185 let an ancilla state ( not necessarily normalized ) @xmath186 be defined by @xmath187 by substituting eq . ( [ bisunitary ] ) into eq . ( [ relocation_vectorform ] ) , we conclude that @xmath188 holds for all @xmath68 . the right hand side is collinear to @xmath189 for all @xmath68 . on the other hand , because @xmath190 is invertible , the left hand side returns linearly independent vectors when @xmath191 are chosen linearly independently . this , however , is a contradiction proving that the assumption that locc one - piece relocalization is possible for two pieces of quantum information delocalized by the controlled - unitary operations @xmath113 must not hold . this proof strongly depends on the fact that bob s input state is kept arbitrary , namely , we considered the situation of delocalized two pieces of quantum information . indeed , if we are allowed to choose bob s input state , one - piece relocation is possible for certain controlled - unitary operations . an example is the controlled - not operation on two qubits . in this paper , we first introduced the concept of pieces of quantum information and reviewed three different characterizations of the globalness of bipartite unitary operations , which were delocalization power , entanglement cost , and entangling power . the first two characterizations are on the globalness of the unitary operations on two pieces of quantum information represented by unknown states , and the last one is on the globalness of the unitary operations on a set of known states . we showed the fundamental difference between these two types of globalness of the unitary operations . next , we extended our analysis on characterization in terms of the delocalization power by introducing a new locc task , locc one - piece relocalization of _ one piece _ of quantum information delocalized by a unitary operation . we showed that there are unitary operations which belong to a higher globalness class in terms of the delocalization power than the local unitary equivalents of controlled - unitary operations , and such operations can be further divided into two subclasses depending on the possibility of this task . we also introduced another new task , locc one - piece relocation of two pieces of delocalized quantum information . we proved that locc one - piece relocation is impossible for any controlled - unitary operations . this confirms that the local unitary equivalents of controlled - unitary operations , which are locc one - piece relocalizable , belong to a class of global operations with relatively weak globalness also in terms of locc relocation of quantum information . in our analysis of the locc tasks , we focused on the locc tasks that transform two pieces of quantum information within the two - qudit hilbert space . this is because our main purpose is to investigate the delocalization power of two - qudit unitary operations . but in general , we can investigate more general properties of delocalized pieces of quantum information by considering locc tasks that transforms @xmath2 pieces of quantum information delocalized in an @xmath192-qudit subspace of a totally @xmath193-qudit hilbert space ( @xmath194 and @xmath195 ) to @xmath196 pieces of quantum information delocalized in an @xmath197-qudit subspace ( @xmath198 and @xmath199 ) , by limiting the allowed operations to be locc for a certain devision of the total hilbert space . self - teleportation @xcite can be interpreted as a special case of this generalized locc task for @xmath200 , @xmath201 and @xmath202 . by denoting the total hilbert space by @xmath203 , only locc is allowed between the division of @xmath38 and @xmath204 in this case . it is shown that asymptotically , @xmath18-copies of any delocalized two pieces of quantum information in @xmath45 can be approximately ` relocated ' to @xmath204 . the error probability of this relocation drops exponentially with the number of copies @xmath18 as long as the two pieces of quantum information are delocalized , namely , not in a product state . in our analysis of the delocalization power , we characterized the order of delocalization power of unitary operations by their ability allowing the locc tasks . for more quantitative analysis of the unitary operations that do not allow the locc tasks , it is important to analyze the entanglement cost of the corresponding entanglement assisted versions of the locc relocalization / relocation tasks . quantum state merging @xcite can be interpreted as evaluating the entanglement cost for performing entanglement assisted approximate generalized locc task for @xmath205 , @xmath206 and @xmath207 ( @xmath208 ) , where only locc is allowed between the division of @xmath38 and @xmath204 and no operation is allowed on @xmath209 . this is an entanglement assisted locc task to achieve relocation of three pieces of quantum information delocalized in @xmath210 to @xmath211 . in the asymptotic limit , it is shown that the entanglement cost coincides with the quantum conditional entropy , which provides an operational interpretation of the quantum conditional entropy . to understand information theoretical meanings of our locc tasks , it is interesting to analyze asymptotic settings of our locc and entanglement assisted locc tasks . we leave these investigations for future works . bennett , c. h. , brassard , g. , crpeau , c. , jozsa , r. and peres , w.k . ( 1993 ) teleporting an unknown quantum state via dual classical and einstein - podolsky - rosen channels . lett . _ * 70 * , 18951899 . nielsen , m. a. , dawson , c.m . , dodd , j. l. , gilchrist , a. , mortimer , d. , osborne , t.j . , bremner , m. j. , harrow , a. w. and hines , a. ( 2003 ) quantum dynamics as a physical resource . a _ * 67 * , 052301 .	we compare three different characterizations of the globalness of bipartite unitary operations , namely , delocalization power , entanglement cost , and entangling power , to investigate global properties of unitary operations . we show that the globalness of the same unitary operation depends on whether input states are given by unknown states representing pieces of quantum information , or a set of known states for the characterization . we extend our analysis on the delocalization power in two ways . first , we show that the delocalization power differs whether the global operation is applied on one piece or two pieces of quantum information . second , by introducing a new task called locc one - piece relocation , we prove that the controlled - unitary operations do not have the delocalization power strong enough to relocate one of two pieces of quantum information by adding locc .
ultracold and interacting dilute alkali - metal vapors trapped by one - dimensional double - well potentials @xcite provides the possibility of studying the formation of macroscopic coherent states @xcite and macroscopic schrdinger - cat states @xcite . the two - site bose - hubbard ( bh ) hamiltonian @xcite efficiently describes the microscopic dynamics of such systems . when the boson - boson interaction is repulsive and the number of bosons is even , the crossover from a delocalized atomic coherent state to a ( fully incoherent ) localized fock state ( the so called twin fock state with the particles equally shared between the two wells ) takes place by increasing the interatomic coupling strength @xcite . for attractively interacting bosons , the two - spatial mode bh model predicts the formation of a macroscopic schrdinger - cat state @xcite when the interatomic attraction becomes sufficiently large . finally , when the attraction between the bosons is sufficiently strong the collapse should take place @xcite . motivated by the concrete possibility to isolate single atomic ions @xcite and manipulate quantum gases at single - atom level @xcite ( note that d. j. wineland was awarded in 2012 with the physics nobel prize for his activity in this sector ) , we focus on the behavior of few trapped bosonic atoms at zero temperature . the aim of the present work , then , is to study the ground state of a system consisting of a low number @xmath0 of bosons confined in a symmetric double - well trap and characterize it from the quantum correlations point of view . to do this we use the two - site bose - hubbard model . we diagonalize the underlying hamiltonian by analytically finding the eigenvector and the eigenvalue of its lowest energetic state for @xmath1 - this case has already been discussed in @xcite - and @xmath7 bosons . hence , we provide analytical formulas for the parameters that describe the correlation properties of the ground state of the system . these parameters are : the fisher information @xmath4 @xcite which is related to the fluctuation of the number of bosons in a given well and achieves its maximum in correspondence to the schrdinger - cat state ; the coherence visibility @xmath5 @xcite which measures the coherence related to the single - particle tunneling across the central barrier and attains its maximum value in correspondence to the atomic coherent state ; the entanglement entropy @xmath6 @xcite which quantifies the amount of the genuine quantum correlations of the ground state from the bi - partition perspective . in particular , we calculate @xmath4 and @xmath5 following two paths : on one hand by taking the average , with respect to the ground state , of the left - right population imbalance variance and the left - well hopping operator , respectively , and on the other hand by applying the hellmann - feynman theorem @xcite . for both the calculations ( that , as we shall comment , provide the same results ) we use the analytically determined ground - state eigenvectors and eigenvalues . we study the ground state and the parameters @xmath4 , @xmath5 , @xmath6 by widely exploring the atom - atom interaction range , from strong attractions to strong repulsions . in this latter regime , we comment about the of @xmath0 even-@xmath0 odd difference : when @xmath0 is even ( the ratio of the number of bosons to the number of wells is a positive integer ) the ground state is a separable fock state with @xmath8 particles in the left well and @xmath8 particles in the right well ( this is , as commented at the beginning , the twin fock state ) , while when @xmath0 is odd ( the total number of bosons is not commensurate with the number of wells ) the ground state is given by a symmetric combination of two separable fock states . when the boson - boson repulsion becomes sufficiently large , the quantities @xmath4 , @xmath5 , @xmath6 , tend to zero for an even number of particles ; they remain , instead , finite when @xmath0 is odd . we analyze a finite number @xmath0 of identical interacting bosonic atoms at zero temperature confined by a trapping potential @xmath9 . we suppose that this potential is given by the superposition of an isotropic harmonic confinement in the radial plane ( @xmath10 ) and a double - well potential @xmath11 in the axial ( @xmath12 ) direction , i.e. @xmath13 where @xmath14 is the mass of the bosons and @xmath15 the trapping frequency in the radial plane . we assume that the double - well is symmetric in the @xmath12 direction and that the system is quasi one - dimensional due to a strong transverse radial harmonic confinement . in the second quantization language , the hamiltonian that controls the microscopic dynamics of the system is @xmath16 the field operator @xmath17 ( @xmath18 ) destroys ( creates ) a boson in the position @xmath19 . @xmath17 and @xmath18 satisfy the usual bosonic commutation rules : @xmath20=\delta^{(3)}({\bf r}-{\bf r'})$ ] , and @xmath21=0=[\hat{\psi}({\bf r})^\dagger,\hat{\psi}^\dagger({\bf r'})]$ ] . we assume that the bosons interact between each other via short - range interactions , so that the atom - atom interaction potential @xmath22 can be described ( in the dilute regime and for ultra - low temperatures ) by a contact potential given by @xmath23 where the coupling constant @xmath24 is equal to @xmath25 with @xmath26 the s - wave scattering length . therefore the hamiltonian ( [ system : ham0 ] ) becomes @xmath27 under the hypothesis that only the lowest energetic doublet of the potential @xmath11 is populated , we expand the field operator @xmath17 according the two - spatial mode decomposition : @xmath28 where @xmath29 ( @xmath30 ) - @xmath31 , with @xmath32 denoting the left ( right ) well - destroys ( creates ) a boson in the @xmath33th well . the single - particle operators @xmath29 and @xmath30 satisfy the bosonic commutation rules : @xmath34=\delta_{k , j } \nonumber\\ & & [ \hat{a}_k,\hat{a}_j]=0=[\hat{a}^\dagger_k,\hat{a}^\dagger_j ] \;.\end{aligned}\ ] ] due to the form of the trapping potential given by eq . ( [ trapping ] ) , the single - particle wave function @xmath35 ( @xmath31 ) can be written according to the factorization @xmath36 where @xmath37 and @xmath38 are the ground - state wave functions of the harmonic oscillator potentials @xmath39 and @xmath40 , respectively . the single - particle wave functions @xmath41 and @xmath42 are tightly localized in the left and right well , respectively , and satisfy the orthonormalization conditions @xmath43 @xmath44 ( with @xmath45 the complex conjugate of @xmath46 ) so that @xmath47 we use the expansion ( [ expansion ] ) and its hermitian conjugate at the right - hand side of eq . ( [ system : ham1 ] ) ; by exploiting the orthonormalization conditions ( [ system : rel2 ] ) and the fact that @xmath11 is symmetric , the well known two - site bose - hubbard hamiltonian @xcite is achieved @xmath48 here @xmath49 is the operator counting the number of bosons in the @xmath33th well . note that the hamiltonian ( [ ham : bh ] ) commutes with the total number operator @xmath50 . the amplitude @xmath51 measures the strength of the boson - boson interaction in the same well ( on - site or intra - well interaction ) @xmath52 with @xmath53 . the sign of @xmath51 is controlled by that of @xmath26 which can be experimentally tuned via the feshbach resonance technique , so that when @xmath26 is positive ( negative ) the bosons are repulsively ( attractively ) interacting . @xmath54 is the tunnel matrix element between the two wells : @xmath55 to capture the main properties of the system , we focus on the eigenproblem @xmath56 for a fixed number @xmath0 of bosons . in this case the hamiltonian @xmath57 can be represented by a @xmath58 matrix in the fock basis @xmath59 ( with @xmath60 denoting the tensor product ) with @xmath61 . for each eigenvalue @xmath62 , with @xmath63 , the corresponding eigenstate @xmath64 will be of the form @xmath65 is the probability to have @xmath66 ( @xmath67 ) bosons in the left ( right ) well when the system is in the @xmath68th eigenstate of the two - site bh hamiltonian . note that since the left - right symmetry of the hamiltonian ( [ ham : bh ] ) , for any eigenstate one has that @xmath69 where the average @xmath70 is taken with respect to the given eigenstate . we are analyzing the system at zero temperature . then , the only two - site bh hamiltonian eigenstate to be occupied is the lowest one , so that in the following we shall denote the corresponding eigenvector and eigenvalue simply by @xmath71 and @xmath72 , respectively . the expansion coefficients with respect to the basis @xmath73 shall be , then , denoted by @xmath74 . as discussed in @xcite , the ground state of the hamiltonian ( [ ham : bh ] ) features different behaviors depending on the interplay between the on - site interaction @xmath51 and the hopping amplitude @xmath54 . following the same path followed in @xcite , we study the ground state in terms of the dimensionless parameter @xmath75 . let us start with some limit cases . * @xmath76 . the ground state is the atomic coherent state @xcite @xmath77 ( the energy associated to this state is @xmath78 ) where @xmath79 is the tensor product between the vacuum of the operator @xmath80 and the vacuum of @xmath81 , i.e. we have no particles in the left well and no particles in the right well . * @xmath82 : @xmath83 . in the case of a strong repulsive interaction and with an even number @xmath0 of bosons , as well known , the ground state tends to the twin fock state @xmath84 if @xmath0 , instead , is odd , when @xmath83 the ground state tends to @xmath85 to understand this , let us consider the extreme case of complete absence of hopping , that is @xmath86 . in this case the eigenvalues of the hamiltonian ( [ ham : bh ] ) are given by those of the intra - well term : @xmath87 . we are here considering the state with @xmath66 @xmath88 bosons in the left ( right ) well . requiring that @xmath89 provides @xmath90 . since @xmath0 is odd and @xmath66 must be an integer , the values of @xmath66 which minimize @xmath72 are those integer closest to @xmath8 , i.e. @xmath91 and @xmath92 that correspond to the two separable fock states @xmath93 these states , although having the ( same ) minimum energy , do not satisfy the condition ( [ lrs ] ) . nevertheless , it is easy to prove that the state @xmath94 has the same energy of @xmath95 ( @xmath96 ) and satisfies the condition ( [ lrs ] ) . * @xmath97 : @xmath98 . in the case of a strong attractive interaction , the ground state tends to the macroscopic superposition state @xmath99 this state , frequently called noon state , is the boson - version of the schrdinger cat state @xcite . at this point it is worth to observe that apart the issue of the possible collapse ( related to attractive interactions ) , the realization of the cat state is not trivial due to the very tiny separation ( in the presence of finite couplings ) between the two lowest levels that makes the cat state very fragile , see , for example , @xcite . in this section we introduce the parameters that we use to characterize the correlations of the ground state of the two - site bh hamiltonian ( [ ham : bh ] ) . these parameters are the fisher information , the coherence visibility , and the entanglement entropy . in the meanwhile , it is useful to remind the well - known properties : @xmath100 and @xmath101 * _ fisher information . _ + the quantum fisher information @xmath102 is the quantity @xcite @xmath103 where the expectation values are taken with respect to the ground state @xmath71 . by using the orthonormality condition ( [ ort ] ) and rules ( [ spaction ] ) in eq . ( [ qfi ] ) , we can express @xmath102 in terms of the expansion coefficients @xmath74 as follows : @xmath104 it is convenient to normalize @xmath102 at its maximum value @xmath105 by defining the fisher information @xmath4 as @xmath106 so that we have a quantity varying in the range @xmath107 $ ] . in terms of the coefficients @xmath74 , @xmath4 is @xmath108 this @xmath4 will be equal to @xmath109 for the noon state ( [ cat ] ) . * _ coherence visibility . _ + in ultracold atom physics , it is customary to investigate the coherence properties in terms of the momentum distribution @xmath110 which is the fourier transform of the one - body density matrix @xmath111 @xcite : @xmath112 where @xmath113 with the operators @xmath114 and @xmath115 - satisfying the standard bosonic commutation rules - annihilating and creating , respectively , a boson at the point @xmath116 , and the average @xmath70 being the ground - state average . following refs . @xcite , it is possible to show that the momentum distribution @xmath110 can be written as @xmath117 here @xmath118 is the momentum distribution in the fully incoherent regime ( @xmath118 depends on the shape of the double - well potential @xmath11 ) , and @xmath119 is the distance between the two minima of @xmath11 . @xmath5 is a real quantity which measures the visibility of the interference fringes . this visibility is given by @xmath120 where the expectation value is taken with respect to the ground state . the quantity @xmath5 characterizes the degree of coherence , between the two wells , related to the left - right ( and back ) tunneling . + we can express the coherence visibility ( [ visibility ] ) in terms of the coefficients @xmath74 by using in eq . ( [ visibility ] ) the rules ( [ spaction ] ) and the orthonormalization condition ( [ ort ] ) , so that one has @xmath121 where @xmath122 is the complex conjugate of @xmath123 . @xmath5 is maximum , that is @xmath109 , for the atomic coherent state ( [ acs ] ) . * _ entanglement entropy . _ + finally , it is interesting to analyze the genuine quantum correlations pertaining to the ground state @xmath71 . in particular , we study the quantum entanglement of @xmath71 from the perspective of the bi - partition . in this framework , the two partitions are given by the left well and right one . when the system is in @xmath71 , the density matrix @xmath124 is @xmath125 an excellent measure of the entanglement between the two wells is provided by the entanglement entropy @xmath6 @xcite . this quantity is the von neumann entropy of the reduced density matrix @xmath126 defined by @xmath127 that is the matrix obtained by partial tracing the total density matrix ( [ dm ] ) over the degrees of freedom of the right ( left ) well ( note that @xmath128 ) . by using the definition of trace of a matrix and the orthonormalization condition ( [ ort ] ) , it is possible to show that the entanglement entropy @xmath129 is given by @xmath130 for a given number of bosons @xmath0 , the theoretical maximum value of @xmath6 is @xmath131 that would correspond to the situation in which the quantities @xmath132 are all equal : @xmath133 whatever @xmath66 . [ cols="^,^,^,^,^",options="header " , ] table 3 . the fisher information @xmath4 , the coherence visibility @xmath5 , and the entanglement entropy @xmath6 , for the atomic coherent state ( [ acs ] ) , the twin fock state ( [ fock ] ) , and the noon state ( [ cat ] ) with @xmath3 bosons . + in this section , we determine the ground state of the two - site bose - hubbard hamiltonian when @xmath134 , @xmath1 , @xmath2 , and @xmath3 . we calculate the fisher information ( [ fi ] ) , the coherence visibility ( [ visibility ] ) , and the entanglement entropy ( [ ee ] ) for a system with @xmath134 , @xmath1 , @xmath2 , and @xmath3 bosons . we analyze the structure of the ground state and @xmath4 , @xmath5 , @xmath6 in terms of the scaled on - site interaction @xmath135 . as first , we represent the hamiltonian @xmath136 with respect to the fock basis @xmath73 . we start from the right - hand side of eq . ( [ ham : bh ] ) and use the rules ( [ spaction ] ) and the orthonormalization condition ( [ ort ] ) . note that , here , we measure the energies in units of @xmath54 . we shall denote by the symbols @xmath137 and @xmath138 the dimensionless energetic quantities , i.e. @xmath139 and @xmath140 . when @xmath134 , the hamiltonian ( [ ham : bh ] ) consists of the only hopping term . in this case , the two - site bose - hubbard hamiltonian , given by eq . ( [ ham : bh ] ) , in the fock basis @xmath73 is @xmath141 the eigenvector @xmath71 associated to the ground state is @xmath142 and the related eigenvalue is @xmath143 then , it is easy to see that the state @xmath71 is the atomic coherent state ( [ acs ] ) with @xmath134 that coincides with the state noon , eq . ( [ cat ] ) , and with the state ( [ pseudo : fock ] ) with @xmath134 . in this case , by using eqs . ( [ fi2 ] ) , ( [ visibility2 ] ) , and ( [ ee ] ) , we immediately see that @xmath144 . at this point , let us focus on a number of bosons larger than one . we therefore consider the cases @xmath1 , @xmath2 , and @xmath3 . the matrices corresponding to the two - site bose - hubbard hamiltonian ( [ ham : bh ] ) with @xmath1 , @xmath2 , and @xmath3 are , respectively @xmath145 @xmath146 @xmath147 where @xmath135 . the ground - state energy @xmath138 pertaining to the case @xmath1 and the corresponding eigenvector @xmath71 are , respectively @xmath148 @xmath149 so that @xmath150 for @xmath2 , we get @xmath151 @xmath152 so that @xmath153 when @xmath3 , for the energy of the ground state we obtain the following result : @xmath154 where @xmath155 . @xmath156 ( the second row of eq . ( [ egs4 ] ) ) when @xmath157 and @xmath158 , and @xmath159 ( the third row of eq . ( [ egs4 ] ) ) when @xmath160 and @xmath161 with @xmath162 . moreover @xmath163 , @xmath164 . note that when @xmath165 , the energy in eq . ( [ egs4 ] ) gives back @xmath166 ( in units of @xmath54 ) for @xmath156 ( @xmath167 ) , i.e. the energy of to the atomic coherent state ( [ acs ] ) . the eigenvector pertaining to the energy in eq . ( [ egs4 ] ) is @xmath168 so that @xmath169 the factors @xmath170 , @xmath171 , and @xmath172 are normalization factors given by the following formulas : @xmath173 @xmath174 @xmath175 with @xmath138 given by eq . ( [ egs4 ] ) . note that eqs . ( [ egs2 ] ) and ( [ sgs2 ] ) are the same that we found in @xcite . we observe that in the limit @xmath176 , the states ( [ sgs2 ] ) , ( [ sgs3 ] ) and ( [ sgs4 ] ) becomes , as expected , @xmath177 i.e. the boson - version of the schrdinger cat state ( [ cat ] ) with @xmath1 , @xmath178 which is the schrdinger cat state ( [ cat ] ) with @xmath2 , and similarly @xmath179 when @xmath180 , from the state ( [ sgs2 ] ) ( @xmath1 ) we retrieve @xmath181 ) ( @xmath2 ) one gets @xmath182 when @xmath3 and @xmath183 , the state ( [ sgs4 ] ) gives @xmath184 these last three states represent the forms assumed by the atomic coherent state @xmath185 ( [ acs ] ) when @xmath1 , @xmath2 , and @xmath3 , respectively . in the limit @xmath186 , the state ( [ sgs2 ] ) ( @xmath1 ) becomes @xmath187 which is the twin fock state ( [ fock ] ) with @xmath1 . instead for @xmath2 , in the deep repulsive regime , @xmath186 , the state ( [ sgs3 ] ) becomes @xmath188 that is the state ( [ pseudo : fock ] ) with @xmath2 . finally , when @xmath3 and @xmath186 , we retrieve the twin fock state @xmath189 . to understand the role of the intra - well interaction - hopping interplay in determining the structure of the ground state of the two - site bh hamiltonian , we have studied the changes experienced by the probabilities @xmath190 by varying the scaled on - site interaction @xmath135 in the presence of @xmath1 , @xmath2 , and @xmath3 bosons , see fig . 1 . from this figure , we can see that a crossover occurs when @xmath191 ranges from @xmath192 ( the two top panels of fig . 1 : the largest probabilities @xmath190 are located in correspondence to @xmath193 and @xmath194 , this being representative of cat - like states ( [ cat ] ) with @xmath195 and @xmath2 bosons ) to @xmath196 ( the two bottom panels of fig . 1 : @xmath190 reaches its largest value in correspondence to @xmath187 when @xmath1 and @xmath189 when @xmath3 - separable twin fock state ( [ fock ] ) - and in correspondence to the states @xmath197 and @xmath198 when @xmath2 , state ( [ pseudo : fock ] ) ) passing for @xmath76 ( the three panels at the fifth row from the top ) describing an almost gaussian distribution of the probabilities @xmath190 , that is the atomic coherent state ( [ acs ] ) . note that at fifth row of fig . 1 , the plot of @xmath190 with @xmath3 has been , nominally , labeled by @xmath76 ; actually , this plot has been obtained by performing the limit @xmath183 in the equations for @xmath190 obtained by eq . ( [ c4 ] ) . + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] + for different values of @xmath135 . horizontal axis : kets @xmath73 ( @xmath61 ) . left : @xmath1 . middle : @xmath2 . right : @xmath3 . at the third row ( from the top ) : plots for @xmath191 signing the maximum of the entanglement entropies ( [ s2 ] ) ( @xmath1 ) , ( [ s3 ] ) ( @xmath2 ) , ( [ s4 ] ) ( @xmath3 ) . all the quantities are dimensionless.,title="fig : " ] in conclusions , we can say that a very strong boson - boson attraction tends to establish in the system a ground state given by a symmetric superposition of two fully populated fock states both with @xmath195 and @xmath2 bosons . on the other hand , a very large interatomic repulsion induces different ground states depending if @xmath1 ( separable twin - fock state ) or @xmath2 ( symmetric superposition of quasi - fully populated fock states ) . on the repulsive side , the above described crossover , for even @xmath0 , is reminescent of the quantum phase transition with optical - lattice - confined bosons theoretically predicted in @xcite and experimentally observed by greiner and co - workers @xcite . this transition - induced by varying the depth of the optical potential - is a transition from the superfluid phase ( the hopping dominates the hamiltonian , @xmath199 . in this case each atom is spread out over the entire lattice ) to the mott insulator one ( on - site interactions dominates the hamiltonian , @xmath200 . in this case , exact numbers of atoms are localized at individual lattice sites ) . note that the even - odd difference ( separable twin fock state versus symmetric superposition of non - fully populated fock states ) which tends to become less relevant for larger particle numbers , indeed , is a well known mott insulators feature , as commented , for example , in ref . as commented before , we characterize the correlations of the ground state by calculating the fisher information , the coherence visibility , and the entanglement entropy . it is possible to achieve , in the case of @xmath201 , @xmath202 , and @xmath203 bosons , analytical formulas for these three parameters . let us start by evaluating the fisher information . to this end we employ at the right - hand side of eq . ( [ fi2 ] ) the expressions for the coefficients @xmath74 given by eq . ( [ c2 ] ) when @xmath1 , by eq . ( [ c3 ] ) when @xmath2 , and by eq . ( [ c4 ] ) when @xmath3 ( with the normalization factors @xmath170 ( @xmath1 ) , @xmath171 ( @xmath2 ) , and @xmath172 ( @xmath3 ) given by eq . ( [ a2 ] ) , eq . ( [ a3 ] ) , and eq . ( [ a4 ] ) respectively ; note that the @xmath74 s are real for any @xmath0 , so that @xmath204 ) . then , for @xmath1 we obtain @xmath205 while for @xmath2 one gets @xmath206 ^ 2 ) } \;.\ ] ] for @xmath3 , the fisher information is given by @xmath207 with @xmath138 being given by eq . ( [ egs4 ] ) . we have studied the fisher informations @xmath4 given by eqs . ( [ f2 ] ) , ( [ f3 ] ) and ( [ f4 ] ) as functions of the dimensionless parameter @xmath135 , see the top panel of fig . 2 . as it can be seen from this figure , when the boson - boson interaction is strongly attractive ( @xmath208 , this being correspondent to states close to the cat state ( [ cat ] ) ) @xmath4 tends to @xmath109 . in the deep repulsive regime , it can be observed that when @xmath209 and @xmath195 ( solid line , dot - dashed ) , when the ground state tends to a separable fock state ( [ fock ] ) , @xmath4 tends to zero . with @xmath2 ( dashed line ) , when the ground state tends to a superposition of two separable fock states given by eq . ( [ pseudo : fock ] ) , the fisher information tends to a finite value ( see also the tabs . 1 - 3 ) . to obtain the coherence visibility @xmath5 , we use at the right - hand side of eq . ( [ visibility2 ] ) the form of the @xmath74 s provided by eq . ( [ c2 ] ) when @xmath1 , by eq . ( [ c3 ] ) when @xmath2 , and by eq . ( [ c4 ] ) when @xmath3 ( with the normalization factors @xmath170 ( @xmath1 ) , @xmath171 ( @xmath2 ) , @xmath172 ( @xmath3 ) given by eq . ( [ a2 ] ) , eq . ( [ a3 ] ) , eq . ( [ a4 ] ) , respectively ) . for @xmath1 , we get @xmath210 for @xmath2 @xmath211 ^ 2 ) } \;,\ ] ] and for @xmath3 @xmath212 with @xmath138 given by eq . ( [ egs4 ] ) and @xmath213 by the second row of eq . ( [ f4 ] ) . we show the behavior of @xmath5 ( eq . ( [ alpha2 ] ) , eq . ( [ alpha3 ] ) ) , eq . ( [ alpha4 ] ) ) , when the scaled on - site interaction is varied , in the middle panel of fig . @xmath5 reaches its maximum value ( @xmath214 ) , both when @xmath195 ( solid line , dot - dashed ) and @xmath2 ( dashed line ) , in the absence of boson - boson interaction that corresponds to the atomic coherent state ( [ acs ] ) . for strongly attractive bosons , @xmath208 , the ground state is a cat - like state , and the coherence visibility approches to zero ( see the solid ( dot - dashed ) line , @xmath215 , and the dashed one , @xmath2 ) . when the repulsion between the bosons is sufficiently strong , @xmath209 , we can see that for @xmath195 ( solid line , dot - dashed line ) - when the ground state tends to a fully incoherent state - @xmath5 approaches to zero , while for @xmath2 ( dashed line ) - when the ground state is close to state ( [ pseudo : fock ] ) - @xmath5 is finite . finally , by employing the expressions of the coefficients @xmath74 provided by eq . ( [ c2 ] ) in eq . ( [ ee ] ) , we calculate the entanglement entropy @xmath6 for @xmath1 , and get @xmath216 + \frac{s_2}{4}\log_2[\frac{s_2a_2^{2}}{4}]\bigg ) \;,\ ] ] where @xmath170 is the normalization factor given by eq . ( [ a2 ] ) and @xmath217 by employing the expressions of the coefficients @xmath74 provided by eq . ( [ c3 ] ) in eq . ( [ ee ] ) , we calculate the entanglement entropy @xmath6 for @xmath2 , which has the following expression @xmath218 + 3\log_2[a_{3}^2]\bigg ) \;,\ ] ] where @xmath171 is the normalization factor given by eq . ( [ a3 ] ) and @xmath219 by following the same path for @xmath3 ( i.e. by using formulas given by eq . ( [ c4 ] ) in eq . ( [ ee ] ) ) , one obtains the following entanglement entropy @xmath220\nonumber\\ & & + 2(18\xi^2 - 9\tilde e\xi+\tilde e^2 - 4)^2\log_2[(18\xi^2 - 9\tilde e\xi+\tilde e^2 - 4)]\bigg)\;,\nonumber\\\end{aligned}\ ] ] where @xmath138 is given by eq . ( [ egs4 ] ) and @xmath221 by the second row of eq . ( [ a4 ] ) . note that the results stated for @xmath1 by eqs . ( [ f2 ] ) ( fisher information ) , ( [ alpha2 ] ) ( coherence visibility ) , and ( [ s2 ] ) ( entanglement entropy ) coincide with those that we obtained in @xcite . the bottom panel of fig . 2 shows the entanglement entropies ( [ s2 ] ) , ( [ s3 ] ) , and ( [ s4 ] ) as functions of the scaled on - site interaction @xmath135 . in the limit @xmath222 , corresponding to the emergence of the cat - like state ( [ cat ] ) , @xmath6 tends to one both when @xmath195 ( solid line , dot - dashed line ) and @xmath2 ( dashed line ) . by analyzing the plot of @xmath6 versus @xmath191 it is possible to observe that - as discussed in @xcite - the greater is the number of bosons the closest to zero is the negative @xmath191 ( which , in fact , is equal to @xmath223 when @xmath1 , @xmath224 when @xmath2 , and is @xmath225 when @xmath3 ) for which @xmath6 attains its maximum value . for @xmath1 and @xmath2 , the latter value coincide with that predicted theoretically , i.e. @xmath131 ( see the comment at the end of the previous section ) : in the ground state of the two - site bose - hubbard hamiltonian all the fock states @xmath73 have the same probability @xmath133 for any @xmath66 , as one can see from the left and the middle panels of the third row ( from the top ) of fig . when @xmath3 , instead , the maximum of @xmath6 does not coincide with @xmath131 , as it can be observed from the right plot of the third row ( from the top ) of fig.1 , where the @xmath190 , corresponding to the interaction signing the maximum of @xmath6 ( [ s4 ] ) , are different from each other . as conclusive remarks , we note that when @xmath2 , @xmath6 approaches to @xmath109 both in the limit @xmath222 and in the limit @xmath226 , as one can observe from the dashed line in the bottom panel of fig . 2 . moreover , we observe that the plot reported in the bottom panel of fig . 2 shows that the cat - like state ( [ cat ] ) ( @xmath208 , deeply attractive bosons ) is not the maximally entangled ground state achievable in our system . a remarkable point emerges from our analysis . the three ground - state characterizing parameters ( @xmath4 , @xmath5 , @xmath6 ) in the deep repulsive regime exhibit very different behavior depending on if @xmath195 or @xmath2 , as it can be seen from fig . 2 and from tabs . 1 - 3 . in fact , when @xmath227 , the fisher information , the coherence visibility , and entanglement entropy are all equal to zero if the ground state is the state ( [ fock ] ) ( @xmath195 ) ; they are finite , instead , if the ground state is the state ( [ pseudo : fock ] ) ( @xmath2 ) . note that this circumstance is quite general . in fact , the states ( [ fock ] ) and ( [ pseudo : fock ] ) are the ground states of the two - site bose - hubbard hamiltonian in the limit @xmath226 for any even @xmath0 and any odd @xmath0 , respectively . in particular , we want to stress that when the boson - boson interaction is strongly repulsive , the ground state of the two - site bh hamiltonian is not quantum entangled when @xmath0 is even , while it is a quantum entangled state when @xmath0 is odd . at this point , it is interesting to observe that it is possible to achieve the above formulas for @xmath4 - eqs . ( [ f2]),([f3]),([f4 ] ) - and @xmath5 - eqs . ( [ alpha2]),([alpha3 ] ) , ( [ alpha4 ] ) - also by exploiting the hellmann - feynman theorem ( hft ) @xcite . by using this theorem a relation can be established between the fisher information @xmath4 and the first partial derivative of the ground - state energy @xmath72 with respect to @xmath51 , @xmath228 . according to the hft , we have that @xcite @xmath229 if the properties @xmath230 ( eq . ( [ lrs ] ) ) and @xmath231 are used in eq . ( [ hft1 ] ) , we get @xmath232 on the other hand , again thanks to @xmath230 and @xmath231 , eqs . ( [ qfi ] ) and ( [ fi ] ) give rise to @xmath233 so that ( as also commented in @xcite ) @xmath234 we have therefore to know the energy of the ground state . when @xmath1 this energy is given by eq . ( [ egs2 ] ) , when @xmath2 by eq . ( [ egs3 ] ) , and when @xmath3 by eq . ( [ egs4 ] ) . by keeping in mind that in these three latter equations @xmath135 and the energies are measured in units of @xmath54 , one can resort from @xmath138 to @xmath72 and obtain eq . ( [ f2 ] ) for @xmath1 , eq . ( [ f3 ] ) for @xmath2 , and eq . ( [ f4 ] ) for @xmath3 . the hellmann - feynman theorem provides a relation between the coherence visibility @xmath5 and the first partial derivative of the ground - state energy @xmath72 with respect to @xmath54 as well . let us start from the fact that , according to the hft , one has that @xcite @xmath235 by using the fact that the coefficients @xmath74 involved in the expansion of @xmath71 ( @xmath236 ) are real ( see eqs . ( [ c2 ] ) , ( [ c3 ] ) , ( [ c4 ] ) jointly to eqs . ( [ a2]),([a3]),([a4 ] ) ) , we can write that @xmath237 , and then @xmath238 on the other hand , again in force to @xmath237 , we have that the coherence visibility ( [ visibility ] ) is @xmath239 so that ( see also @xcite ) @xmath240 also in this case , we use in eq . ( [ alphahf ] ) the results given by eq . ( [ egs2 ] ) , @xmath1 , by eq . ( [ egs3 ] ) , @xmath2 , and by eq . ( [ egs4 ] ) ( by keeping in mind that @xmath135 and the energies are measured in units of @xmath54 ) and get eq . ( [ alpha2 ] ) for @xmath1 , eq . ( [ alpha3 ] ) for @xmath2 , and eq . ( [ alpha4 ] ) for @xmath3 . we have investigated a finite number @xmath0 of ( both attractively and repulsively ) interacting bosonic atoms confined by a one - dimensional double - well shaped potential . within the two - site bose - hubbard model framework , we have carried out the zero - temperature analysis for @xmath1 , @xmath2 , and @xmath3 bosons by finding analytical formulas for the eigenvectors and eigenvalues of the corresponding ground states . these have been characterized by analytically calculating the fisher information , the coherence visibility , and the entanglement entropy . we have studied these parameters by varying the boson - boson interaction strength ( ranging from strong attractions to strong repulsions ) which is the key quantity in determining the kind of ground state sustained by the two - site bose - hubbard hamiltonian . in particular , we have commented on the difference , existing in the deep repulsive regime , between the structure of the ground state in the presence of an even number of bosons and that with an odd number of particles . we have pointed out , in particular , that the ground state of the two - site bose - hubbard hamiltonian is not quantum entangled when @xmath0 is even , while it is a quantum entangled state when @xmath0 is odd . + this work has been supported by miur ( prin 2010llkjbx ) . gm and ls acknowledge financial support from the university of padova ( progetto di ateneo 2011 ) and cariparo foundation ( progetto di eccellenza 2011 ) . gm acknowledges financial support also from progetto giovani 2011 of university of padova . wootters , phys . d. * 23 * , 357 ( 1981 ) ; c. w. helstrom , _ quantum detection and estimation theory _ ( academic press , new york , 1976 ) , chap . viii ; a. s. holevo , _ probalistic and statistical aspect of quantum theory _ ( north - holland , amsterdam , 1982 ) ; s.l . braunstein and c. m. caves , phys . lett . * 72 * , 3439 ( 1994 ) . bennett , h.j . bernstein , s. popescu , b. schumacher , phys . a * 53 * , 2046 ( 1996 ) ; s. hill and w. wootters , phys . 78 * , 5022 ( 1997 ) ; l. amico , r. fazio , a. osterloh , v. vedral , rev . * 80 * , 517 ( 2008 ) ; j. eisert , m. cramer , m. b. plenio , rev . phys . * 82 * , 277 ( 2010 ) .	we consider a finite number @xmath0 of interacting bosonic atoms at zero temperature confined in a one - dimensional double - well trap and study this system by using the two - site bose - hubbard ( bh ) hamiltonian . for systems with @xmath1 and @xmath2 , and @xmath3 bosons we analytically solve the eigenproblem associated to this hamiltonian and find its lowest energetic state . we investigate the structure of the ground state by varying the strength of the boson - boson interaction from the strongly attractive regime to the deep repulsive one . we characterize the ground state of the two - site bh hamiltonian by calculating the fisher information @xmath4 , the coherence visibility @xmath5 , and the entanglement entropy @xmath6 . for these quantities we provide analytical formulas that we use to study @xmath4 , @xmath5 , and @xmath6 as functions of the interaction between the particles . we discuss the difference existing , in the deep repulsive regime , between the case with an even number of bosons and that with an odd number of particles , both in the structure of the lowest energetic state and in the behavior of the three above ground - state characterizing parameters .
-0.3 cm , rosseland optical depth @xmath2(left scale ) and density ( right scale ) are given as a function of distance ( in @xmath3 ) for a typical sne ia at 15 days after the explosion . for comparison , we give the temperature @xmath4 for the grey extended atmosphere . the two dotted , vertical lines indicate the region of spectra formation . , title="fig:",width=117 ] -0.2 cm the last decade has witnessed an explosive growth of high - quality data for supernovae . advances in computational methods provided new insights into the physics of the objects , and advances in cosmology . both trends combined provided spectacular results not only for astronomy and the origin of elements but also for nuclear , high energy and particle physics , and cosmology . further improvements and the quest for the nature of the dark energy requires an increased accuracy for distance determinations from 10 % to about 2 to 3 % ( weller & albrecht 2001 ) making evolutionary effects with redshift a main concern , and a a better understanding of the physics of sne ia a requirement . there is general agreement that type ia supernovae ( sne ia ) are the result of a thermonuclear explosion of a degenerate c / o white dwarf ( wd ) with a mass close to the chandrasekhar limit . these scenarios allow to reproduce optical / infrared light curves ( lc ) and spectra of sne ia reasonably well . nowadays , we understand the basic , observational features . sne ia appear rather homogeneous because nuclear physics determines the structure of the wd , the explosion , light curves and spectra : ( 1 ) the wd is supported by degenerate electron pressure , ( 2 ) the total energy production during the explosion is given by the release of thermonuclear energy , and ( 3 ) the light curves are powered by the radioactive decay of @xmath1 produced during the explosion to first order , the outcome hardly depends on details of the physics , the scenario , or the progenitor ( `` stellar amnesia '' ) . homogeneity of sne ia does not ( ! ) imply a unique scenario , and it took the revolution in observational methods with respect to time and wavelength coverage to reveal differences and expose the diversity of within sne ia . for recent reviews see branch ( 1999 ) and hflich et al . 2003 ) . -0.0 cm and solar composition during the early phase of the explosion . the dominant signature of this interaction is the appearance of a persistent , secondary , high velocity ca ii feature . without ongoing interaction , no h or he lines are detectable . note that , even without a shell , a secondary ca ii feature can be seen for a period of 2 to 3 days during the phase when ca iii recombines to ca ii emphasizing the importance of a good time coverage for the observations . nearby shells mainly change early time lcs ( lower right ) due to blocking by thomson optical depth in the shell . in contrast , ongoing interaction will change the late time luminosities ( from gerardy et al . , title="fig:",width=188 ] -0.0 cm transition at 1.644 @xmath5 are given for the original delayed detonation model ( solid ) and mixed chemistry ( light ) normalized to the maximum line flux ( dotted ) and the wings ( dashed ) , respectively . mixing of the inner iron - rich layers of @xmath0 and stable isotopes ( fig . 4 ) is to be expected from current 3d models during the deflagration phase which is dominated by rt instabilities , and would produce round profiles which seem to be at odds with the observations . possible explanations may be that small - scale , pre - existing velocity fields are important for the propagation of nuclear flames . on the right , the visual light curve and @xmath6 are given for the same delayed detonation model but with and without mixing of the inner layers . differences in v and b - v are @xmath70.2@xmath8 and @xmath9 , respectively . in effect , mixing redistributes @xmath1 from the outer to the inner layers which decreases the photospheric heating at about maximum light but increases the @xmath10-trapping later on ( from hflich et al . title="fig:",width=181 ] 1.0 cm -rays at day 1 ( left ) and 23 ( right ) for a 3d deflagration model based on our full 3-d mc gamma ray transport . at about day 23 , the energy deposition is not confined to the radioactive @xmath1 ruling out clumpiness as a solution to the excessive excitation of s and si lines . * lower , right panel : * optical flux and polarization spectra at day 15 after the explosion for the subluminous 3-d delayed - detonation model in comparison with the sn1999by at about maximum light . in the observations , the polarization angle is constant indicating rotational symmetry of the envelope , and an axis ratio a / b of 1.17 ( from hflich et al . 2002 , hflich 2002 , howell et al . , title="fig:",width=207 ] - consistency is required to link the observable and the progenitor system , progenitor and explosion physics . by the physical conditions , consistency is also demanded for the treatment of hydrodynamics , rate equations , and radiation transport ( fig.1 ) . density structures require detailed hydrodynamics , low densities cause strong non - lte effects throughout the entire envelopes and the radiation field differs from a black body , chemical profiles are depth dependent , energy source and sink terms due to hydrodynamical effects and radioactive decays dominate throughout the photon decoupling region , and all physical properties are time - dependent because the energy diffusion time scales are comparable to the hydrodynamical expansion time scale ( hflich , 1995 ) . our approach significantly reduces the number of free parameters , namely the initial structure of the progenitor , the accretion rate on the wd , and the description of the nuclear burning front . the light curves and spectral evolution follow directly from the explosion model without any further tuning . - the best current explosion scenario is the delayed detonation ( dd ) model ( khokhlov 1991 ) . these models reproduce reasonable well the color evolution of optical and ir - lc , and the spectral evolution of both normal bright ( hflich 1995 ) and subluminous sne ( fig . 4 , hflich et al . 2002 ) within a unified scenario . by enlarge , dd models produce radially stratified chemical structures for elements of explosive carbon , oxygen and si burning in agreement with the observations from the optical and near ir ( e.g. fig . 4 , barbon et al . 1989 , wheeler et al . 1998 , marion et al . 2003 ) the brightness decline relation ( phillips 1993 ) can be understood as an opacity effect ( fig . 3 , hflich et al . 1996 , maeda et al . 2003 ) including the color evolution in b - v and the brightness range ( fig . 2 hflich et al . 2002 ) by varying a single , free parameter for the amount of burning when the deflagration turns into a detonation which determines the pre - expansion of the wd . in spherical models , this quantity is often parameterized by a transition density @xmath11 . other properties produce a dispersion . e.g. changes in the progenitor , i.e. in main sequence mass @xmath12 and z , alter the explosion energy and @xmath1 mass , respectively , but , by enlarge , along the @xmath13 and @xmath14 relations ( hflich et al . 1998 , 2001 ) . however , the diversity in progenitors cause an intrinsic dispersion in b - v of @xmath15 which , in practice , seriously hampers the corrections by interstellar reddening . - the ir is a primary tool to study intermediate mass elements , and spectropolarimetry and line profiles allow to probe for 3-d signatures ( hflich 1991 , wang et al . 1997 , bowers et al . 1997 , wheeler et al . 1998 , howell et al . 2001 , hflich et al.2002 ) . - our models allow to study particular physics and relations to the properties of the progenitors in isolation such as signatures of the progenitor system ( fig . 5 , gerardy et al . 2004 ) and the distribution of neutron rich isotopes ( fig . 6 , hflich et al . 2004 ) . in extensive studies , we investigated the influence of the central density of the wd , the ignition process , and metallicities ( which are expected to change with time ) ( hflich et al . 1998 , 2001 , dominguez et al . 2002 , see also timmes et al . pre - conditioning of the wd is a key to understand the diversity of sneia . metallicity will effect the statistical sample because it changes the life times and radii of stars and , consequently , the binary system with mass overflow , and the physics of individual objects . a change of metallicity or progenitor ms mass effects both the progenitor evolution , wd structure and the observable lc shapes simultaneously , and the combined effect must be considered when comparing to observations . as can be seen from fig . the change of the initial fe changes the core he burning during the stellar evolution which determines the c / o ratio , whereas the @xmath16 directly effects the explosive nuclear burning by shifting the nuclear statistical equilibrium slightly away from @xmath1 . the former effect alters the energetics of the explosion ( explosion energy ) whereas the latter changes the energy gain by radioactive decay of @xmath1 which powers the lcs . for low metallicities , the influence of the energetics dominates the changes in the lc shape . it causes a change @xmath17 of the rise - time for a given decline ratio and , thus , an offset @xmath18 in the brightness decline relation by @xmath19 . these results do not ( ! ) imply that all processes involved are spherical in nature but they are a result of the stellar amnesia mentioned above , i.e. the properties depend mostly on integral quantities and basic nuclear physics . despite the successes of spherical delayed detonation models , it becomes increasingly obvious that multidimensional effects are important towards a better understanding of sne ia . - asymmetries in a sn ia may be present for a variety of reasons . within the framework of the explosion of a massive wd , reasons and all seems to be realized . reasons include instabilities in the burning front ( khokhlov 1995 , 2001 , reinecke et al . 2002 gamezo et al . 2003 ) , namely the deflagration phase , off - center detonations or the transitions from deflagration to detonation ( livne 1999 ) , rapid rotation of the wd , and interaction with the accretion disk and the companion star . whereas we have evidence for the existence of the latter effects , direct evidence is still elusive for the rt instabilities or , more precisely , such evidence has not been seen although it was expected ( figs . 4 & 6 ) . both from the remnants and the lack of strong polarization in thermonuclear supernovae , limited scattering in brightness , we can conclude that the envelopes have rather small global deviations from spherical geometry of the order of 5 to 20 % . for cosmology , deviations from sphericity will introduce a directional dependence of the luminosity of the order of about 4 - 6 % . argast d. , samland m. , gerhard o. e.,thielemann f .- k . 2000 , a&a 356 , 873 barbon r. , benetti s. , rosiono l. , cappellaro e. turatto m.1990 a&a 237 , 79 bowers , e. j. c. , meikle , w. p. s. , geballe , t. r. , walton , n. a. , pinto , p. a. , dhillon , v. s. , howell , , & harrop - allin , m. k. 1997 , , 290 , 663 branch , d. 1999 , , 36 , 17 dominguez , i. , hflich , p. , straniero , o. 2001 , , 557 , 279 gamezo , v. n. et al . 2003 , science , 299 , 77 garnavich et al . 2004 , apj , in press gerardy , c. l. , hflich , p. , fesen , r. a. , marion , g. h. , nomoto , k. , quimby , r. , schaefer , b. e. , wang , l. , & wheeler , j. c. 2004 , , in press ( astro - ph/0309639 ) hflich , p. , gerardy c. , nomoto k. , motohara k. , fesen r.a . , maeda k. , ohkubo t. , tominaga n , apj , in press ( dec . 20 , 2004 ) & astro - ph/2004 hflich , p. , gerardy , c. , linder , e. , & marion , h. 2003 , lecture notes in physics 635 , springer press , p. 203 & astro - ph/0301334 hflich , p. , gerardy , c. , fesen , r. , & sakai , s. 2002 , , 568 , 791 hflich , p. , wheeler , j. c. , thielemann , f. k. 1998 , , 495 , 617 hflich , p. , khokhlov , a. , wheeler , j. c. , phillips , m. m. , suntzeff , n. b. , & hamuy , m. 1996 , , 472 , l81 hflich , p. , khokhlov , a. 1996 , , 457 , 500 hflich , p. 1995 , , 443 , 89 howell , a. , hflich , p. , wang , l. , wheeler , j. c. 2001 , , 556 , 302 kasen d. , et al . 2004 , , in press & astro - ph/0301312 khokhlov , a. 2001 , astro - ph/0008463 khokhlov , a. 1991 , , 245 114 livne e. 1999 , apj 527 , l97 marion h. , hflich p. , vacca w.d . , wheeler c.j . 2003 , apj 591 , 316 maeda , k. , mazzali , p. , deng , j. , et al . 2003 , , 593 , 22 phillips , m.m . , 1993 , , 413 , 105 phillips , m.m . , 2003 , lecture notes in physics 635 , springer press , p. 175 reinecke , m. , hillebrandt , w. , & niemeyer , j. c. 2002 , aap , 391 , 116 timmes f. , brown e. , truran 2003 , apj , 590l wang , l. , wheeler , j. c. , hflich , p. 1997 , , 476 , 27 wang , l. , baade , d. , hflich , p. , wheeler , j. c. , kawabata , k. , nomoto , k. 2003 , apj , in press wheeler c.j . , hflich p. , harkness r. , syromillo j. 1998 , 496 , 908 weller j. , albrecht a. 2001 , physical review letters , 86 , 1939	based on detailed models for the progenitors , explosions , light curves ( lcs ) and spectra , we discuss signatures of thermonuclear explosions , and the implications for cosmology . consistency is needed to link observables and explosion physics . sne ia most probably result from the explosion of a degenerate co - white dwarf ( wd ) close to the chandrasekhar mass . there is strong evidence that most of the wd is burned with an extended outer layer of explosive c - burning products ( o , ne , mg ) and very little c remaining . overall , the chemical structure is radially stratified . this leads to the currently favored delayed detonation model in which a phase of slow nuclear burning as a deflagration front is followed by a detonation phase . the importance of pre - conditioning became obvious . within a unified scenario , spherical models allow to understand both the homogeneity and basic properties of lcs and spectra , and they allow to probe for their diversity which is a key for high precision cosmology by sne ia . we emphasize the relation between lc properties and spectra because , for local sne ia , the diversity becomes apparent the combination of spectra and lcs whereas , by enlarge , we have to for high - z objects . we show how we can actually probe the properties of the progenitor , its environment , and details of the explosion physics . we demonstrate the influence of the metallicity z on the progenitors , explosion physics and the combined effect on light curves . by enlarge , a change of z causes a shift of along the brightness decline relation because z shifts the balance between @xmath0 and non - radioactive isotopes but hardly changes the energetics or the @xmath1 distribution . however , the diversity of the progenitors produces an intrinsic dispersion in b - v which may pose a problem for reddening corrections . we discuss the nature of subluminous sn1999by , and how it can be understood in the same framework as normal - bright sne ia . at the example of sn200du , we show the influence of the progenitor system and distribution of isotopes on light curves . in both objects , we have seen clear evidence for some departure from sphericity probably due to circumstellar interaction and stellar rotation but the 3d signatures of deflagration fronts remain an elusive feature .

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.